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_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
143 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
144 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
145 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
146 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
147 #define VM_ARCH_2 0x02000000
148 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
150 #ifdef CONFIG_MEM_SOFT_DIRTY
151 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
153 # define VM_SOFTDIRTY 0
156 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
157 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
158 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
159 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
161 #if defined(CONFIG_X86)
162 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
163 #elif defined(CONFIG_PPC)
164 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
165 #elif defined(CONFIG_PARISC)
166 # define VM_GROWSUP VM_ARCH_1
167 #elif defined(CONFIG_METAG)
168 # define VM_GROWSUP VM_ARCH_1
169 #elif defined(CONFIG_IA64)
170 # define VM_GROWSUP VM_ARCH_1
171 #elif !defined(CONFIG_MMU)
172 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
175 #if defined(CONFIG_X86)
176 /* MPX specific bounds table or bounds directory */
177 # define VM_MPX VM_ARCH_2
181 # define VM_GROWSUP VM_NONE
184 /* Bits set in the VMA until the stack is in its final location */
185 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
187 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
188 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
191 #ifdef CONFIG_STACK_GROWSUP
192 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
194 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
198 * Special vmas that are non-mergable, non-mlock()able.
199 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
201 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
203 /* This mask defines which mm->def_flags a process can inherit its parent */
204 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
206 /* This mask is used to clear all the VMA flags used by mlock */
207 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
210 * mapping from the currently active vm_flags protection bits (the
211 * low four bits) to a page protection mask..
213 extern pgprot_t protection_map
[16];
215 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
216 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
217 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
218 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
219 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
220 #define FAULT_FLAG_TRIED 0x20 /* Second try */
221 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
224 * vm_fault is filled by the the pagefault handler and passed to the vma's
225 * ->fault function. The vma's ->fault is responsible for returning a bitmask
226 * of VM_FAULT_xxx flags that give details about how the fault was handled.
228 * pgoff should be used in favour of virtual_address, if possible.
231 unsigned int flags
; /* FAULT_FLAG_xxx flags */
232 pgoff_t pgoff
; /* Logical page offset based on vma */
233 void __user
*virtual_address
; /* Faulting virtual address */
235 struct page
*cow_page
; /* Handler may choose to COW */
236 struct page
*page
; /* ->fault handlers should return a
237 * page here, unless VM_FAULT_NOPAGE
238 * is set (which is also implied by
241 /* for ->map_pages() only */
242 pgoff_t max_pgoff
; /* map pages for offset from pgoff till
243 * max_pgoff inclusive */
244 pte_t
*pte
; /* pte entry associated with ->pgoff */
248 * These are the virtual MM functions - opening of an area, closing and
249 * unmapping it (needed to keep files on disk up-to-date etc), pointer
250 * to the functions called when a no-page or a wp-page exception occurs.
252 struct vm_operations_struct
{
253 void (*open
)(struct vm_area_struct
* area
);
254 void (*close
)(struct vm_area_struct
* area
);
255 int (*mremap
)(struct vm_area_struct
* area
);
256 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
257 int (*pmd_fault
)(struct vm_area_struct
*, unsigned long address
,
258 pmd_t
*, unsigned int flags
);
259 void (*map_pages
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
261 /* notification that a previously read-only page is about to become
262 * writable, if an error is returned it will cause a SIGBUS */
263 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
265 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
266 int (*pfn_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
268 /* called by access_process_vm when get_user_pages() fails, typically
269 * for use by special VMAs that can switch between memory and hardware
271 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
272 void *buf
, int len
, int write
);
274 /* Called by the /proc/PID/maps code to ask the vma whether it
275 * has a special name. Returning non-NULL will also cause this
276 * vma to be dumped unconditionally. */
277 const char *(*name
)(struct vm_area_struct
*vma
);
281 * set_policy() op must add a reference to any non-NULL @new mempolicy
282 * to hold the policy upon return. Caller should pass NULL @new to
283 * remove a policy and fall back to surrounding context--i.e. do not
284 * install a MPOL_DEFAULT policy, nor the task or system default
287 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
290 * get_policy() op must add reference [mpol_get()] to any policy at
291 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
292 * in mm/mempolicy.c will do this automatically.
293 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
294 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
295 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
296 * must return NULL--i.e., do not "fallback" to task or system default
299 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
303 * Called by vm_normal_page() for special PTEs to find the
304 * page for @addr. This is useful if the default behavior
305 * (using pte_page()) would not find the correct page.
307 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
314 #define page_private(page) ((page)->private)
315 #define set_page_private(page, v) ((page)->private = (v))
318 * FIXME: take this include out, include page-flags.h in
319 * files which need it (119 of them)
321 #include <linux/page-flags.h>
322 #include <linux/huge_mm.h>
325 * Methods to modify the page usage count.
327 * What counts for a page usage:
328 * - cache mapping (page->mapping)
329 * - private data (page->private)
330 * - page mapped in a task's page tables, each mapping
331 * is counted separately
333 * Also, many kernel routines increase the page count before a critical
334 * routine so they can be sure the page doesn't go away from under them.
338 * Drop a ref, return true if the refcount fell to zero (the page has no users)
340 static inline int put_page_testzero(struct page
*page
)
342 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) == 0, page
);
343 return atomic_dec_and_test(&page
->_count
);
347 * Try to grab a ref unless the page has a refcount of zero, return false if
349 * This can be called when MMU is off so it must not access
350 * any of the virtual mappings.
352 static inline int get_page_unless_zero(struct page
*page
)
354 return atomic_inc_not_zero(&page
->_count
);
357 extern int page_is_ram(unsigned long pfn
);
365 int region_intersects(resource_size_t offset
, size_t size
, const char *type
);
367 /* Support for virtually mapped pages */
368 struct page
*vmalloc_to_page(const void *addr
);
369 unsigned long vmalloc_to_pfn(const void *addr
);
372 * Determine if an address is within the vmalloc range
374 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
375 * is no special casing required.
377 static inline int is_vmalloc_addr(const void *x
)
380 unsigned long addr
= (unsigned long)x
;
382 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
388 extern int is_vmalloc_or_module_addr(const void *x
);
390 static inline int is_vmalloc_or_module_addr(const void *x
)
396 extern void kvfree(const void *addr
);
398 static inline void compound_lock(struct page
*page
)
400 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
401 VM_BUG_ON_PAGE(PageSlab(page
), page
);
402 bit_spin_lock(PG_compound_lock
, &page
->flags
);
406 static inline void compound_unlock(struct page
*page
)
408 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
409 VM_BUG_ON_PAGE(PageSlab(page
), page
);
410 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
414 static inline unsigned long compound_lock_irqsave(struct page
*page
)
416 unsigned long uninitialized_var(flags
);
417 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
418 local_irq_save(flags
);
424 static inline void compound_unlock_irqrestore(struct page
*page
,
427 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
428 compound_unlock(page
);
429 local_irq_restore(flags
);
433 static inline struct page
*compound_head_by_tail(struct page
*tail
)
435 struct page
*head
= tail
->first_page
;
438 * page->first_page may be a dangling pointer to an old
439 * compound page, so recheck that it is still a tail
440 * page before returning.
443 if (likely(PageTail(tail
)))
449 * Since either compound page could be dismantled asynchronously in THP
450 * or we access asynchronously arbitrary positioned struct page, there
451 * would be tail flag race. To handle this race, we should call
452 * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
454 static inline struct page
*compound_head(struct page
*page
)
456 if (unlikely(PageTail(page
)))
457 return compound_head_by_tail(page
);
462 * If we access compound page synchronously such as access to
463 * allocated page, there is no need to handle tail flag race, so we can
464 * check tail flag directly without any synchronization primitive.
466 static inline struct page
*compound_head_fast(struct page
*page
)
468 if (unlikely(PageTail(page
)))
469 return page
->first_page
;
474 * The atomic page->_mapcount, starts from -1: so that transitions
475 * both from it and to it can be tracked, using atomic_inc_and_test
476 * and atomic_add_negative(-1).
478 static inline void page_mapcount_reset(struct page
*page
)
480 atomic_set(&(page
)->_mapcount
, -1);
483 static inline int page_mapcount(struct page
*page
)
485 VM_BUG_ON_PAGE(PageSlab(page
), page
);
486 return atomic_read(&page
->_mapcount
) + 1;
489 static inline int page_count(struct page
*page
)
491 return atomic_read(&compound_head(page
)->_count
);
494 static inline bool __compound_tail_refcounted(struct page
*page
)
496 return PageAnon(page
) && !PageSlab(page
) && !PageHeadHuge(page
);
500 * This takes a head page as parameter and tells if the
501 * tail page reference counting can be skipped.
503 * For this to be safe, PageSlab and PageHeadHuge must remain true on
504 * any given page where they return true here, until all tail pins
505 * have been released.
507 static inline bool compound_tail_refcounted(struct page
*page
)
509 VM_BUG_ON_PAGE(!PageHead(page
), page
);
510 return __compound_tail_refcounted(page
);
513 static inline void get_huge_page_tail(struct page
*page
)
516 * __split_huge_page_refcount() cannot run from under us.
518 VM_BUG_ON_PAGE(!PageTail(page
), page
);
519 VM_BUG_ON_PAGE(page_mapcount(page
) < 0, page
);
520 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) != 0, page
);
521 if (compound_tail_refcounted(page
->first_page
))
522 atomic_inc(&page
->_mapcount
);
525 extern bool __get_page_tail(struct page
*page
);
527 static inline void get_page(struct page
*page
)
529 if (unlikely(PageTail(page
)))
530 if (likely(__get_page_tail(page
)))
533 * Getting a normal page or the head of a compound page
534 * requires to already have an elevated page->_count.
536 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) <= 0, page
);
537 atomic_inc(&page
->_count
);
540 static inline struct page
*virt_to_head_page(const void *x
)
542 struct page
*page
= virt_to_page(x
);
545 * We don't need to worry about synchronization of tail flag
546 * when we call virt_to_head_page() since it is only called for
547 * already allocated page and this page won't be freed until
548 * this virt_to_head_page() is finished. So use _fast variant.
550 return compound_head_fast(page
);
554 * Setup the page count before being freed into the page allocator for
555 * the first time (boot or memory hotplug)
557 static inline void init_page_count(struct page
*page
)
559 atomic_set(&page
->_count
, 1);
562 void put_page(struct page
*page
);
563 void put_pages_list(struct list_head
*pages
);
565 void split_page(struct page
*page
, unsigned int order
);
566 int split_free_page(struct page
*page
);
569 * Compound pages have a destructor function. Provide a
570 * prototype for that function and accessor functions.
571 * These are _only_ valid on the head of a compound page.
573 typedef void compound_page_dtor(struct page
*);
575 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
576 enum compound_dtor_id
{
579 #ifdef CONFIG_HUGETLB_PAGE
584 extern compound_page_dtor
* const compound_page_dtors
[];
586 static inline void set_compound_page_dtor(struct page
*page
,
587 enum compound_dtor_id compound_dtor
)
589 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
590 page
[1].compound_dtor
= compound_dtor
;
593 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
595 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
596 return compound_page_dtors
[page
[1].compound_dtor
];
599 static inline int compound_order(struct page
*page
)
603 return page
[1].compound_order
;
606 static inline void set_compound_order(struct page
*page
, unsigned int order
)
608 page
[1].compound_order
= order
;
613 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
614 * servicing faults for write access. In the normal case, do always want
615 * pte_mkwrite. But get_user_pages can cause write faults for mappings
616 * that do not have writing enabled, when used by access_process_vm.
618 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
620 if (likely(vma
->vm_flags
& VM_WRITE
))
621 pte
= pte_mkwrite(pte
);
625 void do_set_pte(struct vm_area_struct
*vma
, unsigned long address
,
626 struct page
*page
, pte_t
*pte
, bool write
, bool anon
);
630 * Multiple processes may "see" the same page. E.g. for untouched
631 * mappings of /dev/null, all processes see the same page full of
632 * zeroes, and text pages of executables and shared libraries have
633 * only one copy in memory, at most, normally.
635 * For the non-reserved pages, page_count(page) denotes a reference count.
636 * page_count() == 0 means the page is free. page->lru is then used for
637 * freelist management in the buddy allocator.
638 * page_count() > 0 means the page has been allocated.
640 * Pages are allocated by the slab allocator in order to provide memory
641 * to kmalloc and kmem_cache_alloc. In this case, the management of the
642 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
643 * unless a particular usage is carefully commented. (the responsibility of
644 * freeing the kmalloc memory is the caller's, of course).
646 * A page may be used by anyone else who does a __get_free_page().
647 * In this case, page_count still tracks the references, and should only
648 * be used through the normal accessor functions. The top bits of page->flags
649 * and page->virtual store page management information, but all other fields
650 * are unused and could be used privately, carefully. The management of this
651 * page is the responsibility of the one who allocated it, and those who have
652 * subsequently been given references to it.
654 * The other pages (we may call them "pagecache pages") are completely
655 * managed by the Linux memory manager: I/O, buffers, swapping etc.
656 * The following discussion applies only to them.
658 * A pagecache page contains an opaque `private' member, which belongs to the
659 * page's address_space. Usually, this is the address of a circular list of
660 * the page's disk buffers. PG_private must be set to tell the VM to call
661 * into the filesystem to release these pages.
663 * A page may belong to an inode's memory mapping. In this case, page->mapping
664 * is the pointer to the inode, and page->index is the file offset of the page,
665 * in units of PAGE_CACHE_SIZE.
667 * If pagecache pages are not associated with an inode, they are said to be
668 * anonymous pages. These may become associated with the swapcache, and in that
669 * case PG_swapcache is set, and page->private is an offset into the swapcache.
671 * In either case (swapcache or inode backed), the pagecache itself holds one
672 * reference to the page. Setting PG_private should also increment the
673 * refcount. The each user mapping also has a reference to the page.
675 * The pagecache pages are stored in a per-mapping radix tree, which is
676 * rooted at mapping->page_tree, and indexed by offset.
677 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
678 * lists, we instead now tag pages as dirty/writeback in the radix tree.
680 * All pagecache pages may be subject to I/O:
681 * - inode pages may need to be read from disk,
682 * - inode pages which have been modified and are MAP_SHARED may need
683 * to be written back to the inode on disk,
684 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
685 * modified may need to be swapped out to swap space and (later) to be read
690 * The zone field is never updated after free_area_init_core()
691 * sets it, so none of the operations on it need to be atomic.
694 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
695 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
696 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
697 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
698 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
701 * Define the bit shifts to access each section. For non-existent
702 * sections we define the shift as 0; that plus a 0 mask ensures
703 * the compiler will optimise away reference to them.
705 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
706 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
707 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
708 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
710 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
711 #ifdef NODE_NOT_IN_PAGE_FLAGS
712 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
713 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
714 SECTIONS_PGOFF : ZONES_PGOFF)
716 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
717 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
718 NODES_PGOFF : ZONES_PGOFF)
721 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
723 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
724 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
727 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
728 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
729 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
730 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
731 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
733 static inline enum zone_type
page_zonenum(const struct page
*page
)
735 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
738 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
739 #define SECTION_IN_PAGE_FLAGS
743 * The identification function is mainly used by the buddy allocator for
744 * determining if two pages could be buddies. We are not really identifying
745 * the zone since we could be using the section number id if we do not have
746 * node id available in page flags.
747 * We only guarantee that it will return the same value for two combinable
750 static inline int page_zone_id(struct page
*page
)
752 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
755 static inline int zone_to_nid(struct zone
*zone
)
764 #ifdef NODE_NOT_IN_PAGE_FLAGS
765 extern int page_to_nid(const struct page
*page
);
767 static inline int page_to_nid(const struct page
*page
)
769 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
773 #ifdef CONFIG_NUMA_BALANCING
774 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
776 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
779 static inline int cpupid_to_pid(int cpupid
)
781 return cpupid
& LAST__PID_MASK
;
784 static inline int cpupid_to_cpu(int cpupid
)
786 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
789 static inline int cpupid_to_nid(int cpupid
)
791 return cpu_to_node(cpupid_to_cpu(cpupid
));
794 static inline bool cpupid_pid_unset(int cpupid
)
796 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
799 static inline bool cpupid_cpu_unset(int cpupid
)
801 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
804 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
806 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
809 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
810 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
811 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
813 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
816 static inline int page_cpupid_last(struct page
*page
)
818 return page
->_last_cpupid
;
820 static inline void page_cpupid_reset_last(struct page
*page
)
822 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
825 static inline int page_cpupid_last(struct page
*page
)
827 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
830 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
832 static inline void page_cpupid_reset_last(struct page
*page
)
834 int cpupid
= (1 << LAST_CPUPID_SHIFT
) - 1;
836 page
->flags
&= ~(LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
);
837 page
->flags
|= (cpupid
& LAST_CPUPID_MASK
) << LAST_CPUPID_PGSHIFT
;
839 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
840 #else /* !CONFIG_NUMA_BALANCING */
841 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
843 return page_to_nid(page
); /* XXX */
846 static inline int page_cpupid_last(struct page
*page
)
848 return page_to_nid(page
); /* XXX */
851 static inline int cpupid_to_nid(int cpupid
)
856 static inline int cpupid_to_pid(int cpupid
)
861 static inline int cpupid_to_cpu(int cpupid
)
866 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
871 static inline bool cpupid_pid_unset(int cpupid
)
876 static inline void page_cpupid_reset_last(struct page
*page
)
880 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
884 #endif /* CONFIG_NUMA_BALANCING */
886 static inline struct zone
*page_zone(const struct page
*page
)
888 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
891 #ifdef SECTION_IN_PAGE_FLAGS
892 static inline void set_page_section(struct page
*page
, unsigned long section
)
894 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
895 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
898 static inline unsigned long page_to_section(const struct page
*page
)
900 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
904 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
906 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
907 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
910 static inline void set_page_node(struct page
*page
, unsigned long node
)
912 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
913 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
916 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
917 unsigned long node
, unsigned long pfn
)
919 set_page_zone(page
, zone
);
920 set_page_node(page
, node
);
921 #ifdef SECTION_IN_PAGE_FLAGS
922 set_page_section(page
, pfn_to_section_nr(pfn
));
927 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
929 return page
->mem_cgroup
;
932 static inline void set_page_memcg(struct page
*page
, struct mem_cgroup
*memcg
)
934 page
->mem_cgroup
= memcg
;
937 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
942 static inline void set_page_memcg(struct page
*page
, struct mem_cgroup
*memcg
)
948 * Some inline functions in vmstat.h depend on page_zone()
950 #include <linux/vmstat.h>
952 static __always_inline
void *lowmem_page_address(const struct page
*page
)
954 return __va(PFN_PHYS(page_to_pfn(page
)));
957 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
958 #define HASHED_PAGE_VIRTUAL
961 #if defined(WANT_PAGE_VIRTUAL)
962 static inline void *page_address(const struct page
*page
)
964 return page
->virtual;
966 static inline void set_page_address(struct page
*page
, void *address
)
968 page
->virtual = address
;
970 #define page_address_init() do { } while(0)
973 #if defined(HASHED_PAGE_VIRTUAL)
974 void *page_address(const struct page
*page
);
975 void set_page_address(struct page
*page
, void *virtual);
976 void page_address_init(void);
979 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
980 #define page_address(page) lowmem_page_address(page)
981 #define set_page_address(page, address) do { } while(0)
982 #define page_address_init() do { } while(0)
985 extern void *page_rmapping(struct page
*page
);
986 extern struct anon_vma
*page_anon_vma(struct page
*page
);
987 extern struct address_space
*page_mapping(struct page
*page
);
989 extern struct address_space
*__page_file_mapping(struct page
*);
992 struct address_space
*page_file_mapping(struct page
*page
)
994 if (unlikely(PageSwapCache(page
)))
995 return __page_file_mapping(page
);
997 return page
->mapping
;
1001 * Return the pagecache index of the passed page. Regular pagecache pages
1002 * use ->index whereas swapcache pages use ->private
1004 static inline pgoff_t
page_index(struct page
*page
)
1006 if (unlikely(PageSwapCache(page
)))
1007 return page_private(page
);
1011 extern pgoff_t
__page_file_index(struct page
*page
);
1014 * Return the file index of the page. Regular pagecache pages use ->index
1015 * whereas swapcache pages use swp_offset(->private)
1017 static inline pgoff_t
page_file_index(struct page
*page
)
1019 if (unlikely(PageSwapCache(page
)))
1020 return __page_file_index(page
);
1026 * Return true if this page is mapped into pagetables.
1028 static inline int page_mapped(struct page
*page
)
1030 return atomic_read(&(page
)->_mapcount
) >= 0;
1034 * Return true only if the page has been allocated with
1035 * ALLOC_NO_WATERMARKS and the low watermark was not
1036 * met implying that the system is under some pressure.
1038 static inline bool page_is_pfmemalloc(struct page
*page
)
1041 * Page index cannot be this large so this must be
1042 * a pfmemalloc page.
1044 return page
->index
== -1UL;
1048 * Only to be called by the page allocator on a freshly allocated
1051 static inline void set_page_pfmemalloc(struct page
*page
)
1056 static inline void clear_page_pfmemalloc(struct page
*page
)
1062 * Different kinds of faults, as returned by handle_mm_fault().
1063 * Used to decide whether a process gets delivered SIGBUS or
1064 * just gets major/minor fault counters bumped up.
1067 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1069 #define VM_FAULT_OOM 0x0001
1070 #define VM_FAULT_SIGBUS 0x0002
1071 #define VM_FAULT_MAJOR 0x0004
1072 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1073 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1074 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1075 #define VM_FAULT_SIGSEGV 0x0040
1077 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1078 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1079 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1080 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1082 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1084 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1085 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1088 /* Encode hstate index for a hwpoisoned large page */
1089 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1090 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1093 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1095 extern void pagefault_out_of_memory(void);
1097 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1100 * Flags passed to show_mem() and show_free_areas() to suppress output in
1103 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1105 extern void show_free_areas(unsigned int flags
);
1106 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
1108 int shmem_zero_setup(struct vm_area_struct
*);
1110 bool shmem_mapping(struct address_space
*mapping
);
1112 static inline bool shmem_mapping(struct address_space
*mapping
)
1118 extern int can_do_mlock(void);
1119 extern int user_shm_lock(size_t, struct user_struct
*);
1120 extern void user_shm_unlock(size_t, struct user_struct
*);
1123 * Parameter block passed down to zap_pte_range in exceptional cases.
1125 struct zap_details
{
1126 struct address_space
*check_mapping
; /* Check page->mapping if set */
1127 pgoff_t first_index
; /* Lowest page->index to unmap */
1128 pgoff_t last_index
; /* Highest page->index to unmap */
1131 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1134 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1135 unsigned long size
);
1136 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1137 unsigned long size
, struct zap_details
*);
1138 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1139 unsigned long start
, unsigned long end
);
1142 * mm_walk - callbacks for walk_page_range
1143 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1144 * this handler is required to be able to handle
1145 * pmd_trans_huge() pmds. They may simply choose to
1146 * split_huge_page() instead of handling it explicitly.
1147 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1148 * @pte_hole: if set, called for each hole at all levels
1149 * @hugetlb_entry: if set, called for each hugetlb entry
1150 * @test_walk: caller specific callback function to determine whether
1151 * we walk over the current vma or not. A positive returned
1152 * value means "do page table walk over the current vma,"
1153 * and a negative one means "abort current page table walk
1154 * right now." 0 means "skip the current vma."
1155 * @mm: mm_struct representing the target process of page table walk
1156 * @vma: vma currently walked (NULL if walking outside vmas)
1157 * @private: private data for callbacks' usage
1159 * (see the comment on walk_page_range() for more details)
1162 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1163 unsigned long next
, struct mm_walk
*walk
);
1164 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1165 unsigned long next
, struct mm_walk
*walk
);
1166 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1167 struct mm_walk
*walk
);
1168 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1169 unsigned long addr
, unsigned long next
,
1170 struct mm_walk
*walk
);
1171 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1172 struct mm_walk
*walk
);
1173 struct mm_struct
*mm
;
1174 struct vm_area_struct
*vma
;
1178 int walk_page_range(unsigned long addr
, unsigned long end
,
1179 struct mm_walk
*walk
);
1180 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1181 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1182 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1183 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1184 struct vm_area_struct
*vma
);
1185 void unmap_mapping_range(struct address_space
*mapping
,
1186 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1187 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1188 unsigned long *pfn
);
1189 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1190 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1191 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1192 void *buf
, int len
, int write
);
1194 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1195 loff_t
const holebegin
, loff_t
const holelen
)
1197 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1200 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1201 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1202 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1203 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1204 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1205 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1206 int invalidate_inode_page(struct page
*page
);
1209 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1210 unsigned long address
, unsigned int flags
);
1211 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1212 unsigned long address
, unsigned int fault_flags
);
1214 static inline int handle_mm_fault(struct mm_struct
*mm
,
1215 struct vm_area_struct
*vma
, unsigned long address
,
1218 /* should never happen if there's no MMU */
1220 return VM_FAULT_SIGBUS
;
1222 static inline int fixup_user_fault(struct task_struct
*tsk
,
1223 struct mm_struct
*mm
, unsigned long address
,
1224 unsigned int fault_flags
)
1226 /* should never happen if there's no MMU */
1232 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1233 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1234 void *buf
, int len
, int write
);
1236 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1237 unsigned long start
, unsigned long nr_pages
,
1238 unsigned int foll_flags
, struct page
**pages
,
1239 struct vm_area_struct
**vmas
, int *nonblocking
);
1240 long get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1241 unsigned long start
, unsigned long nr_pages
,
1242 int write
, int force
, struct page
**pages
,
1243 struct vm_area_struct
**vmas
);
1244 long get_user_pages_locked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1245 unsigned long start
, unsigned long nr_pages
,
1246 int write
, int force
, struct page
**pages
,
1248 long __get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1249 unsigned long start
, unsigned long nr_pages
,
1250 int write
, int force
, struct page
**pages
,
1251 unsigned int gup_flags
);
1252 long get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1253 unsigned long start
, unsigned long nr_pages
,
1254 int write
, int force
, struct page
**pages
);
1255 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1256 struct page
**pages
);
1258 /* Container for pinned pfns / pages */
1259 struct frame_vector
{
1260 unsigned int nr_allocated
; /* Number of frames we have space for */
1261 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1262 bool got_ref
; /* Did we pin pages by getting page ref? */
1263 bool is_pfns
; /* Does array contain pages or pfns? */
1264 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1265 * pfns_vector_pages() or pfns_vector_pfns()
1269 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1270 void frame_vector_destroy(struct frame_vector
*vec
);
1271 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1272 bool write
, bool force
, struct frame_vector
*vec
);
1273 void put_vaddr_frames(struct frame_vector
*vec
);
1274 int frame_vector_to_pages(struct frame_vector
*vec
);
1275 void frame_vector_to_pfns(struct frame_vector
*vec
);
1277 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1279 return vec
->nr_frames
;
1282 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1285 int err
= frame_vector_to_pages(vec
);
1288 return ERR_PTR(err
);
1290 return (struct page
**)(vec
->ptrs
);
1293 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1296 frame_vector_to_pfns(vec
);
1297 return (unsigned long *)(vec
->ptrs
);
1301 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1302 struct page
**pages
);
1303 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1304 struct page
*get_dump_page(unsigned long addr
);
1306 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1307 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1308 unsigned int length
);
1310 int __set_page_dirty_nobuffers(struct page
*page
);
1311 int __set_page_dirty_no_writeback(struct page
*page
);
1312 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1314 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
,
1315 struct mem_cgroup
*memcg
);
1316 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1317 struct mem_cgroup
*memcg
, struct bdi_writeback
*wb
);
1318 int set_page_dirty(struct page
*page
);
1319 int set_page_dirty_lock(struct page
*page
);
1320 void cancel_dirty_page(struct page
*page
);
1321 int clear_page_dirty_for_io(struct page
*page
);
1323 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1325 /* Is the vma a continuation of the stack vma above it? */
1326 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1328 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1331 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1333 return !vma
->vm_ops
;
1336 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1339 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1340 (vma
->vm_start
== addr
) &&
1341 !vma_growsdown(vma
->vm_prev
, addr
);
1344 /* Is the vma a continuation of the stack vma below it? */
1345 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1347 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1350 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1353 return (vma
->vm_flags
& VM_GROWSUP
) &&
1354 (vma
->vm_end
== addr
) &&
1355 !vma_growsup(vma
->vm_next
, addr
);
1358 extern struct task_struct
*task_of_stack(struct task_struct
*task
,
1359 struct vm_area_struct
*vma
, bool in_group
);
1361 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1362 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1363 unsigned long new_addr
, unsigned long len
,
1364 bool need_rmap_locks
);
1365 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1366 unsigned long end
, pgprot_t newprot
,
1367 int dirty_accountable
, int prot_numa
);
1368 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1369 struct vm_area_struct
**pprev
, unsigned long start
,
1370 unsigned long end
, unsigned long newflags
);
1373 * doesn't attempt to fault and will return short.
1375 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1376 struct page
**pages
);
1378 * per-process(per-mm_struct) statistics.
1380 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1382 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1384 #ifdef SPLIT_RSS_COUNTING
1386 * counter is updated in asynchronous manner and may go to minus.
1387 * But it's never be expected number for users.
1392 return (unsigned long)val
;
1395 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1397 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1400 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1402 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1405 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1407 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1410 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1412 return get_mm_counter(mm
, MM_FILEPAGES
) +
1413 get_mm_counter(mm
, MM_ANONPAGES
);
1416 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1418 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1421 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1423 return max(mm
->hiwater_vm
, mm
->total_vm
);
1426 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1428 unsigned long _rss
= get_mm_rss(mm
);
1430 if ((mm
)->hiwater_rss
< _rss
)
1431 (mm
)->hiwater_rss
= _rss
;
1434 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1436 if (mm
->hiwater_vm
< mm
->total_vm
)
1437 mm
->hiwater_vm
= mm
->total_vm
;
1440 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1442 mm
->hiwater_rss
= get_mm_rss(mm
);
1445 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1446 struct mm_struct
*mm
)
1448 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1450 if (*maxrss
< hiwater_rss
)
1451 *maxrss
= hiwater_rss
;
1454 #if defined(SPLIT_RSS_COUNTING)
1455 void sync_mm_rss(struct mm_struct
*mm
);
1457 static inline void sync_mm_rss(struct mm_struct
*mm
)
1462 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1464 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1466 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1470 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1474 #ifdef __PAGETABLE_PUD_FOLDED
1475 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1476 unsigned long address
)
1481 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1484 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1485 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1486 unsigned long address
)
1491 static inline void mm_nr_pmds_init(struct mm_struct
*mm
) {}
1493 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1498 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1499 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1502 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1504 static inline void mm_nr_pmds_init(struct mm_struct
*mm
)
1506 atomic_long_set(&mm
->nr_pmds
, 0);
1509 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1511 return atomic_long_read(&mm
->nr_pmds
);
1514 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1516 atomic_long_inc(&mm
->nr_pmds
);
1519 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1521 atomic_long_dec(&mm
->nr_pmds
);
1525 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1526 pmd_t
*pmd
, unsigned long address
);
1527 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1530 * The following ifdef needed to get the 4level-fixup.h header to work.
1531 * Remove it when 4level-fixup.h has been removed.
1533 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1534 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1536 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1537 NULL
: pud_offset(pgd
, address
);
1540 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1542 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1543 NULL
: pmd_offset(pud
, address
);
1545 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1547 #if USE_SPLIT_PTE_PTLOCKS
1548 #if ALLOC_SPLIT_PTLOCKS
1549 void __init
ptlock_cache_init(void);
1550 extern bool ptlock_alloc(struct page
*page
);
1551 extern void ptlock_free(struct page
*page
);
1553 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1557 #else /* ALLOC_SPLIT_PTLOCKS */
1558 static inline void ptlock_cache_init(void)
1562 static inline bool ptlock_alloc(struct page
*page
)
1567 static inline void ptlock_free(struct page
*page
)
1571 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1575 #endif /* ALLOC_SPLIT_PTLOCKS */
1577 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1579 return ptlock_ptr(pmd_page(*pmd
));
1582 static inline bool ptlock_init(struct page
*page
)
1585 * prep_new_page() initialize page->private (and therefore page->ptl)
1586 * with 0. Make sure nobody took it in use in between.
1588 * It can happen if arch try to use slab for page table allocation:
1589 * slab code uses page->slab_cache and page->first_page (for tail
1590 * pages), which share storage with page->ptl.
1592 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1593 if (!ptlock_alloc(page
))
1595 spin_lock_init(ptlock_ptr(page
));
1599 /* Reset page->mapping so free_pages_check won't complain. */
1600 static inline void pte_lock_deinit(struct page
*page
)
1602 page
->mapping
= NULL
;
1606 #else /* !USE_SPLIT_PTE_PTLOCKS */
1608 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1610 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1612 return &mm
->page_table_lock
;
1614 static inline void ptlock_cache_init(void) {}
1615 static inline bool ptlock_init(struct page
*page
) { return true; }
1616 static inline void pte_lock_deinit(struct page
*page
) {}
1617 #endif /* USE_SPLIT_PTE_PTLOCKS */
1619 static inline void pgtable_init(void)
1621 ptlock_cache_init();
1622 pgtable_cache_init();
1625 static inline bool pgtable_page_ctor(struct page
*page
)
1627 if (!ptlock_init(page
))
1629 inc_zone_page_state(page
, NR_PAGETABLE
);
1633 static inline void pgtable_page_dtor(struct page
*page
)
1635 pte_lock_deinit(page
);
1636 dec_zone_page_state(page
, NR_PAGETABLE
);
1639 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1641 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1642 pte_t *__pte = pte_offset_map(pmd, address); \
1648 #define pte_unmap_unlock(pte, ptl) do { \
1653 #define pte_alloc_map(mm, vma, pmd, address) \
1654 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1656 NULL: pte_offset_map(pmd, address))
1658 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1659 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1661 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1663 #define pte_alloc_kernel(pmd, address) \
1664 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1665 NULL: pte_offset_kernel(pmd, address))
1667 #if USE_SPLIT_PMD_PTLOCKS
1669 static struct page
*pmd_to_page(pmd_t
*pmd
)
1671 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1672 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1675 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1677 return ptlock_ptr(pmd_to_page(pmd
));
1680 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1682 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1683 page
->pmd_huge_pte
= NULL
;
1685 return ptlock_init(page
);
1688 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1690 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1691 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1696 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1700 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1702 return &mm
->page_table_lock
;
1705 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1706 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1708 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1712 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1714 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1719 extern void free_area_init(unsigned long * zones_size
);
1720 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1721 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1722 extern void free_initmem(void);
1725 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1726 * into the buddy system. The freed pages will be poisoned with pattern
1727 * "poison" if it's within range [0, UCHAR_MAX].
1728 * Return pages freed into the buddy system.
1730 extern unsigned long free_reserved_area(void *start
, void *end
,
1731 int poison
, char *s
);
1733 #ifdef CONFIG_HIGHMEM
1735 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1736 * and totalram_pages.
1738 extern void free_highmem_page(struct page
*page
);
1741 extern void adjust_managed_page_count(struct page
*page
, long count
);
1742 extern void mem_init_print_info(const char *str
);
1744 extern void reserve_bootmem_region(unsigned long start
, unsigned long end
);
1746 /* Free the reserved page into the buddy system, so it gets managed. */
1747 static inline void __free_reserved_page(struct page
*page
)
1749 ClearPageReserved(page
);
1750 init_page_count(page
);
1754 static inline void free_reserved_page(struct page
*page
)
1756 __free_reserved_page(page
);
1757 adjust_managed_page_count(page
, 1);
1760 static inline void mark_page_reserved(struct page
*page
)
1762 SetPageReserved(page
);
1763 adjust_managed_page_count(page
, -1);
1767 * Default method to free all the __init memory into the buddy system.
1768 * The freed pages will be poisoned with pattern "poison" if it's within
1769 * range [0, UCHAR_MAX].
1770 * Return pages freed into the buddy system.
1772 static inline unsigned long free_initmem_default(int poison
)
1774 extern char __init_begin
[], __init_end
[];
1776 return free_reserved_area(&__init_begin
, &__init_end
,
1777 poison
, "unused kernel");
1780 static inline unsigned long get_num_physpages(void)
1783 unsigned long phys_pages
= 0;
1785 for_each_online_node(nid
)
1786 phys_pages
+= node_present_pages(nid
);
1791 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1793 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1794 * zones, allocate the backing mem_map and account for memory holes in a more
1795 * architecture independent manner. This is a substitute for creating the
1796 * zone_sizes[] and zholes_size[] arrays and passing them to
1797 * free_area_init_node()
1799 * An architecture is expected to register range of page frames backed by
1800 * physical memory with memblock_add[_node]() before calling
1801 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1802 * usage, an architecture is expected to do something like
1804 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1806 * for_each_valid_physical_page_range()
1807 * memblock_add_node(base, size, nid)
1808 * free_area_init_nodes(max_zone_pfns);
1810 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1811 * registered physical page range. Similarly
1812 * sparse_memory_present_with_active_regions() calls memory_present() for
1813 * each range when SPARSEMEM is enabled.
1815 * See mm/page_alloc.c for more information on each function exposed by
1816 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1818 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1819 unsigned long node_map_pfn_alignment(void);
1820 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1821 unsigned long end_pfn
);
1822 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1823 unsigned long end_pfn
);
1824 extern void get_pfn_range_for_nid(unsigned int nid
,
1825 unsigned long *start_pfn
, unsigned long *end_pfn
);
1826 extern unsigned long find_min_pfn_with_active_regions(void);
1827 extern void free_bootmem_with_active_regions(int nid
,
1828 unsigned long max_low_pfn
);
1829 extern void sparse_memory_present_with_active_regions(int nid
);
1831 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1833 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1834 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1835 static inline int __early_pfn_to_nid(unsigned long pfn
,
1836 struct mminit_pfnnid_cache
*state
)
1841 /* please see mm/page_alloc.c */
1842 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1843 /* there is a per-arch backend function. */
1844 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
1845 struct mminit_pfnnid_cache
*state
);
1848 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1849 extern void memmap_init_zone(unsigned long, int, unsigned long,
1850 unsigned long, enum memmap_context
);
1851 extern void setup_per_zone_wmarks(void);
1852 extern int __meminit
init_per_zone_wmark_min(void);
1853 extern void mem_init(void);
1854 extern void __init
mmap_init(void);
1855 extern void show_mem(unsigned int flags
);
1856 extern void si_meminfo(struct sysinfo
* val
);
1857 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1859 extern __printf(3, 4)
1860 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1862 extern void setup_per_cpu_pageset(void);
1864 extern void zone_pcp_update(struct zone
*zone
);
1865 extern void zone_pcp_reset(struct zone
*zone
);
1868 extern int min_free_kbytes
;
1871 extern atomic_long_t mmap_pages_allocated
;
1872 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1874 /* interval_tree.c */
1875 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1876 struct rb_root
*root
);
1877 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1878 struct vm_area_struct
*prev
,
1879 struct rb_root
*root
);
1880 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1881 struct rb_root
*root
);
1882 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1883 unsigned long start
, unsigned long last
);
1884 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1885 unsigned long start
, unsigned long last
);
1887 #define vma_interval_tree_foreach(vma, root, start, last) \
1888 for (vma = vma_interval_tree_iter_first(root, start, last); \
1889 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1891 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1892 struct rb_root
*root
);
1893 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1894 struct rb_root
*root
);
1895 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1896 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1897 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1898 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1899 #ifdef CONFIG_DEBUG_VM_RB
1900 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1903 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1904 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1905 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1908 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1909 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1910 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1911 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1912 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1913 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1914 struct mempolicy
*, struct vm_userfaultfd_ctx
);
1915 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1916 extern int split_vma(struct mm_struct
*,
1917 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1918 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1919 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1920 struct rb_node
**, struct rb_node
*);
1921 extern void unlink_file_vma(struct vm_area_struct
*);
1922 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1923 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1924 bool *need_rmap_locks
);
1925 extern void exit_mmap(struct mm_struct
*);
1927 static inline int check_data_rlimit(unsigned long rlim
,
1929 unsigned long start
,
1930 unsigned long end_data
,
1931 unsigned long start_data
)
1933 if (rlim
< RLIM_INFINITY
) {
1934 if (((new - start
) + (end_data
- start_data
)) > rlim
)
1941 extern int mm_take_all_locks(struct mm_struct
*mm
);
1942 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1944 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1945 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1947 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1948 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
1949 unsigned long addr
, unsigned long len
,
1950 unsigned long flags
,
1951 const struct vm_special_mapping
*spec
);
1952 /* This is an obsolete alternative to _install_special_mapping. */
1953 extern int install_special_mapping(struct mm_struct
*mm
,
1954 unsigned long addr
, unsigned long len
,
1955 unsigned long flags
, struct page
**pages
);
1957 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1959 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1960 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
1961 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
1962 unsigned long len
, unsigned long prot
, unsigned long flags
,
1963 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
);
1964 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1966 static inline unsigned long
1967 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1968 unsigned long len
, unsigned long prot
, unsigned long flags
,
1969 unsigned long pgoff
, unsigned long *populate
)
1971 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
);
1975 extern int __mm_populate(unsigned long addr
, unsigned long len
,
1977 static inline void mm_populate(unsigned long addr
, unsigned long len
)
1980 (void) __mm_populate(addr
, len
, 1);
1983 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
1986 /* These take the mm semaphore themselves */
1987 extern unsigned long vm_brk(unsigned long, unsigned long);
1988 extern int vm_munmap(unsigned long, size_t);
1989 extern unsigned long vm_mmap(struct file
*, unsigned long,
1990 unsigned long, unsigned long,
1991 unsigned long, unsigned long);
1993 struct vm_unmapped_area_info
{
1994 #define VM_UNMAPPED_AREA_TOPDOWN 1
1995 unsigned long flags
;
1996 unsigned long length
;
1997 unsigned long low_limit
;
1998 unsigned long high_limit
;
1999 unsigned long align_mask
;
2000 unsigned long align_offset
;
2003 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
2004 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
2007 * Search for an unmapped address range.
2009 * We are looking for a range that:
2010 * - does not intersect with any VMA;
2011 * - is contained within the [low_limit, high_limit) interval;
2012 * - is at least the desired size.
2013 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2015 static inline unsigned long
2016 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
2018 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
2019 return unmapped_area_topdown(info
);
2021 return unmapped_area(info
);
2025 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2026 extern void truncate_inode_pages_range(struct address_space
*,
2027 loff_t lstart
, loff_t lend
);
2028 extern void truncate_inode_pages_final(struct address_space
*);
2030 /* generic vm_area_ops exported for stackable file systems */
2031 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
2032 extern void filemap_map_pages(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
2033 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
2035 /* mm/page-writeback.c */
2036 int write_one_page(struct page
*page
, int wait
);
2037 void task_dirty_inc(struct task_struct
*tsk
);
2040 #define VM_MAX_READAHEAD 128 /* kbytes */
2041 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2043 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2044 pgoff_t offset
, unsigned long nr_to_read
);
2046 void page_cache_sync_readahead(struct address_space
*mapping
,
2047 struct file_ra_state
*ra
,
2050 unsigned long size
);
2052 void page_cache_async_readahead(struct address_space
*mapping
,
2053 struct file_ra_state
*ra
,
2057 unsigned long size
);
2059 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2060 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2062 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2063 extern int expand_downwards(struct vm_area_struct
*vma
,
2064 unsigned long address
);
2066 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2068 #define expand_upwards(vma, address) (0)
2071 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2072 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2073 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2074 struct vm_area_struct
**pprev
);
2076 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2077 NULL if none. Assume start_addr < end_addr. */
2078 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2080 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2082 if (vma
&& end_addr
<= vma
->vm_start
)
2087 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2089 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2092 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2093 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2094 unsigned long vm_start
, unsigned long vm_end
)
2096 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2098 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2105 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2106 void vma_set_page_prot(struct vm_area_struct
*vma
);
2108 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2112 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2114 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2118 #ifdef CONFIG_NUMA_BALANCING
2119 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2120 unsigned long start
, unsigned long end
);
2123 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2124 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2125 unsigned long pfn
, unsigned long size
, pgprot_t
);
2126 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2127 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2129 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2131 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2134 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2135 unsigned long address
, unsigned int foll_flags
,
2136 unsigned int *page_mask
);
2138 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2139 unsigned long address
, unsigned int foll_flags
)
2141 unsigned int unused_page_mask
;
2142 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2145 #define FOLL_WRITE 0x01 /* check pte is writable */
2146 #define FOLL_TOUCH 0x02 /* mark page accessed */
2147 #define FOLL_GET 0x04 /* do get_page on page */
2148 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2149 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2150 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2151 * and return without waiting upon it */
2152 #define FOLL_POPULATE 0x40 /* fault in page */
2153 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2154 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2155 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2156 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2157 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2158 #define FOLL_MLOCK 0x1000 /* lock present pages */
2160 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2162 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2163 unsigned long size
, pte_fn_t fn
, void *data
);
2165 #ifdef CONFIG_PROC_FS
2166 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
2168 static inline void vm_stat_account(struct mm_struct
*mm
,
2169 unsigned long flags
, struct file
*file
, long pages
)
2171 mm
->total_vm
+= pages
;
2173 #endif /* CONFIG_PROC_FS */
2175 #ifdef CONFIG_DEBUG_PAGEALLOC
2176 extern bool _debug_pagealloc_enabled
;
2177 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2179 static inline bool debug_pagealloc_enabled(void)
2181 return _debug_pagealloc_enabled
;
2185 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2187 if (!debug_pagealloc_enabled())
2190 __kernel_map_pages(page
, numpages
, enable
);
2192 #ifdef CONFIG_HIBERNATION
2193 extern bool kernel_page_present(struct page
*page
);
2194 #endif /* CONFIG_HIBERNATION */
2197 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2198 #ifdef CONFIG_HIBERNATION
2199 static inline bool kernel_page_present(struct page
*page
) { return true; }
2200 #endif /* CONFIG_HIBERNATION */
2203 #ifdef __HAVE_ARCH_GATE_AREA
2204 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2205 extern int in_gate_area_no_mm(unsigned long addr
);
2206 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2208 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2212 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2213 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2217 #endif /* __HAVE_ARCH_GATE_AREA */
2219 #ifdef CONFIG_SYSCTL
2220 extern int sysctl_drop_caches
;
2221 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2222 void __user
*, size_t *, loff_t
*);
2225 void drop_slab(void);
2226 void drop_slab_node(int nid
);
2229 #define randomize_va_space 0
2231 extern int randomize_va_space
;
2234 const char * arch_vma_name(struct vm_area_struct
*vma
);
2235 void print_vma_addr(char *prefix
, unsigned long rip
);
2237 void sparse_mem_maps_populate_node(struct page
**map_map
,
2238 unsigned long pnum_begin
,
2239 unsigned long pnum_end
,
2240 unsigned long map_count
,
2243 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2244 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2245 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2246 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2247 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2248 void *vmemmap_alloc_block(unsigned long size
, int node
);
2249 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
2250 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2251 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2253 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2254 void vmemmap_populate_print_last(void);
2255 #ifdef CONFIG_MEMORY_HOTPLUG
2256 void vmemmap_free(unsigned long start
, unsigned long end
);
2258 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2259 unsigned long size
);
2262 MF_COUNT_INCREASED
= 1 << 0,
2263 MF_ACTION_REQUIRED
= 1 << 1,
2264 MF_MUST_KILL
= 1 << 2,
2265 MF_SOFT_OFFLINE
= 1 << 3,
2267 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2268 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2269 extern int unpoison_memory(unsigned long pfn
);
2270 extern int get_hwpoison_page(struct page
*page
);
2271 extern void put_hwpoison_page(struct page
*page
);
2272 extern int sysctl_memory_failure_early_kill
;
2273 extern int sysctl_memory_failure_recovery
;
2274 extern void shake_page(struct page
*p
, int access
);
2275 extern atomic_long_t num_poisoned_pages
;
2276 extern int soft_offline_page(struct page
*page
, int flags
);
2280 * Error handlers for various types of pages.
2283 MF_IGNORED
, /* Error: cannot be handled */
2284 MF_FAILED
, /* Error: handling failed */
2285 MF_DELAYED
, /* Will be handled later */
2286 MF_RECOVERED
, /* Successfully recovered */
2289 enum mf_action_page_type
{
2291 MF_MSG_KERNEL_HIGH_ORDER
,
2293 MF_MSG_DIFFERENT_COMPOUND
,
2294 MF_MSG_POISONED_HUGE
,
2297 MF_MSG_UNMAP_FAILED
,
2298 MF_MSG_DIRTY_SWAPCACHE
,
2299 MF_MSG_CLEAN_SWAPCACHE
,
2300 MF_MSG_DIRTY_MLOCKED_LRU
,
2301 MF_MSG_CLEAN_MLOCKED_LRU
,
2302 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2303 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2306 MF_MSG_TRUNCATED_LRU
,
2312 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2313 extern void clear_huge_page(struct page
*page
,
2315 unsigned int pages_per_huge_page
);
2316 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2317 unsigned long addr
, struct vm_area_struct
*vma
,
2318 unsigned int pages_per_huge_page
);
2319 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2321 extern struct page_ext_operations debug_guardpage_ops
;
2322 extern struct page_ext_operations page_poisoning_ops
;
2324 #ifdef CONFIG_DEBUG_PAGEALLOC
2325 extern unsigned int _debug_guardpage_minorder
;
2326 extern bool _debug_guardpage_enabled
;
2328 static inline unsigned int debug_guardpage_minorder(void)
2330 return _debug_guardpage_minorder
;
2333 static inline bool debug_guardpage_enabled(void)
2335 return _debug_guardpage_enabled
;
2338 static inline bool page_is_guard(struct page
*page
)
2340 struct page_ext
*page_ext
;
2342 if (!debug_guardpage_enabled())
2345 page_ext
= lookup_page_ext(page
);
2346 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2349 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2350 static inline bool debug_guardpage_enabled(void) { return false; }
2351 static inline bool page_is_guard(struct page
*page
) { return false; }
2352 #endif /* CONFIG_DEBUG_PAGEALLOC */
2354 #if MAX_NUMNODES > 1
2355 void __init
setup_nr_node_ids(void);
2357 static inline void setup_nr_node_ids(void) {}
2360 #endif /* __KERNEL__ */
2361 #endif /* _LINUX_MM_H */