4 #include <linux/errno.h>
8 #include <linux/mmdebug.h>
10 #include <linux/bug.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 #include <linux/resource.h>
22 #include <linux/page_ext.h>
23 #include <linux/err.h>
27 struct anon_vma_chain
;
30 struct writeback_control
;
33 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
34 extern unsigned long max_mapnr
;
36 static inline void set_max_mapnr(unsigned long limit
)
41 static inline void set_max_mapnr(unsigned long limit
) { }
44 extern unsigned long totalram_pages
;
45 extern void * high_memory
;
46 extern int page_cluster
;
49 extern int sysctl_legacy_va_layout
;
51 #define sysctl_legacy_va_layout 0
55 #include <asm/pgtable.h>
56 #include <asm/processor.h>
59 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
63 * To prevent common memory management code establishing
64 * a zero page mapping on a read fault.
65 * This macro should be defined within <asm/pgtable.h>.
66 * s390 does this to prevent multiplexing of hardware bits
67 * related to the physical page in case of virtualization.
69 #ifndef mm_forbids_zeropage
70 #define mm_forbids_zeropage(X) (0)
73 extern unsigned long sysctl_user_reserve_kbytes
;
74 extern unsigned long sysctl_admin_reserve_kbytes
;
76 extern int sysctl_overcommit_memory
;
77 extern int sysctl_overcommit_ratio
;
78 extern unsigned long sysctl_overcommit_kbytes
;
80 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
82 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
85 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
87 /* to align the pointer to the (next) page boundary */
88 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
90 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
91 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
94 * Linux kernel virtual memory manager primitives.
95 * The idea being to have a "virtual" mm in the same way
96 * we have a virtual fs - giving a cleaner interface to the
97 * mm details, and allowing different kinds of memory mappings
98 * (from shared memory to executable loading to arbitrary
102 extern struct kmem_cache
*vm_area_cachep
;
105 extern struct rb_root nommu_region_tree
;
106 extern struct rw_semaphore nommu_region_sem
;
108 extern unsigned int kobjsize(const void *objp
);
112 * vm_flags in vm_area_struct, see mm_types.h.
114 #define VM_NONE 0x00000000
116 #define VM_READ 0x00000001 /* currently active flags */
117 #define VM_WRITE 0x00000002
118 #define VM_EXEC 0x00000004
119 #define VM_SHARED 0x00000008
121 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
122 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
123 #define VM_MAYWRITE 0x00000020
124 #define VM_MAYEXEC 0x00000040
125 #define VM_MAYSHARE 0x00000080
127 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
128 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
129 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
130 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
131 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
133 #define VM_LOCKED 0x00002000
134 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
136 /* Used by sys_madvise() */
137 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
138 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
140 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
141 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
142 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
143 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
144 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
145 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
146 #define VM_ARCH_2 0x02000000
147 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
149 #ifdef CONFIG_MEM_SOFT_DIRTY
150 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
152 # define VM_SOFTDIRTY 0
155 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
156 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
157 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
158 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
160 #if defined(CONFIG_X86)
161 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
162 #elif defined(CONFIG_PPC)
163 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
164 #elif defined(CONFIG_PARISC)
165 # define VM_GROWSUP VM_ARCH_1
166 #elif defined(CONFIG_METAG)
167 # define VM_GROWSUP VM_ARCH_1
168 #elif defined(CONFIG_IA64)
169 # define VM_GROWSUP VM_ARCH_1
170 #elif !defined(CONFIG_MMU)
171 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
174 #if defined(CONFIG_X86)
175 /* MPX specific bounds table or bounds directory */
176 # define VM_MPX VM_ARCH_2
180 # define VM_GROWSUP VM_NONE
183 /* Bits set in the VMA until the stack is in its final location */
184 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
186 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
187 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
190 #ifdef CONFIG_STACK_GROWSUP
191 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
193 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
197 * Special vmas that are non-mergable, non-mlock()able.
198 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
200 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
202 /* This mask defines which mm->def_flags a process can inherit its parent */
203 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
206 * mapping from the currently active vm_flags protection bits (the
207 * low four bits) to a page protection mask..
209 extern pgprot_t protection_map
[16];
211 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
212 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
213 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
214 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
215 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
216 #define FAULT_FLAG_TRIED 0x20 /* Second try */
217 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
220 * vm_fault is filled by the the pagefault handler and passed to the vma's
221 * ->fault function. The vma's ->fault is responsible for returning a bitmask
222 * of VM_FAULT_xxx flags that give details about how the fault was handled.
224 * pgoff should be used in favour of virtual_address, if possible.
227 unsigned int flags
; /* FAULT_FLAG_xxx flags */
228 pgoff_t pgoff
; /* Logical page offset based on vma */
229 void __user
*virtual_address
; /* Faulting virtual address */
231 struct page
*cow_page
; /* Handler may choose to COW */
232 struct page
*page
; /* ->fault handlers should return a
233 * page here, unless VM_FAULT_NOPAGE
234 * is set (which is also implied by
237 /* for ->map_pages() only */
238 pgoff_t max_pgoff
; /* map pages for offset from pgoff till
239 * max_pgoff inclusive */
240 pte_t
*pte
; /* pte entry associated with ->pgoff */
244 * These are the virtual MM functions - opening of an area, closing and
245 * unmapping it (needed to keep files on disk up-to-date etc), pointer
246 * to the functions called when a no-page or a wp-page exception occurs.
248 struct vm_operations_struct
{
249 void (*open
)(struct vm_area_struct
* area
);
250 void (*close
)(struct vm_area_struct
* area
);
251 int (*mremap
)(struct vm_area_struct
* area
);
252 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
253 int (*pmd_fault
)(struct vm_area_struct
*, unsigned long address
,
254 pmd_t
*, unsigned int flags
);
255 void (*map_pages
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
257 /* notification that a previously read-only page is about to become
258 * writable, if an error is returned it will cause a SIGBUS */
259 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
261 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
262 int (*pfn_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
264 /* called by access_process_vm when get_user_pages() fails, typically
265 * for use by special VMAs that can switch between memory and hardware
267 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
268 void *buf
, int len
, int write
);
270 /* Called by the /proc/PID/maps code to ask the vma whether it
271 * has a special name. Returning non-NULL will also cause this
272 * vma to be dumped unconditionally. */
273 const char *(*name
)(struct vm_area_struct
*vma
);
277 * set_policy() op must add a reference to any non-NULL @new mempolicy
278 * to hold the policy upon return. Caller should pass NULL @new to
279 * remove a policy and fall back to surrounding context--i.e. do not
280 * install a MPOL_DEFAULT policy, nor the task or system default
283 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
286 * get_policy() op must add reference [mpol_get()] to any policy at
287 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
288 * in mm/mempolicy.c will do this automatically.
289 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
290 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
291 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
292 * must return NULL--i.e., do not "fallback" to task or system default
295 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
299 * Called by vm_normal_page() for special PTEs to find the
300 * page for @addr. This is useful if the default behavior
301 * (using pte_page()) would not find the correct page.
303 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
310 #define page_private(page) ((page)->private)
311 #define set_page_private(page, v) ((page)->private = (v))
314 * FIXME: take this include out, include page-flags.h in
315 * files which need it (119 of them)
317 #include <linux/page-flags.h>
318 #include <linux/huge_mm.h>
321 * Methods to modify the page usage count.
323 * What counts for a page usage:
324 * - cache mapping (page->mapping)
325 * - private data (page->private)
326 * - page mapped in a task's page tables, each mapping
327 * is counted separately
329 * Also, many kernel routines increase the page count before a critical
330 * routine so they can be sure the page doesn't go away from under them.
334 * Drop a ref, return true if the refcount fell to zero (the page has no users)
336 static inline int put_page_testzero(struct page
*page
)
338 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) == 0, page
);
339 return atomic_dec_and_test(&page
->_count
);
343 * Try to grab a ref unless the page has a refcount of zero, return false if
345 * This can be called when MMU is off so it must not access
346 * any of the virtual mappings.
348 static inline int get_page_unless_zero(struct page
*page
)
350 return atomic_inc_not_zero(&page
->_count
);
353 extern int page_is_ram(unsigned long pfn
);
361 int region_intersects(resource_size_t offset
, size_t size
, const char *type
);
363 /* Support for virtually mapped pages */
364 struct page
*vmalloc_to_page(const void *addr
);
365 unsigned long vmalloc_to_pfn(const void *addr
);
368 * Determine if an address is within the vmalloc range
370 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
371 * is no special casing required.
373 static inline int is_vmalloc_addr(const void *x
)
376 unsigned long addr
= (unsigned long)x
;
378 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
384 extern int is_vmalloc_or_module_addr(const void *x
);
386 static inline int is_vmalloc_or_module_addr(const void *x
)
392 extern void kvfree(const void *addr
);
394 static inline void compound_lock(struct page
*page
)
396 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
397 VM_BUG_ON_PAGE(PageSlab(page
), page
);
398 bit_spin_lock(PG_compound_lock
, &page
->flags
);
402 static inline void compound_unlock(struct page
*page
)
404 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
405 VM_BUG_ON_PAGE(PageSlab(page
), page
);
406 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
410 static inline unsigned long compound_lock_irqsave(struct page
*page
)
412 unsigned long uninitialized_var(flags
);
413 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
414 local_irq_save(flags
);
420 static inline void compound_unlock_irqrestore(struct page
*page
,
423 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
424 compound_unlock(page
);
425 local_irq_restore(flags
);
429 static inline struct page
*compound_head_by_tail(struct page
*tail
)
431 struct page
*head
= tail
->first_page
;
434 * page->first_page may be a dangling pointer to an old
435 * compound page, so recheck that it is still a tail
436 * page before returning.
439 if (likely(PageTail(tail
)))
445 * Since either compound page could be dismantled asynchronously in THP
446 * or we access asynchronously arbitrary positioned struct page, there
447 * would be tail flag race. To handle this race, we should call
448 * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
450 static inline struct page
*compound_head(struct page
*page
)
452 if (unlikely(PageTail(page
)))
453 return compound_head_by_tail(page
);
458 * If we access compound page synchronously such as access to
459 * allocated page, there is no need to handle tail flag race, so we can
460 * check tail flag directly without any synchronization primitive.
462 static inline struct page
*compound_head_fast(struct page
*page
)
464 if (unlikely(PageTail(page
)))
465 return page
->first_page
;
470 * The atomic page->_mapcount, starts from -1: so that transitions
471 * both from it and to it can be tracked, using atomic_inc_and_test
472 * and atomic_add_negative(-1).
474 static inline void page_mapcount_reset(struct page
*page
)
476 atomic_set(&(page
)->_mapcount
, -1);
479 static inline int page_mapcount(struct page
*page
)
481 VM_BUG_ON_PAGE(PageSlab(page
), page
);
482 return atomic_read(&page
->_mapcount
) + 1;
485 static inline int page_count(struct page
*page
)
487 return atomic_read(&compound_head(page
)->_count
);
490 static inline bool __compound_tail_refcounted(struct page
*page
)
492 return PageAnon(page
) && !PageSlab(page
) && !PageHeadHuge(page
);
496 * This takes a head page as parameter and tells if the
497 * tail page reference counting can be skipped.
499 * For this to be safe, PageSlab and PageHeadHuge must remain true on
500 * any given page where they return true here, until all tail pins
501 * have been released.
503 static inline bool compound_tail_refcounted(struct page
*page
)
505 VM_BUG_ON_PAGE(!PageHead(page
), page
);
506 return __compound_tail_refcounted(page
);
509 static inline void get_huge_page_tail(struct page
*page
)
512 * __split_huge_page_refcount() cannot run from under us.
514 VM_BUG_ON_PAGE(!PageTail(page
), page
);
515 VM_BUG_ON_PAGE(page_mapcount(page
) < 0, page
);
516 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) != 0, page
);
517 if (compound_tail_refcounted(page
->first_page
))
518 atomic_inc(&page
->_mapcount
);
521 extern bool __get_page_tail(struct page
*page
);
523 static inline void get_page(struct page
*page
)
525 if (unlikely(PageTail(page
)))
526 if (likely(__get_page_tail(page
)))
529 * Getting a normal page or the head of a compound page
530 * requires to already have an elevated page->_count.
532 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) <= 0, page
);
533 atomic_inc(&page
->_count
);
536 static inline struct page
*virt_to_head_page(const void *x
)
538 struct page
*page
= virt_to_page(x
);
541 * We don't need to worry about synchronization of tail flag
542 * when we call virt_to_head_page() since it is only called for
543 * already allocated page and this page won't be freed until
544 * this virt_to_head_page() is finished. So use _fast variant.
546 return compound_head_fast(page
);
550 * Setup the page count before being freed into the page allocator for
551 * the first time (boot or memory hotplug)
553 static inline void init_page_count(struct page
*page
)
555 atomic_set(&page
->_count
, 1);
558 void put_page(struct page
*page
);
559 void put_pages_list(struct list_head
*pages
);
561 void split_page(struct page
*page
, unsigned int order
);
562 int split_free_page(struct page
*page
);
565 * Compound pages have a destructor function. Provide a
566 * prototype for that function and accessor functions.
567 * These are _only_ valid on the head of a PG_compound page.
570 static inline void set_compound_page_dtor(struct page
*page
,
571 compound_page_dtor
*dtor
)
573 page
[1].compound_dtor
= dtor
;
576 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
578 return page
[1].compound_dtor
;
581 static inline int compound_order(struct page
*page
)
585 return page
[1].compound_order
;
588 static inline void set_compound_order(struct page
*page
, unsigned long order
)
590 page
[1].compound_order
= order
;
595 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
596 * servicing faults for write access. In the normal case, do always want
597 * pte_mkwrite. But get_user_pages can cause write faults for mappings
598 * that do not have writing enabled, when used by access_process_vm.
600 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
602 if (likely(vma
->vm_flags
& VM_WRITE
))
603 pte
= pte_mkwrite(pte
);
607 void do_set_pte(struct vm_area_struct
*vma
, unsigned long address
,
608 struct page
*page
, pte_t
*pte
, bool write
, bool anon
);
612 * Multiple processes may "see" the same page. E.g. for untouched
613 * mappings of /dev/null, all processes see the same page full of
614 * zeroes, and text pages of executables and shared libraries have
615 * only one copy in memory, at most, normally.
617 * For the non-reserved pages, page_count(page) denotes a reference count.
618 * page_count() == 0 means the page is free. page->lru is then used for
619 * freelist management in the buddy allocator.
620 * page_count() > 0 means the page has been allocated.
622 * Pages are allocated by the slab allocator in order to provide memory
623 * to kmalloc and kmem_cache_alloc. In this case, the management of the
624 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
625 * unless a particular usage is carefully commented. (the responsibility of
626 * freeing the kmalloc memory is the caller's, of course).
628 * A page may be used by anyone else who does a __get_free_page().
629 * In this case, page_count still tracks the references, and should only
630 * be used through the normal accessor functions. The top bits of page->flags
631 * and page->virtual store page management information, but all other fields
632 * are unused and could be used privately, carefully. The management of this
633 * page is the responsibility of the one who allocated it, and those who have
634 * subsequently been given references to it.
636 * The other pages (we may call them "pagecache pages") are completely
637 * managed by the Linux memory manager: I/O, buffers, swapping etc.
638 * The following discussion applies only to them.
640 * A pagecache page contains an opaque `private' member, which belongs to the
641 * page's address_space. Usually, this is the address of a circular list of
642 * the page's disk buffers. PG_private must be set to tell the VM to call
643 * into the filesystem to release these pages.
645 * A page may belong to an inode's memory mapping. In this case, page->mapping
646 * is the pointer to the inode, and page->index is the file offset of the page,
647 * in units of PAGE_CACHE_SIZE.
649 * If pagecache pages are not associated with an inode, they are said to be
650 * anonymous pages. These may become associated with the swapcache, and in that
651 * case PG_swapcache is set, and page->private is an offset into the swapcache.
653 * In either case (swapcache or inode backed), the pagecache itself holds one
654 * reference to the page. Setting PG_private should also increment the
655 * refcount. The each user mapping also has a reference to the page.
657 * The pagecache pages are stored in a per-mapping radix tree, which is
658 * rooted at mapping->page_tree, and indexed by offset.
659 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
660 * lists, we instead now tag pages as dirty/writeback in the radix tree.
662 * All pagecache pages may be subject to I/O:
663 * - inode pages may need to be read from disk,
664 * - inode pages which have been modified and are MAP_SHARED may need
665 * to be written back to the inode on disk,
666 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
667 * modified may need to be swapped out to swap space and (later) to be read
672 * The zone field is never updated after free_area_init_core()
673 * sets it, so none of the operations on it need to be atomic.
676 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
677 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
678 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
679 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
680 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
683 * Define the bit shifts to access each section. For non-existent
684 * sections we define the shift as 0; that plus a 0 mask ensures
685 * the compiler will optimise away reference to them.
687 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
688 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
689 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
690 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
692 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
693 #ifdef NODE_NOT_IN_PAGE_FLAGS
694 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
695 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
696 SECTIONS_PGOFF : ZONES_PGOFF)
698 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
699 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
700 NODES_PGOFF : ZONES_PGOFF)
703 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
705 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
706 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
709 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
710 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
711 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
712 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
713 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
715 static inline enum zone_type
page_zonenum(const struct page
*page
)
717 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
720 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
721 #define SECTION_IN_PAGE_FLAGS
725 * The identification function is mainly used by the buddy allocator for
726 * determining if two pages could be buddies. We are not really identifying
727 * the zone since we could be using the section number id if we do not have
728 * node id available in page flags.
729 * We only guarantee that it will return the same value for two combinable
732 static inline int page_zone_id(struct page
*page
)
734 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
737 static inline int zone_to_nid(struct zone
*zone
)
746 #ifdef NODE_NOT_IN_PAGE_FLAGS
747 extern int page_to_nid(const struct page
*page
);
749 static inline int page_to_nid(const struct page
*page
)
751 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
755 #ifdef CONFIG_NUMA_BALANCING
756 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
758 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
761 static inline int cpupid_to_pid(int cpupid
)
763 return cpupid
& LAST__PID_MASK
;
766 static inline int cpupid_to_cpu(int cpupid
)
768 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
771 static inline int cpupid_to_nid(int cpupid
)
773 return cpu_to_node(cpupid_to_cpu(cpupid
));
776 static inline bool cpupid_pid_unset(int cpupid
)
778 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
781 static inline bool cpupid_cpu_unset(int cpupid
)
783 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
786 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
788 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
791 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
792 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
793 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
795 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
798 static inline int page_cpupid_last(struct page
*page
)
800 return page
->_last_cpupid
;
802 static inline void page_cpupid_reset_last(struct page
*page
)
804 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
807 static inline int page_cpupid_last(struct page
*page
)
809 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
812 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
814 static inline void page_cpupid_reset_last(struct page
*page
)
816 int cpupid
= (1 << LAST_CPUPID_SHIFT
) - 1;
818 page
->flags
&= ~(LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
);
819 page
->flags
|= (cpupid
& LAST_CPUPID_MASK
) << LAST_CPUPID_PGSHIFT
;
821 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
822 #else /* !CONFIG_NUMA_BALANCING */
823 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
825 return page_to_nid(page
); /* XXX */
828 static inline int page_cpupid_last(struct page
*page
)
830 return page_to_nid(page
); /* XXX */
833 static inline int cpupid_to_nid(int cpupid
)
838 static inline int cpupid_to_pid(int cpupid
)
843 static inline int cpupid_to_cpu(int cpupid
)
848 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
853 static inline bool cpupid_pid_unset(int cpupid
)
858 static inline void page_cpupid_reset_last(struct page
*page
)
862 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
866 #endif /* CONFIG_NUMA_BALANCING */
868 static inline struct zone
*page_zone(const struct page
*page
)
870 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
873 #ifdef SECTION_IN_PAGE_FLAGS
874 static inline void set_page_section(struct page
*page
, unsigned long section
)
876 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
877 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
880 static inline unsigned long page_to_section(const struct page
*page
)
882 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
886 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
888 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
889 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
892 static inline void set_page_node(struct page
*page
, unsigned long node
)
894 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
895 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
898 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
899 unsigned long node
, unsigned long pfn
)
901 set_page_zone(page
, zone
);
902 set_page_node(page
, node
);
903 #ifdef SECTION_IN_PAGE_FLAGS
904 set_page_section(page
, pfn_to_section_nr(pfn
));
909 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
911 return page
->mem_cgroup
;
914 static inline void set_page_memcg(struct page
*page
, struct mem_cgroup
*memcg
)
916 page
->mem_cgroup
= memcg
;
919 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
924 static inline void set_page_memcg(struct page
*page
, struct mem_cgroup
*memcg
)
930 * Some inline functions in vmstat.h depend on page_zone()
932 #include <linux/vmstat.h>
934 static __always_inline
void *lowmem_page_address(const struct page
*page
)
936 return __va(PFN_PHYS(page_to_pfn(page
)));
939 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
940 #define HASHED_PAGE_VIRTUAL
943 #if defined(WANT_PAGE_VIRTUAL)
944 static inline void *page_address(const struct page
*page
)
946 return page
->virtual;
948 static inline void set_page_address(struct page
*page
, void *address
)
950 page
->virtual = address
;
952 #define page_address_init() do { } while(0)
955 #if defined(HASHED_PAGE_VIRTUAL)
956 void *page_address(const struct page
*page
);
957 void set_page_address(struct page
*page
, void *virtual);
958 void page_address_init(void);
961 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
962 #define page_address(page) lowmem_page_address(page)
963 #define set_page_address(page, address) do { } while(0)
964 #define page_address_init() do { } while(0)
967 extern void *page_rmapping(struct page
*page
);
968 extern struct anon_vma
*page_anon_vma(struct page
*page
);
969 extern struct address_space
*page_mapping(struct page
*page
);
971 extern struct address_space
*__page_file_mapping(struct page
*);
974 struct address_space
*page_file_mapping(struct page
*page
)
976 if (unlikely(PageSwapCache(page
)))
977 return __page_file_mapping(page
);
979 return page
->mapping
;
983 * Return the pagecache index of the passed page. Regular pagecache pages
984 * use ->index whereas swapcache pages use ->private
986 static inline pgoff_t
page_index(struct page
*page
)
988 if (unlikely(PageSwapCache(page
)))
989 return page_private(page
);
993 extern pgoff_t
__page_file_index(struct page
*page
);
996 * Return the file index of the page. Regular pagecache pages use ->index
997 * whereas swapcache pages use swp_offset(->private)
999 static inline pgoff_t
page_file_index(struct page
*page
)
1001 if (unlikely(PageSwapCache(page
)))
1002 return __page_file_index(page
);
1008 * Return true if this page is mapped into pagetables.
1010 static inline int page_mapped(struct page
*page
)
1012 return atomic_read(&(page
)->_mapcount
) >= 0;
1016 * Return true only if the page has been allocated with
1017 * ALLOC_NO_WATERMARKS and the low watermark was not
1018 * met implying that the system is under some pressure.
1020 static inline bool page_is_pfmemalloc(struct page
*page
)
1023 * Page index cannot be this large so this must be
1024 * a pfmemalloc page.
1026 return page
->index
== -1UL;
1030 * Only to be called by the page allocator on a freshly allocated
1033 static inline void set_page_pfmemalloc(struct page
*page
)
1038 static inline void clear_page_pfmemalloc(struct page
*page
)
1044 * Different kinds of faults, as returned by handle_mm_fault().
1045 * Used to decide whether a process gets delivered SIGBUS or
1046 * just gets major/minor fault counters bumped up.
1049 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1051 #define VM_FAULT_OOM 0x0001
1052 #define VM_FAULT_SIGBUS 0x0002
1053 #define VM_FAULT_MAJOR 0x0004
1054 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1055 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1056 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1057 #define VM_FAULT_SIGSEGV 0x0040
1059 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1060 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1061 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1062 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1064 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1066 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1067 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1070 /* Encode hstate index for a hwpoisoned large page */
1071 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1072 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1075 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1077 extern void pagefault_out_of_memory(void);
1079 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1082 * Flags passed to show_mem() and show_free_areas() to suppress output in
1085 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1087 extern void show_free_areas(unsigned int flags
);
1088 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
1090 int shmem_zero_setup(struct vm_area_struct
*);
1092 bool shmem_mapping(struct address_space
*mapping
);
1094 static inline bool shmem_mapping(struct address_space
*mapping
)
1100 extern int can_do_mlock(void);
1101 extern int user_shm_lock(size_t, struct user_struct
*);
1102 extern void user_shm_unlock(size_t, struct user_struct
*);
1105 * Parameter block passed down to zap_pte_range in exceptional cases.
1107 struct zap_details
{
1108 struct address_space
*check_mapping
; /* Check page->mapping if set */
1109 pgoff_t first_index
; /* Lowest page->index to unmap */
1110 pgoff_t last_index
; /* Highest page->index to unmap */
1113 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1116 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1117 unsigned long size
);
1118 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1119 unsigned long size
, struct zap_details
*);
1120 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1121 unsigned long start
, unsigned long end
);
1124 * mm_walk - callbacks for walk_page_range
1125 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1126 * this handler is required to be able to handle
1127 * pmd_trans_huge() pmds. They may simply choose to
1128 * split_huge_page() instead of handling it explicitly.
1129 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1130 * @pte_hole: if set, called for each hole at all levels
1131 * @hugetlb_entry: if set, called for each hugetlb entry
1132 * @test_walk: caller specific callback function to determine whether
1133 * we walk over the current vma or not. A positive returned
1134 * value means "do page table walk over the current vma,"
1135 * and a negative one means "abort current page table walk
1136 * right now." 0 means "skip the current vma."
1137 * @mm: mm_struct representing the target process of page table walk
1138 * @vma: vma currently walked (NULL if walking outside vmas)
1139 * @private: private data for callbacks' usage
1141 * (see the comment on walk_page_range() for more details)
1144 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1145 unsigned long next
, struct mm_walk
*walk
);
1146 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1147 unsigned long next
, struct mm_walk
*walk
);
1148 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1149 struct mm_walk
*walk
);
1150 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1151 unsigned long addr
, unsigned long next
,
1152 struct mm_walk
*walk
);
1153 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1154 struct mm_walk
*walk
);
1155 struct mm_struct
*mm
;
1156 struct vm_area_struct
*vma
;
1160 int walk_page_range(unsigned long addr
, unsigned long end
,
1161 struct mm_walk
*walk
);
1162 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1163 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1164 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1165 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1166 struct vm_area_struct
*vma
);
1167 void unmap_mapping_range(struct address_space
*mapping
,
1168 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1169 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1170 unsigned long *pfn
);
1171 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1172 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1173 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1174 void *buf
, int len
, int write
);
1176 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1177 loff_t
const holebegin
, loff_t
const holelen
)
1179 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1182 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1183 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1184 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1185 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1186 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1187 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1188 int invalidate_inode_page(struct page
*page
);
1191 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1192 unsigned long address
, unsigned int flags
);
1193 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1194 unsigned long address
, unsigned int fault_flags
);
1196 static inline int handle_mm_fault(struct mm_struct
*mm
,
1197 struct vm_area_struct
*vma
, unsigned long address
,
1200 /* should never happen if there's no MMU */
1202 return VM_FAULT_SIGBUS
;
1204 static inline int fixup_user_fault(struct task_struct
*tsk
,
1205 struct mm_struct
*mm
, unsigned long address
,
1206 unsigned int fault_flags
)
1208 /* should never happen if there's no MMU */
1214 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1215 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1216 void *buf
, int len
, int write
);
1218 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1219 unsigned long start
, unsigned long nr_pages
,
1220 unsigned int foll_flags
, struct page
**pages
,
1221 struct vm_area_struct
**vmas
, int *nonblocking
);
1222 long get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1223 unsigned long start
, unsigned long nr_pages
,
1224 int write
, int force
, struct page
**pages
,
1225 struct vm_area_struct
**vmas
);
1226 long get_user_pages_locked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1227 unsigned long start
, unsigned long nr_pages
,
1228 int write
, int force
, struct page
**pages
,
1230 long __get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1231 unsigned long start
, unsigned long nr_pages
,
1232 int write
, int force
, struct page
**pages
,
1233 unsigned int gup_flags
);
1234 long get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1235 unsigned long start
, unsigned long nr_pages
,
1236 int write
, int force
, struct page
**pages
);
1237 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1238 struct page
**pages
);
1240 /* Container for pinned pfns / pages */
1241 struct frame_vector
{
1242 unsigned int nr_allocated
; /* Number of frames we have space for */
1243 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1244 bool got_ref
; /* Did we pin pages by getting page ref? */
1245 bool is_pfns
; /* Does array contain pages or pfns? */
1246 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1247 * pfns_vector_pages() or pfns_vector_pfns()
1251 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1252 void frame_vector_destroy(struct frame_vector
*vec
);
1253 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1254 bool write
, bool force
, struct frame_vector
*vec
);
1255 void put_vaddr_frames(struct frame_vector
*vec
);
1256 int frame_vector_to_pages(struct frame_vector
*vec
);
1257 void frame_vector_to_pfns(struct frame_vector
*vec
);
1259 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1261 return vec
->nr_frames
;
1264 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1267 int err
= frame_vector_to_pages(vec
);
1270 return ERR_PTR(err
);
1272 return (struct page
**)(vec
->ptrs
);
1275 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1278 frame_vector_to_pfns(vec
);
1279 return (unsigned long *)(vec
->ptrs
);
1283 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1284 struct page
**pages
);
1285 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1286 struct page
*get_dump_page(unsigned long addr
);
1288 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1289 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1290 unsigned int length
);
1292 int __set_page_dirty_nobuffers(struct page
*page
);
1293 int __set_page_dirty_no_writeback(struct page
*page
);
1294 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1296 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
,
1297 struct mem_cgroup
*memcg
);
1298 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1299 struct mem_cgroup
*memcg
, struct bdi_writeback
*wb
);
1300 int set_page_dirty(struct page
*page
);
1301 int set_page_dirty_lock(struct page
*page
);
1302 void cancel_dirty_page(struct page
*page
);
1303 int clear_page_dirty_for_io(struct page
*page
);
1305 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1307 /* Is the vma a continuation of the stack vma above it? */
1308 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1310 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1313 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1315 return !vma
->vm_ops
;
1318 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1321 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1322 (vma
->vm_start
== addr
) &&
1323 !vma_growsdown(vma
->vm_prev
, addr
);
1326 /* Is the vma a continuation of the stack vma below it? */
1327 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1329 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1332 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1335 return (vma
->vm_flags
& VM_GROWSUP
) &&
1336 (vma
->vm_end
== addr
) &&
1337 !vma_growsup(vma
->vm_next
, addr
);
1340 extern struct task_struct
*task_of_stack(struct task_struct
*task
,
1341 struct vm_area_struct
*vma
, bool in_group
);
1343 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1344 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1345 unsigned long new_addr
, unsigned long len
,
1346 bool need_rmap_locks
);
1347 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1348 unsigned long end
, pgprot_t newprot
,
1349 int dirty_accountable
, int prot_numa
);
1350 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1351 struct vm_area_struct
**pprev
, unsigned long start
,
1352 unsigned long end
, unsigned long newflags
);
1355 * doesn't attempt to fault and will return short.
1357 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1358 struct page
**pages
);
1360 * per-process(per-mm_struct) statistics.
1362 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1364 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1366 #ifdef SPLIT_RSS_COUNTING
1368 * counter is updated in asynchronous manner and may go to minus.
1369 * But it's never be expected number for users.
1374 return (unsigned long)val
;
1377 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1379 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1382 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1384 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1387 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1389 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1392 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1394 return get_mm_counter(mm
, MM_FILEPAGES
) +
1395 get_mm_counter(mm
, MM_ANONPAGES
);
1398 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1400 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1403 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1405 return max(mm
->hiwater_vm
, mm
->total_vm
);
1408 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1410 unsigned long _rss
= get_mm_rss(mm
);
1412 if ((mm
)->hiwater_rss
< _rss
)
1413 (mm
)->hiwater_rss
= _rss
;
1416 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1418 if (mm
->hiwater_vm
< mm
->total_vm
)
1419 mm
->hiwater_vm
= mm
->total_vm
;
1422 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1424 mm
->hiwater_rss
= get_mm_rss(mm
);
1427 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1428 struct mm_struct
*mm
)
1430 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1432 if (*maxrss
< hiwater_rss
)
1433 *maxrss
= hiwater_rss
;
1436 #if defined(SPLIT_RSS_COUNTING)
1437 void sync_mm_rss(struct mm_struct
*mm
);
1439 static inline void sync_mm_rss(struct mm_struct
*mm
)
1444 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1446 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1448 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1452 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1456 #ifdef __PAGETABLE_PUD_FOLDED
1457 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1458 unsigned long address
)
1463 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1466 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1467 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1468 unsigned long address
)
1473 static inline void mm_nr_pmds_init(struct mm_struct
*mm
) {}
1475 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1480 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1481 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1484 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1486 static inline void mm_nr_pmds_init(struct mm_struct
*mm
)
1488 atomic_long_set(&mm
->nr_pmds
, 0);
1491 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1493 return atomic_long_read(&mm
->nr_pmds
);
1496 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1498 atomic_long_inc(&mm
->nr_pmds
);
1501 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1503 atomic_long_dec(&mm
->nr_pmds
);
1507 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1508 pmd_t
*pmd
, unsigned long address
);
1509 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1512 * The following ifdef needed to get the 4level-fixup.h header to work.
1513 * Remove it when 4level-fixup.h has been removed.
1515 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1516 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1518 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1519 NULL
: pud_offset(pgd
, address
);
1522 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1524 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1525 NULL
: pmd_offset(pud
, address
);
1527 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1529 #if USE_SPLIT_PTE_PTLOCKS
1530 #if ALLOC_SPLIT_PTLOCKS
1531 void __init
ptlock_cache_init(void);
1532 extern bool ptlock_alloc(struct page
*page
);
1533 extern void ptlock_free(struct page
*page
);
1535 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1539 #else /* ALLOC_SPLIT_PTLOCKS */
1540 static inline void ptlock_cache_init(void)
1544 static inline bool ptlock_alloc(struct page
*page
)
1549 static inline void ptlock_free(struct page
*page
)
1553 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1557 #endif /* ALLOC_SPLIT_PTLOCKS */
1559 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1561 return ptlock_ptr(pmd_page(*pmd
));
1564 static inline bool ptlock_init(struct page
*page
)
1567 * prep_new_page() initialize page->private (and therefore page->ptl)
1568 * with 0. Make sure nobody took it in use in between.
1570 * It can happen if arch try to use slab for page table allocation:
1571 * slab code uses page->slab_cache and page->first_page (for tail
1572 * pages), which share storage with page->ptl.
1574 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1575 if (!ptlock_alloc(page
))
1577 spin_lock_init(ptlock_ptr(page
));
1581 /* Reset page->mapping so free_pages_check won't complain. */
1582 static inline void pte_lock_deinit(struct page
*page
)
1584 page
->mapping
= NULL
;
1588 #else /* !USE_SPLIT_PTE_PTLOCKS */
1590 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1592 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1594 return &mm
->page_table_lock
;
1596 static inline void ptlock_cache_init(void) {}
1597 static inline bool ptlock_init(struct page
*page
) { return true; }
1598 static inline void pte_lock_deinit(struct page
*page
) {}
1599 #endif /* USE_SPLIT_PTE_PTLOCKS */
1601 static inline void pgtable_init(void)
1603 ptlock_cache_init();
1604 pgtable_cache_init();
1607 static inline bool pgtable_page_ctor(struct page
*page
)
1609 inc_zone_page_state(page
, NR_PAGETABLE
);
1610 return ptlock_init(page
);
1613 static inline void pgtable_page_dtor(struct page
*page
)
1615 pte_lock_deinit(page
);
1616 dec_zone_page_state(page
, NR_PAGETABLE
);
1619 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1621 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1622 pte_t *__pte = pte_offset_map(pmd, address); \
1628 #define pte_unmap_unlock(pte, ptl) do { \
1633 #define pte_alloc_map(mm, vma, pmd, address) \
1634 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1636 NULL: pte_offset_map(pmd, address))
1638 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1639 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1641 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1643 #define pte_alloc_kernel(pmd, address) \
1644 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1645 NULL: pte_offset_kernel(pmd, address))
1647 #if USE_SPLIT_PMD_PTLOCKS
1649 static struct page
*pmd_to_page(pmd_t
*pmd
)
1651 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1652 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1655 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1657 return ptlock_ptr(pmd_to_page(pmd
));
1660 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1662 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1663 page
->pmd_huge_pte
= NULL
;
1665 return ptlock_init(page
);
1668 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1670 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1671 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1676 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1680 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1682 return &mm
->page_table_lock
;
1685 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1686 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1688 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1692 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1694 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1699 extern void free_area_init(unsigned long * zones_size
);
1700 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1701 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1702 extern void free_initmem(void);
1705 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1706 * into the buddy system. The freed pages will be poisoned with pattern
1707 * "poison" if it's within range [0, UCHAR_MAX].
1708 * Return pages freed into the buddy system.
1710 extern unsigned long free_reserved_area(void *start
, void *end
,
1711 int poison
, char *s
);
1713 #ifdef CONFIG_HIGHMEM
1715 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1716 * and totalram_pages.
1718 extern void free_highmem_page(struct page
*page
);
1721 extern void adjust_managed_page_count(struct page
*page
, long count
);
1722 extern void mem_init_print_info(const char *str
);
1724 extern void reserve_bootmem_region(unsigned long start
, unsigned long end
);
1726 /* Free the reserved page into the buddy system, so it gets managed. */
1727 static inline void __free_reserved_page(struct page
*page
)
1729 ClearPageReserved(page
);
1730 init_page_count(page
);
1734 static inline void free_reserved_page(struct page
*page
)
1736 __free_reserved_page(page
);
1737 adjust_managed_page_count(page
, 1);
1740 static inline void mark_page_reserved(struct page
*page
)
1742 SetPageReserved(page
);
1743 adjust_managed_page_count(page
, -1);
1747 * Default method to free all the __init memory into the buddy system.
1748 * The freed pages will be poisoned with pattern "poison" if it's within
1749 * range [0, UCHAR_MAX].
1750 * Return pages freed into the buddy system.
1752 static inline unsigned long free_initmem_default(int poison
)
1754 extern char __init_begin
[], __init_end
[];
1756 return free_reserved_area(&__init_begin
, &__init_end
,
1757 poison
, "unused kernel");
1760 static inline unsigned long get_num_physpages(void)
1763 unsigned long phys_pages
= 0;
1765 for_each_online_node(nid
)
1766 phys_pages
+= node_present_pages(nid
);
1771 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1773 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1774 * zones, allocate the backing mem_map and account for memory holes in a more
1775 * architecture independent manner. This is a substitute for creating the
1776 * zone_sizes[] and zholes_size[] arrays and passing them to
1777 * free_area_init_node()
1779 * An architecture is expected to register range of page frames backed by
1780 * physical memory with memblock_add[_node]() before calling
1781 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1782 * usage, an architecture is expected to do something like
1784 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1786 * for_each_valid_physical_page_range()
1787 * memblock_add_node(base, size, nid)
1788 * free_area_init_nodes(max_zone_pfns);
1790 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1791 * registered physical page range. Similarly
1792 * sparse_memory_present_with_active_regions() calls memory_present() for
1793 * each range when SPARSEMEM is enabled.
1795 * See mm/page_alloc.c for more information on each function exposed by
1796 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1798 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1799 unsigned long node_map_pfn_alignment(void);
1800 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1801 unsigned long end_pfn
);
1802 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1803 unsigned long end_pfn
);
1804 extern void get_pfn_range_for_nid(unsigned int nid
,
1805 unsigned long *start_pfn
, unsigned long *end_pfn
);
1806 extern unsigned long find_min_pfn_with_active_regions(void);
1807 extern void free_bootmem_with_active_regions(int nid
,
1808 unsigned long max_low_pfn
);
1809 extern void sparse_memory_present_with_active_regions(int nid
);
1811 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1813 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1814 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1815 static inline int __early_pfn_to_nid(unsigned long pfn
,
1816 struct mminit_pfnnid_cache
*state
)
1821 /* please see mm/page_alloc.c */
1822 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1823 /* there is a per-arch backend function. */
1824 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
1825 struct mminit_pfnnid_cache
*state
);
1828 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1829 extern void memmap_init_zone(unsigned long, int, unsigned long,
1830 unsigned long, enum memmap_context
);
1831 extern void setup_per_zone_wmarks(void);
1832 extern int __meminit
init_per_zone_wmark_min(void);
1833 extern void mem_init(void);
1834 extern void __init
mmap_init(void);
1835 extern void show_mem(unsigned int flags
);
1836 extern void si_meminfo(struct sysinfo
* val
);
1837 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1839 extern __printf(3, 4)
1840 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1842 extern void setup_per_cpu_pageset(void);
1844 extern void zone_pcp_update(struct zone
*zone
);
1845 extern void zone_pcp_reset(struct zone
*zone
);
1848 extern int min_free_kbytes
;
1851 extern atomic_long_t mmap_pages_allocated
;
1852 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1854 /* interval_tree.c */
1855 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1856 struct rb_root
*root
);
1857 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1858 struct vm_area_struct
*prev
,
1859 struct rb_root
*root
);
1860 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1861 struct rb_root
*root
);
1862 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1863 unsigned long start
, unsigned long last
);
1864 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1865 unsigned long start
, unsigned long last
);
1867 #define vma_interval_tree_foreach(vma, root, start, last) \
1868 for (vma = vma_interval_tree_iter_first(root, start, last); \
1869 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1871 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1872 struct rb_root
*root
);
1873 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1874 struct rb_root
*root
);
1875 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1876 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1877 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1878 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1879 #ifdef CONFIG_DEBUG_VM_RB
1880 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1883 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1884 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1885 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1888 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1889 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1890 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1891 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1892 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1893 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1894 struct mempolicy
*, struct vm_userfaultfd_ctx
);
1895 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1896 extern int split_vma(struct mm_struct
*,
1897 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1898 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1899 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1900 struct rb_node
**, struct rb_node
*);
1901 extern void unlink_file_vma(struct vm_area_struct
*);
1902 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1903 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1904 bool *need_rmap_locks
);
1905 extern void exit_mmap(struct mm_struct
*);
1907 static inline int check_data_rlimit(unsigned long rlim
,
1909 unsigned long start
,
1910 unsigned long end_data
,
1911 unsigned long start_data
)
1913 if (rlim
< RLIM_INFINITY
) {
1914 if (((new - start
) + (end_data
- start_data
)) > rlim
)
1921 extern int mm_take_all_locks(struct mm_struct
*mm
);
1922 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1924 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1925 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1927 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1928 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
1929 unsigned long addr
, unsigned long len
,
1930 unsigned long flags
,
1931 const struct vm_special_mapping
*spec
);
1932 /* This is an obsolete alternative to _install_special_mapping. */
1933 extern int install_special_mapping(struct mm_struct
*mm
,
1934 unsigned long addr
, unsigned long len
,
1935 unsigned long flags
, struct page
**pages
);
1937 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1939 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1940 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
1941 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1942 unsigned long len
, unsigned long prot
, unsigned long flags
,
1943 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
);
1944 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1946 static inline unsigned long
1947 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1948 unsigned long len
, unsigned long prot
, unsigned long flags
,
1949 unsigned long pgoff
, unsigned long *populate
)
1951 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
);
1955 extern int __mm_populate(unsigned long addr
, unsigned long len
,
1957 static inline void mm_populate(unsigned long addr
, unsigned long len
)
1960 (void) __mm_populate(addr
, len
, 1);
1963 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
1966 /* These take the mm semaphore themselves */
1967 extern unsigned long vm_brk(unsigned long, unsigned long);
1968 extern int vm_munmap(unsigned long, size_t);
1969 extern unsigned long vm_mmap(struct file
*, unsigned long,
1970 unsigned long, unsigned long,
1971 unsigned long, unsigned long);
1973 struct vm_unmapped_area_info
{
1974 #define VM_UNMAPPED_AREA_TOPDOWN 1
1975 unsigned long flags
;
1976 unsigned long length
;
1977 unsigned long low_limit
;
1978 unsigned long high_limit
;
1979 unsigned long align_mask
;
1980 unsigned long align_offset
;
1983 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
1984 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
1987 * Search for an unmapped address range.
1989 * We are looking for a range that:
1990 * - does not intersect with any VMA;
1991 * - is contained within the [low_limit, high_limit) interval;
1992 * - is at least the desired size.
1993 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1995 static inline unsigned long
1996 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
1998 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
1999 return unmapped_area_topdown(info
);
2001 return unmapped_area(info
);
2005 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2006 extern void truncate_inode_pages_range(struct address_space
*,
2007 loff_t lstart
, loff_t lend
);
2008 extern void truncate_inode_pages_final(struct address_space
*);
2010 /* generic vm_area_ops exported for stackable file systems */
2011 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
2012 extern void filemap_map_pages(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
2013 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
2015 /* mm/page-writeback.c */
2016 int write_one_page(struct page
*page
, int wait
);
2017 void task_dirty_inc(struct task_struct
*tsk
);
2020 #define VM_MAX_READAHEAD 128 /* kbytes */
2021 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2023 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2024 pgoff_t offset
, unsigned long nr_to_read
);
2026 void page_cache_sync_readahead(struct address_space
*mapping
,
2027 struct file_ra_state
*ra
,
2030 unsigned long size
);
2032 void page_cache_async_readahead(struct address_space
*mapping
,
2033 struct file_ra_state
*ra
,
2037 unsigned long size
);
2039 unsigned long max_sane_readahead(unsigned long nr
);
2041 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2042 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2044 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2045 extern int expand_downwards(struct vm_area_struct
*vma
,
2046 unsigned long address
);
2048 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2050 #define expand_upwards(vma, address) (0)
2053 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2054 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2055 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2056 struct vm_area_struct
**pprev
);
2058 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2059 NULL if none. Assume start_addr < end_addr. */
2060 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2062 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2064 if (vma
&& end_addr
<= vma
->vm_start
)
2069 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2071 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2074 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2075 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2076 unsigned long vm_start
, unsigned long vm_end
)
2078 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2080 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2087 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2088 void vma_set_page_prot(struct vm_area_struct
*vma
);
2090 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2094 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2096 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2100 #ifdef CONFIG_NUMA_BALANCING
2101 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2102 unsigned long start
, unsigned long end
);
2105 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2106 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2107 unsigned long pfn
, unsigned long size
, pgprot_t
);
2108 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2109 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2111 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2113 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2116 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2117 unsigned long address
, unsigned int foll_flags
,
2118 unsigned int *page_mask
);
2120 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2121 unsigned long address
, unsigned int foll_flags
)
2123 unsigned int unused_page_mask
;
2124 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2127 #define FOLL_WRITE 0x01 /* check pte is writable */
2128 #define FOLL_TOUCH 0x02 /* mark page accessed */
2129 #define FOLL_GET 0x04 /* do get_page on page */
2130 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2131 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2132 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2133 * and return without waiting upon it */
2134 #define FOLL_POPULATE 0x40 /* fault in page */
2135 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2136 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2137 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2138 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2139 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2141 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2143 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2144 unsigned long size
, pte_fn_t fn
, void *data
);
2146 #ifdef CONFIG_PROC_FS
2147 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
2149 static inline void vm_stat_account(struct mm_struct
*mm
,
2150 unsigned long flags
, struct file
*file
, long pages
)
2152 mm
->total_vm
+= pages
;
2154 #endif /* CONFIG_PROC_FS */
2156 #ifdef CONFIG_DEBUG_PAGEALLOC
2157 extern bool _debug_pagealloc_enabled
;
2158 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2160 static inline bool debug_pagealloc_enabled(void)
2162 return _debug_pagealloc_enabled
;
2166 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2168 if (!debug_pagealloc_enabled())
2171 __kernel_map_pages(page
, numpages
, enable
);
2173 #ifdef CONFIG_HIBERNATION
2174 extern bool kernel_page_present(struct page
*page
);
2175 #endif /* CONFIG_HIBERNATION */
2178 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2179 #ifdef CONFIG_HIBERNATION
2180 static inline bool kernel_page_present(struct page
*page
) { return true; }
2181 #endif /* CONFIG_HIBERNATION */
2184 #ifdef __HAVE_ARCH_GATE_AREA
2185 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2186 extern int in_gate_area_no_mm(unsigned long addr
);
2187 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2189 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2193 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2194 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2198 #endif /* __HAVE_ARCH_GATE_AREA */
2200 #ifdef CONFIG_SYSCTL
2201 extern int sysctl_drop_caches
;
2202 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2203 void __user
*, size_t *, loff_t
*);
2206 void drop_slab(void);
2207 void drop_slab_node(int nid
);
2210 #define randomize_va_space 0
2212 extern int randomize_va_space
;
2215 const char * arch_vma_name(struct vm_area_struct
*vma
);
2216 void print_vma_addr(char *prefix
, unsigned long rip
);
2218 void sparse_mem_maps_populate_node(struct page
**map_map
,
2219 unsigned long pnum_begin
,
2220 unsigned long pnum_end
,
2221 unsigned long map_count
,
2224 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2225 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2226 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2227 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2228 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2229 void *vmemmap_alloc_block(unsigned long size
, int node
);
2230 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
2231 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2232 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2234 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2235 void vmemmap_populate_print_last(void);
2236 #ifdef CONFIG_MEMORY_HOTPLUG
2237 void vmemmap_free(unsigned long start
, unsigned long end
);
2239 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2240 unsigned long size
);
2243 MF_COUNT_INCREASED
= 1 << 0,
2244 MF_ACTION_REQUIRED
= 1 << 1,
2245 MF_MUST_KILL
= 1 << 2,
2246 MF_SOFT_OFFLINE
= 1 << 3,
2248 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2249 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2250 extern int unpoison_memory(unsigned long pfn
);
2251 extern int get_hwpoison_page(struct page
*page
);
2252 extern void put_hwpoison_page(struct page
*page
);
2253 extern int sysctl_memory_failure_early_kill
;
2254 extern int sysctl_memory_failure_recovery
;
2255 extern void shake_page(struct page
*p
, int access
);
2256 extern atomic_long_t num_poisoned_pages
;
2257 extern int soft_offline_page(struct page
*page
, int flags
);
2261 * Error handlers for various types of pages.
2264 MF_IGNORED
, /* Error: cannot be handled */
2265 MF_FAILED
, /* Error: handling failed */
2266 MF_DELAYED
, /* Will be handled later */
2267 MF_RECOVERED
, /* Successfully recovered */
2270 enum mf_action_page_type
{
2272 MF_MSG_KERNEL_HIGH_ORDER
,
2274 MF_MSG_DIFFERENT_COMPOUND
,
2275 MF_MSG_POISONED_HUGE
,
2278 MF_MSG_UNMAP_FAILED
,
2279 MF_MSG_DIRTY_SWAPCACHE
,
2280 MF_MSG_CLEAN_SWAPCACHE
,
2281 MF_MSG_DIRTY_MLOCKED_LRU
,
2282 MF_MSG_CLEAN_MLOCKED_LRU
,
2283 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2284 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2287 MF_MSG_TRUNCATED_LRU
,
2293 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2294 extern void clear_huge_page(struct page
*page
,
2296 unsigned int pages_per_huge_page
);
2297 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2298 unsigned long addr
, struct vm_area_struct
*vma
,
2299 unsigned int pages_per_huge_page
);
2300 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2302 extern struct page_ext_operations debug_guardpage_ops
;
2303 extern struct page_ext_operations page_poisoning_ops
;
2305 #ifdef CONFIG_DEBUG_PAGEALLOC
2306 extern unsigned int _debug_guardpage_minorder
;
2307 extern bool _debug_guardpage_enabled
;
2309 static inline unsigned int debug_guardpage_minorder(void)
2311 return _debug_guardpage_minorder
;
2314 static inline bool debug_guardpage_enabled(void)
2316 return _debug_guardpage_enabled
;
2319 static inline bool page_is_guard(struct page
*page
)
2321 struct page_ext
*page_ext
;
2323 if (!debug_guardpage_enabled())
2326 page_ext
= lookup_page_ext(page
);
2327 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2330 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2331 static inline bool debug_guardpage_enabled(void) { return false; }
2332 static inline bool page_is_guard(struct page
*page
) { return false; }
2333 #endif /* CONFIG_DEBUG_PAGEALLOC */
2335 #if MAX_NUMNODES > 1
2336 void __init
setup_nr_node_ids(void);
2338 static inline void setup_nr_node_ids(void) {}
2341 #endif /* __KERNEL__ */
2342 #endif /* _LINUX_MM_H */