4 #include <linux/errno.h>
8 #include <linux/mmdebug.h>
10 #include <linux/bug.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/percpu-refcount.h>
20 #include <linux/bit_spinlock.h>
21 #include <linux/shrinker.h>
22 #include <linux/resource.h>
23 #include <linux/page_ext.h>
24 #include <linux/err.h>
25 #include <linux/page_ref.h>
29 struct anon_vma_chain;
32 struct writeback_control;
35 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
36 extern unsigned long max_mapnr;
38 static inline void set_max_mapnr(unsigned long limit)
43 static inline void set_max_mapnr(unsigned long limit) { }
46 extern unsigned long totalram_pages;
47 extern void * high_memory;
48 extern int page_cluster;
51 extern int sysctl_legacy_va_layout;
53 #define sysctl_legacy_va_layout 0
56 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
57 extern const int mmap_rnd_bits_min;
58 extern const int mmap_rnd_bits_max;
59 extern int mmap_rnd_bits __read_mostly;
61 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
62 extern const int mmap_rnd_compat_bits_min;
63 extern const int mmap_rnd_compat_bits_max;
64 extern int mmap_rnd_compat_bits __read_mostly;
68 #include <asm/pgtable.h>
69 #include <asm/processor.h>
72 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
76 * To prevent common memory management code establishing
77 * a zero page mapping on a read fault.
78 * This macro should be defined within <asm/pgtable.h>.
79 * s390 does this to prevent multiplexing of hardware bits
80 * related to the physical page in case of virtualization.
82 #ifndef mm_forbids_zeropage
83 #define mm_forbids_zeropage(X) (0)
87 * Default maximum number of active map areas, this limits the number of vmas
88 * per mm struct. Users can overwrite this number by sysctl but there is a
91 * When a program's coredump is generated as ELF format, a section is created
92 * per a vma. In ELF, the number of sections is represented in unsigned short.
93 * This means the number of sections should be smaller than 65535 at coredump.
94 * Because the kernel adds some informative sections to a image of program at
95 * generating coredump, we need some margin. The number of extra sections is
96 * 1-3 now and depends on arch. We use "5" as safe margin, here.
98 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
99 * not a hard limit any more. Although some userspace tools can be surprised by
102 #define MAPCOUNT_ELF_CORE_MARGIN (5)
103 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
105 extern int sysctl_max_map_count;
107 extern unsigned long sysctl_user_reserve_kbytes;
108 extern unsigned long sysctl_admin_reserve_kbytes;
110 extern int sysctl_overcommit_memory;
111 extern int sysctl_overcommit_ratio;
112 extern unsigned long sysctl_overcommit_kbytes;
114 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
116 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
119 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
121 /* to align the pointer to the (next) page boundary */
122 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
124 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
125 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
128 * Linux kernel virtual memory manager primitives.
129 * The idea being to have a "virtual" mm in the same way
130 * we have a virtual fs - giving a cleaner interface to the
131 * mm details, and allowing different kinds of memory mappings
132 * (from shared memory to executable loading to arbitrary
136 extern struct kmem_cache *vm_area_cachep;
139 extern struct rb_root nommu_region_tree;
140 extern struct rw_semaphore nommu_region_sem;
142 extern unsigned int kobjsize(const void *objp);
146 * vm_flags in vm_area_struct, see mm_types.h.
147 * When changing, update also include/trace/events/mmflags.h
149 #define VM_NONE 0x00000000
151 #define VM_READ 0x00000001 /* currently active flags */
152 #define VM_WRITE 0x00000002
153 #define VM_EXEC 0x00000004
154 #define VM_SHARED 0x00000008
156 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
157 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
158 #define VM_MAYWRITE 0x00000020
159 #define VM_MAYEXEC 0x00000040
160 #define VM_MAYSHARE 0x00000080
162 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
163 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
164 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
165 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
166 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
168 #define VM_LOCKED 0x00002000
169 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
171 /* Used by sys_madvise() */
172 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
173 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
175 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
176 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
177 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
178 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
179 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
180 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
181 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
182 #define VM_ARCH_2 0x02000000
183 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
185 #ifdef CONFIG_MEM_SOFT_DIRTY
186 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
188 # define VM_SOFTDIRTY 0
191 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
192 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
193 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
194 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
196 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
197 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
198 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
199 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
200 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
201 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
202 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
203 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
204 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
205 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
207 #if defined(CONFIG_X86)
208 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
209 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
210 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
211 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
212 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
213 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
214 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
216 #elif defined(CONFIG_PPC)
217 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
218 #elif defined(CONFIG_PARISC)
219 # define VM_GROWSUP VM_ARCH_1
220 #elif defined(CONFIG_METAG)
221 # define VM_GROWSUP VM_ARCH_1
222 #elif defined(CONFIG_IA64)
223 # define VM_GROWSUP VM_ARCH_1
224 #elif !defined(CONFIG_MMU)
225 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
228 #if defined(CONFIG_X86)
229 /* MPX specific bounds table or bounds directory */
230 # define VM_MPX VM_ARCH_2
234 # define VM_GROWSUP VM_NONE
237 /* Bits set in the VMA until the stack is in its final location */
238 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
240 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
241 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
244 #ifdef CONFIG_STACK_GROWSUP
245 #define VM_STACK VM_GROWSUP
247 #define VM_STACK VM_GROWSDOWN
250 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
253 * Special vmas that are non-mergable, non-mlock()able.
254 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
256 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
258 /* This mask defines which mm->def_flags a process can inherit its parent */
259 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
261 /* This mask is used to clear all the VMA flags used by mlock */
262 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
265 * mapping from the currently active vm_flags protection bits (the
266 * low four bits) to a page protection mask..
268 extern pgprot_t protection_map[16];
270 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
271 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
272 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
273 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
274 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
275 #define FAULT_FLAG_TRIED 0x20 /* Second try */
276 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
277 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
278 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
281 * vm_fault is filled by the the pagefault handler and passed to the vma's
282 * ->fault function. The vma's ->fault is responsible for returning a bitmask
283 * of VM_FAULT_xxx flags that give details about how the fault was handled.
285 * MM layer fills up gfp_mask for page allocations but fault handler might
286 * alter it if its implementation requires a different allocation context.
288 * pgoff should be used in favour of virtual_address, if possible.
291 unsigned int flags; /* FAULT_FLAG_xxx flags */
292 gfp_t gfp_mask; /* gfp mask to be used for allocations */
293 pgoff_t pgoff; /* Logical page offset based on vma */
294 void __user *virtual_address; /* Faulting virtual address */
296 struct page *cow_page; /* Handler may choose to COW */
297 struct page *page; /* ->fault handlers should return a
298 * page here, unless VM_FAULT_NOPAGE
299 * is set (which is also implied by
302 /* for ->map_pages() only */
303 pgoff_t max_pgoff; /* map pages for offset from pgoff till
304 * max_pgoff inclusive */
305 pte_t *pte; /* pte entry associated with ->pgoff */
309 * These are the virtual MM functions - opening of an area, closing and
310 * unmapping it (needed to keep files on disk up-to-date etc), pointer
311 * to the functions called when a no-page or a wp-page exception occurs.
313 struct vm_operations_struct {
314 void (*open)(struct vm_area_struct * area);
315 void (*close)(struct vm_area_struct * area);
316 int (*mremap)(struct vm_area_struct * area);
317 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
318 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
319 pmd_t *, unsigned int flags);
320 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
322 /* notification that a previously read-only page is about to become
323 * writable, if an error is returned it will cause a SIGBUS */
324 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
326 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
327 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
329 /* called by access_process_vm when get_user_pages() fails, typically
330 * for use by special VMAs that can switch between memory and hardware
332 int (*access)(struct vm_area_struct *vma, unsigned long addr,
333 void *buf, int len, int write);
335 /* Called by the /proc/PID/maps code to ask the vma whether it
336 * has a special name. Returning non-NULL will also cause this
337 * vma to be dumped unconditionally. */
338 const char *(*name)(struct vm_area_struct *vma);
342 * set_policy() op must add a reference to any non-NULL @new mempolicy
343 * to hold the policy upon return. Caller should pass NULL @new to
344 * remove a policy and fall back to surrounding context--i.e. do not
345 * install a MPOL_DEFAULT policy, nor the task or system default
348 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
351 * get_policy() op must add reference [mpol_get()] to any policy at
352 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
353 * in mm/mempolicy.c will do this automatically.
354 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
355 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
356 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
357 * must return NULL--i.e., do not "fallback" to task or system default
360 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
364 * Called by vm_normal_page() for special PTEs to find the
365 * page for @addr. This is useful if the default behavior
366 * (using pte_page()) would not find the correct page.
368 struct page *(*find_special_page)(struct vm_area_struct *vma,
375 #define page_private(page) ((page)->private)
376 #define set_page_private(page, v) ((page)->private = (v))
378 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
379 static inline int pmd_devmap(pmd_t pmd)
386 * FIXME: take this include out, include page-flags.h in
387 * files which need it (119 of them)
389 #include <linux/page-flags.h>
390 #include <linux/huge_mm.h>
393 * Methods to modify the page usage count.
395 * What counts for a page usage:
396 * - cache mapping (page->mapping)
397 * - private data (page->private)
398 * - page mapped in a task's page tables, each mapping
399 * is counted separately
401 * Also, many kernel routines increase the page count before a critical
402 * routine so they can be sure the page doesn't go away from under them.
406 * Drop a ref, return true if the refcount fell to zero (the page has no users)
408 static inline int put_page_testzero(struct page *page)
410 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
411 return page_ref_dec_and_test(page);
415 * Try to grab a ref unless the page has a refcount of zero, return false if
417 * This can be called when MMU is off so it must not access
418 * any of the virtual mappings.
420 static inline int get_page_unless_zero(struct page *page)
422 return page_ref_add_unless(page, 1, 0);
425 extern int page_is_ram(unsigned long pfn);
433 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
436 /* Support for virtually mapped pages */
437 struct page *vmalloc_to_page(const void *addr);
438 unsigned long vmalloc_to_pfn(const void *addr);
441 * Determine if an address is within the vmalloc range
443 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
444 * is no special casing required.
446 static inline int is_vmalloc_addr(const void *x)
449 unsigned long addr = (unsigned long)x;
451 return addr >= VMALLOC_START && addr < VMALLOC_END;
457 extern int is_vmalloc_or_module_addr(const void *x);
459 static inline int is_vmalloc_or_module_addr(const void *x)
465 extern void kvfree(const void *addr);
467 static inline atomic_t *compound_mapcount_ptr(struct page *page)
469 return &page[1].compound_mapcount;
472 static inline int compound_mapcount(struct page *page)
474 if (!PageCompound(page))
476 page = compound_head(page);
477 return atomic_read(compound_mapcount_ptr(page)) + 1;
481 * The atomic page->_mapcount, starts from -1: so that transitions
482 * both from it and to it can be tracked, using atomic_inc_and_test
483 * and atomic_add_negative(-1).
485 static inline void page_mapcount_reset(struct page *page)
487 atomic_set(&(page)->_mapcount, -1);
490 int __page_mapcount(struct page *page);
492 static inline int page_mapcount(struct page *page)
494 VM_BUG_ON_PAGE(PageSlab(page), page);
496 if (unlikely(PageCompound(page)))
497 return __page_mapcount(page);
498 return atomic_read(&page->_mapcount) + 1;
501 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
502 int total_mapcount(struct page *page);
503 int page_trans_huge_mapcount(struct page *page, int *total_mapcount);
505 static inline int total_mapcount(struct page *page)
507 return page_mapcount(page);
509 static inline int page_trans_huge_mapcount(struct page *page,
512 int mapcount = page_mapcount(page);
514 *total_mapcount = mapcount;
519 static inline struct page *virt_to_head_page(const void *x)
521 struct page *page = virt_to_page(x);
523 return compound_head(page);
526 void __put_page(struct page *page);
528 void put_pages_list(struct list_head *pages);
530 void split_page(struct page *page, unsigned int order);
531 int split_free_page(struct page *page);
534 * Compound pages have a destructor function. Provide a
535 * prototype for that function and accessor functions.
536 * These are _only_ valid on the head of a compound page.
538 typedef void compound_page_dtor(struct page *);
540 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
541 enum compound_dtor_id {
544 #ifdef CONFIG_HUGETLB_PAGE
547 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
552 extern compound_page_dtor * const compound_page_dtors[];
554 static inline void set_compound_page_dtor(struct page *page,
555 enum compound_dtor_id compound_dtor)
557 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
558 page[1].compound_dtor = compound_dtor;
561 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
563 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
564 return compound_page_dtors[page[1].compound_dtor];
567 static inline unsigned int compound_order(struct page *page)
571 return page[1].compound_order;
574 static inline void set_compound_order(struct page *page, unsigned int order)
576 page[1].compound_order = order;
579 void free_compound_page(struct page *page);
583 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
584 * servicing faults for write access. In the normal case, do always want
585 * pte_mkwrite. But get_user_pages can cause write faults for mappings
586 * that do not have writing enabled, when used by access_process_vm.
588 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
590 if (likely(vma->vm_flags & VM_WRITE))
591 pte = pte_mkwrite(pte);
595 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
596 struct page *page, pte_t *pte, bool write, bool anon);
600 * Multiple processes may "see" the same page. E.g. for untouched
601 * mappings of /dev/null, all processes see the same page full of
602 * zeroes, and text pages of executables and shared libraries have
603 * only one copy in memory, at most, normally.
605 * For the non-reserved pages, page_count(page) denotes a reference count.
606 * page_count() == 0 means the page is free. page->lru is then used for
607 * freelist management in the buddy allocator.
608 * page_count() > 0 means the page has been allocated.
610 * Pages are allocated by the slab allocator in order to provide memory
611 * to kmalloc and kmem_cache_alloc. In this case, the management of the
612 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
613 * unless a particular usage is carefully commented. (the responsibility of
614 * freeing the kmalloc memory is the caller's, of course).
616 * A page may be used by anyone else who does a __get_free_page().
617 * In this case, page_count still tracks the references, and should only
618 * be used through the normal accessor functions. The top bits of page->flags
619 * and page->virtual store page management information, but all other fields
620 * are unused and could be used privately, carefully. The management of this
621 * page is the responsibility of the one who allocated it, and those who have
622 * subsequently been given references to it.
624 * The other pages (we may call them "pagecache pages") are completely
625 * managed by the Linux memory manager: I/O, buffers, swapping etc.
626 * The following discussion applies only to them.
628 * A pagecache page contains an opaque `private' member, which belongs to the
629 * page's address_space. Usually, this is the address of a circular list of
630 * the page's disk buffers. PG_private must be set to tell the VM to call
631 * into the filesystem to release these pages.
633 * A page may belong to an inode's memory mapping. In this case, page->mapping
634 * is the pointer to the inode, and page->index is the file offset of the page,
635 * in units of PAGE_SIZE.
637 * If pagecache pages are not associated with an inode, they are said to be
638 * anonymous pages. These may become associated with the swapcache, and in that
639 * case PG_swapcache is set, and page->private is an offset into the swapcache.
641 * In either case (swapcache or inode backed), the pagecache itself holds one
642 * reference to the page. Setting PG_private should also increment the
643 * refcount. The each user mapping also has a reference to the page.
645 * The pagecache pages are stored in a per-mapping radix tree, which is
646 * rooted at mapping->page_tree, and indexed by offset.
647 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
648 * lists, we instead now tag pages as dirty/writeback in the radix tree.
650 * All pagecache pages may be subject to I/O:
651 * - inode pages may need to be read from disk,
652 * - inode pages which have been modified and are MAP_SHARED may need
653 * to be written back to the inode on disk,
654 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
655 * modified may need to be swapped out to swap space and (later) to be read
660 * The zone field is never updated after free_area_init_core()
661 * sets it, so none of the operations on it need to be atomic.
664 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
665 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
666 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
667 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
668 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
671 * Define the bit shifts to access each section. For non-existent
672 * sections we define the shift as 0; that plus a 0 mask ensures
673 * the compiler will optimise away reference to them.
675 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
676 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
677 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
678 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
680 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
681 #ifdef NODE_NOT_IN_PAGE_FLAGS
682 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
683 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
684 SECTIONS_PGOFF : ZONES_PGOFF)
686 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
687 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
688 NODES_PGOFF : ZONES_PGOFF)
691 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
693 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
694 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
697 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
698 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
699 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
700 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
701 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
703 static inline enum zone_type page_zonenum(const struct page *page)
705 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
708 #ifdef CONFIG_ZONE_DEVICE
709 void get_zone_device_page(struct page *page);
710 void put_zone_device_page(struct page *page);
711 static inline bool is_zone_device_page(const struct page *page)
713 return page_zonenum(page) == ZONE_DEVICE;
716 static inline void get_zone_device_page(struct page *page)
719 static inline void put_zone_device_page(struct page *page)
722 static inline bool is_zone_device_page(const struct page *page)
728 static inline void get_page(struct page *page)
730 page = compound_head(page);
732 * Getting a normal page or the head of a compound page
733 * requires to already have an elevated page->_count.
735 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
738 if (unlikely(is_zone_device_page(page)))
739 get_zone_device_page(page);
742 static inline void put_page(struct page *page)
744 page = compound_head(page);
746 if (put_page_testzero(page))
749 if (unlikely(is_zone_device_page(page)))
750 put_zone_device_page(page);
753 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
754 #define SECTION_IN_PAGE_FLAGS
758 * The identification function is mainly used by the buddy allocator for
759 * determining if two pages could be buddies. We are not really identifying
760 * the zone since we could be using the section number id if we do not have
761 * node id available in page flags.
762 * We only guarantee that it will return the same value for two combinable
765 static inline int page_zone_id(struct page *page)
767 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
770 static inline int zone_to_nid(struct zone *zone)
779 #ifdef NODE_NOT_IN_PAGE_FLAGS
780 extern int page_to_nid(const struct page *page);
782 static inline int page_to_nid(const struct page *page)
784 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
788 #ifdef CONFIG_NUMA_BALANCING
789 static inline int cpu_pid_to_cpupid(int cpu, int pid)
791 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
794 static inline int cpupid_to_pid(int cpupid)
796 return cpupid & LAST__PID_MASK;
799 static inline int cpupid_to_cpu(int cpupid)
801 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
804 static inline int cpupid_to_nid(int cpupid)
806 return cpu_to_node(cpupid_to_cpu(cpupid));
809 static inline bool cpupid_pid_unset(int cpupid)
811 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
814 static inline bool cpupid_cpu_unset(int cpupid)
816 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
819 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
821 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
824 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
825 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
826 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
828 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
831 static inline int page_cpupid_last(struct page *page)
833 return page->_last_cpupid;
835 static inline void page_cpupid_reset_last(struct page *page)
837 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
840 static inline int page_cpupid_last(struct page *page)
842 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
845 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
847 static inline void page_cpupid_reset_last(struct page *page)
849 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
851 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
852 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
854 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
855 #else /* !CONFIG_NUMA_BALANCING */
856 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
858 return page_to_nid(page); /* XXX */
861 static inline int page_cpupid_last(struct page *page)
863 return page_to_nid(page); /* XXX */
866 static inline int cpupid_to_nid(int cpupid)
871 static inline int cpupid_to_pid(int cpupid)
876 static inline int cpupid_to_cpu(int cpupid)
881 static inline int cpu_pid_to_cpupid(int nid, int pid)
886 static inline bool cpupid_pid_unset(int cpupid)
891 static inline void page_cpupid_reset_last(struct page *page)
895 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
899 #endif /* CONFIG_NUMA_BALANCING */
901 static inline struct zone *page_zone(const struct page *page)
903 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
906 #ifdef SECTION_IN_PAGE_FLAGS
907 static inline void set_page_section(struct page *page, unsigned long section)
909 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
910 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
913 static inline unsigned long page_to_section(const struct page *page)
915 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
919 static inline void set_page_zone(struct page *page, enum zone_type zone)
921 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
922 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
925 static inline void set_page_node(struct page *page, unsigned long node)
927 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
928 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
931 static inline void set_page_links(struct page *page, enum zone_type zone,
932 unsigned long node, unsigned long pfn)
934 set_page_zone(page, zone);
935 set_page_node(page, node);
936 #ifdef SECTION_IN_PAGE_FLAGS
937 set_page_section(page, pfn_to_section_nr(pfn));
942 static inline struct mem_cgroup *page_memcg(struct page *page)
944 return page->mem_cgroup;
947 static inline struct mem_cgroup *page_memcg(struct page *page)
954 * Some inline functions in vmstat.h depend on page_zone()
956 #include <linux/vmstat.h>
958 static __always_inline void *lowmem_page_address(const struct page *page)
960 return __va(PFN_PHYS(page_to_pfn(page)));
963 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
964 #define HASHED_PAGE_VIRTUAL
967 #if defined(WANT_PAGE_VIRTUAL)
968 static inline void *page_address(const struct page *page)
970 return page->virtual;
972 static inline void set_page_address(struct page *page, void *address)
974 page->virtual = address;
976 #define page_address_init() do { } while(0)
979 #if defined(HASHED_PAGE_VIRTUAL)
980 void *page_address(const struct page *page);
981 void set_page_address(struct page *page, void *virtual);
982 void page_address_init(void);
985 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
986 #define page_address(page) lowmem_page_address(page)
987 #define set_page_address(page, address) do { } while(0)
988 #define page_address_init() do { } while(0)
991 extern void *page_rmapping(struct page *page);
992 extern struct anon_vma *page_anon_vma(struct page *page);
993 extern struct address_space *page_mapping(struct page *page);
995 extern struct address_space *__page_file_mapping(struct page *);
998 struct address_space *page_file_mapping(struct page *page)
1000 if (unlikely(PageSwapCache(page)))
1001 return __page_file_mapping(page);
1003 return page->mapping;
1007 * Return the pagecache index of the passed page. Regular pagecache pages
1008 * use ->index whereas swapcache pages use ->private
1010 static inline pgoff_t page_index(struct page *page)
1012 if (unlikely(PageSwapCache(page)))
1013 return page_private(page);
1017 extern pgoff_t __page_file_index(struct page *page);
1020 * Return the file index of the page. Regular pagecache pages use ->index
1021 * whereas swapcache pages use swp_offset(->private)
1023 static inline pgoff_t page_file_index(struct page *page)
1025 if (unlikely(PageSwapCache(page)))
1026 return __page_file_index(page);
1032 * Return true if this page is mapped into pagetables.
1033 * For compound page it returns true if any subpage of compound page is mapped.
1035 static inline bool page_mapped(struct page *page)
1038 if (likely(!PageCompound(page)))
1039 return atomic_read(&page->_mapcount) >= 0;
1040 page = compound_head(page);
1041 if (atomic_read(compound_mapcount_ptr(page)) >= 0)
1045 for (i = 0; i < hpage_nr_pages(page); i++) {
1046 if (atomic_read(&page[i]._mapcount) >= 0)
1053 * Return true only if the page has been allocated with
1054 * ALLOC_NO_WATERMARKS and the low watermark was not
1055 * met implying that the system is under some pressure.
1057 static inline bool page_is_pfmemalloc(struct page *page)
1060 * Page index cannot be this large so this must be
1061 * a pfmemalloc page.
1063 return page->index == -1UL;
1067 * Only to be called by the page allocator on a freshly allocated
1070 static inline void set_page_pfmemalloc(struct page *page)
1075 static inline void clear_page_pfmemalloc(struct page *page)
1081 * Different kinds of faults, as returned by handle_mm_fault().
1082 * Used to decide whether a process gets delivered SIGBUS or
1083 * just gets major/minor fault counters bumped up.
1086 #define VM_FAULT_OOM 0x0001
1087 #define VM_FAULT_SIGBUS 0x0002
1088 #define VM_FAULT_MAJOR 0x0004
1089 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1090 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1091 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1092 #define VM_FAULT_SIGSEGV 0x0040
1094 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1095 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1096 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1097 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1099 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1101 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1102 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1105 /* Encode hstate index for a hwpoisoned large page */
1106 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1107 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1110 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1112 extern void pagefault_out_of_memory(void);
1114 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1117 * Flags passed to show_mem() and show_free_areas() to suppress output in
1120 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1122 extern void show_free_areas(unsigned int flags);
1123 extern bool skip_free_areas_node(unsigned int flags, int nid);
1125 int shmem_zero_setup(struct vm_area_struct *);
1127 bool shmem_mapping(struct address_space *mapping);
1129 static inline bool shmem_mapping(struct address_space *mapping)
1135 extern bool can_do_mlock(void);
1136 extern int user_shm_lock(size_t, struct user_struct *);
1137 extern void user_shm_unlock(size_t, struct user_struct *);
1140 * Parameter block passed down to zap_pte_range in exceptional cases.
1142 struct zap_details {
1143 struct address_space *check_mapping; /* Check page->mapping if set */
1144 pgoff_t first_index; /* Lowest page->index to unmap */
1145 pgoff_t last_index; /* Highest page->index to unmap */
1146 bool ignore_dirty; /* Ignore dirty pages */
1147 bool check_swap_entries; /* Check also swap entries */
1150 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1152 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1155 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1156 unsigned long size);
1157 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1158 unsigned long size, struct zap_details *);
1159 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1160 unsigned long start, unsigned long end);
1163 * mm_walk - callbacks for walk_page_range
1164 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1165 * this handler is required to be able to handle
1166 * pmd_trans_huge() pmds. They may simply choose to
1167 * split_huge_page() instead of handling it explicitly.
1168 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1169 * @pte_hole: if set, called for each hole at all levels
1170 * @hugetlb_entry: if set, called for each hugetlb entry
1171 * @test_walk: caller specific callback function to determine whether
1172 * we walk over the current vma or not. A positive returned
1173 * value means "do page table walk over the current vma,"
1174 * and a negative one means "abort current page table walk
1175 * right now." 0 means "skip the current vma."
1176 * @mm: mm_struct representing the target process of page table walk
1177 * @vma: vma currently walked (NULL if walking outside vmas)
1178 * @private: private data for callbacks' usage
1180 * (see the comment on walk_page_range() for more details)
1183 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1184 unsigned long next, struct mm_walk *walk);
1185 int (*pte_entry)(pte_t *pte, unsigned long addr,
1186 unsigned long next, struct mm_walk *walk);
1187 int (*pte_hole)(unsigned long addr, unsigned long next,
1188 struct mm_walk *walk);
1189 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1190 unsigned long addr, unsigned long next,
1191 struct mm_walk *walk);
1192 int (*test_walk)(unsigned long addr, unsigned long next,
1193 struct mm_walk *walk);
1194 struct mm_struct *mm;
1195 struct vm_area_struct *vma;
1199 int walk_page_range(unsigned long addr, unsigned long end,
1200 struct mm_walk *walk);
1201 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1202 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1203 unsigned long end, unsigned long floor, unsigned long ceiling);
1204 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1205 struct vm_area_struct *vma);
1206 void unmap_mapping_range(struct address_space *mapping,
1207 loff_t const holebegin, loff_t const holelen, int even_cows);
1208 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1209 unsigned long *pfn);
1210 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1211 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1212 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1213 void *buf, int len, int write);
1215 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1216 loff_t const holebegin, loff_t const holelen)
1218 unmap_mapping_range(mapping, holebegin, holelen, 0);
1221 extern void truncate_pagecache(struct inode *inode, loff_t new);
1222 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1223 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1224 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1225 int truncate_inode_page(struct address_space *mapping, struct page *page);
1226 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1227 int invalidate_inode_page(struct page *page);
1230 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1231 unsigned long address, unsigned int flags);
1232 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1233 unsigned long address, unsigned int fault_flags,
1236 static inline int handle_mm_fault(struct mm_struct *mm,
1237 struct vm_area_struct *vma, unsigned long address,
1240 /* should never happen if there's no MMU */
1242 return VM_FAULT_SIGBUS;
1244 static inline int fixup_user_fault(struct task_struct *tsk,
1245 struct mm_struct *mm, unsigned long address,
1246 unsigned int fault_flags, bool *unlocked)
1248 /* should never happen if there's no MMU */
1254 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1255 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1256 void *buf, int len, int write);
1258 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1259 unsigned long start, unsigned long nr_pages,
1260 unsigned int foll_flags, struct page **pages,
1261 struct vm_area_struct **vmas, int *nonblocking);
1262 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1263 unsigned long start, unsigned long nr_pages,
1264 int write, int force, struct page **pages,
1265 struct vm_area_struct **vmas);
1266 long get_user_pages(unsigned long start, unsigned long nr_pages,
1267 int write, int force, struct page **pages,
1268 struct vm_area_struct **vmas);
1269 long get_user_pages_locked(unsigned long start, unsigned long nr_pages,
1270 int write, int force, struct page **pages, int *locked);
1271 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1272 unsigned long start, unsigned long nr_pages,
1273 int write, int force, struct page **pages,
1274 unsigned int gup_flags);
1275 long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages,
1276 int write, int force, struct page **pages);
1277 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1278 struct page **pages);
1280 /* Container for pinned pfns / pages */
1281 struct frame_vector {
1282 unsigned int nr_allocated; /* Number of frames we have space for */
1283 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1284 bool got_ref; /* Did we pin pages by getting page ref? */
1285 bool is_pfns; /* Does array contain pages or pfns? */
1286 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1287 * pfns_vector_pages() or pfns_vector_pfns()
1291 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1292 void frame_vector_destroy(struct frame_vector *vec);
1293 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1294 bool write, bool force, struct frame_vector *vec);
1295 void put_vaddr_frames(struct frame_vector *vec);
1296 int frame_vector_to_pages(struct frame_vector *vec);
1297 void frame_vector_to_pfns(struct frame_vector *vec);
1299 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1301 return vec->nr_frames;
1304 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1307 int err = frame_vector_to_pages(vec);
1310 return ERR_PTR(err);
1312 return (struct page **)(vec->ptrs);
1315 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1318 frame_vector_to_pfns(vec);
1319 return (unsigned long *)(vec->ptrs);
1323 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1324 struct page **pages);
1325 int get_kernel_page(unsigned long start, int write, struct page **pages);
1326 struct page *get_dump_page(unsigned long addr);
1328 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1329 extern void do_invalidatepage(struct page *page, unsigned int offset,
1330 unsigned int length);
1332 int __set_page_dirty_nobuffers(struct page *page);
1333 int __set_page_dirty_no_writeback(struct page *page);
1334 int redirty_page_for_writepage(struct writeback_control *wbc,
1336 void account_page_dirtied(struct page *page, struct address_space *mapping);
1337 void account_page_cleaned(struct page *page, struct address_space *mapping,
1338 struct bdi_writeback *wb);
1339 int set_page_dirty(struct page *page);
1340 int set_page_dirty_lock(struct page *page);
1341 void cancel_dirty_page(struct page *page);
1342 int clear_page_dirty_for_io(struct page *page);
1344 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1346 /* Is the vma a continuation of the stack vma above it? */
1347 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1349 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1352 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1354 return !vma->vm_ops;
1357 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1360 return (vma->vm_flags & VM_GROWSDOWN) &&
1361 (vma->vm_start == addr) &&
1362 !vma_growsdown(vma->vm_prev, addr);
1365 /* Is the vma a continuation of the stack vma below it? */
1366 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1368 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1371 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1374 return (vma->vm_flags & VM_GROWSUP) &&
1375 (vma->vm_end == addr) &&
1376 !vma_growsup(vma->vm_next, addr);
1379 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1381 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1382 unsigned long old_addr, struct vm_area_struct *new_vma,
1383 unsigned long new_addr, unsigned long len,
1384 bool need_rmap_locks);
1385 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1386 unsigned long end, pgprot_t newprot,
1387 int dirty_accountable, int prot_numa);
1388 extern int mprotect_fixup(struct vm_area_struct *vma,
1389 struct vm_area_struct **pprev, unsigned long start,
1390 unsigned long end, unsigned long newflags);
1393 * doesn't attempt to fault and will return short.
1395 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1396 struct page **pages);
1398 * per-process(per-mm_struct) statistics.
1400 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1402 long val = atomic_long_read(&mm->rss_stat.count[member]);
1404 #ifdef SPLIT_RSS_COUNTING
1406 * counter is updated in asynchronous manner and may go to minus.
1407 * But it's never be expected number for users.
1412 return (unsigned long)val;
1415 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1417 atomic_long_add(value, &mm->rss_stat.count[member]);
1420 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1422 atomic_long_inc(&mm->rss_stat.count[member]);
1425 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1427 atomic_long_dec(&mm->rss_stat.count[member]);
1430 /* Optimized variant when page is already known not to be PageAnon */
1431 static inline int mm_counter_file(struct page *page)
1433 if (PageSwapBacked(page))
1434 return MM_SHMEMPAGES;
1435 return MM_FILEPAGES;
1438 static inline int mm_counter(struct page *page)
1441 return MM_ANONPAGES;
1442 return mm_counter_file(page);
1445 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1447 return get_mm_counter(mm, MM_FILEPAGES) +
1448 get_mm_counter(mm, MM_ANONPAGES) +
1449 get_mm_counter(mm, MM_SHMEMPAGES);
1452 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1454 return max(mm->hiwater_rss, get_mm_rss(mm));
1457 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1459 return max(mm->hiwater_vm, mm->total_vm);
1462 static inline void update_hiwater_rss(struct mm_struct *mm)
1464 unsigned long _rss = get_mm_rss(mm);
1466 if ((mm)->hiwater_rss < _rss)
1467 (mm)->hiwater_rss = _rss;
1470 static inline void update_hiwater_vm(struct mm_struct *mm)
1472 if (mm->hiwater_vm < mm->total_vm)
1473 mm->hiwater_vm = mm->total_vm;
1476 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1478 mm->hiwater_rss = get_mm_rss(mm);
1481 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1482 struct mm_struct *mm)
1484 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1486 if (*maxrss < hiwater_rss)
1487 *maxrss = hiwater_rss;
1490 #if defined(SPLIT_RSS_COUNTING)
1491 void sync_mm_rss(struct mm_struct *mm);
1493 static inline void sync_mm_rss(struct mm_struct *mm)
1498 #ifndef __HAVE_ARCH_PTE_DEVMAP
1499 static inline int pte_devmap(pte_t pte)
1505 int vma_wants_writenotify(struct vm_area_struct *vma);
1507 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1509 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1513 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1517 #ifdef __PAGETABLE_PUD_FOLDED
1518 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1519 unsigned long address)
1524 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1527 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1528 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1529 unsigned long address)
1534 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1536 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1541 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1542 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1545 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1547 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1549 atomic_long_set(&mm->nr_pmds, 0);
1552 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1554 return atomic_long_read(&mm->nr_pmds);
1557 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1559 atomic_long_inc(&mm->nr_pmds);
1562 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1564 atomic_long_dec(&mm->nr_pmds);
1568 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1569 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1572 * The following ifdef needed to get the 4level-fixup.h header to work.
1573 * Remove it when 4level-fixup.h has been removed.
1575 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1576 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1578 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1579 NULL: pud_offset(pgd, address);
1582 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1584 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1585 NULL: pmd_offset(pud, address);
1587 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1589 #if USE_SPLIT_PTE_PTLOCKS
1590 #if ALLOC_SPLIT_PTLOCKS
1591 void __init ptlock_cache_init(void);
1592 extern bool ptlock_alloc(struct page *page);
1593 extern void ptlock_free(struct page *page);
1595 static inline spinlock_t *ptlock_ptr(struct page *page)
1599 #else /* ALLOC_SPLIT_PTLOCKS */
1600 static inline void ptlock_cache_init(void)
1604 static inline bool ptlock_alloc(struct page *page)
1609 static inline void ptlock_free(struct page *page)
1613 static inline spinlock_t *ptlock_ptr(struct page *page)
1617 #endif /* ALLOC_SPLIT_PTLOCKS */
1619 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1621 return ptlock_ptr(pmd_page(*pmd));
1624 static inline bool ptlock_init(struct page *page)
1627 * prep_new_page() initialize page->private (and therefore page->ptl)
1628 * with 0. Make sure nobody took it in use in between.
1630 * It can happen if arch try to use slab for page table allocation:
1631 * slab code uses page->slab_cache, which share storage with page->ptl.
1633 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1634 if (!ptlock_alloc(page))
1636 spin_lock_init(ptlock_ptr(page));
1640 /* Reset page->mapping so free_pages_check won't complain. */
1641 static inline void pte_lock_deinit(struct page *page)
1643 page->mapping = NULL;
1647 #else /* !USE_SPLIT_PTE_PTLOCKS */
1649 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1651 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1653 return &mm->page_table_lock;
1655 static inline void ptlock_cache_init(void) {}
1656 static inline bool ptlock_init(struct page *page) { return true; }
1657 static inline void pte_lock_deinit(struct page *page) {}
1658 #endif /* USE_SPLIT_PTE_PTLOCKS */
1660 static inline void pgtable_init(void)
1662 ptlock_cache_init();
1663 pgtable_cache_init();
1666 static inline bool pgtable_page_ctor(struct page *page)
1668 if (!ptlock_init(page))
1670 inc_zone_page_state(page, NR_PAGETABLE);
1674 static inline void pgtable_page_dtor(struct page *page)
1676 pte_lock_deinit(page);
1677 dec_zone_page_state(page, NR_PAGETABLE);
1680 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1682 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1683 pte_t *__pte = pte_offset_map(pmd, address); \
1689 #define pte_unmap_unlock(pte, ptl) do { \
1694 #define pte_alloc(mm, pmd, address) \
1695 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1697 #define pte_alloc_map(mm, pmd, address) \
1698 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1700 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1701 (pte_alloc(mm, pmd, address) ? \
1702 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1704 #define pte_alloc_kernel(pmd, address) \
1705 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1706 NULL: pte_offset_kernel(pmd, address))
1708 #if USE_SPLIT_PMD_PTLOCKS
1710 static struct page *pmd_to_page(pmd_t *pmd)
1712 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1713 return virt_to_page((void *)((unsigned long) pmd & mask));
1716 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1718 return ptlock_ptr(pmd_to_page(pmd));
1721 static inline bool pgtable_pmd_page_ctor(struct page *page)
1723 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1724 page->pmd_huge_pte = NULL;
1726 return ptlock_init(page);
1729 static inline void pgtable_pmd_page_dtor(struct page *page)
1731 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1732 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1737 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1741 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1743 return &mm->page_table_lock;
1746 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1747 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1749 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1753 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1755 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1760 extern void free_area_init(unsigned long * zones_size);
1761 extern void free_area_init_node(int nid, unsigned long * zones_size,
1762 unsigned long zone_start_pfn, unsigned long *zholes_size);
1763 extern void free_initmem(void);
1766 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1767 * into the buddy system. The freed pages will be poisoned with pattern
1768 * "poison" if it's within range [0, UCHAR_MAX].
1769 * Return pages freed into the buddy system.
1771 extern unsigned long free_reserved_area(void *start, void *end,
1772 int poison, char *s);
1774 #ifdef CONFIG_HIGHMEM
1776 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1777 * and totalram_pages.
1779 extern void free_highmem_page(struct page *page);
1782 extern void adjust_managed_page_count(struct page *page, long count);
1783 extern void mem_init_print_info(const char *str);
1785 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1787 /* Free the reserved page into the buddy system, so it gets managed. */
1788 static inline void __free_reserved_page(struct page *page)
1790 ClearPageReserved(page);
1791 init_page_count(page);
1795 static inline void free_reserved_page(struct page *page)
1797 __free_reserved_page(page);
1798 adjust_managed_page_count(page, 1);
1801 static inline void mark_page_reserved(struct page *page)
1803 SetPageReserved(page);
1804 adjust_managed_page_count(page, -1);
1808 * Default method to free all the __init memory into the buddy system.
1809 * The freed pages will be poisoned with pattern "poison" if it's within
1810 * range [0, UCHAR_MAX].
1811 * Return pages freed into the buddy system.
1813 static inline unsigned long free_initmem_default(int poison)
1815 extern char __init_begin[], __init_end[];
1817 return free_reserved_area(&__init_begin, &__init_end,
1818 poison, "unused kernel");
1821 static inline unsigned long get_num_physpages(void)
1824 unsigned long phys_pages = 0;
1826 for_each_online_node(nid)
1827 phys_pages += node_present_pages(nid);
1832 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1834 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1835 * zones, allocate the backing mem_map and account for memory holes in a more
1836 * architecture independent manner. This is a substitute for creating the
1837 * zone_sizes[] and zholes_size[] arrays and passing them to
1838 * free_area_init_node()
1840 * An architecture is expected to register range of page frames backed by
1841 * physical memory with memblock_add[_node]() before calling
1842 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1843 * usage, an architecture is expected to do something like
1845 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1847 * for_each_valid_physical_page_range()
1848 * memblock_add_node(base, size, nid)
1849 * free_area_init_nodes(max_zone_pfns);
1851 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1852 * registered physical page range. Similarly
1853 * sparse_memory_present_with_active_regions() calls memory_present() for
1854 * each range when SPARSEMEM is enabled.
1856 * See mm/page_alloc.c for more information on each function exposed by
1857 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1859 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1860 unsigned long node_map_pfn_alignment(void);
1861 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1862 unsigned long end_pfn);
1863 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1864 unsigned long end_pfn);
1865 extern void get_pfn_range_for_nid(unsigned int nid,
1866 unsigned long *start_pfn, unsigned long *end_pfn);
1867 extern unsigned long find_min_pfn_with_active_regions(void);
1868 extern void free_bootmem_with_active_regions(int nid,
1869 unsigned long max_low_pfn);
1870 extern void sparse_memory_present_with_active_regions(int nid);
1872 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1874 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1875 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1876 static inline int __early_pfn_to_nid(unsigned long pfn,
1877 struct mminit_pfnnid_cache *state)
1882 /* please see mm/page_alloc.c */
1883 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1884 /* there is a per-arch backend function. */
1885 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1886 struct mminit_pfnnid_cache *state);
1889 extern void set_dma_reserve(unsigned long new_dma_reserve);
1890 extern void memmap_init_zone(unsigned long, int, unsigned long,
1891 unsigned long, enum memmap_context);
1892 extern void setup_per_zone_wmarks(void);
1893 extern int __meminit init_per_zone_wmark_min(void);
1894 extern void mem_init(void);
1895 extern void __init mmap_init(void);
1896 extern void show_mem(unsigned int flags);
1897 extern long si_mem_available(void);
1898 extern void si_meminfo(struct sysinfo * val);
1899 extern void si_meminfo_node(struct sysinfo *val, int nid);
1901 extern __printf(3, 4)
1902 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1903 const char *fmt, ...);
1905 extern void setup_per_cpu_pageset(void);
1907 extern void zone_pcp_update(struct zone *zone);
1908 extern void zone_pcp_reset(struct zone *zone);
1911 extern int min_free_kbytes;
1912 extern int watermark_scale_factor;
1915 extern atomic_long_t mmap_pages_allocated;
1916 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1918 /* interval_tree.c */
1919 void vma_interval_tree_insert(struct vm_area_struct *node,
1920 struct rb_root *root);
1921 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1922 struct vm_area_struct *prev,
1923 struct rb_root *root);
1924 void vma_interval_tree_remove(struct vm_area_struct *node,
1925 struct rb_root *root);
1926 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1927 unsigned long start, unsigned long last);
1928 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1929 unsigned long start, unsigned long last);
1931 #define vma_interval_tree_foreach(vma, root, start, last) \
1932 for (vma = vma_interval_tree_iter_first(root, start, last); \
1933 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1935 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1936 struct rb_root *root);
1937 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1938 struct rb_root *root);
1939 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1940 struct rb_root *root, unsigned long start, unsigned long last);
1941 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1942 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1943 #ifdef CONFIG_DEBUG_VM_RB
1944 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1947 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1948 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1949 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1952 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1953 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1954 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1955 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1956 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1957 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1958 struct mempolicy *, struct vm_userfaultfd_ctx);
1959 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1960 extern int split_vma(struct mm_struct *,
1961 struct vm_area_struct *, unsigned long addr, int new_below);
1962 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1963 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1964 struct rb_node **, struct rb_node *);
1965 extern void unlink_file_vma(struct vm_area_struct *);
1966 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1967 unsigned long addr, unsigned long len, pgoff_t pgoff,
1968 bool *need_rmap_locks);
1969 extern void exit_mmap(struct mm_struct *);
1971 static inline int check_data_rlimit(unsigned long rlim,
1973 unsigned long start,
1974 unsigned long end_data,
1975 unsigned long start_data)
1977 if (rlim < RLIM_INFINITY) {
1978 if (((new - start) + (end_data - start_data)) > rlim)
1985 extern int mm_take_all_locks(struct mm_struct *mm);
1986 extern void mm_drop_all_locks(struct mm_struct *mm);
1988 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1989 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1991 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
1992 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
1994 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1995 unsigned long addr, unsigned long len,
1996 unsigned long flags,
1997 const struct vm_special_mapping *spec);
1998 /* This is an obsolete alternative to _install_special_mapping. */
1999 extern int install_special_mapping(struct mm_struct *mm,
2000 unsigned long addr, unsigned long len,
2001 unsigned long flags, struct page **pages);
2003 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2005 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2006 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2007 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2008 unsigned long len, unsigned long prot, unsigned long flags,
2009 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2010 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2012 static inline unsigned long
2013 do_mmap_pgoff(struct file *file, unsigned long addr,
2014 unsigned long len, unsigned long prot, unsigned long flags,
2015 unsigned long pgoff, unsigned long *populate)
2017 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2021 extern int __mm_populate(unsigned long addr, unsigned long len,
2023 static inline void mm_populate(unsigned long addr, unsigned long len)
2026 (void) __mm_populate(addr, len, 1);
2029 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2032 /* These take the mm semaphore themselves */
2033 extern unsigned long vm_brk(unsigned long, unsigned long);
2034 extern int vm_munmap(unsigned long, size_t);
2035 extern unsigned long vm_mmap(struct file *, unsigned long,
2036 unsigned long, unsigned long,
2037 unsigned long, unsigned long);
2039 struct vm_unmapped_area_info {
2040 #define VM_UNMAPPED_AREA_TOPDOWN 1
2041 unsigned long flags;
2042 unsigned long length;
2043 unsigned long low_limit;
2044 unsigned long high_limit;
2045 unsigned long align_mask;
2046 unsigned long align_offset;
2049 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2050 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2053 * Search for an unmapped address range.
2055 * We are looking for a range that:
2056 * - does not intersect with any VMA;
2057 * - is contained within the [low_limit, high_limit) interval;
2058 * - is at least the desired size.
2059 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2061 static inline unsigned long
2062 vm_unmapped_area(struct vm_unmapped_area_info *info)
2064 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2065 return unmapped_area_topdown(info);
2067 return unmapped_area(info);
2071 extern void truncate_inode_pages(struct address_space *, loff_t);
2072 extern void truncate_inode_pages_range(struct address_space *,
2073 loff_t lstart, loff_t lend);
2074 extern void truncate_inode_pages_final(struct address_space *);
2076 /* generic vm_area_ops exported for stackable file systems */
2077 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2078 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2079 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2081 /* mm/page-writeback.c */
2082 int write_one_page(struct page *page, int wait);
2083 void task_dirty_inc(struct task_struct *tsk);
2086 #define VM_MAX_READAHEAD 128 /* kbytes */
2087 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2089 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2090 pgoff_t offset, unsigned long nr_to_read);
2092 void page_cache_sync_readahead(struct address_space *mapping,
2093 struct file_ra_state *ra,
2096 unsigned long size);
2098 void page_cache_async_readahead(struct address_space *mapping,
2099 struct file_ra_state *ra,
2103 unsigned long size);
2105 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2106 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2108 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2109 extern int expand_downwards(struct vm_area_struct *vma,
2110 unsigned long address);
2112 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2114 #define expand_upwards(vma, address) (0)
2117 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2118 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2119 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2120 struct vm_area_struct **pprev);
2122 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2123 NULL if none. Assume start_addr < end_addr. */
2124 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2126 struct vm_area_struct * vma = find_vma(mm,start_addr);
2128 if (vma && end_addr <= vma->vm_start)
2133 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2135 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2138 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2139 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2140 unsigned long vm_start, unsigned long vm_end)
2142 struct vm_area_struct *vma = find_vma(mm, vm_start);
2144 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2151 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2152 void vma_set_page_prot(struct vm_area_struct *vma);
2154 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2158 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2160 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2164 #ifdef CONFIG_NUMA_BALANCING
2165 unsigned long change_prot_numa(struct vm_area_struct *vma,
2166 unsigned long start, unsigned long end);
2169 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2170 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2171 unsigned long pfn, unsigned long size, pgprot_t);
2172 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2173 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2175 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2176 unsigned long pfn, pgprot_t pgprot);
2177 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2179 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2182 struct page *follow_page_mask(struct vm_area_struct *vma,
2183 unsigned long address, unsigned int foll_flags,
2184 unsigned int *page_mask);
2186 static inline struct page *follow_page(struct vm_area_struct *vma,
2187 unsigned long address, unsigned int foll_flags)
2189 unsigned int unused_page_mask;
2190 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2193 #define FOLL_WRITE 0x01 /* check pte is writable */
2194 #define FOLL_TOUCH 0x02 /* mark page accessed */
2195 #define FOLL_GET 0x04 /* do get_page on page */
2196 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2197 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2198 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2199 * and return without waiting upon it */
2200 #define FOLL_POPULATE 0x40 /* fault in page */
2201 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2202 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2203 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2204 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2205 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2206 #define FOLL_MLOCK 0x1000 /* lock present pages */
2207 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2209 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2211 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2212 unsigned long size, pte_fn_t fn, void *data);
2215 #ifdef CONFIG_PAGE_POISONING
2216 extern bool page_poisoning_enabled(void);
2217 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2218 extern bool page_is_poisoned(struct page *page);
2220 static inline bool page_poisoning_enabled(void) { return false; }
2221 static inline void kernel_poison_pages(struct page *page, int numpages,
2223 static inline bool page_is_poisoned(struct page *page) { return false; }
2226 #ifdef CONFIG_DEBUG_PAGEALLOC
2227 extern bool _debug_pagealloc_enabled;
2228 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2230 static inline bool debug_pagealloc_enabled(void)
2232 return _debug_pagealloc_enabled;
2236 kernel_map_pages(struct page *page, int numpages, int enable)
2238 if (!debug_pagealloc_enabled())
2241 __kernel_map_pages(page, numpages, enable);
2243 #ifdef CONFIG_HIBERNATION
2244 extern bool kernel_page_present(struct page *page);
2245 #endif /* CONFIG_HIBERNATION */
2246 #else /* CONFIG_DEBUG_PAGEALLOC */
2248 kernel_map_pages(struct page *page, int numpages, int enable) {}
2249 #ifdef CONFIG_HIBERNATION
2250 static inline bool kernel_page_present(struct page *page) { return true; }
2251 #endif /* CONFIG_HIBERNATION */
2252 static inline bool debug_pagealloc_enabled(void)
2256 #endif /* CONFIG_DEBUG_PAGEALLOC */
2258 #ifdef __HAVE_ARCH_GATE_AREA
2259 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2260 extern int in_gate_area_no_mm(unsigned long addr);
2261 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2263 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2267 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2268 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2272 #endif /* __HAVE_ARCH_GATE_AREA */
2274 #ifdef CONFIG_SYSCTL
2275 extern int sysctl_drop_caches;
2276 int drop_caches_sysctl_handler(struct ctl_table *, int,
2277 void __user *, size_t *, loff_t *);
2280 void drop_slab(void);
2281 void drop_slab_node(int nid);
2284 #define randomize_va_space 0
2286 extern int randomize_va_space;
2289 const char * arch_vma_name(struct vm_area_struct *vma);
2290 void print_vma_addr(char *prefix, unsigned long rip);
2292 void sparse_mem_maps_populate_node(struct page **map_map,
2293 unsigned long pnum_begin,
2294 unsigned long pnum_end,
2295 unsigned long map_count,
2298 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2299 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2300 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2301 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2302 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2303 void *vmemmap_alloc_block(unsigned long size, int node);
2305 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2306 struct vmem_altmap *altmap);
2307 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2309 return __vmemmap_alloc_block_buf(size, node, NULL);
2312 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2313 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2315 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2316 void vmemmap_populate_print_last(void);
2317 #ifdef CONFIG_MEMORY_HOTPLUG
2318 void vmemmap_free(unsigned long start, unsigned long end);
2320 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2321 unsigned long size);
2324 MF_COUNT_INCREASED = 1 << 0,
2325 MF_ACTION_REQUIRED = 1 << 1,
2326 MF_MUST_KILL = 1 << 2,
2327 MF_SOFT_OFFLINE = 1 << 3,
2329 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2330 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2331 extern int unpoison_memory(unsigned long pfn);
2332 extern int get_hwpoison_page(struct page *page);
2333 #define put_hwpoison_page(page) put_page(page)
2334 extern int sysctl_memory_failure_early_kill;
2335 extern int sysctl_memory_failure_recovery;
2336 extern void shake_page(struct page *p, int access);
2337 extern atomic_long_t num_poisoned_pages;
2338 extern int soft_offline_page(struct page *page, int flags);
2342 * Error handlers for various types of pages.
2345 MF_IGNORED, /* Error: cannot be handled */
2346 MF_FAILED, /* Error: handling failed */
2347 MF_DELAYED, /* Will be handled later */
2348 MF_RECOVERED, /* Successfully recovered */
2351 enum mf_action_page_type {
2353 MF_MSG_KERNEL_HIGH_ORDER,
2355 MF_MSG_DIFFERENT_COMPOUND,
2356 MF_MSG_POISONED_HUGE,
2359 MF_MSG_UNMAP_FAILED,
2360 MF_MSG_DIRTY_SWAPCACHE,
2361 MF_MSG_CLEAN_SWAPCACHE,
2362 MF_MSG_DIRTY_MLOCKED_LRU,
2363 MF_MSG_CLEAN_MLOCKED_LRU,
2364 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2365 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2368 MF_MSG_TRUNCATED_LRU,
2374 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2375 extern void clear_huge_page(struct page *page,
2377 unsigned int pages_per_huge_page);
2378 extern void copy_user_huge_page(struct page *dst, struct page *src,
2379 unsigned long addr, struct vm_area_struct *vma,
2380 unsigned int pages_per_huge_page);
2381 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2383 extern struct page_ext_operations debug_guardpage_ops;
2384 extern struct page_ext_operations page_poisoning_ops;
2386 #ifdef CONFIG_DEBUG_PAGEALLOC
2387 extern unsigned int _debug_guardpage_minorder;
2388 extern bool _debug_guardpage_enabled;
2390 static inline unsigned int debug_guardpage_minorder(void)
2392 return _debug_guardpage_minorder;
2395 static inline bool debug_guardpage_enabled(void)
2397 return _debug_guardpage_enabled;
2400 static inline bool page_is_guard(struct page *page)
2402 struct page_ext *page_ext;
2404 if (!debug_guardpage_enabled())
2407 page_ext = lookup_page_ext(page);
2408 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2411 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2412 static inline bool debug_guardpage_enabled(void) { return false; }
2413 static inline bool page_is_guard(struct page *page) { return false; }
2414 #endif /* CONFIG_DEBUG_PAGEALLOC */
2416 #if MAX_NUMNODES > 1
2417 void __init setup_nr_node_ids(void);
2419 static inline void setup_nr_node_ids(void) {}
2422 #endif /* __KERNEL__ */
2423 #endif /* _LINUX_MM_H */