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;
33 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
34 extern unsigned long max_mapnr;
36 static inline void set_max_mapnr(unsigned long limit)
41 static inline void set_max_mapnr(unsigned long limit) { }
44 extern unsigned long totalram_pages;
45 extern void * high_memory;
46 extern int page_cluster;
49 extern int sysctl_legacy_va_layout;
51 #define sysctl_legacy_va_layout 0
55 #include <asm/pgtable.h>
56 #include <asm/processor.h>
59 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
63 * To prevent common memory management code establishing
64 * a zero page mapping on a read fault.
65 * This macro should be defined within <asm/pgtable.h>.
66 * s390 does this to prevent multiplexing of hardware bits
67 * related to the physical page in case of virtualization.
69 #ifndef mm_forbids_zeropage
70 #define mm_forbids_zeropage(X) (0)
73 extern unsigned long sysctl_user_reserve_kbytes;
74 extern unsigned long sysctl_admin_reserve_kbytes;
76 extern int sysctl_overcommit_memory;
77 extern int sysctl_overcommit_ratio;
78 extern unsigned long sysctl_overcommit_kbytes;
80 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
82 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
85 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
87 /* to align the pointer to the (next) page boundary */
88 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
90 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
91 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
94 * Linux kernel virtual memory manager primitives.
95 * The idea being to have a "virtual" mm in the same way
96 * we have a virtual fs - giving a cleaner interface to the
97 * mm details, and allowing different kinds of memory mappings
98 * (from shared memory to executable loading to arbitrary
102 extern struct kmem_cache *vm_area_cachep;
105 extern struct rb_root nommu_region_tree;
106 extern struct rw_semaphore nommu_region_sem;
108 extern unsigned int kobjsize(const void *objp);
112 * vm_flags in vm_area_struct, see mm_types.h.
114 #define VM_NONE 0x00000000
116 #define VM_READ 0x00000001 /* currently active flags */
117 #define VM_WRITE 0x00000002
118 #define VM_EXEC 0x00000004
119 #define VM_SHARED 0x00000008
121 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
122 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
123 #define VM_MAYWRITE 0x00000020
124 #define VM_MAYEXEC 0x00000040
125 #define VM_MAYSHARE 0x00000080
127 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
128 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
129 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
130 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
131 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
133 #define VM_LOCKED 0x00002000
134 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
136 /* Used by sys_madvise() */
137 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
138 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
140 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
141 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
142 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
143 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
144 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
145 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
146 #define VM_ARCH_2 0x02000000
147 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
149 #ifdef CONFIG_MEM_SOFT_DIRTY
150 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
152 # define VM_SOFTDIRTY 0
155 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
156 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
157 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
158 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
160 #if defined(CONFIG_X86)
161 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
162 #elif defined(CONFIG_PPC)
163 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
164 #elif defined(CONFIG_PARISC)
165 # define VM_GROWSUP VM_ARCH_1
166 #elif defined(CONFIG_METAG)
167 # define VM_GROWSUP VM_ARCH_1
168 #elif defined(CONFIG_IA64)
169 # define VM_GROWSUP VM_ARCH_1
170 #elif !defined(CONFIG_MMU)
171 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
174 #if defined(CONFIG_X86)
175 /* MPX specific bounds table or bounds directory */
176 # define VM_MPX VM_ARCH_2
180 # define VM_GROWSUP VM_NONE
183 /* Bits set in the VMA until the stack is in its final location */
184 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
186 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
187 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
190 #ifdef CONFIG_STACK_GROWSUP
191 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
193 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
197 * Special vmas that are non-mergable, non-mlock()able.
198 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
200 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
202 /* This mask defines which mm->def_flags a process can inherit its parent */
203 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
206 * mapping from the currently active vm_flags protection bits (the
207 * low four bits) to a page protection mask..
209 extern pgprot_t protection_map[16];
211 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
212 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
213 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
214 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
215 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
216 #define FAULT_FLAG_TRIED 0x20 /* Second try */
217 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
220 * vm_fault is filled by the the pagefault handler and passed to the vma's
221 * ->fault function. The vma's ->fault is responsible for returning a bitmask
222 * of VM_FAULT_xxx flags that give details about how the fault was handled.
224 * pgoff should be used in favour of virtual_address, if possible.
227 unsigned int flags; /* FAULT_FLAG_xxx flags */
228 pgoff_t pgoff; /* Logical page offset based on vma */
229 void __user *virtual_address; /* Faulting virtual address */
231 struct page *cow_page; /* Handler may choose to COW */
232 struct page *page; /* ->fault handlers should return a
233 * page here, unless VM_FAULT_NOPAGE
234 * is set (which is also implied by
237 /* for ->map_pages() only */
238 pgoff_t max_pgoff; /* map pages for offset from pgoff till
239 * max_pgoff inclusive */
240 pte_t *pte; /* pte entry associated with ->pgoff */
244 * These are the virtual MM functions - opening of an area, closing and
245 * unmapping it (needed to keep files on disk up-to-date etc), pointer
246 * to the functions called when a no-page or a wp-page exception occurs.
248 struct vm_operations_struct {
249 void (*open)(struct vm_area_struct * area);
250 void (*close)(struct vm_area_struct * area);
251 int (*mremap)(struct vm_area_struct * area);
252 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
253 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
254 pmd_t *, unsigned int flags);
255 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
257 /* notification that a previously read-only page is about to become
258 * writable, if an error is returned it will cause a SIGBUS */
259 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
261 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
262 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
264 /* called by access_process_vm when get_user_pages() fails, typically
265 * for use by special VMAs that can switch between memory and hardware
267 int (*access)(struct vm_area_struct *vma, unsigned long addr,
268 void *buf, int len, int write);
270 /* Called by the /proc/PID/maps code to ask the vma whether it
271 * has a special name. Returning non-NULL will also cause this
272 * vma to be dumped unconditionally. */
273 const char *(*name)(struct vm_area_struct *vma);
277 * set_policy() op must add a reference to any non-NULL @new mempolicy
278 * to hold the policy upon return. Caller should pass NULL @new to
279 * remove a policy and fall back to surrounding context--i.e. do not
280 * install a MPOL_DEFAULT policy, nor the task or system default
283 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
286 * get_policy() op must add reference [mpol_get()] to any policy at
287 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
288 * in mm/mempolicy.c will do this automatically.
289 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
290 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
291 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
292 * must return NULL--i.e., do not "fallback" to task or system default
295 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
299 * Called by vm_normal_page() for special PTEs to find the
300 * page for @addr. This is useful if the default behavior
301 * (using pte_page()) would not find the correct page.
303 struct page *(*find_special_page)(struct vm_area_struct *vma,
310 #define page_private(page) ((page)->private)
311 #define set_page_private(page, v) ((page)->private = (v))
314 * FIXME: take this include out, include page-flags.h in
315 * files which need it (119 of them)
317 #include <linux/page-flags.h>
318 #include <linux/huge_mm.h>
321 * Methods to modify the page usage count.
323 * What counts for a page usage:
324 * - cache mapping (page->mapping)
325 * - private data (page->private)
326 * - page mapped in a task's page tables, each mapping
327 * is counted separately
329 * Also, many kernel routines increase the page count before a critical
330 * routine so they can be sure the page doesn't go away from under them.
334 * Drop a ref, return true if the refcount fell to zero (the page has no users)
336 static inline int put_page_testzero(struct page *page)
338 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
339 return atomic_dec_and_test(&page->_count);
343 * Try to grab a ref unless the page has a refcount of zero, return false if
345 * This can be called when MMU is off so it must not access
346 * any of the virtual mappings.
348 static inline int get_page_unless_zero(struct page *page)
350 return atomic_inc_not_zero(&page->_count);
353 extern int page_is_ram(unsigned long pfn);
361 int region_intersects(resource_size_t offset, size_t size, const char *type);
363 /* Support for virtually mapped pages */
364 struct page *vmalloc_to_page(const void *addr);
365 unsigned long vmalloc_to_pfn(const void *addr);
368 * Determine if an address is within the vmalloc range
370 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
371 * is no special casing required.
373 static inline int is_vmalloc_addr(const void *x)
376 unsigned long addr = (unsigned long)x;
378 return addr >= VMALLOC_START && addr < VMALLOC_END;
384 extern int is_vmalloc_or_module_addr(const void *x);
386 static inline int is_vmalloc_or_module_addr(const void *x)
392 extern void kvfree(const void *addr);
394 static inline void compound_lock(struct page *page)
396 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
397 VM_BUG_ON_PAGE(PageSlab(page), page);
398 bit_spin_lock(PG_compound_lock, &page->flags);
402 static inline void compound_unlock(struct page *page)
404 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
405 VM_BUG_ON_PAGE(PageSlab(page), page);
406 bit_spin_unlock(PG_compound_lock, &page->flags);
410 static inline unsigned long compound_lock_irqsave(struct page *page)
412 unsigned long uninitialized_var(flags);
413 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
414 local_irq_save(flags);
420 static inline void compound_unlock_irqrestore(struct page *page,
423 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
424 compound_unlock(page);
425 local_irq_restore(flags);
429 static inline struct page *compound_head_by_tail(struct page *tail)
431 struct page *head = tail->first_page;
434 * page->first_page may be a dangling pointer to an old
435 * compound page, so recheck that it is still a tail
436 * page before returning.
439 if (likely(PageTail(tail)))
445 * Since either compound page could be dismantled asynchronously in THP
446 * or we access asynchronously arbitrary positioned struct page, there
447 * would be tail flag race. To handle this race, we should call
448 * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
450 static inline struct page *compound_head(struct page *page)
452 if (unlikely(PageTail(page)))
453 return compound_head_by_tail(page);
458 * If we access compound page synchronously such as access to
459 * allocated page, there is no need to handle tail flag race, so we can
460 * check tail flag directly without any synchronization primitive.
462 static inline struct page *compound_head_fast(struct page *page)
464 if (unlikely(PageTail(page)))
465 return page->first_page;
470 * The atomic page->_mapcount, starts from -1: so that transitions
471 * both from it and to it can be tracked, using atomic_inc_and_test
472 * and atomic_add_negative(-1).
474 static inline void page_mapcount_reset(struct page *page)
476 atomic_set(&(page)->_mapcount, -1);
479 static inline int page_mapcount(struct page *page)
481 VM_BUG_ON_PAGE(PageSlab(page), page);
482 return atomic_read(&page->_mapcount) + 1;
485 static inline int page_count(struct page *page)
487 return atomic_read(&compound_head(page)->_count);
490 static inline bool __compound_tail_refcounted(struct page *page)
492 return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
496 * This takes a head page as parameter and tells if the
497 * tail page reference counting can be skipped.
499 * For this to be safe, PageSlab and PageHeadHuge must remain true on
500 * any given page where they return true here, until all tail pins
501 * have been released.
503 static inline bool compound_tail_refcounted(struct page *page)
505 VM_BUG_ON_PAGE(!PageHead(page), page);
506 return __compound_tail_refcounted(page);
509 static inline void get_huge_page_tail(struct page *page)
512 * __split_huge_page_refcount() cannot run from under us.
514 VM_BUG_ON_PAGE(!PageTail(page), page);
515 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
516 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
517 if (compound_tail_refcounted(page->first_page))
518 atomic_inc(&page->_mapcount);
521 extern bool __get_page_tail(struct page *page);
523 static inline void get_page(struct page *page)
525 if (unlikely(PageTail(page)))
526 if (likely(__get_page_tail(page)))
529 * Getting a normal page or the head of a compound page
530 * requires to already have an elevated page->_count.
532 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
533 atomic_inc(&page->_count);
536 static inline struct page *virt_to_head_page(const void *x)
538 struct page *page = virt_to_page(x);
541 * We don't need to worry about synchronization of tail flag
542 * when we call virt_to_head_page() since it is only called for
543 * already allocated page and this page won't be freed until
544 * this virt_to_head_page() is finished. So use _fast variant.
546 return compound_head_fast(page);
550 * Setup the page count before being freed into the page allocator for
551 * the first time (boot or memory hotplug)
553 static inline void init_page_count(struct page *page)
555 atomic_set(&page->_count, 1);
558 void put_page(struct page *page);
559 void put_pages_list(struct list_head *pages);
561 void split_page(struct page *page, unsigned int order);
562 int split_free_page(struct page *page);
565 * Compound pages have a destructor function. Provide a
566 * prototype for that function and accessor functions.
567 * These are _only_ valid on the head of a PG_compound page.
570 static inline void set_compound_page_dtor(struct page *page,
571 compound_page_dtor *dtor)
573 page[1].compound_dtor = dtor;
576 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
578 return page[1].compound_dtor;
581 static inline int compound_order(struct page *page)
585 return page[1].compound_order;
588 static inline void set_compound_order(struct page *page, unsigned long order)
590 page[1].compound_order = order;
595 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
596 * servicing faults for write access. In the normal case, do always want
597 * pte_mkwrite. But get_user_pages can cause write faults for mappings
598 * that do not have writing enabled, when used by access_process_vm.
600 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
602 if (likely(vma->vm_flags & VM_WRITE))
603 pte = pte_mkwrite(pte);
607 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
608 struct page *page, pte_t *pte, bool write, bool anon);
612 * Multiple processes may "see" the same page. E.g. for untouched
613 * mappings of /dev/null, all processes see the same page full of
614 * zeroes, and text pages of executables and shared libraries have
615 * only one copy in memory, at most, normally.
617 * For the non-reserved pages, page_count(page) denotes a reference count.
618 * page_count() == 0 means the page is free. page->lru is then used for
619 * freelist management in the buddy allocator.
620 * page_count() > 0 means the page has been allocated.
622 * Pages are allocated by the slab allocator in order to provide memory
623 * to kmalloc and kmem_cache_alloc. In this case, the management of the
624 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
625 * unless a particular usage is carefully commented. (the responsibility of
626 * freeing the kmalloc memory is the caller's, of course).
628 * A page may be used by anyone else who does a __get_free_page().
629 * In this case, page_count still tracks the references, and should only
630 * be used through the normal accessor functions. The top bits of page->flags
631 * and page->virtual store page management information, but all other fields
632 * are unused and could be used privately, carefully. The management of this
633 * page is the responsibility of the one who allocated it, and those who have
634 * subsequently been given references to it.
636 * The other pages (we may call them "pagecache pages") are completely
637 * managed by the Linux memory manager: I/O, buffers, swapping etc.
638 * The following discussion applies only to them.
640 * A pagecache page contains an opaque `private' member, which belongs to the
641 * page's address_space. Usually, this is the address of a circular list of
642 * the page's disk buffers. PG_private must be set to tell the VM to call
643 * into the filesystem to release these pages.
645 * A page may belong to an inode's memory mapping. In this case, page->mapping
646 * is the pointer to the inode, and page->index is the file offset of the page,
647 * in units of PAGE_CACHE_SIZE.
649 * If pagecache pages are not associated with an inode, they are said to be
650 * anonymous pages. These may become associated with the swapcache, and in that
651 * case PG_swapcache is set, and page->private is an offset into the swapcache.
653 * In either case (swapcache or inode backed), the pagecache itself holds one
654 * reference to the page. Setting PG_private should also increment the
655 * refcount. The each user mapping also has a reference to the page.
657 * The pagecache pages are stored in a per-mapping radix tree, which is
658 * rooted at mapping->page_tree, and indexed by offset.
659 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
660 * lists, we instead now tag pages as dirty/writeback in the radix tree.
662 * All pagecache pages may be subject to I/O:
663 * - inode pages may need to be read from disk,
664 * - inode pages which have been modified and are MAP_SHARED may need
665 * to be written back to the inode on disk,
666 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
667 * modified may need to be swapped out to swap space and (later) to be read
672 * The zone field is never updated after free_area_init_core()
673 * sets it, so none of the operations on it need to be atomic.
676 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
677 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
678 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
679 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
680 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
683 * Define the bit shifts to access each section. For non-existent
684 * sections we define the shift as 0; that plus a 0 mask ensures
685 * the compiler will optimise away reference to them.
687 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
688 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
689 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
690 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
692 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
693 #ifdef NODE_NOT_IN_PAGE_FLAGS
694 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
695 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
696 SECTIONS_PGOFF : ZONES_PGOFF)
698 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
699 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
700 NODES_PGOFF : ZONES_PGOFF)
703 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
705 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
706 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
709 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
710 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
711 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
712 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
713 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
715 static inline enum zone_type page_zonenum(const struct page *page)
717 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
720 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
721 #define SECTION_IN_PAGE_FLAGS
725 * The identification function is mainly used by the buddy allocator for
726 * determining if two pages could be buddies. We are not really identifying
727 * the zone since we could be using the section number id if we do not have
728 * node id available in page flags.
729 * We only guarantee that it will return the same value for two combinable
732 static inline int page_zone_id(struct page *page)
734 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
737 static inline int zone_to_nid(struct zone *zone)
746 #ifdef NODE_NOT_IN_PAGE_FLAGS
747 extern int page_to_nid(const struct page *page);
749 static inline int page_to_nid(const struct page *page)
751 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
755 #ifdef CONFIG_NUMA_BALANCING
756 static inline int cpu_pid_to_cpupid(int cpu, int pid)
758 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
761 static inline int cpupid_to_pid(int cpupid)
763 return cpupid & LAST__PID_MASK;
766 static inline int cpupid_to_cpu(int cpupid)
768 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
771 static inline int cpupid_to_nid(int cpupid)
773 return cpu_to_node(cpupid_to_cpu(cpupid));
776 static inline bool cpupid_pid_unset(int cpupid)
778 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
781 static inline bool cpupid_cpu_unset(int cpupid)
783 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
786 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
788 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
791 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
792 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
793 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
795 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
798 static inline int page_cpupid_last(struct page *page)
800 return page->_last_cpupid;
802 static inline void page_cpupid_reset_last(struct page *page)
804 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
807 static inline int page_cpupid_last(struct page *page)
809 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
812 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
814 static inline void page_cpupid_reset_last(struct page *page)
816 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
818 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
819 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
821 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
822 #else /* !CONFIG_NUMA_BALANCING */
823 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
825 return page_to_nid(page); /* XXX */
828 static inline int page_cpupid_last(struct page *page)
830 return page_to_nid(page); /* XXX */
833 static inline int cpupid_to_nid(int cpupid)
838 static inline int cpupid_to_pid(int cpupid)
843 static inline int cpupid_to_cpu(int cpupid)
848 static inline int cpu_pid_to_cpupid(int nid, int pid)
853 static inline bool cpupid_pid_unset(int cpupid)
858 static inline void page_cpupid_reset_last(struct page *page)
862 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
866 #endif /* CONFIG_NUMA_BALANCING */
868 static inline struct zone *page_zone(const struct page *page)
870 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
873 #ifdef SECTION_IN_PAGE_FLAGS
874 static inline void set_page_section(struct page *page, unsigned long section)
876 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
877 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
880 static inline unsigned long page_to_section(const struct page *page)
882 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
886 static inline void set_page_zone(struct page *page, enum zone_type zone)
888 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
889 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
892 static inline void set_page_node(struct page *page, unsigned long node)
894 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
895 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
898 static inline void set_page_links(struct page *page, enum zone_type zone,
899 unsigned long node, unsigned long pfn)
901 set_page_zone(page, zone);
902 set_page_node(page, node);
903 #ifdef SECTION_IN_PAGE_FLAGS
904 set_page_section(page, pfn_to_section_nr(pfn));
909 * Some inline functions in vmstat.h depend on page_zone()
911 #include <linux/vmstat.h>
913 static __always_inline void *lowmem_page_address(const struct page *page)
915 return __va(PFN_PHYS(page_to_pfn(page)));
918 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
919 #define HASHED_PAGE_VIRTUAL
922 #if defined(WANT_PAGE_VIRTUAL)
923 static inline void *page_address(const struct page *page)
925 return page->virtual;
927 static inline void set_page_address(struct page *page, void *address)
929 page->virtual = address;
931 #define page_address_init() do { } while(0)
934 #if defined(HASHED_PAGE_VIRTUAL)
935 void *page_address(const struct page *page);
936 void set_page_address(struct page *page, void *virtual);
937 void page_address_init(void);
940 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
941 #define page_address(page) lowmem_page_address(page)
942 #define set_page_address(page, address) do { } while(0)
943 #define page_address_init() do { } while(0)
946 extern void *page_rmapping(struct page *page);
947 extern struct anon_vma *page_anon_vma(struct page *page);
948 extern struct address_space *page_mapping(struct page *page);
950 extern struct address_space *__page_file_mapping(struct page *);
953 struct address_space *page_file_mapping(struct page *page)
955 if (unlikely(PageSwapCache(page)))
956 return __page_file_mapping(page);
958 return page->mapping;
962 * Return the pagecache index of the passed page. Regular pagecache pages
963 * use ->index whereas swapcache pages use ->private
965 static inline pgoff_t page_index(struct page *page)
967 if (unlikely(PageSwapCache(page)))
968 return page_private(page);
972 extern pgoff_t __page_file_index(struct page *page);
975 * Return the file index of the page. Regular pagecache pages use ->index
976 * whereas swapcache pages use swp_offset(->private)
978 static inline pgoff_t page_file_index(struct page *page)
980 if (unlikely(PageSwapCache(page)))
981 return __page_file_index(page);
987 * Return true if this page is mapped into pagetables.
989 static inline int page_mapped(struct page *page)
991 return atomic_read(&(page)->_mapcount) >= 0;
995 * Return true only if the page has been allocated with
996 * ALLOC_NO_WATERMARKS and the low watermark was not
997 * met implying that the system is under some pressure.
999 static inline bool page_is_pfmemalloc(struct page *page)
1002 * Page index cannot be this large so this must be
1003 * a pfmemalloc page.
1005 return page->index == -1UL;
1009 * Only to be called by the page allocator on a freshly allocated
1012 static inline void set_page_pfmemalloc(struct page *page)
1017 static inline void clear_page_pfmemalloc(struct page *page)
1023 * Different kinds of faults, as returned by handle_mm_fault().
1024 * Used to decide whether a process gets delivered SIGBUS or
1025 * just gets major/minor fault counters bumped up.
1028 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1030 #define VM_FAULT_OOM 0x0001
1031 #define VM_FAULT_SIGBUS 0x0002
1032 #define VM_FAULT_MAJOR 0x0004
1033 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1034 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1035 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1036 #define VM_FAULT_SIGSEGV 0x0040
1038 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1039 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1040 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1041 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1043 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1045 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1046 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1049 /* Encode hstate index for a hwpoisoned large page */
1050 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1051 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1054 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1056 extern void pagefault_out_of_memory(void);
1058 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1061 * Flags passed to show_mem() and show_free_areas() to suppress output in
1064 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1066 extern void show_free_areas(unsigned int flags);
1067 extern bool skip_free_areas_node(unsigned int flags, int nid);
1069 int shmem_zero_setup(struct vm_area_struct *);
1071 bool shmem_mapping(struct address_space *mapping);
1073 static inline bool shmem_mapping(struct address_space *mapping)
1079 extern int can_do_mlock(void);
1080 extern int user_shm_lock(size_t, struct user_struct *);
1081 extern void user_shm_unlock(size_t, struct user_struct *);
1084 * Parameter block passed down to zap_pte_range in exceptional cases.
1086 struct zap_details {
1087 struct address_space *check_mapping; /* Check page->mapping if set */
1088 pgoff_t first_index; /* Lowest page->index to unmap */
1089 pgoff_t last_index; /* Highest page->index to unmap */
1092 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1095 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1096 unsigned long size);
1097 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1098 unsigned long size, struct zap_details *);
1099 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1100 unsigned long start, unsigned long end);
1103 * mm_walk - callbacks for walk_page_range
1104 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1105 * this handler is required to be able to handle
1106 * pmd_trans_huge() pmds. They may simply choose to
1107 * split_huge_page() instead of handling it explicitly.
1108 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1109 * @pte_hole: if set, called for each hole at all levels
1110 * @hugetlb_entry: if set, called for each hugetlb entry
1111 * @test_walk: caller specific callback function to determine whether
1112 * we walk over the current vma or not. A positive returned
1113 * value means "do page table walk over the current vma,"
1114 * and a negative one means "abort current page table walk
1115 * right now." 0 means "skip the current vma."
1116 * @mm: mm_struct representing the target process of page table walk
1117 * @vma: vma currently walked (NULL if walking outside vmas)
1118 * @private: private data for callbacks' usage
1120 * (see the comment on walk_page_range() for more details)
1123 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1124 unsigned long next, struct mm_walk *walk);
1125 int (*pte_entry)(pte_t *pte, unsigned long addr,
1126 unsigned long next, struct mm_walk *walk);
1127 int (*pte_hole)(unsigned long addr, unsigned long next,
1128 struct mm_walk *walk);
1129 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1130 unsigned long addr, unsigned long next,
1131 struct mm_walk *walk);
1132 int (*test_walk)(unsigned long addr, unsigned long next,
1133 struct mm_walk *walk);
1134 struct mm_struct *mm;
1135 struct vm_area_struct *vma;
1139 int walk_page_range(unsigned long addr, unsigned long end,
1140 struct mm_walk *walk);
1141 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1142 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1143 unsigned long end, unsigned long floor, unsigned long ceiling);
1144 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1145 struct vm_area_struct *vma);
1146 void unmap_mapping_range(struct address_space *mapping,
1147 loff_t const holebegin, loff_t const holelen, int even_cows);
1148 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1149 unsigned long *pfn);
1150 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1151 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1152 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1153 void *buf, int len, int write);
1155 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1156 loff_t const holebegin, loff_t const holelen)
1158 unmap_mapping_range(mapping, holebegin, holelen, 0);
1161 extern void truncate_pagecache(struct inode *inode, loff_t new);
1162 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1163 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1164 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1165 int truncate_inode_page(struct address_space *mapping, struct page *page);
1166 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1167 int invalidate_inode_page(struct page *page);
1170 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1171 unsigned long address, unsigned int flags);
1172 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1173 unsigned long address, unsigned int fault_flags);
1175 static inline int handle_mm_fault(struct mm_struct *mm,
1176 struct vm_area_struct *vma, unsigned long address,
1179 /* should never happen if there's no MMU */
1181 return VM_FAULT_SIGBUS;
1183 static inline int fixup_user_fault(struct task_struct *tsk,
1184 struct mm_struct *mm, unsigned long address,
1185 unsigned int fault_flags)
1187 /* should never happen if there's no MMU */
1193 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1194 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1195 void *buf, int len, int write);
1197 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1198 unsigned long start, unsigned long nr_pages,
1199 unsigned int foll_flags, struct page **pages,
1200 struct vm_area_struct **vmas, int *nonblocking);
1201 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1202 unsigned long start, unsigned long nr_pages,
1203 int write, int force, struct page **pages,
1204 struct vm_area_struct **vmas);
1205 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1206 unsigned long start, unsigned long nr_pages,
1207 int write, int force, struct page **pages,
1209 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1210 unsigned long start, unsigned long nr_pages,
1211 int write, int force, struct page **pages,
1212 unsigned int gup_flags);
1213 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1214 unsigned long start, unsigned long nr_pages,
1215 int write, int force, struct page **pages);
1216 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1217 struct page **pages);
1219 /* Container for pinned pfns / pages */
1220 struct frame_vector {
1221 unsigned int nr_allocated; /* Number of frames we have space for */
1222 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1223 bool got_ref; /* Did we pin pages by getting page ref? */
1224 bool is_pfns; /* Does array contain pages or pfns? */
1225 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1226 * pfns_vector_pages() or pfns_vector_pfns()
1230 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1231 void frame_vector_destroy(struct frame_vector *vec);
1232 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1233 bool write, bool force, struct frame_vector *vec);
1234 void put_vaddr_frames(struct frame_vector *vec);
1235 int frame_vector_to_pages(struct frame_vector *vec);
1236 void frame_vector_to_pfns(struct frame_vector *vec);
1238 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1240 return vec->nr_frames;
1243 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1246 int err = frame_vector_to_pages(vec);
1249 return ERR_PTR(err);
1251 return (struct page **)(vec->ptrs);
1254 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1257 frame_vector_to_pfns(vec);
1258 return (unsigned long *)(vec->ptrs);
1262 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1263 struct page **pages);
1264 int get_kernel_page(unsigned long start, int write, struct page **pages);
1265 struct page *get_dump_page(unsigned long addr);
1267 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1268 extern void do_invalidatepage(struct page *page, unsigned int offset,
1269 unsigned int length);
1271 int __set_page_dirty_nobuffers(struct page *page);
1272 int __set_page_dirty_no_writeback(struct page *page);
1273 int redirty_page_for_writepage(struct writeback_control *wbc,
1275 void account_page_dirtied(struct page *page, struct address_space *mapping,
1276 struct mem_cgroup *memcg);
1277 void account_page_cleaned(struct page *page, struct address_space *mapping,
1278 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1279 int set_page_dirty(struct page *page);
1280 int set_page_dirty_lock(struct page *page);
1281 void cancel_dirty_page(struct page *page);
1282 int clear_page_dirty_for_io(struct page *page);
1284 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1286 /* Is the vma a continuation of the stack vma above it? */
1287 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1289 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1292 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1294 return !vma->vm_ops;
1297 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1300 return (vma->vm_flags & VM_GROWSDOWN) &&
1301 (vma->vm_start == addr) &&
1302 !vma_growsdown(vma->vm_prev, addr);
1305 /* Is the vma a continuation of the stack vma below it? */
1306 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1308 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1311 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1314 return (vma->vm_flags & VM_GROWSUP) &&
1315 (vma->vm_end == addr) &&
1316 !vma_growsup(vma->vm_next, addr);
1319 extern struct task_struct *task_of_stack(struct task_struct *task,
1320 struct vm_area_struct *vma, bool in_group);
1322 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1323 unsigned long old_addr, struct vm_area_struct *new_vma,
1324 unsigned long new_addr, unsigned long len,
1325 bool need_rmap_locks);
1326 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1327 unsigned long end, pgprot_t newprot,
1328 int dirty_accountable, int prot_numa);
1329 extern int mprotect_fixup(struct vm_area_struct *vma,
1330 struct vm_area_struct **pprev, unsigned long start,
1331 unsigned long end, unsigned long newflags);
1334 * doesn't attempt to fault and will return short.
1336 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1337 struct page **pages);
1339 * per-process(per-mm_struct) statistics.
1341 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1343 long val = atomic_long_read(&mm->rss_stat.count[member]);
1345 #ifdef SPLIT_RSS_COUNTING
1347 * counter is updated in asynchronous manner and may go to minus.
1348 * But it's never be expected number for users.
1353 return (unsigned long)val;
1356 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1358 atomic_long_add(value, &mm->rss_stat.count[member]);
1361 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1363 atomic_long_inc(&mm->rss_stat.count[member]);
1366 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1368 atomic_long_dec(&mm->rss_stat.count[member]);
1371 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1373 return get_mm_counter(mm, MM_FILEPAGES) +
1374 get_mm_counter(mm, MM_ANONPAGES);
1377 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1379 return max(mm->hiwater_rss, get_mm_rss(mm));
1382 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1384 return max(mm->hiwater_vm, mm->total_vm);
1387 static inline void update_hiwater_rss(struct mm_struct *mm)
1389 unsigned long _rss = get_mm_rss(mm);
1391 if ((mm)->hiwater_rss < _rss)
1392 (mm)->hiwater_rss = _rss;
1395 static inline void update_hiwater_vm(struct mm_struct *mm)
1397 if (mm->hiwater_vm < mm->total_vm)
1398 mm->hiwater_vm = mm->total_vm;
1401 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1403 mm->hiwater_rss = get_mm_rss(mm);
1406 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1407 struct mm_struct *mm)
1409 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1411 if (*maxrss < hiwater_rss)
1412 *maxrss = hiwater_rss;
1415 #if defined(SPLIT_RSS_COUNTING)
1416 void sync_mm_rss(struct mm_struct *mm);
1418 static inline void sync_mm_rss(struct mm_struct *mm)
1423 int vma_wants_writenotify(struct vm_area_struct *vma);
1425 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1427 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1431 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1435 #ifdef __PAGETABLE_PUD_FOLDED
1436 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1437 unsigned long address)
1442 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1445 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1446 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1447 unsigned long address)
1452 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1454 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1459 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1460 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1463 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1465 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1467 atomic_long_set(&mm->nr_pmds, 0);
1470 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1472 return atomic_long_read(&mm->nr_pmds);
1475 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1477 atomic_long_inc(&mm->nr_pmds);
1480 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1482 atomic_long_dec(&mm->nr_pmds);
1486 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1487 pmd_t *pmd, unsigned long address);
1488 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1491 * The following ifdef needed to get the 4level-fixup.h header to work.
1492 * Remove it when 4level-fixup.h has been removed.
1494 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1495 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1497 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1498 NULL: pud_offset(pgd, address);
1501 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1503 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1504 NULL: pmd_offset(pud, address);
1506 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1508 #if USE_SPLIT_PTE_PTLOCKS
1509 #if ALLOC_SPLIT_PTLOCKS
1510 void __init ptlock_cache_init(void);
1511 extern bool ptlock_alloc(struct page *page);
1512 extern void ptlock_free(struct page *page);
1514 static inline spinlock_t *ptlock_ptr(struct page *page)
1518 #else /* ALLOC_SPLIT_PTLOCKS */
1519 static inline void ptlock_cache_init(void)
1523 static inline bool ptlock_alloc(struct page *page)
1528 static inline void ptlock_free(struct page *page)
1532 static inline spinlock_t *ptlock_ptr(struct page *page)
1536 #endif /* ALLOC_SPLIT_PTLOCKS */
1538 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1540 return ptlock_ptr(pmd_page(*pmd));
1543 static inline bool ptlock_init(struct page *page)
1546 * prep_new_page() initialize page->private (and therefore page->ptl)
1547 * with 0. Make sure nobody took it in use in between.
1549 * It can happen if arch try to use slab for page table allocation:
1550 * slab code uses page->slab_cache and page->first_page (for tail
1551 * pages), which share storage with page->ptl.
1553 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1554 if (!ptlock_alloc(page))
1556 spin_lock_init(ptlock_ptr(page));
1560 /* Reset page->mapping so free_pages_check won't complain. */
1561 static inline void pte_lock_deinit(struct page *page)
1563 page->mapping = NULL;
1567 #else /* !USE_SPLIT_PTE_PTLOCKS */
1569 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1571 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1573 return &mm->page_table_lock;
1575 static inline void ptlock_cache_init(void) {}
1576 static inline bool ptlock_init(struct page *page) { return true; }
1577 static inline void pte_lock_deinit(struct page *page) {}
1578 #endif /* USE_SPLIT_PTE_PTLOCKS */
1580 static inline void pgtable_init(void)
1582 ptlock_cache_init();
1583 pgtable_cache_init();
1586 static inline bool pgtable_page_ctor(struct page *page)
1588 inc_zone_page_state(page, NR_PAGETABLE);
1589 return ptlock_init(page);
1592 static inline void pgtable_page_dtor(struct page *page)
1594 pte_lock_deinit(page);
1595 dec_zone_page_state(page, NR_PAGETABLE);
1598 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1600 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1601 pte_t *__pte = pte_offset_map(pmd, address); \
1607 #define pte_unmap_unlock(pte, ptl) do { \
1612 #define pte_alloc_map(mm, vma, pmd, address) \
1613 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1615 NULL: pte_offset_map(pmd, address))
1617 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1618 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1620 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1622 #define pte_alloc_kernel(pmd, address) \
1623 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1624 NULL: pte_offset_kernel(pmd, address))
1626 #if USE_SPLIT_PMD_PTLOCKS
1628 static struct page *pmd_to_page(pmd_t *pmd)
1630 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1631 return virt_to_page((void *)((unsigned long) pmd & mask));
1634 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1636 return ptlock_ptr(pmd_to_page(pmd));
1639 static inline bool pgtable_pmd_page_ctor(struct page *page)
1641 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1642 page->pmd_huge_pte = NULL;
1644 return ptlock_init(page);
1647 static inline void pgtable_pmd_page_dtor(struct page *page)
1649 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1650 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1655 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1659 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1661 return &mm->page_table_lock;
1664 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1665 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1667 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1671 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1673 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1678 extern void free_area_init(unsigned long * zones_size);
1679 extern void free_area_init_node(int nid, unsigned long * zones_size,
1680 unsigned long zone_start_pfn, unsigned long *zholes_size);
1681 extern void free_initmem(void);
1684 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1685 * into the buddy system. The freed pages will be poisoned with pattern
1686 * "poison" if it's within range [0, UCHAR_MAX].
1687 * Return pages freed into the buddy system.
1689 extern unsigned long free_reserved_area(void *start, void *end,
1690 int poison, char *s);
1692 #ifdef CONFIG_HIGHMEM
1694 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1695 * and totalram_pages.
1697 extern void free_highmem_page(struct page *page);
1700 extern void adjust_managed_page_count(struct page *page, long count);
1701 extern void mem_init_print_info(const char *str);
1703 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1705 /* Free the reserved page into the buddy system, so it gets managed. */
1706 static inline void __free_reserved_page(struct page *page)
1708 ClearPageReserved(page);
1709 init_page_count(page);
1713 static inline void free_reserved_page(struct page *page)
1715 __free_reserved_page(page);
1716 adjust_managed_page_count(page, 1);
1719 static inline void mark_page_reserved(struct page *page)
1721 SetPageReserved(page);
1722 adjust_managed_page_count(page, -1);
1726 * Default method to free all the __init memory into the buddy system.
1727 * The freed pages will be poisoned with pattern "poison" if it's within
1728 * range [0, UCHAR_MAX].
1729 * Return pages freed into the buddy system.
1731 static inline unsigned long free_initmem_default(int poison)
1733 extern char __init_begin[], __init_end[];
1735 return free_reserved_area(&__init_begin, &__init_end,
1736 poison, "unused kernel");
1739 static inline unsigned long get_num_physpages(void)
1742 unsigned long phys_pages = 0;
1744 for_each_online_node(nid)
1745 phys_pages += node_present_pages(nid);
1750 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1752 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1753 * zones, allocate the backing mem_map and account for memory holes in a more
1754 * architecture independent manner. This is a substitute for creating the
1755 * zone_sizes[] and zholes_size[] arrays and passing them to
1756 * free_area_init_node()
1758 * An architecture is expected to register range of page frames backed by
1759 * physical memory with memblock_add[_node]() before calling
1760 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1761 * usage, an architecture is expected to do something like
1763 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1765 * for_each_valid_physical_page_range()
1766 * memblock_add_node(base, size, nid)
1767 * free_area_init_nodes(max_zone_pfns);
1769 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1770 * registered physical page range. Similarly
1771 * sparse_memory_present_with_active_regions() calls memory_present() for
1772 * each range when SPARSEMEM is enabled.
1774 * See mm/page_alloc.c for more information on each function exposed by
1775 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1777 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1778 unsigned long node_map_pfn_alignment(void);
1779 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1780 unsigned long end_pfn);
1781 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1782 unsigned long end_pfn);
1783 extern void get_pfn_range_for_nid(unsigned int nid,
1784 unsigned long *start_pfn, unsigned long *end_pfn);
1785 extern unsigned long find_min_pfn_with_active_regions(void);
1786 extern void free_bootmem_with_active_regions(int nid,
1787 unsigned long max_low_pfn);
1788 extern void sparse_memory_present_with_active_regions(int nid);
1790 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1792 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1793 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1794 static inline int __early_pfn_to_nid(unsigned long pfn,
1795 struct mminit_pfnnid_cache *state)
1800 /* please see mm/page_alloc.c */
1801 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1802 /* there is a per-arch backend function. */
1803 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1804 struct mminit_pfnnid_cache *state);
1807 extern void set_dma_reserve(unsigned long new_dma_reserve);
1808 extern void memmap_init_zone(unsigned long, int, unsigned long,
1809 unsigned long, enum memmap_context);
1810 extern void setup_per_zone_wmarks(void);
1811 extern int __meminit init_per_zone_wmark_min(void);
1812 extern void mem_init(void);
1813 extern void __init mmap_init(void);
1814 extern void show_mem(unsigned int flags);
1815 extern void si_meminfo(struct sysinfo * val);
1816 extern void si_meminfo_node(struct sysinfo *val, int nid);
1818 extern __printf(3, 4)
1819 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1821 extern void setup_per_cpu_pageset(void);
1823 extern void zone_pcp_update(struct zone *zone);
1824 extern void zone_pcp_reset(struct zone *zone);
1827 extern int min_free_kbytes;
1830 extern atomic_long_t mmap_pages_allocated;
1831 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1833 /* interval_tree.c */
1834 void vma_interval_tree_insert(struct vm_area_struct *node,
1835 struct rb_root *root);
1836 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1837 struct vm_area_struct *prev,
1838 struct rb_root *root);
1839 void vma_interval_tree_remove(struct vm_area_struct *node,
1840 struct rb_root *root);
1841 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1842 unsigned long start, unsigned long last);
1843 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1844 unsigned long start, unsigned long last);
1846 #define vma_interval_tree_foreach(vma, root, start, last) \
1847 for (vma = vma_interval_tree_iter_first(root, start, last); \
1848 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1850 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1851 struct rb_root *root);
1852 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1853 struct rb_root *root);
1854 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1855 struct rb_root *root, unsigned long start, unsigned long last);
1856 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1857 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1858 #ifdef CONFIG_DEBUG_VM_RB
1859 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1862 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1863 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1864 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1867 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1868 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1869 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1870 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1871 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1872 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1873 struct mempolicy *, struct vm_userfaultfd_ctx);
1874 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1875 extern int split_vma(struct mm_struct *,
1876 struct vm_area_struct *, unsigned long addr, int new_below);
1877 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1878 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1879 struct rb_node **, struct rb_node *);
1880 extern void unlink_file_vma(struct vm_area_struct *);
1881 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1882 unsigned long addr, unsigned long len, pgoff_t pgoff,
1883 bool *need_rmap_locks);
1884 extern void exit_mmap(struct mm_struct *);
1886 static inline int check_data_rlimit(unsigned long rlim,
1888 unsigned long start,
1889 unsigned long end_data,
1890 unsigned long start_data)
1892 if (rlim < RLIM_INFINITY) {
1893 if (((new - start) + (end_data - start_data)) > rlim)
1900 extern int mm_take_all_locks(struct mm_struct *mm);
1901 extern void mm_drop_all_locks(struct mm_struct *mm);
1903 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1904 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1906 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1907 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1908 unsigned long addr, unsigned long len,
1909 unsigned long flags,
1910 const struct vm_special_mapping *spec);
1911 /* This is an obsolete alternative to _install_special_mapping. */
1912 extern int install_special_mapping(struct mm_struct *mm,
1913 unsigned long addr, unsigned long len,
1914 unsigned long flags, struct page **pages);
1916 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1918 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1919 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1920 extern unsigned long do_mmap(struct file *file, unsigned long addr,
1921 unsigned long len, unsigned long prot, unsigned long flags,
1922 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
1923 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1925 static inline unsigned long
1926 do_mmap_pgoff(struct file *file, unsigned long addr,
1927 unsigned long len, unsigned long prot, unsigned long flags,
1928 unsigned long pgoff, unsigned long *populate)
1930 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
1934 extern int __mm_populate(unsigned long addr, unsigned long len,
1936 static inline void mm_populate(unsigned long addr, unsigned long len)
1939 (void) __mm_populate(addr, len, 1);
1942 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1945 /* These take the mm semaphore themselves */
1946 extern unsigned long vm_brk(unsigned long, unsigned long);
1947 extern int vm_munmap(unsigned long, size_t);
1948 extern unsigned long vm_mmap(struct file *, unsigned long,
1949 unsigned long, unsigned long,
1950 unsigned long, unsigned long);
1952 struct vm_unmapped_area_info {
1953 #define VM_UNMAPPED_AREA_TOPDOWN 1
1954 unsigned long flags;
1955 unsigned long length;
1956 unsigned long low_limit;
1957 unsigned long high_limit;
1958 unsigned long align_mask;
1959 unsigned long align_offset;
1962 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1963 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1966 * Search for an unmapped address range.
1968 * We are looking for a range that:
1969 * - does not intersect with any VMA;
1970 * - is contained within the [low_limit, high_limit) interval;
1971 * - is at least the desired size.
1972 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1974 static inline unsigned long
1975 vm_unmapped_area(struct vm_unmapped_area_info *info)
1977 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
1978 return unmapped_area_topdown(info);
1980 return unmapped_area(info);
1984 extern void truncate_inode_pages(struct address_space *, loff_t);
1985 extern void truncate_inode_pages_range(struct address_space *,
1986 loff_t lstart, loff_t lend);
1987 extern void truncate_inode_pages_final(struct address_space *);
1989 /* generic vm_area_ops exported for stackable file systems */
1990 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1991 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1992 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1994 /* mm/page-writeback.c */
1995 int write_one_page(struct page *page, int wait);
1996 void task_dirty_inc(struct task_struct *tsk);
1999 #define VM_MAX_READAHEAD 128 /* kbytes */
2000 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2002 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2003 pgoff_t offset, unsigned long nr_to_read);
2005 void page_cache_sync_readahead(struct address_space *mapping,
2006 struct file_ra_state *ra,
2009 unsigned long size);
2011 void page_cache_async_readahead(struct address_space *mapping,
2012 struct file_ra_state *ra,
2016 unsigned long size);
2018 unsigned long max_sane_readahead(unsigned long nr);
2020 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2021 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2023 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2024 extern int expand_downwards(struct vm_area_struct *vma,
2025 unsigned long address);
2027 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2029 #define expand_upwards(vma, address) (0)
2032 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2033 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2034 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2035 struct vm_area_struct **pprev);
2037 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2038 NULL if none. Assume start_addr < end_addr. */
2039 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2041 struct vm_area_struct * vma = find_vma(mm,start_addr);
2043 if (vma && end_addr <= vma->vm_start)
2048 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2050 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2053 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2054 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2055 unsigned long vm_start, unsigned long vm_end)
2057 struct vm_area_struct *vma = find_vma(mm, vm_start);
2059 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2066 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2067 void vma_set_page_prot(struct vm_area_struct *vma);
2069 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2073 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2075 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2079 #ifdef CONFIG_NUMA_BALANCING
2080 unsigned long change_prot_numa(struct vm_area_struct *vma,
2081 unsigned long start, unsigned long end);
2084 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2085 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2086 unsigned long pfn, unsigned long size, pgprot_t);
2087 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2088 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2090 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2092 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2095 struct page *follow_page_mask(struct vm_area_struct *vma,
2096 unsigned long address, unsigned int foll_flags,
2097 unsigned int *page_mask);
2099 static inline struct page *follow_page(struct vm_area_struct *vma,
2100 unsigned long address, unsigned int foll_flags)
2102 unsigned int unused_page_mask;
2103 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2106 #define FOLL_WRITE 0x01 /* check pte is writable */
2107 #define FOLL_TOUCH 0x02 /* mark page accessed */
2108 #define FOLL_GET 0x04 /* do get_page on page */
2109 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2110 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2111 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2112 * and return without waiting upon it */
2113 #define FOLL_POPULATE 0x40 /* fault in page */
2114 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2115 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2116 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2117 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2118 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2120 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2122 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2123 unsigned long size, pte_fn_t fn, void *data);
2125 #ifdef CONFIG_PROC_FS
2126 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2128 static inline void vm_stat_account(struct mm_struct *mm,
2129 unsigned long flags, struct file *file, long pages)
2131 mm->total_vm += pages;
2133 #endif /* CONFIG_PROC_FS */
2135 #ifdef CONFIG_DEBUG_PAGEALLOC
2136 extern bool _debug_pagealloc_enabled;
2137 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2139 static inline bool debug_pagealloc_enabled(void)
2141 return _debug_pagealloc_enabled;
2145 kernel_map_pages(struct page *page, int numpages, int enable)
2147 if (!debug_pagealloc_enabled())
2150 __kernel_map_pages(page, numpages, enable);
2152 #ifdef CONFIG_HIBERNATION
2153 extern bool kernel_page_present(struct page *page);
2154 #endif /* CONFIG_HIBERNATION */
2157 kernel_map_pages(struct page *page, int numpages, int enable) {}
2158 #ifdef CONFIG_HIBERNATION
2159 static inline bool kernel_page_present(struct page *page) { return true; }
2160 #endif /* CONFIG_HIBERNATION */
2163 #ifdef __HAVE_ARCH_GATE_AREA
2164 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2165 extern int in_gate_area_no_mm(unsigned long addr);
2166 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2168 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2172 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2173 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2177 #endif /* __HAVE_ARCH_GATE_AREA */
2179 #ifdef CONFIG_SYSCTL
2180 extern int sysctl_drop_caches;
2181 int drop_caches_sysctl_handler(struct ctl_table *, int,
2182 void __user *, size_t *, loff_t *);
2185 void drop_slab(void);
2186 void drop_slab_node(int nid);
2189 #define randomize_va_space 0
2191 extern int randomize_va_space;
2194 const char * arch_vma_name(struct vm_area_struct *vma);
2195 void print_vma_addr(char *prefix, unsigned long rip);
2197 void sparse_mem_maps_populate_node(struct page **map_map,
2198 unsigned long pnum_begin,
2199 unsigned long pnum_end,
2200 unsigned long map_count,
2203 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2204 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2205 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2206 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2207 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2208 void *vmemmap_alloc_block(unsigned long size, int node);
2209 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2210 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2211 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2213 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2214 void vmemmap_populate_print_last(void);
2215 #ifdef CONFIG_MEMORY_HOTPLUG
2216 void vmemmap_free(unsigned long start, unsigned long end);
2218 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2219 unsigned long size);
2222 MF_COUNT_INCREASED = 1 << 0,
2223 MF_ACTION_REQUIRED = 1 << 1,
2224 MF_MUST_KILL = 1 << 2,
2225 MF_SOFT_OFFLINE = 1 << 3,
2227 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2228 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2229 extern int unpoison_memory(unsigned long pfn);
2230 extern int get_hwpoison_page(struct page *page);
2231 extern void put_hwpoison_page(struct page *page);
2232 extern int sysctl_memory_failure_early_kill;
2233 extern int sysctl_memory_failure_recovery;
2234 extern void shake_page(struct page *p, int access);
2235 extern atomic_long_t num_poisoned_pages;
2236 extern int soft_offline_page(struct page *page, int flags);
2240 * Error handlers for various types of pages.
2243 MF_IGNORED, /* Error: cannot be handled */
2244 MF_FAILED, /* Error: handling failed */
2245 MF_DELAYED, /* Will be handled later */
2246 MF_RECOVERED, /* Successfully recovered */
2249 enum mf_action_page_type {
2251 MF_MSG_KERNEL_HIGH_ORDER,
2253 MF_MSG_DIFFERENT_COMPOUND,
2254 MF_MSG_POISONED_HUGE,
2257 MF_MSG_UNMAP_FAILED,
2258 MF_MSG_DIRTY_SWAPCACHE,
2259 MF_MSG_CLEAN_SWAPCACHE,
2260 MF_MSG_DIRTY_MLOCKED_LRU,
2261 MF_MSG_CLEAN_MLOCKED_LRU,
2262 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2263 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2266 MF_MSG_TRUNCATED_LRU,
2272 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2273 extern void clear_huge_page(struct page *page,
2275 unsigned int pages_per_huge_page);
2276 extern void copy_user_huge_page(struct page *dst, struct page *src,
2277 unsigned long addr, struct vm_area_struct *vma,
2278 unsigned int pages_per_huge_page);
2279 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2281 extern struct page_ext_operations debug_guardpage_ops;
2282 extern struct page_ext_operations page_poisoning_ops;
2284 #ifdef CONFIG_DEBUG_PAGEALLOC
2285 extern unsigned int _debug_guardpage_minorder;
2286 extern bool _debug_guardpage_enabled;
2288 static inline unsigned int debug_guardpage_minorder(void)
2290 return _debug_guardpage_minorder;
2293 static inline bool debug_guardpage_enabled(void)
2295 return _debug_guardpage_enabled;
2298 static inline bool page_is_guard(struct page *page)
2300 struct page_ext *page_ext;
2302 if (!debug_guardpage_enabled())
2305 page_ext = lookup_page_ext(page);
2306 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2309 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2310 static inline bool debug_guardpage_enabled(void) { return false; }
2311 static inline bool page_is_guard(struct page *page) { return false; }
2312 #endif /* CONFIG_DEBUG_PAGEALLOC */
2314 #if MAX_NUMNODES > 1
2315 void __init setup_nr_node_ids(void);
2317 static inline void setup_nr_node_ids(void) {}
2320 #endif /* __KERNEL__ */
2321 #endif /* _LINUX_MM_H */