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>
28 struct anon_vma_chain;
31 struct writeback_control;
34 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
35 extern unsigned long max_mapnr;
37 static inline void set_max_mapnr(unsigned long limit)
42 static inline void set_max_mapnr(unsigned long limit) { }
45 extern unsigned long totalram_pages;
46 extern void * high_memory;
47 extern int page_cluster;
50 extern int sysctl_legacy_va_layout;
52 #define sysctl_legacy_va_layout 0
55 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
56 extern const int mmap_rnd_bits_min;
57 extern const int mmap_rnd_bits_max;
58 extern int mmap_rnd_bits __read_mostly;
60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
61 extern const int mmap_rnd_compat_bits_min;
62 extern const int mmap_rnd_compat_bits_max;
63 extern int mmap_rnd_compat_bits __read_mostly;
67 #include <asm/pgtable.h>
68 #include <asm/processor.h>
71 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
75 * To prevent common memory management code establishing
76 * a zero page mapping on a read fault.
77 * This macro should be defined within <asm/pgtable.h>.
78 * s390 does this to prevent multiplexing of hardware bits
79 * related to the physical page in case of virtualization.
81 #ifndef mm_forbids_zeropage
82 #define mm_forbids_zeropage(X) (0)
85 extern unsigned long sysctl_user_reserve_kbytes;
86 extern unsigned long sysctl_admin_reserve_kbytes;
88 extern int sysctl_overcommit_memory;
89 extern int sysctl_overcommit_ratio;
90 extern unsigned long sysctl_overcommit_kbytes;
92 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
94 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
97 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
99 /* to align the pointer to the (next) page boundary */
100 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
102 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
103 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
106 * Linux kernel virtual memory manager primitives.
107 * The idea being to have a "virtual" mm in the same way
108 * we have a virtual fs - giving a cleaner interface to the
109 * mm details, and allowing different kinds of memory mappings
110 * (from shared memory to executable loading to arbitrary
114 extern struct kmem_cache *vm_area_cachep;
117 extern struct rb_root nommu_region_tree;
118 extern struct rw_semaphore nommu_region_sem;
120 extern unsigned int kobjsize(const void *objp);
124 * vm_flags in vm_area_struct, see mm_types.h.
126 #define VM_NONE 0x00000000
128 #define VM_READ 0x00000001 /* currently active flags */
129 #define VM_WRITE 0x00000002
130 #define VM_EXEC 0x00000004
131 #define VM_SHARED 0x00000008
133 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
134 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
135 #define VM_MAYWRITE 0x00000020
136 #define VM_MAYEXEC 0x00000040
137 #define VM_MAYSHARE 0x00000080
139 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
140 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
141 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
142 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
143 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
145 #define VM_LOCKED 0x00002000
146 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
148 /* Used by sys_madvise() */
149 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
150 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
152 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
153 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
154 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
155 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
156 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
157 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
158 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
159 #define VM_ARCH_2 0x02000000
160 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
162 #ifdef CONFIG_MEM_SOFT_DIRTY
163 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
165 # define VM_SOFTDIRTY 0
168 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
169 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
170 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
171 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
173 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
174 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
175 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
176 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
177 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
178 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
179 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
180 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
181 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
182 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
184 #if defined(CONFIG_X86)
185 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
186 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
187 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
188 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
189 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
190 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
191 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
193 #elif defined(CONFIG_PPC)
194 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
195 #elif defined(CONFIG_PARISC)
196 # define VM_GROWSUP VM_ARCH_1
197 #elif defined(CONFIG_METAG)
198 # define VM_GROWSUP VM_ARCH_1
199 #elif defined(CONFIG_IA64)
200 # define VM_GROWSUP VM_ARCH_1
201 #elif !defined(CONFIG_MMU)
202 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
205 #if defined(CONFIG_X86)
206 /* MPX specific bounds table or bounds directory */
207 # define VM_MPX VM_ARCH_2
211 # define VM_GROWSUP VM_NONE
214 /* Bits set in the VMA until the stack is in its final location */
215 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
217 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
218 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
221 #ifdef CONFIG_STACK_GROWSUP
222 #define VM_STACK VM_GROWSUP
224 #define VM_STACK VM_GROWSDOWN
227 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
230 * Special vmas that are non-mergable, non-mlock()able.
231 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
233 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
235 /* This mask defines which mm->def_flags a process can inherit its parent */
236 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
238 /* This mask is used to clear all the VMA flags used by mlock */
239 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
242 * mapping from the currently active vm_flags protection bits (the
243 * low four bits) to a page protection mask..
245 extern pgprot_t protection_map[16];
247 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
248 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
249 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
250 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
251 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
252 #define FAULT_FLAG_TRIED 0x20 /* Second try */
253 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
254 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
257 * vm_fault is filled by the the pagefault handler and passed to the vma's
258 * ->fault function. The vma's ->fault is responsible for returning a bitmask
259 * of VM_FAULT_xxx flags that give details about how the fault was handled.
261 * MM layer fills up gfp_mask for page allocations but fault handler might
262 * alter it if its implementation requires a different allocation context.
264 * pgoff should be used in favour of virtual_address, if possible.
267 unsigned int flags; /* FAULT_FLAG_xxx flags */
268 gfp_t gfp_mask; /* gfp mask to be used for allocations */
269 pgoff_t pgoff; /* Logical page offset based on vma */
270 void __user *virtual_address; /* Faulting virtual address */
272 struct page *cow_page; /* Handler may choose to COW */
273 struct page *page; /* ->fault handlers should return a
274 * page here, unless VM_FAULT_NOPAGE
275 * is set (which is also implied by
278 /* for ->map_pages() only */
279 pgoff_t max_pgoff; /* map pages for offset from pgoff till
280 * max_pgoff inclusive */
281 pte_t *pte; /* pte entry associated with ->pgoff */
285 * These are the virtual MM functions - opening of an area, closing and
286 * unmapping it (needed to keep files on disk up-to-date etc), pointer
287 * to the functions called when a no-page or a wp-page exception occurs.
289 struct vm_operations_struct {
290 void (*open)(struct vm_area_struct * area);
291 void (*close)(struct vm_area_struct * area);
292 int (*mremap)(struct vm_area_struct * area);
293 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
294 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
295 pmd_t *, unsigned int flags);
296 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
298 /* notification that a previously read-only page is about to become
299 * writable, if an error is returned it will cause a SIGBUS */
300 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
302 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
303 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
305 /* called by access_process_vm when get_user_pages() fails, typically
306 * for use by special VMAs that can switch between memory and hardware
308 int (*access)(struct vm_area_struct *vma, unsigned long addr,
309 void *buf, int len, int write);
311 /* Called by the /proc/PID/maps code to ask the vma whether it
312 * has a special name. Returning non-NULL will also cause this
313 * vma to be dumped unconditionally. */
314 const char *(*name)(struct vm_area_struct *vma);
318 * set_policy() op must add a reference to any non-NULL @new mempolicy
319 * to hold the policy upon return. Caller should pass NULL @new to
320 * remove a policy and fall back to surrounding context--i.e. do not
321 * install a MPOL_DEFAULT policy, nor the task or system default
324 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
327 * get_policy() op must add reference [mpol_get()] to any policy at
328 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
329 * in mm/mempolicy.c will do this automatically.
330 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
331 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
332 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
333 * must return NULL--i.e., do not "fallback" to task or system default
336 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
340 * Called by vm_normal_page() for special PTEs to find the
341 * page for @addr. This is useful if the default behavior
342 * (using pte_page()) would not find the correct page.
344 struct page *(*find_special_page)(struct vm_area_struct *vma,
351 #define page_private(page) ((page)->private)
352 #define set_page_private(page, v) ((page)->private = (v))
354 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
355 static inline int pmd_devmap(pmd_t pmd)
362 * FIXME: take this include out, include page-flags.h in
363 * files which need it (119 of them)
365 #include <linux/page-flags.h>
366 #include <linux/huge_mm.h>
369 * Methods to modify the page usage count.
371 * What counts for a page usage:
372 * - cache mapping (page->mapping)
373 * - private data (page->private)
374 * - page mapped in a task's page tables, each mapping
375 * is counted separately
377 * Also, many kernel routines increase the page count before a critical
378 * routine so they can be sure the page doesn't go away from under them.
382 * Drop a ref, return true if the refcount fell to zero (the page has no users)
384 static inline int put_page_testzero(struct page *page)
386 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
387 return atomic_dec_and_test(&page->_count);
391 * Try to grab a ref unless the page has a refcount of zero, return false if
393 * This can be called when MMU is off so it must not access
394 * any of the virtual mappings.
396 static inline int get_page_unless_zero(struct page *page)
398 return atomic_inc_not_zero(&page->_count);
401 extern int page_is_ram(unsigned long pfn);
409 int region_intersects(resource_size_t offset, size_t size, const char *type);
411 /* Support for virtually mapped pages */
412 struct page *vmalloc_to_page(const void *addr);
413 unsigned long vmalloc_to_pfn(const void *addr);
416 * Determine if an address is within the vmalloc range
418 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
419 * is no special casing required.
421 static inline int is_vmalloc_addr(const void *x)
424 unsigned long addr = (unsigned long)x;
426 return addr >= VMALLOC_START && addr < VMALLOC_END;
432 extern int is_vmalloc_or_module_addr(const void *x);
434 static inline int is_vmalloc_or_module_addr(const void *x)
440 extern void kvfree(const void *addr);
442 static inline atomic_t *compound_mapcount_ptr(struct page *page)
444 return &page[1].compound_mapcount;
447 static inline int compound_mapcount(struct page *page)
449 if (!PageCompound(page))
451 page = compound_head(page);
452 return atomic_read(compound_mapcount_ptr(page)) + 1;
456 * The atomic page->_mapcount, starts from -1: so that transitions
457 * both from it and to it can be tracked, using atomic_inc_and_test
458 * and atomic_add_negative(-1).
460 static inline void page_mapcount_reset(struct page *page)
462 atomic_set(&(page)->_mapcount, -1);
465 int __page_mapcount(struct page *page);
467 static inline int page_mapcount(struct page *page)
469 VM_BUG_ON_PAGE(PageSlab(page), page);
471 if (unlikely(PageCompound(page)))
472 return __page_mapcount(page);
473 return atomic_read(&page->_mapcount) + 1;
476 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
477 int total_mapcount(struct page *page);
479 static inline int total_mapcount(struct page *page)
481 return page_mapcount(page);
485 static inline int page_count(struct page *page)
487 return atomic_read(&compound_head(page)->_count);
490 static inline struct page *virt_to_head_page(const void *x)
492 struct page *page = virt_to_page(x);
494 return compound_head(page);
498 * Setup the page count before being freed into the page allocator for
499 * the first time (boot or memory hotplug)
501 static inline void init_page_count(struct page *page)
503 atomic_set(&page->_count, 1);
506 void __put_page(struct page *page);
508 void put_pages_list(struct list_head *pages);
510 void split_page(struct page *page, unsigned int order);
511 int split_free_page(struct page *page);
514 * Compound pages have a destructor function. Provide a
515 * prototype for that function and accessor functions.
516 * These are _only_ valid on the head of a compound page.
518 typedef void compound_page_dtor(struct page *);
520 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
521 enum compound_dtor_id {
524 #ifdef CONFIG_HUGETLB_PAGE
527 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
532 extern compound_page_dtor * const compound_page_dtors[];
534 static inline void set_compound_page_dtor(struct page *page,
535 enum compound_dtor_id compound_dtor)
537 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
538 page[1].compound_dtor = compound_dtor;
541 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
543 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
544 return compound_page_dtors[page[1].compound_dtor];
547 static inline unsigned int compound_order(struct page *page)
551 return page[1].compound_order;
554 static inline void set_compound_order(struct page *page, unsigned int order)
556 page[1].compound_order = order;
559 void free_compound_page(struct page *page);
563 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
564 * servicing faults for write access. In the normal case, do always want
565 * pte_mkwrite. But get_user_pages can cause write faults for mappings
566 * that do not have writing enabled, when used by access_process_vm.
568 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
570 if (likely(vma->vm_flags & VM_WRITE))
571 pte = pte_mkwrite(pte);
575 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
576 struct page *page, pte_t *pte, bool write, bool anon);
580 * Multiple processes may "see" the same page. E.g. for untouched
581 * mappings of /dev/null, all processes see the same page full of
582 * zeroes, and text pages of executables and shared libraries have
583 * only one copy in memory, at most, normally.
585 * For the non-reserved pages, page_count(page) denotes a reference count.
586 * page_count() == 0 means the page is free. page->lru is then used for
587 * freelist management in the buddy allocator.
588 * page_count() > 0 means the page has been allocated.
590 * Pages are allocated by the slab allocator in order to provide memory
591 * to kmalloc and kmem_cache_alloc. In this case, the management of the
592 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
593 * unless a particular usage is carefully commented. (the responsibility of
594 * freeing the kmalloc memory is the caller's, of course).
596 * A page may be used by anyone else who does a __get_free_page().
597 * In this case, page_count still tracks the references, and should only
598 * be used through the normal accessor functions. The top bits of page->flags
599 * and page->virtual store page management information, but all other fields
600 * are unused and could be used privately, carefully. The management of this
601 * page is the responsibility of the one who allocated it, and those who have
602 * subsequently been given references to it.
604 * The other pages (we may call them "pagecache pages") are completely
605 * managed by the Linux memory manager: I/O, buffers, swapping etc.
606 * The following discussion applies only to them.
608 * A pagecache page contains an opaque `private' member, which belongs to the
609 * page's address_space. Usually, this is the address of a circular list of
610 * the page's disk buffers. PG_private must be set to tell the VM to call
611 * into the filesystem to release these pages.
613 * A page may belong to an inode's memory mapping. In this case, page->mapping
614 * is the pointer to the inode, and page->index is the file offset of the page,
615 * in units of PAGE_CACHE_SIZE.
617 * If pagecache pages are not associated with an inode, they are said to be
618 * anonymous pages. These may become associated with the swapcache, and in that
619 * case PG_swapcache is set, and page->private is an offset into the swapcache.
621 * In either case (swapcache or inode backed), the pagecache itself holds one
622 * reference to the page. Setting PG_private should also increment the
623 * refcount. The each user mapping also has a reference to the page.
625 * The pagecache pages are stored in a per-mapping radix tree, which is
626 * rooted at mapping->page_tree, and indexed by offset.
627 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
628 * lists, we instead now tag pages as dirty/writeback in the radix tree.
630 * All pagecache pages may be subject to I/O:
631 * - inode pages may need to be read from disk,
632 * - inode pages which have been modified and are MAP_SHARED may need
633 * to be written back to the inode on disk,
634 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
635 * modified may need to be swapped out to swap space and (later) to be read
640 * The zone field is never updated after free_area_init_core()
641 * sets it, so none of the operations on it need to be atomic.
644 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
645 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
646 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
647 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
648 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
651 * Define the bit shifts to access each section. For non-existent
652 * sections we define the shift as 0; that plus a 0 mask ensures
653 * the compiler will optimise away reference to them.
655 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
656 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
657 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
658 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
660 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
661 #ifdef NODE_NOT_IN_PAGE_FLAGS
662 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
663 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
664 SECTIONS_PGOFF : ZONES_PGOFF)
666 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
667 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
668 NODES_PGOFF : ZONES_PGOFF)
671 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
673 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
674 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
677 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
678 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
679 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
680 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
681 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
683 static inline enum zone_type page_zonenum(const struct page *page)
685 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
688 #ifdef CONFIG_ZONE_DEVICE
689 void get_zone_device_page(struct page *page);
690 void put_zone_device_page(struct page *page);
691 static inline bool is_zone_device_page(const struct page *page)
693 return page_zonenum(page) == ZONE_DEVICE;
696 static inline void get_zone_device_page(struct page *page)
699 static inline void put_zone_device_page(struct page *page)
702 static inline bool is_zone_device_page(const struct page *page)
708 static inline void get_page(struct page *page)
710 page = compound_head(page);
712 * Getting a normal page or the head of a compound page
713 * requires to already have an elevated page->_count.
715 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
716 atomic_inc(&page->_count);
718 if (unlikely(is_zone_device_page(page)))
719 get_zone_device_page(page);
722 static inline void put_page(struct page *page)
724 page = compound_head(page);
726 if (put_page_testzero(page))
729 if (unlikely(is_zone_device_page(page)))
730 put_zone_device_page(page);
733 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
734 #define SECTION_IN_PAGE_FLAGS
738 * The identification function is mainly used by the buddy allocator for
739 * determining if two pages could be buddies. We are not really identifying
740 * the zone since we could be using the section number id if we do not have
741 * node id available in page flags.
742 * We only guarantee that it will return the same value for two combinable
745 static inline int page_zone_id(struct page *page)
747 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
750 static inline int zone_to_nid(struct zone *zone)
759 #ifdef NODE_NOT_IN_PAGE_FLAGS
760 extern int page_to_nid(const struct page *page);
762 static inline int page_to_nid(const struct page *page)
764 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
768 #ifdef CONFIG_NUMA_BALANCING
769 static inline int cpu_pid_to_cpupid(int cpu, int pid)
771 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
774 static inline int cpupid_to_pid(int cpupid)
776 return cpupid & LAST__PID_MASK;
779 static inline int cpupid_to_cpu(int cpupid)
781 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
784 static inline int cpupid_to_nid(int cpupid)
786 return cpu_to_node(cpupid_to_cpu(cpupid));
789 static inline bool cpupid_pid_unset(int cpupid)
791 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
794 static inline bool cpupid_cpu_unset(int cpupid)
796 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
799 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
801 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
804 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
805 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
806 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
808 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
811 static inline int page_cpupid_last(struct page *page)
813 return page->_last_cpupid;
815 static inline void page_cpupid_reset_last(struct page *page)
817 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
820 static inline int page_cpupid_last(struct page *page)
822 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
825 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
827 static inline void page_cpupid_reset_last(struct page *page)
829 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
831 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
832 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
834 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
835 #else /* !CONFIG_NUMA_BALANCING */
836 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
838 return page_to_nid(page); /* XXX */
841 static inline int page_cpupid_last(struct page *page)
843 return page_to_nid(page); /* XXX */
846 static inline int cpupid_to_nid(int cpupid)
851 static inline int cpupid_to_pid(int cpupid)
856 static inline int cpupid_to_cpu(int cpupid)
861 static inline int cpu_pid_to_cpupid(int nid, int pid)
866 static inline bool cpupid_pid_unset(int cpupid)
871 static inline void page_cpupid_reset_last(struct page *page)
875 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
879 #endif /* CONFIG_NUMA_BALANCING */
881 static inline struct zone *page_zone(const struct page *page)
883 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
886 #ifdef SECTION_IN_PAGE_FLAGS
887 static inline void set_page_section(struct page *page, unsigned long section)
889 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
890 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
893 static inline unsigned long page_to_section(const struct page *page)
895 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
899 static inline void set_page_zone(struct page *page, enum zone_type zone)
901 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
902 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
905 static inline void set_page_node(struct page *page, unsigned long node)
907 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
908 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
911 static inline void set_page_links(struct page *page, enum zone_type zone,
912 unsigned long node, unsigned long pfn)
914 set_page_zone(page, zone);
915 set_page_node(page, node);
916 #ifdef SECTION_IN_PAGE_FLAGS
917 set_page_section(page, pfn_to_section_nr(pfn));
922 static inline struct mem_cgroup *page_memcg(struct page *page)
924 return page->mem_cgroup;
927 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
929 page->mem_cgroup = memcg;
932 static inline struct mem_cgroup *page_memcg(struct page *page)
937 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
943 * Some inline functions in vmstat.h depend on page_zone()
945 #include <linux/vmstat.h>
947 static __always_inline void *lowmem_page_address(const struct page *page)
949 return __va(PFN_PHYS(page_to_pfn(page)));
952 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
953 #define HASHED_PAGE_VIRTUAL
956 #if defined(WANT_PAGE_VIRTUAL)
957 static inline void *page_address(const struct page *page)
959 return page->virtual;
961 static inline void set_page_address(struct page *page, void *address)
963 page->virtual = address;
965 #define page_address_init() do { } while(0)
968 #if defined(HASHED_PAGE_VIRTUAL)
969 void *page_address(const struct page *page);
970 void set_page_address(struct page *page, void *virtual);
971 void page_address_init(void);
974 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
975 #define page_address(page) lowmem_page_address(page)
976 #define set_page_address(page, address) do { } while(0)
977 #define page_address_init() do { } while(0)
980 extern void *page_rmapping(struct page *page);
981 extern struct anon_vma *page_anon_vma(struct page *page);
982 extern struct address_space *page_mapping(struct page *page);
984 extern struct address_space *__page_file_mapping(struct page *);
987 struct address_space *page_file_mapping(struct page *page)
989 if (unlikely(PageSwapCache(page)))
990 return __page_file_mapping(page);
992 return page->mapping;
996 * Return the pagecache index of the passed page. Regular pagecache pages
997 * use ->index whereas swapcache pages use ->private
999 static inline pgoff_t page_index(struct page *page)
1001 if (unlikely(PageSwapCache(page)))
1002 return page_private(page);
1006 extern pgoff_t __page_file_index(struct page *page);
1009 * Return the file index of the page. Regular pagecache pages use ->index
1010 * whereas swapcache pages use swp_offset(->private)
1012 static inline pgoff_t page_file_index(struct page *page)
1014 if (unlikely(PageSwapCache(page)))
1015 return __page_file_index(page);
1021 * Return true if this page is mapped into pagetables.
1022 * For compound page it returns true if any subpage of compound page is mapped.
1024 static inline bool page_mapped(struct page *page)
1027 if (likely(!PageCompound(page)))
1028 return atomic_read(&page->_mapcount) >= 0;
1029 page = compound_head(page);
1030 if (atomic_read(compound_mapcount_ptr(page)) >= 0)
1032 for (i = 0; i < hpage_nr_pages(page); i++) {
1033 if (atomic_read(&page[i]._mapcount) >= 0)
1040 * Return true only if the page has been allocated with
1041 * ALLOC_NO_WATERMARKS and the low watermark was not
1042 * met implying that the system is under some pressure.
1044 static inline bool page_is_pfmemalloc(struct page *page)
1047 * Page index cannot be this large so this must be
1048 * a pfmemalloc page.
1050 return page->index == -1UL;
1054 * Only to be called by the page allocator on a freshly allocated
1057 static inline void set_page_pfmemalloc(struct page *page)
1062 static inline void clear_page_pfmemalloc(struct page *page)
1068 * Different kinds of faults, as returned by handle_mm_fault().
1069 * Used to decide whether a process gets delivered SIGBUS or
1070 * just gets major/minor fault counters bumped up.
1073 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1075 #define VM_FAULT_OOM 0x0001
1076 #define VM_FAULT_SIGBUS 0x0002
1077 #define VM_FAULT_MAJOR 0x0004
1078 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1079 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1080 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1081 #define VM_FAULT_SIGSEGV 0x0040
1083 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1084 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1085 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1086 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1088 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1090 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1091 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1094 /* Encode hstate index for a hwpoisoned large page */
1095 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1096 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1099 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1101 extern void pagefault_out_of_memory(void);
1103 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1106 * Flags passed to show_mem() and show_free_areas() to suppress output in
1109 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1111 extern void show_free_areas(unsigned int flags);
1112 extern bool skip_free_areas_node(unsigned int flags, int nid);
1114 int shmem_zero_setup(struct vm_area_struct *);
1116 bool shmem_mapping(struct address_space *mapping);
1118 static inline bool shmem_mapping(struct address_space *mapping)
1124 extern bool can_do_mlock(void);
1125 extern int user_shm_lock(size_t, struct user_struct *);
1126 extern void user_shm_unlock(size_t, struct user_struct *);
1129 * Parameter block passed down to zap_pte_range in exceptional cases.
1131 struct zap_details {
1132 struct address_space *check_mapping; /* Check page->mapping if set */
1133 pgoff_t first_index; /* Lowest page->index to unmap */
1134 pgoff_t last_index; /* Highest page->index to unmap */
1137 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1140 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1141 unsigned long size);
1142 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1143 unsigned long size, struct zap_details *);
1144 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1145 unsigned long start, unsigned long end);
1148 * mm_walk - callbacks for walk_page_range
1149 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1150 * this handler is required to be able to handle
1151 * pmd_trans_huge() pmds. They may simply choose to
1152 * split_huge_page() instead of handling it explicitly.
1153 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1154 * @pte_hole: if set, called for each hole at all levels
1155 * @hugetlb_entry: if set, called for each hugetlb entry
1156 * @test_walk: caller specific callback function to determine whether
1157 * we walk over the current vma or not. A positive returned
1158 * value means "do page table walk over the current vma,"
1159 * and a negative one means "abort current page table walk
1160 * right now." 0 means "skip the current vma."
1161 * @mm: mm_struct representing the target process of page table walk
1162 * @vma: vma currently walked (NULL if walking outside vmas)
1163 * @private: private data for callbacks' usage
1165 * (see the comment on walk_page_range() for more details)
1168 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1169 unsigned long next, struct mm_walk *walk);
1170 int (*pte_entry)(pte_t *pte, unsigned long addr,
1171 unsigned long next, struct mm_walk *walk);
1172 int (*pte_hole)(unsigned long addr, unsigned long next,
1173 struct mm_walk *walk);
1174 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1175 unsigned long addr, unsigned long next,
1176 struct mm_walk *walk);
1177 int (*test_walk)(unsigned long addr, unsigned long next,
1178 struct mm_walk *walk);
1179 struct mm_struct *mm;
1180 struct vm_area_struct *vma;
1184 int walk_page_range(unsigned long addr, unsigned long end,
1185 struct mm_walk *walk);
1186 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1187 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1188 unsigned long end, unsigned long floor, unsigned long ceiling);
1189 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1190 struct vm_area_struct *vma);
1191 void unmap_mapping_range(struct address_space *mapping,
1192 loff_t const holebegin, loff_t const holelen, int even_cows);
1193 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1194 unsigned long *pfn);
1195 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1196 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1197 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1198 void *buf, int len, int write);
1200 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1201 loff_t const holebegin, loff_t const holelen)
1203 unmap_mapping_range(mapping, holebegin, holelen, 0);
1206 extern void truncate_pagecache(struct inode *inode, loff_t new);
1207 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1208 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1209 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1210 int truncate_inode_page(struct address_space *mapping, struct page *page);
1211 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1212 int invalidate_inode_page(struct page *page);
1215 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1216 unsigned long address, unsigned int flags);
1217 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1218 unsigned long address, unsigned int fault_flags,
1221 static inline int handle_mm_fault(struct mm_struct *mm,
1222 struct vm_area_struct *vma, unsigned long address,
1225 /* should never happen if there's no MMU */
1227 return VM_FAULT_SIGBUS;
1229 static inline int fixup_user_fault(struct task_struct *tsk,
1230 struct mm_struct *mm, unsigned long address,
1231 unsigned int fault_flags, bool *unlocked)
1233 /* should never happen if there's no MMU */
1239 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1240 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1241 void *buf, int len, int write);
1243 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1244 unsigned long start, unsigned long nr_pages,
1245 unsigned int foll_flags, struct page **pages,
1246 struct vm_area_struct **vmas, int *nonblocking);
1247 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1248 unsigned long start, unsigned long nr_pages,
1249 int write, int force, struct page **pages,
1250 struct vm_area_struct **vmas);
1251 long get_user_pages6(unsigned long start, unsigned long nr_pages,
1252 int write, int force, struct page **pages,
1253 struct vm_area_struct **vmas);
1254 long get_user_pages_locked6(unsigned long start, unsigned long nr_pages,
1255 int write, int force, struct page **pages, int *locked);
1256 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1257 unsigned long start, unsigned long nr_pages,
1258 int write, int force, struct page **pages,
1259 unsigned int gup_flags);
1260 long get_user_pages_unlocked5(unsigned long start, unsigned long nr_pages,
1261 int write, int force, struct page **pages);
1262 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1263 struct page **pages);
1265 /* suppress warnings from use in EXPORT_SYMBOL() */
1266 #ifndef __DISABLE_GUP_DEPRECATED
1267 #define __gup_deprecated __deprecated
1269 #define __gup_deprecated
1272 * These macros provide backward-compatibility with the old
1273 * get_user_pages() variants which took tsk/mm. These
1274 * functions/macros provide both compile-time __deprecated so we
1275 * can catch old-style use and not break the build. The actual
1276 * functions also have WARN_ON()s to let us know at runtime if
1277 * the get_user_pages() should have been the "remote" variant.
1279 * These are hideous, but temporary.
1281 * If you run into one of these __deprecated warnings, look
1282 * at how you are calling get_user_pages(). If you are calling
1283 * it with current/current->mm as the first two arguments,
1284 * simply remove those arguments. The behavior will be the same
1285 * as it is now. If you are calling it on another task, use
1286 * get_user_pages_remote() instead.
1288 * Any questions? Ask Dave Hansen <dave@sr71.net>
1292 get_user_pages8(struct task_struct *tsk, struct mm_struct *mm,
1293 unsigned long start, unsigned long nr_pages,
1294 int write, int force, struct page **pages,
1295 struct vm_area_struct **vmas);
1296 #define GUP_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, get_user_pages, ...) \
1298 #define get_user_pages(...) GUP_MACRO(__VA_ARGS__, \
1299 get_user_pages8, x, \
1300 get_user_pages6, x, x, x, x, x)(__VA_ARGS__)
1303 long get_user_pages_locked8(struct task_struct *tsk, struct mm_struct *mm,
1304 unsigned long start, unsigned long nr_pages,
1305 int write, int force, struct page **pages,
1307 #define GUPL_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, get_user_pages_locked, ...) \
1308 get_user_pages_locked
1309 #define get_user_pages_locked(...) GUPL_MACRO(__VA_ARGS__, \
1310 get_user_pages_locked8, x, \
1311 get_user_pages_locked6, x, x, x, x)(__VA_ARGS__)
1314 long get_user_pages_unlocked7(struct task_struct *tsk, struct mm_struct *mm,
1315 unsigned long start, unsigned long nr_pages,
1316 int write, int force, struct page **pages);
1317 #define GUPU_MACRO(_1, _2, _3, _4, _5, _6, _7, get_user_pages_unlocked, ...) \
1318 get_user_pages_unlocked
1319 #define get_user_pages_unlocked(...) GUPU_MACRO(__VA_ARGS__, \
1320 get_user_pages_unlocked7, x, \
1321 get_user_pages_unlocked5, x, x, x, x)(__VA_ARGS__)
1323 /* Container for pinned pfns / pages */
1324 struct frame_vector {
1325 unsigned int nr_allocated; /* Number of frames we have space for */
1326 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1327 bool got_ref; /* Did we pin pages by getting page ref? */
1328 bool is_pfns; /* Does array contain pages or pfns? */
1329 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1330 * pfns_vector_pages() or pfns_vector_pfns()
1334 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1335 void frame_vector_destroy(struct frame_vector *vec);
1336 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1337 bool write, bool force, struct frame_vector *vec);
1338 void put_vaddr_frames(struct frame_vector *vec);
1339 int frame_vector_to_pages(struct frame_vector *vec);
1340 void frame_vector_to_pfns(struct frame_vector *vec);
1342 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1344 return vec->nr_frames;
1347 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1350 int err = frame_vector_to_pages(vec);
1353 return ERR_PTR(err);
1355 return (struct page **)(vec->ptrs);
1358 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1361 frame_vector_to_pfns(vec);
1362 return (unsigned long *)(vec->ptrs);
1366 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1367 struct page **pages);
1368 int get_kernel_page(unsigned long start, int write, struct page **pages);
1369 struct page *get_dump_page(unsigned long addr);
1371 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1372 extern void do_invalidatepage(struct page *page, unsigned int offset,
1373 unsigned int length);
1375 int __set_page_dirty_nobuffers(struct page *page);
1376 int __set_page_dirty_no_writeback(struct page *page);
1377 int redirty_page_for_writepage(struct writeback_control *wbc,
1379 void account_page_dirtied(struct page *page, struct address_space *mapping,
1380 struct mem_cgroup *memcg);
1381 void account_page_cleaned(struct page *page, struct address_space *mapping,
1382 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1383 int set_page_dirty(struct page *page);
1384 int set_page_dirty_lock(struct page *page);
1385 void cancel_dirty_page(struct page *page);
1386 int clear_page_dirty_for_io(struct page *page);
1388 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1390 /* Is the vma a continuation of the stack vma above it? */
1391 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1393 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1396 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1398 return !vma->vm_ops;
1401 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1404 return (vma->vm_flags & VM_GROWSDOWN) &&
1405 (vma->vm_start == addr) &&
1406 !vma_growsdown(vma->vm_prev, addr);
1409 /* Is the vma a continuation of the stack vma below it? */
1410 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1412 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1415 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1418 return (vma->vm_flags & VM_GROWSUP) &&
1419 (vma->vm_end == addr) &&
1420 !vma_growsup(vma->vm_next, addr);
1423 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1425 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1426 unsigned long old_addr, struct vm_area_struct *new_vma,
1427 unsigned long new_addr, unsigned long len,
1428 bool need_rmap_locks);
1429 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1430 unsigned long end, pgprot_t newprot,
1431 int dirty_accountable, int prot_numa);
1432 extern int mprotect_fixup(struct vm_area_struct *vma,
1433 struct vm_area_struct **pprev, unsigned long start,
1434 unsigned long end, unsigned long newflags);
1437 * doesn't attempt to fault and will return short.
1439 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1440 struct page **pages);
1442 * per-process(per-mm_struct) statistics.
1444 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1446 long val = atomic_long_read(&mm->rss_stat.count[member]);
1448 #ifdef SPLIT_RSS_COUNTING
1450 * counter is updated in asynchronous manner and may go to minus.
1451 * But it's never be expected number for users.
1456 return (unsigned long)val;
1459 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1461 atomic_long_add(value, &mm->rss_stat.count[member]);
1464 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1466 atomic_long_inc(&mm->rss_stat.count[member]);
1469 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1471 atomic_long_dec(&mm->rss_stat.count[member]);
1474 /* Optimized variant when page is already known not to be PageAnon */
1475 static inline int mm_counter_file(struct page *page)
1477 if (PageSwapBacked(page))
1478 return MM_SHMEMPAGES;
1479 return MM_FILEPAGES;
1482 static inline int mm_counter(struct page *page)
1485 return MM_ANONPAGES;
1486 return mm_counter_file(page);
1489 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1491 return get_mm_counter(mm, MM_FILEPAGES) +
1492 get_mm_counter(mm, MM_ANONPAGES) +
1493 get_mm_counter(mm, MM_SHMEMPAGES);
1496 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1498 return max(mm->hiwater_rss, get_mm_rss(mm));
1501 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1503 return max(mm->hiwater_vm, mm->total_vm);
1506 static inline void update_hiwater_rss(struct mm_struct *mm)
1508 unsigned long _rss = get_mm_rss(mm);
1510 if ((mm)->hiwater_rss < _rss)
1511 (mm)->hiwater_rss = _rss;
1514 static inline void update_hiwater_vm(struct mm_struct *mm)
1516 if (mm->hiwater_vm < mm->total_vm)
1517 mm->hiwater_vm = mm->total_vm;
1520 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1522 mm->hiwater_rss = get_mm_rss(mm);
1525 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1526 struct mm_struct *mm)
1528 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1530 if (*maxrss < hiwater_rss)
1531 *maxrss = hiwater_rss;
1534 #if defined(SPLIT_RSS_COUNTING)
1535 void sync_mm_rss(struct mm_struct *mm);
1537 static inline void sync_mm_rss(struct mm_struct *mm)
1542 #ifndef __HAVE_ARCH_PTE_DEVMAP
1543 static inline int pte_devmap(pte_t pte)
1549 int vma_wants_writenotify(struct vm_area_struct *vma);
1551 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1553 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1557 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1561 #ifdef __PAGETABLE_PUD_FOLDED
1562 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1563 unsigned long address)
1568 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1571 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1572 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1573 unsigned long address)
1578 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1580 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1585 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1586 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1589 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1591 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1593 atomic_long_set(&mm->nr_pmds, 0);
1596 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1598 return atomic_long_read(&mm->nr_pmds);
1601 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1603 atomic_long_inc(&mm->nr_pmds);
1606 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1608 atomic_long_dec(&mm->nr_pmds);
1612 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1613 pmd_t *pmd, unsigned long address);
1614 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1617 * The following ifdef needed to get the 4level-fixup.h header to work.
1618 * Remove it when 4level-fixup.h has been removed.
1620 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1621 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1623 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1624 NULL: pud_offset(pgd, address);
1627 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1629 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1630 NULL: pmd_offset(pud, address);
1632 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1634 #if USE_SPLIT_PTE_PTLOCKS
1635 #if ALLOC_SPLIT_PTLOCKS
1636 void __init ptlock_cache_init(void);
1637 extern bool ptlock_alloc(struct page *page);
1638 extern void ptlock_free(struct page *page);
1640 static inline spinlock_t *ptlock_ptr(struct page *page)
1644 #else /* ALLOC_SPLIT_PTLOCKS */
1645 static inline void ptlock_cache_init(void)
1649 static inline bool ptlock_alloc(struct page *page)
1654 static inline void ptlock_free(struct page *page)
1658 static inline spinlock_t *ptlock_ptr(struct page *page)
1662 #endif /* ALLOC_SPLIT_PTLOCKS */
1664 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1666 return ptlock_ptr(pmd_page(*pmd));
1669 static inline bool ptlock_init(struct page *page)
1672 * prep_new_page() initialize page->private (and therefore page->ptl)
1673 * with 0. Make sure nobody took it in use in between.
1675 * It can happen if arch try to use slab for page table allocation:
1676 * slab code uses page->slab_cache, which share storage with page->ptl.
1678 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1679 if (!ptlock_alloc(page))
1681 spin_lock_init(ptlock_ptr(page));
1685 /* Reset page->mapping so free_pages_check won't complain. */
1686 static inline void pte_lock_deinit(struct page *page)
1688 page->mapping = NULL;
1692 #else /* !USE_SPLIT_PTE_PTLOCKS */
1694 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1696 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1698 return &mm->page_table_lock;
1700 static inline void ptlock_cache_init(void) {}
1701 static inline bool ptlock_init(struct page *page) { return true; }
1702 static inline void pte_lock_deinit(struct page *page) {}
1703 #endif /* USE_SPLIT_PTE_PTLOCKS */
1705 static inline void pgtable_init(void)
1707 ptlock_cache_init();
1708 pgtable_cache_init();
1711 static inline bool pgtable_page_ctor(struct page *page)
1713 if (!ptlock_init(page))
1715 inc_zone_page_state(page, NR_PAGETABLE);
1719 static inline void pgtable_page_dtor(struct page *page)
1721 pte_lock_deinit(page);
1722 dec_zone_page_state(page, NR_PAGETABLE);
1725 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1727 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1728 pte_t *__pte = pte_offset_map(pmd, address); \
1734 #define pte_unmap_unlock(pte, ptl) do { \
1739 #define pte_alloc_map(mm, vma, pmd, address) \
1740 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1742 NULL: pte_offset_map(pmd, address))
1744 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1745 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1747 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1749 #define pte_alloc_kernel(pmd, address) \
1750 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1751 NULL: pte_offset_kernel(pmd, address))
1753 #if USE_SPLIT_PMD_PTLOCKS
1755 static struct page *pmd_to_page(pmd_t *pmd)
1757 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1758 return virt_to_page((void *)((unsigned long) pmd & mask));
1761 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1763 return ptlock_ptr(pmd_to_page(pmd));
1766 static inline bool pgtable_pmd_page_ctor(struct page *page)
1768 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1769 page->pmd_huge_pte = NULL;
1771 return ptlock_init(page);
1774 static inline void pgtable_pmd_page_dtor(struct page *page)
1776 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1777 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1782 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1786 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1788 return &mm->page_table_lock;
1791 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1792 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1794 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1798 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1800 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1805 extern void free_area_init(unsigned long * zones_size);
1806 extern void free_area_init_node(int nid, unsigned long * zones_size,
1807 unsigned long zone_start_pfn, unsigned long *zholes_size);
1808 extern void free_initmem(void);
1811 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1812 * into the buddy system. The freed pages will be poisoned with pattern
1813 * "poison" if it's within range [0, UCHAR_MAX].
1814 * Return pages freed into the buddy system.
1816 extern unsigned long free_reserved_area(void *start, void *end,
1817 int poison, char *s);
1819 #ifdef CONFIG_HIGHMEM
1821 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1822 * and totalram_pages.
1824 extern void free_highmem_page(struct page *page);
1827 extern void adjust_managed_page_count(struct page *page, long count);
1828 extern void mem_init_print_info(const char *str);
1830 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1832 /* Free the reserved page into the buddy system, so it gets managed. */
1833 static inline void __free_reserved_page(struct page *page)
1835 ClearPageReserved(page);
1836 init_page_count(page);
1840 static inline void free_reserved_page(struct page *page)
1842 __free_reserved_page(page);
1843 adjust_managed_page_count(page, 1);
1846 static inline void mark_page_reserved(struct page *page)
1848 SetPageReserved(page);
1849 adjust_managed_page_count(page, -1);
1853 * Default method to free all the __init memory into the buddy system.
1854 * The freed pages will be poisoned with pattern "poison" if it's within
1855 * range [0, UCHAR_MAX].
1856 * Return pages freed into the buddy system.
1858 static inline unsigned long free_initmem_default(int poison)
1860 extern char __init_begin[], __init_end[];
1862 return free_reserved_area(&__init_begin, &__init_end,
1863 poison, "unused kernel");
1866 static inline unsigned long get_num_physpages(void)
1869 unsigned long phys_pages = 0;
1871 for_each_online_node(nid)
1872 phys_pages += node_present_pages(nid);
1877 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1879 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1880 * zones, allocate the backing mem_map and account for memory holes in a more
1881 * architecture independent manner. This is a substitute for creating the
1882 * zone_sizes[] and zholes_size[] arrays and passing them to
1883 * free_area_init_node()
1885 * An architecture is expected to register range of page frames backed by
1886 * physical memory with memblock_add[_node]() before calling
1887 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1888 * usage, an architecture is expected to do something like
1890 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1892 * for_each_valid_physical_page_range()
1893 * memblock_add_node(base, size, nid)
1894 * free_area_init_nodes(max_zone_pfns);
1896 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1897 * registered physical page range. Similarly
1898 * sparse_memory_present_with_active_regions() calls memory_present() for
1899 * each range when SPARSEMEM is enabled.
1901 * See mm/page_alloc.c for more information on each function exposed by
1902 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1904 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1905 unsigned long node_map_pfn_alignment(void);
1906 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1907 unsigned long end_pfn);
1908 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1909 unsigned long end_pfn);
1910 extern void get_pfn_range_for_nid(unsigned int nid,
1911 unsigned long *start_pfn, unsigned long *end_pfn);
1912 extern unsigned long find_min_pfn_with_active_regions(void);
1913 extern void free_bootmem_with_active_regions(int nid,
1914 unsigned long max_low_pfn);
1915 extern void sparse_memory_present_with_active_regions(int nid);
1917 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1919 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1920 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1921 static inline int __early_pfn_to_nid(unsigned long pfn,
1922 struct mminit_pfnnid_cache *state)
1927 /* please see mm/page_alloc.c */
1928 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1929 /* there is a per-arch backend function. */
1930 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1931 struct mminit_pfnnid_cache *state);
1934 extern void set_dma_reserve(unsigned long new_dma_reserve);
1935 extern void memmap_init_zone(unsigned long, int, unsigned long,
1936 unsigned long, enum memmap_context);
1937 extern void setup_per_zone_wmarks(void);
1938 extern int __meminit init_per_zone_wmark_min(void);
1939 extern void mem_init(void);
1940 extern void __init mmap_init(void);
1941 extern void show_mem(unsigned int flags);
1942 extern void si_meminfo(struct sysinfo * val);
1943 extern void si_meminfo_node(struct sysinfo *val, int nid);
1945 extern __printf(3, 4)
1946 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1947 const char *fmt, ...);
1949 extern void setup_per_cpu_pageset(void);
1951 extern void zone_pcp_update(struct zone *zone);
1952 extern void zone_pcp_reset(struct zone *zone);
1955 extern int min_free_kbytes;
1958 extern atomic_long_t mmap_pages_allocated;
1959 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1961 /* interval_tree.c */
1962 void vma_interval_tree_insert(struct vm_area_struct *node,
1963 struct rb_root *root);
1964 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1965 struct vm_area_struct *prev,
1966 struct rb_root *root);
1967 void vma_interval_tree_remove(struct vm_area_struct *node,
1968 struct rb_root *root);
1969 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1970 unsigned long start, unsigned long last);
1971 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1972 unsigned long start, unsigned long last);
1974 #define vma_interval_tree_foreach(vma, root, start, last) \
1975 for (vma = vma_interval_tree_iter_first(root, start, last); \
1976 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1978 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1979 struct rb_root *root);
1980 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1981 struct rb_root *root);
1982 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1983 struct rb_root *root, unsigned long start, unsigned long last);
1984 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1985 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1986 #ifdef CONFIG_DEBUG_VM_RB
1987 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1990 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1991 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1992 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1995 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1996 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1997 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1998 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1999 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2000 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2001 struct mempolicy *, struct vm_userfaultfd_ctx);
2002 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2003 extern int split_vma(struct mm_struct *,
2004 struct vm_area_struct *, unsigned long addr, int new_below);
2005 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2006 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2007 struct rb_node **, struct rb_node *);
2008 extern void unlink_file_vma(struct vm_area_struct *);
2009 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2010 unsigned long addr, unsigned long len, pgoff_t pgoff,
2011 bool *need_rmap_locks);
2012 extern void exit_mmap(struct mm_struct *);
2014 static inline int check_data_rlimit(unsigned long rlim,
2016 unsigned long start,
2017 unsigned long end_data,
2018 unsigned long start_data)
2020 if (rlim < RLIM_INFINITY) {
2021 if (((new - start) + (end_data - start_data)) > rlim)
2028 extern int mm_take_all_locks(struct mm_struct *mm);
2029 extern void mm_drop_all_locks(struct mm_struct *mm);
2031 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2032 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2034 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2035 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2037 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2038 unsigned long addr, unsigned long len,
2039 unsigned long flags,
2040 const struct vm_special_mapping *spec);
2041 /* This is an obsolete alternative to _install_special_mapping. */
2042 extern int install_special_mapping(struct mm_struct *mm,
2043 unsigned long addr, unsigned long len,
2044 unsigned long flags, struct page **pages);
2046 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2048 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2049 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2050 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2051 unsigned long len, unsigned long prot, unsigned long flags,
2052 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2053 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2055 static inline unsigned long
2056 do_mmap_pgoff(struct file *file, unsigned long addr,
2057 unsigned long len, unsigned long prot, unsigned long flags,
2058 unsigned long pgoff, unsigned long *populate)
2060 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2064 extern int __mm_populate(unsigned long addr, unsigned long len,
2066 static inline void mm_populate(unsigned long addr, unsigned long len)
2069 (void) __mm_populate(addr, len, 1);
2072 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2075 /* These take the mm semaphore themselves */
2076 extern unsigned long vm_brk(unsigned long, unsigned long);
2077 extern int vm_munmap(unsigned long, size_t);
2078 extern unsigned long vm_mmap(struct file *, unsigned long,
2079 unsigned long, unsigned long,
2080 unsigned long, unsigned long);
2082 struct vm_unmapped_area_info {
2083 #define VM_UNMAPPED_AREA_TOPDOWN 1
2084 unsigned long flags;
2085 unsigned long length;
2086 unsigned long low_limit;
2087 unsigned long high_limit;
2088 unsigned long align_mask;
2089 unsigned long align_offset;
2092 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2093 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2096 * Search for an unmapped address range.
2098 * We are looking for a range that:
2099 * - does not intersect with any VMA;
2100 * - is contained within the [low_limit, high_limit) interval;
2101 * - is at least the desired size.
2102 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2104 static inline unsigned long
2105 vm_unmapped_area(struct vm_unmapped_area_info *info)
2107 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2108 return unmapped_area_topdown(info);
2110 return unmapped_area(info);
2114 extern void truncate_inode_pages(struct address_space *, loff_t);
2115 extern void truncate_inode_pages_range(struct address_space *,
2116 loff_t lstart, loff_t lend);
2117 extern void truncate_inode_pages_final(struct address_space *);
2119 /* generic vm_area_ops exported for stackable file systems */
2120 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2121 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2122 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2124 /* mm/page-writeback.c */
2125 int write_one_page(struct page *page, int wait);
2126 void task_dirty_inc(struct task_struct *tsk);
2129 #define VM_MAX_READAHEAD 128 /* kbytes */
2130 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2132 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2133 pgoff_t offset, unsigned long nr_to_read);
2135 void page_cache_sync_readahead(struct address_space *mapping,
2136 struct file_ra_state *ra,
2139 unsigned long size);
2141 void page_cache_async_readahead(struct address_space *mapping,
2142 struct file_ra_state *ra,
2146 unsigned long size);
2148 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2149 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2151 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2152 extern int expand_downwards(struct vm_area_struct *vma,
2153 unsigned long address);
2155 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2157 #define expand_upwards(vma, address) (0)
2160 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2161 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2162 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2163 struct vm_area_struct **pprev);
2165 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2166 NULL if none. Assume start_addr < end_addr. */
2167 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2169 struct vm_area_struct * vma = find_vma(mm,start_addr);
2171 if (vma && end_addr <= vma->vm_start)
2176 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2178 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2181 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2182 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2183 unsigned long vm_start, unsigned long vm_end)
2185 struct vm_area_struct *vma = find_vma(mm, vm_start);
2187 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2194 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2195 void vma_set_page_prot(struct vm_area_struct *vma);
2197 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2201 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2203 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2207 #ifdef CONFIG_NUMA_BALANCING
2208 unsigned long change_prot_numa(struct vm_area_struct *vma,
2209 unsigned long start, unsigned long end);
2212 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2213 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2214 unsigned long pfn, unsigned long size, pgprot_t);
2215 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2216 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2218 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2219 unsigned long pfn, pgprot_t pgprot);
2220 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2222 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2225 struct page *follow_page_mask(struct vm_area_struct *vma,
2226 unsigned long address, unsigned int foll_flags,
2227 unsigned int *page_mask);
2229 static inline struct page *follow_page(struct vm_area_struct *vma,
2230 unsigned long address, unsigned int foll_flags)
2232 unsigned int unused_page_mask;
2233 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2236 #define FOLL_WRITE 0x01 /* check pte is writable */
2237 #define FOLL_TOUCH 0x02 /* mark page accessed */
2238 #define FOLL_GET 0x04 /* do get_page on page */
2239 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2240 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2241 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2242 * and return without waiting upon it */
2243 #define FOLL_POPULATE 0x40 /* fault in page */
2244 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2245 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2246 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2247 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2248 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2249 #define FOLL_MLOCK 0x1000 /* lock present pages */
2250 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2252 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2254 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2255 unsigned long size, pte_fn_t fn, void *data);
2258 #ifdef CONFIG_DEBUG_PAGEALLOC
2259 extern bool _debug_pagealloc_enabled;
2260 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2262 static inline bool debug_pagealloc_enabled(void)
2264 return _debug_pagealloc_enabled;
2268 kernel_map_pages(struct page *page, int numpages, int enable)
2270 if (!debug_pagealloc_enabled())
2273 __kernel_map_pages(page, numpages, enable);
2275 #ifdef CONFIG_HIBERNATION
2276 extern bool kernel_page_present(struct page *page);
2277 #endif /* CONFIG_HIBERNATION */
2280 kernel_map_pages(struct page *page, int numpages, int enable) {}
2281 #ifdef CONFIG_HIBERNATION
2282 static inline bool kernel_page_present(struct page *page) { return true; }
2283 #endif /* CONFIG_HIBERNATION */
2286 #ifdef __HAVE_ARCH_GATE_AREA
2287 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2288 extern int in_gate_area_no_mm(unsigned long addr);
2289 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2291 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2295 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2296 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2300 #endif /* __HAVE_ARCH_GATE_AREA */
2302 #ifdef CONFIG_SYSCTL
2303 extern int sysctl_drop_caches;
2304 int drop_caches_sysctl_handler(struct ctl_table *, int,
2305 void __user *, size_t *, loff_t *);
2308 void drop_slab(void);
2309 void drop_slab_node(int nid);
2312 #define randomize_va_space 0
2314 extern int randomize_va_space;
2317 const char * arch_vma_name(struct vm_area_struct *vma);
2318 void print_vma_addr(char *prefix, unsigned long rip);
2320 void sparse_mem_maps_populate_node(struct page **map_map,
2321 unsigned long pnum_begin,
2322 unsigned long pnum_end,
2323 unsigned long map_count,
2326 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2327 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2328 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2329 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2330 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2331 void *vmemmap_alloc_block(unsigned long size, int node);
2333 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2334 struct vmem_altmap *altmap);
2335 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2337 return __vmemmap_alloc_block_buf(size, node, NULL);
2340 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2341 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2343 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2344 void vmemmap_populate_print_last(void);
2345 #ifdef CONFIG_MEMORY_HOTPLUG
2346 void vmemmap_free(unsigned long start, unsigned long end);
2348 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2349 unsigned long size);
2352 MF_COUNT_INCREASED = 1 << 0,
2353 MF_ACTION_REQUIRED = 1 << 1,
2354 MF_MUST_KILL = 1 << 2,
2355 MF_SOFT_OFFLINE = 1 << 3,
2357 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2358 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2359 extern int unpoison_memory(unsigned long pfn);
2360 extern int get_hwpoison_page(struct page *page);
2361 #define put_hwpoison_page(page) put_page(page)
2362 extern int sysctl_memory_failure_early_kill;
2363 extern int sysctl_memory_failure_recovery;
2364 extern void shake_page(struct page *p, int access);
2365 extern atomic_long_t num_poisoned_pages;
2366 extern int soft_offline_page(struct page *page, int flags);
2370 * Error handlers for various types of pages.
2373 MF_IGNORED, /* Error: cannot be handled */
2374 MF_FAILED, /* Error: handling failed */
2375 MF_DELAYED, /* Will be handled later */
2376 MF_RECOVERED, /* Successfully recovered */
2379 enum mf_action_page_type {
2381 MF_MSG_KERNEL_HIGH_ORDER,
2383 MF_MSG_DIFFERENT_COMPOUND,
2384 MF_MSG_POISONED_HUGE,
2387 MF_MSG_UNMAP_FAILED,
2388 MF_MSG_DIRTY_SWAPCACHE,
2389 MF_MSG_CLEAN_SWAPCACHE,
2390 MF_MSG_DIRTY_MLOCKED_LRU,
2391 MF_MSG_CLEAN_MLOCKED_LRU,
2392 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2393 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2396 MF_MSG_TRUNCATED_LRU,
2402 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2403 extern void clear_huge_page(struct page *page,
2405 unsigned int pages_per_huge_page);
2406 extern void copy_user_huge_page(struct page *dst, struct page *src,
2407 unsigned long addr, struct vm_area_struct *vma,
2408 unsigned int pages_per_huge_page);
2409 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2411 extern struct page_ext_operations debug_guardpage_ops;
2412 extern struct page_ext_operations page_poisoning_ops;
2414 #ifdef CONFIG_DEBUG_PAGEALLOC
2415 extern unsigned int _debug_guardpage_minorder;
2416 extern bool _debug_guardpage_enabled;
2418 static inline unsigned int debug_guardpage_minorder(void)
2420 return _debug_guardpage_minorder;
2423 static inline bool debug_guardpage_enabled(void)
2425 return _debug_guardpage_enabled;
2428 static inline bool page_is_guard(struct page *page)
2430 struct page_ext *page_ext;
2432 if (!debug_guardpage_enabled())
2435 page_ext = lookup_page_ext(page);
2436 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2439 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2440 static inline bool debug_guardpage_enabled(void) { return false; }
2441 static inline bool page_is_guard(struct page *page) { return false; }
2442 #endif /* CONFIG_DEBUG_PAGEALLOC */
2444 #if MAX_NUMNODES > 1
2445 void __init setup_nr_node_ids(void);
2447 static inline void setup_nr_node_ids(void) {}
2450 #endif /* __KERNEL__ */
2451 #endif /* _LINUX_MM_H */