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>
31 struct anon_vma_chain;
34 struct writeback_control;
37 void init_mm_internals(void);
39 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
40 extern unsigned long max_mapnr;
42 static inline void set_max_mapnr(unsigned long limit)
47 static inline void set_max_mapnr(unsigned long limit) { }
50 extern unsigned long totalram_pages;
51 extern void * high_memory;
52 extern int page_cluster;
55 extern int sysctl_legacy_va_layout;
57 #define sysctl_legacy_va_layout 0
60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
61 extern const int mmap_rnd_bits_min;
62 extern const int mmap_rnd_bits_max;
63 extern int mmap_rnd_bits __read_mostly;
65 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
66 extern const int mmap_rnd_compat_bits_min;
67 extern const int mmap_rnd_compat_bits_max;
68 extern int mmap_rnd_compat_bits __read_mostly;
72 #include <asm/pgtable.h>
73 #include <asm/processor.h>
76 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
80 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
84 #define lm_alias(x) __va(__pa_symbol(x))
88 * To prevent common memory management code establishing
89 * a zero page mapping on a read fault.
90 * This macro should be defined within <asm/pgtable.h>.
91 * s390 does this to prevent multiplexing of hardware bits
92 * related to the physical page in case of virtualization.
94 #ifndef mm_forbids_zeropage
95 #define mm_forbids_zeropage(X) (0)
99 * Default maximum number of active map areas, this limits the number of vmas
100 * per mm struct. Users can overwrite this number by sysctl but there is a
103 * When a program's coredump is generated as ELF format, a section is created
104 * per a vma. In ELF, the number of sections is represented in unsigned short.
105 * This means the number of sections should be smaller than 65535 at coredump.
106 * Because the kernel adds some informative sections to a image of program at
107 * generating coredump, we need some margin. The number of extra sections is
108 * 1-3 now and depends on arch. We use "5" as safe margin, here.
110 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
111 * not a hard limit any more. Although some userspace tools can be surprised by
114 #define MAPCOUNT_ELF_CORE_MARGIN (5)
115 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
117 extern int sysctl_max_map_count;
119 extern unsigned long sysctl_user_reserve_kbytes;
120 extern unsigned long sysctl_admin_reserve_kbytes;
122 extern int sysctl_overcommit_memory;
123 extern int sysctl_overcommit_ratio;
124 extern unsigned long sysctl_overcommit_kbytes;
126 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
128 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
131 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
133 /* to align the pointer to the (next) page boundary */
134 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
136 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
137 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
140 * Linux kernel virtual memory manager primitives.
141 * The idea being to have a "virtual" mm in the same way
142 * we have a virtual fs - giving a cleaner interface to the
143 * mm details, and allowing different kinds of memory mappings
144 * (from shared memory to executable loading to arbitrary
148 extern struct kmem_cache *vm_area_cachep;
151 extern struct rb_root nommu_region_tree;
152 extern struct rw_semaphore nommu_region_sem;
154 extern unsigned int kobjsize(const void *objp);
158 * vm_flags in vm_area_struct, see mm_types.h.
159 * When changing, update also include/trace/events/mmflags.h
161 #define VM_NONE 0x00000000
163 #define VM_READ 0x00000001 /* currently active flags */
164 #define VM_WRITE 0x00000002
165 #define VM_EXEC 0x00000004
166 #define VM_SHARED 0x00000008
168 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
169 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
170 #define VM_MAYWRITE 0x00000020
171 #define VM_MAYEXEC 0x00000040
172 #define VM_MAYSHARE 0x00000080
174 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
175 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
176 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
177 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
178 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
180 #define VM_LOCKED 0x00002000
181 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
183 /* Used by sys_madvise() */
184 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
185 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
187 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
188 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
189 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
190 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
191 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
192 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
193 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
194 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
195 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
197 #ifdef CONFIG_MEM_SOFT_DIRTY
198 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
200 # define VM_SOFTDIRTY 0
203 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
204 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
205 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
206 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
208 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
209 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
210 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
211 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
212 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
213 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
214 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
215 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
216 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
217 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
218 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
219 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
221 #if defined(CONFIG_X86)
222 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
223 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
224 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
225 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
226 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
227 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
228 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
230 #elif defined(CONFIG_PPC)
231 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
232 #elif defined(CONFIG_PARISC)
233 # define VM_GROWSUP VM_ARCH_1
234 #elif defined(CONFIG_METAG)
235 # define VM_GROWSUP VM_ARCH_1
236 #elif defined(CONFIG_IA64)
237 # define VM_GROWSUP VM_ARCH_1
238 #elif !defined(CONFIG_MMU)
239 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
242 #if defined(CONFIG_X86_INTEL_MPX)
243 /* MPX specific bounds table or bounds directory */
244 # define VM_MPX VM_HIGH_ARCH_4
246 # define VM_MPX VM_NONE
250 # define VM_GROWSUP VM_NONE
253 /* Bits set in the VMA until the stack is in its final location */
254 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
256 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
257 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
260 #ifdef CONFIG_STACK_GROWSUP
261 #define VM_STACK VM_GROWSUP
263 #define VM_STACK VM_GROWSDOWN
266 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
269 * Special vmas that are non-mergable, non-mlock()able.
270 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
272 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
274 /* This mask defines which mm->def_flags a process can inherit its parent */
275 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
277 /* This mask is used to clear all the VMA flags used by mlock */
278 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
281 * mapping from the currently active vm_flags protection bits (the
282 * low four bits) to a page protection mask..
284 extern pgprot_t protection_map[16];
286 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
287 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
288 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
289 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
290 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
291 #define FAULT_FLAG_TRIED 0x20 /* Second try */
292 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
293 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
294 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
296 #define FAULT_FLAG_TRACE \
297 { FAULT_FLAG_WRITE, "WRITE" }, \
298 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
299 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
300 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
301 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
302 { FAULT_FLAG_TRIED, "TRIED" }, \
303 { FAULT_FLAG_USER, "USER" }, \
304 { FAULT_FLAG_REMOTE, "REMOTE" }, \
305 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
308 * vm_fault is filled by the the pagefault handler and passed to the vma's
309 * ->fault function. The vma's ->fault is responsible for returning a bitmask
310 * of VM_FAULT_xxx flags that give details about how the fault was handled.
312 * MM layer fills up gfp_mask for page allocations but fault handler might
313 * alter it if its implementation requires a different allocation context.
315 * pgoff should be used in favour of virtual_address, if possible.
318 struct vm_area_struct *vma; /* Target VMA */
319 unsigned int flags; /* FAULT_FLAG_xxx flags */
320 gfp_t gfp_mask; /* gfp mask to be used for allocations */
321 pgoff_t pgoff; /* Logical page offset based on vma */
322 unsigned long address; /* Faulting virtual address */
323 pmd_t *pmd; /* Pointer to pmd entry matching
325 pud_t *pud; /* Pointer to pud entry matching
328 pte_t orig_pte; /* Value of PTE at the time of fault */
330 struct page *cow_page; /* Page handler may use for COW fault */
331 struct mem_cgroup *memcg; /* Cgroup cow_page belongs to */
332 struct page *page; /* ->fault handlers should return a
333 * page here, unless VM_FAULT_NOPAGE
334 * is set (which is also implied by
337 /* These three entries are valid only while holding ptl lock */
338 pte_t *pte; /* Pointer to pte entry matching
339 * the 'address'. NULL if the page
340 * table hasn't been allocated.
342 spinlock_t *ptl; /* Page table lock.
343 * Protects pte page table if 'pte'
344 * is not NULL, otherwise pmd.
346 pgtable_t prealloc_pte; /* Pre-allocated pte page table.
347 * vm_ops->map_pages() calls
348 * alloc_set_pte() from atomic context.
349 * do_fault_around() pre-allocates
350 * page table to avoid allocation from
355 /* page entry size for vm->huge_fault() */
356 enum page_entry_size {
363 * These are the virtual MM functions - opening of an area, closing and
364 * unmapping it (needed to keep files on disk up-to-date etc), pointer
365 * to the functions called when a no-page or a wp-page exception occurs.
367 struct vm_operations_struct {
368 void (*open)(struct vm_area_struct * area);
369 void (*close)(struct vm_area_struct * area);
370 int (*split)(struct vm_area_struct * area, unsigned long addr);
371 int (*mremap)(struct vm_area_struct * area);
372 int (*fault)(struct vm_fault *vmf);
373 int (*huge_fault)(struct vm_fault *vmf, enum page_entry_size pe_size);
374 void (*map_pages)(struct vm_fault *vmf,
375 pgoff_t start_pgoff, pgoff_t end_pgoff);
377 /* notification that a previously read-only page is about to become
378 * writable, if an error is returned it will cause a SIGBUS */
379 int (*page_mkwrite)(struct vm_fault *vmf);
381 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
382 int (*pfn_mkwrite)(struct vm_fault *vmf);
384 /* called by access_process_vm when get_user_pages() fails, typically
385 * for use by special VMAs that can switch between memory and hardware
387 int (*access)(struct vm_area_struct *vma, unsigned long addr,
388 void *buf, int len, int write);
390 /* Called by the /proc/PID/maps code to ask the vma whether it
391 * has a special name. Returning non-NULL will also cause this
392 * vma to be dumped unconditionally. */
393 const char *(*name)(struct vm_area_struct *vma);
397 * set_policy() op must add a reference to any non-NULL @new mempolicy
398 * to hold the policy upon return. Caller should pass NULL @new to
399 * remove a policy and fall back to surrounding context--i.e. do not
400 * install a MPOL_DEFAULT policy, nor the task or system default
403 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
406 * get_policy() op must add reference [mpol_get()] to any policy at
407 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
408 * in mm/mempolicy.c will do this automatically.
409 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
410 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
411 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
412 * must return NULL--i.e., do not "fallback" to task or system default
415 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
419 * Called by vm_normal_page() for special PTEs to find the
420 * page for @addr. This is useful if the default behavior
421 * (using pte_page()) would not find the correct page.
423 struct page *(*find_special_page)(struct vm_area_struct *vma,
430 #define page_private(page) ((page)->private)
431 #define set_page_private(page, v) ((page)->private = (v))
433 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
434 static inline int pmd_devmap(pmd_t pmd)
438 static inline int pud_devmap(pud_t pud)
442 static inline int pgd_devmap(pgd_t pgd)
449 * FIXME: take this include out, include page-flags.h in
450 * files which need it (119 of them)
452 #include <linux/page-flags.h>
453 #include <linux/huge_mm.h>
456 * Methods to modify the page usage count.
458 * What counts for a page usage:
459 * - cache mapping (page->mapping)
460 * - private data (page->private)
461 * - page mapped in a task's page tables, each mapping
462 * is counted separately
464 * Also, many kernel routines increase the page count before a critical
465 * routine so they can be sure the page doesn't go away from under them.
469 * Drop a ref, return true if the refcount fell to zero (the page has no users)
471 static inline int put_page_testzero(struct page *page)
473 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
474 return page_ref_dec_and_test(page);
478 * Try to grab a ref unless the page has a refcount of zero, return false if
480 * This can be called when MMU is off so it must not access
481 * any of the virtual mappings.
483 static inline int get_page_unless_zero(struct page *page)
485 return page_ref_add_unless(page, 1, 0);
488 extern int page_is_ram(unsigned long pfn);
496 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
499 /* Support for virtually mapped pages */
500 struct page *vmalloc_to_page(const void *addr);
501 unsigned long vmalloc_to_pfn(const void *addr);
504 * Determine if an address is within the vmalloc range
506 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
507 * is no special casing required.
509 static inline bool is_vmalloc_addr(const void *x)
512 unsigned long addr = (unsigned long)x;
514 return addr >= VMALLOC_START && addr < VMALLOC_END;
520 extern int is_vmalloc_or_module_addr(const void *x);
522 static inline int is_vmalloc_or_module_addr(const void *x)
528 extern void *kvmalloc_node(size_t size, gfp_t flags, int node);
529 static inline void *kvmalloc(size_t size, gfp_t flags)
531 return kvmalloc_node(size, flags, NUMA_NO_NODE);
533 static inline void *kvzalloc_node(size_t size, gfp_t flags, int node)
535 return kvmalloc_node(size, flags | __GFP_ZERO, node);
537 static inline void *kvzalloc(size_t size, gfp_t flags)
539 return kvmalloc(size, flags | __GFP_ZERO);
542 static inline void *kvmalloc_array(size_t n, size_t size, gfp_t flags)
544 if (size != 0 && n > SIZE_MAX / size)
547 return kvmalloc(n * size, flags);
550 extern void kvfree(const void *addr);
551 extern void kvfree_sensitive(const void *addr, size_t len);
554 * Mapcount of compound page as a whole, does not include mapped sub-pages.
556 * Must be called only for compound pages or any their tail sub-pages.
558 static inline int compound_mapcount(struct page *page)
560 VM_BUG_ON_PAGE(!PageCompound(page), page);
561 page = compound_head(page);
562 return atomic_read(compound_mapcount_ptr(page)) + 1;
566 * The atomic page->_mapcount, starts from -1: so that transitions
567 * both from it and to it can be tracked, using atomic_inc_and_test
568 * and atomic_add_negative(-1).
570 static inline void page_mapcount_reset(struct page *page)
572 atomic_set(&(page)->_mapcount, -1);
575 int __page_mapcount(struct page *page);
578 * Mapcount of 0-order page; when compound sub-page, includes
579 * compound_mapcount().
581 * Result is undefined for pages which cannot be mapped into userspace.
582 * For example SLAB or special types of pages. See function page_has_type().
583 * They use this place in struct page differently.
585 static inline int page_mapcount(struct page *page)
587 if (unlikely(PageCompound(page)))
588 return __page_mapcount(page);
589 return atomic_read(&page->_mapcount) + 1;
592 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
593 int total_mapcount(struct page *page);
594 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
596 static inline int total_mapcount(struct page *page)
598 return page_mapcount(page);
600 static inline int page_trans_huge_mapcount(struct page *page,
603 int mapcount = page_mapcount(page);
605 *total_mapcount = mapcount;
610 static inline struct page *virt_to_head_page(const void *x)
612 struct page *page = virt_to_page(x);
614 return compound_head(page);
617 void __put_page(struct page *page);
619 void put_pages_list(struct list_head *pages);
621 void split_page(struct page *page, unsigned int order);
624 * Compound pages have a destructor function. Provide a
625 * prototype for that function and accessor functions.
626 * These are _only_ valid on the head of a compound page.
628 typedef void compound_page_dtor(struct page *);
630 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
631 enum compound_dtor_id {
634 #ifdef CONFIG_HUGETLB_PAGE
637 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
642 extern compound_page_dtor * const compound_page_dtors[];
644 static inline void set_compound_page_dtor(struct page *page,
645 enum compound_dtor_id compound_dtor)
647 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
648 page[1].compound_dtor = compound_dtor;
651 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
653 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
654 return compound_page_dtors[page[1].compound_dtor];
657 static inline unsigned int compound_order(struct page *page)
661 return page[1].compound_order;
664 static inline void set_compound_order(struct page *page, unsigned int order)
666 page[1].compound_order = order;
669 void free_compound_page(struct page *page);
673 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
674 * servicing faults for write access. In the normal case, do always want
675 * pte_mkwrite. But get_user_pages can cause write faults for mappings
676 * that do not have writing enabled, when used by access_process_vm.
678 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
680 if (likely(vma->vm_flags & VM_WRITE))
681 pte = pte_mkwrite(pte);
685 int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg,
687 int finish_fault(struct vm_fault *vmf);
688 int finish_mkwrite_fault(struct vm_fault *vmf);
692 * Multiple processes may "see" the same page. E.g. for untouched
693 * mappings of /dev/null, all processes see the same page full of
694 * zeroes, and text pages of executables and shared libraries have
695 * only one copy in memory, at most, normally.
697 * For the non-reserved pages, page_count(page) denotes a reference count.
698 * page_count() == 0 means the page is free. page->lru is then used for
699 * freelist management in the buddy allocator.
700 * page_count() > 0 means the page has been allocated.
702 * Pages are allocated by the slab allocator in order to provide memory
703 * to kmalloc and kmem_cache_alloc. In this case, the management of the
704 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
705 * unless a particular usage is carefully commented. (the responsibility of
706 * freeing the kmalloc memory is the caller's, of course).
708 * A page may be used by anyone else who does a __get_free_page().
709 * In this case, page_count still tracks the references, and should only
710 * be used through the normal accessor functions. The top bits of page->flags
711 * and page->virtual store page management information, but all other fields
712 * are unused and could be used privately, carefully. The management of this
713 * page is the responsibility of the one who allocated it, and those who have
714 * subsequently been given references to it.
716 * The other pages (we may call them "pagecache pages") are completely
717 * managed by the Linux memory manager: I/O, buffers, swapping etc.
718 * The following discussion applies only to them.
720 * A pagecache page contains an opaque `private' member, which belongs to the
721 * page's address_space. Usually, this is the address of a circular list of
722 * the page's disk buffers. PG_private must be set to tell the VM to call
723 * into the filesystem to release these pages.
725 * A page may belong to an inode's memory mapping. In this case, page->mapping
726 * is the pointer to the inode, and page->index is the file offset of the page,
727 * in units of PAGE_SIZE.
729 * If pagecache pages are not associated with an inode, they are said to be
730 * anonymous pages. These may become associated with the swapcache, and in that
731 * case PG_swapcache is set, and page->private is an offset into the swapcache.
733 * In either case (swapcache or inode backed), the pagecache itself holds one
734 * reference to the page. Setting PG_private should also increment the
735 * refcount. The each user mapping also has a reference to the page.
737 * The pagecache pages are stored in a per-mapping radix tree, which is
738 * rooted at mapping->page_tree, and indexed by offset.
739 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
740 * lists, we instead now tag pages as dirty/writeback in the radix tree.
742 * All pagecache pages may be subject to I/O:
743 * - inode pages may need to be read from disk,
744 * - inode pages which have been modified and are MAP_SHARED may need
745 * to be written back to the inode on disk,
746 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
747 * modified may need to be swapped out to swap space and (later) to be read
752 * The zone field is never updated after free_area_init_core()
753 * sets it, so none of the operations on it need to be atomic.
756 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
757 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
758 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
759 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
760 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
763 * Define the bit shifts to access each section. For non-existent
764 * sections we define the shift as 0; that plus a 0 mask ensures
765 * the compiler will optimise away reference to them.
767 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
768 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
769 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
770 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
772 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
773 #ifdef NODE_NOT_IN_PAGE_FLAGS
774 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
775 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
776 SECTIONS_PGOFF : ZONES_PGOFF)
778 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
779 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
780 NODES_PGOFF : ZONES_PGOFF)
783 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
785 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
786 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
789 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
790 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
791 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
792 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
793 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
795 static inline enum zone_type page_zonenum(const struct page *page)
797 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
800 #ifdef CONFIG_ZONE_DEVICE
801 static inline bool is_zone_device_page(const struct page *page)
803 return page_zonenum(page) == ZONE_DEVICE;
806 static inline bool is_zone_device_page(const struct page *page)
812 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
813 void put_zone_device_private_or_public_page(struct page *page);
814 DECLARE_STATIC_KEY_FALSE(device_private_key);
815 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
816 static inline bool is_device_private_page(const struct page *page);
817 static inline bool is_device_public_page(const struct page *page);
818 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
819 static inline void put_zone_device_private_or_public_page(struct page *page)
822 #define IS_HMM_ENABLED 0
823 static inline bool is_device_private_page(const struct page *page)
827 static inline bool is_device_public_page(const struct page *page)
831 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
834 /* 127: arbitrary random number, small enough to assemble well */
835 #define page_ref_zero_or_close_to_overflow(page) \
836 ((unsigned int) page_ref_count(page) + 127u <= 127u)
838 static inline void get_page(struct page *page)
840 page = compound_head(page);
842 * Getting a normal page or the head of a compound page
843 * requires to already have an elevated page->_refcount.
845 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
849 static inline __must_check bool try_get_page(struct page *page)
851 page = compound_head(page);
852 if (WARN_ON_ONCE(page_ref_count(page) <= 0))
858 static inline void put_page(struct page *page)
860 page = compound_head(page);
863 * For private device pages we need to catch refcount transition from
864 * 2 to 1, when refcount reach one it means the private device page is
865 * free and we need to inform the device driver through callback. See
866 * include/linux/memremap.h and HMM for details.
868 if (IS_HMM_ENABLED && unlikely(is_device_private_page(page) ||
869 unlikely(is_device_public_page(page)))) {
870 put_zone_device_private_or_public_page(page);
874 if (put_page_testzero(page))
878 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
879 #define SECTION_IN_PAGE_FLAGS
883 * The identification function is mainly used by the buddy allocator for
884 * determining if two pages could be buddies. We are not really identifying
885 * the zone since we could be using the section number id if we do not have
886 * node id available in page flags.
887 * We only guarantee that it will return the same value for two combinable
890 static inline int page_zone_id(struct page *page)
892 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
895 static inline int zone_to_nid(struct zone *zone)
904 #ifdef NODE_NOT_IN_PAGE_FLAGS
905 extern int page_to_nid(const struct page *page);
907 static inline int page_to_nid(const struct page *page)
909 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
913 #ifdef CONFIG_NUMA_BALANCING
914 static inline int cpu_pid_to_cpupid(int cpu, int pid)
916 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
919 static inline int cpupid_to_pid(int cpupid)
921 return cpupid & LAST__PID_MASK;
924 static inline int cpupid_to_cpu(int cpupid)
926 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
929 static inline int cpupid_to_nid(int cpupid)
931 return cpu_to_node(cpupid_to_cpu(cpupid));
934 static inline bool cpupid_pid_unset(int cpupid)
936 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
939 static inline bool cpupid_cpu_unset(int cpupid)
941 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
944 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
946 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
949 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
950 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
951 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
953 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
956 static inline int page_cpupid_last(struct page *page)
958 return page->_last_cpupid;
960 static inline void page_cpupid_reset_last(struct page *page)
962 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
965 static inline int page_cpupid_last(struct page *page)
967 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
970 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
972 static inline void page_cpupid_reset_last(struct page *page)
974 page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT;
976 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
977 #else /* !CONFIG_NUMA_BALANCING */
978 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
980 return page_to_nid(page); /* XXX */
983 static inline int page_cpupid_last(struct page *page)
985 return page_to_nid(page); /* XXX */
988 static inline int cpupid_to_nid(int cpupid)
993 static inline int cpupid_to_pid(int cpupid)
998 static inline int cpupid_to_cpu(int cpupid)
1003 static inline int cpu_pid_to_cpupid(int nid, int pid)
1008 static inline bool cpupid_pid_unset(int cpupid)
1013 static inline void page_cpupid_reset_last(struct page *page)
1017 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
1021 #endif /* CONFIG_NUMA_BALANCING */
1023 static inline struct zone *page_zone(const struct page *page)
1025 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
1028 static inline pg_data_t *page_pgdat(const struct page *page)
1030 return NODE_DATA(page_to_nid(page));
1033 #ifdef SECTION_IN_PAGE_FLAGS
1034 static inline void set_page_section(struct page *page, unsigned long section)
1036 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
1037 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
1040 static inline unsigned long page_to_section(const struct page *page)
1042 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
1046 static inline void set_page_zone(struct page *page, enum zone_type zone)
1048 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
1049 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
1052 static inline void set_page_node(struct page *page, unsigned long node)
1054 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
1055 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
1058 static inline void set_page_links(struct page *page, enum zone_type zone,
1059 unsigned long node, unsigned long pfn)
1061 set_page_zone(page, zone);
1062 set_page_node(page, node);
1063 #ifdef SECTION_IN_PAGE_FLAGS
1064 set_page_section(page, pfn_to_section_nr(pfn));
1069 static inline struct mem_cgroup *page_memcg(struct page *page)
1071 return page->mem_cgroup;
1073 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1075 WARN_ON_ONCE(!rcu_read_lock_held());
1076 return READ_ONCE(page->mem_cgroup);
1079 static inline struct mem_cgroup *page_memcg(struct page *page)
1083 static inline struct mem_cgroup *page_memcg_rcu(struct page *page)
1085 WARN_ON_ONCE(!rcu_read_lock_held());
1091 * Some inline functions in vmstat.h depend on page_zone()
1093 #include <linux/vmstat.h>
1095 static __always_inline void *lowmem_page_address(const struct page *page)
1097 return page_to_virt(page);
1100 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1101 #define HASHED_PAGE_VIRTUAL
1104 #if defined(WANT_PAGE_VIRTUAL)
1105 static inline void *page_address(const struct page *page)
1107 return page->virtual;
1109 static inline void set_page_address(struct page *page, void *address)
1111 page->virtual = address;
1113 #define page_address_init() do { } while(0)
1116 #if defined(HASHED_PAGE_VIRTUAL)
1117 void *page_address(const struct page *page);
1118 void set_page_address(struct page *page, void *virtual);
1119 void page_address_init(void);
1122 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1123 #define page_address(page) lowmem_page_address(page)
1124 #define set_page_address(page, address) do { } while(0)
1125 #define page_address_init() do { } while(0)
1128 extern void *page_rmapping(struct page *page);
1129 extern struct anon_vma *page_anon_vma(struct page *page);
1130 extern struct address_space *page_mapping(struct page *page);
1132 extern struct address_space *__page_file_mapping(struct page *);
1135 struct address_space *page_file_mapping(struct page *page)
1137 if (unlikely(PageSwapCache(page)))
1138 return __page_file_mapping(page);
1140 return page->mapping;
1143 extern pgoff_t __page_file_index(struct page *page);
1146 * Return the pagecache index of the passed page. Regular pagecache pages
1147 * use ->index whereas swapcache pages use swp_offset(->private)
1149 static inline pgoff_t page_index(struct page *page)
1151 if (unlikely(PageSwapCache(page)))
1152 return __page_file_index(page);
1156 bool page_mapped(struct page *page);
1157 struct address_space *page_mapping(struct page *page);
1160 * Return true only if the page has been allocated with
1161 * ALLOC_NO_WATERMARKS and the low watermark was not
1162 * met implying that the system is under some pressure.
1164 static inline bool page_is_pfmemalloc(struct page *page)
1167 * Page index cannot be this large so this must be
1168 * a pfmemalloc page.
1170 return page->index == -1UL;
1174 * Only to be called by the page allocator on a freshly allocated
1177 static inline void set_page_pfmemalloc(struct page *page)
1182 static inline void clear_page_pfmemalloc(struct page *page)
1188 * Different kinds of faults, as returned by handle_mm_fault().
1189 * Used to decide whether a process gets delivered SIGBUS or
1190 * just gets major/minor fault counters bumped up.
1193 #define VM_FAULT_OOM 0x0001
1194 #define VM_FAULT_SIGBUS 0x0002
1195 #define VM_FAULT_MAJOR 0x0004
1196 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1197 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1198 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1199 #define VM_FAULT_SIGSEGV 0x0040
1201 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1202 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1203 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1204 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1205 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1207 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1209 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1210 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1213 #define VM_FAULT_RESULT_TRACE \
1214 { VM_FAULT_OOM, "OOM" }, \
1215 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1216 { VM_FAULT_MAJOR, "MAJOR" }, \
1217 { VM_FAULT_WRITE, "WRITE" }, \
1218 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1219 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1220 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1221 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1222 { VM_FAULT_LOCKED, "LOCKED" }, \
1223 { VM_FAULT_RETRY, "RETRY" }, \
1224 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1225 { VM_FAULT_DONE_COW, "DONE_COW" }
1227 /* Encode hstate index for a hwpoisoned large page */
1228 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1229 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1232 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1234 extern void pagefault_out_of_memory(void);
1236 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1239 * Flags passed to show_mem() and show_free_areas() to suppress output in
1242 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1244 extern void show_free_areas(unsigned int flags, nodemask_t *nodemask);
1246 extern bool can_do_mlock(void);
1247 extern int user_shm_lock(size_t, struct user_struct *);
1248 extern void user_shm_unlock(size_t, struct user_struct *);
1251 * Parameter block passed down to zap_pte_range in exceptional cases.
1253 struct zap_details {
1254 struct address_space *check_mapping; /* Check page->mapping if set */
1255 pgoff_t first_index; /* Lowest page->index to unmap */
1256 pgoff_t last_index; /* Highest page->index to unmap */
1259 struct page *_vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1260 pte_t pte, bool with_public_device);
1261 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1263 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1266 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1267 unsigned long size);
1268 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1269 unsigned long size);
1270 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1271 unsigned long start, unsigned long end);
1274 * mm_walk - callbacks for walk_page_range
1275 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1276 * this handler should only handle pud_trans_huge() puds.
1277 * the pmd_entry or pte_entry callbacks will be used for
1279 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1280 * this handler is required to be able to handle
1281 * pmd_trans_huge() pmds. They may simply choose to
1282 * split_huge_page() instead of handling it explicitly.
1283 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1284 * @pte_hole: if set, called for each hole at all levels
1285 * @hugetlb_entry: if set, called for each hugetlb entry
1286 * @test_walk: caller specific callback function to determine whether
1287 * we walk over the current vma or not. Returning 0
1288 * value means "do page table walk over the current vma,"
1289 * and a negative one means "abort current page table walk
1290 * right now." 1 means "skip the current vma."
1291 * @mm: mm_struct representing the target process of page table walk
1292 * @vma: vma currently walked (NULL if walking outside vmas)
1293 * @private: private data for callbacks' usage
1295 * (see the comment on walk_page_range() for more details)
1298 int (*pud_entry)(pud_t *pud, unsigned long addr,
1299 unsigned long next, struct mm_walk *walk);
1300 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1301 unsigned long next, struct mm_walk *walk);
1302 int (*pte_entry)(pte_t *pte, unsigned long addr,
1303 unsigned long next, struct mm_walk *walk);
1304 int (*pte_hole)(unsigned long addr, unsigned long next,
1305 struct mm_walk *walk);
1306 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1307 unsigned long addr, unsigned long next,
1308 struct mm_walk *walk);
1309 int (*test_walk)(unsigned long addr, unsigned long next,
1310 struct mm_walk *walk);
1311 struct mm_struct *mm;
1312 struct vm_area_struct *vma;
1316 int walk_page_range(unsigned long addr, unsigned long end,
1317 struct mm_walk *walk);
1318 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1319 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1320 unsigned long end, unsigned long floor, unsigned long ceiling);
1321 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1322 struct vm_area_struct *vma);
1323 void unmap_mapping_range(struct address_space *mapping,
1324 loff_t const holebegin, loff_t const holelen, int even_cows);
1325 int follow_pte_pmd(struct mm_struct *mm, unsigned long address,
1326 unsigned long *start, unsigned long *end,
1327 pte_t **ptepp, pmd_t **pmdpp, spinlock_t **ptlp);
1328 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1329 unsigned long *pfn);
1330 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1331 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1332 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1333 void *buf, int len, int write);
1335 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1336 loff_t const holebegin, loff_t const holelen)
1338 unmap_mapping_range(mapping, holebegin, holelen, 0);
1341 extern void truncate_pagecache(struct inode *inode, loff_t new);
1342 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1343 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1344 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1345 int truncate_inode_page(struct address_space *mapping, struct page *page);
1346 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1347 int invalidate_inode_page(struct page *page);
1350 extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address,
1351 unsigned int flags);
1352 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1353 unsigned long address, unsigned int fault_flags,
1356 static inline int handle_mm_fault(struct vm_area_struct *vma,
1357 unsigned long address, unsigned int flags)
1359 /* should never happen if there's no MMU */
1361 return VM_FAULT_SIGBUS;
1363 static inline int fixup_user_fault(struct task_struct *tsk,
1364 struct mm_struct *mm, unsigned long address,
1365 unsigned int fault_flags, bool *unlocked)
1367 /* should never happen if there's no MMU */
1373 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len,
1374 unsigned int gup_flags);
1375 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1376 void *buf, int len, unsigned int gup_flags);
1377 extern int __access_remote_vm(struct task_struct *tsk, struct mm_struct *mm,
1378 unsigned long addr, void *buf, int len, unsigned int gup_flags);
1380 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1381 unsigned long start, unsigned long nr_pages,
1382 unsigned int gup_flags, struct page **pages,
1383 struct vm_area_struct **vmas, int *locked);
1384 long get_user_pages(unsigned long start, unsigned long nr_pages,
1385 unsigned int gup_flags, struct page **pages,
1386 struct vm_area_struct **vmas);
1387 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1388 unsigned int gup_flags, struct page **pages, int *locked);
1389 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1390 struct page **pages, unsigned int gup_flags);
1391 #ifdef CONFIG_FS_DAX
1392 long get_user_pages_longterm(unsigned long start, unsigned long nr_pages,
1393 unsigned int gup_flags, struct page **pages,
1394 struct vm_area_struct **vmas);
1396 static inline long get_user_pages_longterm(unsigned long start,
1397 unsigned long nr_pages, unsigned int gup_flags,
1398 struct page **pages, struct vm_area_struct **vmas)
1400 return get_user_pages(start, nr_pages, gup_flags, pages, vmas);
1402 #endif /* CONFIG_FS_DAX */
1404 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1405 struct page **pages);
1407 /* Container for pinned pfns / pages */
1408 struct frame_vector {
1409 unsigned int nr_allocated; /* Number of frames we have space for */
1410 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1411 bool got_ref; /* Did we pin pages by getting page ref? */
1412 bool is_pfns; /* Does array contain pages or pfns? */
1413 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1414 * pfns_vector_pages() or pfns_vector_pfns()
1418 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1419 void frame_vector_destroy(struct frame_vector *vec);
1420 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1421 unsigned int gup_flags, struct frame_vector *vec);
1422 void put_vaddr_frames(struct frame_vector *vec);
1423 int frame_vector_to_pages(struct frame_vector *vec);
1424 void frame_vector_to_pfns(struct frame_vector *vec);
1426 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1428 return vec->nr_frames;
1431 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1434 int err = frame_vector_to_pages(vec);
1437 return ERR_PTR(err);
1439 return (struct page **)(vec->ptrs);
1442 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1445 frame_vector_to_pfns(vec);
1446 return (unsigned long *)(vec->ptrs);
1450 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1451 struct page **pages);
1452 int get_kernel_page(unsigned long start, int write, struct page **pages);
1453 struct page *get_dump_page(unsigned long addr);
1455 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1456 extern void do_invalidatepage(struct page *page, unsigned int offset,
1457 unsigned int length);
1459 int __set_page_dirty_nobuffers(struct page *page);
1460 int __set_page_dirty_no_writeback(struct page *page);
1461 int redirty_page_for_writepage(struct writeback_control *wbc,
1463 void account_page_dirtied(struct page *page, struct address_space *mapping);
1464 void account_page_cleaned(struct page *page, struct address_space *mapping,
1465 struct bdi_writeback *wb);
1466 int set_page_dirty(struct page *page);
1467 int set_page_dirty_lock(struct page *page);
1468 void cancel_dirty_page(struct page *page);
1469 int clear_page_dirty_for_io(struct page *page);
1471 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1473 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1475 return !vma->vm_ops;
1480 * The vma_is_shmem is not inline because it is used only by slow
1481 * paths in userfault.
1483 bool vma_is_shmem(struct vm_area_struct *vma);
1485 static inline bool vma_is_shmem(struct vm_area_struct *vma) { return false; }
1488 int vma_is_stack_for_current(struct vm_area_struct *vma);
1490 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1491 unsigned long old_addr, struct vm_area_struct *new_vma,
1492 unsigned long new_addr, unsigned long len,
1493 bool need_rmap_locks);
1494 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1495 unsigned long end, pgprot_t newprot,
1496 int dirty_accountable, int prot_numa);
1497 extern int mprotect_fixup(struct vm_area_struct *vma,
1498 struct vm_area_struct **pprev, unsigned long start,
1499 unsigned long end, unsigned long newflags);
1502 * doesn't attempt to fault and will return short.
1504 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1505 struct page **pages);
1507 * per-process(per-mm_struct) statistics.
1509 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1511 long val = atomic_long_read(&mm->rss_stat.count[member]);
1513 #ifdef SPLIT_RSS_COUNTING
1515 * counter is updated in asynchronous manner and may go to minus.
1516 * But it's never be expected number for users.
1521 return (unsigned long)val;
1524 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1526 atomic_long_add(value, &mm->rss_stat.count[member]);
1529 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1531 atomic_long_inc(&mm->rss_stat.count[member]);
1534 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1536 atomic_long_dec(&mm->rss_stat.count[member]);
1539 /* Optimized variant when page is already known not to be PageAnon */
1540 static inline int mm_counter_file(struct page *page)
1542 if (PageSwapBacked(page))
1543 return MM_SHMEMPAGES;
1544 return MM_FILEPAGES;
1547 static inline int mm_counter(struct page *page)
1550 return MM_ANONPAGES;
1551 return mm_counter_file(page);
1554 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1556 return get_mm_counter(mm, MM_FILEPAGES) +
1557 get_mm_counter(mm, MM_ANONPAGES) +
1558 get_mm_counter(mm, MM_SHMEMPAGES);
1561 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1563 return max(mm->hiwater_rss, get_mm_rss(mm));
1566 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1568 return max(mm->hiwater_vm, mm->total_vm);
1571 static inline void update_hiwater_rss(struct mm_struct *mm)
1573 unsigned long _rss = get_mm_rss(mm);
1575 if ((mm)->hiwater_rss < _rss)
1576 (mm)->hiwater_rss = _rss;
1579 static inline void update_hiwater_vm(struct mm_struct *mm)
1581 if (mm->hiwater_vm < mm->total_vm)
1582 mm->hiwater_vm = mm->total_vm;
1585 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1587 mm->hiwater_rss = get_mm_rss(mm);
1590 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1591 struct mm_struct *mm)
1593 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1595 if (*maxrss < hiwater_rss)
1596 *maxrss = hiwater_rss;
1599 #if defined(SPLIT_RSS_COUNTING)
1600 void sync_mm_rss(struct mm_struct *mm);
1602 static inline void sync_mm_rss(struct mm_struct *mm)
1607 #ifndef __HAVE_ARCH_PTE_DEVMAP
1608 static inline int pte_devmap(pte_t pte)
1614 int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot);
1616 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1618 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1622 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1626 #ifdef __PAGETABLE_P4D_FOLDED
1627 static inline int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1628 unsigned long address)
1633 int __p4d_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1636 #ifdef __PAGETABLE_PUD_FOLDED
1637 static inline int __pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1638 unsigned long address)
1643 int __pud_alloc(struct mm_struct *mm, p4d_t *p4d, unsigned long address);
1646 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1647 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1648 unsigned long address)
1653 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1655 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1660 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1661 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1664 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1666 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1668 atomic_long_set(&mm->nr_pmds, 0);
1671 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1673 return atomic_long_read(&mm->nr_pmds);
1676 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1678 atomic_long_inc(&mm->nr_pmds);
1681 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1683 atomic_long_dec(&mm->nr_pmds);
1687 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1688 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1691 * The following ifdef needed to get the 4level-fixup.h header to work.
1692 * Remove it when 4level-fixup.h has been removed.
1694 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1696 #ifndef __ARCH_HAS_5LEVEL_HACK
1697 static inline p4d_t *p4d_alloc(struct mm_struct *mm, pgd_t *pgd,
1698 unsigned long address)
1700 return (unlikely(pgd_none(*pgd)) && __p4d_alloc(mm, pgd, address)) ?
1701 NULL : p4d_offset(pgd, address);
1704 static inline pud_t *pud_alloc(struct mm_struct *mm, p4d_t *p4d,
1705 unsigned long address)
1707 return (unlikely(p4d_none(*p4d)) && __pud_alloc(mm, p4d, address)) ?
1708 NULL : pud_offset(p4d, address);
1710 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1712 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1714 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1715 NULL: pmd_offset(pud, address);
1717 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1719 #if USE_SPLIT_PTE_PTLOCKS
1720 #if ALLOC_SPLIT_PTLOCKS
1721 void __init ptlock_cache_init(void);
1722 extern bool ptlock_alloc(struct page *page);
1723 extern void ptlock_free(struct page *page);
1725 static inline spinlock_t *ptlock_ptr(struct page *page)
1729 #else /* ALLOC_SPLIT_PTLOCKS */
1730 static inline void ptlock_cache_init(void)
1734 static inline bool ptlock_alloc(struct page *page)
1739 static inline void ptlock_free(struct page *page)
1743 static inline spinlock_t *ptlock_ptr(struct page *page)
1747 #endif /* ALLOC_SPLIT_PTLOCKS */
1749 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1751 return ptlock_ptr(pmd_page(*pmd));
1754 static inline bool ptlock_init(struct page *page)
1757 * prep_new_page() initialize page->private (and therefore page->ptl)
1758 * with 0. Make sure nobody took it in use in between.
1760 * It can happen if arch try to use slab for page table allocation:
1761 * slab code uses page->slab_cache, which share storage with page->ptl.
1763 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1764 if (!ptlock_alloc(page))
1766 spin_lock_init(ptlock_ptr(page));
1770 /* Reset page->mapping so free_pages_check won't complain. */
1771 static inline void pte_lock_deinit(struct page *page)
1773 page->mapping = NULL;
1777 #else /* !USE_SPLIT_PTE_PTLOCKS */
1779 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1781 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1783 return &mm->page_table_lock;
1785 static inline void ptlock_cache_init(void) {}
1786 static inline bool ptlock_init(struct page *page) { return true; }
1787 static inline void pte_lock_deinit(struct page *page) {}
1788 #endif /* USE_SPLIT_PTE_PTLOCKS */
1790 static inline void pgtable_init(void)
1792 ptlock_cache_init();
1793 pgtable_cache_init();
1796 static inline bool pgtable_page_ctor(struct page *page)
1798 if (!ptlock_init(page))
1800 inc_zone_page_state(page, NR_PAGETABLE);
1804 static inline void pgtable_page_dtor(struct page *page)
1806 pte_lock_deinit(page);
1807 dec_zone_page_state(page, NR_PAGETABLE);
1810 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1812 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1813 pte_t *__pte = pte_offset_map(pmd, address); \
1819 #define pte_unmap_unlock(pte, ptl) do { \
1824 #define pte_alloc(mm, pmd, address) \
1825 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1827 #define pte_alloc_map(mm, pmd, address) \
1828 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1830 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1831 (pte_alloc(mm, pmd, address) ? \
1832 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1834 #define pte_alloc_kernel(pmd, address) \
1835 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1836 NULL: pte_offset_kernel(pmd, address))
1838 #if USE_SPLIT_PMD_PTLOCKS
1840 static struct page *pmd_to_page(pmd_t *pmd)
1842 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1843 return virt_to_page((void *)((unsigned long) pmd & mask));
1846 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1848 return ptlock_ptr(pmd_to_page(pmd));
1851 static inline bool pgtable_pmd_page_ctor(struct page *page)
1853 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1854 page->pmd_huge_pte = NULL;
1856 return ptlock_init(page);
1859 static inline void pgtable_pmd_page_dtor(struct page *page)
1861 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1862 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1867 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1871 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1873 return &mm->page_table_lock;
1876 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1877 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1879 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1883 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1885 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1891 * No scalability reason to split PUD locks yet, but follow the same pattern
1892 * as the PMD locks to make it easier if we decide to. The VM should not be
1893 * considered ready to switch to split PUD locks yet; there may be places
1894 * which need to be converted from page_table_lock.
1896 static inline spinlock_t *pud_lockptr(struct mm_struct *mm, pud_t *pud)
1898 return &mm->page_table_lock;
1901 static inline spinlock_t *pud_lock(struct mm_struct *mm, pud_t *pud)
1903 spinlock_t *ptl = pud_lockptr(mm, pud);
1909 extern void __init pagecache_init(void);
1910 extern void free_area_init(unsigned long * zones_size);
1911 extern void free_area_init_node(int nid, unsigned long * zones_size,
1912 unsigned long zone_start_pfn, unsigned long *zholes_size);
1913 extern void free_initmem(void);
1916 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1917 * into the buddy system. The freed pages will be poisoned with pattern
1918 * "poison" if it's within range [0, UCHAR_MAX].
1919 * Return pages freed into the buddy system.
1921 extern unsigned long free_reserved_area(void *start, void *end,
1922 int poison, char *s);
1924 #ifdef CONFIG_HIGHMEM
1926 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1927 * and totalram_pages.
1929 extern void free_highmem_page(struct page *page);
1932 extern void adjust_managed_page_count(struct page *page, long count);
1933 extern void mem_init_print_info(const char *str);
1935 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1937 /* Free the reserved page into the buddy system, so it gets managed. */
1938 static inline void __free_reserved_page(struct page *page)
1940 ClearPageReserved(page);
1941 init_page_count(page);
1945 static inline void free_reserved_page(struct page *page)
1947 __free_reserved_page(page);
1948 adjust_managed_page_count(page, 1);
1951 static inline void mark_page_reserved(struct page *page)
1953 SetPageReserved(page);
1954 adjust_managed_page_count(page, -1);
1958 * Default method to free all the __init memory into the buddy system.
1959 * The freed pages will be poisoned with pattern "poison" if it's within
1960 * range [0, UCHAR_MAX].
1961 * Return pages freed into the buddy system.
1963 static inline unsigned long free_initmem_default(int poison)
1965 extern char __init_begin[], __init_end[];
1967 return free_reserved_area(&__init_begin, &__init_end,
1968 poison, "unused kernel");
1971 static inline unsigned long get_num_physpages(void)
1974 unsigned long phys_pages = 0;
1976 for_each_online_node(nid)
1977 phys_pages += node_present_pages(nid);
1982 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1984 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1985 * zones, allocate the backing mem_map and account for memory holes in a more
1986 * architecture independent manner. This is a substitute for creating the
1987 * zone_sizes[] and zholes_size[] arrays and passing them to
1988 * free_area_init_node()
1990 * An architecture is expected to register range of page frames backed by
1991 * physical memory with memblock_add[_node]() before calling
1992 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1993 * usage, an architecture is expected to do something like
1995 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1997 * for_each_valid_physical_page_range()
1998 * memblock_add_node(base, size, nid)
1999 * free_area_init_nodes(max_zone_pfns);
2001 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2002 * registered physical page range. Similarly
2003 * sparse_memory_present_with_active_regions() calls memory_present() for
2004 * each range when SPARSEMEM is enabled.
2006 * See mm/page_alloc.c for more information on each function exposed by
2007 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2009 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
2010 unsigned long node_map_pfn_alignment(void);
2011 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
2012 unsigned long end_pfn);
2013 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
2014 unsigned long end_pfn);
2015 extern void get_pfn_range_for_nid(unsigned int nid,
2016 unsigned long *start_pfn, unsigned long *end_pfn);
2017 extern unsigned long find_min_pfn_with_active_regions(void);
2018 extern void free_bootmem_with_active_regions(int nid,
2019 unsigned long max_low_pfn);
2020 extern void sparse_memory_present_with_active_regions(int nid);
2022 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2024 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2025 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2026 static inline int __early_pfn_to_nid(unsigned long pfn,
2027 struct mminit_pfnnid_cache *state)
2032 /* please see mm/page_alloc.c */
2033 extern int __meminit early_pfn_to_nid(unsigned long pfn);
2034 /* there is a per-arch backend function. */
2035 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
2036 struct mminit_pfnnid_cache *state);
2039 extern void set_dma_reserve(unsigned long new_dma_reserve);
2040 extern void memmap_init_zone(unsigned long, int, unsigned long,
2041 unsigned long, enum memmap_context);
2042 extern void setup_per_zone_wmarks(void);
2043 extern int __meminit init_per_zone_wmark_min(void);
2044 extern void mem_init(void);
2045 extern void __init mmap_init(void);
2046 extern void show_mem(unsigned int flags, nodemask_t *nodemask);
2047 extern long si_mem_available(void);
2048 extern void si_meminfo(struct sysinfo * val);
2049 extern void si_meminfo_node(struct sysinfo *val, int nid);
2050 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2051 extern unsigned long arch_reserved_kernel_pages(void);
2054 extern __printf(3, 4)
2055 void warn_alloc(gfp_t gfp_mask, nodemask_t *nodemask, const char *fmt, ...);
2057 extern void setup_per_cpu_pageset(void);
2059 extern void zone_pcp_update(struct zone *zone);
2060 extern void zone_pcp_reset(struct zone *zone);
2063 extern int min_free_kbytes;
2064 extern int watermark_scale_factor;
2067 extern atomic_long_t mmap_pages_allocated;
2068 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
2070 /* interval_tree.c */
2071 void vma_interval_tree_insert(struct vm_area_struct *node,
2072 struct rb_root_cached *root);
2073 void vma_interval_tree_insert_after(struct vm_area_struct *node,
2074 struct vm_area_struct *prev,
2075 struct rb_root_cached *root);
2076 void vma_interval_tree_remove(struct vm_area_struct *node,
2077 struct rb_root_cached *root);
2078 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root_cached *root,
2079 unsigned long start, unsigned long last);
2080 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
2081 unsigned long start, unsigned long last);
2083 #define vma_interval_tree_foreach(vma, root, start, last) \
2084 for (vma = vma_interval_tree_iter_first(root, start, last); \
2085 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2087 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
2088 struct rb_root_cached *root);
2089 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
2090 struct rb_root_cached *root);
2091 struct anon_vma_chain *
2092 anon_vma_interval_tree_iter_first(struct rb_root_cached *root,
2093 unsigned long start, unsigned long last);
2094 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
2095 struct anon_vma_chain *node, unsigned long start, unsigned long last);
2096 #ifdef CONFIG_DEBUG_VM_RB
2097 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
2100 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2101 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2102 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2105 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2106 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
2107 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert,
2108 struct vm_area_struct *expand);
2109 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2110 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
2112 return __vma_adjust(vma, start, end, pgoff, insert, NULL);
2114 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2115 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2116 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2117 struct mempolicy *, struct vm_userfaultfd_ctx);
2118 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2119 extern int __split_vma(struct mm_struct *, struct vm_area_struct *,
2120 unsigned long addr, int new_below);
2121 extern int split_vma(struct mm_struct *, struct vm_area_struct *,
2122 unsigned long addr, int new_below);
2123 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2124 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2125 struct rb_node **, struct rb_node *);
2126 extern void unlink_file_vma(struct vm_area_struct *);
2127 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2128 unsigned long addr, unsigned long len, pgoff_t pgoff,
2129 bool *need_rmap_locks);
2130 extern void exit_mmap(struct mm_struct *);
2132 static inline int check_data_rlimit(unsigned long rlim,
2134 unsigned long start,
2135 unsigned long end_data,
2136 unsigned long start_data)
2138 if (rlim < RLIM_INFINITY) {
2139 if (((new - start) + (end_data - start_data)) > rlim)
2146 extern int mm_take_all_locks(struct mm_struct *mm);
2147 extern void mm_drop_all_locks(struct mm_struct *mm);
2149 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2150 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2151 extern struct file *get_task_exe_file(struct task_struct *task);
2153 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2154 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2156 extern bool vma_is_special_mapping(const struct vm_area_struct *vma,
2157 const struct vm_special_mapping *sm);
2158 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2159 unsigned long addr, unsigned long len,
2160 unsigned long flags,
2161 const struct vm_special_mapping *spec);
2162 /* This is an obsolete alternative to _install_special_mapping. */
2163 extern int install_special_mapping(struct mm_struct *mm,
2164 unsigned long addr, unsigned long len,
2165 unsigned long flags, struct page **pages);
2167 unsigned long randomize_page(unsigned long start, unsigned long range);
2169 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2171 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2172 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff,
2173 struct list_head *uf);
2174 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2175 unsigned long len, unsigned long prot, unsigned long flags,
2176 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate,
2177 struct list_head *uf);
2178 extern int do_munmap(struct mm_struct *, unsigned long, size_t,
2179 struct list_head *uf);
2181 static inline unsigned long
2182 do_mmap_pgoff(struct file *file, unsigned long addr,
2183 unsigned long len, unsigned long prot, unsigned long flags,
2184 unsigned long pgoff, unsigned long *populate,
2185 struct list_head *uf)
2187 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate, uf);
2191 extern int __mm_populate(unsigned long addr, unsigned long len,
2193 static inline void mm_populate(unsigned long addr, unsigned long len)
2196 (void) __mm_populate(addr, len, 1);
2199 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2202 /* These take the mm semaphore themselves */
2203 extern int __must_check vm_brk(unsigned long, unsigned long);
2204 extern int __must_check vm_brk_flags(unsigned long, unsigned long, unsigned long);
2205 extern int vm_munmap(unsigned long, size_t);
2206 extern unsigned long __must_check vm_mmap(struct file *, unsigned long,
2207 unsigned long, unsigned long,
2208 unsigned long, unsigned long);
2210 struct vm_unmapped_area_info {
2211 #define VM_UNMAPPED_AREA_TOPDOWN 1
2212 unsigned long flags;
2213 unsigned long length;
2214 unsigned long low_limit;
2215 unsigned long high_limit;
2216 unsigned long align_mask;
2217 unsigned long align_offset;
2220 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2221 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2224 * Search for an unmapped address range.
2226 * We are looking for a range that:
2227 * - does not intersect with any VMA;
2228 * - is contained within the [low_limit, high_limit) interval;
2229 * - is at least the desired size.
2230 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2232 static inline unsigned long
2233 vm_unmapped_area(struct vm_unmapped_area_info *info)
2235 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2236 return unmapped_area_topdown(info);
2238 return unmapped_area(info);
2242 extern void truncate_inode_pages(struct address_space *, loff_t);
2243 extern void truncate_inode_pages_range(struct address_space *,
2244 loff_t lstart, loff_t lend);
2245 extern void truncate_inode_pages_final(struct address_space *);
2247 /* generic vm_area_ops exported for stackable file systems */
2248 extern int filemap_fault(struct vm_fault *vmf);
2249 extern void filemap_map_pages(struct vm_fault *vmf,
2250 pgoff_t start_pgoff, pgoff_t end_pgoff);
2251 extern int filemap_page_mkwrite(struct vm_fault *vmf);
2253 /* mm/page-writeback.c */
2254 int __must_check write_one_page(struct page *page);
2255 void task_dirty_inc(struct task_struct *tsk);
2258 #define VM_MAX_READAHEAD 128 /* kbytes */
2259 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2261 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2262 pgoff_t offset, unsigned long nr_to_read);
2264 void page_cache_sync_readahead(struct address_space *mapping,
2265 struct file_ra_state *ra,
2268 unsigned long size);
2270 void page_cache_async_readahead(struct address_space *mapping,
2271 struct file_ra_state *ra,
2275 unsigned long size);
2277 extern unsigned long stack_guard_gap;
2278 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2279 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2281 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2282 extern int expand_downwards(struct vm_area_struct *vma,
2283 unsigned long address);
2285 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2287 #define expand_upwards(vma, address) (0)
2290 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2291 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2292 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2293 struct vm_area_struct **pprev);
2295 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2296 NULL if none. Assume start_addr < end_addr. */
2297 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2299 struct vm_area_struct * vma = find_vma(mm,start_addr);
2301 if (vma && end_addr <= vma->vm_start)
2306 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2308 unsigned long vm_start = vma->vm_start;
2310 if (vma->vm_flags & VM_GROWSDOWN) {
2311 vm_start -= stack_guard_gap;
2312 if (vm_start > vma->vm_start)
2318 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2320 unsigned long vm_end = vma->vm_end;
2322 if (vma->vm_flags & VM_GROWSUP) {
2323 vm_end += stack_guard_gap;
2324 if (vm_end < vma->vm_end)
2325 vm_end = -PAGE_SIZE;
2330 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2332 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2335 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2336 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2337 unsigned long vm_start, unsigned long vm_end)
2339 struct vm_area_struct *vma = find_vma(mm, vm_start);
2341 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2347 static inline bool range_in_vma(struct vm_area_struct *vma,
2348 unsigned long start, unsigned long end)
2350 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2354 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2355 void vma_set_page_prot(struct vm_area_struct *vma);
2357 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2361 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2363 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2367 #ifdef CONFIG_NUMA_BALANCING
2368 unsigned long change_prot_numa(struct vm_area_struct *vma,
2369 unsigned long start, unsigned long end);
2372 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2373 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2374 unsigned long pfn, unsigned long size, pgprot_t);
2375 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2376 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2378 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2379 unsigned long pfn, pgprot_t pgprot);
2380 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2382 int vm_insert_mixed_mkwrite(struct vm_area_struct *vma, unsigned long addr,
2384 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2387 struct page *follow_page_mask(struct vm_area_struct *vma,
2388 unsigned long address, unsigned int foll_flags,
2389 unsigned int *page_mask);
2391 static inline struct page *follow_page(struct vm_area_struct *vma,
2392 unsigned long address, unsigned int foll_flags)
2394 unsigned int unused_page_mask;
2395 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2398 #define FOLL_WRITE 0x01 /* check pte is writable */
2399 #define FOLL_TOUCH 0x02 /* mark page accessed */
2400 #define FOLL_GET 0x04 /* do get_page on page */
2401 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2402 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2403 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2404 * and return without waiting upon it */
2405 #define FOLL_POPULATE 0x40 /* fault in page */
2406 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2407 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2408 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2409 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2410 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2411 #define FOLL_MLOCK 0x1000 /* lock present pages */
2412 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2413 #define FOLL_COW 0x4000 /* internal GUP flag */
2414 #define FOLL_ANON 0x8000 /* don't do file mappings */
2416 static inline int vm_fault_to_errno(int vm_fault, int foll_flags)
2418 if (vm_fault & VM_FAULT_OOM)
2420 if (vm_fault & (VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE))
2421 return (foll_flags & FOLL_HWPOISON) ? -EHWPOISON : -EFAULT;
2422 if (vm_fault & (VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV))
2427 #ifndef io_remap_pfn_range
2428 static inline int io_remap_pfn_range(struct vm_area_struct *vma,
2429 unsigned long addr, unsigned long pfn,
2430 unsigned long size, pgprot_t prot)
2432 return remap_pfn_range(vma, addr, pfn, size, pgprot_decrypted(prot));
2436 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2438 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2439 unsigned long size, pte_fn_t fn, void *data);
2442 #ifdef CONFIG_PAGE_POISONING
2443 extern bool page_poisoning_enabled(void);
2444 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2445 extern bool page_is_poisoned(struct page *page);
2447 static inline bool page_poisoning_enabled(void) { return false; }
2448 static inline void kernel_poison_pages(struct page *page, int numpages,
2450 static inline bool page_is_poisoned(struct page *page) { return false; }
2453 #ifdef CONFIG_DEBUG_PAGEALLOC
2454 extern bool _debug_pagealloc_enabled;
2455 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2457 static inline bool debug_pagealloc_enabled(void)
2459 return _debug_pagealloc_enabled;
2463 kernel_map_pages(struct page *page, int numpages, int enable)
2465 if (!debug_pagealloc_enabled())
2468 __kernel_map_pages(page, numpages, enable);
2470 #ifdef CONFIG_HIBERNATION
2471 extern bool kernel_page_present(struct page *page);
2472 #endif /* CONFIG_HIBERNATION */
2473 #else /* CONFIG_DEBUG_PAGEALLOC */
2475 kernel_map_pages(struct page *page, int numpages, int enable) {}
2476 #ifdef CONFIG_HIBERNATION
2477 static inline bool kernel_page_present(struct page *page) { return true; }
2478 #endif /* CONFIG_HIBERNATION */
2479 static inline bool debug_pagealloc_enabled(void)
2483 #endif /* CONFIG_DEBUG_PAGEALLOC */
2485 #ifdef __HAVE_ARCH_GATE_AREA
2486 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2487 extern int in_gate_area_no_mm(unsigned long addr);
2488 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2490 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2494 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2495 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2499 #endif /* __HAVE_ARCH_GATE_AREA */
2501 extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm);
2503 #ifdef CONFIG_SYSCTL
2504 extern int sysctl_drop_caches;
2505 int drop_caches_sysctl_handler(struct ctl_table *, int,
2506 void __user *, size_t *, loff_t *);
2509 void drop_slab(void);
2510 void drop_slab_node(int nid);
2513 #define randomize_va_space 0
2515 extern int randomize_va_space;
2518 const char * arch_vma_name(struct vm_area_struct *vma);
2519 void print_vma_addr(char *prefix, unsigned long rip);
2521 void sparse_mem_maps_populate_node(struct page **map_map,
2522 unsigned long pnum_begin,
2523 unsigned long pnum_end,
2524 unsigned long map_count,
2527 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2528 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2529 p4d_t *vmemmap_p4d_populate(pgd_t *pgd, unsigned long addr, int node);
2530 pud_t *vmemmap_pud_populate(p4d_t *p4d, unsigned long addr, int node);
2531 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2532 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2533 void *vmemmap_alloc_block(unsigned long size, int node);
2535 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2536 struct vmem_altmap *altmap);
2537 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2539 return __vmemmap_alloc_block_buf(size, node, NULL);
2542 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2543 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2545 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2546 void vmemmap_populate_print_last(void);
2547 #ifdef CONFIG_MEMORY_HOTPLUG
2548 void vmemmap_free(unsigned long start, unsigned long end);
2550 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2551 unsigned long nr_pages);
2554 MF_COUNT_INCREASED = 1 << 0,
2555 MF_ACTION_REQUIRED = 1 << 1,
2556 MF_MUST_KILL = 1 << 2,
2557 MF_SOFT_OFFLINE = 1 << 3,
2559 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2560 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2561 extern int unpoison_memory(unsigned long pfn);
2562 extern int get_hwpoison_page(struct page *page);
2563 #define put_hwpoison_page(page) put_page(page)
2564 extern int sysctl_memory_failure_early_kill;
2565 extern int sysctl_memory_failure_recovery;
2566 extern void shake_page(struct page *p, int access);
2567 extern atomic_long_t num_poisoned_pages;
2568 extern int soft_offline_page(struct page *page, int flags);
2572 * Error handlers for various types of pages.
2575 MF_IGNORED, /* Error: cannot be handled */
2576 MF_FAILED, /* Error: handling failed */
2577 MF_DELAYED, /* Will be handled later */
2578 MF_RECOVERED, /* Successfully recovered */
2581 enum mf_action_page_type {
2583 MF_MSG_KERNEL_HIGH_ORDER,
2585 MF_MSG_DIFFERENT_COMPOUND,
2586 MF_MSG_POISONED_HUGE,
2589 MF_MSG_NON_PMD_HUGE,
2590 MF_MSG_UNMAP_FAILED,
2591 MF_MSG_DIRTY_SWAPCACHE,
2592 MF_MSG_CLEAN_SWAPCACHE,
2593 MF_MSG_DIRTY_MLOCKED_LRU,
2594 MF_MSG_CLEAN_MLOCKED_LRU,
2595 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2596 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2599 MF_MSG_TRUNCATED_LRU,
2605 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2606 extern void clear_huge_page(struct page *page,
2607 unsigned long addr_hint,
2608 unsigned int pages_per_huge_page);
2609 extern void copy_user_huge_page(struct page *dst, struct page *src,
2610 unsigned long addr, struct vm_area_struct *vma,
2611 unsigned int pages_per_huge_page);
2612 extern long copy_huge_page_from_user(struct page *dst_page,
2613 const void __user *usr_src,
2614 unsigned int pages_per_huge_page,
2615 bool allow_pagefault);
2616 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2618 extern struct page_ext_operations debug_guardpage_ops;
2620 #ifdef CONFIG_DEBUG_PAGEALLOC
2621 extern unsigned int _debug_guardpage_minorder;
2622 extern bool _debug_guardpage_enabled;
2624 static inline unsigned int debug_guardpage_minorder(void)
2626 return _debug_guardpage_minorder;
2629 static inline bool debug_guardpage_enabled(void)
2631 return _debug_guardpage_enabled;
2634 static inline bool page_is_guard(struct page *page)
2636 struct page_ext *page_ext;
2638 if (!debug_guardpage_enabled())
2641 page_ext = lookup_page_ext(page);
2642 if (unlikely(!page_ext))
2645 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2648 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2649 static inline bool debug_guardpage_enabled(void) { return false; }
2650 static inline bool page_is_guard(struct page *page) { return false; }
2651 #endif /* CONFIG_DEBUG_PAGEALLOC */
2653 #if MAX_NUMNODES > 1
2654 void __init setup_nr_node_ids(void);
2656 static inline void setup_nr_node_ids(void) {}
2659 #endif /* __KERNEL__ */
2660 #endif /* _LINUX_MM_H */