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 */
255 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
258 * vm_fault is filled by the the pagefault handler and passed to the vma's
259 * ->fault function. The vma's ->fault is responsible for returning a bitmask
260 * of VM_FAULT_xxx flags that give details about how the fault was handled.
262 * MM layer fills up gfp_mask for page allocations but fault handler might
263 * alter it if its implementation requires a different allocation context.
265 * pgoff should be used in favour of virtual_address, if possible.
268 unsigned int flags; /* FAULT_FLAG_xxx flags */
269 gfp_t gfp_mask; /* gfp mask to be used for allocations */
270 pgoff_t pgoff; /* Logical page offset based on vma */
271 void __user *virtual_address; /* Faulting virtual address */
273 struct page *cow_page; /* Handler may choose to COW */
274 struct page *page; /* ->fault handlers should return a
275 * page here, unless VM_FAULT_NOPAGE
276 * is set (which is also implied by
279 /* for ->map_pages() only */
280 pgoff_t max_pgoff; /* map pages for offset from pgoff till
281 * max_pgoff inclusive */
282 pte_t *pte; /* pte entry associated with ->pgoff */
286 * These are the virtual MM functions - opening of an area, closing and
287 * unmapping it (needed to keep files on disk up-to-date etc), pointer
288 * to the functions called when a no-page or a wp-page exception occurs.
290 struct vm_operations_struct {
291 void (*open)(struct vm_area_struct * area);
292 void (*close)(struct vm_area_struct * area);
293 int (*mremap)(struct vm_area_struct * area);
294 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
295 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
296 pmd_t *, unsigned int flags);
297 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
299 /* notification that a previously read-only page is about to become
300 * writable, if an error is returned it will cause a SIGBUS */
301 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
303 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
304 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
306 /* called by access_process_vm when get_user_pages() fails, typically
307 * for use by special VMAs that can switch between memory and hardware
309 int (*access)(struct vm_area_struct *vma, unsigned long addr,
310 void *buf, int len, int write);
312 /* Called by the /proc/PID/maps code to ask the vma whether it
313 * has a special name. Returning non-NULL will also cause this
314 * vma to be dumped unconditionally. */
315 const char *(*name)(struct vm_area_struct *vma);
319 * set_policy() op must add a reference to any non-NULL @new mempolicy
320 * to hold the policy upon return. Caller should pass NULL @new to
321 * remove a policy and fall back to surrounding context--i.e. do not
322 * install a MPOL_DEFAULT policy, nor the task or system default
325 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
328 * get_policy() op must add reference [mpol_get()] to any policy at
329 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
330 * in mm/mempolicy.c will do this automatically.
331 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
332 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
333 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
334 * must return NULL--i.e., do not "fallback" to task or system default
337 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
341 * Called by vm_normal_page() for special PTEs to find the
342 * page for @addr. This is useful if the default behavior
343 * (using pte_page()) would not find the correct page.
345 struct page *(*find_special_page)(struct vm_area_struct *vma,
352 #define page_private(page) ((page)->private)
353 #define set_page_private(page, v) ((page)->private = (v))
355 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
356 static inline int pmd_devmap(pmd_t pmd)
363 * FIXME: take this include out, include page-flags.h in
364 * files which need it (119 of them)
366 #include <linux/page-flags.h>
367 #include <linux/huge_mm.h>
370 * Methods to modify the page usage count.
372 * What counts for a page usage:
373 * - cache mapping (page->mapping)
374 * - private data (page->private)
375 * - page mapped in a task's page tables, each mapping
376 * is counted separately
378 * Also, many kernel routines increase the page count before a critical
379 * routine so they can be sure the page doesn't go away from under them.
383 * Drop a ref, return true if the refcount fell to zero (the page has no users)
385 static inline int put_page_testzero(struct page *page)
387 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
388 return atomic_dec_and_test(&page->_count);
392 * Try to grab a ref unless the page has a refcount of zero, return false if
394 * This can be called when MMU is off so it must not access
395 * any of the virtual mappings.
397 static inline int get_page_unless_zero(struct page *page)
399 return atomic_inc_not_zero(&page->_count);
402 extern int page_is_ram(unsigned long pfn);
410 int region_intersects(resource_size_t offset, size_t size, const char *type);
412 /* Support for virtually mapped pages */
413 struct page *vmalloc_to_page(const void *addr);
414 unsigned long vmalloc_to_pfn(const void *addr);
417 * Determine if an address is within the vmalloc range
419 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
420 * is no special casing required.
422 static inline int is_vmalloc_addr(const void *x)
425 unsigned long addr = (unsigned long)x;
427 return addr >= VMALLOC_START && addr < VMALLOC_END;
433 extern int is_vmalloc_or_module_addr(const void *x);
435 static inline int is_vmalloc_or_module_addr(const void *x)
441 extern void kvfree(const void *addr);
443 static inline atomic_t *compound_mapcount_ptr(struct page *page)
445 return &page[1].compound_mapcount;
448 static inline int compound_mapcount(struct page *page)
450 if (!PageCompound(page))
452 page = compound_head(page);
453 return atomic_read(compound_mapcount_ptr(page)) + 1;
457 * The atomic page->_mapcount, starts from -1: so that transitions
458 * both from it and to it can be tracked, using atomic_inc_and_test
459 * and atomic_add_negative(-1).
461 static inline void page_mapcount_reset(struct page *page)
463 atomic_set(&(page)->_mapcount, -1);
466 int __page_mapcount(struct page *page);
468 static inline int page_mapcount(struct page *page)
470 VM_BUG_ON_PAGE(PageSlab(page), page);
472 if (unlikely(PageCompound(page)))
473 return __page_mapcount(page);
474 return atomic_read(&page->_mapcount) + 1;
477 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
478 int total_mapcount(struct page *page);
480 static inline int total_mapcount(struct page *page)
482 return page_mapcount(page);
486 static inline int page_count(struct page *page)
488 return atomic_read(&compound_head(page)->_count);
491 static inline struct page *virt_to_head_page(const void *x)
493 struct page *page = virt_to_page(x);
495 return compound_head(page);
499 * Setup the page count before being freed into the page allocator for
500 * the first time (boot or memory hotplug)
502 static inline void init_page_count(struct page *page)
504 atomic_set(&page->_count, 1);
507 void __put_page(struct page *page);
509 void put_pages_list(struct list_head *pages);
511 void split_page(struct page *page, unsigned int order);
512 int split_free_page(struct page *page);
515 * Compound pages have a destructor function. Provide a
516 * prototype for that function and accessor functions.
517 * These are _only_ valid on the head of a compound page.
519 typedef void compound_page_dtor(struct page *);
521 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
522 enum compound_dtor_id {
525 #ifdef CONFIG_HUGETLB_PAGE
528 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
533 extern compound_page_dtor * const compound_page_dtors[];
535 static inline void set_compound_page_dtor(struct page *page,
536 enum compound_dtor_id compound_dtor)
538 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
539 page[1].compound_dtor = compound_dtor;
542 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
544 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
545 return compound_page_dtors[page[1].compound_dtor];
548 static inline unsigned int compound_order(struct page *page)
552 return page[1].compound_order;
555 static inline void set_compound_order(struct page *page, unsigned int order)
557 page[1].compound_order = order;
560 void free_compound_page(struct page *page);
564 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
565 * servicing faults for write access. In the normal case, do always want
566 * pte_mkwrite. But get_user_pages can cause write faults for mappings
567 * that do not have writing enabled, when used by access_process_vm.
569 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
571 if (likely(vma->vm_flags & VM_WRITE))
572 pte = pte_mkwrite(pte);
576 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
577 struct page *page, pte_t *pte, bool write, bool anon);
581 * Multiple processes may "see" the same page. E.g. for untouched
582 * mappings of /dev/null, all processes see the same page full of
583 * zeroes, and text pages of executables and shared libraries have
584 * only one copy in memory, at most, normally.
586 * For the non-reserved pages, page_count(page) denotes a reference count.
587 * page_count() == 0 means the page is free. page->lru is then used for
588 * freelist management in the buddy allocator.
589 * page_count() > 0 means the page has been allocated.
591 * Pages are allocated by the slab allocator in order to provide memory
592 * to kmalloc and kmem_cache_alloc. In this case, the management of the
593 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
594 * unless a particular usage is carefully commented. (the responsibility of
595 * freeing the kmalloc memory is the caller's, of course).
597 * A page may be used by anyone else who does a __get_free_page().
598 * In this case, page_count still tracks the references, and should only
599 * be used through the normal accessor functions. The top bits of page->flags
600 * and page->virtual store page management information, but all other fields
601 * are unused and could be used privately, carefully. The management of this
602 * page is the responsibility of the one who allocated it, and those who have
603 * subsequently been given references to it.
605 * The other pages (we may call them "pagecache pages") are completely
606 * managed by the Linux memory manager: I/O, buffers, swapping etc.
607 * The following discussion applies only to them.
609 * A pagecache page contains an opaque `private' member, which belongs to the
610 * page's address_space. Usually, this is the address of a circular list of
611 * the page's disk buffers. PG_private must be set to tell the VM to call
612 * into the filesystem to release these pages.
614 * A page may belong to an inode's memory mapping. In this case, page->mapping
615 * is the pointer to the inode, and page->index is the file offset of the page,
616 * in units of PAGE_CACHE_SIZE.
618 * If pagecache pages are not associated with an inode, they are said to be
619 * anonymous pages. These may become associated with the swapcache, and in that
620 * case PG_swapcache is set, and page->private is an offset into the swapcache.
622 * In either case (swapcache or inode backed), the pagecache itself holds one
623 * reference to the page. Setting PG_private should also increment the
624 * refcount. The each user mapping also has a reference to the page.
626 * The pagecache pages are stored in a per-mapping radix tree, which is
627 * rooted at mapping->page_tree, and indexed by offset.
628 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
629 * lists, we instead now tag pages as dirty/writeback in the radix tree.
631 * All pagecache pages may be subject to I/O:
632 * - inode pages may need to be read from disk,
633 * - inode pages which have been modified and are MAP_SHARED may need
634 * to be written back to the inode on disk,
635 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
636 * modified may need to be swapped out to swap space and (later) to be read
641 * The zone field is never updated after free_area_init_core()
642 * sets it, so none of the operations on it need to be atomic.
645 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
646 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
647 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
648 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
649 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
652 * Define the bit shifts to access each section. For non-existent
653 * sections we define the shift as 0; that plus a 0 mask ensures
654 * the compiler will optimise away reference to them.
656 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
657 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
658 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
659 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
661 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
662 #ifdef NODE_NOT_IN_PAGE_FLAGS
663 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
664 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
665 SECTIONS_PGOFF : ZONES_PGOFF)
667 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
668 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
669 NODES_PGOFF : ZONES_PGOFF)
672 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
674 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
675 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
678 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
679 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
680 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
681 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
682 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
684 static inline enum zone_type page_zonenum(const struct page *page)
686 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
689 #ifdef CONFIG_ZONE_DEVICE
690 void get_zone_device_page(struct page *page);
691 void put_zone_device_page(struct page *page);
692 static inline bool is_zone_device_page(const struct page *page)
694 return page_zonenum(page) == ZONE_DEVICE;
697 static inline void get_zone_device_page(struct page *page)
700 static inline void put_zone_device_page(struct page *page)
703 static inline bool is_zone_device_page(const struct page *page)
709 static inline void get_page(struct page *page)
711 page = compound_head(page);
713 * Getting a normal page or the head of a compound page
714 * requires to already have an elevated page->_count.
716 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
717 atomic_inc(&page->_count);
719 if (unlikely(is_zone_device_page(page)))
720 get_zone_device_page(page);
723 static inline void put_page(struct page *page)
725 page = compound_head(page);
727 if (put_page_testzero(page))
730 if (unlikely(is_zone_device_page(page)))
731 put_zone_device_page(page);
734 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
735 #define SECTION_IN_PAGE_FLAGS
739 * The identification function is mainly used by the buddy allocator for
740 * determining if two pages could be buddies. We are not really identifying
741 * the zone since we could be using the section number id if we do not have
742 * node id available in page flags.
743 * We only guarantee that it will return the same value for two combinable
746 static inline int page_zone_id(struct page *page)
748 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
751 static inline int zone_to_nid(struct zone *zone)
760 #ifdef NODE_NOT_IN_PAGE_FLAGS
761 extern int page_to_nid(const struct page *page);
763 static inline int page_to_nid(const struct page *page)
765 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
769 #ifdef CONFIG_NUMA_BALANCING
770 static inline int cpu_pid_to_cpupid(int cpu, int pid)
772 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
775 static inline int cpupid_to_pid(int cpupid)
777 return cpupid & LAST__PID_MASK;
780 static inline int cpupid_to_cpu(int cpupid)
782 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
785 static inline int cpupid_to_nid(int cpupid)
787 return cpu_to_node(cpupid_to_cpu(cpupid));
790 static inline bool cpupid_pid_unset(int cpupid)
792 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
795 static inline bool cpupid_cpu_unset(int cpupid)
797 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
800 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
802 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
805 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
806 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
807 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
809 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
812 static inline int page_cpupid_last(struct page *page)
814 return page->_last_cpupid;
816 static inline void page_cpupid_reset_last(struct page *page)
818 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
821 static inline int page_cpupid_last(struct page *page)
823 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
826 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
828 static inline void page_cpupid_reset_last(struct page *page)
830 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
832 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
833 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
835 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
836 #else /* !CONFIG_NUMA_BALANCING */
837 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
839 return page_to_nid(page); /* XXX */
842 static inline int page_cpupid_last(struct page *page)
844 return page_to_nid(page); /* XXX */
847 static inline int cpupid_to_nid(int cpupid)
852 static inline int cpupid_to_pid(int cpupid)
857 static inline int cpupid_to_cpu(int cpupid)
862 static inline int cpu_pid_to_cpupid(int nid, int pid)
867 static inline bool cpupid_pid_unset(int cpupid)
872 static inline void page_cpupid_reset_last(struct page *page)
876 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
880 #endif /* CONFIG_NUMA_BALANCING */
882 static inline struct zone *page_zone(const struct page *page)
884 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
887 #ifdef SECTION_IN_PAGE_FLAGS
888 static inline void set_page_section(struct page *page, unsigned long section)
890 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
891 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
894 static inline unsigned long page_to_section(const struct page *page)
896 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
900 static inline void set_page_zone(struct page *page, enum zone_type zone)
902 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
903 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
906 static inline void set_page_node(struct page *page, unsigned long node)
908 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
909 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
912 static inline void set_page_links(struct page *page, enum zone_type zone,
913 unsigned long node, unsigned long pfn)
915 set_page_zone(page, zone);
916 set_page_node(page, node);
917 #ifdef SECTION_IN_PAGE_FLAGS
918 set_page_section(page, pfn_to_section_nr(pfn));
923 static inline struct mem_cgroup *page_memcg(struct page *page)
925 return page->mem_cgroup;
928 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
930 page->mem_cgroup = memcg;
933 static inline struct mem_cgroup *page_memcg(struct page *page)
938 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
944 * Some inline functions in vmstat.h depend on page_zone()
946 #include <linux/vmstat.h>
948 static __always_inline void *lowmem_page_address(const struct page *page)
950 return __va(PFN_PHYS(page_to_pfn(page)));
953 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
954 #define HASHED_PAGE_VIRTUAL
957 #if defined(WANT_PAGE_VIRTUAL)
958 static inline void *page_address(const struct page *page)
960 return page->virtual;
962 static inline void set_page_address(struct page *page, void *address)
964 page->virtual = address;
966 #define page_address_init() do { } while(0)
969 #if defined(HASHED_PAGE_VIRTUAL)
970 void *page_address(const struct page *page);
971 void set_page_address(struct page *page, void *virtual);
972 void page_address_init(void);
975 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
976 #define page_address(page) lowmem_page_address(page)
977 #define set_page_address(page, address) do { } while(0)
978 #define page_address_init() do { } while(0)
981 extern void *page_rmapping(struct page *page);
982 extern struct anon_vma *page_anon_vma(struct page *page);
983 extern struct address_space *page_mapping(struct page *page);
985 extern struct address_space *__page_file_mapping(struct page *);
988 struct address_space *page_file_mapping(struct page *page)
990 if (unlikely(PageSwapCache(page)))
991 return __page_file_mapping(page);
993 return page->mapping;
997 * Return the pagecache index of the passed page. Regular pagecache pages
998 * use ->index whereas swapcache pages use ->private
1000 static inline pgoff_t page_index(struct page *page)
1002 if (unlikely(PageSwapCache(page)))
1003 return page_private(page);
1007 extern pgoff_t __page_file_index(struct page *page);
1010 * Return the file index of the page. Regular pagecache pages use ->index
1011 * whereas swapcache pages use swp_offset(->private)
1013 static inline pgoff_t page_file_index(struct page *page)
1015 if (unlikely(PageSwapCache(page)))
1016 return __page_file_index(page);
1022 * Return true if this page is mapped into pagetables.
1023 * For compound page it returns true if any subpage of compound page is mapped.
1025 static inline bool page_mapped(struct page *page)
1028 if (likely(!PageCompound(page)))
1029 return atomic_read(&page->_mapcount) >= 0;
1030 page = compound_head(page);
1031 if (atomic_read(compound_mapcount_ptr(page)) >= 0)
1033 for (i = 0; i < hpage_nr_pages(page); i++) {
1034 if (atomic_read(&page[i]._mapcount) >= 0)
1041 * Return true only if the page has been allocated with
1042 * ALLOC_NO_WATERMARKS and the low watermark was not
1043 * met implying that the system is under some pressure.
1045 static inline bool page_is_pfmemalloc(struct page *page)
1048 * Page index cannot be this large so this must be
1049 * a pfmemalloc page.
1051 return page->index == -1UL;
1055 * Only to be called by the page allocator on a freshly allocated
1058 static inline void set_page_pfmemalloc(struct page *page)
1063 static inline void clear_page_pfmemalloc(struct page *page)
1069 * Different kinds of faults, as returned by handle_mm_fault().
1070 * Used to decide whether a process gets delivered SIGBUS or
1071 * just gets major/minor fault counters bumped up.
1074 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1076 #define VM_FAULT_OOM 0x0001
1077 #define VM_FAULT_SIGBUS 0x0002
1078 #define VM_FAULT_MAJOR 0x0004
1079 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1080 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1081 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1082 #define VM_FAULT_SIGSEGV 0x0040
1084 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1085 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1086 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1087 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1089 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1091 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1092 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1095 /* Encode hstate index for a hwpoisoned large page */
1096 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1097 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1100 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1102 extern void pagefault_out_of_memory(void);
1104 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1107 * Flags passed to show_mem() and show_free_areas() to suppress output in
1110 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1112 extern void show_free_areas(unsigned int flags);
1113 extern bool skip_free_areas_node(unsigned int flags, int nid);
1115 int shmem_zero_setup(struct vm_area_struct *);
1117 bool shmem_mapping(struct address_space *mapping);
1119 static inline bool shmem_mapping(struct address_space *mapping)
1125 extern bool can_do_mlock(void);
1126 extern int user_shm_lock(size_t, struct user_struct *);
1127 extern void user_shm_unlock(size_t, struct user_struct *);
1130 * Parameter block passed down to zap_pte_range in exceptional cases.
1132 struct zap_details {
1133 struct address_space *check_mapping; /* Check page->mapping if set */
1134 pgoff_t first_index; /* Lowest page->index to unmap */
1135 pgoff_t last_index; /* Highest page->index to unmap */
1138 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1141 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1142 unsigned long size);
1143 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1144 unsigned long size, struct zap_details *);
1145 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1146 unsigned long start, unsigned long end);
1149 * mm_walk - callbacks for walk_page_range
1150 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1151 * this handler is required to be able to handle
1152 * pmd_trans_huge() pmds. They may simply choose to
1153 * split_huge_page() instead of handling it explicitly.
1154 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1155 * @pte_hole: if set, called for each hole at all levels
1156 * @hugetlb_entry: if set, called for each hugetlb entry
1157 * @test_walk: caller specific callback function to determine whether
1158 * we walk over the current vma or not. A positive returned
1159 * value means "do page table walk over the current vma,"
1160 * and a negative one means "abort current page table walk
1161 * right now." 0 means "skip the current vma."
1162 * @mm: mm_struct representing the target process of page table walk
1163 * @vma: vma currently walked (NULL if walking outside vmas)
1164 * @private: private data for callbacks' usage
1166 * (see the comment on walk_page_range() for more details)
1169 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1170 unsigned long next, struct mm_walk *walk);
1171 int (*pte_entry)(pte_t *pte, unsigned long addr,
1172 unsigned long next, struct mm_walk *walk);
1173 int (*pte_hole)(unsigned long addr, unsigned long next,
1174 struct mm_walk *walk);
1175 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1176 unsigned long addr, unsigned long next,
1177 struct mm_walk *walk);
1178 int (*test_walk)(unsigned long addr, unsigned long next,
1179 struct mm_walk *walk);
1180 struct mm_struct *mm;
1181 struct vm_area_struct *vma;
1185 int walk_page_range(unsigned long addr, unsigned long end,
1186 struct mm_walk *walk);
1187 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1188 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1189 unsigned long end, unsigned long floor, unsigned long ceiling);
1190 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1191 struct vm_area_struct *vma);
1192 void unmap_mapping_range(struct address_space *mapping,
1193 loff_t const holebegin, loff_t const holelen, int even_cows);
1194 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1195 unsigned long *pfn);
1196 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1197 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1198 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1199 void *buf, int len, int write);
1201 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1202 loff_t const holebegin, loff_t const holelen)
1204 unmap_mapping_range(mapping, holebegin, holelen, 0);
1207 extern void truncate_pagecache(struct inode *inode, loff_t new);
1208 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1209 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1210 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1211 int truncate_inode_page(struct address_space *mapping, struct page *page);
1212 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1213 int invalidate_inode_page(struct page *page);
1216 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1217 unsigned long address, unsigned int flags);
1218 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1219 unsigned long address, unsigned int fault_flags,
1222 static inline int handle_mm_fault(struct mm_struct *mm,
1223 struct vm_area_struct *vma, unsigned long address,
1226 /* should never happen if there's no MMU */
1228 return VM_FAULT_SIGBUS;
1230 static inline int fixup_user_fault(struct task_struct *tsk,
1231 struct mm_struct *mm, unsigned long address,
1232 unsigned int fault_flags, bool *unlocked)
1234 /* should never happen if there's no MMU */
1240 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1241 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1242 void *buf, int len, int write);
1244 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1245 unsigned long start, unsigned long nr_pages,
1246 unsigned int foll_flags, struct page **pages,
1247 struct vm_area_struct **vmas, int *nonblocking);
1248 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1249 unsigned long start, unsigned long nr_pages,
1250 int write, int force, struct page **pages,
1251 struct vm_area_struct **vmas);
1252 long get_user_pages6(unsigned long start, unsigned long nr_pages,
1253 int write, int force, struct page **pages,
1254 struct vm_area_struct **vmas);
1255 long get_user_pages_locked6(unsigned long start, unsigned long nr_pages,
1256 int write, int force, struct page **pages, int *locked);
1257 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1258 unsigned long start, unsigned long nr_pages,
1259 int write, int force, struct page **pages,
1260 unsigned int gup_flags);
1261 long get_user_pages_unlocked5(unsigned long start, unsigned long nr_pages,
1262 int write, int force, struct page **pages);
1263 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1264 struct page **pages);
1266 /* suppress warnings from use in EXPORT_SYMBOL() */
1267 #ifndef __DISABLE_GUP_DEPRECATED
1268 #define __gup_deprecated __deprecated
1270 #define __gup_deprecated
1273 * These macros provide backward-compatibility with the old
1274 * get_user_pages() variants which took tsk/mm. These
1275 * functions/macros provide both compile-time __deprecated so we
1276 * can catch old-style use and not break the build. The actual
1277 * functions also have WARN_ON()s to let us know at runtime if
1278 * the get_user_pages() should have been the "remote" variant.
1280 * These are hideous, but temporary.
1282 * If you run into one of these __deprecated warnings, look
1283 * at how you are calling get_user_pages(). If you are calling
1284 * it with current/current->mm as the first two arguments,
1285 * simply remove those arguments. The behavior will be the same
1286 * as it is now. If you are calling it on another task, use
1287 * get_user_pages_remote() instead.
1289 * Any questions? Ask Dave Hansen <dave@sr71.net>
1293 get_user_pages8(struct task_struct *tsk, struct mm_struct *mm,
1294 unsigned long start, unsigned long nr_pages,
1295 int write, int force, struct page **pages,
1296 struct vm_area_struct **vmas);
1297 #define GUP_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, get_user_pages, ...) \
1299 #define get_user_pages(...) GUP_MACRO(__VA_ARGS__, \
1300 get_user_pages8, x, \
1301 get_user_pages6, x, x, x, x, x)(__VA_ARGS__)
1304 long get_user_pages_locked8(struct task_struct *tsk, struct mm_struct *mm,
1305 unsigned long start, unsigned long nr_pages,
1306 int write, int force, struct page **pages,
1308 #define GUPL_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, get_user_pages_locked, ...) \
1309 get_user_pages_locked
1310 #define get_user_pages_locked(...) GUPL_MACRO(__VA_ARGS__, \
1311 get_user_pages_locked8, x, \
1312 get_user_pages_locked6, x, x, x, x)(__VA_ARGS__)
1315 long get_user_pages_unlocked7(struct task_struct *tsk, struct mm_struct *mm,
1316 unsigned long start, unsigned long nr_pages,
1317 int write, int force, struct page **pages);
1318 #define GUPU_MACRO(_1, _2, _3, _4, _5, _6, _7, get_user_pages_unlocked, ...) \
1319 get_user_pages_unlocked
1320 #define get_user_pages_unlocked(...) GUPU_MACRO(__VA_ARGS__, \
1321 get_user_pages_unlocked7, x, \
1322 get_user_pages_unlocked5, x, x, x, x)(__VA_ARGS__)
1324 /* Container for pinned pfns / pages */
1325 struct frame_vector {
1326 unsigned int nr_allocated; /* Number of frames we have space for */
1327 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1328 bool got_ref; /* Did we pin pages by getting page ref? */
1329 bool is_pfns; /* Does array contain pages or pfns? */
1330 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1331 * pfns_vector_pages() or pfns_vector_pfns()
1335 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1336 void frame_vector_destroy(struct frame_vector *vec);
1337 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1338 bool write, bool force, struct frame_vector *vec);
1339 void put_vaddr_frames(struct frame_vector *vec);
1340 int frame_vector_to_pages(struct frame_vector *vec);
1341 void frame_vector_to_pfns(struct frame_vector *vec);
1343 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1345 return vec->nr_frames;
1348 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1351 int err = frame_vector_to_pages(vec);
1354 return ERR_PTR(err);
1356 return (struct page **)(vec->ptrs);
1359 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1362 frame_vector_to_pfns(vec);
1363 return (unsigned long *)(vec->ptrs);
1367 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1368 struct page **pages);
1369 int get_kernel_page(unsigned long start, int write, struct page **pages);
1370 struct page *get_dump_page(unsigned long addr);
1372 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1373 extern void do_invalidatepage(struct page *page, unsigned int offset,
1374 unsigned int length);
1376 int __set_page_dirty_nobuffers(struct page *page);
1377 int __set_page_dirty_no_writeback(struct page *page);
1378 int redirty_page_for_writepage(struct writeback_control *wbc,
1380 void account_page_dirtied(struct page *page, struct address_space *mapping,
1381 struct mem_cgroup *memcg);
1382 void account_page_cleaned(struct page *page, struct address_space *mapping,
1383 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1384 int set_page_dirty(struct page *page);
1385 int set_page_dirty_lock(struct page *page);
1386 void cancel_dirty_page(struct page *page);
1387 int clear_page_dirty_for_io(struct page *page);
1389 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1391 /* Is the vma a continuation of the stack vma above it? */
1392 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1394 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1397 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1399 return !vma->vm_ops;
1402 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1405 return (vma->vm_flags & VM_GROWSDOWN) &&
1406 (vma->vm_start == addr) &&
1407 !vma_growsdown(vma->vm_prev, addr);
1410 /* Is the vma a continuation of the stack vma below it? */
1411 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1413 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1416 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1419 return (vma->vm_flags & VM_GROWSUP) &&
1420 (vma->vm_end == addr) &&
1421 !vma_growsup(vma->vm_next, addr);
1424 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1426 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1427 unsigned long old_addr, struct vm_area_struct *new_vma,
1428 unsigned long new_addr, unsigned long len,
1429 bool need_rmap_locks);
1430 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1431 unsigned long end, pgprot_t newprot,
1432 int dirty_accountable, int prot_numa);
1433 extern int mprotect_fixup(struct vm_area_struct *vma,
1434 struct vm_area_struct **pprev, unsigned long start,
1435 unsigned long end, unsigned long newflags);
1438 * doesn't attempt to fault and will return short.
1440 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1441 struct page **pages);
1443 * per-process(per-mm_struct) statistics.
1445 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1447 long val = atomic_long_read(&mm->rss_stat.count[member]);
1449 #ifdef SPLIT_RSS_COUNTING
1451 * counter is updated in asynchronous manner and may go to minus.
1452 * But it's never be expected number for users.
1457 return (unsigned long)val;
1460 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1462 atomic_long_add(value, &mm->rss_stat.count[member]);
1465 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1467 atomic_long_inc(&mm->rss_stat.count[member]);
1470 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1472 atomic_long_dec(&mm->rss_stat.count[member]);
1475 /* Optimized variant when page is already known not to be PageAnon */
1476 static inline int mm_counter_file(struct page *page)
1478 if (PageSwapBacked(page))
1479 return MM_SHMEMPAGES;
1480 return MM_FILEPAGES;
1483 static inline int mm_counter(struct page *page)
1486 return MM_ANONPAGES;
1487 return mm_counter_file(page);
1490 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1492 return get_mm_counter(mm, MM_FILEPAGES) +
1493 get_mm_counter(mm, MM_ANONPAGES) +
1494 get_mm_counter(mm, MM_SHMEMPAGES);
1497 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1499 return max(mm->hiwater_rss, get_mm_rss(mm));
1502 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1504 return max(mm->hiwater_vm, mm->total_vm);
1507 static inline void update_hiwater_rss(struct mm_struct *mm)
1509 unsigned long _rss = get_mm_rss(mm);
1511 if ((mm)->hiwater_rss < _rss)
1512 (mm)->hiwater_rss = _rss;
1515 static inline void update_hiwater_vm(struct mm_struct *mm)
1517 if (mm->hiwater_vm < mm->total_vm)
1518 mm->hiwater_vm = mm->total_vm;
1521 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1523 mm->hiwater_rss = get_mm_rss(mm);
1526 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1527 struct mm_struct *mm)
1529 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1531 if (*maxrss < hiwater_rss)
1532 *maxrss = hiwater_rss;
1535 #if defined(SPLIT_RSS_COUNTING)
1536 void sync_mm_rss(struct mm_struct *mm);
1538 static inline void sync_mm_rss(struct mm_struct *mm)
1543 #ifndef __HAVE_ARCH_PTE_DEVMAP
1544 static inline int pte_devmap(pte_t pte)
1550 int vma_wants_writenotify(struct vm_area_struct *vma);
1552 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1554 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1558 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1562 #ifdef __PAGETABLE_PUD_FOLDED
1563 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1564 unsigned long address)
1569 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1572 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1573 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1574 unsigned long address)
1579 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1581 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1586 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1587 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1590 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1592 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1594 atomic_long_set(&mm->nr_pmds, 0);
1597 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1599 return atomic_long_read(&mm->nr_pmds);
1602 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1604 atomic_long_inc(&mm->nr_pmds);
1607 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1609 atomic_long_dec(&mm->nr_pmds);
1613 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1614 pmd_t *pmd, unsigned long address);
1615 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1618 * The following ifdef needed to get the 4level-fixup.h header to work.
1619 * Remove it when 4level-fixup.h has been removed.
1621 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1622 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1624 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1625 NULL: pud_offset(pgd, address);
1628 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1630 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1631 NULL: pmd_offset(pud, address);
1633 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1635 #if USE_SPLIT_PTE_PTLOCKS
1636 #if ALLOC_SPLIT_PTLOCKS
1637 void __init ptlock_cache_init(void);
1638 extern bool ptlock_alloc(struct page *page);
1639 extern void ptlock_free(struct page *page);
1641 static inline spinlock_t *ptlock_ptr(struct page *page)
1645 #else /* ALLOC_SPLIT_PTLOCKS */
1646 static inline void ptlock_cache_init(void)
1650 static inline bool ptlock_alloc(struct page *page)
1655 static inline void ptlock_free(struct page *page)
1659 static inline spinlock_t *ptlock_ptr(struct page *page)
1663 #endif /* ALLOC_SPLIT_PTLOCKS */
1665 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1667 return ptlock_ptr(pmd_page(*pmd));
1670 static inline bool ptlock_init(struct page *page)
1673 * prep_new_page() initialize page->private (and therefore page->ptl)
1674 * with 0. Make sure nobody took it in use in between.
1676 * It can happen if arch try to use slab for page table allocation:
1677 * slab code uses page->slab_cache, which share storage with page->ptl.
1679 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1680 if (!ptlock_alloc(page))
1682 spin_lock_init(ptlock_ptr(page));
1686 /* Reset page->mapping so free_pages_check won't complain. */
1687 static inline void pte_lock_deinit(struct page *page)
1689 page->mapping = NULL;
1693 #else /* !USE_SPLIT_PTE_PTLOCKS */
1695 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1697 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1699 return &mm->page_table_lock;
1701 static inline void ptlock_cache_init(void) {}
1702 static inline bool ptlock_init(struct page *page) { return true; }
1703 static inline void pte_lock_deinit(struct page *page) {}
1704 #endif /* USE_SPLIT_PTE_PTLOCKS */
1706 static inline void pgtable_init(void)
1708 ptlock_cache_init();
1709 pgtable_cache_init();
1712 static inline bool pgtable_page_ctor(struct page *page)
1714 if (!ptlock_init(page))
1716 inc_zone_page_state(page, NR_PAGETABLE);
1720 static inline void pgtable_page_dtor(struct page *page)
1722 pte_lock_deinit(page);
1723 dec_zone_page_state(page, NR_PAGETABLE);
1726 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1728 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1729 pte_t *__pte = pte_offset_map(pmd, address); \
1735 #define pte_unmap_unlock(pte, ptl) do { \
1740 #define pte_alloc_map(mm, vma, pmd, address) \
1741 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1743 NULL: pte_offset_map(pmd, address))
1745 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1746 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1748 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1750 #define pte_alloc_kernel(pmd, address) \
1751 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1752 NULL: pte_offset_kernel(pmd, address))
1754 #if USE_SPLIT_PMD_PTLOCKS
1756 static struct page *pmd_to_page(pmd_t *pmd)
1758 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1759 return virt_to_page((void *)((unsigned long) pmd & mask));
1762 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1764 return ptlock_ptr(pmd_to_page(pmd));
1767 static inline bool pgtable_pmd_page_ctor(struct page *page)
1769 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1770 page->pmd_huge_pte = NULL;
1772 return ptlock_init(page);
1775 static inline void pgtable_pmd_page_dtor(struct page *page)
1777 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1778 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1783 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1787 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1789 return &mm->page_table_lock;
1792 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1793 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1795 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1799 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1801 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1806 extern void free_area_init(unsigned long * zones_size);
1807 extern void free_area_init_node(int nid, unsigned long * zones_size,
1808 unsigned long zone_start_pfn, unsigned long *zholes_size);
1809 extern void free_initmem(void);
1812 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1813 * into the buddy system. The freed pages will be poisoned with pattern
1814 * "poison" if it's within range [0, UCHAR_MAX].
1815 * Return pages freed into the buddy system.
1817 extern unsigned long free_reserved_area(void *start, void *end,
1818 int poison, char *s);
1820 #ifdef CONFIG_HIGHMEM
1822 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1823 * and totalram_pages.
1825 extern void free_highmem_page(struct page *page);
1828 extern void adjust_managed_page_count(struct page *page, long count);
1829 extern void mem_init_print_info(const char *str);
1831 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1833 /* Free the reserved page into the buddy system, so it gets managed. */
1834 static inline void __free_reserved_page(struct page *page)
1836 ClearPageReserved(page);
1837 init_page_count(page);
1841 static inline void free_reserved_page(struct page *page)
1843 __free_reserved_page(page);
1844 adjust_managed_page_count(page, 1);
1847 static inline void mark_page_reserved(struct page *page)
1849 SetPageReserved(page);
1850 adjust_managed_page_count(page, -1);
1854 * Default method to free all the __init memory into the buddy system.
1855 * The freed pages will be poisoned with pattern "poison" if it's within
1856 * range [0, UCHAR_MAX].
1857 * Return pages freed into the buddy system.
1859 static inline unsigned long free_initmem_default(int poison)
1861 extern char __init_begin[], __init_end[];
1863 return free_reserved_area(&__init_begin, &__init_end,
1864 poison, "unused kernel");
1867 static inline unsigned long get_num_physpages(void)
1870 unsigned long phys_pages = 0;
1872 for_each_online_node(nid)
1873 phys_pages += node_present_pages(nid);
1878 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1880 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1881 * zones, allocate the backing mem_map and account for memory holes in a more
1882 * architecture independent manner. This is a substitute for creating the
1883 * zone_sizes[] and zholes_size[] arrays and passing them to
1884 * free_area_init_node()
1886 * An architecture is expected to register range of page frames backed by
1887 * physical memory with memblock_add[_node]() before calling
1888 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1889 * usage, an architecture is expected to do something like
1891 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1893 * for_each_valid_physical_page_range()
1894 * memblock_add_node(base, size, nid)
1895 * free_area_init_nodes(max_zone_pfns);
1897 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1898 * registered physical page range. Similarly
1899 * sparse_memory_present_with_active_regions() calls memory_present() for
1900 * each range when SPARSEMEM is enabled.
1902 * See mm/page_alloc.c for more information on each function exposed by
1903 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1905 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1906 unsigned long node_map_pfn_alignment(void);
1907 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1908 unsigned long end_pfn);
1909 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1910 unsigned long end_pfn);
1911 extern void get_pfn_range_for_nid(unsigned int nid,
1912 unsigned long *start_pfn, unsigned long *end_pfn);
1913 extern unsigned long find_min_pfn_with_active_regions(void);
1914 extern void free_bootmem_with_active_regions(int nid,
1915 unsigned long max_low_pfn);
1916 extern void sparse_memory_present_with_active_regions(int nid);
1918 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1920 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1921 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1922 static inline int __early_pfn_to_nid(unsigned long pfn,
1923 struct mminit_pfnnid_cache *state)
1928 /* please see mm/page_alloc.c */
1929 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1930 /* there is a per-arch backend function. */
1931 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1932 struct mminit_pfnnid_cache *state);
1935 extern void set_dma_reserve(unsigned long new_dma_reserve);
1936 extern void memmap_init_zone(unsigned long, int, unsigned long,
1937 unsigned long, enum memmap_context);
1938 extern void setup_per_zone_wmarks(void);
1939 extern int __meminit init_per_zone_wmark_min(void);
1940 extern void mem_init(void);
1941 extern void __init mmap_init(void);
1942 extern void show_mem(unsigned int flags);
1943 extern void si_meminfo(struct sysinfo * val);
1944 extern void si_meminfo_node(struct sysinfo *val, int nid);
1946 extern __printf(3, 4)
1947 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1948 const char *fmt, ...);
1950 extern void setup_per_cpu_pageset(void);
1952 extern void zone_pcp_update(struct zone *zone);
1953 extern void zone_pcp_reset(struct zone *zone);
1956 extern int min_free_kbytes;
1959 extern atomic_long_t mmap_pages_allocated;
1960 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1962 /* interval_tree.c */
1963 void vma_interval_tree_insert(struct vm_area_struct *node,
1964 struct rb_root *root);
1965 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1966 struct vm_area_struct *prev,
1967 struct rb_root *root);
1968 void vma_interval_tree_remove(struct vm_area_struct *node,
1969 struct rb_root *root);
1970 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1971 unsigned long start, unsigned long last);
1972 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1973 unsigned long start, unsigned long last);
1975 #define vma_interval_tree_foreach(vma, root, start, last) \
1976 for (vma = vma_interval_tree_iter_first(root, start, last); \
1977 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1979 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1980 struct rb_root *root);
1981 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1982 struct rb_root *root);
1983 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1984 struct rb_root *root, unsigned long start, unsigned long last);
1985 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1986 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1987 #ifdef CONFIG_DEBUG_VM_RB
1988 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1991 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1992 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1993 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1996 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1997 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1998 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1999 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2000 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2001 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2002 struct mempolicy *, struct vm_userfaultfd_ctx);
2003 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2004 extern int split_vma(struct mm_struct *,
2005 struct vm_area_struct *, unsigned long addr, int new_below);
2006 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2007 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2008 struct rb_node **, struct rb_node *);
2009 extern void unlink_file_vma(struct vm_area_struct *);
2010 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2011 unsigned long addr, unsigned long len, pgoff_t pgoff,
2012 bool *need_rmap_locks);
2013 extern void exit_mmap(struct mm_struct *);
2015 static inline int check_data_rlimit(unsigned long rlim,
2017 unsigned long start,
2018 unsigned long end_data,
2019 unsigned long start_data)
2021 if (rlim < RLIM_INFINITY) {
2022 if (((new - start) + (end_data - start_data)) > rlim)
2029 extern int mm_take_all_locks(struct mm_struct *mm);
2030 extern void mm_drop_all_locks(struct mm_struct *mm);
2032 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2033 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2035 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2036 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2038 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2039 unsigned long addr, unsigned long len,
2040 unsigned long flags,
2041 const struct vm_special_mapping *spec);
2042 /* This is an obsolete alternative to _install_special_mapping. */
2043 extern int install_special_mapping(struct mm_struct *mm,
2044 unsigned long addr, unsigned long len,
2045 unsigned long flags, struct page **pages);
2047 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2049 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2050 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2051 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2052 unsigned long len, unsigned long prot, unsigned long flags,
2053 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2054 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2056 static inline unsigned long
2057 do_mmap_pgoff(struct file *file, unsigned long addr,
2058 unsigned long len, unsigned long prot, unsigned long flags,
2059 unsigned long pgoff, unsigned long *populate)
2061 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2065 extern int __mm_populate(unsigned long addr, unsigned long len,
2067 static inline void mm_populate(unsigned long addr, unsigned long len)
2070 (void) __mm_populate(addr, len, 1);
2073 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2076 /* These take the mm semaphore themselves */
2077 extern unsigned long vm_brk(unsigned long, unsigned long);
2078 extern int vm_munmap(unsigned long, size_t);
2079 extern unsigned long vm_mmap(struct file *, unsigned long,
2080 unsigned long, unsigned long,
2081 unsigned long, unsigned long);
2083 struct vm_unmapped_area_info {
2084 #define VM_UNMAPPED_AREA_TOPDOWN 1
2085 unsigned long flags;
2086 unsigned long length;
2087 unsigned long low_limit;
2088 unsigned long high_limit;
2089 unsigned long align_mask;
2090 unsigned long align_offset;
2093 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2094 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2097 * Search for an unmapped address range.
2099 * We are looking for a range that:
2100 * - does not intersect with any VMA;
2101 * - is contained within the [low_limit, high_limit) interval;
2102 * - is at least the desired size.
2103 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2105 static inline unsigned long
2106 vm_unmapped_area(struct vm_unmapped_area_info *info)
2108 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2109 return unmapped_area_topdown(info);
2111 return unmapped_area(info);
2115 extern void truncate_inode_pages(struct address_space *, loff_t);
2116 extern void truncate_inode_pages_range(struct address_space *,
2117 loff_t lstart, loff_t lend);
2118 extern void truncate_inode_pages_final(struct address_space *);
2120 /* generic vm_area_ops exported for stackable file systems */
2121 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2122 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2123 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2125 /* mm/page-writeback.c */
2126 int write_one_page(struct page *page, int wait);
2127 void task_dirty_inc(struct task_struct *tsk);
2130 #define VM_MAX_READAHEAD 128 /* kbytes */
2131 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2133 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2134 pgoff_t offset, unsigned long nr_to_read);
2136 void page_cache_sync_readahead(struct address_space *mapping,
2137 struct file_ra_state *ra,
2140 unsigned long size);
2142 void page_cache_async_readahead(struct address_space *mapping,
2143 struct file_ra_state *ra,
2147 unsigned long size);
2149 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2150 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2152 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2153 extern int expand_downwards(struct vm_area_struct *vma,
2154 unsigned long address);
2156 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2158 #define expand_upwards(vma, address) (0)
2161 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2162 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2163 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2164 struct vm_area_struct **pprev);
2166 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2167 NULL if none. Assume start_addr < end_addr. */
2168 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2170 struct vm_area_struct * vma = find_vma(mm,start_addr);
2172 if (vma && end_addr <= vma->vm_start)
2177 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2179 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2182 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2183 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2184 unsigned long vm_start, unsigned long vm_end)
2186 struct vm_area_struct *vma = find_vma(mm, vm_start);
2188 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2195 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2196 void vma_set_page_prot(struct vm_area_struct *vma);
2198 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2202 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2204 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2208 #ifdef CONFIG_NUMA_BALANCING
2209 unsigned long change_prot_numa(struct vm_area_struct *vma,
2210 unsigned long start, unsigned long end);
2213 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2214 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2215 unsigned long pfn, unsigned long size, pgprot_t);
2216 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2217 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2219 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2220 unsigned long pfn, pgprot_t pgprot);
2221 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2223 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2226 struct page *follow_page_mask(struct vm_area_struct *vma,
2227 unsigned long address, unsigned int foll_flags,
2228 unsigned int *page_mask);
2230 static inline struct page *follow_page(struct vm_area_struct *vma,
2231 unsigned long address, unsigned int foll_flags)
2233 unsigned int unused_page_mask;
2234 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2237 #define FOLL_WRITE 0x01 /* check pte is writable */
2238 #define FOLL_TOUCH 0x02 /* mark page accessed */
2239 #define FOLL_GET 0x04 /* do get_page on page */
2240 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2241 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2242 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2243 * and return without waiting upon it */
2244 #define FOLL_POPULATE 0x40 /* fault in page */
2245 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2246 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2247 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2248 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2249 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2250 #define FOLL_MLOCK 0x1000 /* lock present pages */
2251 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2253 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2255 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2256 unsigned long size, pte_fn_t fn, void *data);
2259 #ifdef CONFIG_DEBUG_PAGEALLOC
2260 extern bool _debug_pagealloc_enabled;
2261 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2263 static inline bool debug_pagealloc_enabled(void)
2265 return _debug_pagealloc_enabled;
2269 kernel_map_pages(struct page *page, int numpages, int enable)
2271 if (!debug_pagealloc_enabled())
2274 __kernel_map_pages(page, numpages, enable);
2276 #ifdef CONFIG_HIBERNATION
2277 extern bool kernel_page_present(struct page *page);
2278 #endif /* CONFIG_HIBERNATION */
2281 kernel_map_pages(struct page *page, int numpages, int enable) {}
2282 #ifdef CONFIG_HIBERNATION
2283 static inline bool kernel_page_present(struct page *page) { return true; }
2284 #endif /* CONFIG_HIBERNATION */
2287 #ifdef __HAVE_ARCH_GATE_AREA
2288 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2289 extern int in_gate_area_no_mm(unsigned long addr);
2290 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2292 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2296 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2297 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2301 #endif /* __HAVE_ARCH_GATE_AREA */
2303 #ifdef CONFIG_SYSCTL
2304 extern int sysctl_drop_caches;
2305 int drop_caches_sysctl_handler(struct ctl_table *, int,
2306 void __user *, size_t *, loff_t *);
2309 void drop_slab(void);
2310 void drop_slab_node(int nid);
2313 #define randomize_va_space 0
2315 extern int randomize_va_space;
2318 const char * arch_vma_name(struct vm_area_struct *vma);
2319 void print_vma_addr(char *prefix, unsigned long rip);
2321 void sparse_mem_maps_populate_node(struct page **map_map,
2322 unsigned long pnum_begin,
2323 unsigned long pnum_end,
2324 unsigned long map_count,
2327 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2328 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2329 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2330 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2331 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2332 void *vmemmap_alloc_block(unsigned long size, int node);
2334 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2335 struct vmem_altmap *altmap);
2336 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2338 return __vmemmap_alloc_block_buf(size, node, NULL);
2341 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2342 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2344 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2345 void vmemmap_populate_print_last(void);
2346 #ifdef CONFIG_MEMORY_HOTPLUG
2347 void vmemmap_free(unsigned long start, unsigned long end);
2349 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2350 unsigned long size);
2353 MF_COUNT_INCREASED = 1 << 0,
2354 MF_ACTION_REQUIRED = 1 << 1,
2355 MF_MUST_KILL = 1 << 2,
2356 MF_SOFT_OFFLINE = 1 << 3,
2358 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2359 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2360 extern int unpoison_memory(unsigned long pfn);
2361 extern int get_hwpoison_page(struct page *page);
2362 #define put_hwpoison_page(page) put_page(page)
2363 extern int sysctl_memory_failure_early_kill;
2364 extern int sysctl_memory_failure_recovery;
2365 extern void shake_page(struct page *p, int access);
2366 extern atomic_long_t num_poisoned_pages;
2367 extern int soft_offline_page(struct page *page, int flags);
2371 * Error handlers for various types of pages.
2374 MF_IGNORED, /* Error: cannot be handled */
2375 MF_FAILED, /* Error: handling failed */
2376 MF_DELAYED, /* Will be handled later */
2377 MF_RECOVERED, /* Successfully recovered */
2380 enum mf_action_page_type {
2382 MF_MSG_KERNEL_HIGH_ORDER,
2384 MF_MSG_DIFFERENT_COMPOUND,
2385 MF_MSG_POISONED_HUGE,
2388 MF_MSG_UNMAP_FAILED,
2389 MF_MSG_DIRTY_SWAPCACHE,
2390 MF_MSG_CLEAN_SWAPCACHE,
2391 MF_MSG_DIRTY_MLOCKED_LRU,
2392 MF_MSG_CLEAN_MLOCKED_LRU,
2393 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2394 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2397 MF_MSG_TRUNCATED_LRU,
2403 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2404 extern void clear_huge_page(struct page *page,
2406 unsigned int pages_per_huge_page);
2407 extern void copy_user_huge_page(struct page *dst, struct page *src,
2408 unsigned long addr, struct vm_area_struct *vma,
2409 unsigned int pages_per_huge_page);
2410 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2412 extern struct page_ext_operations debug_guardpage_ops;
2413 extern struct page_ext_operations page_poisoning_ops;
2415 #ifdef CONFIG_DEBUG_PAGEALLOC
2416 extern unsigned int _debug_guardpage_minorder;
2417 extern bool _debug_guardpage_enabled;
2419 static inline unsigned int debug_guardpage_minorder(void)
2421 return _debug_guardpage_minorder;
2424 static inline bool debug_guardpage_enabled(void)
2426 return _debug_guardpage_enabled;
2429 static inline bool page_is_guard(struct page *page)
2431 struct page_ext *page_ext;
2433 if (!debug_guardpage_enabled())
2436 page_ext = lookup_page_ext(page);
2437 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2440 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2441 static inline bool debug_guardpage_enabled(void) { return false; }
2442 static inline bool page_is_guard(struct page *page) { return false; }
2443 #endif /* CONFIG_DEBUG_PAGEALLOC */
2445 #if MAX_NUMNODES > 1
2446 void __init setup_nr_node_ids(void);
2448 static inline void setup_nr_node_ids(void) {}
2451 #endif /* __KERNEL__ */
2452 #endif /* _LINUX_MM_H */