MAINTAINERS: add CHAR and MISC driver maintainers
[zen-stable.git] / include / linux / mm.h
blob3dc3a8c2c4858a1d3400aa2d5fd029d36a1177c6
1 #ifndef _LINUX_MM_H
2 #define _LINUX_MM_H
4 #include <linux/errno.h>
6 #ifdef __KERNEL__
8 #include <linux/gfp.h>
9 #include <linux/list.h>
10 #include <linux/mmzone.h>
11 #include <linux/rbtree.h>
12 #include <linux/prio_tree.h>
13 #include <linux/debug_locks.h>
14 #include <linux/mm_types.h>
15 #include <linux/range.h>
16 #include <linux/pfn.h>
17 #include <linux/bit_spinlock.h>
18 #include <linux/shrinker.h>
20 struct mempolicy;
21 struct anon_vma;
22 struct file_ra_state;
23 struct user_struct;
24 struct writeback_control;
26 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
27 extern unsigned long max_mapnr;
28 #endif
30 extern unsigned long num_physpages;
31 extern unsigned long totalram_pages;
32 extern void * high_memory;
33 extern int page_cluster;
35 #ifdef CONFIG_SYSCTL
36 extern int sysctl_legacy_va_layout;
37 #else
38 #define sysctl_legacy_va_layout 0
39 #endif
41 #include <asm/page.h>
42 #include <asm/pgtable.h>
43 #include <asm/processor.h>
45 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
47 /* to align the pointer to the (next) page boundary */
48 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
51 * Linux kernel virtual memory manager primitives.
52 * The idea being to have a "virtual" mm in the same way
53 * we have a virtual fs - giving a cleaner interface to the
54 * mm details, and allowing different kinds of memory mappings
55 * (from shared memory to executable loading to arbitrary
56 * mmap() functions).
59 extern struct kmem_cache *vm_area_cachep;
61 #ifndef CONFIG_MMU
62 extern struct rb_root nommu_region_tree;
63 extern struct rw_semaphore nommu_region_sem;
65 extern unsigned int kobjsize(const void *objp);
66 #endif
69 * vm_flags in vm_area_struct, see mm_types.h.
71 #define VM_READ 0x00000001 /* currently active flags */
72 #define VM_WRITE 0x00000002
73 #define VM_EXEC 0x00000004
74 #define VM_SHARED 0x00000008
76 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
77 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
78 #define VM_MAYWRITE 0x00000020
79 #define VM_MAYEXEC 0x00000040
80 #define VM_MAYSHARE 0x00000080
82 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
83 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
84 #define VM_GROWSUP 0x00000200
85 #else
86 #define VM_GROWSUP 0x00000000
87 #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
88 #endif
89 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
90 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
92 #define VM_EXECUTABLE 0x00001000
93 #define VM_LOCKED 0x00002000
94 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
96 /* Used by sys_madvise() */
97 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
98 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
100 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
101 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
102 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
103 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
104 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
105 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
106 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
107 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
108 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
109 #else
110 #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
111 #endif
112 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
113 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
115 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
116 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
117 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
118 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
119 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
121 /* Bits set in the VMA until the stack is in its final location */
122 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
124 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
125 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
126 #endif
128 #ifdef CONFIG_STACK_GROWSUP
129 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
130 #else
131 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
132 #endif
134 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
135 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
136 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
137 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
138 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
141 * Special vmas that are non-mergable, non-mlock()able.
142 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
144 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
147 * mapping from the currently active vm_flags protection bits (the
148 * low four bits) to a page protection mask..
150 extern pgprot_t protection_map[16];
152 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
153 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
154 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
155 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
156 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
157 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
160 * This interface is used by x86 PAT code to identify a pfn mapping that is
161 * linear over entire vma. This is to optimize PAT code that deals with
162 * marking the physical region with a particular prot. This is not for generic
163 * mm use. Note also that this check will not work if the pfn mapping is
164 * linear for a vma starting at physical address 0. In which case PAT code
165 * falls back to slow path of reserving physical range page by page.
167 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
169 return !!(vma->vm_flags & VM_PFN_AT_MMAP);
172 static inline int is_pfn_mapping(struct vm_area_struct *vma)
174 return !!(vma->vm_flags & VM_PFNMAP);
178 * vm_fault is filled by the the pagefault handler and passed to the vma's
179 * ->fault function. The vma's ->fault is responsible for returning a bitmask
180 * of VM_FAULT_xxx flags that give details about how the fault was handled.
182 * pgoff should be used in favour of virtual_address, if possible. If pgoff
183 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
184 * mapping support.
186 struct vm_fault {
187 unsigned int flags; /* FAULT_FLAG_xxx flags */
188 pgoff_t pgoff; /* Logical page offset based on vma */
189 void __user *virtual_address; /* Faulting virtual address */
191 struct page *page; /* ->fault handlers should return a
192 * page here, unless VM_FAULT_NOPAGE
193 * is set (which is also implied by
194 * VM_FAULT_ERROR).
199 * These are the virtual MM functions - opening of an area, closing and
200 * unmapping it (needed to keep files on disk up-to-date etc), pointer
201 * to the functions called when a no-page or a wp-page exception occurs.
203 struct vm_operations_struct {
204 void (*open)(struct vm_area_struct * area);
205 void (*close)(struct vm_area_struct * area);
206 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
208 /* notification that a previously read-only page is about to become
209 * writable, if an error is returned it will cause a SIGBUS */
210 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
212 /* called by access_process_vm when get_user_pages() fails, typically
213 * for use by special VMAs that can switch between memory and hardware
215 int (*access)(struct vm_area_struct *vma, unsigned long addr,
216 void *buf, int len, int write);
217 #ifdef CONFIG_NUMA
219 * set_policy() op must add a reference to any non-NULL @new mempolicy
220 * to hold the policy upon return. Caller should pass NULL @new to
221 * remove a policy and fall back to surrounding context--i.e. do not
222 * install a MPOL_DEFAULT policy, nor the task or system default
223 * mempolicy.
225 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
228 * get_policy() op must add reference [mpol_get()] to any policy at
229 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
230 * in mm/mempolicy.c will do this automatically.
231 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
232 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
233 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
234 * must return NULL--i.e., do not "fallback" to task or system default
235 * policy.
237 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
238 unsigned long addr);
239 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
240 const nodemask_t *to, unsigned long flags);
241 #endif
244 struct mmu_gather;
245 struct inode;
247 #define page_private(page) ((page)->private)
248 #define set_page_private(page, v) ((page)->private = (v))
251 * FIXME: take this include out, include page-flags.h in
252 * files which need it (119 of them)
254 #include <linux/page-flags.h>
255 #include <linux/huge_mm.h>
258 * Methods to modify the page usage count.
260 * What counts for a page usage:
261 * - cache mapping (page->mapping)
262 * - private data (page->private)
263 * - page mapped in a task's page tables, each mapping
264 * is counted separately
266 * Also, many kernel routines increase the page count before a critical
267 * routine so they can be sure the page doesn't go away from under them.
271 * Drop a ref, return true if the refcount fell to zero (the page has no users)
273 static inline int put_page_testzero(struct page *page)
275 VM_BUG_ON(atomic_read(&page->_count) == 0);
276 return atomic_dec_and_test(&page->_count);
280 * Try to grab a ref unless the page has a refcount of zero, return false if
281 * that is the case.
283 static inline int get_page_unless_zero(struct page *page)
285 return atomic_inc_not_zero(&page->_count);
288 extern int page_is_ram(unsigned long pfn);
290 /* Support for virtually mapped pages */
291 struct page *vmalloc_to_page(const void *addr);
292 unsigned long vmalloc_to_pfn(const void *addr);
295 * Determine if an address is within the vmalloc range
297 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
298 * is no special casing required.
300 static inline int is_vmalloc_addr(const void *x)
302 #ifdef CONFIG_MMU
303 unsigned long addr = (unsigned long)x;
305 return addr >= VMALLOC_START && addr < VMALLOC_END;
306 #else
307 return 0;
308 #endif
310 #ifdef CONFIG_MMU
311 extern int is_vmalloc_or_module_addr(const void *x);
312 #else
313 static inline int is_vmalloc_or_module_addr(const void *x)
315 return 0;
317 #endif
319 static inline void compound_lock(struct page *page)
321 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
322 bit_spin_lock(PG_compound_lock, &page->flags);
323 #endif
326 static inline void compound_unlock(struct page *page)
328 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
329 bit_spin_unlock(PG_compound_lock, &page->flags);
330 #endif
333 static inline unsigned long compound_lock_irqsave(struct page *page)
335 unsigned long uninitialized_var(flags);
336 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
337 local_irq_save(flags);
338 compound_lock(page);
339 #endif
340 return flags;
343 static inline void compound_unlock_irqrestore(struct page *page,
344 unsigned long flags)
346 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
347 compound_unlock(page);
348 local_irq_restore(flags);
349 #endif
352 static inline struct page *compound_head(struct page *page)
354 if (unlikely(PageTail(page)))
355 return page->first_page;
356 return page;
360 * The atomic page->_mapcount, starts from -1: so that transitions
361 * both from it and to it can be tracked, using atomic_inc_and_test
362 * and atomic_add_negative(-1).
364 static inline void reset_page_mapcount(struct page *page)
366 atomic_set(&(page)->_mapcount, -1);
369 static inline int page_mapcount(struct page *page)
371 return atomic_read(&(page)->_mapcount) + 1;
374 static inline int page_count(struct page *page)
376 return atomic_read(&compound_head(page)->_count);
379 static inline void get_huge_page_tail(struct page *page)
382 * __split_huge_page_refcount() cannot run
383 * from under us.
385 VM_BUG_ON(page_mapcount(page) < 0);
386 VM_BUG_ON(atomic_read(&page->_count) != 0);
387 atomic_inc(&page->_mapcount);
390 extern bool __get_page_tail(struct page *page);
392 static inline void get_page(struct page *page)
394 if (unlikely(PageTail(page)))
395 if (likely(__get_page_tail(page)))
396 return;
398 * Getting a normal page or the head of a compound page
399 * requires to already have an elevated page->_count.
401 VM_BUG_ON(atomic_read(&page->_count) <= 0);
402 atomic_inc(&page->_count);
405 static inline struct page *virt_to_head_page(const void *x)
407 struct page *page = virt_to_page(x);
408 return compound_head(page);
412 * Setup the page count before being freed into the page allocator for
413 * the first time (boot or memory hotplug)
415 static inline void init_page_count(struct page *page)
417 atomic_set(&page->_count, 1);
421 * PageBuddy() indicate that the page is free and in the buddy system
422 * (see mm/page_alloc.c).
424 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
425 * -2 so that an underflow of the page_mapcount() won't be mistaken
426 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
427 * efficiently by most CPU architectures.
429 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
431 static inline int PageBuddy(struct page *page)
433 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
436 static inline void __SetPageBuddy(struct page *page)
438 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
439 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
442 static inline void __ClearPageBuddy(struct page *page)
444 VM_BUG_ON(!PageBuddy(page));
445 atomic_set(&page->_mapcount, -1);
448 void put_page(struct page *page);
449 void put_pages_list(struct list_head *pages);
451 void split_page(struct page *page, unsigned int order);
452 int split_free_page(struct page *page);
455 * Compound pages have a destructor function. Provide a
456 * prototype for that function and accessor functions.
457 * These are _only_ valid on the head of a PG_compound page.
459 typedef void compound_page_dtor(struct page *);
461 static inline void set_compound_page_dtor(struct page *page,
462 compound_page_dtor *dtor)
464 page[1].lru.next = (void *)dtor;
467 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
469 return (compound_page_dtor *)page[1].lru.next;
472 static inline int compound_order(struct page *page)
474 if (!PageHead(page))
475 return 0;
476 return (unsigned long)page[1].lru.prev;
479 static inline int compound_trans_order(struct page *page)
481 int order;
482 unsigned long flags;
484 if (!PageHead(page))
485 return 0;
487 flags = compound_lock_irqsave(page);
488 order = compound_order(page);
489 compound_unlock_irqrestore(page, flags);
490 return order;
493 static inline void set_compound_order(struct page *page, unsigned long order)
495 page[1].lru.prev = (void *)order;
498 #ifdef CONFIG_MMU
500 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
501 * servicing faults for write access. In the normal case, do always want
502 * pte_mkwrite. But get_user_pages can cause write faults for mappings
503 * that do not have writing enabled, when used by access_process_vm.
505 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
507 if (likely(vma->vm_flags & VM_WRITE))
508 pte = pte_mkwrite(pte);
509 return pte;
511 #endif
514 * Multiple processes may "see" the same page. E.g. for untouched
515 * mappings of /dev/null, all processes see the same page full of
516 * zeroes, and text pages of executables and shared libraries have
517 * only one copy in memory, at most, normally.
519 * For the non-reserved pages, page_count(page) denotes a reference count.
520 * page_count() == 0 means the page is free. page->lru is then used for
521 * freelist management in the buddy allocator.
522 * page_count() > 0 means the page has been allocated.
524 * Pages are allocated by the slab allocator in order to provide memory
525 * to kmalloc and kmem_cache_alloc. In this case, the management of the
526 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
527 * unless a particular usage is carefully commented. (the responsibility of
528 * freeing the kmalloc memory is the caller's, of course).
530 * A page may be used by anyone else who does a __get_free_page().
531 * In this case, page_count still tracks the references, and should only
532 * be used through the normal accessor functions. The top bits of page->flags
533 * and page->virtual store page management information, but all other fields
534 * are unused and could be used privately, carefully. The management of this
535 * page is the responsibility of the one who allocated it, and those who have
536 * subsequently been given references to it.
538 * The other pages (we may call them "pagecache pages") are completely
539 * managed by the Linux memory manager: I/O, buffers, swapping etc.
540 * The following discussion applies only to them.
542 * A pagecache page contains an opaque `private' member, which belongs to the
543 * page's address_space. Usually, this is the address of a circular list of
544 * the page's disk buffers. PG_private must be set to tell the VM to call
545 * into the filesystem to release these pages.
547 * A page may belong to an inode's memory mapping. In this case, page->mapping
548 * is the pointer to the inode, and page->index is the file offset of the page,
549 * in units of PAGE_CACHE_SIZE.
551 * If pagecache pages are not associated with an inode, they are said to be
552 * anonymous pages. These may become associated with the swapcache, and in that
553 * case PG_swapcache is set, and page->private is an offset into the swapcache.
555 * In either case (swapcache or inode backed), the pagecache itself holds one
556 * reference to the page. Setting PG_private should also increment the
557 * refcount. The each user mapping also has a reference to the page.
559 * The pagecache pages are stored in a per-mapping radix tree, which is
560 * rooted at mapping->page_tree, and indexed by offset.
561 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
562 * lists, we instead now tag pages as dirty/writeback in the radix tree.
564 * All pagecache pages may be subject to I/O:
565 * - inode pages may need to be read from disk,
566 * - inode pages which have been modified and are MAP_SHARED may need
567 * to be written back to the inode on disk,
568 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
569 * modified may need to be swapped out to swap space and (later) to be read
570 * back into memory.
574 * The zone field is never updated after free_area_init_core()
575 * sets it, so none of the operations on it need to be atomic.
580 * page->flags layout:
582 * There are three possibilities for how page->flags get
583 * laid out. The first is for the normal case, without
584 * sparsemem. The second is for sparsemem when there is
585 * plenty of space for node and section. The last is when
586 * we have run out of space and have to fall back to an
587 * alternate (slower) way of determining the node.
589 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
590 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
591 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
593 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
594 #define SECTIONS_WIDTH SECTIONS_SHIFT
595 #else
596 #define SECTIONS_WIDTH 0
597 #endif
599 #define ZONES_WIDTH ZONES_SHIFT
601 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
602 #define NODES_WIDTH NODES_SHIFT
603 #else
604 #ifdef CONFIG_SPARSEMEM_VMEMMAP
605 #error "Vmemmap: No space for nodes field in page flags"
606 #endif
607 #define NODES_WIDTH 0
608 #endif
610 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
611 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
612 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
613 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
616 * We are going to use the flags for the page to node mapping if its in
617 * there. This includes the case where there is no node, so it is implicit.
619 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
620 #define NODE_NOT_IN_PAGE_FLAGS
621 #endif
624 * Define the bit shifts to access each section. For non-existent
625 * sections we define the shift as 0; that plus a 0 mask ensures
626 * the compiler will optimise away reference to them.
628 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
629 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
630 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
632 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
633 #ifdef NODE_NOT_IN_PAGE_FLAGS
634 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
635 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
636 SECTIONS_PGOFF : ZONES_PGOFF)
637 #else
638 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
639 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
640 NODES_PGOFF : ZONES_PGOFF)
641 #endif
643 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
645 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
646 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
647 #endif
649 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
650 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
651 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
652 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
654 static inline enum zone_type page_zonenum(const struct page *page)
656 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
660 * The identification function is only used by the buddy allocator for
661 * determining if two pages could be buddies. We are not really
662 * identifying a zone since we could be using a the section number
663 * id if we have not node id available in page flags.
664 * We guarantee only that it will return the same value for two
665 * combinable pages in a zone.
667 static inline int page_zone_id(struct page *page)
669 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
672 static inline int zone_to_nid(struct zone *zone)
674 #ifdef CONFIG_NUMA
675 return zone->node;
676 #else
677 return 0;
678 #endif
681 #ifdef NODE_NOT_IN_PAGE_FLAGS
682 extern int page_to_nid(const struct page *page);
683 #else
684 static inline int page_to_nid(const struct page *page)
686 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
688 #endif
690 static inline struct zone *page_zone(const struct page *page)
692 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
695 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
696 static inline void set_page_section(struct page *page, unsigned long section)
698 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
699 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
702 static inline unsigned long page_to_section(const struct page *page)
704 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
706 #endif
708 static inline void set_page_zone(struct page *page, enum zone_type zone)
710 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
711 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
714 static inline void set_page_node(struct page *page, unsigned long node)
716 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
717 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
720 static inline void set_page_links(struct page *page, enum zone_type zone,
721 unsigned long node, unsigned long pfn)
723 set_page_zone(page, zone);
724 set_page_node(page, node);
725 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
726 set_page_section(page, pfn_to_section_nr(pfn));
727 #endif
731 * Some inline functions in vmstat.h depend on page_zone()
733 #include <linux/vmstat.h>
735 static __always_inline void *lowmem_page_address(const struct page *page)
737 return __va(PFN_PHYS(page_to_pfn(page)));
740 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
741 #define HASHED_PAGE_VIRTUAL
742 #endif
744 #if defined(WANT_PAGE_VIRTUAL)
745 #define page_address(page) ((page)->virtual)
746 #define set_page_address(page, address) \
747 do { \
748 (page)->virtual = (address); \
749 } while(0)
750 #define page_address_init() do { } while(0)
751 #endif
753 #if defined(HASHED_PAGE_VIRTUAL)
754 void *page_address(const struct page *page);
755 void set_page_address(struct page *page, void *virtual);
756 void page_address_init(void);
757 #endif
759 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
760 #define page_address(page) lowmem_page_address(page)
761 #define set_page_address(page, address) do { } while(0)
762 #define page_address_init() do { } while(0)
763 #endif
766 * On an anonymous page mapped into a user virtual memory area,
767 * page->mapping points to its anon_vma, not to a struct address_space;
768 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
770 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
771 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
772 * and then page->mapping points, not to an anon_vma, but to a private
773 * structure which KSM associates with that merged page. See ksm.h.
775 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
777 * Please note that, confusingly, "page_mapping" refers to the inode
778 * address_space which maps the page from disk; whereas "page_mapped"
779 * refers to user virtual address space into which the page is mapped.
781 #define PAGE_MAPPING_ANON 1
782 #define PAGE_MAPPING_KSM 2
783 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
785 extern struct address_space swapper_space;
786 static inline struct address_space *page_mapping(struct page *page)
788 struct address_space *mapping = page->mapping;
790 VM_BUG_ON(PageSlab(page));
791 if (unlikely(PageSwapCache(page)))
792 mapping = &swapper_space;
793 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
794 mapping = NULL;
795 return mapping;
798 /* Neutral page->mapping pointer to address_space or anon_vma or other */
799 static inline void *page_rmapping(struct page *page)
801 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
804 static inline int PageAnon(struct page *page)
806 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
810 * Return the pagecache index of the passed page. Regular pagecache pages
811 * use ->index whereas swapcache pages use ->private
813 static inline pgoff_t page_index(struct page *page)
815 if (unlikely(PageSwapCache(page)))
816 return page_private(page);
817 return page->index;
821 * Return true if this page is mapped into pagetables.
823 static inline int page_mapped(struct page *page)
825 return atomic_read(&(page)->_mapcount) >= 0;
829 * Different kinds of faults, as returned by handle_mm_fault().
830 * Used to decide whether a process gets delivered SIGBUS or
831 * just gets major/minor fault counters bumped up.
834 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
836 #define VM_FAULT_OOM 0x0001
837 #define VM_FAULT_SIGBUS 0x0002
838 #define VM_FAULT_MAJOR 0x0004
839 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
840 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
841 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
843 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
844 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
845 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
847 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
849 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
850 VM_FAULT_HWPOISON_LARGE)
852 /* Encode hstate index for a hwpoisoned large page */
853 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
854 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
857 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
859 extern void pagefault_out_of_memory(void);
861 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
864 * Flags passed to show_mem() and show_free_areas() to suppress output in
865 * various contexts.
867 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
869 extern void show_free_areas(unsigned int flags);
870 extern bool skip_free_areas_node(unsigned int flags, int nid);
872 int shmem_lock(struct file *file, int lock, struct user_struct *user);
873 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
874 int shmem_zero_setup(struct vm_area_struct *);
876 extern int can_do_mlock(void);
877 extern int user_shm_lock(size_t, struct user_struct *);
878 extern void user_shm_unlock(size_t, struct user_struct *);
881 * Parameter block passed down to zap_pte_range in exceptional cases.
883 struct zap_details {
884 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
885 struct address_space *check_mapping; /* Check page->mapping if set */
886 pgoff_t first_index; /* Lowest page->index to unmap */
887 pgoff_t last_index; /* Highest page->index to unmap */
890 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
891 pte_t pte);
893 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
894 unsigned long size);
895 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
896 unsigned long size, struct zap_details *);
897 unsigned long unmap_vmas(struct mmu_gather *tlb,
898 struct vm_area_struct *start_vma, unsigned long start_addr,
899 unsigned long end_addr, unsigned long *nr_accounted,
900 struct zap_details *);
903 * mm_walk - callbacks for walk_page_range
904 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
905 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
906 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
907 * this handler is required to be able to handle
908 * pmd_trans_huge() pmds. They may simply choose to
909 * split_huge_page() instead of handling it explicitly.
910 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
911 * @pte_hole: if set, called for each hole at all levels
912 * @hugetlb_entry: if set, called for each hugetlb entry
913 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
914 * is used.
916 * (see walk_page_range for more details)
918 struct mm_walk {
919 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
920 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
921 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
922 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
923 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
924 int (*hugetlb_entry)(pte_t *, unsigned long,
925 unsigned long, unsigned long, struct mm_walk *);
926 struct mm_struct *mm;
927 void *private;
930 int walk_page_range(unsigned long addr, unsigned long end,
931 struct mm_walk *walk);
932 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
933 unsigned long end, unsigned long floor, unsigned long ceiling);
934 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
935 struct vm_area_struct *vma);
936 void unmap_mapping_range(struct address_space *mapping,
937 loff_t const holebegin, loff_t const holelen, int even_cows);
938 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
939 unsigned long *pfn);
940 int follow_phys(struct vm_area_struct *vma, unsigned long address,
941 unsigned int flags, unsigned long *prot, resource_size_t *phys);
942 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
943 void *buf, int len, int write);
945 static inline void unmap_shared_mapping_range(struct address_space *mapping,
946 loff_t const holebegin, loff_t const holelen)
948 unmap_mapping_range(mapping, holebegin, holelen, 0);
951 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
952 extern void truncate_setsize(struct inode *inode, loff_t newsize);
953 extern int vmtruncate(struct inode *inode, loff_t offset);
954 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
956 int truncate_inode_page(struct address_space *mapping, struct page *page);
957 int generic_error_remove_page(struct address_space *mapping, struct page *page);
959 int invalidate_inode_page(struct page *page);
961 #ifdef CONFIG_MMU
962 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
963 unsigned long address, unsigned int flags);
964 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
965 unsigned long address, unsigned int fault_flags);
966 #else
967 static inline int handle_mm_fault(struct mm_struct *mm,
968 struct vm_area_struct *vma, unsigned long address,
969 unsigned int flags)
971 /* should never happen if there's no MMU */
972 BUG();
973 return VM_FAULT_SIGBUS;
975 static inline int fixup_user_fault(struct task_struct *tsk,
976 struct mm_struct *mm, unsigned long address,
977 unsigned int fault_flags)
979 /* should never happen if there's no MMU */
980 BUG();
981 return -EFAULT;
983 #endif
985 extern int make_pages_present(unsigned long addr, unsigned long end);
986 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
987 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
988 void *buf, int len, int write);
990 int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
991 unsigned long start, int len, unsigned int foll_flags,
992 struct page **pages, struct vm_area_struct **vmas,
993 int *nonblocking);
994 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
995 unsigned long start, int nr_pages, int write, int force,
996 struct page **pages, struct vm_area_struct **vmas);
997 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
998 struct page **pages);
999 struct page *get_dump_page(unsigned long addr);
1001 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1002 extern void do_invalidatepage(struct page *page, unsigned long offset);
1004 int __set_page_dirty_nobuffers(struct page *page);
1005 int __set_page_dirty_no_writeback(struct page *page);
1006 int redirty_page_for_writepage(struct writeback_control *wbc,
1007 struct page *page);
1008 void account_page_dirtied(struct page *page, struct address_space *mapping);
1009 void account_page_writeback(struct page *page);
1010 int set_page_dirty(struct page *page);
1011 int set_page_dirty_lock(struct page *page);
1012 int clear_page_dirty_for_io(struct page *page);
1014 /* Is the vma a continuation of the stack vma above it? */
1015 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1017 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1020 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1021 unsigned long addr)
1023 return (vma->vm_flags & VM_GROWSDOWN) &&
1024 (vma->vm_start == addr) &&
1025 !vma_growsdown(vma->vm_prev, addr);
1028 /* Is the vma a continuation of the stack vma below it? */
1029 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1031 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1034 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1035 unsigned long addr)
1037 return (vma->vm_flags & VM_GROWSUP) &&
1038 (vma->vm_end == addr) &&
1039 !vma_growsup(vma->vm_next, addr);
1042 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1043 unsigned long old_addr, struct vm_area_struct *new_vma,
1044 unsigned long new_addr, unsigned long len);
1045 extern unsigned long do_mremap(unsigned long addr,
1046 unsigned long old_len, unsigned long new_len,
1047 unsigned long flags, unsigned long new_addr);
1048 extern int mprotect_fixup(struct vm_area_struct *vma,
1049 struct vm_area_struct **pprev, unsigned long start,
1050 unsigned long end, unsigned long newflags);
1053 * doesn't attempt to fault and will return short.
1055 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1056 struct page **pages);
1058 * per-process(per-mm_struct) statistics.
1060 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
1062 atomic_long_set(&mm->rss_stat.count[member], value);
1065 #if defined(SPLIT_RSS_COUNTING)
1066 unsigned long get_mm_counter(struct mm_struct *mm, int member);
1067 #else
1068 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1070 return atomic_long_read(&mm->rss_stat.count[member]);
1072 #endif
1074 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1076 atomic_long_add(value, &mm->rss_stat.count[member]);
1079 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1081 atomic_long_inc(&mm->rss_stat.count[member]);
1084 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1086 atomic_long_dec(&mm->rss_stat.count[member]);
1089 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1091 return get_mm_counter(mm, MM_FILEPAGES) +
1092 get_mm_counter(mm, MM_ANONPAGES);
1095 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1097 return max(mm->hiwater_rss, get_mm_rss(mm));
1100 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1102 return max(mm->hiwater_vm, mm->total_vm);
1105 static inline void update_hiwater_rss(struct mm_struct *mm)
1107 unsigned long _rss = get_mm_rss(mm);
1109 if ((mm)->hiwater_rss < _rss)
1110 (mm)->hiwater_rss = _rss;
1113 static inline void update_hiwater_vm(struct mm_struct *mm)
1115 if (mm->hiwater_vm < mm->total_vm)
1116 mm->hiwater_vm = mm->total_vm;
1119 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1120 struct mm_struct *mm)
1122 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1124 if (*maxrss < hiwater_rss)
1125 *maxrss = hiwater_rss;
1128 #if defined(SPLIT_RSS_COUNTING)
1129 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
1130 #else
1131 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
1134 #endif
1136 int vma_wants_writenotify(struct vm_area_struct *vma);
1138 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1139 spinlock_t **ptl);
1140 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1141 spinlock_t **ptl)
1143 pte_t *ptep;
1144 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1145 return ptep;
1148 #ifdef __PAGETABLE_PUD_FOLDED
1149 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1150 unsigned long address)
1152 return 0;
1154 #else
1155 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1156 #endif
1158 #ifdef __PAGETABLE_PMD_FOLDED
1159 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1160 unsigned long address)
1162 return 0;
1164 #else
1165 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1166 #endif
1168 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1169 pmd_t *pmd, unsigned long address);
1170 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1173 * The following ifdef needed to get the 4level-fixup.h header to work.
1174 * Remove it when 4level-fixup.h has been removed.
1176 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1177 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1179 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1180 NULL: pud_offset(pgd, address);
1183 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1185 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1186 NULL: pmd_offset(pud, address);
1188 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1190 #if USE_SPLIT_PTLOCKS
1192 * We tuck a spinlock to guard each pagetable page into its struct page,
1193 * at page->private, with BUILD_BUG_ON to make sure that this will not
1194 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1195 * When freeing, reset page->mapping so free_pages_check won't complain.
1197 #define __pte_lockptr(page) &((page)->ptl)
1198 #define pte_lock_init(_page) do { \
1199 spin_lock_init(__pte_lockptr(_page)); \
1200 } while (0)
1201 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1202 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1203 #else /* !USE_SPLIT_PTLOCKS */
1205 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1207 #define pte_lock_init(page) do {} while (0)
1208 #define pte_lock_deinit(page) do {} while (0)
1209 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1210 #endif /* USE_SPLIT_PTLOCKS */
1212 static inline void pgtable_page_ctor(struct page *page)
1214 pte_lock_init(page);
1215 inc_zone_page_state(page, NR_PAGETABLE);
1218 static inline void pgtable_page_dtor(struct page *page)
1220 pte_lock_deinit(page);
1221 dec_zone_page_state(page, NR_PAGETABLE);
1224 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1225 ({ \
1226 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1227 pte_t *__pte = pte_offset_map(pmd, address); \
1228 *(ptlp) = __ptl; \
1229 spin_lock(__ptl); \
1230 __pte; \
1233 #define pte_unmap_unlock(pte, ptl) do { \
1234 spin_unlock(ptl); \
1235 pte_unmap(pte); \
1236 } while (0)
1238 #define pte_alloc_map(mm, vma, pmd, address) \
1239 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1240 pmd, address))? \
1241 NULL: pte_offset_map(pmd, address))
1243 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1244 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1245 pmd, address))? \
1246 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1248 #define pte_alloc_kernel(pmd, address) \
1249 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1250 NULL: pte_offset_kernel(pmd, address))
1252 extern void free_area_init(unsigned long * zones_size);
1253 extern void free_area_init_node(int nid, unsigned long * zones_size,
1254 unsigned long zone_start_pfn, unsigned long *zholes_size);
1255 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1257 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1258 * zones, allocate the backing mem_map and account for memory holes in a more
1259 * architecture independent manner. This is a substitute for creating the
1260 * zone_sizes[] and zholes_size[] arrays and passing them to
1261 * free_area_init_node()
1263 * An architecture is expected to register range of page frames backed by
1264 * physical memory with add_active_range() before calling
1265 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1266 * usage, an architecture is expected to do something like
1268 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1269 * max_highmem_pfn};
1270 * for_each_valid_physical_page_range()
1271 * add_active_range(node_id, start_pfn, end_pfn)
1272 * free_area_init_nodes(max_zone_pfns);
1274 * If the architecture guarantees that there are no holes in the ranges
1275 * registered with add_active_range(), free_bootmem_active_regions()
1276 * will call free_bootmem_node() for each registered physical page range.
1277 * Similarly sparse_memory_present_with_active_regions() calls
1278 * memory_present() for each range when SPARSEMEM is enabled.
1280 * See mm/page_alloc.c for more information on each function exposed by
1281 * CONFIG_ARCH_POPULATES_NODE_MAP
1283 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1284 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1285 unsigned long end_pfn);
1286 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1287 unsigned long end_pfn);
1288 extern void remove_all_active_ranges(void);
1289 void sort_node_map(void);
1290 unsigned long node_map_pfn_alignment(void);
1291 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1292 unsigned long end_pfn);
1293 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1294 unsigned long end_pfn);
1295 extern void get_pfn_range_for_nid(unsigned int nid,
1296 unsigned long *start_pfn, unsigned long *end_pfn);
1297 extern unsigned long find_min_pfn_with_active_regions(void);
1298 extern void free_bootmem_with_active_regions(int nid,
1299 unsigned long max_low_pfn);
1300 int add_from_early_node_map(struct range *range, int az,
1301 int nr_range, int nid);
1302 u64 __init find_memory_core_early(int nid, u64 size, u64 align,
1303 u64 goal, u64 limit);
1304 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1305 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1306 extern void sparse_memory_present_with_active_regions(int nid);
1307 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1309 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1310 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1311 static inline int __early_pfn_to_nid(unsigned long pfn)
1313 return 0;
1315 #else
1316 /* please see mm/page_alloc.c */
1317 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1318 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1319 /* there is a per-arch backend function. */
1320 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1321 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1322 #endif
1324 extern void set_dma_reserve(unsigned long new_dma_reserve);
1325 extern void memmap_init_zone(unsigned long, int, unsigned long,
1326 unsigned long, enum memmap_context);
1327 extern void setup_per_zone_wmarks(void);
1328 extern int __meminit init_per_zone_wmark_min(void);
1329 extern void mem_init(void);
1330 extern void __init mmap_init(void);
1331 extern void show_mem(unsigned int flags);
1332 extern void si_meminfo(struct sysinfo * val);
1333 extern void si_meminfo_node(struct sysinfo *val, int nid);
1334 extern int after_bootmem;
1336 extern __printf(3, 4)
1337 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1339 extern void setup_per_cpu_pageset(void);
1341 extern void zone_pcp_update(struct zone *zone);
1343 /* nommu.c */
1344 extern atomic_long_t mmap_pages_allocated;
1345 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1347 /* prio_tree.c */
1348 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1349 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1350 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1351 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1352 struct prio_tree_iter *iter);
1354 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1355 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1356 (vma = vma_prio_tree_next(vma, iter)); )
1358 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1359 struct list_head *list)
1361 vma->shared.vm_set.parent = NULL;
1362 list_add_tail(&vma->shared.vm_set.list, list);
1365 /* mmap.c */
1366 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1367 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1368 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1369 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1370 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1371 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1372 struct mempolicy *);
1373 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1374 extern int split_vma(struct mm_struct *,
1375 struct vm_area_struct *, unsigned long addr, int new_below);
1376 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1377 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1378 struct rb_node **, struct rb_node *);
1379 extern void unlink_file_vma(struct vm_area_struct *);
1380 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1381 unsigned long addr, unsigned long len, pgoff_t pgoff);
1382 extern void exit_mmap(struct mm_struct *);
1384 extern int mm_take_all_locks(struct mm_struct *mm);
1385 extern void mm_drop_all_locks(struct mm_struct *mm);
1387 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1388 extern void added_exe_file_vma(struct mm_struct *mm);
1389 extern void removed_exe_file_vma(struct mm_struct *mm);
1390 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1391 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1393 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1394 extern int install_special_mapping(struct mm_struct *mm,
1395 unsigned long addr, unsigned long len,
1396 unsigned long flags, struct page **pages);
1398 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1400 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1401 unsigned long len, unsigned long prot,
1402 unsigned long flag, unsigned long pgoff);
1403 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1404 unsigned long len, unsigned long flags,
1405 vm_flags_t vm_flags, unsigned long pgoff);
1407 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1408 unsigned long len, unsigned long prot,
1409 unsigned long flag, unsigned long offset)
1411 unsigned long ret = -EINVAL;
1412 if ((offset + PAGE_ALIGN(len)) < offset)
1413 goto out;
1414 if (!(offset & ~PAGE_MASK))
1415 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1416 out:
1417 return ret;
1420 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1422 extern unsigned long do_brk(unsigned long, unsigned long);
1424 /* truncate.c */
1425 extern void truncate_inode_pages(struct address_space *, loff_t);
1426 extern void truncate_inode_pages_range(struct address_space *,
1427 loff_t lstart, loff_t lend);
1429 /* generic vm_area_ops exported for stackable file systems */
1430 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1432 /* mm/page-writeback.c */
1433 int write_one_page(struct page *page, int wait);
1434 void task_dirty_inc(struct task_struct *tsk);
1436 /* readahead.c */
1437 #define VM_MAX_READAHEAD 128 /* kbytes */
1438 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1440 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1441 pgoff_t offset, unsigned long nr_to_read);
1443 void page_cache_sync_readahead(struct address_space *mapping,
1444 struct file_ra_state *ra,
1445 struct file *filp,
1446 pgoff_t offset,
1447 unsigned long size);
1449 void page_cache_async_readahead(struct address_space *mapping,
1450 struct file_ra_state *ra,
1451 struct file *filp,
1452 struct page *pg,
1453 pgoff_t offset,
1454 unsigned long size);
1456 unsigned long max_sane_readahead(unsigned long nr);
1457 unsigned long ra_submit(struct file_ra_state *ra,
1458 struct address_space *mapping,
1459 struct file *filp);
1461 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1462 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1464 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1465 extern int expand_downwards(struct vm_area_struct *vma,
1466 unsigned long address);
1467 #if VM_GROWSUP
1468 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1469 #else
1470 #define expand_upwards(vma, address) do { } while (0)
1471 #endif
1473 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1474 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1475 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1476 struct vm_area_struct **pprev);
1478 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1479 NULL if none. Assume start_addr < end_addr. */
1480 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1482 struct vm_area_struct * vma = find_vma(mm,start_addr);
1484 if (vma && end_addr <= vma->vm_start)
1485 vma = NULL;
1486 return vma;
1489 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1491 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1494 #ifdef CONFIG_MMU
1495 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1496 #else
1497 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1499 return __pgprot(0);
1501 #endif
1503 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1504 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1505 unsigned long pfn, unsigned long size, pgprot_t);
1506 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1507 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1508 unsigned long pfn);
1509 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1510 unsigned long pfn);
1512 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1513 unsigned int foll_flags);
1514 #define FOLL_WRITE 0x01 /* check pte is writable */
1515 #define FOLL_TOUCH 0x02 /* mark page accessed */
1516 #define FOLL_GET 0x04 /* do get_page on page */
1517 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1518 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1519 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1520 * and return without waiting upon it */
1521 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1522 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1523 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1525 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1526 void *data);
1527 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1528 unsigned long size, pte_fn_t fn, void *data);
1530 #ifdef CONFIG_PROC_FS
1531 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1532 #else
1533 static inline void vm_stat_account(struct mm_struct *mm,
1534 unsigned long flags, struct file *file, long pages)
1537 #endif /* CONFIG_PROC_FS */
1539 #ifdef CONFIG_DEBUG_PAGEALLOC
1540 extern int debug_pagealloc_enabled;
1542 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1544 static inline void enable_debug_pagealloc(void)
1546 debug_pagealloc_enabled = 1;
1548 #ifdef CONFIG_HIBERNATION
1549 extern bool kernel_page_present(struct page *page);
1550 #endif /* CONFIG_HIBERNATION */
1551 #else
1552 static inline void
1553 kernel_map_pages(struct page *page, int numpages, int enable) {}
1554 static inline void enable_debug_pagealloc(void)
1557 #ifdef CONFIG_HIBERNATION
1558 static inline bool kernel_page_present(struct page *page) { return true; }
1559 #endif /* CONFIG_HIBERNATION */
1560 #endif
1562 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1563 #ifdef __HAVE_ARCH_GATE_AREA
1564 int in_gate_area_no_mm(unsigned long addr);
1565 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1566 #else
1567 int in_gate_area_no_mm(unsigned long addr);
1568 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1569 #endif /* __HAVE_ARCH_GATE_AREA */
1571 int drop_caches_sysctl_handler(struct ctl_table *, int,
1572 void __user *, size_t *, loff_t *);
1573 unsigned long shrink_slab(struct shrink_control *shrink,
1574 unsigned long nr_pages_scanned,
1575 unsigned long lru_pages);
1577 #ifndef CONFIG_MMU
1578 #define randomize_va_space 0
1579 #else
1580 extern int randomize_va_space;
1581 #endif
1583 const char * arch_vma_name(struct vm_area_struct *vma);
1584 void print_vma_addr(char *prefix, unsigned long rip);
1586 void sparse_mem_maps_populate_node(struct page **map_map,
1587 unsigned long pnum_begin,
1588 unsigned long pnum_end,
1589 unsigned long map_count,
1590 int nodeid);
1592 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1593 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1594 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1595 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1596 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1597 void *vmemmap_alloc_block(unsigned long size, int node);
1598 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1599 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1600 int vmemmap_populate_basepages(struct page *start_page,
1601 unsigned long pages, int node);
1602 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1603 void vmemmap_populate_print_last(void);
1606 enum mf_flags {
1607 MF_COUNT_INCREASED = 1 << 0,
1609 extern void memory_failure(unsigned long pfn, int trapno);
1610 extern int __memory_failure(unsigned long pfn, int trapno, int flags);
1611 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1612 extern int unpoison_memory(unsigned long pfn);
1613 extern int sysctl_memory_failure_early_kill;
1614 extern int sysctl_memory_failure_recovery;
1615 extern void shake_page(struct page *p, int access);
1616 extern atomic_long_t mce_bad_pages;
1617 extern int soft_offline_page(struct page *page, int flags);
1619 extern void dump_page(struct page *page);
1621 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1622 extern void clear_huge_page(struct page *page,
1623 unsigned long addr,
1624 unsigned int pages_per_huge_page);
1625 extern void copy_user_huge_page(struct page *dst, struct page *src,
1626 unsigned long addr, struct vm_area_struct *vma,
1627 unsigned int pages_per_huge_page);
1628 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1630 #endif /* __KERNEL__ */
1631 #endif /* _LINUX_MM_H */