4 #include <linux/errno.h>
10 #include <linux/list.h>
11 #include <linux/mmzone.h>
12 #include <linux/rbtree.h>
13 #include <linux/atomic.h>
14 #include <linux/debug_locks.h>
15 #include <linux/mm_types.h>
16 #include <linux/range.h>
17 #include <linux/pfn.h>
18 #include <linux/bit_spinlock.h>
19 #include <linux/shrinker.h>
23 struct anon_vma_chain;
26 struct writeback_control;
28 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
29 extern unsigned long max_mapnr;
31 static inline void set_max_mapnr(unsigned long limit)
36 static inline void set_max_mapnr(unsigned long limit) { }
39 extern unsigned long totalram_pages;
40 extern void * high_memory;
41 extern int page_cluster;
44 extern int sysctl_legacy_va_layout;
46 #define sysctl_legacy_va_layout 0
50 #include <asm/pgtable.h>
51 #include <asm/processor.h>
54 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
57 extern unsigned long sysctl_user_reserve_kbytes;
58 extern unsigned long sysctl_admin_reserve_kbytes;
60 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
62 /* to align the pointer to the (next) page boundary */
63 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
65 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
66 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
69 * Linux kernel virtual memory manager primitives.
70 * The idea being to have a "virtual" mm in the same way
71 * we have a virtual fs - giving a cleaner interface to the
72 * mm details, and allowing different kinds of memory mappings
73 * (from shared memory to executable loading to arbitrary
77 extern struct kmem_cache *vm_area_cachep;
80 extern struct rb_root nommu_region_tree;
81 extern struct rw_semaphore nommu_region_sem;
83 extern unsigned int kobjsize(const void *objp);
87 * vm_flags in vm_area_struct, see mm_types.h.
89 #define VM_NONE 0x00000000
91 #define VM_READ 0x00000001 /* currently active flags */
92 #define VM_WRITE 0x00000002
93 #define VM_EXEC 0x00000004
94 #define VM_SHARED 0x00000008
96 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
97 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
98 #define VM_MAYWRITE 0x00000020
99 #define VM_MAYEXEC 0x00000040
100 #define VM_MAYSHARE 0x00000080
102 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
103 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
104 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
106 #define VM_LOCKED 0x00002000
107 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
109 /* Used by sys_madvise() */
110 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
111 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
113 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
114 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
115 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
116 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
117 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
118 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
119 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
120 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
122 #ifdef CONFIG_MEM_SOFT_DIRTY
123 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
125 # define VM_SOFTDIRTY 0
128 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
129 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
130 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
131 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
133 #if defined(CONFIG_X86)
134 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
135 #elif defined(CONFIG_PPC)
136 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
137 #elif defined(CONFIG_PARISC)
138 # define VM_GROWSUP VM_ARCH_1
139 #elif defined(CONFIG_METAG)
140 # define VM_GROWSUP VM_ARCH_1
141 #elif defined(CONFIG_IA64)
142 # define VM_GROWSUP VM_ARCH_1
143 #elif !defined(CONFIG_MMU)
144 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
148 # define VM_GROWSUP VM_NONE
151 /* Bits set in the VMA until the stack is in its final location */
152 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
154 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
155 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
158 #ifdef CONFIG_STACK_GROWSUP
159 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
161 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
165 * Special vmas that are non-mergable, non-mlock()able.
166 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
168 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP)
171 * mapping from the currently active vm_flags protection bits (the
172 * low four bits) to a page protection mask..
174 extern pgprot_t protection_map[16];
176 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
177 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
178 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
179 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
180 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
181 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
182 #define FAULT_FLAG_TRIED 0x40 /* second try */
183 #define FAULT_FLAG_USER 0x80 /* The fault originated in userspace */
186 * vm_fault is filled by the the pagefault handler and passed to the vma's
187 * ->fault function. The vma's ->fault is responsible for returning a bitmask
188 * of VM_FAULT_xxx flags that give details about how the fault was handled.
190 * pgoff should be used in favour of virtual_address, if possible. If pgoff
191 * is used, one may implement ->remap_pages to get nonlinear mapping support.
194 unsigned int flags; /* FAULT_FLAG_xxx flags */
195 pgoff_t pgoff; /* Logical page offset based on vma */
196 void __user *virtual_address; /* Faulting virtual address */
198 struct page *page; /* ->fault handlers should return a
199 * page here, unless VM_FAULT_NOPAGE
200 * is set (which is also implied by
206 * These are the virtual MM functions - opening of an area, closing and
207 * unmapping it (needed to keep files on disk up-to-date etc), pointer
208 * to the functions called when a no-page or a wp-page exception occurs.
210 struct vm_operations_struct {
211 void (*open)(struct vm_area_struct * area);
212 void (*close)(struct vm_area_struct * area);
213 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
215 /* notification that a previously read-only page is about to become
216 * writable, if an error is returned it will cause a SIGBUS */
217 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
219 /* called by access_process_vm when get_user_pages() fails, typically
220 * for use by special VMAs that can switch between memory and hardware
222 int (*access)(struct vm_area_struct *vma, unsigned long addr,
223 void *buf, int len, int write);
226 * set_policy() op must add a reference to any non-NULL @new mempolicy
227 * to hold the policy upon return. Caller should pass NULL @new to
228 * remove a policy and fall back to surrounding context--i.e. do not
229 * install a MPOL_DEFAULT policy, nor the task or system default
232 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
235 * get_policy() op must add reference [mpol_get()] to any policy at
236 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
237 * in mm/mempolicy.c will do this automatically.
238 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
239 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
240 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
241 * must return NULL--i.e., do not "fallback" to task or system default
244 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
246 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
247 const nodemask_t *to, unsigned long flags);
249 /* called by sys_remap_file_pages() to populate non-linear mapping */
250 int (*remap_pages)(struct vm_area_struct *vma, unsigned long addr,
251 unsigned long size, pgoff_t pgoff);
257 #define page_private(page) ((page)->private)
258 #define set_page_private(page, v) ((page)->private = (v))
260 /* It's valid only if the page is free path or free_list */
261 static inline void set_freepage_migratetype(struct page *page, int migratetype)
263 page->index = migratetype;
266 /* It's valid only if the page is free path or free_list */
267 static inline int get_freepage_migratetype(struct page *page)
273 * FIXME: take this include out, include page-flags.h in
274 * files which need it (119 of them)
276 #include <linux/page-flags.h>
277 #include <linux/huge_mm.h>
280 * Methods to modify the page usage count.
282 * What counts for a page usage:
283 * - cache mapping (page->mapping)
284 * - private data (page->private)
285 * - page mapped in a task's page tables, each mapping
286 * is counted separately
288 * Also, many kernel routines increase the page count before a critical
289 * routine so they can be sure the page doesn't go away from under them.
293 * Drop a ref, return true if the refcount fell to zero (the page has no users)
295 static inline int put_page_testzero(struct page *page)
297 VM_BUG_ON(atomic_read(&page->_count) == 0);
298 return atomic_dec_and_test(&page->_count);
302 * Try to grab a ref unless the page has a refcount of zero, return false if
304 * This can be called when MMU is off so it must not access
305 * any of the virtual mappings.
307 static inline int get_page_unless_zero(struct page *page)
309 return atomic_inc_not_zero(&page->_count);
313 * Try to drop a ref unless the page has a refcount of one, return false if
315 * This is to make sure that the refcount won't become zero after this drop.
316 * This can be called when MMU is off so it must not access
317 * any of the virtual mappings.
319 static inline int put_page_unless_one(struct page *page)
321 return atomic_add_unless(&page->_count, -1, 1);
324 extern int page_is_ram(unsigned long pfn);
326 /* Support for virtually mapped pages */
327 struct page *vmalloc_to_page(const void *addr);
328 unsigned long vmalloc_to_pfn(const void *addr);
331 * Determine if an address is within the vmalloc range
333 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
334 * is no special casing required.
336 static inline int is_vmalloc_addr(const void *x)
339 unsigned long addr = (unsigned long)x;
341 return addr >= VMALLOC_START && addr < VMALLOC_END;
347 extern int is_vmalloc_or_module_addr(const void *x);
349 static inline int is_vmalloc_or_module_addr(const void *x)
355 static inline void compound_lock(struct page *page)
357 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
358 VM_BUG_ON(PageSlab(page));
359 bit_spin_lock(PG_compound_lock, &page->flags);
363 static inline void compound_unlock(struct page *page)
365 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
366 VM_BUG_ON(PageSlab(page));
367 bit_spin_unlock(PG_compound_lock, &page->flags);
371 static inline unsigned long compound_lock_irqsave(struct page *page)
373 unsigned long uninitialized_var(flags);
374 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
375 local_irq_save(flags);
381 static inline void compound_unlock_irqrestore(struct page *page,
384 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
385 compound_unlock(page);
386 local_irq_restore(flags);
390 static inline struct page *compound_head(struct page *page)
392 if (unlikely(PageTail(page)))
393 return page->first_page;
398 * The atomic page->_mapcount, starts from -1: so that transitions
399 * both from it and to it can be tracked, using atomic_inc_and_test
400 * and atomic_add_negative(-1).
402 static inline void page_mapcount_reset(struct page *page)
404 atomic_set(&(page)->_mapcount, -1);
407 static inline int page_mapcount(struct page *page)
409 return atomic_read(&(page)->_mapcount) + 1;
412 static inline int page_count(struct page *page)
414 return atomic_read(&compound_head(page)->_count);
417 static inline void get_huge_page_tail(struct page *page)
420 * __split_huge_page_refcount() cannot run
423 VM_BUG_ON(page_mapcount(page) < 0);
424 VM_BUG_ON(atomic_read(&page->_count) != 0);
425 atomic_inc(&page->_mapcount);
428 extern bool __get_page_tail(struct page *page);
430 static inline void get_page(struct page *page)
432 if (unlikely(PageTail(page)))
433 if (likely(__get_page_tail(page)))
436 * Getting a normal page or the head of a compound page
437 * requires to already have an elevated page->_count.
439 VM_BUG_ON(atomic_read(&page->_count) <= 0);
440 atomic_inc(&page->_count);
443 static inline struct page *virt_to_head_page(const void *x)
445 struct page *page = virt_to_page(x);
446 return compound_head(page);
450 * Setup the page count before being freed into the page allocator for
451 * the first time (boot or memory hotplug)
453 static inline void init_page_count(struct page *page)
455 atomic_set(&page->_count, 1);
459 * PageBuddy() indicate that the page is free and in the buddy system
460 * (see mm/page_alloc.c).
462 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
463 * -2 so that an underflow of the page_mapcount() won't be mistaken
464 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
465 * efficiently by most CPU architectures.
467 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
469 static inline int PageBuddy(struct page *page)
471 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
474 static inline void __SetPageBuddy(struct page *page)
476 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
477 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
480 static inline void __ClearPageBuddy(struct page *page)
482 VM_BUG_ON(!PageBuddy(page));
483 atomic_set(&page->_mapcount, -1);
486 void put_page(struct page *page);
487 void put_pages_list(struct list_head *pages);
489 void split_page(struct page *page, unsigned int order);
490 int split_free_page(struct page *page);
493 * Compound pages have a destructor function. Provide a
494 * prototype for that function and accessor functions.
495 * These are _only_ valid on the head of a PG_compound page.
497 typedef void compound_page_dtor(struct page *);
499 static inline void set_compound_page_dtor(struct page *page,
500 compound_page_dtor *dtor)
502 page[1].lru.next = (void *)dtor;
505 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
507 return (compound_page_dtor *)page[1].lru.next;
510 static inline int compound_order(struct page *page)
514 return (unsigned long)page[1].lru.prev;
517 static inline void set_compound_order(struct page *page, unsigned long order)
519 page[1].lru.prev = (void *)order;
524 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
525 * servicing faults for write access. In the normal case, do always want
526 * pte_mkwrite. But get_user_pages can cause write faults for mappings
527 * that do not have writing enabled, when used by access_process_vm.
529 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
531 if (likely(vma->vm_flags & VM_WRITE))
532 pte = pte_mkwrite(pte);
538 * Multiple processes may "see" the same page. E.g. for untouched
539 * mappings of /dev/null, all processes see the same page full of
540 * zeroes, and text pages of executables and shared libraries have
541 * only one copy in memory, at most, normally.
543 * For the non-reserved pages, page_count(page) denotes a reference count.
544 * page_count() == 0 means the page is free. page->lru is then used for
545 * freelist management in the buddy allocator.
546 * page_count() > 0 means the page has been allocated.
548 * Pages are allocated by the slab allocator in order to provide memory
549 * to kmalloc and kmem_cache_alloc. In this case, the management of the
550 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
551 * unless a particular usage is carefully commented. (the responsibility of
552 * freeing the kmalloc memory is the caller's, of course).
554 * A page may be used by anyone else who does a __get_free_page().
555 * In this case, page_count still tracks the references, and should only
556 * be used through the normal accessor functions. The top bits of page->flags
557 * and page->virtual store page management information, but all other fields
558 * are unused and could be used privately, carefully. The management of this
559 * page is the responsibility of the one who allocated it, and those who have
560 * subsequently been given references to it.
562 * The other pages (we may call them "pagecache pages") are completely
563 * managed by the Linux memory manager: I/O, buffers, swapping etc.
564 * The following discussion applies only to them.
566 * A pagecache page contains an opaque `private' member, which belongs to the
567 * page's address_space. Usually, this is the address of a circular list of
568 * the page's disk buffers. PG_private must be set to tell the VM to call
569 * into the filesystem to release these pages.
571 * A page may belong to an inode's memory mapping. In this case, page->mapping
572 * is the pointer to the inode, and page->index is the file offset of the page,
573 * in units of PAGE_CACHE_SIZE.
575 * If pagecache pages are not associated with an inode, they are said to be
576 * anonymous pages. These may become associated with the swapcache, and in that
577 * case PG_swapcache is set, and page->private is an offset into the swapcache.
579 * In either case (swapcache or inode backed), the pagecache itself holds one
580 * reference to the page. Setting PG_private should also increment the
581 * refcount. The each user mapping also has a reference to the page.
583 * The pagecache pages are stored in a per-mapping radix tree, which is
584 * rooted at mapping->page_tree, and indexed by offset.
585 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
586 * lists, we instead now tag pages as dirty/writeback in the radix tree.
588 * All pagecache pages may be subject to I/O:
589 * - inode pages may need to be read from disk,
590 * - inode pages which have been modified and are MAP_SHARED may need
591 * to be written back to the inode on disk,
592 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
593 * modified may need to be swapped out to swap space and (later) to be read
598 * The zone field is never updated after free_area_init_core()
599 * sets it, so none of the operations on it need to be atomic.
602 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
603 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
604 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
605 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
606 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
609 * Define the bit shifts to access each section. For non-existent
610 * sections we define the shift as 0; that plus a 0 mask ensures
611 * the compiler will optimise away reference to them.
613 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
614 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
615 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
616 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
618 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
619 #ifdef NODE_NOT_IN_PAGE_FLAGS
620 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
621 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
622 SECTIONS_PGOFF : ZONES_PGOFF)
624 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
625 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
626 NODES_PGOFF : ZONES_PGOFF)
629 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
631 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
632 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
635 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
636 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
637 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
638 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_WIDTH) - 1)
639 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
641 static inline enum zone_type page_zonenum(const struct page *page)
643 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
646 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
647 #define SECTION_IN_PAGE_FLAGS
651 * The identification function is mainly used by the buddy allocator for
652 * determining if two pages could be buddies. We are not really identifying
653 * the zone since we could be using the section number id if we do not have
654 * node id available in page flags.
655 * We only guarantee that it will return the same value for two combinable
658 static inline int page_zone_id(struct page *page)
660 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
663 static inline int zone_to_nid(struct zone *zone)
672 #ifdef NODE_NOT_IN_PAGE_FLAGS
673 extern int page_to_nid(const struct page *page);
675 static inline int page_to_nid(const struct page *page)
677 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
681 #ifdef CONFIG_NUMA_BALANCING
682 static inline int cpu_pid_to_cpupid(int cpu, int pid)
684 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
687 static inline int cpupid_to_pid(int cpupid)
689 return cpupid & LAST__PID_MASK;
692 static inline int cpupid_to_cpu(int cpupid)
694 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
697 static inline int cpupid_to_nid(int cpupid)
699 return cpu_to_node(cpupid_to_cpu(cpupid));
702 static inline bool cpupid_pid_unset(int cpupid)
704 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
707 static inline bool cpupid_cpu_unset(int cpupid)
709 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
712 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
714 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
717 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
718 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
719 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
721 return xchg(&page->_last_cpupid, cpupid);
724 static inline int page_cpupid_last(struct page *page)
726 return page->_last_cpupid;
728 static inline void page_cpupid_reset_last(struct page *page)
730 page->_last_cpupid = -1;
733 static inline int page_cpupid_last(struct page *page)
735 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
738 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
740 static inline void page_cpupid_reset_last(struct page *page)
742 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
744 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
745 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
747 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
748 #else /* !CONFIG_NUMA_BALANCING */
749 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
751 return page_to_nid(page); /* XXX */
754 static inline int page_cpupid_last(struct page *page)
756 return page_to_nid(page); /* XXX */
759 static inline int cpupid_to_nid(int cpupid)
764 static inline int cpupid_to_pid(int cpupid)
769 static inline int cpupid_to_cpu(int cpupid)
774 static inline int cpu_pid_to_cpupid(int nid, int pid)
779 static inline bool cpupid_pid_unset(int cpupid)
784 static inline void page_cpupid_reset_last(struct page *page)
788 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
792 #endif /* CONFIG_NUMA_BALANCING */
794 static inline struct zone *page_zone(const struct page *page)
796 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
799 #ifdef SECTION_IN_PAGE_FLAGS
800 static inline void set_page_section(struct page *page, unsigned long section)
802 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
803 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
806 static inline unsigned long page_to_section(const struct page *page)
808 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
812 static inline void set_page_zone(struct page *page, enum zone_type zone)
814 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
815 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
818 static inline void set_page_node(struct page *page, unsigned long node)
820 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
821 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
824 static inline void set_page_links(struct page *page, enum zone_type zone,
825 unsigned long node, unsigned long pfn)
827 set_page_zone(page, zone);
828 set_page_node(page, node);
829 #ifdef SECTION_IN_PAGE_FLAGS
830 set_page_section(page, pfn_to_section_nr(pfn));
835 * Some inline functions in vmstat.h depend on page_zone()
837 #include <linux/vmstat.h>
839 static __always_inline void *lowmem_page_address(const struct page *page)
841 return __va(PFN_PHYS(page_to_pfn(page)));
844 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
845 #define HASHED_PAGE_VIRTUAL
848 #if defined(WANT_PAGE_VIRTUAL)
849 #define page_address(page) ((page)->virtual)
850 #define set_page_address(page, address) \
852 (page)->virtual = (address); \
854 #define page_address_init() do { } while(0)
857 #if defined(HASHED_PAGE_VIRTUAL)
858 void *page_address(const struct page *page);
859 void set_page_address(struct page *page, void *virtual);
860 void page_address_init(void);
863 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
864 #define page_address(page) lowmem_page_address(page)
865 #define set_page_address(page, address) do { } while(0)
866 #define page_address_init() do { } while(0)
870 * On an anonymous page mapped into a user virtual memory area,
871 * page->mapping points to its anon_vma, not to a struct address_space;
872 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
874 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
875 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
876 * and then page->mapping points, not to an anon_vma, but to a private
877 * structure which KSM associates with that merged page. See ksm.h.
879 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
881 * Please note that, confusingly, "page_mapping" refers to the inode
882 * address_space which maps the page from disk; whereas "page_mapped"
883 * refers to user virtual address space into which the page is mapped.
885 #define PAGE_MAPPING_ANON 1
886 #define PAGE_MAPPING_KSM 2
887 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
889 extern struct address_space *page_mapping(struct page *page);
891 /* Neutral page->mapping pointer to address_space or anon_vma or other */
892 static inline void *page_rmapping(struct page *page)
894 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
897 extern struct address_space *__page_file_mapping(struct page *);
900 struct address_space *page_file_mapping(struct page *page)
902 if (unlikely(PageSwapCache(page)))
903 return __page_file_mapping(page);
905 return page->mapping;
908 static inline int PageAnon(struct page *page)
910 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
914 * Return the pagecache index of the passed page. Regular pagecache pages
915 * use ->index whereas swapcache pages use ->private
917 static inline pgoff_t page_index(struct page *page)
919 if (unlikely(PageSwapCache(page)))
920 return page_private(page);
924 extern pgoff_t __page_file_index(struct page *page);
927 * Return the file index of the page. Regular pagecache pages use ->index
928 * whereas swapcache pages use swp_offset(->private)
930 static inline pgoff_t page_file_index(struct page *page)
932 if (unlikely(PageSwapCache(page)))
933 return __page_file_index(page);
939 * Return true if this page is mapped into pagetables.
941 static inline int page_mapped(struct page *page)
943 return atomic_read(&(page)->_mapcount) >= 0;
947 * Different kinds of faults, as returned by handle_mm_fault().
948 * Used to decide whether a process gets delivered SIGBUS or
949 * just gets major/minor fault counters bumped up.
952 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
954 #define VM_FAULT_OOM 0x0001
955 #define VM_FAULT_SIGBUS 0x0002
956 #define VM_FAULT_MAJOR 0x0004
957 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
958 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
959 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
961 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
962 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
963 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
964 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
966 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
968 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
969 VM_FAULT_FALLBACK | VM_FAULT_HWPOISON_LARGE)
971 /* Encode hstate index for a hwpoisoned large page */
972 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
973 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
976 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
978 extern void pagefault_out_of_memory(void);
980 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
983 * Flags passed to show_mem() and show_free_areas() to suppress output in
986 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
987 #define SHOW_MEM_FILTER_PAGE_COUNT (0x0002u) /* page type count */
989 extern void show_free_areas(unsigned int flags);
990 extern bool skip_free_areas_node(unsigned int flags, int nid);
992 int shmem_zero_setup(struct vm_area_struct *);
994 extern int can_do_mlock(void);
995 extern int user_shm_lock(size_t, struct user_struct *);
996 extern void user_shm_unlock(size_t, struct user_struct *);
999 * Parameter block passed down to zap_pte_range in exceptional cases.
1001 struct zap_details {
1002 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
1003 struct address_space *check_mapping; /* Check page->mapping if set */
1004 pgoff_t first_index; /* Lowest page->index to unmap */
1005 pgoff_t last_index; /* Highest page->index to unmap */
1008 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1011 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1012 unsigned long size);
1013 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1014 unsigned long size, struct zap_details *);
1015 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1016 unsigned long start, unsigned long end);
1019 * mm_walk - callbacks for walk_page_range
1020 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
1021 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1022 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1023 * this handler is required to be able to handle
1024 * pmd_trans_huge() pmds. They may simply choose to
1025 * split_huge_page() instead of handling it explicitly.
1026 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1027 * @pte_hole: if set, called for each hole at all levels
1028 * @hugetlb_entry: if set, called for each hugetlb entry
1029 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
1032 * (see walk_page_range for more details)
1035 int (*pgd_entry)(pgd_t *pgd, unsigned long addr,
1036 unsigned long next, struct mm_walk *walk);
1037 int (*pud_entry)(pud_t *pud, unsigned long addr,
1038 unsigned long next, struct mm_walk *walk);
1039 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1040 unsigned long next, struct mm_walk *walk);
1041 int (*pte_entry)(pte_t *pte, unsigned long addr,
1042 unsigned long next, struct mm_walk *walk);
1043 int (*pte_hole)(unsigned long addr, unsigned long next,
1044 struct mm_walk *walk);
1045 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1046 unsigned long addr, unsigned long next,
1047 struct mm_walk *walk);
1048 struct mm_struct *mm;
1052 int walk_page_range(unsigned long addr, unsigned long end,
1053 struct mm_walk *walk);
1054 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1055 unsigned long end, unsigned long floor, unsigned long ceiling);
1056 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1057 struct vm_area_struct *vma);
1058 void unmap_mapping_range(struct address_space *mapping,
1059 loff_t const holebegin, loff_t const holelen, int even_cows);
1060 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1061 unsigned long *pfn);
1062 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1063 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1064 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1065 void *buf, int len, int write);
1067 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1068 loff_t const holebegin, loff_t const holelen)
1070 unmap_mapping_range(mapping, holebegin, holelen, 0);
1073 extern void truncate_pagecache(struct inode *inode, loff_t new);
1074 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1075 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1076 int truncate_inode_page(struct address_space *mapping, struct page *page);
1077 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1078 int invalidate_inode_page(struct page *page);
1081 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1082 unsigned long address, unsigned int flags);
1083 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1084 unsigned long address, unsigned int fault_flags);
1086 static inline int handle_mm_fault(struct mm_struct *mm,
1087 struct vm_area_struct *vma, unsigned long address,
1090 /* should never happen if there's no MMU */
1092 return VM_FAULT_SIGBUS;
1094 static inline int fixup_user_fault(struct task_struct *tsk,
1095 struct mm_struct *mm, unsigned long address,
1096 unsigned int fault_flags)
1098 /* should never happen if there's no MMU */
1104 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1105 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1106 void *buf, int len, int write);
1108 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1109 unsigned long start, unsigned long nr_pages,
1110 unsigned int foll_flags, struct page **pages,
1111 struct vm_area_struct **vmas, int *nonblocking);
1112 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1113 unsigned long start, unsigned long nr_pages,
1114 int write, int force, struct page **pages,
1115 struct vm_area_struct **vmas);
1116 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1117 struct page **pages);
1119 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1120 struct page **pages);
1121 int get_kernel_page(unsigned long start, int write, struct page **pages);
1122 struct page *get_dump_page(unsigned long addr);
1124 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1125 extern void do_invalidatepage(struct page *page, unsigned int offset,
1126 unsigned int length);
1128 int __set_page_dirty_nobuffers(struct page *page);
1129 int __set_page_dirty_no_writeback(struct page *page);
1130 int redirty_page_for_writepage(struct writeback_control *wbc,
1132 void account_page_dirtied(struct page *page, struct address_space *mapping);
1133 void account_page_writeback(struct page *page);
1134 int set_page_dirty(struct page *page);
1135 int set_page_dirty_lock(struct page *page);
1136 int clear_page_dirty_for_io(struct page *page);
1138 /* Is the vma a continuation of the stack vma above it? */
1139 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1141 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1144 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1147 return (vma->vm_flags & VM_GROWSDOWN) &&
1148 (vma->vm_start == addr) &&
1149 !vma_growsdown(vma->vm_prev, addr);
1152 /* Is the vma a continuation of the stack vma below it? */
1153 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1155 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1158 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1161 return (vma->vm_flags & VM_GROWSUP) &&
1162 (vma->vm_end == addr) &&
1163 !vma_growsup(vma->vm_next, addr);
1167 vm_is_stack(struct task_struct *task, struct vm_area_struct *vma, int in_group);
1169 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1170 unsigned long old_addr, struct vm_area_struct *new_vma,
1171 unsigned long new_addr, unsigned long len,
1172 bool need_rmap_locks);
1173 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1174 unsigned long end, pgprot_t newprot,
1175 int dirty_accountable, int prot_numa);
1176 extern int mprotect_fixup(struct vm_area_struct *vma,
1177 struct vm_area_struct **pprev, unsigned long start,
1178 unsigned long end, unsigned long newflags);
1181 * doesn't attempt to fault and will return short.
1183 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1184 struct page **pages);
1186 * per-process(per-mm_struct) statistics.
1188 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1190 long val = atomic_long_read(&mm->rss_stat.count[member]);
1192 #ifdef SPLIT_RSS_COUNTING
1194 * counter is updated in asynchronous manner and may go to minus.
1195 * But it's never be expected number for users.
1200 return (unsigned long)val;
1203 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1205 atomic_long_add(value, &mm->rss_stat.count[member]);
1208 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1210 atomic_long_inc(&mm->rss_stat.count[member]);
1213 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1215 atomic_long_dec(&mm->rss_stat.count[member]);
1218 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1220 return get_mm_counter(mm, MM_FILEPAGES) +
1221 get_mm_counter(mm, MM_ANONPAGES);
1224 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1226 return max(mm->hiwater_rss, get_mm_rss(mm));
1229 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1231 return max(mm->hiwater_vm, mm->total_vm);
1234 static inline void update_hiwater_rss(struct mm_struct *mm)
1236 unsigned long _rss = get_mm_rss(mm);
1238 if ((mm)->hiwater_rss < _rss)
1239 (mm)->hiwater_rss = _rss;
1242 static inline void update_hiwater_vm(struct mm_struct *mm)
1244 if (mm->hiwater_vm < mm->total_vm)
1245 mm->hiwater_vm = mm->total_vm;
1248 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1249 struct mm_struct *mm)
1251 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1253 if (*maxrss < hiwater_rss)
1254 *maxrss = hiwater_rss;
1257 #if defined(SPLIT_RSS_COUNTING)
1258 void sync_mm_rss(struct mm_struct *mm);
1260 static inline void sync_mm_rss(struct mm_struct *mm)
1265 int vma_wants_writenotify(struct vm_area_struct *vma);
1267 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1269 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1273 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1277 #ifdef __PAGETABLE_PUD_FOLDED
1278 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1279 unsigned long address)
1284 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1287 #ifdef __PAGETABLE_PMD_FOLDED
1288 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1289 unsigned long address)
1294 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1297 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1298 pmd_t *pmd, unsigned long address);
1299 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1302 * The following ifdef needed to get the 4level-fixup.h header to work.
1303 * Remove it when 4level-fixup.h has been removed.
1305 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1306 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1308 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1309 NULL: pud_offset(pgd, address);
1312 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1314 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1315 NULL: pmd_offset(pud, address);
1317 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1319 #if USE_SPLIT_PTE_PTLOCKS
1320 #if ALLOC_SPLIT_PTLOCKS
1321 extern bool ptlock_alloc(struct page *page);
1322 extern void ptlock_free(struct page *page);
1324 static inline spinlock_t *ptlock_ptr(struct page *page)
1328 #else /* ALLOC_SPLIT_PTLOCKS */
1329 static inline bool ptlock_alloc(struct page *page)
1334 static inline void ptlock_free(struct page *page)
1338 static inline spinlock_t *ptlock_ptr(struct page *page)
1342 #endif /* ALLOC_SPLIT_PTLOCKS */
1344 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1346 return ptlock_ptr(pmd_page(*pmd));
1349 static inline bool ptlock_init(struct page *page)
1352 * prep_new_page() initialize page->private (and therefore page->ptl)
1353 * with 0. Make sure nobody took it in use in between.
1355 * It can happen if arch try to use slab for page table allocation:
1356 * slab code uses page->slab_cache and page->first_page (for tail
1357 * pages), which share storage with page->ptl.
1359 VM_BUG_ON(*(unsigned long *)&page->ptl);
1360 if (!ptlock_alloc(page))
1362 spin_lock_init(ptlock_ptr(page));
1366 /* Reset page->mapping so free_pages_check won't complain. */
1367 static inline void pte_lock_deinit(struct page *page)
1369 page->mapping = NULL;
1373 #else /* !USE_SPLIT_PTE_PTLOCKS */
1375 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1377 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1379 return &mm->page_table_lock;
1381 static inline bool ptlock_init(struct page *page) { return true; }
1382 static inline void pte_lock_deinit(struct page *page) {}
1383 #endif /* USE_SPLIT_PTE_PTLOCKS */
1385 static inline bool pgtable_page_ctor(struct page *page)
1387 inc_zone_page_state(page, NR_PAGETABLE);
1388 return ptlock_init(page);
1391 static inline void pgtable_page_dtor(struct page *page)
1393 pte_lock_deinit(page);
1394 dec_zone_page_state(page, NR_PAGETABLE);
1397 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1399 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1400 pte_t *__pte = pte_offset_map(pmd, address); \
1406 #define pte_unmap_unlock(pte, ptl) do { \
1411 #define pte_alloc_map(mm, vma, pmd, address) \
1412 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1414 NULL: pte_offset_map(pmd, address))
1416 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1417 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1419 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1421 #define pte_alloc_kernel(pmd, address) \
1422 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1423 NULL: pte_offset_kernel(pmd, address))
1425 #if USE_SPLIT_PMD_PTLOCKS
1427 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1429 return ptlock_ptr(virt_to_page(pmd));
1432 static inline bool pgtable_pmd_page_ctor(struct page *page)
1434 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1435 page->pmd_huge_pte = NULL;
1437 return ptlock_init(page);
1440 static inline void pgtable_pmd_page_dtor(struct page *page)
1442 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1443 VM_BUG_ON(page->pmd_huge_pte);
1448 #define pmd_huge_pte(mm, pmd) (virt_to_page(pmd)->pmd_huge_pte)
1452 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1454 return &mm->page_table_lock;
1457 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1458 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1460 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1464 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1466 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1471 extern void free_area_init(unsigned long * zones_size);
1472 extern void free_area_init_node(int nid, unsigned long * zones_size,
1473 unsigned long zone_start_pfn, unsigned long *zholes_size);
1474 extern void free_initmem(void);
1477 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1478 * into the buddy system. The freed pages will be poisoned with pattern
1479 * "poison" if it's within range [0, UCHAR_MAX].
1480 * Return pages freed into the buddy system.
1482 extern unsigned long free_reserved_area(void *start, void *end,
1483 int poison, char *s);
1485 #ifdef CONFIG_HIGHMEM
1487 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1488 * and totalram_pages.
1490 extern void free_highmem_page(struct page *page);
1493 extern void adjust_managed_page_count(struct page *page, long count);
1494 extern void mem_init_print_info(const char *str);
1496 /* Free the reserved page into the buddy system, so it gets managed. */
1497 static inline void __free_reserved_page(struct page *page)
1499 ClearPageReserved(page);
1500 init_page_count(page);
1504 static inline void free_reserved_page(struct page *page)
1506 __free_reserved_page(page);
1507 adjust_managed_page_count(page, 1);
1510 static inline void mark_page_reserved(struct page *page)
1512 SetPageReserved(page);
1513 adjust_managed_page_count(page, -1);
1517 * Default method to free all the __init memory into the buddy system.
1518 * The freed pages will be poisoned with pattern "poison" if it's within
1519 * range [0, UCHAR_MAX].
1520 * Return pages freed into the buddy system.
1522 static inline unsigned long free_initmem_default(int poison)
1524 extern char __init_begin[], __init_end[];
1526 return free_reserved_area(&__init_begin, &__init_end,
1527 poison, "unused kernel");
1530 static inline unsigned long get_num_physpages(void)
1533 unsigned long phys_pages = 0;
1535 for_each_online_node(nid)
1536 phys_pages += node_present_pages(nid);
1541 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1543 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1544 * zones, allocate the backing mem_map and account for memory holes in a more
1545 * architecture independent manner. This is a substitute for creating the
1546 * zone_sizes[] and zholes_size[] arrays and passing them to
1547 * free_area_init_node()
1549 * An architecture is expected to register range of page frames backed by
1550 * physical memory with memblock_add[_node]() before calling
1551 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1552 * usage, an architecture is expected to do something like
1554 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1556 * for_each_valid_physical_page_range()
1557 * memblock_add_node(base, size, nid)
1558 * free_area_init_nodes(max_zone_pfns);
1560 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1561 * registered physical page range. Similarly
1562 * sparse_memory_present_with_active_regions() calls memory_present() for
1563 * each range when SPARSEMEM is enabled.
1565 * See mm/page_alloc.c for more information on each function exposed by
1566 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1568 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1569 unsigned long node_map_pfn_alignment(void);
1570 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1571 unsigned long end_pfn);
1572 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1573 unsigned long end_pfn);
1574 extern void get_pfn_range_for_nid(unsigned int nid,
1575 unsigned long *start_pfn, unsigned long *end_pfn);
1576 extern unsigned long find_min_pfn_with_active_regions(void);
1577 extern void free_bootmem_with_active_regions(int nid,
1578 unsigned long max_low_pfn);
1579 extern void sparse_memory_present_with_active_regions(int nid);
1581 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1583 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1584 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1585 static inline int __early_pfn_to_nid(unsigned long pfn)
1590 /* please see mm/page_alloc.c */
1591 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1592 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1593 /* there is a per-arch backend function. */
1594 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1595 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1598 extern void set_dma_reserve(unsigned long new_dma_reserve);
1599 extern void memmap_init_zone(unsigned long, int, unsigned long,
1600 unsigned long, enum memmap_context);
1601 extern void setup_per_zone_wmarks(void);
1602 extern int __meminit init_per_zone_wmark_min(void);
1603 extern void mem_init(void);
1604 extern void __init mmap_init(void);
1605 extern void show_mem(unsigned int flags);
1606 extern void si_meminfo(struct sysinfo * val);
1607 extern void si_meminfo_node(struct sysinfo *val, int nid);
1609 extern __printf(3, 4)
1610 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1612 extern void setup_per_cpu_pageset(void);
1614 extern void zone_pcp_update(struct zone *zone);
1615 extern void zone_pcp_reset(struct zone *zone);
1618 extern int min_free_kbytes;
1621 extern atomic_long_t mmap_pages_allocated;
1622 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1624 /* interval_tree.c */
1625 void vma_interval_tree_insert(struct vm_area_struct *node,
1626 struct rb_root *root);
1627 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1628 struct vm_area_struct *prev,
1629 struct rb_root *root);
1630 void vma_interval_tree_remove(struct vm_area_struct *node,
1631 struct rb_root *root);
1632 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1633 unsigned long start, unsigned long last);
1634 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1635 unsigned long start, unsigned long last);
1637 #define vma_interval_tree_foreach(vma, root, start, last) \
1638 for (vma = vma_interval_tree_iter_first(root, start, last); \
1639 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1641 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1642 struct list_head *list)
1644 list_add_tail(&vma->shared.nonlinear, list);
1647 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1648 struct rb_root *root);
1649 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1650 struct rb_root *root);
1651 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1652 struct rb_root *root, unsigned long start, unsigned long last);
1653 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1654 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1655 #ifdef CONFIG_DEBUG_VM_RB
1656 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1659 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1660 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1661 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1664 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1665 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1666 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1667 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1668 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1669 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1670 struct mempolicy *);
1671 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1672 extern int split_vma(struct mm_struct *,
1673 struct vm_area_struct *, unsigned long addr, int new_below);
1674 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1675 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1676 struct rb_node **, struct rb_node *);
1677 extern void unlink_file_vma(struct vm_area_struct *);
1678 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1679 unsigned long addr, unsigned long len, pgoff_t pgoff,
1680 bool *need_rmap_locks);
1681 extern void exit_mmap(struct mm_struct *);
1683 extern int mm_take_all_locks(struct mm_struct *mm);
1684 extern void mm_drop_all_locks(struct mm_struct *mm);
1686 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1687 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1689 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1690 extern int install_special_mapping(struct mm_struct *mm,
1691 unsigned long addr, unsigned long len,
1692 unsigned long flags, struct page **pages);
1694 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1696 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1697 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1698 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1699 unsigned long len, unsigned long prot, unsigned long flags,
1700 unsigned long pgoff, unsigned long *populate);
1701 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1704 extern int __mm_populate(unsigned long addr, unsigned long len,
1706 static inline void mm_populate(unsigned long addr, unsigned long len)
1709 (void) __mm_populate(addr, len, 1);
1712 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1715 /* These take the mm semaphore themselves */
1716 extern unsigned long vm_brk(unsigned long, unsigned long);
1717 extern int vm_munmap(unsigned long, size_t);
1718 extern unsigned long vm_mmap(struct file *, unsigned long,
1719 unsigned long, unsigned long,
1720 unsigned long, unsigned long);
1722 struct vm_unmapped_area_info {
1723 #define VM_UNMAPPED_AREA_TOPDOWN 1
1724 unsigned long flags;
1725 unsigned long length;
1726 unsigned long low_limit;
1727 unsigned long high_limit;
1728 unsigned long align_mask;
1729 unsigned long align_offset;
1732 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1733 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1736 * Search for an unmapped address range.
1738 * We are looking for a range that:
1739 * - does not intersect with any VMA;
1740 * - is contained within the [low_limit, high_limit) interval;
1741 * - is at least the desired size.
1742 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1744 static inline unsigned long
1745 vm_unmapped_area(struct vm_unmapped_area_info *info)
1747 if (!(info->flags & VM_UNMAPPED_AREA_TOPDOWN))
1748 return unmapped_area(info);
1750 return unmapped_area_topdown(info);
1754 extern void truncate_inode_pages(struct address_space *, loff_t);
1755 extern void truncate_inode_pages_range(struct address_space *,
1756 loff_t lstart, loff_t lend);
1758 /* generic vm_area_ops exported for stackable file systems */
1759 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1760 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1762 /* mm/page-writeback.c */
1763 int write_one_page(struct page *page, int wait);
1764 void task_dirty_inc(struct task_struct *tsk);
1767 #define VM_MAX_READAHEAD 128 /* kbytes */
1768 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1770 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1771 pgoff_t offset, unsigned long nr_to_read);
1773 void page_cache_sync_readahead(struct address_space *mapping,
1774 struct file_ra_state *ra,
1777 unsigned long size);
1779 void page_cache_async_readahead(struct address_space *mapping,
1780 struct file_ra_state *ra,
1784 unsigned long size);
1786 unsigned long max_sane_readahead(unsigned long nr);
1787 unsigned long ra_submit(struct file_ra_state *ra,
1788 struct address_space *mapping,
1791 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1792 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1794 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1795 extern int expand_downwards(struct vm_area_struct *vma,
1796 unsigned long address);
1798 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1800 #define expand_upwards(vma, address) do { } while (0)
1803 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1804 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1805 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1806 struct vm_area_struct **pprev);
1808 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1809 NULL if none. Assume start_addr < end_addr. */
1810 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1812 struct vm_area_struct * vma = find_vma(mm,start_addr);
1814 if (vma && end_addr <= vma->vm_start)
1819 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1821 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1824 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1825 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
1826 unsigned long vm_start, unsigned long vm_end)
1828 struct vm_area_struct *vma = find_vma(mm, vm_start);
1830 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
1837 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1839 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1845 #ifdef CONFIG_ARCH_USES_NUMA_PROT_NONE
1846 unsigned long change_prot_numa(struct vm_area_struct *vma,
1847 unsigned long start, unsigned long end);
1850 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1851 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1852 unsigned long pfn, unsigned long size, pgprot_t);
1853 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1854 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1856 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1858 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
1861 struct page *follow_page_mask(struct vm_area_struct *vma,
1862 unsigned long address, unsigned int foll_flags,
1863 unsigned int *page_mask);
1865 static inline struct page *follow_page(struct vm_area_struct *vma,
1866 unsigned long address, unsigned int foll_flags)
1868 unsigned int unused_page_mask;
1869 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
1872 #define FOLL_WRITE 0x01 /* check pte is writable */
1873 #define FOLL_TOUCH 0x02 /* mark page accessed */
1874 #define FOLL_GET 0x04 /* do get_page on page */
1875 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1876 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1877 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1878 * and return without waiting upon it */
1879 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1880 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1881 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1882 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
1883 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
1885 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1887 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1888 unsigned long size, pte_fn_t fn, void *data);
1890 #ifdef CONFIG_PROC_FS
1891 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1893 static inline void vm_stat_account(struct mm_struct *mm,
1894 unsigned long flags, struct file *file, long pages)
1896 mm->total_vm += pages;
1898 #endif /* CONFIG_PROC_FS */
1900 #ifdef CONFIG_DEBUG_PAGEALLOC
1901 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1902 #ifdef CONFIG_HIBERNATION
1903 extern bool kernel_page_present(struct page *page);
1904 #endif /* CONFIG_HIBERNATION */
1907 kernel_map_pages(struct page *page, int numpages, int enable) {}
1908 #ifdef CONFIG_HIBERNATION
1909 static inline bool kernel_page_present(struct page *page) { return true; }
1910 #endif /* CONFIG_HIBERNATION */
1913 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1914 #ifdef __HAVE_ARCH_GATE_AREA
1915 int in_gate_area_no_mm(unsigned long addr);
1916 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1918 int in_gate_area_no_mm(unsigned long addr);
1919 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1920 #endif /* __HAVE_ARCH_GATE_AREA */
1922 #ifdef CONFIG_SYSCTL
1923 extern int sysctl_drop_caches;
1924 int drop_caches_sysctl_handler(struct ctl_table *, int,
1925 void __user *, size_t *, loff_t *);
1928 unsigned long shrink_slab(struct shrink_control *shrink,
1929 unsigned long nr_pages_scanned,
1930 unsigned long lru_pages);
1933 #define randomize_va_space 0
1935 extern int randomize_va_space;
1938 const char * arch_vma_name(struct vm_area_struct *vma);
1939 void print_vma_addr(char *prefix, unsigned long rip);
1941 void sparse_mem_maps_populate_node(struct page **map_map,
1942 unsigned long pnum_begin,
1943 unsigned long pnum_end,
1944 unsigned long map_count,
1947 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1948 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1949 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1950 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1951 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1952 void *vmemmap_alloc_block(unsigned long size, int node);
1953 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1954 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1955 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
1957 int vmemmap_populate(unsigned long start, unsigned long end, int node);
1958 void vmemmap_populate_print_last(void);
1959 #ifdef CONFIG_MEMORY_HOTPLUG
1960 void vmemmap_free(unsigned long start, unsigned long end);
1962 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
1963 unsigned long size);
1966 MF_COUNT_INCREASED = 1 << 0,
1967 MF_ACTION_REQUIRED = 1 << 1,
1968 MF_MUST_KILL = 1 << 2,
1969 MF_SOFT_OFFLINE = 1 << 3,
1971 extern int memory_failure(unsigned long pfn, int trapno, int flags);
1972 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1973 extern int unpoison_memory(unsigned long pfn);
1974 extern int sysctl_memory_failure_early_kill;
1975 extern int sysctl_memory_failure_recovery;
1976 extern void shake_page(struct page *p, int access);
1977 extern atomic_long_t num_poisoned_pages;
1978 extern int soft_offline_page(struct page *page, int flags);
1980 extern void dump_page(struct page *page);
1982 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1983 extern void clear_huge_page(struct page *page,
1985 unsigned int pages_per_huge_page);
1986 extern void copy_user_huge_page(struct page *dst, struct page *src,
1987 unsigned long addr, struct vm_area_struct *vma,
1988 unsigned int pages_per_huge_page);
1989 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1991 #ifdef CONFIG_DEBUG_PAGEALLOC
1992 extern unsigned int _debug_guardpage_minorder;
1994 static inline unsigned int debug_guardpage_minorder(void)
1996 return _debug_guardpage_minorder;
1999 static inline bool page_is_guard(struct page *page)
2001 return test_bit(PAGE_DEBUG_FLAG_GUARD, &page->debug_flags);
2004 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2005 static inline bool page_is_guard(struct page *page) { return false; }
2006 #endif /* CONFIG_DEBUG_PAGEALLOC */
2008 #if MAX_NUMNODES > 1
2009 void __init setup_nr_node_ids(void);
2011 static inline void setup_nr_node_ids(void) {}
2014 #endif /* __KERNEL__ */
2015 #endif /* _LINUX_MM_H */