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 * MM layer fills up gfp_mask for page allocations but fault handler might
229 * alter it if its implementation requires a different allocation context.
231 * pgoff should be used in favour of virtual_address, if possible.
234 unsigned int flags; /* FAULT_FLAG_xxx flags */
235 gfp_t gfp_mask; /* gfp mask to be used for allocations */
236 pgoff_t pgoff; /* Logical page offset based on vma */
237 void __user *virtual_address; /* Faulting virtual address */
239 struct page *cow_page; /* Handler may choose to COW */
240 struct page *page; /* ->fault handlers should return a
241 * page here, unless VM_FAULT_NOPAGE
242 * is set (which is also implied by
245 /* for ->map_pages() only */
246 pgoff_t max_pgoff; /* map pages for offset from pgoff till
247 * max_pgoff inclusive */
248 pte_t *pte; /* pte entry associated with ->pgoff */
252 * These are the virtual MM functions - opening of an area, closing and
253 * unmapping it (needed to keep files on disk up-to-date etc), pointer
254 * to the functions called when a no-page or a wp-page exception occurs.
256 struct vm_operations_struct {
257 void (*open)(struct vm_area_struct * area);
258 void (*close)(struct vm_area_struct * area);
259 int (*mremap)(struct vm_area_struct * area);
260 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
261 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
262 pmd_t *, unsigned int flags);
263 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
265 /* notification that a previously read-only page is about to become
266 * writable, if an error is returned it will cause a SIGBUS */
267 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
269 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
270 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
272 /* called by access_process_vm when get_user_pages() fails, typically
273 * for use by special VMAs that can switch between memory and hardware
275 int (*access)(struct vm_area_struct *vma, unsigned long addr,
276 void *buf, int len, int write);
278 /* Called by the /proc/PID/maps code to ask the vma whether it
279 * has a special name. Returning non-NULL will also cause this
280 * vma to be dumped unconditionally. */
281 const char *(*name)(struct vm_area_struct *vma);
285 * set_policy() op must add a reference to any non-NULL @new mempolicy
286 * to hold the policy upon return. Caller should pass NULL @new to
287 * remove a policy and fall back to surrounding context--i.e. do not
288 * install a MPOL_DEFAULT policy, nor the task or system default
291 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
294 * get_policy() op must add reference [mpol_get()] to any policy at
295 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
296 * in mm/mempolicy.c will do this automatically.
297 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
298 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
299 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
300 * must return NULL--i.e., do not "fallback" to task or system default
303 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
307 * Called by vm_normal_page() for special PTEs to find the
308 * page for @addr. This is useful if the default behavior
309 * (using pte_page()) would not find the correct page.
311 struct page *(*find_special_page)(struct vm_area_struct *vma,
318 #define page_private(page) ((page)->private)
319 #define set_page_private(page, v) ((page)->private = (v))
322 * FIXME: take this include out, include page-flags.h in
323 * files which need it (119 of them)
325 #include <linux/page-flags.h>
326 #include <linux/huge_mm.h>
329 * Methods to modify the page usage count.
331 * What counts for a page usage:
332 * - cache mapping (page->mapping)
333 * - private data (page->private)
334 * - page mapped in a task's page tables, each mapping
335 * is counted separately
337 * Also, many kernel routines increase the page count before a critical
338 * routine so they can be sure the page doesn't go away from under them.
342 * Drop a ref, return true if the refcount fell to zero (the page has no users)
344 static inline int put_page_testzero(struct page *page)
346 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
347 return atomic_dec_and_test(&page->_count);
351 * Try to grab a ref unless the page has a refcount of zero, return false if
353 * This can be called when MMU is off so it must not access
354 * any of the virtual mappings.
356 static inline int get_page_unless_zero(struct page *page)
358 return atomic_inc_not_zero(&page->_count);
361 extern int page_is_ram(unsigned long pfn);
369 int region_intersects(resource_size_t offset, size_t size, const char *type);
371 /* Support for virtually mapped pages */
372 struct page *vmalloc_to_page(const void *addr);
373 unsigned long vmalloc_to_pfn(const void *addr);
376 * Determine if an address is within the vmalloc range
378 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
379 * is no special casing required.
381 static inline int is_vmalloc_addr(const void *x)
384 unsigned long addr = (unsigned long)x;
386 return addr >= VMALLOC_START && addr < VMALLOC_END;
392 extern int is_vmalloc_or_module_addr(const void *x);
394 static inline int is_vmalloc_or_module_addr(const void *x)
400 extern void kvfree(const void *addr);
402 static inline void compound_lock(struct page *page)
404 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
405 VM_BUG_ON_PAGE(PageSlab(page), page);
406 bit_spin_lock(PG_compound_lock, &page->flags);
410 static inline void compound_unlock(struct page *page)
412 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
413 VM_BUG_ON_PAGE(PageSlab(page), page);
414 bit_spin_unlock(PG_compound_lock, &page->flags);
418 static inline unsigned long compound_lock_irqsave(struct page *page)
420 unsigned long uninitialized_var(flags);
421 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
422 local_irq_save(flags);
428 static inline void compound_unlock_irqrestore(struct page *page,
431 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
432 compound_unlock(page);
433 local_irq_restore(flags);
438 * The atomic page->_mapcount, starts from -1: so that transitions
439 * both from it and to it can be tracked, using atomic_inc_and_test
440 * and atomic_add_negative(-1).
442 static inline void page_mapcount_reset(struct page *page)
444 atomic_set(&(page)->_mapcount, -1);
447 static inline int page_mapcount(struct page *page)
449 return atomic_read(&page->_mapcount) + 1;
452 static inline int page_count(struct page *page)
454 return atomic_read(&compound_head(page)->_count);
457 static inline bool __compound_tail_refcounted(struct page *page)
459 return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
463 * This takes a head page as parameter and tells if the
464 * tail page reference counting can be skipped.
466 * For this to be safe, PageSlab and PageHeadHuge must remain true on
467 * any given page where they return true here, until all tail pins
468 * have been released.
470 static inline bool compound_tail_refcounted(struct page *page)
472 VM_BUG_ON_PAGE(!PageHead(page), page);
473 return __compound_tail_refcounted(page);
476 static inline void get_huge_page_tail(struct page *page)
479 * __split_huge_page_refcount() cannot run from under us.
481 VM_BUG_ON_PAGE(!PageTail(page), page);
482 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
483 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
484 if (compound_tail_refcounted(compound_head(page)))
485 atomic_inc(&page->_mapcount);
488 extern bool __get_page_tail(struct page *page);
490 static inline int page_ref_count(struct page *page)
492 return atomic_read(&page->_count);
495 /* 127: arbitrary random number, small enough to assemble well */
496 #define page_ref_zero_or_close_to_overflow(page) \
497 ((unsigned int) atomic_read(&page->_count) + 127u <= 127u)
499 static inline void get_page(struct page *page)
501 if (unlikely(PageTail(page)))
502 if (likely(__get_page_tail(page)))
505 * Getting a normal page or the head of a compound page
506 * requires to already have an elevated page->_count.
508 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page), page);
509 atomic_inc(&page->_count);
512 static inline __must_check bool try_get_page(struct page *page)
514 if (unlikely(PageTail(page)))
515 if (likely(__get_page_tail(page)))
518 if (WARN_ON_ONCE(atomic_read(&page->_count) <= 0))
520 atomic_inc(&page->_count);
524 static inline struct page *virt_to_head_page(const void *x)
526 struct page *page = virt_to_page(x);
528 return compound_head(page);
532 * Setup the page count before being freed into the page allocator for
533 * the first time (boot or memory hotplug)
535 static inline void init_page_count(struct page *page)
537 atomic_set(&page->_count, 1);
540 void put_page(struct page *page);
541 void put_pages_list(struct list_head *pages);
543 void split_page(struct page *page, unsigned int order);
544 int split_free_page(struct page *page);
547 * Compound pages have a destructor function. Provide a
548 * prototype for that function and accessor functions.
549 * These are _only_ valid on the head of a compound page.
551 typedef void compound_page_dtor(struct page *);
553 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
554 enum compound_dtor_id {
557 #ifdef CONFIG_HUGETLB_PAGE
562 extern compound_page_dtor * const compound_page_dtors[];
564 static inline void set_compound_page_dtor(struct page *page,
565 enum compound_dtor_id compound_dtor)
567 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
568 page[1].compound_dtor = compound_dtor;
571 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
573 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
574 return compound_page_dtors[page[1].compound_dtor];
577 static inline unsigned int compound_order(struct page *page)
581 return page[1].compound_order;
584 static inline void set_compound_order(struct page *page, unsigned int order)
586 page[1].compound_order = order;
591 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
592 * servicing faults for write access. In the normal case, do always want
593 * pte_mkwrite. But get_user_pages can cause write faults for mappings
594 * that do not have writing enabled, when used by access_process_vm.
596 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
598 if (likely(vma->vm_flags & VM_WRITE))
599 pte = pte_mkwrite(pte);
603 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
604 struct page *page, pte_t *pte, bool write, bool anon);
608 * Multiple processes may "see" the same page. E.g. for untouched
609 * mappings of /dev/null, all processes see the same page full of
610 * zeroes, and text pages of executables and shared libraries have
611 * only one copy in memory, at most, normally.
613 * For the non-reserved pages, page_count(page) denotes a reference count.
614 * page_count() == 0 means the page is free. page->lru is then used for
615 * freelist management in the buddy allocator.
616 * page_count() > 0 means the page has been allocated.
618 * Pages are allocated by the slab allocator in order to provide memory
619 * to kmalloc and kmem_cache_alloc. In this case, the management of the
620 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
621 * unless a particular usage is carefully commented. (the responsibility of
622 * freeing the kmalloc memory is the caller's, of course).
624 * A page may be used by anyone else who does a __get_free_page().
625 * In this case, page_count still tracks the references, and should only
626 * be used through the normal accessor functions. The top bits of page->flags
627 * and page->virtual store page management information, but all other fields
628 * are unused and could be used privately, carefully. The management of this
629 * page is the responsibility of the one who allocated it, and those who have
630 * subsequently been given references to it.
632 * The other pages (we may call them "pagecache pages") are completely
633 * managed by the Linux memory manager: I/O, buffers, swapping etc.
634 * The following discussion applies only to them.
636 * A pagecache page contains an opaque `private' member, which belongs to the
637 * page's address_space. Usually, this is the address of a circular list of
638 * the page's disk buffers. PG_private must be set to tell the VM to call
639 * into the filesystem to release these pages.
641 * A page may belong to an inode's memory mapping. In this case, page->mapping
642 * is the pointer to the inode, and page->index is the file offset of the page,
643 * in units of PAGE_CACHE_SIZE.
645 * If pagecache pages are not associated with an inode, they are said to be
646 * anonymous pages. These may become associated with the swapcache, and in that
647 * case PG_swapcache is set, and page->private is an offset into the swapcache.
649 * In either case (swapcache or inode backed), the pagecache itself holds one
650 * reference to the page. Setting PG_private should also increment the
651 * refcount. The each user mapping also has a reference to the page.
653 * The pagecache pages are stored in a per-mapping radix tree, which is
654 * rooted at mapping->page_tree, and indexed by offset.
655 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
656 * lists, we instead now tag pages as dirty/writeback in the radix tree.
658 * All pagecache pages may be subject to I/O:
659 * - inode pages may need to be read from disk,
660 * - inode pages which have been modified and are MAP_SHARED may need
661 * to be written back to the inode on disk,
662 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
663 * modified may need to be swapped out to swap space and (later) to be read
668 * The zone field is never updated after free_area_init_core()
669 * sets it, so none of the operations on it need to be atomic.
672 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
673 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
674 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
675 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
676 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
679 * Define the bit shifts to access each section. For non-existent
680 * sections we define the shift as 0; that plus a 0 mask ensures
681 * the compiler will optimise away reference to them.
683 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
684 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
685 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
686 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
688 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
689 #ifdef NODE_NOT_IN_PAGE_FLAGS
690 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
691 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
692 SECTIONS_PGOFF : ZONES_PGOFF)
694 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
695 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
696 NODES_PGOFF : ZONES_PGOFF)
699 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
701 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
702 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
705 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
706 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
707 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
708 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
709 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
711 static inline enum zone_type page_zonenum(const struct page *page)
713 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
716 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
717 #define SECTION_IN_PAGE_FLAGS
721 * The identification function is mainly used by the buddy allocator for
722 * determining if two pages could be buddies. We are not really identifying
723 * the zone since we could be using the section number id if we do not have
724 * node id available in page flags.
725 * We only guarantee that it will return the same value for two combinable
728 static inline int page_zone_id(struct page *page)
730 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
733 static inline int zone_to_nid(struct zone *zone)
742 #ifdef NODE_NOT_IN_PAGE_FLAGS
743 extern int page_to_nid(const struct page *page);
745 static inline int page_to_nid(const struct page *page)
747 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
751 #ifdef CONFIG_NUMA_BALANCING
752 static inline int cpu_pid_to_cpupid(int cpu, int pid)
754 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
757 static inline int cpupid_to_pid(int cpupid)
759 return cpupid & LAST__PID_MASK;
762 static inline int cpupid_to_cpu(int cpupid)
764 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
767 static inline int cpupid_to_nid(int cpupid)
769 return cpu_to_node(cpupid_to_cpu(cpupid));
772 static inline bool cpupid_pid_unset(int cpupid)
774 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
777 static inline bool cpupid_cpu_unset(int cpupid)
779 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
782 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
784 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
787 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
788 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
789 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
791 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
794 static inline int page_cpupid_last(struct page *page)
796 return page->_last_cpupid;
798 static inline void page_cpupid_reset_last(struct page *page)
800 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
803 static inline int page_cpupid_last(struct page *page)
805 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
808 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
810 static inline void page_cpupid_reset_last(struct page *page)
812 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
814 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
815 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
817 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
818 #else /* !CONFIG_NUMA_BALANCING */
819 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
821 return page_to_nid(page); /* XXX */
824 static inline int page_cpupid_last(struct page *page)
826 return page_to_nid(page); /* XXX */
829 static inline int cpupid_to_nid(int cpupid)
834 static inline int cpupid_to_pid(int cpupid)
839 static inline int cpupid_to_cpu(int cpupid)
844 static inline int cpu_pid_to_cpupid(int nid, int pid)
849 static inline bool cpupid_pid_unset(int cpupid)
854 static inline void page_cpupid_reset_last(struct page *page)
858 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
862 #endif /* CONFIG_NUMA_BALANCING */
864 static inline struct zone *page_zone(const struct page *page)
866 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
869 #ifdef SECTION_IN_PAGE_FLAGS
870 static inline void set_page_section(struct page *page, unsigned long section)
872 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
873 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
876 static inline unsigned long page_to_section(const struct page *page)
878 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
882 static inline void set_page_zone(struct page *page, enum zone_type zone)
884 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
885 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
888 static inline void set_page_node(struct page *page, unsigned long node)
890 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
891 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
894 static inline void set_page_links(struct page *page, enum zone_type zone,
895 unsigned long node, unsigned long pfn)
897 set_page_zone(page, zone);
898 set_page_node(page, node);
899 #ifdef SECTION_IN_PAGE_FLAGS
900 set_page_section(page, pfn_to_section_nr(pfn));
905 static inline struct mem_cgroup *page_memcg(struct page *page)
907 return page->mem_cgroup;
910 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
912 page->mem_cgroup = memcg;
915 static inline struct mem_cgroup *page_memcg(struct page *page)
920 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
926 * Some inline functions in vmstat.h depend on page_zone()
928 #include <linux/vmstat.h>
930 static __always_inline void *lowmem_page_address(const struct page *page)
932 return __va(PFN_PHYS(page_to_pfn(page)));
935 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
936 #define HASHED_PAGE_VIRTUAL
939 #if defined(WANT_PAGE_VIRTUAL)
940 static inline void *page_address(const struct page *page)
942 return page->virtual;
944 static inline void set_page_address(struct page *page, void *address)
946 page->virtual = address;
948 #define page_address_init() do { } while(0)
951 #if defined(HASHED_PAGE_VIRTUAL)
952 void *page_address(const struct page *page);
953 void set_page_address(struct page *page, void *virtual);
954 void page_address_init(void);
957 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
958 #define page_address(page) lowmem_page_address(page)
959 #define set_page_address(page, address) do { } while(0)
960 #define page_address_init() do { } while(0)
963 extern void *page_rmapping(struct page *page);
964 extern struct anon_vma *page_anon_vma(struct page *page);
965 extern struct address_space *page_mapping(struct page *page);
967 extern struct address_space *__page_file_mapping(struct page *);
970 struct address_space *page_file_mapping(struct page *page)
972 if (unlikely(PageSwapCache(page)))
973 return __page_file_mapping(page);
975 return page->mapping;
979 * Return the pagecache index of the passed page. Regular pagecache pages
980 * use ->index whereas swapcache pages use ->private
982 static inline pgoff_t page_index(struct page *page)
984 if (unlikely(PageSwapCache(page)))
985 return page_private(page);
989 extern pgoff_t __page_file_index(struct page *page);
992 * Return the file index of the page. Regular pagecache pages use ->index
993 * whereas swapcache pages use swp_offset(->private)
995 static inline pgoff_t page_file_index(struct page *page)
997 if (unlikely(PageSwapCache(page)))
998 return __page_file_index(page);
1004 * Return true if this page is mapped into pagetables.
1006 static inline int page_mapped(struct page *page)
1008 return atomic_read(&(page)->_mapcount) >= 0;
1012 * Return true only if the page has been allocated with
1013 * ALLOC_NO_WATERMARKS and the low watermark was not
1014 * met implying that the system is under some pressure.
1016 static inline bool page_is_pfmemalloc(struct page *page)
1019 * Page index cannot be this large so this must be
1020 * a pfmemalloc page.
1022 return page->index == -1UL;
1026 * Only to be called by the page allocator on a freshly allocated
1029 static inline void set_page_pfmemalloc(struct page *page)
1034 static inline void clear_page_pfmemalloc(struct page *page)
1040 * Different kinds of faults, as returned by handle_mm_fault().
1041 * Used to decide whether a process gets delivered SIGBUS or
1042 * just gets major/minor fault counters bumped up.
1045 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1047 #define VM_FAULT_OOM 0x0001
1048 #define VM_FAULT_SIGBUS 0x0002
1049 #define VM_FAULT_MAJOR 0x0004
1050 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1051 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1052 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1053 #define VM_FAULT_SIGSEGV 0x0040
1055 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1056 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1057 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1058 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1060 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1062 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1063 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1066 /* Encode hstate index for a hwpoisoned large page */
1067 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1068 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1071 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1073 extern void pagefault_out_of_memory(void);
1075 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1078 * Flags passed to show_mem() and show_free_areas() to suppress output in
1081 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1083 extern void show_free_areas(unsigned int flags);
1084 extern bool skip_free_areas_node(unsigned int flags, int nid);
1086 int shmem_zero_setup(struct vm_area_struct *);
1088 bool shmem_mapping(struct address_space *mapping);
1090 static inline bool shmem_mapping(struct address_space *mapping)
1096 extern int can_do_mlock(void);
1097 extern int user_shm_lock(size_t, struct user_struct *);
1098 extern void user_shm_unlock(size_t, struct user_struct *);
1101 * Parameter block passed down to zap_pte_range in exceptional cases.
1103 struct zap_details {
1104 struct address_space *check_mapping; /* Check page->mapping if set */
1105 pgoff_t first_index; /* Lowest page->index to unmap */
1106 pgoff_t last_index; /* Highest page->index to unmap */
1109 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1111 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1114 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1115 unsigned long size);
1116 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1117 unsigned long size, struct zap_details *);
1118 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1119 unsigned long start, unsigned long end);
1122 * This has to be called after a get_task_mm()/mmget_not_zero()
1123 * followed by taking the mmap_sem for writing before modifying the
1124 * vmas or anything the coredump pretends not to change from under it.
1126 * It also has to be called when mmgrab() is used in the context of
1127 * the process, but then the mm_count refcount is transferred outside
1128 * the context of the process to run down_write() on that pinned mm.
1130 * NOTE: find_extend_vma() called from GUP context is the only place
1131 * that can modify the "mm" (notably the vm_start/end) under mmap_sem
1132 * for reading and outside the context of the process, so it is also
1133 * the only case that holds the mmap_sem for reading that must call
1134 * this function. Generally if the mmap_sem is hold for reading
1135 * there's no need of this check after get_task_mm()/mmget_not_zero().
1137 * This function can be obsoleted and the check can be removed, after
1138 * the coredump code will hold the mmap_sem for writing before
1139 * invoking the ->core_dump methods.
1141 static inline bool mmget_still_valid(struct mm_struct *mm)
1143 return likely(!mm->core_state);
1147 * mm_walk - callbacks for walk_page_range
1148 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1149 * this handler is required to be able to handle
1150 * pmd_trans_huge() pmds. They may simply choose to
1151 * split_huge_page() instead of handling it explicitly.
1152 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1153 * @pte_hole: if set, called for each hole at all levels
1154 * @hugetlb_entry: if set, called for each hugetlb entry
1155 * @test_walk: caller specific callback function to determine whether
1156 * we walk over the current vma or not. A positive returned
1157 * value means "do page table walk over the current vma,"
1158 * and a negative one means "abort current page table walk
1159 * right now." 0 means "skip the current vma."
1160 * @mm: mm_struct representing the target process of page table walk
1161 * @vma: vma currently walked (NULL if walking outside vmas)
1162 * @private: private data for callbacks' usage
1164 * (see the comment on walk_page_range() for more details)
1167 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1168 unsigned long next, struct mm_walk *walk);
1169 int (*pte_entry)(pte_t *pte, unsigned long addr,
1170 unsigned long next, struct mm_walk *walk);
1171 int (*pte_hole)(unsigned long addr, unsigned long next,
1172 struct mm_walk *walk);
1173 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1174 unsigned long addr, unsigned long next,
1175 struct mm_walk *walk);
1176 int (*test_walk)(unsigned long addr, unsigned long next,
1177 struct mm_walk *walk);
1178 struct mm_struct *mm;
1179 struct vm_area_struct *vma;
1183 int walk_page_range(unsigned long addr, unsigned long end,
1184 struct mm_walk *walk);
1185 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1186 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1187 unsigned long end, unsigned long floor, unsigned long ceiling);
1188 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1189 struct vm_area_struct *vma);
1190 void unmap_mapping_range(struct address_space *mapping,
1191 loff_t const holebegin, loff_t const holelen, int even_cows);
1192 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1193 unsigned long *pfn);
1194 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1195 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1196 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1197 void *buf, int len, int write);
1199 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1200 loff_t const holebegin, loff_t const holelen)
1202 unmap_mapping_range(mapping, holebegin, holelen, 0);
1205 extern void truncate_pagecache(struct inode *inode, loff_t new);
1206 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1207 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1208 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1209 int truncate_inode_page(struct address_space *mapping, struct page *page);
1210 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1211 int invalidate_inode_page(struct page *page);
1214 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1215 unsigned long address, unsigned int flags);
1216 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1217 unsigned long address, unsigned int fault_flags);
1219 static inline int handle_mm_fault(struct mm_struct *mm,
1220 struct vm_area_struct *vma, unsigned long address,
1223 /* should never happen if there's no MMU */
1225 return VM_FAULT_SIGBUS;
1227 static inline int fixup_user_fault(struct task_struct *tsk,
1228 struct mm_struct *mm, unsigned long address,
1229 unsigned int fault_flags)
1231 /* should never happen if there's no MMU */
1237 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1238 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1239 void *buf, int len, unsigned int gup_flags);
1241 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1242 unsigned long start, unsigned long nr_pages,
1243 unsigned int foll_flags, struct page **pages,
1244 struct vm_area_struct **vmas, int *nonblocking);
1245 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1246 unsigned long start, unsigned long nr_pages,
1247 unsigned int gup_flags, struct page **pages,
1248 struct vm_area_struct **vmas);
1249 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1250 unsigned long start, unsigned long nr_pages,
1251 unsigned int gup_flags, struct page **pages, int *locked);
1252 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1253 unsigned long start, unsigned long nr_pages,
1254 struct page **pages, unsigned int gup_flags);
1255 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1256 unsigned long start, unsigned long nr_pages,
1257 struct page **pages, unsigned int gup_flags);
1258 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1259 struct page **pages);
1261 /* Container for pinned pfns / pages */
1262 struct frame_vector {
1263 unsigned int nr_allocated; /* Number of frames we have space for */
1264 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1265 bool got_ref; /* Did we pin pages by getting page ref? */
1266 bool is_pfns; /* Does array contain pages or pfns? */
1267 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1268 * pfns_vector_pages() or pfns_vector_pfns()
1272 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1273 void frame_vector_destroy(struct frame_vector *vec);
1274 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1275 unsigned int gup_flags, struct frame_vector *vec);
1276 void put_vaddr_frames(struct frame_vector *vec);
1277 int frame_vector_to_pages(struct frame_vector *vec);
1278 void frame_vector_to_pfns(struct frame_vector *vec);
1280 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1282 return vec->nr_frames;
1285 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1288 int err = frame_vector_to_pages(vec);
1291 return ERR_PTR(err);
1293 return (struct page **)(vec->ptrs);
1296 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1299 frame_vector_to_pfns(vec);
1300 return (unsigned long *)(vec->ptrs);
1304 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1305 struct page **pages);
1306 int get_kernel_page(unsigned long start, int write, struct page **pages);
1307 struct page *get_dump_page(unsigned long addr);
1309 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1310 extern void do_invalidatepage(struct page *page, unsigned int offset,
1311 unsigned int length);
1313 int __set_page_dirty_nobuffers(struct page *page);
1314 int __set_page_dirty_no_writeback(struct page *page);
1315 int redirty_page_for_writepage(struct writeback_control *wbc,
1317 void account_page_dirtied(struct page *page, struct address_space *mapping,
1318 struct mem_cgroup *memcg);
1319 void account_page_cleaned(struct page *page, struct address_space *mapping,
1320 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1321 int set_page_dirty(struct page *page);
1322 int set_page_dirty_lock(struct page *page);
1323 void cancel_dirty_page(struct page *page);
1324 int clear_page_dirty_for_io(struct page *page);
1326 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1328 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1330 return !vma->vm_ops;
1333 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1335 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1336 unsigned long old_addr, struct vm_area_struct *new_vma,
1337 unsigned long new_addr, unsigned long len,
1338 bool need_rmap_locks);
1339 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1340 unsigned long end, pgprot_t newprot,
1341 int dirty_accountable, int prot_numa);
1342 extern int mprotect_fixup(struct vm_area_struct *vma,
1343 struct vm_area_struct **pprev, unsigned long start,
1344 unsigned long end, unsigned long newflags);
1347 * doesn't attempt to fault and will return short.
1349 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1350 struct page **pages);
1352 * per-process(per-mm_struct) statistics.
1354 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1356 long val = atomic_long_read(&mm->rss_stat.count[member]);
1358 #ifdef SPLIT_RSS_COUNTING
1360 * counter is updated in asynchronous manner and may go to minus.
1361 * But it's never be expected number for users.
1366 return (unsigned long)val;
1369 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1371 atomic_long_add(value, &mm->rss_stat.count[member]);
1374 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1376 atomic_long_inc(&mm->rss_stat.count[member]);
1379 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1381 atomic_long_dec(&mm->rss_stat.count[member]);
1384 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1386 return get_mm_counter(mm, MM_FILEPAGES) +
1387 get_mm_counter(mm, MM_ANONPAGES);
1390 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1392 return max(mm->hiwater_rss, get_mm_rss(mm));
1395 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1397 return max(mm->hiwater_vm, mm->total_vm);
1400 static inline void update_hiwater_rss(struct mm_struct *mm)
1402 unsigned long _rss = get_mm_rss(mm);
1404 if ((mm)->hiwater_rss < _rss)
1405 (mm)->hiwater_rss = _rss;
1408 static inline void update_hiwater_vm(struct mm_struct *mm)
1410 if (mm->hiwater_vm < mm->total_vm)
1411 mm->hiwater_vm = mm->total_vm;
1414 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1416 mm->hiwater_rss = get_mm_rss(mm);
1419 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1420 struct mm_struct *mm)
1422 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1424 if (*maxrss < hiwater_rss)
1425 *maxrss = hiwater_rss;
1428 #if defined(SPLIT_RSS_COUNTING)
1429 void sync_mm_rss(struct mm_struct *mm);
1431 static inline void sync_mm_rss(struct mm_struct *mm)
1436 int vma_wants_writenotify(struct vm_area_struct *vma);
1438 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1440 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1444 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1448 #ifdef __PAGETABLE_PUD_FOLDED
1449 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1450 unsigned long address)
1455 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1458 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1459 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1460 unsigned long address)
1465 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1467 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1472 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1473 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1476 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1478 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1480 atomic_long_set(&mm->nr_pmds, 0);
1483 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1485 return atomic_long_read(&mm->nr_pmds);
1488 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1490 atomic_long_inc(&mm->nr_pmds);
1493 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1495 atomic_long_dec(&mm->nr_pmds);
1499 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1500 pmd_t *pmd, unsigned long address);
1501 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1504 * The following ifdef needed to get the 4level-fixup.h header to work.
1505 * Remove it when 4level-fixup.h has been removed.
1507 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1508 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1510 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1511 NULL: pud_offset(pgd, address);
1514 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1516 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1517 NULL: pmd_offset(pud, address);
1519 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1521 #if USE_SPLIT_PTE_PTLOCKS
1522 #if ALLOC_SPLIT_PTLOCKS
1523 void __init ptlock_cache_init(void);
1524 extern bool ptlock_alloc(struct page *page);
1525 extern void ptlock_free(struct page *page);
1527 static inline spinlock_t *ptlock_ptr(struct page *page)
1531 #else /* ALLOC_SPLIT_PTLOCKS */
1532 static inline void ptlock_cache_init(void)
1536 static inline bool ptlock_alloc(struct page *page)
1541 static inline void ptlock_free(struct page *page)
1545 static inline spinlock_t *ptlock_ptr(struct page *page)
1549 #endif /* ALLOC_SPLIT_PTLOCKS */
1551 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1553 return ptlock_ptr(pmd_page(*pmd));
1556 static inline bool ptlock_init(struct page *page)
1559 * prep_new_page() initialize page->private (and therefore page->ptl)
1560 * with 0. Make sure nobody took it in use in between.
1562 * It can happen if arch try to use slab for page table allocation:
1563 * slab code uses page->slab_cache, which share storage with page->ptl.
1565 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1566 if (!ptlock_alloc(page))
1568 spin_lock_init(ptlock_ptr(page));
1572 /* Reset page->mapping so free_pages_check won't complain. */
1573 static inline void pte_lock_deinit(struct page *page)
1575 page->mapping = NULL;
1579 #else /* !USE_SPLIT_PTE_PTLOCKS */
1581 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1583 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1585 return &mm->page_table_lock;
1587 static inline void ptlock_cache_init(void) {}
1588 static inline bool ptlock_init(struct page *page) { return true; }
1589 static inline void pte_lock_deinit(struct page *page) {}
1590 #endif /* USE_SPLIT_PTE_PTLOCKS */
1592 static inline void pgtable_init(void)
1594 ptlock_cache_init();
1595 pgtable_cache_init();
1598 static inline bool pgtable_page_ctor(struct page *page)
1600 if (!ptlock_init(page))
1602 inc_zone_page_state(page, NR_PAGETABLE);
1606 static inline void pgtable_page_dtor(struct page *page)
1608 pte_lock_deinit(page);
1609 dec_zone_page_state(page, NR_PAGETABLE);
1612 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1614 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1615 pte_t *__pte = pte_offset_map(pmd, address); \
1621 #define pte_unmap_unlock(pte, ptl) do { \
1626 #define pte_alloc_map(mm, vma, pmd, address) \
1627 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1629 NULL: pte_offset_map(pmd, address))
1631 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1632 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1634 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1636 #define pte_alloc_kernel(pmd, address) \
1637 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1638 NULL: pte_offset_kernel(pmd, address))
1640 #if USE_SPLIT_PMD_PTLOCKS
1642 static struct page *pmd_to_page(pmd_t *pmd)
1644 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1645 return virt_to_page((void *)((unsigned long) pmd & mask));
1648 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1650 return ptlock_ptr(pmd_to_page(pmd));
1653 static inline bool pgtable_pmd_page_ctor(struct page *page)
1655 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1656 page->pmd_huge_pte = NULL;
1658 return ptlock_init(page);
1661 static inline void pgtable_pmd_page_dtor(struct page *page)
1663 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1664 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1669 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1673 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1675 return &mm->page_table_lock;
1678 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1679 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1681 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1685 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1687 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1692 extern void free_area_init(unsigned long * zones_size);
1693 extern void free_area_init_node(int nid, unsigned long * zones_size,
1694 unsigned long zone_start_pfn, unsigned long *zholes_size);
1695 extern void free_initmem(void);
1698 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1699 * into the buddy system. The freed pages will be poisoned with pattern
1700 * "poison" if it's within range [0, UCHAR_MAX].
1701 * Return pages freed into the buddy system.
1703 extern unsigned long free_reserved_area(void *start, void *end,
1704 int poison, char *s);
1706 #ifdef CONFIG_HIGHMEM
1708 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1709 * and totalram_pages.
1711 extern void free_highmem_page(struct page *page);
1714 extern void adjust_managed_page_count(struct page *page, long count);
1715 extern void mem_init_print_info(const char *str);
1717 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1719 /* Free the reserved page into the buddy system, so it gets managed. */
1720 static inline void __free_reserved_page(struct page *page)
1722 ClearPageReserved(page);
1723 init_page_count(page);
1727 static inline void free_reserved_page(struct page *page)
1729 __free_reserved_page(page);
1730 adjust_managed_page_count(page, 1);
1733 static inline void mark_page_reserved(struct page *page)
1735 SetPageReserved(page);
1736 adjust_managed_page_count(page, -1);
1740 * Default method to free all the __init memory into the buddy system.
1741 * The freed pages will be poisoned with pattern "poison" if it's within
1742 * range [0, UCHAR_MAX].
1743 * Return pages freed into the buddy system.
1745 static inline unsigned long free_initmem_default(int poison)
1747 extern char __init_begin[], __init_end[];
1749 return free_reserved_area(&__init_begin, &__init_end,
1750 poison, "unused kernel");
1753 static inline unsigned long get_num_physpages(void)
1756 unsigned long phys_pages = 0;
1758 for_each_online_node(nid)
1759 phys_pages += node_present_pages(nid);
1764 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1766 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1767 * zones, allocate the backing mem_map and account for memory holes in a more
1768 * architecture independent manner. This is a substitute for creating the
1769 * zone_sizes[] and zholes_size[] arrays and passing them to
1770 * free_area_init_node()
1772 * An architecture is expected to register range of page frames backed by
1773 * physical memory with memblock_add[_node]() before calling
1774 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1775 * usage, an architecture is expected to do something like
1777 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1779 * for_each_valid_physical_page_range()
1780 * memblock_add_node(base, size, nid)
1781 * free_area_init_nodes(max_zone_pfns);
1783 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1784 * registered physical page range. Similarly
1785 * sparse_memory_present_with_active_regions() calls memory_present() for
1786 * each range when SPARSEMEM is enabled.
1788 * See mm/page_alloc.c for more information on each function exposed by
1789 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1791 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1792 unsigned long node_map_pfn_alignment(void);
1793 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1794 unsigned long end_pfn);
1795 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1796 unsigned long end_pfn);
1797 extern void get_pfn_range_for_nid(unsigned int nid,
1798 unsigned long *start_pfn, unsigned long *end_pfn);
1799 extern unsigned long find_min_pfn_with_active_regions(void);
1800 extern void free_bootmem_with_active_regions(int nid,
1801 unsigned long max_low_pfn);
1802 extern void sparse_memory_present_with_active_regions(int nid);
1804 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1806 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1807 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1808 static inline int __early_pfn_to_nid(unsigned long pfn,
1809 struct mminit_pfnnid_cache *state)
1814 /* please see mm/page_alloc.c */
1815 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1816 /* there is a per-arch backend function. */
1817 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1818 struct mminit_pfnnid_cache *state);
1821 extern void set_dma_reserve(unsigned long new_dma_reserve);
1822 extern void memmap_init_zone(unsigned long, int, unsigned long,
1823 unsigned long, enum memmap_context);
1824 extern void setup_per_zone_wmarks(void);
1825 extern int __meminit init_per_zone_wmark_min(void);
1826 extern void mem_init(void);
1827 extern void __init mmap_init(void);
1828 extern void show_mem(unsigned int flags);
1829 extern long si_mem_available(void);
1830 extern void si_meminfo(struct sysinfo * val);
1831 extern void si_meminfo_node(struct sysinfo *val, int nid);
1833 extern __printf(3, 4)
1834 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1835 const char *fmt, ...);
1837 extern void setup_per_cpu_pageset(void);
1839 extern void zone_pcp_update(struct zone *zone);
1840 extern void zone_pcp_reset(struct zone *zone);
1843 extern int min_free_kbytes;
1846 extern atomic_long_t mmap_pages_allocated;
1847 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1849 /* interval_tree.c */
1850 void vma_interval_tree_insert(struct vm_area_struct *node,
1851 struct rb_root *root);
1852 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1853 struct vm_area_struct *prev,
1854 struct rb_root *root);
1855 void vma_interval_tree_remove(struct vm_area_struct *node,
1856 struct rb_root *root);
1857 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1858 unsigned long start, unsigned long last);
1859 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1860 unsigned long start, unsigned long last);
1862 #define vma_interval_tree_foreach(vma, root, start, last) \
1863 for (vma = vma_interval_tree_iter_first(root, start, last); \
1864 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1866 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1867 struct rb_root *root);
1868 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1869 struct rb_root *root);
1870 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1871 struct rb_root *root, unsigned long start, unsigned long last);
1872 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1873 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1874 #ifdef CONFIG_DEBUG_VM_RB
1875 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1878 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1879 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1880 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1883 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1884 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1885 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1886 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1887 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1888 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1889 struct mempolicy *, struct vm_userfaultfd_ctx);
1890 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1891 extern int split_vma(struct mm_struct *,
1892 struct vm_area_struct *, unsigned long addr, int new_below);
1893 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1894 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1895 struct rb_node **, struct rb_node *);
1896 extern void unlink_file_vma(struct vm_area_struct *);
1897 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1898 unsigned long addr, unsigned long len, pgoff_t pgoff,
1899 bool *need_rmap_locks);
1900 extern void exit_mmap(struct mm_struct *);
1902 static inline int check_data_rlimit(unsigned long rlim,
1904 unsigned long start,
1905 unsigned long end_data,
1906 unsigned long start_data)
1908 if (rlim < RLIM_INFINITY) {
1909 if (((new - start) + (end_data - start_data)) > rlim)
1916 extern int mm_take_all_locks(struct mm_struct *mm);
1917 extern void mm_drop_all_locks(struct mm_struct *mm);
1919 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1920 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1921 extern struct file *get_task_exe_file(struct task_struct *task);
1923 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1924 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1925 unsigned long addr, unsigned long len,
1926 unsigned long flags,
1927 const struct vm_special_mapping *spec);
1928 /* This is an obsolete alternative to _install_special_mapping. */
1929 extern int install_special_mapping(struct mm_struct *mm,
1930 unsigned long addr, unsigned long len,
1931 unsigned long flags, struct page **pages);
1933 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1935 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1936 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1937 extern unsigned long do_mmap(struct file *file, unsigned long addr,
1938 unsigned long len, unsigned long prot, unsigned long flags,
1939 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
1940 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1942 static inline unsigned long
1943 do_mmap_pgoff(struct file *file, unsigned long addr,
1944 unsigned long len, unsigned long prot, unsigned long flags,
1945 unsigned long pgoff, unsigned long *populate)
1947 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
1951 extern int __mm_populate(unsigned long addr, unsigned long len,
1953 static inline void mm_populate(unsigned long addr, unsigned long len)
1956 (void) __mm_populate(addr, len, 1);
1959 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1962 /* These take the mm semaphore themselves */
1963 extern unsigned long vm_brk(unsigned long, unsigned long);
1964 extern int vm_munmap(unsigned long, size_t);
1965 extern unsigned long vm_mmap(struct file *, unsigned long,
1966 unsigned long, unsigned long,
1967 unsigned long, unsigned long);
1969 struct vm_unmapped_area_info {
1970 #define VM_UNMAPPED_AREA_TOPDOWN 1
1971 unsigned long flags;
1972 unsigned long length;
1973 unsigned long low_limit;
1974 unsigned long high_limit;
1975 unsigned long align_mask;
1976 unsigned long align_offset;
1979 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1980 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1983 * Search for an unmapped address range.
1985 * We are looking for a range that:
1986 * - does not intersect with any VMA;
1987 * - is contained within the [low_limit, high_limit) interval;
1988 * - is at least the desired size.
1989 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1991 static inline unsigned long
1992 vm_unmapped_area(struct vm_unmapped_area_info *info)
1994 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
1995 return unmapped_area_topdown(info);
1997 return unmapped_area(info);
2001 extern void truncate_inode_pages(struct address_space *, loff_t);
2002 extern void truncate_inode_pages_range(struct address_space *,
2003 loff_t lstart, loff_t lend);
2004 extern void truncate_inode_pages_final(struct address_space *);
2006 /* generic vm_area_ops exported for stackable file systems */
2007 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2008 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2009 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2011 /* mm/page-writeback.c */
2012 int write_one_page(struct page *page, int wait);
2013 void task_dirty_inc(struct task_struct *tsk);
2016 #define VM_MAX_READAHEAD 128 /* kbytes */
2017 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2019 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2020 pgoff_t offset, unsigned long nr_to_read);
2022 void page_cache_sync_readahead(struct address_space *mapping,
2023 struct file_ra_state *ra,
2026 unsigned long size);
2028 void page_cache_async_readahead(struct address_space *mapping,
2029 struct file_ra_state *ra,
2033 unsigned long size);
2035 extern unsigned long stack_guard_gap;
2036 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2037 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2039 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2040 extern int expand_downwards(struct vm_area_struct *vma,
2041 unsigned long address);
2043 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2045 #define expand_upwards(vma, address) (0)
2048 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2049 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2050 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2051 struct vm_area_struct **pprev);
2053 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2054 NULL if none. Assume start_addr < end_addr. */
2055 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2057 struct vm_area_struct * vma = find_vma(mm,start_addr);
2059 if (vma && end_addr <= vma->vm_start)
2064 static inline unsigned long vm_start_gap(struct vm_area_struct *vma)
2066 unsigned long vm_start = vma->vm_start;
2068 if (vma->vm_flags & VM_GROWSDOWN) {
2069 vm_start -= stack_guard_gap;
2070 if (vm_start > vma->vm_start)
2076 static inline unsigned long vm_end_gap(struct vm_area_struct *vma)
2078 unsigned long vm_end = vma->vm_end;
2080 if (vma->vm_flags & VM_GROWSUP) {
2081 vm_end += stack_guard_gap;
2082 if (vm_end < vma->vm_end)
2083 vm_end = -PAGE_SIZE;
2088 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2090 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2093 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2094 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2095 unsigned long vm_start, unsigned long vm_end)
2097 struct vm_area_struct *vma = find_vma(mm, vm_start);
2099 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2105 static inline bool range_in_vma(struct vm_area_struct *vma,
2106 unsigned long start, unsigned long end)
2108 return (vma && vma->vm_start <= start && end <= vma->vm_end);
2112 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2113 void vma_set_page_prot(struct vm_area_struct *vma);
2115 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2119 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2121 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2125 #ifdef CONFIG_NUMA_BALANCING
2126 unsigned long change_prot_numa(struct vm_area_struct *vma,
2127 unsigned long start, unsigned long end);
2130 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2131 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2132 unsigned long pfn, unsigned long size, pgprot_t);
2133 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2134 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2136 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2137 unsigned long pfn, pgprot_t pgprot);
2138 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2140 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2143 struct page *follow_page_mask(struct vm_area_struct *vma,
2144 unsigned long address, unsigned int foll_flags,
2145 unsigned int *page_mask);
2147 static inline struct page *follow_page(struct vm_area_struct *vma,
2148 unsigned long address, unsigned int foll_flags)
2150 unsigned int unused_page_mask;
2151 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2154 #define FOLL_WRITE 0x01 /* check pte is writable */
2155 #define FOLL_TOUCH 0x02 /* mark page accessed */
2156 #define FOLL_GET 0x04 /* do get_page on page */
2157 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2158 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2159 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2160 * and return without waiting upon it */
2161 #define FOLL_POPULATE 0x40 /* fault in page */
2162 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2163 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2164 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2165 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2166 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2167 #define FOLL_MLOCK 0x1000 /* lock present pages */
2168 #define FOLL_COW 0x4000 /* internal GUP flag */
2169 #define FOLL_ANON 0x8000 /* don't do file mappings */
2171 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2173 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2174 unsigned long size, pte_fn_t fn, void *data);
2176 #ifdef CONFIG_PROC_FS
2177 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2179 static inline void vm_stat_account(struct mm_struct *mm,
2180 unsigned long flags, struct file *file, long pages)
2182 mm->total_vm += pages;
2184 #endif /* CONFIG_PROC_FS */
2186 #ifdef CONFIG_DEBUG_PAGEALLOC
2187 extern bool _debug_pagealloc_enabled;
2188 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2190 static inline bool debug_pagealloc_enabled(void)
2192 return _debug_pagealloc_enabled;
2196 kernel_map_pages(struct page *page, int numpages, int enable)
2198 if (!debug_pagealloc_enabled())
2201 __kernel_map_pages(page, numpages, enable);
2203 #ifdef CONFIG_HIBERNATION
2204 extern bool kernel_page_present(struct page *page);
2205 #endif /* CONFIG_HIBERNATION */
2208 kernel_map_pages(struct page *page, int numpages, int enable) {}
2209 #ifdef CONFIG_HIBERNATION
2210 static inline bool kernel_page_present(struct page *page) { return true; }
2211 #endif /* CONFIG_HIBERNATION */
2214 #ifdef __HAVE_ARCH_GATE_AREA
2215 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2216 extern int in_gate_area_no_mm(unsigned long addr);
2217 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2219 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2223 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2224 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2228 #endif /* __HAVE_ARCH_GATE_AREA */
2230 #ifdef CONFIG_SYSCTL
2231 extern int sysctl_drop_caches;
2232 int drop_caches_sysctl_handler(struct ctl_table *, int,
2233 void __user *, size_t *, loff_t *);
2236 void drop_slab(void);
2237 void drop_slab_node(int nid);
2240 #define randomize_va_space 0
2242 extern int randomize_va_space;
2245 const char * arch_vma_name(struct vm_area_struct *vma);
2246 void print_vma_addr(char *prefix, unsigned long rip);
2248 void sparse_mem_maps_populate_node(struct page **map_map,
2249 unsigned long pnum_begin,
2250 unsigned long pnum_end,
2251 unsigned long map_count,
2254 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2255 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2256 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2257 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2258 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2259 void *vmemmap_alloc_block(unsigned long size, int node);
2260 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2261 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2262 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2264 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2265 void vmemmap_populate_print_last(void);
2266 #ifdef CONFIG_MEMORY_HOTPLUG
2267 void vmemmap_free(unsigned long start, unsigned long end);
2269 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2270 unsigned long size);
2273 MF_COUNT_INCREASED = 1 << 0,
2274 MF_ACTION_REQUIRED = 1 << 1,
2275 MF_MUST_KILL = 1 << 2,
2276 MF_SOFT_OFFLINE = 1 << 3,
2278 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2279 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2280 extern int unpoison_memory(unsigned long pfn);
2281 extern int get_hwpoison_page(struct page *page);
2282 extern void put_hwpoison_page(struct page *page);
2283 extern int sysctl_memory_failure_early_kill;
2284 extern int sysctl_memory_failure_recovery;
2285 extern void shake_page(struct page *p, int access);
2286 extern atomic_long_t num_poisoned_pages;
2287 extern int soft_offline_page(struct page *page, int flags);
2291 * Error handlers for various types of pages.
2294 MF_IGNORED, /* Error: cannot be handled */
2295 MF_FAILED, /* Error: handling failed */
2296 MF_DELAYED, /* Will be handled later */
2297 MF_RECOVERED, /* Successfully recovered */
2300 enum mf_action_page_type {
2302 MF_MSG_KERNEL_HIGH_ORDER,
2304 MF_MSG_DIFFERENT_COMPOUND,
2305 MF_MSG_POISONED_HUGE,
2308 MF_MSG_UNMAP_FAILED,
2309 MF_MSG_DIRTY_SWAPCACHE,
2310 MF_MSG_CLEAN_SWAPCACHE,
2311 MF_MSG_DIRTY_MLOCKED_LRU,
2312 MF_MSG_CLEAN_MLOCKED_LRU,
2313 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2314 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2317 MF_MSG_TRUNCATED_LRU,
2323 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2324 extern void clear_huge_page(struct page *page,
2326 unsigned int pages_per_huge_page);
2327 extern void copy_user_huge_page(struct page *dst, struct page *src,
2328 unsigned long addr, struct vm_area_struct *vma,
2329 unsigned int pages_per_huge_page);
2330 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2332 extern struct page_ext_operations debug_guardpage_ops;
2333 extern struct page_ext_operations page_poisoning_ops;
2335 #ifdef CONFIG_DEBUG_PAGEALLOC
2336 extern unsigned int _debug_guardpage_minorder;
2337 extern bool _debug_guardpage_enabled;
2339 static inline unsigned int debug_guardpage_minorder(void)
2341 return _debug_guardpage_minorder;
2344 static inline bool debug_guardpage_enabled(void)
2346 return _debug_guardpage_enabled;
2349 static inline bool page_is_guard(struct page *page)
2351 struct page_ext *page_ext;
2353 if (!debug_guardpage_enabled())
2356 page_ext = lookup_page_ext(page);
2357 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2360 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2361 static inline bool debug_guardpage_enabled(void) { return false; }
2362 static inline bool page_is_guard(struct page *page) { return false; }
2363 #endif /* CONFIG_DEBUG_PAGEALLOC */
2365 #if MAX_NUMNODES > 1
2366 void __init setup_nr_node_ids(void);
2368 static inline void setup_nr_node_ids(void) {}
2371 #endif /* __KERNEL__ */
2372 #endif /* _LINUX_MM_H */