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/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 #include <linux/resource.h>
22 #include <linux/page_ext.h>
23 #include <linux/err.h>
27 struct anon_vma_chain;
30 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 #define lm_alias(x) __va(__pa_symbol(x))
79 * To prevent common memory management code establishing
80 * a zero page mapping on a read fault.
81 * This macro should be defined within <asm/pgtable.h>.
82 * s390 does this to prevent multiplexing of hardware bits
83 * related to the physical page in case of virtualization.
85 #ifndef mm_forbids_zeropage
86 #define mm_forbids_zeropage(X) (0)
89 extern unsigned long sysctl_user_reserve_kbytes;
90 extern unsigned long sysctl_admin_reserve_kbytes;
92 extern int sysctl_overcommit_memory;
93 extern int sysctl_overcommit_ratio;
94 extern unsigned long sysctl_overcommit_kbytes;
96 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
98 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
101 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
103 /* to align the pointer to the (next) page boundary */
104 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
106 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
107 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
110 * Linux kernel virtual memory manager primitives.
111 * The idea being to have a "virtual" mm in the same way
112 * we have a virtual fs - giving a cleaner interface to the
113 * mm details, and allowing different kinds of memory mappings
114 * (from shared memory to executable loading to arbitrary
118 extern struct kmem_cache *vm_area_cachep;
121 extern struct rb_root nommu_region_tree;
122 extern struct rw_semaphore nommu_region_sem;
124 extern unsigned int kobjsize(const void *objp);
128 * vm_flags in vm_area_struct, see mm_types.h.
130 #define VM_NONE 0x00000000
132 #define VM_READ 0x00000001 /* currently active flags */
133 #define VM_WRITE 0x00000002
134 #define VM_EXEC 0x00000004
135 #define VM_SHARED 0x00000008
137 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
138 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
139 #define VM_MAYWRITE 0x00000020
140 #define VM_MAYEXEC 0x00000040
141 #define VM_MAYSHARE 0x00000080
143 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
144 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
145 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
146 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
147 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
149 #define VM_LOCKED 0x00002000
150 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
152 /* Used by sys_madvise() */
153 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
154 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
156 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
157 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
158 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
159 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
160 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
161 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
162 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
163 #define VM_ARCH_2 0x02000000
164 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
166 #ifdef CONFIG_MEM_SOFT_DIRTY
167 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
169 # define VM_SOFTDIRTY 0
172 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
173 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
174 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
175 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
177 #if defined(CONFIG_X86)
178 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
179 #elif defined(CONFIG_PPC)
180 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
181 #elif defined(CONFIG_PARISC)
182 # define VM_GROWSUP VM_ARCH_1
183 #elif defined(CONFIG_METAG)
184 # define VM_GROWSUP VM_ARCH_1
185 #elif defined(CONFIG_IA64)
186 # define VM_GROWSUP VM_ARCH_1
187 #elif !defined(CONFIG_MMU)
188 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
191 #if defined(CONFIG_X86)
192 /* MPX specific bounds table or bounds directory */
193 # define VM_MPX VM_ARCH_2
197 # define VM_GROWSUP VM_NONE
200 /* Bits set in the VMA until the stack is in its final location */
201 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
203 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
204 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
207 #ifdef CONFIG_STACK_GROWSUP
208 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
210 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
214 * Special vmas that are non-mergable, non-mlock()able.
215 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
217 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
219 /* This mask defines which mm->def_flags a process can inherit its parent */
220 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
222 /* This mask is used to clear all the VMA flags used by mlock */
223 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
226 * mapping from the currently active vm_flags protection bits (the
227 * low four bits) to a page protection mask..
229 extern pgprot_t protection_map[16];
231 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
232 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
233 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
234 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
235 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
236 #define FAULT_FLAG_TRIED 0x20 /* Second try */
237 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
238 #define FAULT_FLAG_SPECULATIVE 0x200 /* Speculative fault, not holding mmap_sem */
241 * vm_fault is filled by the the pagefault handler and passed to the vma's
242 * ->fault function. The vma's ->fault is responsible for returning a bitmask
243 * of VM_FAULT_xxx flags that give details about how the fault was handled.
245 * MM layer fills up gfp_mask for page allocations but fault handler might
246 * alter it if its implementation requires a different allocation context.
248 * pgoff should be used in favour of virtual_address, if possible.
251 struct vm_area_struct *vma; /* Target VMA */
252 pmd_t *pmd; /* Pointer to pmd entry matching
254 pud_t *pud; /* Pointer to pud entry matching
257 unsigned long address; /* Faulting virtual address */
258 spinlock_t *ptl; /* Page table lock.
259 * Protects pte page table if 'pte'
260 * is not NULL, otherwise pmd.
262 pte_t orig_pte; /* Value of PTE at the time of fault */
264 * These entries are required when handling speculative page fault.
265 * This way the page handling is done using consistent field values.
267 unsigned long vma_flags;
268 pgprot_t vma_page_prot;
269 #ifdef CONFIG_SPECULATIVE_PAGE_FAULT
270 unsigned int sequence;
271 pmd_t orig_pmd; /* value of PMD at the time of fault */
274 unsigned int flags; /* FAULT_FLAG_xxx flags */
275 gfp_t gfp_mask; /* gfp mask to be used for allocations */
276 pgoff_t pgoff; /* Logical page offset based on vma */
277 void __user *virtual_address; /* Faulting virtual address */
279 struct page *cow_page; /* Handler may choose to COW */
280 struct page *page; /* ->fault handlers should return a
281 * page here, unless VM_FAULT_NOPAGE
282 * is set (which is also implied by
285 /* for ->map_pages() only */
286 pgoff_t max_pgoff; /* map pages for offset from pgoff till
287 * max_pgoff inclusive */
288 pte_t *pte; /* pte entry associated with ->pgoff */
292 * These are the virtual MM functions - opening of an area, closing and
293 * unmapping it (needed to keep files on disk up-to-date etc), pointer
294 * to the functions called when a no-page or a wp-page exception occurs.
296 struct vm_operations_struct {
297 void (*open)(struct vm_area_struct * area);
298 void (*close)(struct vm_area_struct * area);
299 int (*mremap)(struct vm_area_struct * area);
300 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
301 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
302 pmd_t *, unsigned int flags);
303 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
305 /* notification that a previously read-only page is about to become
306 * writable, if an error is returned it will cause a SIGBUS */
307 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
309 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
310 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
312 /* called by access_process_vm when get_user_pages() fails, typically
313 * for use by special VMAs that can switch between memory and hardware
315 int (*access)(struct vm_area_struct *vma, unsigned long addr,
316 void *buf, int len, int write);
318 /* Called by the /proc/PID/maps code to ask the vma whether it
319 * has a special name. Returning non-NULL will also cause this
320 * vma to be dumped unconditionally. */
321 const char *(*name)(struct vm_area_struct *vma);
325 * set_policy() op must add a reference to any non-NULL @new mempolicy
326 * to hold the policy upon return. Caller should pass NULL @new to
327 * remove a policy and fall back to surrounding context--i.e. do not
328 * install a MPOL_DEFAULT policy, nor the task or system default
331 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
334 * get_policy() op must add reference [mpol_get()] to any policy at
335 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
336 * in mm/mempolicy.c will do this automatically.
337 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
338 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
339 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
340 * must return NULL--i.e., do not "fallback" to task or system default
343 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
347 * Called by vm_normal_page() for special PTEs to find the
348 * page for @addr. This is useful if the default behavior
349 * (using pte_page()) would not find the correct page.
351 struct page *(*find_special_page)(struct vm_area_struct *vma,
358 #define page_private(page) ((page)->private)
359 #define set_page_private(page, v) ((page)->private = (v))
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 bool 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 void compound_lock(struct page *page)
444 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
445 VM_BUG_ON_PAGE(PageSlab(page), page);
446 bit_spin_lock(PG_compound_lock, &page->flags);
450 static inline void compound_unlock(struct page *page)
452 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
453 VM_BUG_ON_PAGE(PageSlab(page), page);
454 bit_spin_unlock(PG_compound_lock, &page->flags);
458 static inline unsigned long compound_lock_irqsave(struct page *page)
460 unsigned long uninitialized_var(flags);
461 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
462 local_irq_save(flags);
468 static inline void compound_unlock_irqrestore(struct page *page,
471 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
472 compound_unlock(page);
473 local_irq_restore(flags);
478 * The atomic page->_mapcount, starts from -1: so that transitions
479 * both from it and to it can be tracked, using atomic_inc_and_test
480 * and atomic_add_negative(-1).
482 static inline void page_mapcount_reset(struct page *page)
484 atomic_set(&(page)->_mapcount, -1);
487 static inline int page_mapcount(struct page *page)
489 VM_BUG_ON_PAGE(PageSlab(page), page);
490 return atomic_read(&page->_mapcount) + 1;
493 static inline int page_count(struct page *page)
495 return atomic_read(&compound_head(page)->_count);
498 static inline bool __compound_tail_refcounted(struct page *page)
500 return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
504 * This takes a head page as parameter and tells if the
505 * tail page reference counting can be skipped.
507 * For this to be safe, PageSlab and PageHeadHuge must remain true on
508 * any given page where they return true here, until all tail pins
509 * have been released.
511 static inline bool compound_tail_refcounted(struct page *page)
513 VM_BUG_ON_PAGE(!PageHead(page), page);
514 return __compound_tail_refcounted(page);
517 static inline void get_huge_page_tail(struct page *page)
520 * __split_huge_page_refcount() cannot run from under us.
522 VM_BUG_ON_PAGE(!PageTail(page), page);
523 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
524 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
525 if (compound_tail_refcounted(compound_head(page)))
526 atomic_inc(&page->_mapcount);
529 extern bool __get_page_tail(struct page *page);
531 static inline void get_page(struct page *page)
533 if (unlikely(PageTail(page)))
534 if (likely(__get_page_tail(page)))
537 * Getting a normal page or the head of a compound page
538 * requires to already have an elevated page->_count.
540 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
541 atomic_inc(&page->_count);
544 static inline struct page *virt_to_head_page(const void *x)
546 struct page *page = virt_to_page(x);
548 return compound_head(page);
552 * Setup the page count before being freed into the page allocator for
553 * the first time (boot or memory hotplug)
555 static inline void init_page_count(struct page *page)
557 atomic_set(&page->_count, 1);
560 void put_page(struct page *page);
561 void put_pages_list(struct list_head *pages);
563 void split_page(struct page *page, unsigned int order);
566 * Compound pages have a destructor function. Provide a
567 * prototype for that function and accessor functions.
568 * These are _only_ valid on the head of a compound page.
570 typedef void compound_page_dtor(struct page *);
572 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
573 enum compound_dtor_id {
576 #ifdef CONFIG_HUGETLB_PAGE
581 extern compound_page_dtor * const compound_page_dtors[];
583 static inline void set_compound_page_dtor(struct page *page,
584 enum compound_dtor_id compound_dtor)
586 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
587 page[1].compound_dtor = compound_dtor;
590 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
592 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
593 return compound_page_dtors[page[1].compound_dtor];
596 static inline unsigned int compound_order(struct page *page)
600 return page[1].compound_order;
603 static inline void set_compound_order(struct page *page, unsigned int order)
605 page[1].compound_order = order;
610 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
611 * servicing faults for write access. In the normal case, do always want
612 * pte_mkwrite. But get_user_pages can cause write faults for mappings
613 * that do not have writing enabled, when used by access_process_vm.
615 static inline pte_t maybe_mkwrite(pte_t pte, unsigned long vma_flags)
617 if (likely(vma_flags & VM_WRITE))
618 pte = pte_mkwrite(pte);
622 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
623 struct page *page, pte_t *pte, bool write, bool anon, struct vm_fault *vmf2);
627 * Multiple processes may "see" the same page. E.g. for untouched
628 * mappings of /dev/null, all processes see the same page full of
629 * zeroes, and text pages of executables and shared libraries have
630 * only one copy in memory, at most, normally.
632 * For the non-reserved pages, page_count(page) denotes a reference count.
633 * page_count() == 0 means the page is free. page->lru is then used for
634 * freelist management in the buddy allocator.
635 * page_count() > 0 means the page has been allocated.
637 * Pages are allocated by the slab allocator in order to provide memory
638 * to kmalloc and kmem_cache_alloc. In this case, the management of the
639 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
640 * unless a particular usage is carefully commented. (the responsibility of
641 * freeing the kmalloc memory is the caller's, of course).
643 * A page may be used by anyone else who does a __get_free_page().
644 * In this case, page_count still tracks the references, and should only
645 * be used through the normal accessor functions. The top bits of page->flags
646 * and page->virtual store page management information, but all other fields
647 * are unused and could be used privately, carefully. The management of this
648 * page is the responsibility of the one who allocated it, and those who have
649 * subsequently been given references to it.
651 * The other pages (we may call them "pagecache pages") are completely
652 * managed by the Linux memory manager: I/O, buffers, swapping etc.
653 * The following discussion applies only to them.
655 * A pagecache page contains an opaque `private' member, which belongs to the
656 * page's address_space. Usually, this is the address of a circular list of
657 * the page's disk buffers. PG_private must be set to tell the VM to call
658 * into the filesystem to release these pages.
660 * A page may belong to an inode's memory mapping. In this case, page->mapping
661 * is the pointer to the inode, and page->index is the file offset of the page,
662 * in units of PAGE_CACHE_SIZE.
664 * If pagecache pages are not associated with an inode, they are said to be
665 * anonymous pages. These may become associated with the swapcache, and in that
666 * case PG_swapcache is set, and page->private is an offset into the swapcache.
668 * In either case (swapcache or inode backed), the pagecache itself holds one
669 * reference to the page. Setting PG_private should also increment the
670 * refcount. The each user mapping also has a reference to the page.
672 * The pagecache pages are stored in a per-mapping radix tree, which is
673 * rooted at mapping->page_tree, and indexed by offset.
674 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
675 * lists, we instead now tag pages as dirty/writeback in the radix tree.
677 * All pagecache pages may be subject to I/O:
678 * - inode pages may need to be read from disk,
679 * - inode pages which have been modified and are MAP_SHARED may need
680 * to be written back to the inode on disk,
681 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
682 * modified may need to be swapped out to swap space and (later) to be read
687 * The zone field is never updated after free_area_init_core()
688 * sets it, so none of the operations on it need to be atomic.
691 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
692 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
693 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
694 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
695 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
698 * Define the bit shifts to access each section. For non-existent
699 * sections we define the shift as 0; that plus a 0 mask ensures
700 * the compiler will optimise away reference to them.
702 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
703 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
704 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
705 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
707 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
708 #ifdef NODE_NOT_IN_PAGE_FLAGS
709 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
710 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
711 SECTIONS_PGOFF : ZONES_PGOFF)
713 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
714 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
715 NODES_PGOFF : ZONES_PGOFF)
718 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
720 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
721 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
724 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
725 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
726 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
727 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
728 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
730 static inline enum zone_type page_zonenum(const struct page *page)
732 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
735 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
736 #define SECTION_IN_PAGE_FLAGS
740 * The identification function is mainly used by the buddy allocator for
741 * determining if two pages could be buddies. We are not really identifying
742 * the zone since we could be using the section number id if we do not have
743 * node id available in page flags.
744 * We only guarantee that it will return the same value for two combinable
747 static inline int page_zone_id(struct page *page)
749 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
752 static inline int zone_to_nid(struct zone *zone)
761 #ifdef NODE_NOT_IN_PAGE_FLAGS
762 extern int page_to_nid(const struct page *page);
764 static inline int page_to_nid(const struct page *page)
766 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
770 #ifdef CONFIG_NUMA_BALANCING
771 static inline int cpu_pid_to_cpupid(int cpu, int pid)
773 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
776 static inline int cpupid_to_pid(int cpupid)
778 return cpupid & LAST__PID_MASK;
781 static inline int cpupid_to_cpu(int cpupid)
783 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
786 static inline int cpupid_to_nid(int cpupid)
788 return cpu_to_node(cpupid_to_cpu(cpupid));
791 static inline bool cpupid_pid_unset(int cpupid)
793 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
796 static inline bool cpupid_cpu_unset(int cpupid)
798 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
801 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
803 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
806 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
807 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
808 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
810 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
813 static inline int page_cpupid_last(struct page *page)
815 return page->_last_cpupid;
817 static inline void page_cpupid_reset_last(struct page *page)
819 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
822 static inline int page_cpupid_last(struct page *page)
824 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
827 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
829 static inline void page_cpupid_reset_last(struct page *page)
831 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
833 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
834 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
836 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
837 #else /* !CONFIG_NUMA_BALANCING */
838 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
840 return page_to_nid(page); /* XXX */
843 static inline int page_cpupid_last(struct page *page)
845 return page_to_nid(page); /* XXX */
848 static inline int cpupid_to_nid(int cpupid)
853 static inline int cpupid_to_pid(int cpupid)
858 static inline int cpupid_to_cpu(int cpupid)
863 static inline int cpu_pid_to_cpupid(int nid, int pid)
868 static inline bool cpupid_pid_unset(int cpupid)
873 static inline void page_cpupid_reset_last(struct page *page)
877 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
881 #endif /* CONFIG_NUMA_BALANCING */
883 static inline struct zone *page_zone(const struct page *page)
885 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
888 #ifdef SECTION_IN_PAGE_FLAGS
889 static inline void set_page_section(struct page *page, unsigned long section)
891 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
892 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
895 static inline unsigned long page_to_section(const struct page *page)
897 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
901 static inline void set_page_zone(struct page *page, enum zone_type zone)
903 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
904 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
907 static inline void set_page_node(struct page *page, unsigned long node)
909 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
910 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
913 static inline void set_page_links(struct page *page, enum zone_type zone,
914 unsigned long node, unsigned long pfn)
916 set_page_zone(page, zone);
917 set_page_node(page, node);
918 #ifdef SECTION_IN_PAGE_FLAGS
919 set_page_section(page, pfn_to_section_nr(pfn));
924 static inline struct mem_cgroup *page_memcg(struct page *page)
926 return page->mem_cgroup;
929 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
931 page->mem_cgroup = memcg;
934 static inline struct mem_cgroup *page_memcg(struct page *page)
939 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
945 * Some inline functions in vmstat.h depend on page_zone()
947 #include <linux/vmstat.h>
949 static __always_inline void *lowmem_page_address(const struct page *page)
951 return __va(PFN_PHYS(page_to_pfn(page)));
954 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
955 #define HASHED_PAGE_VIRTUAL
958 #if defined(WANT_PAGE_VIRTUAL)
959 static inline void *page_address(const struct page *page)
961 return page->virtual;
963 static inline void set_page_address(struct page *page, void *address)
965 page->virtual = address;
967 #define page_address_init() do { } while(0)
970 #if defined(HASHED_PAGE_VIRTUAL)
971 void *page_address(const struct page *page);
972 void set_page_address(struct page *page, void *virtual);
973 void page_address_init(void);
976 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
977 #define page_address(page) lowmem_page_address(page)
978 #define set_page_address(page, address) do { } while(0)
979 #define page_address_init() do { } while(0)
982 extern void *page_rmapping(struct page *page);
983 extern struct anon_vma *page_anon_vma(struct page *page);
984 extern struct address_space *page_mapping(struct page *page);
986 extern struct address_space *__page_file_mapping(struct page *);
989 struct address_space *page_file_mapping(struct page *page)
991 if (unlikely(PageSwapCache(page)))
992 return __page_file_mapping(page);
994 return page->mapping;
998 * Return the pagecache index of the passed page. Regular pagecache pages
999 * use ->index whereas swapcache pages use ->private
1001 static inline pgoff_t page_index(struct page *page)
1003 if (unlikely(PageSwapCache(page)))
1004 return page_private(page);
1008 extern pgoff_t __page_file_index(struct page *page);
1011 * Return the file index of the page. Regular pagecache pages use ->index
1012 * whereas swapcache pages use swp_offset(->private)
1014 static inline pgoff_t page_file_index(struct page *page)
1016 if (unlikely(PageSwapCache(page)))
1017 return __page_file_index(page);
1023 * Return true if this page is mapped into pagetables.
1025 static inline int page_mapped(struct page *page)
1027 return atomic_read(&(page)->_mapcount) >= 0;
1029 struct address_space *page_mapping(struct page *page);
1032 * Return true only if the page has been allocated with
1033 * ALLOC_NO_WATERMARKS and the low watermark was not
1034 * met implying that the system is under some pressure.
1036 static inline bool page_is_pfmemalloc(struct page *page)
1039 * Page index cannot be this large so this must be
1040 * a pfmemalloc page.
1042 return page->index == -1UL;
1046 * Only to be called by the page allocator on a freshly allocated
1049 static inline void set_page_pfmemalloc(struct page *page)
1054 static inline void clear_page_pfmemalloc(struct page *page)
1060 * Different kinds of faults, as returned by handle_mm_fault().
1061 * Used to decide whether a process gets delivered SIGBUS or
1062 * just gets major/minor fault counters bumped up.
1065 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1067 #define VM_FAULT_OOM 0x0001
1068 #define VM_FAULT_SIGBUS 0x0002
1069 #define VM_FAULT_MAJOR 0x0004
1070 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1071 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1072 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1073 #define VM_FAULT_SIGSEGV 0x0040
1075 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1076 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1077 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1078 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1079 #define VM_FAULT_PTNOTSAME 0x4000 /* Page table entries have changed */
1081 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1083 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1084 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1087 /* Encode hstate index for a hwpoisoned large page */
1088 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1089 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1092 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1094 extern void pagefault_out_of_memory(void);
1096 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1099 * Flags passed to show_mem() and show_free_areas() to suppress output in
1102 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1104 extern void show_free_areas(unsigned int flags);
1105 extern bool skip_free_areas_node(unsigned int flags, int nid);
1107 void shmem_set_file(struct vm_area_struct *vma, struct file *file);
1108 int shmem_zero_setup(struct vm_area_struct *);
1110 bool shmem_mapping(struct address_space *mapping);
1112 static inline bool shmem_mapping(struct address_space *mapping)
1118 extern int can_do_mlock(void);
1119 extern int user_shm_lock(size_t, struct user_struct *);
1120 extern void user_shm_unlock(size_t, struct user_struct *);
1123 * Parameter block passed down to zap_pte_range in exceptional cases.
1125 struct zap_details {
1126 struct address_space *check_mapping; /* Check page->mapping if set */
1127 pgoff_t first_index; /* Lowest page->index to unmap */
1128 pgoff_t last_index; /* Highest page->index to unmap */
1131 static inline void INIT_VMA(struct vm_area_struct *vma)
1133 INIT_LIST_HEAD(&vma->anon_vma_chain);
1134 #ifdef CONFIG_SPECULATIVE_PAGE_FAULT
1135 seqcount_init(&vma->vm_sequence);
1136 atomic_set(&vma->vm_ref_count, 1);
1140 struct page *__vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1141 pte_t pte, unsigned long vma_flags);
1142 static inline struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1145 return __vm_normal_page(vma, addr, pte, vma->vm_flags);
1148 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1151 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1152 unsigned long size);
1153 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1154 unsigned long size, struct zap_details *);
1155 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1156 unsigned long start, unsigned long end);
1159 * This has to be called after a get_task_mm()/mmget_not_zero()
1160 * followed by taking the mmap_sem for writing before modifying the
1161 * vmas or anything the coredump pretends not to change from under it.
1163 * NOTE: find_extend_vma() called from GUP context is the only place
1164 * that can modify the "mm" (notably the vm_start/end) under mmap_sem
1165 * for reading and outside the context of the process, so it is also
1166 * the only case that holds the mmap_sem for reading that must call
1167 * this function. Generally if the mmap_sem is hold for reading
1168 * there's no need of this check after get_task_mm()/mmget_not_zero().
1170 * This function can be obsoleted and the check can be removed, after
1171 * the coredump code will hold the mmap_sem for writing before
1172 * invoking the ->core_dump methods.
1174 static inline bool mmget_still_valid(struct mm_struct *mm)
1176 return likely(!mm->core_state);
1180 * mm_walk - callbacks for walk_page_range
1181 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1182 * this handler is required to be able to handle
1183 * pmd_trans_huge() pmds. They may simply choose to
1184 * split_huge_page() instead of handling it explicitly.
1185 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1186 * @pte_hole: if set, called for each hole at all levels
1187 * @hugetlb_entry: if set, called for each hugetlb entry
1188 * @test_walk: caller specific callback function to determine whether
1189 * we walk over the current vma or not. A positive returned
1190 * value means "do page table walk over the current vma,"
1191 * and a negative one means "abort current page table walk
1192 * right now." 0 means "skip the current vma."
1193 * @mm: mm_struct representing the target process of page table walk
1194 * @vma: vma currently walked (NULL if walking outside vmas)
1195 * @private: private data for callbacks' usage
1197 * (see the comment on walk_page_range() for more details)
1200 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1201 unsigned long next, struct mm_walk *walk);
1202 int (*pte_entry)(pte_t *pte, unsigned long addr,
1203 unsigned long next, struct mm_walk *walk);
1204 int (*pte_hole)(unsigned long addr, unsigned long next,
1205 struct mm_walk *walk);
1206 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1207 unsigned long addr, unsigned long next,
1208 struct mm_walk *walk);
1209 int (*test_walk)(unsigned long addr, unsigned long next,
1210 struct mm_walk *walk);
1211 struct mm_struct *mm;
1212 struct vm_area_struct *vma;
1216 int walk_page_range(unsigned long addr, unsigned long end,
1217 struct mm_walk *walk);
1218 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1219 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1220 unsigned long end, unsigned long floor, unsigned long ceiling);
1221 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1222 struct vm_area_struct *vma);
1223 void unmap_mapping_range(struct address_space *mapping,
1224 loff_t const holebegin, loff_t const holelen, int even_cows);
1225 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1226 unsigned long *pfn);
1227 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1228 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1229 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1230 void *buf, int len, int write);
1232 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1233 loff_t const holebegin, loff_t const holelen)
1235 unmap_mapping_range(mapping, holebegin, holelen, 0);
1238 extern void truncate_pagecache(struct inode *inode, loff_t new);
1239 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1240 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1241 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1242 int truncate_inode_page(struct address_space *mapping, struct page *page);
1243 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1244 int invalidate_inode_page(struct page *page);
1247 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1248 unsigned long address, unsigned int flags);
1250 #ifdef CONFIG_SPECULATIVE_PAGE_FAULT
1251 extern int __handle_speculative_fault(struct mm_struct *mm,
1252 unsigned long address,
1253 unsigned int flags);
1254 static inline int handle_speculative_fault(struct mm_struct *mm,
1255 unsigned long address,
1259 * Try speculative page fault for multithreaded user space task only.
1261 if (!(flags & FAULT_FLAG_USER) || atomic_read(&mm->mm_users) == 1)
1262 return VM_FAULT_RETRY;
1263 return __handle_speculative_fault(mm, address, flags);
1266 static inline int handle_speculative_fault(struct mm_struct *mm,
1267 unsigned long address,
1270 return VM_FAULT_RETRY;
1272 #endif /* CONFIG_SPECULATIVE_PAGE_FAULT */
1274 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1275 unsigned long address, unsigned int fault_flags);
1277 static inline int handle_mm_fault(struct mm_struct *mm,
1278 struct vm_area_struct *vma, unsigned long address,
1281 /* should never happen if there's no MMU */
1283 return VM_FAULT_SIGBUS;
1285 static inline int fixup_user_fault(struct task_struct *tsk,
1286 struct mm_struct *mm, unsigned long address,
1287 unsigned int fault_flags)
1289 /* should never happen if there's no MMU */
1295 #ifdef CONFIG_SPECULATIVE_PAGE_FAULT
1296 static inline void vm_write_begin(struct vm_area_struct *vma)
1298 write_seqcount_begin(&vma->vm_sequence);
1300 static inline void vm_write_begin_nested(struct vm_area_struct *vma,
1303 write_seqcount_begin_nested(&vma->vm_sequence, subclass);
1305 static inline void vm_write_end(struct vm_area_struct *vma)
1307 write_seqcount_end(&vma->vm_sequence);
1309 static inline void vm_raw_write_begin(struct vm_area_struct *vma)
1311 raw_write_seqcount_begin(&vma->vm_sequence);
1313 static inline void vm_raw_write_end(struct vm_area_struct *vma)
1315 raw_write_seqcount_end(&vma->vm_sequence);
1318 static inline void vm_write_begin(struct vm_area_struct *vma)
1321 static inline void vm_write_begin_nested(struct vm_area_struct *vma,
1325 static inline void vm_write_end(struct vm_area_struct *vma)
1328 static inline void vm_raw_write_begin(struct vm_area_struct *vma)
1331 static inline void vm_raw_write_end(struct vm_area_struct *vma)
1334 #endif /* CONFIG_SPECULATIVE_PAGE_FAULT */
1336 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1337 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1338 void *buf, int len, unsigned int gup_flags);
1340 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1341 unsigned long start, unsigned long nr_pages,
1342 unsigned int foll_flags, struct page **pages,
1343 struct vm_area_struct **vmas, int *nonblocking);
1344 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1345 unsigned long start, unsigned long nr_pages,
1346 unsigned int gup_flags, struct page **pages,
1347 struct vm_area_struct **vmas);
1348 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1349 unsigned long start, unsigned long nr_pages,
1350 unsigned int gup_flags, struct page **pages, int *locked);
1351 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1352 unsigned long start, unsigned long nr_pages,
1353 struct page **pages, unsigned int gup_flags);
1354 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1355 unsigned long start, unsigned long nr_pages,
1356 struct page **pages, unsigned int gup_flags);
1357 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1358 struct page **pages);
1360 /* Container for pinned pfns / pages */
1361 struct frame_vector {
1362 unsigned int nr_allocated; /* Number of frames we have space for */
1363 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1364 bool got_ref; /* Did we pin pages by getting page ref? */
1365 bool is_pfns; /* Does array contain pages or pfns? */
1366 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1367 * pfns_vector_pages() or pfns_vector_pfns()
1371 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1372 void frame_vector_destroy(struct frame_vector *vec);
1373 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1374 unsigned int gup_flags, struct frame_vector *vec);
1375 void put_vaddr_frames(struct frame_vector *vec);
1376 int frame_vector_to_pages(struct frame_vector *vec);
1377 void frame_vector_to_pfns(struct frame_vector *vec);
1379 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1381 return vec->nr_frames;
1384 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1387 int err = frame_vector_to_pages(vec);
1390 return ERR_PTR(err);
1392 return (struct page **)(vec->ptrs);
1395 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1398 frame_vector_to_pfns(vec);
1399 return (unsigned long *)(vec->ptrs);
1403 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1404 struct page **pages);
1405 int get_kernel_page(unsigned long start, int write, struct page **pages);
1406 struct page *get_dump_page(unsigned long addr);
1408 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1409 extern void do_invalidatepage(struct page *page, unsigned int offset,
1410 unsigned int length);
1412 int __set_page_dirty_nobuffers(struct page *page);
1413 int __set_page_dirty_no_writeback(struct page *page);
1414 int redirty_page_for_writepage(struct writeback_control *wbc,
1416 void account_page_dirtied(struct page *page, struct address_space *mapping,
1417 struct mem_cgroup *memcg);
1418 void account_page_cleaned(struct page *page, struct address_space *mapping,
1419 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1420 int set_page_dirty(struct page *page);
1421 int set_page_dirty_lock(struct page *page);
1422 void cancel_dirty_page(struct page *page);
1423 int clear_page_dirty_for_io(struct page *page);
1425 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1427 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1429 return !vma->vm_ops;
1432 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1434 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1435 unsigned long old_addr, struct vm_area_struct *new_vma,
1436 unsigned long new_addr, unsigned long len,
1437 bool need_rmap_locks);
1438 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1439 unsigned long end, pgprot_t newprot,
1440 int dirty_accountable, int prot_numa);
1441 extern int mprotect_fixup(struct vm_area_struct *vma,
1442 struct vm_area_struct **pprev, unsigned long start,
1443 unsigned long end, unsigned long newflags);
1446 * doesn't attempt to fault and will return short.
1448 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1449 struct page **pages);
1451 * per-process(per-mm_struct) statistics.
1453 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1455 long val = atomic_long_read(&mm->rss_stat.count[member]);
1457 #ifdef SPLIT_RSS_COUNTING
1459 * counter is updated in asynchronous manner and may go to minus.
1460 * But it's never be expected number for users.
1465 return (unsigned long)val;
1468 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1470 atomic_long_add(value, &mm->rss_stat.count[member]);
1473 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1475 atomic_long_inc(&mm->rss_stat.count[member]);
1478 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1480 atomic_long_dec(&mm->rss_stat.count[member]);
1483 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1485 return get_mm_counter(mm, MM_FILEPAGES) +
1486 get_mm_counter(mm, MM_ANONPAGES);
1489 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1491 return max(mm->hiwater_rss, get_mm_rss(mm));
1494 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1496 return max(mm->hiwater_vm, mm->total_vm);
1499 static inline void update_hiwater_rss(struct mm_struct *mm)
1501 unsigned long _rss = get_mm_rss(mm);
1503 if ((mm)->hiwater_rss < _rss)
1504 (mm)->hiwater_rss = _rss;
1507 static inline void update_hiwater_vm(struct mm_struct *mm)
1509 if (mm->hiwater_vm < mm->total_vm)
1510 mm->hiwater_vm = mm->total_vm;
1513 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1515 mm->hiwater_rss = get_mm_rss(mm);
1518 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1519 struct mm_struct *mm)
1521 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1523 if (*maxrss < hiwater_rss)
1524 *maxrss = hiwater_rss;
1527 #if defined(SPLIT_RSS_COUNTING)
1528 void sync_mm_rss(struct mm_struct *mm);
1530 static inline void sync_mm_rss(struct mm_struct *mm)
1535 int vma_wants_writenotify(struct vm_area_struct *vma);
1537 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1539 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1543 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1547 #ifdef __PAGETABLE_PUD_FOLDED
1548 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1549 unsigned long address)
1554 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1557 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1558 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1559 unsigned long address)
1564 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1566 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1571 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1572 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1575 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1577 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1579 atomic_long_set(&mm->nr_pmds, 0);
1582 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1584 return atomic_long_read(&mm->nr_pmds);
1587 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1589 atomic_long_inc(&mm->nr_pmds);
1592 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1594 atomic_long_dec(&mm->nr_pmds);
1598 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1599 pmd_t *pmd, unsigned long address);
1600 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1603 * The following ifdef needed to get the 4level-fixup.h header to work.
1604 * Remove it when 4level-fixup.h has been removed.
1606 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1607 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1609 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1610 NULL: pud_offset(pgd, address);
1613 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1615 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1616 NULL: pmd_offset(pud, address);
1618 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1620 #if USE_SPLIT_PTE_PTLOCKS
1621 #if ALLOC_SPLIT_PTLOCKS
1622 void __init ptlock_cache_init(void);
1623 extern bool ptlock_alloc(struct page *page);
1624 extern void ptlock_free(struct page *page);
1626 static inline spinlock_t *ptlock_ptr(struct page *page)
1630 #else /* ALLOC_SPLIT_PTLOCKS */
1631 static inline void ptlock_cache_init(void)
1635 static inline bool ptlock_alloc(struct page *page)
1640 static inline void ptlock_free(struct page *page)
1644 static inline spinlock_t *ptlock_ptr(struct page *page)
1648 #endif /* ALLOC_SPLIT_PTLOCKS */
1650 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1652 return ptlock_ptr(pmd_page(*pmd));
1655 static inline bool ptlock_init(struct page *page)
1658 * prep_new_page() initialize page->private (and therefore page->ptl)
1659 * with 0. Make sure nobody took it in use in between.
1661 * It can happen if arch try to use slab for page table allocation:
1662 * slab code uses page->slab_cache, which share storage with page->ptl.
1664 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1665 if (!ptlock_alloc(page))
1667 spin_lock_init(ptlock_ptr(page));
1671 /* Reset page->mapping so free_pages_check won't complain. */
1672 static inline void pte_lock_deinit(struct page *page)
1674 page->mapping = NULL;
1678 #else /* !USE_SPLIT_PTE_PTLOCKS */
1680 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1682 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1684 return &mm->page_table_lock;
1686 static inline void ptlock_cache_init(void) {}
1687 static inline bool ptlock_init(struct page *page) { return true; }
1688 static inline void pte_lock_deinit(struct page *page) {}
1689 #endif /* USE_SPLIT_PTE_PTLOCKS */
1691 static inline void pgtable_init(void)
1693 ptlock_cache_init();
1694 pgtable_cache_init();
1697 static inline bool pgtable_page_ctor(struct page *page)
1699 if (!ptlock_init(page))
1701 inc_zone_page_state(page, NR_PAGETABLE);
1705 static inline void pgtable_page_dtor(struct page *page)
1707 pte_lock_deinit(page);
1708 dec_zone_page_state(page, NR_PAGETABLE);
1711 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1713 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1714 pte_t *__pte = pte_offset_map(pmd, address); \
1720 #define pte_unmap_unlock(pte, ptl) do { \
1725 #define pte_alloc_map(mm, vma, pmd, address) \
1726 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1728 NULL: pte_offset_map(pmd, address))
1730 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1731 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1733 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1735 #define pte_alloc_kernel(pmd, address) \
1736 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1737 NULL: pte_offset_kernel(pmd, address))
1739 #if USE_SPLIT_PMD_PTLOCKS
1741 static struct page *pmd_to_page(pmd_t *pmd)
1743 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1744 return virt_to_page((void *)((unsigned long) pmd & mask));
1747 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1749 return ptlock_ptr(pmd_to_page(pmd));
1752 static inline bool pgtable_pmd_page_ctor(struct page *page)
1754 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1755 page->pmd_huge_pte = NULL;
1757 return ptlock_init(page);
1760 static inline void pgtable_pmd_page_dtor(struct page *page)
1762 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1763 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1768 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1772 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1774 return &mm->page_table_lock;
1777 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1778 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1780 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1784 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1786 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1791 extern void free_area_init(unsigned long * zones_size);
1792 extern void free_area_init_node(int nid, unsigned long * zones_size,
1793 unsigned long zone_start_pfn, unsigned long *zholes_size);
1794 extern void free_initmem(void);
1797 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1798 * into the buddy system. The freed pages will be poisoned with pattern
1799 * "poison" if it's within range [0, UCHAR_MAX].
1800 * Return pages freed into the buddy system.
1802 extern unsigned long free_reserved_area(void *start, void *end,
1803 int poison, char *s);
1805 #ifdef CONFIG_HIGHMEM
1807 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1808 * and totalram_pages.
1810 extern void free_highmem_page(struct page *page);
1813 extern void adjust_managed_page_count(struct page *page, long count);
1814 extern void mem_init_print_info(const char *str);
1816 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1818 /* Free the reserved page into the buddy system, so it gets managed. */
1819 static inline void __free_reserved_page(struct page *page)
1821 ClearPageReserved(page);
1822 init_page_count(page);
1826 static inline void free_reserved_page(struct page *page)
1828 __free_reserved_page(page);
1829 adjust_managed_page_count(page, 1);
1832 static inline void mark_page_reserved(struct page *page)
1834 SetPageReserved(page);
1835 adjust_managed_page_count(page, -1);
1839 * Default method to free all the __init memory into the buddy system.
1840 * The freed pages will be poisoned with pattern "poison" if it's within
1841 * range [0, UCHAR_MAX].
1842 * Return pages freed into the buddy system.
1844 static inline unsigned long free_initmem_default(int poison)
1846 extern char __init_begin[], __init_end[];
1848 return free_reserved_area(&__init_begin, &__init_end,
1849 poison, "unused kernel");
1852 static inline unsigned long get_num_physpages(void)
1855 unsigned long phys_pages = 0;
1857 for_each_online_node(nid)
1858 phys_pages += node_present_pages(nid);
1863 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1865 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1866 * zones, allocate the backing mem_map and account for memory holes in a more
1867 * architecture independent manner. This is a substitute for creating the
1868 * zone_sizes[] and zholes_size[] arrays and passing them to
1869 * free_area_init_node()
1871 * An architecture is expected to register range of page frames backed by
1872 * physical memory with memblock_add[_node]() before calling
1873 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1874 * usage, an architecture is expected to do something like
1876 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1878 * for_each_valid_physical_page_range()
1879 * memblock_add_node(base, size, nid)
1880 * free_area_init_nodes(max_zone_pfns);
1882 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1883 * registered physical page range. Similarly
1884 * sparse_memory_present_with_active_regions() calls memory_present() for
1885 * each range when SPARSEMEM is enabled.
1887 * See mm/page_alloc.c for more information on each function exposed by
1888 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1890 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1891 unsigned long node_map_pfn_alignment(void);
1892 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1893 unsigned long end_pfn);
1894 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1895 unsigned long end_pfn);
1896 extern void get_pfn_range_for_nid(unsigned int nid,
1897 unsigned long *start_pfn, unsigned long *end_pfn);
1898 extern unsigned long find_min_pfn_with_active_regions(void);
1899 extern void free_bootmem_with_active_regions(int nid,
1900 unsigned long max_low_pfn);
1901 extern void sparse_memory_present_with_active_regions(int nid);
1903 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1905 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1906 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1907 static inline int __early_pfn_to_nid(unsigned long pfn,
1908 struct mminit_pfnnid_cache *state)
1913 /* please see mm/page_alloc.c */
1914 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1915 /* there is a per-arch backend function. */
1916 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1917 struct mminit_pfnnid_cache *state);
1920 extern void set_dma_reserve(unsigned long new_dma_reserve);
1921 extern void memmap_init_zone(unsigned long, int, unsigned long,
1922 unsigned long, enum memmap_context);
1923 extern void setup_per_zone_wmarks(void);
1924 extern int __meminit init_per_zone_wmark_min(void);
1925 extern void mem_init(void);
1926 extern void __init mmap_init(void);
1927 extern void show_mem(unsigned int flags);
1928 extern void si_meminfo(struct sysinfo * val);
1929 extern void si_meminfo_node(struct sysinfo *val, int nid);
1931 extern __printf(3, 4)
1932 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1933 const char *fmt, ...);
1935 extern void setup_per_cpu_pageset(void);
1937 extern void zone_pcp_update(struct zone *zone);
1938 extern void zone_pcp_reset(struct zone *zone);
1941 extern int min_free_kbytes;
1944 extern atomic_long_t mmap_pages_allocated;
1945 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1947 /* interval_tree.c */
1948 void vma_interval_tree_insert(struct vm_area_struct *node,
1949 struct rb_root *root);
1950 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1951 struct vm_area_struct *prev,
1952 struct rb_root *root);
1953 void vma_interval_tree_remove(struct vm_area_struct *node,
1954 struct rb_root *root);
1955 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1956 unsigned long start, unsigned long last);
1957 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1958 unsigned long start, unsigned long last);
1960 #define vma_interval_tree_foreach(vma, root, start, last) \
1961 for (vma = vma_interval_tree_iter_first(root, start, last); \
1962 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1964 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1965 struct rb_root *root);
1966 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1967 struct rb_root *root);
1968 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1969 struct rb_root *root, unsigned long start, unsigned long last);
1970 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1971 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1972 #ifdef CONFIG_DEBUG_VM_RB
1973 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1976 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1977 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1978 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1981 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1982 extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start,
1983 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert, bool keep_locked);
1984 static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1985 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert)
1987 return __vma_adjust(vma, start, end, pgoff, insert, false);
1989 extern struct vm_area_struct *__vma_merge(struct mm_struct *,
1990 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1991 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1992 struct mempolicy *, struct vm_userfaultfd_ctx, const char __user *, bool keep_locked);
1993 static inline struct vm_area_struct *vma_merge(struct mm_struct *mm,
1994 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1995 unsigned long vm_flags, struct anon_vma *anon, struct file *file, pgoff_t off,
1996 struct mempolicy *pol, struct vm_userfaultfd_ctx ctx, const char __user *anon_name)
1998 return __vma_merge(mm, prev, addr, end, vm_flags, anon, file,
1999 off, pol, ctx, anon_name, false);
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);
2033 extern struct file *get_task_exe_file(struct task_struct *task);
2035 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
2036 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2037 unsigned long addr, unsigned long len,
2038 unsigned long flags,
2039 const struct vm_special_mapping *spec);
2040 /* This is an obsolete alternative to _install_special_mapping. */
2041 extern int install_special_mapping(struct mm_struct *mm,
2042 unsigned long addr, unsigned long len,
2043 unsigned long flags, struct page **pages);
2045 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2047 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2048 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2049 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2050 unsigned long len, unsigned long prot, unsigned long flags,
2051 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2052 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2054 static inline unsigned long
2055 do_mmap_pgoff(struct file *file, unsigned long addr,
2056 unsigned long len, unsigned long prot, unsigned long flags,
2057 unsigned long pgoff, unsigned long *populate)
2059 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2063 extern int __mm_populate(unsigned long addr, unsigned long len,
2065 static inline void mm_populate(unsigned long addr, unsigned long len)
2068 (void) __mm_populate(addr, len, 1);
2071 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2074 /* These take the mm semaphore themselves */
2075 extern unsigned long vm_brk(unsigned long, unsigned long);
2076 extern int vm_munmap(unsigned long, size_t);
2077 extern unsigned long vm_mmap(struct file *, unsigned long,
2078 unsigned long, unsigned long,
2079 unsigned long, unsigned long);
2081 struct vm_unmapped_area_info {
2082 #define VM_UNMAPPED_AREA_TOPDOWN 1
2083 unsigned long flags;
2084 unsigned long length;
2085 unsigned long low_limit;
2086 unsigned long high_limit;
2087 unsigned long align_mask;
2088 unsigned long align_offset;
2091 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2092 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2095 * Search for an unmapped address range.
2097 * We are looking for a range that:
2098 * - does not intersect with any VMA;
2099 * - is contained within the [low_limit, high_limit) interval;
2100 * - is at least the desired size.
2101 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2103 static inline unsigned long
2104 vm_unmapped_area(struct vm_unmapped_area_info *info)
2106 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2107 return unmapped_area_topdown(info);
2109 return unmapped_area(info);
2113 extern void truncate_inode_pages(struct address_space *, loff_t);
2114 extern void truncate_inode_pages_fill_zero(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_range_fill_zero(struct address_space *,
2118 loff_t lstart, loff_t lend);
2119 extern void truncate_inode_pages_final(struct address_space *);
2121 /* generic vm_area_ops exported for stackable file systems */
2122 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2123 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2124 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2126 /* mm/page-writeback.c */
2127 int write_one_page(struct page *page, int wait);
2128 void task_dirty_inc(struct task_struct *tsk);
2131 #define VM_MAX_READAHEAD CONFIG_VM_MAX_READAHEAD /* kbytes */
2132 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2134 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2135 pgoff_t offset, unsigned long nr_to_read);
2137 void page_cache_sync_readahead(struct address_space *mapping,
2138 struct file_ra_state *ra,
2141 unsigned long size);
2143 void page_cache_async_readahead(struct address_space *mapping,
2144 struct file_ra_state *ra,
2148 unsigned long size);
2150 extern unsigned long stack_guard_gap;
2151 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2152 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2154 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2155 extern int expand_downwards(struct vm_area_struct *vma,
2156 unsigned long address);
2158 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2160 #define expand_upwards(vma, address) (0)
2163 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2164 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2165 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2166 struct vm_area_struct **pprev);
2168 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2169 NULL if none. Assume start_addr < end_addr. */
2170 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2172 struct vm_area_struct * vma = find_vma(mm,start_addr);
2174 if (vma && end_addr <= vma->vm_start)
2179 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2181 unsigned long vm_start = vma->vm_start;
2183 if (vma->vm_flags & VM_GROWSDOWN) {
2184 vm_start -= stack_guard_gap;
2185 if (vm_start > vma->vm_start)
2191 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2193 unsigned long vm_end = vma->vm_end;
2195 if (vma->vm_flags & VM_GROWSUP) {
2196 vm_end += stack_guard_gap;
2197 if (vm_end < vma->vm_end)
2198 vm_end = -PAGE_SIZE;
2203 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2205 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2208 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2209 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2210 unsigned long vm_start, unsigned long vm_end)
2212 struct vm_area_struct *vma = find_vma(mm, vm_start);
2214 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2220 static inline bool range_in_vma(struct vm_area_struct *vma,
2221 unsigned long start, unsigned long end)
2223 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2227 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2228 void vma_set_page_prot(struct vm_area_struct *vma);
2230 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2234 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2236 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2240 #ifdef CONFIG_NUMA_BALANCING
2241 unsigned long change_prot_numa(struct vm_area_struct *vma,
2242 unsigned long start, unsigned long end);
2245 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2246 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2247 unsigned long pfn, unsigned long size, pgprot_t);
2248 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2249 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2251 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2252 unsigned long pfn, pgprot_t pgprot);
2253 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2255 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2258 struct page *follow_page_mask(struct vm_area_struct *vma,
2259 unsigned long address, unsigned int foll_flags,
2260 unsigned int *page_mask);
2262 static inline struct page *follow_page(struct vm_area_struct *vma,
2263 unsigned long address, unsigned int foll_flags)
2265 unsigned int unused_page_mask;
2266 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2269 #define FOLL_WRITE 0x01 /* check pte is writable */
2270 #define FOLL_TOUCH 0x02 /* mark page accessed */
2271 #define FOLL_GET 0x04 /* do get_page on page */
2272 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2273 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2274 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2275 * and return without waiting upon it */
2276 #define FOLL_POPULATE 0x40 /* fault in page */
2277 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2278 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2279 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2280 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2281 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2282 #define FOLL_MLOCK 0x1000 /* lock present pages */
2283 #define FOLL_COW 0x4000 /* internal GUP flag */
2284 #define FOLL_ANON 0x8000 /* don't do file mappings */
2286 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2288 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2289 unsigned long size, pte_fn_t fn, void *data);
2291 #ifdef CONFIG_PROC_FS
2292 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2294 static inline void vm_stat_account(struct mm_struct *mm,
2295 unsigned long flags, struct file *file, long pages)
2297 mm->total_vm += pages;
2299 #endif /* CONFIG_PROC_FS */
2301 #ifdef CONFIG_PAGE_POISONING
2302 extern bool page_poisoning_enabled(void);
2303 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2304 extern bool page_is_poisoned(struct page *page);
2306 static inline bool page_poisoning_enabled(void) { return false; }
2307 static inline void kernel_poison_pages(struct page *page, int numpages,
2309 static inline bool page_is_poisoned(struct page *page) { return false; }
2312 #ifdef CONFIG_DEBUG_PAGEALLOC
2313 extern bool _debug_pagealloc_enabled;
2314 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2316 static inline bool debug_pagealloc_enabled(void)
2318 return _debug_pagealloc_enabled;
2322 kernel_map_pages(struct page *page, int numpages, int enable)
2324 if (!debug_pagealloc_enabled())
2327 __kernel_map_pages(page, numpages, enable);
2329 #ifdef CONFIG_HIBERNATION
2330 extern bool kernel_page_present(struct page *page);
2331 #endif /* CONFIG_HIBERNATION */
2332 #else /* CONFIG_DEBUG_PAGEALLOC */
2334 kernel_map_pages(struct page *page, int numpages, int enable) {}
2335 #ifdef CONFIG_HIBERNATION
2336 static inline bool kernel_page_present(struct page *page) { return true; }
2337 #endif /* CONFIG_HIBERNATION */
2338 static inline bool debug_pagealloc_enabled(void)
2342 #endif /* CONFIG_DEBUG_PAGEALLOC */
2344 #ifdef __HAVE_ARCH_GATE_AREA
2345 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2346 extern int in_gate_area_no_mm(unsigned long addr);
2347 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2349 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2353 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2354 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2358 #endif /* __HAVE_ARCH_GATE_AREA */
2360 #ifdef CONFIG_SYSCTL
2361 extern int sysctl_drop_caches;
2362 int drop_caches_sysctl_handler(struct ctl_table *, int,
2363 void __user *, size_t *, loff_t *);
2366 void drop_slab(void);
2367 void drop_slab_node(int nid);
2369 void drop_pagecache_sb(struct super_block *sb, void *unused);
2372 #define randomize_va_space 0
2374 extern int randomize_va_space;
2377 const char * arch_vma_name(struct vm_area_struct *vma);
2378 void print_vma_addr(char *prefix, unsigned long rip);
2380 void sparse_mem_maps_populate_node(struct page **map_map,
2381 unsigned long pnum_begin,
2382 unsigned long pnum_end,
2383 unsigned long map_count,
2386 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2387 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2388 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2389 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2390 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2391 void *vmemmap_alloc_block(unsigned long size, int node);
2392 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2393 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2394 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2396 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2397 void vmemmap_populate_print_last(void);
2398 #ifdef CONFIG_MEMORY_HOTPLUG
2399 void vmemmap_free(unsigned long start, unsigned long end);
2401 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2402 unsigned long size);
2405 MF_COUNT_INCREASED = 1 << 0,
2406 MF_ACTION_REQUIRED = 1 << 1,
2407 MF_MUST_KILL = 1 << 2,
2408 MF_SOFT_OFFLINE = 1 << 3,
2410 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2411 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2412 extern int unpoison_memory(unsigned long pfn);
2413 extern int get_hwpoison_page(struct page *page);
2414 extern void put_hwpoison_page(struct page *page);
2415 extern int sysctl_memory_failure_early_kill;
2416 extern int sysctl_memory_failure_recovery;
2417 extern void shake_page(struct page *p, int access);
2418 extern atomic_long_t num_poisoned_pages;
2419 extern int soft_offline_page(struct page *page, int flags);
2423 * Error handlers for various types of pages.
2426 MF_IGNORED, /* Error: cannot be handled */
2427 MF_FAILED, /* Error: handling failed */
2428 MF_DELAYED, /* Will be handled later */
2429 MF_RECOVERED, /* Successfully recovered */
2432 enum mf_action_page_type {
2434 MF_MSG_KERNEL_HIGH_ORDER,
2436 MF_MSG_DIFFERENT_COMPOUND,
2437 MF_MSG_POISONED_HUGE,
2440 MF_MSG_UNMAP_FAILED,
2441 MF_MSG_DIRTY_SWAPCACHE,
2442 MF_MSG_CLEAN_SWAPCACHE,
2443 MF_MSG_DIRTY_MLOCKED_LRU,
2444 MF_MSG_CLEAN_MLOCKED_LRU,
2445 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2446 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2449 MF_MSG_TRUNCATED_LRU,
2455 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2456 extern void clear_huge_page(struct page *page,
2458 unsigned int pages_per_huge_page);
2459 extern void copy_user_huge_page(struct page *dst, struct page *src,
2460 unsigned long addr, struct vm_area_struct *vma,
2461 unsigned int pages_per_huge_page);
2462 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2464 extern struct page_ext_operations debug_guardpage_ops;
2466 #ifdef CONFIG_DEBUG_PAGEALLOC
2467 extern unsigned int _debug_guardpage_minorder;
2468 extern bool _debug_guardpage_enabled;
2470 static inline unsigned int debug_guardpage_minorder(void)
2472 return _debug_guardpage_minorder;
2475 static inline bool debug_guardpage_enabled(void)
2477 return _debug_guardpage_enabled;
2480 static inline bool page_is_guard(struct page *page)
2482 struct page_ext *page_ext;
2484 if (!debug_guardpage_enabled())
2487 page_ext = lookup_page_ext(page);
2488 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2491 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2492 static inline bool debug_guardpage_enabled(void) { return false; }
2493 static inline bool page_is_guard(struct page *page) { return false; }
2494 #endif /* CONFIG_DEBUG_PAGEALLOC */
2496 #if MAX_NUMNODES > 1
2497 void __init setup_nr_node_ids(void);
2499 static inline void setup_nr_node_ids(void) {}
2502 #ifdef CONFIG_PROCESS_RECLAIM
2503 struct reclaim_param {
2504 struct vm_area_struct *vma;
2505 /* Number of pages scanned */
2507 /* max pages to reclaim */
2509 /* pages reclaimed */
2512 extern struct reclaim_param reclaim_task_anon(struct task_struct *task,
2514 extern int reclaim_pte_range(pmd_t *pmd, unsigned long addr,
2515 unsigned long end, struct mm_walk *walk);
2516 extern unsigned long reclaim_global(unsigned long nr_to_reclaim);
2519 #endif /* __KERNEL__ */
2520 #endif /* _LINUX_MM_H */