The discovered bit in PGCCSR register indicates if the device has been
[linux-2.6/next.git] / include / linux / mm.h
blob23d95283cac4191b0f8869c7e1f549b498b5df4e
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/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>
21 struct mempolicy;
22 struct anon_vma;
23 struct file_ra_state;
24 struct user_struct;
25 struct writeback_control;
27 #ifndef CONFIG_DISCONTIGMEM /* Don't use mapnrs, do it properly */
28 extern unsigned long max_mapnr;
29 #endif
31 extern unsigned long num_physpages;
32 extern unsigned long totalram_pages;
33 extern void * high_memory;
34 extern int page_cluster;
36 #ifdef CONFIG_SYSCTL
37 extern int sysctl_legacy_va_layout;
38 #else
39 #define sysctl_legacy_va_layout 0
40 #endif
42 #include <asm/page.h>
43 #include <asm/pgtable.h>
44 #include <asm/processor.h>
46 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
48 /* to align the pointer to the (next) page boundary */
49 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
52 * Linux kernel virtual memory manager primitives.
53 * The idea being to have a "virtual" mm in the same way
54 * we have a virtual fs - giving a cleaner interface to the
55 * mm details, and allowing different kinds of memory mappings
56 * (from shared memory to executable loading to arbitrary
57 * mmap() functions).
60 extern struct kmem_cache *vm_area_cachep;
62 #ifndef CONFIG_MMU
63 extern struct rb_root nommu_region_tree;
64 extern struct rw_semaphore nommu_region_sem;
66 extern unsigned int kobjsize(const void *objp);
67 #endif
70 * vm_flags in vm_area_struct, see mm_types.h.
72 #define VM_READ 0x00000001 /* currently active flags */
73 #define VM_WRITE 0x00000002
74 #define VM_EXEC 0x00000004
75 #define VM_SHARED 0x00000008
77 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
78 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
79 #define VM_MAYWRITE 0x00000020
80 #define VM_MAYEXEC 0x00000040
81 #define VM_MAYSHARE 0x00000080
83 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
84 #if defined(CONFIG_STACK_GROWSUP) || defined(CONFIG_IA64)
85 #define VM_GROWSUP 0x00000200
86 #else
87 #define VM_GROWSUP 0x00000000
88 #define VM_NOHUGEPAGE 0x00000200 /* MADV_NOHUGEPAGE marked this vma */
89 #endif
90 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
91 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
93 #define VM_EXECUTABLE 0x00001000
94 #define VM_LOCKED 0x00002000
95 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
97 /* Used by sys_madvise() */
98 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
99 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
101 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
102 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
103 #define VM_RESERVED 0x00080000 /* Count as reserved_vm like IO */
104 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
105 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
106 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
107 #define VM_NONLINEAR 0x00800000 /* Is non-linear (remap_file_pages) */
108 #ifndef CONFIG_TRANSPARENT_HUGEPAGE
109 #define VM_MAPPED_COPY 0x01000000 /* T if mapped copy of data (nommu mmap) */
110 #else
111 #define VM_HUGEPAGE 0x01000000 /* MADV_HUGEPAGE marked this vma */
112 #endif
113 #define VM_INSERTPAGE 0x02000000 /* The vma has had "vm_insert_page()" done on it */
114 #define VM_ALWAYSDUMP 0x04000000 /* Always include in core dumps */
116 #define VM_CAN_NONLINEAR 0x08000000 /* Has ->fault & does nonlinear pages */
117 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
118 #define VM_SAO 0x20000000 /* Strong Access Ordering (powerpc) */
119 #define VM_PFN_AT_MMAP 0x40000000 /* PFNMAP vma that is fully mapped at mmap time */
120 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
122 /* Bits set in the VMA until the stack is in its final location */
123 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
125 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
126 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
127 #endif
129 #ifdef CONFIG_STACK_GROWSUP
130 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
131 #else
132 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
133 #endif
135 #define VM_READHINTMASK (VM_SEQ_READ | VM_RAND_READ)
136 #define VM_ClearReadHint(v) (v)->vm_flags &= ~VM_READHINTMASK
137 #define VM_NormalReadHint(v) (!((v)->vm_flags & VM_READHINTMASK))
138 #define VM_SequentialReadHint(v) ((v)->vm_flags & VM_SEQ_READ)
139 #define VM_RandomReadHint(v) ((v)->vm_flags & VM_RAND_READ)
142 * Special vmas that are non-mergable, non-mlock()able.
143 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
145 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_RESERVED | VM_PFNMAP)
148 * mapping from the currently active vm_flags protection bits (the
149 * low four bits) to a page protection mask..
151 extern pgprot_t protection_map[16];
153 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
154 #define FAULT_FLAG_NONLINEAR 0x02 /* Fault was via a nonlinear mapping */
155 #define FAULT_FLAG_MKWRITE 0x04 /* Fault was mkwrite of existing pte */
156 #define FAULT_FLAG_ALLOW_RETRY 0x08 /* Retry fault if blocking */
157 #define FAULT_FLAG_RETRY_NOWAIT 0x10 /* Don't drop mmap_sem and wait when retrying */
158 #define FAULT_FLAG_KILLABLE 0x20 /* The fault task is in SIGKILL killable region */
161 * This interface is used by x86 PAT code to identify a pfn mapping that is
162 * linear over entire vma. This is to optimize PAT code that deals with
163 * marking the physical region with a particular prot. This is not for generic
164 * mm use. Note also that this check will not work if the pfn mapping is
165 * linear for a vma starting at physical address 0. In which case PAT code
166 * falls back to slow path of reserving physical range page by page.
168 static inline int is_linear_pfn_mapping(struct vm_area_struct *vma)
170 return !!(vma->vm_flags & VM_PFN_AT_MMAP);
173 static inline int is_pfn_mapping(struct vm_area_struct *vma)
175 return !!(vma->vm_flags & VM_PFNMAP);
179 * vm_fault is filled by the the pagefault handler and passed to the vma's
180 * ->fault function. The vma's ->fault is responsible for returning a bitmask
181 * of VM_FAULT_xxx flags that give details about how the fault was handled.
183 * pgoff should be used in favour of virtual_address, if possible. If pgoff
184 * is used, one may set VM_CAN_NONLINEAR in the vma->vm_flags to get nonlinear
185 * mapping support.
187 struct vm_fault {
188 unsigned int flags; /* FAULT_FLAG_xxx flags */
189 pgoff_t pgoff; /* Logical page offset based on vma */
190 void __user *virtual_address; /* Faulting virtual address */
192 struct page *page; /* ->fault handlers should return a
193 * page here, unless VM_FAULT_NOPAGE
194 * is set (which is also implied by
195 * VM_FAULT_ERROR).
200 * These are the virtual MM functions - opening of an area, closing and
201 * unmapping it (needed to keep files on disk up-to-date etc), pointer
202 * to the functions called when a no-page or a wp-page exception occurs.
204 struct vm_operations_struct {
205 void (*open)(struct vm_area_struct * area);
206 void (*close)(struct vm_area_struct * area);
207 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
209 /* notification that a previously read-only page is about to become
210 * writable, if an error is returned it will cause a SIGBUS */
211 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
213 /* called by access_process_vm when get_user_pages() fails, typically
214 * for use by special VMAs that can switch between memory and hardware
216 int (*access)(struct vm_area_struct *vma, unsigned long addr,
217 void *buf, int len, int write);
218 #ifdef CONFIG_NUMA
220 * set_policy() op must add a reference to any non-NULL @new mempolicy
221 * to hold the policy upon return. Caller should pass NULL @new to
222 * remove a policy and fall back to surrounding context--i.e. do not
223 * install a MPOL_DEFAULT policy, nor the task or system default
224 * mempolicy.
226 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
229 * get_policy() op must add reference [mpol_get()] to any policy at
230 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
231 * in mm/mempolicy.c will do this automatically.
232 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
233 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
234 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
235 * must return NULL--i.e., do not "fallback" to task or system default
236 * policy.
238 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
239 unsigned long addr);
240 int (*migrate)(struct vm_area_struct *vma, const nodemask_t *from,
241 const nodemask_t *to, unsigned long flags);
242 #endif
245 struct mmu_gather;
246 struct inode;
248 #define page_private(page) ((page)->private)
249 #define set_page_private(page, v) ((page)->private = (v))
252 * FIXME: take this include out, include page-flags.h in
253 * files which need it (119 of them)
255 #include <linux/page-flags.h>
256 #include <linux/huge_mm.h>
259 * Methods to modify the page usage count.
261 * What counts for a page usage:
262 * - cache mapping (page->mapping)
263 * - private data (page->private)
264 * - page mapped in a task's page tables, each mapping
265 * is counted separately
267 * Also, many kernel routines increase the page count before a critical
268 * routine so they can be sure the page doesn't go away from under them.
272 * Drop a ref, return true if the refcount fell to zero (the page has no users)
274 static inline int put_page_testzero(struct page *page)
276 VM_BUG_ON(atomic_read(&page->_count) == 0);
277 return atomic_dec_and_test(&page->_count);
281 * Try to grab a ref unless the page has a refcount of zero, return false if
282 * that is the case.
284 static inline int get_page_unless_zero(struct page *page)
286 return atomic_inc_not_zero(&page->_count);
289 extern int page_is_ram(unsigned long pfn);
291 /* Support for virtually mapped pages */
292 struct page *vmalloc_to_page(const void *addr);
293 unsigned long vmalloc_to_pfn(const void *addr);
296 * Determine if an address is within the vmalloc range
298 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
299 * is no special casing required.
301 static inline int is_vmalloc_addr(const void *x)
303 #ifdef CONFIG_MMU
304 unsigned long addr = (unsigned long)x;
306 return addr >= VMALLOC_START && addr < VMALLOC_END;
307 #else
308 return 0;
309 #endif
311 #ifdef CONFIG_MMU
312 extern int is_vmalloc_or_module_addr(const void *x);
313 #else
314 static inline int is_vmalloc_or_module_addr(const void *x)
316 return 0;
318 #endif
320 static inline void compound_lock(struct page *page)
322 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
323 bit_spin_lock(PG_compound_lock, &page->flags);
324 #endif
327 static inline void compound_unlock(struct page *page)
329 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
330 bit_spin_unlock(PG_compound_lock, &page->flags);
331 #endif
334 static inline unsigned long compound_lock_irqsave(struct page *page)
336 unsigned long uninitialized_var(flags);
337 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
338 local_irq_save(flags);
339 compound_lock(page);
340 #endif
341 return flags;
344 static inline void compound_unlock_irqrestore(struct page *page,
345 unsigned long flags)
347 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
348 compound_unlock(page);
349 local_irq_restore(flags);
350 #endif
353 static inline struct page *compound_head(struct page *page)
355 if (unlikely(PageTail(page)))
356 return page->first_page;
357 return page;
360 static inline int page_count(struct page *page)
362 return atomic_read(&compound_head(page)->_count);
365 static inline void get_page(struct page *page)
368 * Getting a normal page or the head of a compound page
369 * requires to already have an elevated page->_count. Only if
370 * we're getting a tail page, the elevated page->_count is
371 * required only in the head page, so for tail pages the
372 * bugcheck only verifies that the page->_count isn't
373 * negative.
375 VM_BUG_ON(atomic_read(&page->_count) < !PageTail(page));
376 atomic_inc(&page->_count);
378 * Getting a tail page will elevate both the head and tail
379 * page->_count(s).
381 if (unlikely(PageTail(page))) {
383 * This is safe only because
384 * __split_huge_page_refcount can't run under
385 * get_page().
387 VM_BUG_ON(atomic_read(&page->first_page->_count) <= 0);
388 atomic_inc(&page->first_page->_count);
392 static inline struct page *virt_to_head_page(const void *x)
394 struct page *page = virt_to_page(x);
395 return compound_head(page);
399 * Setup the page count before being freed into the page allocator for
400 * the first time (boot or memory hotplug)
402 static inline void init_page_count(struct page *page)
404 atomic_set(&page->_count, 1);
408 * PageBuddy() indicate that the page is free and in the buddy system
409 * (see mm/page_alloc.c).
411 * PAGE_BUDDY_MAPCOUNT_VALUE must be <= -2 but better not too close to
412 * -2 so that an underflow of the page_mapcount() won't be mistaken
413 * for a genuine PAGE_BUDDY_MAPCOUNT_VALUE. -128 can be created very
414 * efficiently by most CPU architectures.
416 #define PAGE_BUDDY_MAPCOUNT_VALUE (-128)
418 static inline int PageBuddy(struct page *page)
420 return atomic_read(&page->_mapcount) == PAGE_BUDDY_MAPCOUNT_VALUE;
423 static inline void __SetPageBuddy(struct page *page)
425 VM_BUG_ON(atomic_read(&page->_mapcount) != -1);
426 atomic_set(&page->_mapcount, PAGE_BUDDY_MAPCOUNT_VALUE);
429 static inline void __ClearPageBuddy(struct page *page)
431 VM_BUG_ON(!PageBuddy(page));
432 atomic_set(&page->_mapcount, -1);
435 void put_page(struct page *page);
436 void put_pages_list(struct list_head *pages);
438 void split_page(struct page *page, unsigned int order);
439 int split_free_page(struct page *page);
442 * Compound pages have a destructor function. Provide a
443 * prototype for that function and accessor functions.
444 * These are _only_ valid on the head of a PG_compound page.
446 typedef void compound_page_dtor(struct page *);
448 static inline void set_compound_page_dtor(struct page *page,
449 compound_page_dtor *dtor)
451 page[1].lru.next = (void *)dtor;
454 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
456 return (compound_page_dtor *)page[1].lru.next;
459 static inline int compound_order(struct page *page)
461 if (!PageHead(page))
462 return 0;
463 return (unsigned long)page[1].lru.prev;
466 static inline int compound_trans_order(struct page *page)
468 int order;
469 unsigned long flags;
471 if (!PageHead(page))
472 return 0;
474 flags = compound_lock_irqsave(page);
475 order = compound_order(page);
476 compound_unlock_irqrestore(page, flags);
477 return order;
480 static inline void set_compound_order(struct page *page, unsigned long order)
482 page[1].lru.prev = (void *)order;
485 #ifdef CONFIG_MMU
487 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
488 * servicing faults for write access. In the normal case, do always want
489 * pte_mkwrite. But get_user_pages can cause write faults for mappings
490 * that do not have writing enabled, when used by access_process_vm.
492 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
494 if (likely(vma->vm_flags & VM_WRITE))
495 pte = pte_mkwrite(pte);
496 return pte;
498 #endif
501 * Multiple processes may "see" the same page. E.g. for untouched
502 * mappings of /dev/null, all processes see the same page full of
503 * zeroes, and text pages of executables and shared libraries have
504 * only one copy in memory, at most, normally.
506 * For the non-reserved pages, page_count(page) denotes a reference count.
507 * page_count() == 0 means the page is free. page->lru is then used for
508 * freelist management in the buddy allocator.
509 * page_count() > 0 means the page has been allocated.
511 * Pages are allocated by the slab allocator in order to provide memory
512 * to kmalloc and kmem_cache_alloc. In this case, the management of the
513 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
514 * unless a particular usage is carefully commented. (the responsibility of
515 * freeing the kmalloc memory is the caller's, of course).
517 * A page may be used by anyone else who does a __get_free_page().
518 * In this case, page_count still tracks the references, and should only
519 * be used through the normal accessor functions. The top bits of page->flags
520 * and page->virtual store page management information, but all other fields
521 * are unused and could be used privately, carefully. The management of this
522 * page is the responsibility of the one who allocated it, and those who have
523 * subsequently been given references to it.
525 * The other pages (we may call them "pagecache pages") are completely
526 * managed by the Linux memory manager: I/O, buffers, swapping etc.
527 * The following discussion applies only to them.
529 * A pagecache page contains an opaque `private' member, which belongs to the
530 * page's address_space. Usually, this is the address of a circular list of
531 * the page's disk buffers. PG_private must be set to tell the VM to call
532 * into the filesystem to release these pages.
534 * A page may belong to an inode's memory mapping. In this case, page->mapping
535 * is the pointer to the inode, and page->index is the file offset of the page,
536 * in units of PAGE_CACHE_SIZE.
538 * If pagecache pages are not associated with an inode, they are said to be
539 * anonymous pages. These may become associated with the swapcache, and in that
540 * case PG_swapcache is set, and page->private is an offset into the swapcache.
542 * In either case (swapcache or inode backed), the pagecache itself holds one
543 * reference to the page. Setting PG_private should also increment the
544 * refcount. The each user mapping also has a reference to the page.
546 * The pagecache pages are stored in a per-mapping radix tree, which is
547 * rooted at mapping->page_tree, and indexed by offset.
548 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
549 * lists, we instead now tag pages as dirty/writeback in the radix tree.
551 * All pagecache pages may be subject to I/O:
552 * - inode pages may need to be read from disk,
553 * - inode pages which have been modified and are MAP_SHARED may need
554 * to be written back to the inode on disk,
555 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
556 * modified may need to be swapped out to swap space and (later) to be read
557 * back into memory.
561 * The zone field is never updated after free_area_init_core()
562 * sets it, so none of the operations on it need to be atomic.
567 * page->flags layout:
569 * There are three possibilities for how page->flags get
570 * laid out. The first is for the normal case, without
571 * sparsemem. The second is for sparsemem when there is
572 * plenty of space for node and section. The last is when
573 * we have run out of space and have to fall back to an
574 * alternate (slower) way of determining the node.
576 * No sparsemem or sparsemem vmemmap: | NODE | ZONE | ... | FLAGS |
577 * classic sparse with space for node:| SECTION | NODE | ZONE | ... | FLAGS |
578 * classic sparse no space for node: | SECTION | ZONE | ... | FLAGS |
580 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
581 #define SECTIONS_WIDTH SECTIONS_SHIFT
582 #else
583 #define SECTIONS_WIDTH 0
584 #endif
586 #define ZONES_WIDTH ZONES_SHIFT
588 #if SECTIONS_WIDTH+ZONES_WIDTH+NODES_SHIFT <= BITS_PER_LONG - NR_PAGEFLAGS
589 #define NODES_WIDTH NODES_SHIFT
590 #else
591 #ifdef CONFIG_SPARSEMEM_VMEMMAP
592 #error "Vmemmap: No space for nodes field in page flags"
593 #endif
594 #define NODES_WIDTH 0
595 #endif
597 /* Page flags: | [SECTION] | [NODE] | ZONE | ... | FLAGS | */
598 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
599 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
600 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
603 * We are going to use the flags for the page to node mapping if its in
604 * there. This includes the case where there is no node, so it is implicit.
606 #if !(NODES_WIDTH > 0 || NODES_SHIFT == 0)
607 #define NODE_NOT_IN_PAGE_FLAGS
608 #endif
611 * Define the bit shifts to access each section. For non-existent
612 * sections we define the shift as 0; that plus a 0 mask ensures
613 * the compiler will optimise away reference to them.
615 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
616 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
617 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
619 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
620 #ifdef NODE_NOT_IN_PAGE_FLAGS
621 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
622 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
623 SECTIONS_PGOFF : ZONES_PGOFF)
624 #else
625 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
626 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
627 NODES_PGOFF : ZONES_PGOFF)
628 #endif
630 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
632 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
633 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
634 #endif
636 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
637 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
638 #define SECTIONS_MASK ((1UL << SECTIONS_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;
647 * The identification function is only used by the buddy allocator for
648 * determining if two pages could be buddies. We are not really
649 * identifying a zone since we could be using a the section number
650 * id if we have not node id available in page flags.
651 * We guarantee only that it will return the same value for two
652 * combinable pages in a zone.
654 static inline int page_zone_id(struct page *page)
656 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
659 static inline int zone_to_nid(struct zone *zone)
661 #ifdef CONFIG_NUMA
662 return zone->node;
663 #else
664 return 0;
665 #endif
668 #ifdef NODE_NOT_IN_PAGE_FLAGS
669 extern int page_to_nid(const struct page *page);
670 #else
671 static inline int page_to_nid(const struct page *page)
673 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
675 #endif
677 static inline struct zone *page_zone(const struct page *page)
679 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
682 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
683 static inline void set_page_section(struct page *page, unsigned long section)
685 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
686 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
689 static inline unsigned long page_to_section(const struct page *page)
691 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
693 #endif
695 static inline void set_page_zone(struct page *page, enum zone_type zone)
697 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
698 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
701 static inline void set_page_node(struct page *page, unsigned long node)
703 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
704 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
707 static inline void set_page_links(struct page *page, enum zone_type zone,
708 unsigned long node, unsigned long pfn)
710 set_page_zone(page, zone);
711 set_page_node(page, node);
712 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
713 set_page_section(page, pfn_to_section_nr(pfn));
714 #endif
718 * Some inline functions in vmstat.h depend on page_zone()
720 #include <linux/vmstat.h>
722 static __always_inline void *lowmem_page_address(const struct page *page)
724 return __va(PFN_PHYS(page_to_pfn(page)));
727 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
728 #define HASHED_PAGE_VIRTUAL
729 #endif
731 #if defined(WANT_PAGE_VIRTUAL)
732 #define page_address(page) ((page)->virtual)
733 #define set_page_address(page, address) \
734 do { \
735 (page)->virtual = (address); \
736 } while(0)
737 #define page_address_init() do { } while(0)
738 #endif
740 #if defined(HASHED_PAGE_VIRTUAL)
741 void *page_address(const struct page *page);
742 void set_page_address(struct page *page, void *virtual);
743 void page_address_init(void);
744 #endif
746 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
747 #define page_address(page) lowmem_page_address(page)
748 #define set_page_address(page, address) do { } while(0)
749 #define page_address_init() do { } while(0)
750 #endif
753 * On an anonymous page mapped into a user virtual memory area,
754 * page->mapping points to its anon_vma, not to a struct address_space;
755 * with the PAGE_MAPPING_ANON bit set to distinguish it. See rmap.h.
757 * On an anonymous page in a VM_MERGEABLE area, if CONFIG_KSM is enabled,
758 * the PAGE_MAPPING_KSM bit may be set along with the PAGE_MAPPING_ANON bit;
759 * and then page->mapping points, not to an anon_vma, but to a private
760 * structure which KSM associates with that merged page. See ksm.h.
762 * PAGE_MAPPING_KSM without PAGE_MAPPING_ANON is currently never used.
764 * Please note that, confusingly, "page_mapping" refers to the inode
765 * address_space which maps the page from disk; whereas "page_mapped"
766 * refers to user virtual address space into which the page is mapped.
768 #define PAGE_MAPPING_ANON 1
769 #define PAGE_MAPPING_KSM 2
770 #define PAGE_MAPPING_FLAGS (PAGE_MAPPING_ANON | PAGE_MAPPING_KSM)
772 extern struct address_space swapper_space;
773 static inline struct address_space *page_mapping(struct page *page)
775 struct address_space *mapping = page->mapping;
777 VM_BUG_ON(PageSlab(page));
778 if (unlikely(PageSwapCache(page)))
779 mapping = &swapper_space;
780 else if ((unsigned long)mapping & PAGE_MAPPING_ANON)
781 mapping = NULL;
782 return mapping;
785 /* Neutral page->mapping pointer to address_space or anon_vma or other */
786 static inline void *page_rmapping(struct page *page)
788 return (void *)((unsigned long)page->mapping & ~PAGE_MAPPING_FLAGS);
791 static inline int PageAnon(struct page *page)
793 return ((unsigned long)page->mapping & PAGE_MAPPING_ANON) != 0;
797 * Return the pagecache index of the passed page. Regular pagecache pages
798 * use ->index whereas swapcache pages use ->private
800 static inline pgoff_t page_index(struct page *page)
802 if (unlikely(PageSwapCache(page)))
803 return page_private(page);
804 return page->index;
808 * The atomic page->_mapcount, like _count, starts from -1:
809 * so that transitions both from it and to it can be tracked,
810 * using atomic_inc_and_test and atomic_add_negative(-1).
812 static inline void reset_page_mapcount(struct page *page)
814 atomic_set(&(page)->_mapcount, -1);
817 static inline int page_mapcount(struct page *page)
819 return atomic_read(&(page)->_mapcount) + 1;
823 * Return true if this page is mapped into pagetables.
825 static inline int page_mapped(struct page *page)
827 return atomic_read(&(page)->_mapcount) >= 0;
831 * Different kinds of faults, as returned by handle_mm_fault().
832 * Used to decide whether a process gets delivered SIGBUS or
833 * just gets major/minor fault counters bumped up.
836 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
838 #define VM_FAULT_OOM 0x0001
839 #define VM_FAULT_SIGBUS 0x0002
840 #define VM_FAULT_MAJOR 0x0004
841 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
842 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
843 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
845 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
846 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
847 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
849 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
851 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_HWPOISON | \
852 VM_FAULT_HWPOISON_LARGE)
854 /* Encode hstate index for a hwpoisoned large page */
855 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
856 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
859 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
861 extern void pagefault_out_of_memory(void);
863 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
866 * Flags passed to show_mem() and show_free_areas() to suppress output in
867 * various contexts.
869 #define SHOW_MEM_FILTER_NODES (0x0001u) /* filter disallowed nodes */
871 extern void show_free_areas(unsigned int flags);
872 extern bool skip_free_areas_node(unsigned int flags, int nid);
874 int shmem_lock(struct file *file, int lock, struct user_struct *user);
875 struct file *shmem_file_setup(const char *name, loff_t size, unsigned long flags);
876 int shmem_zero_setup(struct vm_area_struct *);
878 extern int can_do_mlock(void);
879 extern int user_shm_lock(size_t, struct user_struct *);
880 extern void user_shm_unlock(size_t, struct user_struct *);
883 * Parameter block passed down to zap_pte_range in exceptional cases.
885 struct zap_details {
886 struct vm_area_struct *nonlinear_vma; /* Check page->index if set */
887 struct address_space *check_mapping; /* Check page->mapping if set */
888 pgoff_t first_index; /* Lowest page->index to unmap */
889 pgoff_t last_index; /* Highest page->index to unmap */
892 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
893 pte_t pte);
895 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
896 unsigned long size);
897 unsigned long zap_page_range(struct vm_area_struct *vma, unsigned long address,
898 unsigned long size, struct zap_details *);
899 unsigned long unmap_vmas(struct mmu_gather *tlb,
900 struct vm_area_struct *start_vma, unsigned long start_addr,
901 unsigned long end_addr, unsigned long *nr_accounted,
902 struct zap_details *);
905 * mm_walk - callbacks for walk_page_range
906 * @pgd_entry: if set, called for each non-empty PGD (top-level) entry
907 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
908 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
909 * this handler is required to be able to handle
910 * pmd_trans_huge() pmds. They may simply choose to
911 * split_huge_page() instead of handling it explicitly.
912 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
913 * @pte_hole: if set, called for each hole at all levels
914 * @hugetlb_entry: if set, called for each hugetlb entry
915 * *Caution*: The caller must hold mmap_sem() if @hugetlb_entry
916 * is used.
918 * (see walk_page_range for more details)
920 struct mm_walk {
921 int (*pgd_entry)(pgd_t *, unsigned long, unsigned long, struct mm_walk *);
922 int (*pud_entry)(pud_t *, unsigned long, unsigned long, struct mm_walk *);
923 int (*pmd_entry)(pmd_t *, unsigned long, unsigned long, struct mm_walk *);
924 int (*pte_entry)(pte_t *, unsigned long, unsigned long, struct mm_walk *);
925 int (*pte_hole)(unsigned long, unsigned long, struct mm_walk *);
926 int (*hugetlb_entry)(pte_t *, unsigned long,
927 unsigned long, unsigned long, struct mm_walk *);
928 struct mm_struct *mm;
929 void *private;
932 int walk_page_range(unsigned long addr, unsigned long end,
933 struct mm_walk *walk);
934 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
935 unsigned long end, unsigned long floor, unsigned long ceiling);
936 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
937 struct vm_area_struct *vma);
938 void unmap_mapping_range(struct address_space *mapping,
939 loff_t const holebegin, loff_t const holelen, int even_cows);
940 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
941 unsigned long *pfn);
942 int follow_phys(struct vm_area_struct *vma, unsigned long address,
943 unsigned int flags, unsigned long *prot, resource_size_t *phys);
944 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
945 void *buf, int len, int write);
947 static inline void unmap_shared_mapping_range(struct address_space *mapping,
948 loff_t const holebegin, loff_t const holelen)
950 unmap_mapping_range(mapping, holebegin, holelen, 0);
953 extern void truncate_pagecache(struct inode *inode, loff_t old, loff_t new);
954 extern void truncate_setsize(struct inode *inode, loff_t newsize);
955 extern int vmtruncate(struct inode *inode, loff_t offset);
956 extern int vmtruncate_range(struct inode *inode, loff_t offset, loff_t end);
958 int truncate_inode_page(struct address_space *mapping, struct page *page);
959 int generic_error_remove_page(struct address_space *mapping, struct page *page);
961 int invalidate_inode_page(struct page *page);
963 #ifdef CONFIG_MMU
964 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
965 unsigned long address, unsigned int flags);
966 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
967 unsigned long address, unsigned int fault_flags);
968 #else
969 static inline int handle_mm_fault(struct mm_struct *mm,
970 struct vm_area_struct *vma, unsigned long address,
971 unsigned int flags)
973 /* should never happen if there's no MMU */
974 BUG();
975 return VM_FAULT_SIGBUS;
977 static inline int fixup_user_fault(struct task_struct *tsk,
978 struct mm_struct *mm, unsigned long address,
979 unsigned int fault_flags)
981 /* should never happen if there's no MMU */
982 BUG();
983 return -EFAULT;
985 #endif
987 extern int make_pages_present(unsigned long addr, unsigned long end);
988 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
989 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
990 void *buf, int len, int write);
992 int __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
993 unsigned long start, int len, unsigned int foll_flags,
994 struct page **pages, struct vm_area_struct **vmas,
995 int *nonblocking);
996 int get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
997 unsigned long start, int nr_pages, int write, int force,
998 struct page **pages, struct vm_area_struct **vmas);
999 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1000 struct page **pages);
1001 struct page *get_dump_page(unsigned long addr);
1003 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1004 extern void do_invalidatepage(struct page *page, unsigned long offset);
1006 int __set_page_dirty_nobuffers(struct page *page);
1007 int __set_page_dirty_no_writeback(struct page *page);
1008 int redirty_page_for_writepage(struct writeback_control *wbc,
1009 struct page *page);
1010 void account_page_dirtied(struct page *page, struct address_space *mapping);
1011 void account_page_writeback(struct page *page);
1012 int set_page_dirty(struct page *page);
1013 int set_page_dirty_lock(struct page *page);
1014 int clear_page_dirty_for_io(struct page *page);
1016 /* Is the vma a continuation of the stack vma above it? */
1017 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1019 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1022 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1023 unsigned long addr)
1025 return (vma->vm_flags & VM_GROWSDOWN) &&
1026 (vma->vm_start == addr) &&
1027 !vma_growsdown(vma->vm_prev, addr);
1030 /* Is the vma a continuation of the stack vma below it? */
1031 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1033 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1036 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1037 unsigned long addr)
1039 return (vma->vm_flags & VM_GROWSUP) &&
1040 (vma->vm_end == addr) &&
1041 !vma_growsup(vma->vm_next, addr);
1044 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1045 unsigned long old_addr, struct vm_area_struct *new_vma,
1046 unsigned long new_addr, unsigned long len);
1047 extern unsigned long do_mremap(unsigned long addr,
1048 unsigned long old_len, unsigned long new_len,
1049 unsigned long flags, unsigned long new_addr);
1050 extern int mprotect_fixup(struct vm_area_struct *vma,
1051 struct vm_area_struct **pprev, unsigned long start,
1052 unsigned long end, unsigned long newflags);
1055 * doesn't attempt to fault and will return short.
1057 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1058 struct page **pages);
1060 * per-process(per-mm_struct) statistics.
1062 static inline void set_mm_counter(struct mm_struct *mm, int member, long value)
1064 atomic_long_set(&mm->rss_stat.count[member], value);
1067 #if defined(SPLIT_RSS_COUNTING)
1068 unsigned long get_mm_counter(struct mm_struct *mm, int member);
1069 #else
1070 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1072 return atomic_long_read(&mm->rss_stat.count[member]);
1074 #endif
1076 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1078 atomic_long_add(value, &mm->rss_stat.count[member]);
1081 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1083 atomic_long_inc(&mm->rss_stat.count[member]);
1086 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1088 atomic_long_dec(&mm->rss_stat.count[member]);
1091 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1093 return get_mm_counter(mm, MM_FILEPAGES) +
1094 get_mm_counter(mm, MM_ANONPAGES);
1097 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1099 return max(mm->hiwater_rss, get_mm_rss(mm));
1102 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1104 return max(mm->hiwater_vm, mm->total_vm);
1107 static inline void update_hiwater_rss(struct mm_struct *mm)
1109 unsigned long _rss = get_mm_rss(mm);
1111 if ((mm)->hiwater_rss < _rss)
1112 (mm)->hiwater_rss = _rss;
1115 static inline void update_hiwater_vm(struct mm_struct *mm)
1117 if (mm->hiwater_vm < mm->total_vm)
1118 mm->hiwater_vm = mm->total_vm;
1121 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1122 struct mm_struct *mm)
1124 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1126 if (*maxrss < hiwater_rss)
1127 *maxrss = hiwater_rss;
1130 #if defined(SPLIT_RSS_COUNTING)
1131 void sync_mm_rss(struct task_struct *task, struct mm_struct *mm);
1132 #else
1133 static inline void sync_mm_rss(struct task_struct *task, struct mm_struct *mm)
1136 #endif
1138 int vma_wants_writenotify(struct vm_area_struct *vma);
1140 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1141 spinlock_t **ptl);
1142 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1143 spinlock_t **ptl)
1145 pte_t *ptep;
1146 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1147 return ptep;
1150 #ifdef __PAGETABLE_PUD_FOLDED
1151 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1152 unsigned long address)
1154 return 0;
1156 #else
1157 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1158 #endif
1160 #ifdef __PAGETABLE_PMD_FOLDED
1161 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1162 unsigned long address)
1164 return 0;
1166 #else
1167 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1168 #endif
1170 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1171 pmd_t *pmd, unsigned long address);
1172 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1175 * The following ifdef needed to get the 4level-fixup.h header to work.
1176 * Remove it when 4level-fixup.h has been removed.
1178 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1179 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1181 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1182 NULL: pud_offset(pgd, address);
1185 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1187 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1188 NULL: pmd_offset(pud, address);
1190 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1192 #if USE_SPLIT_PTLOCKS
1194 * We tuck a spinlock to guard each pagetable page into its struct page,
1195 * at page->private, with BUILD_BUG_ON to make sure that this will not
1196 * overflow into the next struct page (as it might with DEBUG_SPINLOCK).
1197 * When freeing, reset page->mapping so free_pages_check won't complain.
1199 #define __pte_lockptr(page) &((page)->ptl)
1200 #define pte_lock_init(_page) do { \
1201 spin_lock_init(__pte_lockptr(_page)); \
1202 } while (0)
1203 #define pte_lock_deinit(page) ((page)->mapping = NULL)
1204 #define pte_lockptr(mm, pmd) ({(void)(mm); __pte_lockptr(pmd_page(*(pmd)));})
1205 #else /* !USE_SPLIT_PTLOCKS */
1207 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1209 #define pte_lock_init(page) do {} while (0)
1210 #define pte_lock_deinit(page) do {} while (0)
1211 #define pte_lockptr(mm, pmd) ({(void)(pmd); &(mm)->page_table_lock;})
1212 #endif /* USE_SPLIT_PTLOCKS */
1214 static inline void pgtable_page_ctor(struct page *page)
1216 pte_lock_init(page);
1217 inc_zone_page_state(page, NR_PAGETABLE);
1220 static inline void pgtable_page_dtor(struct page *page)
1222 pte_lock_deinit(page);
1223 dec_zone_page_state(page, NR_PAGETABLE);
1226 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1227 ({ \
1228 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1229 pte_t *__pte = pte_offset_map(pmd, address); \
1230 *(ptlp) = __ptl; \
1231 spin_lock(__ptl); \
1232 __pte; \
1235 #define pte_unmap_unlock(pte, ptl) do { \
1236 spin_unlock(ptl); \
1237 pte_unmap(pte); \
1238 } while (0)
1240 #define pte_alloc_map(mm, vma, pmd, address) \
1241 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1242 pmd, address))? \
1243 NULL: pte_offset_map(pmd, address))
1245 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1246 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1247 pmd, address))? \
1248 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1250 #define pte_alloc_kernel(pmd, address) \
1251 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1252 NULL: pte_offset_kernel(pmd, address))
1254 extern void free_area_init(unsigned long * zones_size);
1255 extern void free_area_init_node(int nid, unsigned long * zones_size,
1256 unsigned long zone_start_pfn, unsigned long *zholes_size);
1257 #ifdef CONFIG_ARCH_POPULATES_NODE_MAP
1259 * With CONFIG_ARCH_POPULATES_NODE_MAP set, an architecture may initialise its
1260 * zones, allocate the backing mem_map and account for memory holes in a more
1261 * architecture independent manner. This is a substitute for creating the
1262 * zone_sizes[] and zholes_size[] arrays and passing them to
1263 * free_area_init_node()
1265 * An architecture is expected to register range of page frames backed by
1266 * physical memory with add_active_range() before calling
1267 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1268 * usage, an architecture is expected to do something like
1270 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1271 * max_highmem_pfn};
1272 * for_each_valid_physical_page_range()
1273 * add_active_range(node_id, start_pfn, end_pfn)
1274 * free_area_init_nodes(max_zone_pfns);
1276 * If the architecture guarantees that there are no holes in the ranges
1277 * registered with add_active_range(), free_bootmem_active_regions()
1278 * will call free_bootmem_node() for each registered physical page range.
1279 * Similarly sparse_memory_present_with_active_regions() calls
1280 * memory_present() for each range when SPARSEMEM is enabled.
1282 * See mm/page_alloc.c for more information on each function exposed by
1283 * CONFIG_ARCH_POPULATES_NODE_MAP
1285 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1286 extern void add_active_range(unsigned int nid, unsigned long start_pfn,
1287 unsigned long end_pfn);
1288 extern void remove_active_range(unsigned int nid, unsigned long start_pfn,
1289 unsigned long end_pfn);
1290 extern void remove_all_active_ranges(void);
1291 void sort_node_map(void);
1292 unsigned long node_map_pfn_alignment(void);
1293 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1294 unsigned long end_pfn);
1295 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1296 unsigned long end_pfn);
1297 extern void get_pfn_range_for_nid(unsigned int nid,
1298 unsigned long *start_pfn, unsigned long *end_pfn);
1299 extern unsigned long find_min_pfn_with_active_regions(void);
1300 extern void free_bootmem_with_active_regions(int nid,
1301 unsigned long max_low_pfn);
1302 int add_from_early_node_map(struct range *range, int az,
1303 int nr_range, int nid);
1304 u64 __init find_memory_core_early(int nid, u64 size, u64 align,
1305 u64 goal, u64 limit);
1306 typedef int (*work_fn_t)(unsigned long, unsigned long, void *);
1307 extern void work_with_active_regions(int nid, work_fn_t work_fn, void *data);
1308 extern void sparse_memory_present_with_active_regions(int nid);
1309 #endif /* CONFIG_ARCH_POPULATES_NODE_MAP */
1311 #if !defined(CONFIG_ARCH_POPULATES_NODE_MAP) && \
1312 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1313 static inline int __early_pfn_to_nid(unsigned long pfn)
1315 return 0;
1317 #else
1318 /* please see mm/page_alloc.c */
1319 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1320 #ifdef CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID
1321 /* there is a per-arch backend function. */
1322 extern int __meminit __early_pfn_to_nid(unsigned long pfn);
1323 #endif /* CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID */
1324 #endif
1326 extern void set_dma_reserve(unsigned long new_dma_reserve);
1327 extern void memmap_init_zone(unsigned long, int, unsigned long,
1328 unsigned long, enum memmap_context);
1329 extern void setup_per_zone_wmarks(void);
1330 extern int __meminit init_per_zone_wmark_min(void);
1331 extern void mem_init(void);
1332 extern void __init mmap_init(void);
1333 extern void show_mem(unsigned int flags);
1334 extern void si_meminfo(struct sysinfo * val);
1335 extern void si_meminfo_node(struct sysinfo *val, int nid);
1336 extern int after_bootmem;
1338 extern void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1340 extern void setup_per_cpu_pageset(void);
1342 extern void zone_pcp_update(struct zone *zone);
1344 /* nommu.c */
1345 extern atomic_long_t mmap_pages_allocated;
1346 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1348 /* prio_tree.c */
1349 void vma_prio_tree_add(struct vm_area_struct *, struct vm_area_struct *old);
1350 void vma_prio_tree_insert(struct vm_area_struct *, struct prio_tree_root *);
1351 void vma_prio_tree_remove(struct vm_area_struct *, struct prio_tree_root *);
1352 struct vm_area_struct *vma_prio_tree_next(struct vm_area_struct *vma,
1353 struct prio_tree_iter *iter);
1355 #define vma_prio_tree_foreach(vma, iter, root, begin, end) \
1356 for (prio_tree_iter_init(iter, root, begin, end), vma = NULL; \
1357 (vma = vma_prio_tree_next(vma, iter)); )
1359 static inline void vma_nonlinear_insert(struct vm_area_struct *vma,
1360 struct list_head *list)
1362 vma->shared.vm_set.parent = NULL;
1363 list_add_tail(&vma->shared.vm_set.list, list);
1366 /* mmap.c */
1367 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1368 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1369 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1370 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1371 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1372 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1373 struct mempolicy *);
1374 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1375 extern int split_vma(struct mm_struct *,
1376 struct vm_area_struct *, unsigned long addr, int new_below);
1377 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1378 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1379 struct rb_node **, struct rb_node *);
1380 extern void unlink_file_vma(struct vm_area_struct *);
1381 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1382 unsigned long addr, unsigned long len, pgoff_t pgoff);
1383 extern void exit_mmap(struct mm_struct *);
1385 extern int mm_take_all_locks(struct mm_struct *mm);
1386 extern void mm_drop_all_locks(struct mm_struct *mm);
1388 /* From fs/proc/base.c. callers must _not_ hold the mm's exe_file_lock */
1389 extern void added_exe_file_vma(struct mm_struct *mm);
1390 extern void removed_exe_file_vma(struct mm_struct *mm);
1391 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1392 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1394 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1395 extern int install_special_mapping(struct mm_struct *mm,
1396 unsigned long addr, unsigned long len,
1397 unsigned long flags, struct page **pages);
1399 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1401 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1402 unsigned long len, unsigned long prot,
1403 unsigned long flag, unsigned long pgoff);
1404 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1405 unsigned long len, unsigned long flags,
1406 vm_flags_t vm_flags, unsigned long pgoff);
1408 static inline unsigned long do_mmap(struct file *file, unsigned long addr,
1409 unsigned long len, unsigned long prot,
1410 unsigned long flag, unsigned long offset)
1412 unsigned long ret = -EINVAL;
1413 if ((offset + PAGE_ALIGN(len)) < offset)
1414 goto out;
1415 if (!(offset & ~PAGE_MASK))
1416 ret = do_mmap_pgoff(file, addr, len, prot, flag, offset >> PAGE_SHIFT);
1417 out:
1418 return ret;
1421 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1423 extern unsigned long do_brk(unsigned long, unsigned long);
1425 /* truncate.c */
1426 extern void truncate_inode_pages(struct address_space *, loff_t);
1427 extern void truncate_inode_pages_range(struct address_space *,
1428 loff_t lstart, loff_t lend);
1430 /* generic vm_area_ops exported for stackable file systems */
1431 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1433 /* mm/page-writeback.c */
1434 int write_one_page(struct page *page, int wait);
1435 void task_dirty_inc(struct task_struct *tsk);
1437 /* readahead.c */
1438 #define VM_MAX_READAHEAD 128 /* kbytes */
1439 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1441 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1442 pgoff_t offset, unsigned long nr_to_read);
1444 void page_cache_sync_readahead(struct address_space *mapping,
1445 struct file_ra_state *ra,
1446 struct file *filp,
1447 pgoff_t offset,
1448 unsigned long size);
1450 void page_cache_async_readahead(struct address_space *mapping,
1451 struct file_ra_state *ra,
1452 struct file *filp,
1453 struct page *pg,
1454 pgoff_t offset,
1455 unsigned long size);
1457 unsigned long max_sane_readahead(unsigned long nr);
1458 unsigned long ra_submit(struct file_ra_state *ra,
1459 struct address_space *mapping,
1460 struct file *filp);
1462 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1463 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1465 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1466 extern int expand_downwards(struct vm_area_struct *vma,
1467 unsigned long address);
1468 #if VM_GROWSUP
1469 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1470 #else
1471 #define expand_upwards(vma, address) do { } while (0)
1472 #endif
1474 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1475 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1476 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1477 struct vm_area_struct **pprev);
1479 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1480 NULL if none. Assume start_addr < end_addr. */
1481 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1483 struct vm_area_struct * vma = find_vma(mm,start_addr);
1485 if (vma && end_addr <= vma->vm_start)
1486 vma = NULL;
1487 return vma;
1490 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1492 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
1495 #ifdef CONFIG_MMU
1496 pgprot_t vm_get_page_prot(unsigned long vm_flags);
1497 #else
1498 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
1500 return __pgprot(0);
1502 #endif
1504 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
1505 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
1506 unsigned long pfn, unsigned long size, pgprot_t);
1507 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
1508 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
1509 unsigned long pfn);
1510 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
1511 unsigned long pfn);
1513 struct page *follow_page(struct vm_area_struct *, unsigned long address,
1514 unsigned int foll_flags);
1515 #define FOLL_WRITE 0x01 /* check pte is writable */
1516 #define FOLL_TOUCH 0x02 /* mark page accessed */
1517 #define FOLL_GET 0x04 /* do get_page on page */
1518 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
1519 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
1520 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
1521 * and return without waiting upon it */
1522 #define FOLL_MLOCK 0x40 /* mark page as mlocked */
1523 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
1524 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
1526 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
1527 void *data);
1528 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
1529 unsigned long size, pte_fn_t fn, void *data);
1531 #ifdef CONFIG_PROC_FS
1532 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
1533 #else
1534 static inline void vm_stat_account(struct mm_struct *mm,
1535 unsigned long flags, struct file *file, long pages)
1538 #endif /* CONFIG_PROC_FS */
1540 #ifdef CONFIG_DEBUG_PAGEALLOC
1541 extern int debug_pagealloc_enabled;
1543 extern void kernel_map_pages(struct page *page, int numpages, int enable);
1545 static inline void enable_debug_pagealloc(void)
1547 debug_pagealloc_enabled = 1;
1549 #ifdef CONFIG_HIBERNATION
1550 extern bool kernel_page_present(struct page *page);
1551 #endif /* CONFIG_HIBERNATION */
1552 #else
1553 static inline void
1554 kernel_map_pages(struct page *page, int numpages, int enable) {}
1555 static inline void enable_debug_pagealloc(void)
1558 #ifdef CONFIG_HIBERNATION
1559 static inline bool kernel_page_present(struct page *page) { return true; }
1560 #endif /* CONFIG_HIBERNATION */
1561 #endif
1563 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
1564 #ifdef __HAVE_ARCH_GATE_AREA
1565 int in_gate_area_no_mm(unsigned long addr);
1566 int in_gate_area(struct mm_struct *mm, unsigned long addr);
1567 #else
1568 int in_gate_area_no_mm(unsigned long addr);
1569 #define in_gate_area(mm, addr) ({(void)mm; in_gate_area_no_mm(addr);})
1570 #endif /* __HAVE_ARCH_GATE_AREA */
1572 int drop_caches_sysctl_handler(struct ctl_table *, int,
1573 void __user *, size_t *, loff_t *);
1574 unsigned long shrink_slab(struct shrink_control *shrink,
1575 unsigned long nr_pages_scanned,
1576 unsigned long lru_pages);
1578 #ifndef CONFIG_MMU
1579 #define randomize_va_space 0
1580 #else
1581 extern int randomize_va_space;
1582 #endif
1584 const char * arch_vma_name(struct vm_area_struct *vma);
1585 void print_vma_addr(char *prefix, unsigned long rip);
1587 void sparse_mem_maps_populate_node(struct page **map_map,
1588 unsigned long pnum_begin,
1589 unsigned long pnum_end,
1590 unsigned long map_count,
1591 int nodeid);
1593 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
1594 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
1595 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
1596 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
1597 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
1598 void *vmemmap_alloc_block(unsigned long size, int node);
1599 void *vmemmap_alloc_block_buf(unsigned long size, int node);
1600 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
1601 int vmemmap_populate_basepages(struct page *start_page,
1602 unsigned long pages, int node);
1603 int vmemmap_populate(struct page *start_page, unsigned long pages, int node);
1604 void vmemmap_populate_print_last(void);
1607 enum mf_flags {
1608 MF_COUNT_INCREASED = 1 << 0,
1610 extern void memory_failure(unsigned long pfn, int trapno);
1611 extern int __memory_failure(unsigned long pfn, int trapno, int flags);
1612 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
1613 extern int unpoison_memory(unsigned long pfn);
1614 extern int sysctl_memory_failure_early_kill;
1615 extern int sysctl_memory_failure_recovery;
1616 extern void shake_page(struct page *p, int access);
1617 extern atomic_long_t mce_bad_pages;
1618 extern int soft_offline_page(struct page *page, int flags);
1620 extern void dump_page(struct page *page);
1622 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
1623 extern void clear_huge_page(struct page *page,
1624 unsigned long addr,
1625 unsigned int pages_per_huge_page);
1626 extern void copy_user_huge_page(struct page *dst, struct page *src,
1627 unsigned long addr, struct vm_area_struct *vma,
1628 unsigned int pages_per_huge_page);
1629 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
1631 #endif /* __KERNEL__ */
1632 #endif /* _LINUX_MM_H */