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>
26 struct anon_vma_chain
;
29 struct writeback_control
;
32 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
33 extern unsigned long max_mapnr
;
35 static inline void set_max_mapnr(unsigned long limit
)
40 static inline void set_max_mapnr(unsigned long limit
) { }
43 extern unsigned long totalram_pages
;
44 extern void * high_memory
;
45 extern int page_cluster
;
48 extern int sysctl_legacy_va_layout
;
50 #define sysctl_legacy_va_layout 0
54 #include <asm/pgtable.h>
55 #include <asm/processor.h>
58 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
62 * To prevent common memory management code establishing
63 * a zero page mapping on a read fault.
64 * This macro should be defined within <asm/pgtable.h>.
65 * s390 does this to prevent multiplexing of hardware bits
66 * related to the physical page in case of virtualization.
68 #ifndef mm_forbids_zeropage
69 #define mm_forbids_zeropage(X) (0)
72 extern unsigned long sysctl_user_reserve_kbytes
;
73 extern unsigned long sysctl_admin_reserve_kbytes
;
75 extern int sysctl_overcommit_memory
;
76 extern int sysctl_overcommit_ratio
;
77 extern unsigned long sysctl_overcommit_kbytes
;
79 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
81 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
84 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
86 /* to align the pointer to the (next) page boundary */
87 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
89 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
90 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
93 * Linux kernel virtual memory manager primitives.
94 * The idea being to have a "virtual" mm in the same way
95 * we have a virtual fs - giving a cleaner interface to the
96 * mm details, and allowing different kinds of memory mappings
97 * (from shared memory to executable loading to arbitrary
101 extern struct kmem_cache
*vm_area_cachep
;
104 extern struct rb_root nommu_region_tree
;
105 extern struct rw_semaphore nommu_region_sem
;
107 extern unsigned int kobjsize(const void *objp
);
111 * vm_flags in vm_area_struct, see mm_types.h.
113 #define VM_NONE 0x00000000
115 #define VM_READ 0x00000001 /* currently active flags */
116 #define VM_WRITE 0x00000002
117 #define VM_EXEC 0x00000004
118 #define VM_SHARED 0x00000008
120 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
121 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
122 #define VM_MAYWRITE 0x00000020
123 #define VM_MAYEXEC 0x00000040
124 #define VM_MAYSHARE 0x00000080
126 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
127 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
128 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
129 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
130 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
132 #define VM_LOCKED 0x00002000
133 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
135 /* Used by sys_madvise() */
136 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
137 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
139 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
140 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
141 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
142 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
143 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
144 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
145 #define VM_ARCH_2 0x02000000
146 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
148 #ifdef CONFIG_MEM_SOFT_DIRTY
149 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
151 # define VM_SOFTDIRTY 0
154 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
155 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
156 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
157 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
159 #if defined(CONFIG_X86)
160 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
161 #elif defined(CONFIG_PPC)
162 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
163 #elif defined(CONFIG_PARISC)
164 # define VM_GROWSUP VM_ARCH_1
165 #elif defined(CONFIG_METAG)
166 # define VM_GROWSUP VM_ARCH_1
167 #elif defined(CONFIG_IA64)
168 # define VM_GROWSUP VM_ARCH_1
169 #elif !defined(CONFIG_MMU)
170 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
173 #if defined(CONFIG_X86)
174 /* MPX specific bounds table or bounds directory */
175 # define VM_MPX VM_ARCH_2
179 # define VM_GROWSUP VM_NONE
182 /* Bits set in the VMA until the stack is in its final location */
183 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
185 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
186 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
189 #ifdef CONFIG_STACK_GROWSUP
190 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
192 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
196 * Special vmas that are non-mergable, non-mlock()able.
197 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
199 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
201 /* This mask defines which mm->def_flags a process can inherit its parent */
202 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
205 * mapping from the currently active vm_flags protection bits (the
206 * low four bits) to a page protection mask..
208 extern pgprot_t protection_map
[16];
210 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
211 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
212 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
213 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
214 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
215 #define FAULT_FLAG_TRIED 0x20 /* Second try */
216 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
219 * vm_fault is filled by the the pagefault handler and passed to the vma's
220 * ->fault function. The vma's ->fault is responsible for returning a bitmask
221 * of VM_FAULT_xxx flags that give details about how the fault was handled.
223 * pgoff should be used in favour of virtual_address, if possible.
226 unsigned int flags
; /* FAULT_FLAG_xxx flags */
227 pgoff_t pgoff
; /* Logical page offset based on vma */
228 void __user
*virtual_address
; /* Faulting virtual address */
230 struct page
*cow_page
; /* Handler may choose to COW */
231 struct page
*page
; /* ->fault handlers should return a
232 * page here, unless VM_FAULT_NOPAGE
233 * is set (which is also implied by
236 /* for ->map_pages() only */
237 pgoff_t max_pgoff
; /* map pages for offset from pgoff till
238 * max_pgoff inclusive */
239 pte_t
*pte
; /* pte entry associated with ->pgoff */
243 * These are the virtual MM functions - opening of an area, closing and
244 * unmapping it (needed to keep files on disk up-to-date etc), pointer
245 * to the functions called when a no-page or a wp-page exception occurs.
247 struct vm_operations_struct
{
248 void (*open
)(struct vm_area_struct
* area
);
249 void (*close
)(struct vm_area_struct
* area
);
250 int (*mremap
)(struct vm_area_struct
* area
);
251 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
252 int (*pmd_fault
)(struct vm_area_struct
*, unsigned long address
,
253 pmd_t
*, unsigned int flags
);
254 void (*map_pages
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
256 /* notification that a previously read-only page is about to become
257 * writable, if an error is returned it will cause a SIGBUS */
258 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
260 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
261 int (*pfn_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
263 /* called by access_process_vm when get_user_pages() fails, typically
264 * for use by special VMAs that can switch between memory and hardware
266 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
267 void *buf
, int len
, int write
);
269 /* Called by the /proc/PID/maps code to ask the vma whether it
270 * has a special name. Returning non-NULL will also cause this
271 * vma to be dumped unconditionally. */
272 const char *(*name
)(struct vm_area_struct
*vma
);
276 * set_policy() op must add a reference to any non-NULL @new mempolicy
277 * to hold the policy upon return. Caller should pass NULL @new to
278 * remove a policy and fall back to surrounding context--i.e. do not
279 * install a MPOL_DEFAULT policy, nor the task or system default
282 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
285 * get_policy() op must add reference [mpol_get()] to any policy at
286 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
287 * in mm/mempolicy.c will do this automatically.
288 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
289 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
290 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
291 * must return NULL--i.e., do not "fallback" to task or system default
294 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
298 * Called by vm_normal_page() for special PTEs to find the
299 * page for @addr. This is useful if the default behavior
300 * (using pte_page()) would not find the correct page.
302 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
309 #define page_private(page) ((page)->private)
310 #define set_page_private(page, v) ((page)->private = (v))
313 * FIXME: take this include out, include page-flags.h in
314 * files which need it (119 of them)
316 #include <linux/page-flags.h>
317 #include <linux/huge_mm.h>
320 * Methods to modify the page usage count.
322 * What counts for a page usage:
323 * - cache mapping (page->mapping)
324 * - private data (page->private)
325 * - page mapped in a task's page tables, each mapping
326 * is counted separately
328 * Also, many kernel routines increase the page count before a critical
329 * routine so they can be sure the page doesn't go away from under them.
333 * Drop a ref, return true if the refcount fell to zero (the page has no users)
335 static inline int put_page_testzero(struct page
*page
)
337 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) == 0, page
);
338 return atomic_dec_and_test(&page
->_count
);
342 * Try to grab a ref unless the page has a refcount of zero, return false if
344 * This can be called when MMU is off so it must not access
345 * any of the virtual mappings.
347 static inline int get_page_unless_zero(struct page
*page
)
349 return atomic_inc_not_zero(&page
->_count
);
352 extern int page_is_ram(unsigned long pfn
);
353 extern int region_is_ram(resource_size_t phys_addr
, unsigned long size
);
355 /* Support for virtually mapped pages */
356 struct page
*vmalloc_to_page(const void *addr
);
357 unsigned long vmalloc_to_pfn(const void *addr
);
360 * Determine if an address is within the vmalloc range
362 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
363 * is no special casing required.
365 static inline int is_vmalloc_addr(const void *x
)
368 unsigned long addr
= (unsigned long)x
;
370 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
376 extern int is_vmalloc_or_module_addr(const void *x
);
378 static inline int is_vmalloc_or_module_addr(const void *x
)
384 extern void kvfree(const void *addr
);
386 static inline void compound_lock(struct page
*page
)
388 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
389 VM_BUG_ON_PAGE(PageSlab(page
), page
);
390 bit_spin_lock(PG_compound_lock
, &page
->flags
);
394 static inline void compound_unlock(struct page
*page
)
396 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
397 VM_BUG_ON_PAGE(PageSlab(page
), page
);
398 bit_spin_unlock(PG_compound_lock
, &page
->flags
);
402 static inline unsigned long compound_lock_irqsave(struct page
*page
)
404 unsigned long uninitialized_var(flags
);
405 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
406 local_irq_save(flags
);
412 static inline void compound_unlock_irqrestore(struct page
*page
,
415 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
416 compound_unlock(page
);
417 local_irq_restore(flags
);
421 static inline struct page
*compound_head_by_tail(struct page
*tail
)
423 struct page
*head
= tail
->first_page
;
426 * page->first_page may be a dangling pointer to an old
427 * compound page, so recheck that it is still a tail
428 * page before returning.
431 if (likely(PageTail(tail
)))
437 * Since either compound page could be dismantled asynchronously in THP
438 * or we access asynchronously arbitrary positioned struct page, there
439 * would be tail flag race. To handle this race, we should call
440 * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
442 static inline struct page
*compound_head(struct page
*page
)
444 if (unlikely(PageTail(page
)))
445 return compound_head_by_tail(page
);
450 * If we access compound page synchronously such as access to
451 * allocated page, there is no need to handle tail flag race, so we can
452 * check tail flag directly without any synchronization primitive.
454 static inline struct page
*compound_head_fast(struct page
*page
)
456 if (unlikely(PageTail(page
)))
457 return page
->first_page
;
462 * The atomic page->_mapcount, starts from -1: so that transitions
463 * both from it and to it can be tracked, using atomic_inc_and_test
464 * and atomic_add_negative(-1).
466 static inline void page_mapcount_reset(struct page
*page
)
468 atomic_set(&(page
)->_mapcount
, -1);
471 static inline int page_mapcount(struct page
*page
)
473 VM_BUG_ON_PAGE(PageSlab(page
), page
);
474 return atomic_read(&page
->_mapcount
) + 1;
477 static inline int page_count(struct page
*page
)
479 return atomic_read(&compound_head(page
)->_count
);
482 static inline bool __compound_tail_refcounted(struct page
*page
)
484 return PageAnon(page
) && !PageSlab(page
) && !PageHeadHuge(page
);
488 * This takes a head page as parameter and tells if the
489 * tail page reference counting can be skipped.
491 * For this to be safe, PageSlab and PageHeadHuge must remain true on
492 * any given page where they return true here, until all tail pins
493 * have been released.
495 static inline bool compound_tail_refcounted(struct page
*page
)
497 VM_BUG_ON_PAGE(!PageHead(page
), page
);
498 return __compound_tail_refcounted(page
);
501 static inline void get_huge_page_tail(struct page
*page
)
504 * __split_huge_page_refcount() cannot run from under us.
506 VM_BUG_ON_PAGE(!PageTail(page
), page
);
507 VM_BUG_ON_PAGE(page_mapcount(page
) < 0, page
);
508 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) != 0, page
);
509 if (compound_tail_refcounted(page
->first_page
))
510 atomic_inc(&page
->_mapcount
);
513 extern bool __get_page_tail(struct page
*page
);
515 static inline void get_page(struct page
*page
)
517 if (unlikely(PageTail(page
)))
518 if (likely(__get_page_tail(page
)))
521 * Getting a normal page or the head of a compound page
522 * requires to already have an elevated page->_count.
524 VM_BUG_ON_PAGE(atomic_read(&page
->_count
) <= 0, page
);
525 atomic_inc(&page
->_count
);
528 static inline struct page
*virt_to_head_page(const void *x
)
530 struct page
*page
= virt_to_page(x
);
533 * We don't need to worry about synchronization of tail flag
534 * when we call virt_to_head_page() since it is only called for
535 * already allocated page and this page won't be freed until
536 * this virt_to_head_page() is finished. So use _fast variant.
538 return compound_head_fast(page
);
542 * Setup the page count before being freed into the page allocator for
543 * the first time (boot or memory hotplug)
545 static inline void init_page_count(struct page
*page
)
547 atomic_set(&page
->_count
, 1);
550 void put_page(struct page
*page
);
551 void put_pages_list(struct list_head
*pages
);
553 void split_page(struct page
*page
, unsigned int order
);
554 int split_free_page(struct page
*page
);
557 * Compound pages have a destructor function. Provide a
558 * prototype for that function and accessor functions.
559 * These are _only_ valid on the head of a PG_compound page.
562 static inline void set_compound_page_dtor(struct page
*page
,
563 compound_page_dtor
*dtor
)
565 page
[1].compound_dtor
= dtor
;
568 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
570 return page
[1].compound_dtor
;
573 static inline int compound_order(struct page
*page
)
577 return page
[1].compound_order
;
580 static inline void set_compound_order(struct page
*page
, unsigned long order
)
582 page
[1].compound_order
= order
;
587 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
588 * servicing faults for write access. In the normal case, do always want
589 * pte_mkwrite. But get_user_pages can cause write faults for mappings
590 * that do not have writing enabled, when used by access_process_vm.
592 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
594 if (likely(vma
->vm_flags
& VM_WRITE
))
595 pte
= pte_mkwrite(pte
);
599 void do_set_pte(struct vm_area_struct
*vma
, unsigned long address
,
600 struct page
*page
, pte_t
*pte
, bool write
, bool anon
);
604 * Multiple processes may "see" the same page. E.g. for untouched
605 * mappings of /dev/null, all processes see the same page full of
606 * zeroes, and text pages of executables and shared libraries have
607 * only one copy in memory, at most, normally.
609 * For the non-reserved pages, page_count(page) denotes a reference count.
610 * page_count() == 0 means the page is free. page->lru is then used for
611 * freelist management in the buddy allocator.
612 * page_count() > 0 means the page has been allocated.
614 * Pages are allocated by the slab allocator in order to provide memory
615 * to kmalloc and kmem_cache_alloc. In this case, the management of the
616 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
617 * unless a particular usage is carefully commented. (the responsibility of
618 * freeing the kmalloc memory is the caller's, of course).
620 * A page may be used by anyone else who does a __get_free_page().
621 * In this case, page_count still tracks the references, and should only
622 * be used through the normal accessor functions. The top bits of page->flags
623 * and page->virtual store page management information, but all other fields
624 * are unused and could be used privately, carefully. The management of this
625 * page is the responsibility of the one who allocated it, and those who have
626 * subsequently been given references to it.
628 * The other pages (we may call them "pagecache pages") are completely
629 * managed by the Linux memory manager: I/O, buffers, swapping etc.
630 * The following discussion applies only to them.
632 * A pagecache page contains an opaque `private' member, which belongs to the
633 * page's address_space. Usually, this is the address of a circular list of
634 * the page's disk buffers. PG_private must be set to tell the VM to call
635 * into the filesystem to release these pages.
637 * A page may belong to an inode's memory mapping. In this case, page->mapping
638 * is the pointer to the inode, and page->index is the file offset of the page,
639 * in units of PAGE_CACHE_SIZE.
641 * If pagecache pages are not associated with an inode, they are said to be
642 * anonymous pages. These may become associated with the swapcache, and in that
643 * case PG_swapcache is set, and page->private is an offset into the swapcache.
645 * In either case (swapcache or inode backed), the pagecache itself holds one
646 * reference to the page. Setting PG_private should also increment the
647 * refcount. The each user mapping also has a reference to the page.
649 * The pagecache pages are stored in a per-mapping radix tree, which is
650 * rooted at mapping->page_tree, and indexed by offset.
651 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
652 * lists, we instead now tag pages as dirty/writeback in the radix tree.
654 * All pagecache pages may be subject to I/O:
655 * - inode pages may need to be read from disk,
656 * - inode pages which have been modified and are MAP_SHARED may need
657 * to be written back to the inode on disk,
658 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
659 * modified may need to be swapped out to swap space and (later) to be read
664 * The zone field is never updated after free_area_init_core()
665 * sets it, so none of the operations on it need to be atomic.
668 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
669 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
670 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
671 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
672 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
675 * Define the bit shifts to access each section. For non-existent
676 * sections we define the shift as 0; that plus a 0 mask ensures
677 * the compiler will optimise away reference to them.
679 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
680 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
681 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
682 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
684 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
685 #ifdef NODE_NOT_IN_PAGE_FLAGS
686 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
687 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
688 SECTIONS_PGOFF : ZONES_PGOFF)
690 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
691 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
692 NODES_PGOFF : ZONES_PGOFF)
695 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
697 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
698 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
701 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
702 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
703 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
704 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
705 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
707 static inline enum zone_type
page_zonenum(const struct page
*page
)
709 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
712 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
713 #define SECTION_IN_PAGE_FLAGS
717 * The identification function is mainly used by the buddy allocator for
718 * determining if two pages could be buddies. We are not really identifying
719 * the zone since we could be using the section number id if we do not have
720 * node id available in page flags.
721 * We only guarantee that it will return the same value for two combinable
724 static inline int page_zone_id(struct page
*page
)
726 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
729 static inline int zone_to_nid(struct zone
*zone
)
738 #ifdef NODE_NOT_IN_PAGE_FLAGS
739 extern int page_to_nid(const struct page
*page
);
741 static inline int page_to_nid(const struct page
*page
)
743 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
747 #ifdef CONFIG_NUMA_BALANCING
748 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
750 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
753 static inline int cpupid_to_pid(int cpupid
)
755 return cpupid
& LAST__PID_MASK
;
758 static inline int cpupid_to_cpu(int cpupid
)
760 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
763 static inline int cpupid_to_nid(int cpupid
)
765 return cpu_to_node(cpupid_to_cpu(cpupid
));
768 static inline bool cpupid_pid_unset(int cpupid
)
770 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
773 static inline bool cpupid_cpu_unset(int cpupid
)
775 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
778 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
780 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
783 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
784 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
785 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
787 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
790 static inline int page_cpupid_last(struct page
*page
)
792 return page
->_last_cpupid
;
794 static inline void page_cpupid_reset_last(struct page
*page
)
796 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
799 static inline int page_cpupid_last(struct page
*page
)
801 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
804 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
806 static inline void page_cpupid_reset_last(struct page
*page
)
808 int cpupid
= (1 << LAST_CPUPID_SHIFT
) - 1;
810 page
->flags
&= ~(LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
);
811 page
->flags
|= (cpupid
& LAST_CPUPID_MASK
) << LAST_CPUPID_PGSHIFT
;
813 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
814 #else /* !CONFIG_NUMA_BALANCING */
815 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
817 return page_to_nid(page
); /* XXX */
820 static inline int page_cpupid_last(struct page
*page
)
822 return page_to_nid(page
); /* XXX */
825 static inline int cpupid_to_nid(int cpupid
)
830 static inline int cpupid_to_pid(int cpupid
)
835 static inline int cpupid_to_cpu(int cpupid
)
840 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
845 static inline bool cpupid_pid_unset(int cpupid
)
850 static inline void page_cpupid_reset_last(struct page
*page
)
854 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
858 #endif /* CONFIG_NUMA_BALANCING */
860 static inline struct zone
*page_zone(const struct page
*page
)
862 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
865 #ifdef SECTION_IN_PAGE_FLAGS
866 static inline void set_page_section(struct page
*page
, unsigned long section
)
868 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
869 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
872 static inline unsigned long page_to_section(const struct page
*page
)
874 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
878 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
880 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
881 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
884 static inline void set_page_node(struct page
*page
, unsigned long node
)
886 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
887 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
890 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
891 unsigned long node
, unsigned long pfn
)
893 set_page_zone(page
, zone
);
894 set_page_node(page
, node
);
895 #ifdef SECTION_IN_PAGE_FLAGS
896 set_page_section(page
, pfn_to_section_nr(pfn
));
901 * Some inline functions in vmstat.h depend on page_zone()
903 #include <linux/vmstat.h>
905 static __always_inline
void *lowmem_page_address(const struct page
*page
)
907 return __va(PFN_PHYS(page_to_pfn(page
)));
910 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
911 #define HASHED_PAGE_VIRTUAL
914 #if defined(WANT_PAGE_VIRTUAL)
915 static inline void *page_address(const struct page
*page
)
917 return page
->virtual;
919 static inline void set_page_address(struct page
*page
, void *address
)
921 page
->virtual = address
;
923 #define page_address_init() do { } while(0)
926 #if defined(HASHED_PAGE_VIRTUAL)
927 void *page_address(const struct page
*page
);
928 void set_page_address(struct page
*page
, void *virtual);
929 void page_address_init(void);
932 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
933 #define page_address(page) lowmem_page_address(page)
934 #define set_page_address(page, address) do { } while(0)
935 #define page_address_init() do { } while(0)
938 extern void *page_rmapping(struct page
*page
);
939 extern struct anon_vma
*page_anon_vma(struct page
*page
);
940 extern struct address_space
*page_mapping(struct page
*page
);
942 extern struct address_space
*__page_file_mapping(struct page
*);
945 struct address_space
*page_file_mapping(struct page
*page
)
947 if (unlikely(PageSwapCache(page
)))
948 return __page_file_mapping(page
);
950 return page
->mapping
;
954 * Return the pagecache index of the passed page. Regular pagecache pages
955 * use ->index whereas swapcache pages use ->private
957 static inline pgoff_t
page_index(struct page
*page
)
959 if (unlikely(PageSwapCache(page
)))
960 return page_private(page
);
964 extern pgoff_t
__page_file_index(struct page
*page
);
967 * Return the file index of the page. Regular pagecache pages use ->index
968 * whereas swapcache pages use swp_offset(->private)
970 static inline pgoff_t
page_file_index(struct page
*page
)
972 if (unlikely(PageSwapCache(page
)))
973 return __page_file_index(page
);
979 * Return true if this page is mapped into pagetables.
981 static inline int page_mapped(struct page
*page
)
983 return atomic_read(&(page
)->_mapcount
) >= 0;
987 * Return true only if the page has been allocated with
988 * ALLOC_NO_WATERMARKS and the low watermark was not
989 * met implying that the system is under some pressure.
991 static inline bool page_is_pfmemalloc(struct page
*page
)
994 * Page index cannot be this large so this must be
997 return page
->index
== -1UL;
1001 * Only to be called by the page allocator on a freshly allocated
1004 static inline void set_page_pfmemalloc(struct page
*page
)
1009 static inline void clear_page_pfmemalloc(struct page
*page
)
1015 * Different kinds of faults, as returned by handle_mm_fault().
1016 * Used to decide whether a process gets delivered SIGBUS or
1017 * just gets major/minor fault counters bumped up.
1020 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1022 #define VM_FAULT_OOM 0x0001
1023 #define VM_FAULT_SIGBUS 0x0002
1024 #define VM_FAULT_MAJOR 0x0004
1025 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1026 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1027 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1028 #define VM_FAULT_SIGSEGV 0x0040
1030 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1031 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1032 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1033 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1035 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1037 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1038 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1041 /* Encode hstate index for a hwpoisoned large page */
1042 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1043 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1046 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1048 extern void pagefault_out_of_memory(void);
1050 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1053 * Flags passed to show_mem() and show_free_areas() to suppress output in
1056 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1058 extern void show_free_areas(unsigned int flags
);
1059 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
1061 int shmem_zero_setup(struct vm_area_struct
*);
1063 bool shmem_mapping(struct address_space
*mapping
);
1065 static inline bool shmem_mapping(struct address_space
*mapping
)
1071 extern int can_do_mlock(void);
1072 extern int user_shm_lock(size_t, struct user_struct
*);
1073 extern void user_shm_unlock(size_t, struct user_struct
*);
1076 * Parameter block passed down to zap_pte_range in exceptional cases.
1078 struct zap_details
{
1079 struct address_space
*check_mapping
; /* Check page->mapping if set */
1080 pgoff_t first_index
; /* Lowest page->index to unmap */
1081 pgoff_t last_index
; /* Highest page->index to unmap */
1084 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1087 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1088 unsigned long size
);
1089 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1090 unsigned long size
, struct zap_details
*);
1091 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1092 unsigned long start
, unsigned long end
);
1095 * mm_walk - callbacks for walk_page_range
1096 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1097 * this handler is required to be able to handle
1098 * pmd_trans_huge() pmds. They may simply choose to
1099 * split_huge_page() instead of handling it explicitly.
1100 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1101 * @pte_hole: if set, called for each hole at all levels
1102 * @hugetlb_entry: if set, called for each hugetlb entry
1103 * @test_walk: caller specific callback function to determine whether
1104 * we walk over the current vma or not. A positive returned
1105 * value means "do page table walk over the current vma,"
1106 * and a negative one means "abort current page table walk
1107 * right now." 0 means "skip the current vma."
1108 * @mm: mm_struct representing the target process of page table walk
1109 * @vma: vma currently walked (NULL if walking outside vmas)
1110 * @private: private data for callbacks' usage
1112 * (see the comment on walk_page_range() for more details)
1115 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1116 unsigned long next
, struct mm_walk
*walk
);
1117 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1118 unsigned long next
, struct mm_walk
*walk
);
1119 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1120 struct mm_walk
*walk
);
1121 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1122 unsigned long addr
, unsigned long next
,
1123 struct mm_walk
*walk
);
1124 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1125 struct mm_walk
*walk
);
1126 struct mm_struct
*mm
;
1127 struct vm_area_struct
*vma
;
1131 int walk_page_range(unsigned long addr
, unsigned long end
,
1132 struct mm_walk
*walk
);
1133 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1134 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1135 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1136 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1137 struct vm_area_struct
*vma
);
1138 void unmap_mapping_range(struct address_space
*mapping
,
1139 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1140 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1141 unsigned long *pfn
);
1142 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1143 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1144 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1145 void *buf
, int len
, int write
);
1147 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1148 loff_t
const holebegin
, loff_t
const holelen
)
1150 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1153 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1154 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1155 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1156 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1157 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1158 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1159 int invalidate_inode_page(struct page
*page
);
1162 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1163 unsigned long address
, unsigned int flags
);
1164 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1165 unsigned long address
, unsigned int fault_flags
);
1167 static inline int handle_mm_fault(struct mm_struct
*mm
,
1168 struct vm_area_struct
*vma
, unsigned long address
,
1171 /* should never happen if there's no MMU */
1173 return VM_FAULT_SIGBUS
;
1175 static inline int fixup_user_fault(struct task_struct
*tsk
,
1176 struct mm_struct
*mm
, unsigned long address
,
1177 unsigned int fault_flags
)
1179 /* should never happen if there's no MMU */
1185 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1186 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1187 void *buf
, int len
, int write
);
1189 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1190 unsigned long start
, unsigned long nr_pages
,
1191 unsigned int foll_flags
, struct page
**pages
,
1192 struct vm_area_struct
**vmas
, int *nonblocking
);
1193 long get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1194 unsigned long start
, unsigned long nr_pages
,
1195 int write
, int force
, struct page
**pages
,
1196 struct vm_area_struct
**vmas
);
1197 long get_user_pages_locked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1198 unsigned long start
, unsigned long nr_pages
,
1199 int write
, int force
, struct page
**pages
,
1201 long __get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1202 unsigned long start
, unsigned long nr_pages
,
1203 int write
, int force
, struct page
**pages
,
1204 unsigned int gup_flags
);
1205 long get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1206 unsigned long start
, unsigned long nr_pages
,
1207 int write
, int force
, struct page
**pages
);
1208 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1209 struct page
**pages
);
1211 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1212 struct page
**pages
);
1213 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1214 struct page
*get_dump_page(unsigned long addr
);
1216 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1217 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1218 unsigned int length
);
1220 int __set_page_dirty_nobuffers(struct page
*page
);
1221 int __set_page_dirty_no_writeback(struct page
*page
);
1222 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1224 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
,
1225 struct mem_cgroup
*memcg
);
1226 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1227 struct mem_cgroup
*memcg
, struct bdi_writeback
*wb
);
1228 int set_page_dirty(struct page
*page
);
1229 int set_page_dirty_lock(struct page
*page
);
1230 void cancel_dirty_page(struct page
*page
);
1231 int clear_page_dirty_for_io(struct page
*page
);
1233 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1235 /* Is the vma a continuation of the stack vma above it? */
1236 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1238 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1241 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1243 return !vma
->vm_ops
;
1246 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1249 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1250 (vma
->vm_start
== addr
) &&
1251 !vma_growsdown(vma
->vm_prev
, addr
);
1254 /* Is the vma a continuation of the stack vma below it? */
1255 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1257 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1260 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1263 return (vma
->vm_flags
& VM_GROWSUP
) &&
1264 (vma
->vm_end
== addr
) &&
1265 !vma_growsup(vma
->vm_next
, addr
);
1268 extern struct task_struct
*task_of_stack(struct task_struct
*task
,
1269 struct vm_area_struct
*vma
, bool in_group
);
1271 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1272 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1273 unsigned long new_addr
, unsigned long len
,
1274 bool need_rmap_locks
);
1275 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1276 unsigned long end
, pgprot_t newprot
,
1277 int dirty_accountable
, int prot_numa
);
1278 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1279 struct vm_area_struct
**pprev
, unsigned long start
,
1280 unsigned long end
, unsigned long newflags
);
1283 * doesn't attempt to fault and will return short.
1285 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1286 struct page
**pages
);
1288 * per-process(per-mm_struct) statistics.
1290 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1292 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1294 #ifdef SPLIT_RSS_COUNTING
1296 * counter is updated in asynchronous manner and may go to minus.
1297 * But it's never be expected number for users.
1302 return (unsigned long)val
;
1305 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1307 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1310 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1312 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1315 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1317 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1320 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1322 return get_mm_counter(mm
, MM_FILEPAGES
) +
1323 get_mm_counter(mm
, MM_ANONPAGES
);
1326 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1328 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1331 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1333 return max(mm
->hiwater_vm
, mm
->total_vm
);
1336 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1338 unsigned long _rss
= get_mm_rss(mm
);
1340 if ((mm
)->hiwater_rss
< _rss
)
1341 (mm
)->hiwater_rss
= _rss
;
1344 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1346 if (mm
->hiwater_vm
< mm
->total_vm
)
1347 mm
->hiwater_vm
= mm
->total_vm
;
1350 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1352 mm
->hiwater_rss
= get_mm_rss(mm
);
1355 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1356 struct mm_struct
*mm
)
1358 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1360 if (*maxrss
< hiwater_rss
)
1361 *maxrss
= hiwater_rss
;
1364 #if defined(SPLIT_RSS_COUNTING)
1365 void sync_mm_rss(struct mm_struct
*mm
);
1367 static inline void sync_mm_rss(struct mm_struct
*mm
)
1372 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1374 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1376 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1380 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1384 #ifdef __PAGETABLE_PUD_FOLDED
1385 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1386 unsigned long address
)
1391 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1394 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1395 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1396 unsigned long address
)
1401 static inline void mm_nr_pmds_init(struct mm_struct
*mm
) {}
1403 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1408 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1409 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1412 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1414 static inline void mm_nr_pmds_init(struct mm_struct
*mm
)
1416 atomic_long_set(&mm
->nr_pmds
, 0);
1419 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1421 return atomic_long_read(&mm
->nr_pmds
);
1424 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1426 atomic_long_inc(&mm
->nr_pmds
);
1429 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1431 atomic_long_dec(&mm
->nr_pmds
);
1435 int __pte_alloc(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1436 pmd_t
*pmd
, unsigned long address
);
1437 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1440 * The following ifdef needed to get the 4level-fixup.h header to work.
1441 * Remove it when 4level-fixup.h has been removed.
1443 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1444 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1446 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1447 NULL
: pud_offset(pgd
, address
);
1450 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1452 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1453 NULL
: pmd_offset(pud
, address
);
1455 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1457 #if USE_SPLIT_PTE_PTLOCKS
1458 #if ALLOC_SPLIT_PTLOCKS
1459 void __init
ptlock_cache_init(void);
1460 extern bool ptlock_alloc(struct page
*page
);
1461 extern void ptlock_free(struct page
*page
);
1463 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1467 #else /* ALLOC_SPLIT_PTLOCKS */
1468 static inline void ptlock_cache_init(void)
1472 static inline bool ptlock_alloc(struct page
*page
)
1477 static inline void ptlock_free(struct page
*page
)
1481 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1485 #endif /* ALLOC_SPLIT_PTLOCKS */
1487 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1489 return ptlock_ptr(pmd_page(*pmd
));
1492 static inline bool ptlock_init(struct page
*page
)
1495 * prep_new_page() initialize page->private (and therefore page->ptl)
1496 * with 0. Make sure nobody took it in use in between.
1498 * It can happen if arch try to use slab for page table allocation:
1499 * slab code uses page->slab_cache and page->first_page (for tail
1500 * pages), which share storage with page->ptl.
1502 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1503 if (!ptlock_alloc(page
))
1505 spin_lock_init(ptlock_ptr(page
));
1509 /* Reset page->mapping so free_pages_check won't complain. */
1510 static inline void pte_lock_deinit(struct page
*page
)
1512 page
->mapping
= NULL
;
1516 #else /* !USE_SPLIT_PTE_PTLOCKS */
1518 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1520 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1522 return &mm
->page_table_lock
;
1524 static inline void ptlock_cache_init(void) {}
1525 static inline bool ptlock_init(struct page
*page
) { return true; }
1526 static inline void pte_lock_deinit(struct page
*page
) {}
1527 #endif /* USE_SPLIT_PTE_PTLOCKS */
1529 static inline void pgtable_init(void)
1531 ptlock_cache_init();
1532 pgtable_cache_init();
1535 static inline bool pgtable_page_ctor(struct page
*page
)
1537 inc_zone_page_state(page
, NR_PAGETABLE
);
1538 return ptlock_init(page
);
1541 static inline void pgtable_page_dtor(struct page
*page
)
1543 pte_lock_deinit(page
);
1544 dec_zone_page_state(page
, NR_PAGETABLE
);
1547 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1549 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1550 pte_t *__pte = pte_offset_map(pmd, address); \
1556 #define pte_unmap_unlock(pte, ptl) do { \
1561 #define pte_alloc_map(mm, vma, pmd, address) \
1562 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1564 NULL: pte_offset_map(pmd, address))
1566 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1567 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1569 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1571 #define pte_alloc_kernel(pmd, address) \
1572 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1573 NULL: pte_offset_kernel(pmd, address))
1575 #if USE_SPLIT_PMD_PTLOCKS
1577 static struct page
*pmd_to_page(pmd_t
*pmd
)
1579 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1580 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1583 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1585 return ptlock_ptr(pmd_to_page(pmd
));
1588 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1590 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1591 page
->pmd_huge_pte
= NULL
;
1593 return ptlock_init(page
);
1596 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1598 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1599 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1604 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1608 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1610 return &mm
->page_table_lock
;
1613 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1614 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1616 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1620 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1622 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1627 extern void free_area_init(unsigned long * zones_size
);
1628 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1629 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1630 extern void free_initmem(void);
1633 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1634 * into the buddy system. The freed pages will be poisoned with pattern
1635 * "poison" if it's within range [0, UCHAR_MAX].
1636 * Return pages freed into the buddy system.
1638 extern unsigned long free_reserved_area(void *start
, void *end
,
1639 int poison
, char *s
);
1641 #ifdef CONFIG_HIGHMEM
1643 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1644 * and totalram_pages.
1646 extern void free_highmem_page(struct page
*page
);
1649 extern void adjust_managed_page_count(struct page
*page
, long count
);
1650 extern void mem_init_print_info(const char *str
);
1652 extern void reserve_bootmem_region(unsigned long start
, unsigned long end
);
1654 /* Free the reserved page into the buddy system, so it gets managed. */
1655 static inline void __free_reserved_page(struct page
*page
)
1657 ClearPageReserved(page
);
1658 init_page_count(page
);
1662 static inline void free_reserved_page(struct page
*page
)
1664 __free_reserved_page(page
);
1665 adjust_managed_page_count(page
, 1);
1668 static inline void mark_page_reserved(struct page
*page
)
1670 SetPageReserved(page
);
1671 adjust_managed_page_count(page
, -1);
1675 * Default method to free all the __init memory into the buddy system.
1676 * The freed pages will be poisoned with pattern "poison" if it's within
1677 * range [0, UCHAR_MAX].
1678 * Return pages freed into the buddy system.
1680 static inline unsigned long free_initmem_default(int poison
)
1682 extern char __init_begin
[], __init_end
[];
1684 return free_reserved_area(&__init_begin
, &__init_end
,
1685 poison
, "unused kernel");
1688 static inline unsigned long get_num_physpages(void)
1691 unsigned long phys_pages
= 0;
1693 for_each_online_node(nid
)
1694 phys_pages
+= node_present_pages(nid
);
1699 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1701 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1702 * zones, allocate the backing mem_map and account for memory holes in a more
1703 * architecture independent manner. This is a substitute for creating the
1704 * zone_sizes[] and zholes_size[] arrays and passing them to
1705 * free_area_init_node()
1707 * An architecture is expected to register range of page frames backed by
1708 * physical memory with memblock_add[_node]() before calling
1709 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1710 * usage, an architecture is expected to do something like
1712 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1714 * for_each_valid_physical_page_range()
1715 * memblock_add_node(base, size, nid)
1716 * free_area_init_nodes(max_zone_pfns);
1718 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1719 * registered physical page range. Similarly
1720 * sparse_memory_present_with_active_regions() calls memory_present() for
1721 * each range when SPARSEMEM is enabled.
1723 * See mm/page_alloc.c for more information on each function exposed by
1724 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1726 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1727 unsigned long node_map_pfn_alignment(void);
1728 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1729 unsigned long end_pfn
);
1730 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1731 unsigned long end_pfn
);
1732 extern void get_pfn_range_for_nid(unsigned int nid
,
1733 unsigned long *start_pfn
, unsigned long *end_pfn
);
1734 extern unsigned long find_min_pfn_with_active_regions(void);
1735 extern void free_bootmem_with_active_regions(int nid
,
1736 unsigned long max_low_pfn
);
1737 extern void sparse_memory_present_with_active_regions(int nid
);
1739 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1741 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1742 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1743 static inline int __early_pfn_to_nid(unsigned long pfn
,
1744 struct mminit_pfnnid_cache
*state
)
1749 /* please see mm/page_alloc.c */
1750 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1751 /* there is a per-arch backend function. */
1752 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
1753 struct mminit_pfnnid_cache
*state
);
1756 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1757 extern void memmap_init_zone(unsigned long, int, unsigned long,
1758 unsigned long, enum memmap_context
);
1759 extern void setup_per_zone_wmarks(void);
1760 extern int __meminit
init_per_zone_wmark_min(void);
1761 extern void mem_init(void);
1762 extern void __init
mmap_init(void);
1763 extern void show_mem(unsigned int flags
);
1764 extern void si_meminfo(struct sysinfo
* val
);
1765 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1767 extern __printf(3, 4)
1768 void warn_alloc_failed(gfp_t gfp_mask
, int order
, const char *fmt
, ...);
1770 extern void setup_per_cpu_pageset(void);
1772 extern void zone_pcp_update(struct zone
*zone
);
1773 extern void zone_pcp_reset(struct zone
*zone
);
1776 extern int min_free_kbytes
;
1779 extern atomic_long_t mmap_pages_allocated
;
1780 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1782 /* interval_tree.c */
1783 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1784 struct rb_root
*root
);
1785 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1786 struct vm_area_struct
*prev
,
1787 struct rb_root
*root
);
1788 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1789 struct rb_root
*root
);
1790 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1791 unsigned long start
, unsigned long last
);
1792 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1793 unsigned long start
, unsigned long last
);
1795 #define vma_interval_tree_foreach(vma, root, start, last) \
1796 for (vma = vma_interval_tree_iter_first(root, start, last); \
1797 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1799 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1800 struct rb_root
*root
);
1801 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1802 struct rb_root
*root
);
1803 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1804 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1805 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1806 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1807 #ifdef CONFIG_DEBUG_VM_RB
1808 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1811 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1812 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1813 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1816 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1817 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1818 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
1819 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1820 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1821 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1822 struct mempolicy
*, struct vm_userfaultfd_ctx
);
1823 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1824 extern int split_vma(struct mm_struct
*,
1825 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1826 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1827 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1828 struct rb_node
**, struct rb_node
*);
1829 extern void unlink_file_vma(struct vm_area_struct
*);
1830 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1831 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1832 bool *need_rmap_locks
);
1833 extern void exit_mmap(struct mm_struct
*);
1835 static inline int check_data_rlimit(unsigned long rlim
,
1837 unsigned long start
,
1838 unsigned long end_data
,
1839 unsigned long start_data
)
1841 if (rlim
< RLIM_INFINITY
) {
1842 if (((new - start
) + (end_data
- start_data
)) > rlim
)
1849 extern int mm_take_all_locks(struct mm_struct
*mm
);
1850 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1852 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1853 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
1855 extern int may_expand_vm(struct mm_struct
*mm
, unsigned long npages
);
1856 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
1857 unsigned long addr
, unsigned long len
,
1858 unsigned long flags
,
1859 const struct vm_special_mapping
*spec
);
1860 /* This is an obsolete alternative to _install_special_mapping. */
1861 extern int install_special_mapping(struct mm_struct
*mm
,
1862 unsigned long addr
, unsigned long len
,
1863 unsigned long flags
, struct page
**pages
);
1865 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
1867 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
1868 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
1869 extern unsigned long do_mmap_pgoff(struct file
*file
, unsigned long addr
,
1870 unsigned long len
, unsigned long prot
, unsigned long flags
,
1871 unsigned long pgoff
, unsigned long *populate
);
1872 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
1875 extern int __mm_populate(unsigned long addr
, unsigned long len
,
1877 static inline void mm_populate(unsigned long addr
, unsigned long len
)
1880 (void) __mm_populate(addr
, len
, 1);
1883 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
1886 /* These take the mm semaphore themselves */
1887 extern unsigned long vm_brk(unsigned long, unsigned long);
1888 extern int vm_munmap(unsigned long, size_t);
1889 extern unsigned long vm_mmap(struct file
*, unsigned long,
1890 unsigned long, unsigned long,
1891 unsigned long, unsigned long);
1893 struct vm_unmapped_area_info
{
1894 #define VM_UNMAPPED_AREA_TOPDOWN 1
1895 unsigned long flags
;
1896 unsigned long length
;
1897 unsigned long low_limit
;
1898 unsigned long high_limit
;
1899 unsigned long align_mask
;
1900 unsigned long align_offset
;
1903 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
1904 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
1907 * Search for an unmapped address range.
1909 * We are looking for a range that:
1910 * - does not intersect with any VMA;
1911 * - is contained within the [low_limit, high_limit) interval;
1912 * - is at least the desired size.
1913 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1915 static inline unsigned long
1916 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
1918 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
1919 return unmapped_area_topdown(info
);
1921 return unmapped_area(info
);
1925 extern void truncate_inode_pages(struct address_space
*, loff_t
);
1926 extern void truncate_inode_pages_range(struct address_space
*,
1927 loff_t lstart
, loff_t lend
);
1928 extern void truncate_inode_pages_final(struct address_space
*);
1930 /* generic vm_area_ops exported for stackable file systems */
1931 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
1932 extern void filemap_map_pages(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1933 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
1935 /* mm/page-writeback.c */
1936 int write_one_page(struct page
*page
, int wait
);
1937 void task_dirty_inc(struct task_struct
*tsk
);
1940 #define VM_MAX_READAHEAD 128 /* kbytes */
1941 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1943 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
1944 pgoff_t offset
, unsigned long nr_to_read
);
1946 void page_cache_sync_readahead(struct address_space
*mapping
,
1947 struct file_ra_state
*ra
,
1950 unsigned long size
);
1952 void page_cache_async_readahead(struct address_space
*mapping
,
1953 struct file_ra_state
*ra
,
1957 unsigned long size
);
1959 unsigned long max_sane_readahead(unsigned long nr
);
1961 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1962 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
1964 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1965 extern int expand_downwards(struct vm_area_struct
*vma
,
1966 unsigned long address
);
1968 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
1970 #define expand_upwards(vma, address) (0)
1973 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1974 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
1975 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
1976 struct vm_area_struct
**pprev
);
1978 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1979 NULL if none. Assume start_addr < end_addr. */
1980 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
1982 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
1984 if (vma
&& end_addr
<= vma
->vm_start
)
1989 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
1991 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
1994 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
1995 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
1996 unsigned long vm_start
, unsigned long vm_end
)
1998 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2000 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2007 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2008 void vma_set_page_prot(struct vm_area_struct
*vma
);
2010 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2014 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2016 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2020 #ifdef CONFIG_NUMA_BALANCING
2021 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2022 unsigned long start
, unsigned long end
);
2025 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2026 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2027 unsigned long pfn
, unsigned long size
, pgprot_t
);
2028 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2029 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2031 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2033 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2036 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2037 unsigned long address
, unsigned int foll_flags
,
2038 unsigned int *page_mask
);
2040 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2041 unsigned long address
, unsigned int foll_flags
)
2043 unsigned int unused_page_mask
;
2044 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2047 #define FOLL_WRITE 0x01 /* check pte is writable */
2048 #define FOLL_TOUCH 0x02 /* mark page accessed */
2049 #define FOLL_GET 0x04 /* do get_page on page */
2050 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2051 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2052 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2053 * and return without waiting upon it */
2054 #define FOLL_POPULATE 0x40 /* fault in page */
2055 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2056 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2057 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2058 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2059 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2061 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2063 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2064 unsigned long size
, pte_fn_t fn
, void *data
);
2066 #ifdef CONFIG_PROC_FS
2067 void vm_stat_account(struct mm_struct
*, unsigned long, struct file
*, long);
2069 static inline void vm_stat_account(struct mm_struct
*mm
,
2070 unsigned long flags
, struct file
*file
, long pages
)
2072 mm
->total_vm
+= pages
;
2074 #endif /* CONFIG_PROC_FS */
2076 #ifdef CONFIG_DEBUG_PAGEALLOC
2077 extern bool _debug_pagealloc_enabled
;
2078 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2080 static inline bool debug_pagealloc_enabled(void)
2082 return _debug_pagealloc_enabled
;
2086 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2088 if (!debug_pagealloc_enabled())
2091 __kernel_map_pages(page
, numpages
, enable
);
2093 #ifdef CONFIG_HIBERNATION
2094 extern bool kernel_page_present(struct page
*page
);
2095 #endif /* CONFIG_HIBERNATION */
2098 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2099 #ifdef CONFIG_HIBERNATION
2100 static inline bool kernel_page_present(struct page
*page
) { return true; }
2101 #endif /* CONFIG_HIBERNATION */
2104 #ifdef __HAVE_ARCH_GATE_AREA
2105 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2106 extern int in_gate_area_no_mm(unsigned long addr
);
2107 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2109 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2113 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2114 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2118 #endif /* __HAVE_ARCH_GATE_AREA */
2120 #ifdef CONFIG_SYSCTL
2121 extern int sysctl_drop_caches
;
2122 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2123 void __user
*, size_t *, loff_t
*);
2126 void drop_slab(void);
2127 void drop_slab_node(int nid
);
2130 #define randomize_va_space 0
2132 extern int randomize_va_space
;
2135 const char * arch_vma_name(struct vm_area_struct
*vma
);
2136 void print_vma_addr(char *prefix
, unsigned long rip
);
2138 void sparse_mem_maps_populate_node(struct page
**map_map
,
2139 unsigned long pnum_begin
,
2140 unsigned long pnum_end
,
2141 unsigned long map_count
,
2144 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2145 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2146 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2147 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2148 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2149 void *vmemmap_alloc_block(unsigned long size
, int node
);
2150 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
2151 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2152 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2154 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2155 void vmemmap_populate_print_last(void);
2156 #ifdef CONFIG_MEMORY_HOTPLUG
2157 void vmemmap_free(unsigned long start
, unsigned long end
);
2159 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2160 unsigned long size
);
2163 MF_COUNT_INCREASED
= 1 << 0,
2164 MF_ACTION_REQUIRED
= 1 << 1,
2165 MF_MUST_KILL
= 1 << 2,
2166 MF_SOFT_OFFLINE
= 1 << 3,
2168 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2169 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2170 extern int unpoison_memory(unsigned long pfn
);
2171 extern int get_hwpoison_page(struct page
*page
);
2172 extern int sysctl_memory_failure_early_kill
;
2173 extern int sysctl_memory_failure_recovery
;
2174 extern void shake_page(struct page
*p
, int access
);
2175 extern atomic_long_t num_poisoned_pages
;
2176 extern int soft_offline_page(struct page
*page
, int flags
);
2180 * Error handlers for various types of pages.
2183 MF_IGNORED
, /* Error: cannot be handled */
2184 MF_FAILED
, /* Error: handling failed */
2185 MF_DELAYED
, /* Will be handled later */
2186 MF_RECOVERED
, /* Successfully recovered */
2189 enum mf_action_page_type
{
2191 MF_MSG_KERNEL_HIGH_ORDER
,
2193 MF_MSG_DIFFERENT_COMPOUND
,
2194 MF_MSG_POISONED_HUGE
,
2197 MF_MSG_UNMAP_FAILED
,
2198 MF_MSG_DIRTY_SWAPCACHE
,
2199 MF_MSG_CLEAN_SWAPCACHE
,
2200 MF_MSG_DIRTY_MLOCKED_LRU
,
2201 MF_MSG_CLEAN_MLOCKED_LRU
,
2202 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2203 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2206 MF_MSG_TRUNCATED_LRU
,
2212 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2213 extern void clear_huge_page(struct page
*page
,
2215 unsigned int pages_per_huge_page
);
2216 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2217 unsigned long addr
, struct vm_area_struct
*vma
,
2218 unsigned int pages_per_huge_page
);
2219 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2221 extern struct page_ext_operations debug_guardpage_ops
;
2222 extern struct page_ext_operations page_poisoning_ops
;
2224 #ifdef CONFIG_DEBUG_PAGEALLOC
2225 extern unsigned int _debug_guardpage_minorder
;
2226 extern bool _debug_guardpage_enabled
;
2228 static inline unsigned int debug_guardpage_minorder(void)
2230 return _debug_guardpage_minorder
;
2233 static inline bool debug_guardpage_enabled(void)
2235 return _debug_guardpage_enabled
;
2238 static inline bool page_is_guard(struct page
*page
)
2240 struct page_ext
*page_ext
;
2242 if (!debug_guardpage_enabled())
2245 page_ext
= lookup_page_ext(page
);
2246 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2249 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2250 static inline bool debug_guardpage_enabled(void) { return false; }
2251 static inline bool page_is_guard(struct page
*page
) { return false; }
2252 #endif /* CONFIG_DEBUG_PAGEALLOC */
2254 #if MAX_NUMNODES > 1
2255 void __init
setup_nr_node_ids(void);
2257 static inline void setup_nr_node_ids(void) {}
2260 #endif /* __KERNEL__ */
2261 #endif /* _LINUX_MM_H */