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/percpu-refcount.h>
20 #include <linux/bit_spinlock.h>
21 #include <linux/shrinker.h>
22 #include <linux/resource.h>
23 #include <linux/page_ext.h>
24 #include <linux/err.h>
28 struct anon_vma_chain;
31 struct writeback_control;
34 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
35 extern unsigned long max_mapnr;
37 static inline void set_max_mapnr(unsigned long limit)
42 static inline void set_max_mapnr(unsigned long limit) { }
45 extern unsigned long totalram_pages;
46 extern void * high_memory;
47 extern int page_cluster;
50 extern int sysctl_legacy_va_layout;
52 #define sysctl_legacy_va_layout 0
55 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
56 extern const int mmap_rnd_bits_min;
57 extern const int mmap_rnd_bits_max;
58 extern int mmap_rnd_bits __read_mostly;
60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
61 extern const int mmap_rnd_compat_bits_min;
62 extern const int mmap_rnd_compat_bits_max;
63 extern int mmap_rnd_compat_bits __read_mostly;
67 #include <asm/pgtable.h>
68 #include <asm/processor.h>
71 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
75 * To prevent common memory management code establishing
76 * a zero page mapping on a read fault.
77 * This macro should be defined within <asm/pgtable.h>.
78 * s390 does this to prevent multiplexing of hardware bits
79 * related to the physical page in case of virtualization.
81 #ifndef mm_forbids_zeropage
82 #define mm_forbids_zeropage(X) (0)
86 * Default maximum number of active map areas, this limits the number of vmas
87 * per mm struct. Users can overwrite this number by sysctl but there is a
90 * When a program's coredump is generated as ELF format, a section is created
91 * per a vma. In ELF, the number of sections is represented in unsigned short.
92 * This means the number of sections should be smaller than 65535 at coredump.
93 * Because the kernel adds some informative sections to a image of program at
94 * generating coredump, we need some margin. The number of extra sections is
95 * 1-3 now and depends on arch. We use "5" as safe margin, here.
97 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
98 * not a hard limit any more. Although some userspace tools can be surprised by
101 #define MAPCOUNT_ELF_CORE_MARGIN (5)
102 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
104 extern int sysctl_max_map_count;
106 extern unsigned long sysctl_user_reserve_kbytes;
107 extern unsigned long sysctl_admin_reserve_kbytes;
109 extern int sysctl_overcommit_memory;
110 extern int sysctl_overcommit_ratio;
111 extern unsigned long sysctl_overcommit_kbytes;
113 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
115 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
118 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
120 /* to align the pointer to the (next) page boundary */
121 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
123 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
124 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
127 * Linux kernel virtual memory manager primitives.
128 * The idea being to have a "virtual" mm in the same way
129 * we have a virtual fs - giving a cleaner interface to the
130 * mm details, and allowing different kinds of memory mappings
131 * (from shared memory to executable loading to arbitrary
135 extern struct kmem_cache *vm_area_cachep;
138 extern struct rb_root nommu_region_tree;
139 extern struct rw_semaphore nommu_region_sem;
141 extern unsigned int kobjsize(const void *objp);
145 * vm_flags in vm_area_struct, see mm_types.h.
146 * When changing, update also include/trace/events/mmflags.h
148 #define VM_NONE 0x00000000
150 #define VM_READ 0x00000001 /* currently active flags */
151 #define VM_WRITE 0x00000002
152 #define VM_EXEC 0x00000004
153 #define VM_SHARED 0x00000008
155 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
156 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
157 #define VM_MAYWRITE 0x00000020
158 #define VM_MAYEXEC 0x00000040
159 #define VM_MAYSHARE 0x00000080
161 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
162 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
163 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
164 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
165 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
167 #define VM_LOCKED 0x00002000
168 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
170 /* Used by sys_madvise() */
171 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
172 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
174 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
175 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
176 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
177 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
178 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
179 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
180 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
181 #define VM_ARCH_2 0x02000000
182 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
184 #ifdef CONFIG_MEM_SOFT_DIRTY
185 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
187 # define VM_SOFTDIRTY 0
190 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
191 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
192 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
193 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
195 #if defined(CONFIG_X86)
196 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
197 #elif defined(CONFIG_PPC)
198 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
199 #elif defined(CONFIG_PARISC)
200 # define VM_GROWSUP VM_ARCH_1
201 #elif defined(CONFIG_METAG)
202 # define VM_GROWSUP VM_ARCH_1
203 #elif defined(CONFIG_IA64)
204 # define VM_GROWSUP VM_ARCH_1
205 #elif !defined(CONFIG_MMU)
206 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
209 #if defined(CONFIG_X86)
210 /* MPX specific bounds table or bounds directory */
211 # define VM_MPX VM_ARCH_2
215 # define VM_GROWSUP VM_NONE
218 /* Bits set in the VMA until the stack is in its final location */
219 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
221 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
222 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
225 #ifdef CONFIG_STACK_GROWSUP
226 #define VM_STACK VM_GROWSUP
228 #define VM_STACK VM_GROWSDOWN
231 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
234 * Special vmas that are non-mergable, non-mlock()able.
235 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
237 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
239 /* This mask defines which mm->def_flags a process can inherit its parent */
240 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
242 /* This mask is used to clear all the VMA flags used by mlock */
243 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
246 * mapping from the currently active vm_flags protection bits (the
247 * low four bits) to a page protection mask..
249 extern pgprot_t protection_map[16];
251 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
252 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
253 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
254 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
255 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
256 #define FAULT_FLAG_TRIED 0x20 /* Second try */
257 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
260 * vm_fault is filled by the the pagefault handler and passed to the vma's
261 * ->fault function. The vma's ->fault is responsible for returning a bitmask
262 * of VM_FAULT_xxx flags that give details about how the fault was handled.
264 * MM layer fills up gfp_mask for page allocations but fault handler might
265 * alter it if its implementation requires a different allocation context.
267 * pgoff should be used in favour of virtual_address, if possible.
270 unsigned int flags; /* FAULT_FLAG_xxx flags */
271 gfp_t gfp_mask; /* gfp mask to be used for allocations */
272 pgoff_t pgoff; /* Logical page offset based on vma */
273 void __user *virtual_address; /* Faulting virtual address */
275 struct page *cow_page; /* Handler may choose to COW */
276 struct page *page; /* ->fault handlers should return a
277 * page here, unless VM_FAULT_NOPAGE
278 * is set (which is also implied by
281 /* for ->map_pages() only */
282 pgoff_t max_pgoff; /* map pages for offset from pgoff till
283 * max_pgoff inclusive */
284 pte_t *pte; /* pte entry associated with ->pgoff */
288 * These are the virtual MM functions - opening of an area, closing and
289 * unmapping it (needed to keep files on disk up-to-date etc), pointer
290 * to the functions called when a no-page or a wp-page exception occurs.
292 struct vm_operations_struct {
293 void (*open)(struct vm_area_struct * area);
294 void (*close)(struct vm_area_struct * area);
295 int (*mremap)(struct vm_area_struct * area);
296 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
297 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
298 pmd_t *, unsigned int flags);
299 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
301 /* notification that a previously read-only page is about to become
302 * writable, if an error is returned it will cause a SIGBUS */
303 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
305 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
306 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
308 /* called by access_process_vm when get_user_pages() fails, typically
309 * for use by special VMAs that can switch between memory and hardware
311 int (*access)(struct vm_area_struct *vma, unsigned long addr,
312 void *buf, int len, int write);
314 /* Called by the /proc/PID/maps code to ask the vma whether it
315 * has a special name. Returning non-NULL will also cause this
316 * vma to be dumped unconditionally. */
317 const char *(*name)(struct vm_area_struct *vma);
321 * set_policy() op must add a reference to any non-NULL @new mempolicy
322 * to hold the policy upon return. Caller should pass NULL @new to
323 * remove a policy and fall back to surrounding context--i.e. do not
324 * install a MPOL_DEFAULT policy, nor the task or system default
327 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
330 * get_policy() op must add reference [mpol_get()] to any policy at
331 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
332 * in mm/mempolicy.c will do this automatically.
333 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
334 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
335 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
336 * must return NULL--i.e., do not "fallback" to task or system default
339 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
343 * Called by vm_normal_page() for special PTEs to find the
344 * page for @addr. This is useful if the default behavior
345 * (using pte_page()) would not find the correct page.
347 struct page *(*find_special_page)(struct vm_area_struct *vma,
354 #define page_private(page) ((page)->private)
355 #define set_page_private(page, v) ((page)->private = (v))
357 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
358 static inline int pmd_devmap(pmd_t pmd)
365 * FIXME: take this include out, include page-flags.h in
366 * files which need it (119 of them)
368 #include <linux/page-flags.h>
369 #include <linux/huge_mm.h>
372 * Methods to modify the page usage count.
374 * What counts for a page usage:
375 * - cache mapping (page->mapping)
376 * - private data (page->private)
377 * - page mapped in a task's page tables, each mapping
378 * is counted separately
380 * Also, many kernel routines increase the page count before a critical
381 * routine so they can be sure the page doesn't go away from under them.
385 * Drop a ref, return true if the refcount fell to zero (the page has no users)
387 static inline int put_page_testzero(struct page *page)
389 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
390 return atomic_dec_and_test(&page->_count);
394 * Try to grab a ref unless the page has a refcount of zero, return false if
396 * This can be called when MMU is off so it must not access
397 * any of the virtual mappings.
399 static inline int get_page_unless_zero(struct page *page)
401 return atomic_inc_not_zero(&page->_count);
404 extern int page_is_ram(unsigned long pfn);
412 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
415 /* Support for virtually mapped pages */
416 struct page *vmalloc_to_page(const void *addr);
417 unsigned long vmalloc_to_pfn(const void *addr);
420 * Determine if an address is within the vmalloc range
422 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
423 * is no special casing required.
425 static inline int is_vmalloc_addr(const void *x)
428 unsigned long addr = (unsigned long)x;
430 return addr >= VMALLOC_START && addr < VMALLOC_END;
436 extern int is_vmalloc_or_module_addr(const void *x);
438 static inline int is_vmalloc_or_module_addr(const void *x)
444 extern void kvfree(const void *addr);
446 static inline atomic_t *compound_mapcount_ptr(struct page *page)
448 return &page[1].compound_mapcount;
451 static inline int compound_mapcount(struct page *page)
453 if (!PageCompound(page))
455 page = compound_head(page);
456 return atomic_read(compound_mapcount_ptr(page)) + 1;
460 * The atomic page->_mapcount, starts from -1: so that transitions
461 * both from it and to it can be tracked, using atomic_inc_and_test
462 * and atomic_add_negative(-1).
464 static inline void page_mapcount_reset(struct page *page)
466 atomic_set(&(page)->_mapcount, -1);
469 int __page_mapcount(struct page *page);
471 static inline int page_mapcount(struct page *page)
473 VM_BUG_ON_PAGE(PageSlab(page), page);
475 if (unlikely(PageCompound(page)))
476 return __page_mapcount(page);
477 return atomic_read(&page->_mapcount) + 1;
480 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
481 int total_mapcount(struct page *page);
483 static inline int total_mapcount(struct page *page)
485 return page_mapcount(page);
489 static inline int page_count(struct page *page)
491 return atomic_read(&compound_head(page)->_count);
494 static inline struct page *virt_to_head_page(const void *x)
496 struct page *page = virt_to_page(x);
498 return compound_head(page);
502 * Setup the page count before being freed into the page allocator for
503 * the first time (boot or memory hotplug)
505 static inline void init_page_count(struct page *page)
507 atomic_set(&page->_count, 1);
510 void __put_page(struct page *page);
512 void put_pages_list(struct list_head *pages);
514 void split_page(struct page *page, unsigned int order);
515 int split_free_page(struct page *page);
518 * Compound pages have a destructor function. Provide a
519 * prototype for that function and accessor functions.
520 * These are _only_ valid on the head of a compound page.
522 typedef void compound_page_dtor(struct page *);
524 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
525 enum compound_dtor_id {
528 #ifdef CONFIG_HUGETLB_PAGE
531 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
536 extern compound_page_dtor * const compound_page_dtors[];
538 static inline void set_compound_page_dtor(struct page *page,
539 enum compound_dtor_id compound_dtor)
541 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
542 page[1].compound_dtor = compound_dtor;
545 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
547 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
548 return compound_page_dtors[page[1].compound_dtor];
551 static inline unsigned int compound_order(struct page *page)
555 return page[1].compound_order;
558 static inline void set_compound_order(struct page *page, unsigned int order)
560 page[1].compound_order = order;
563 void free_compound_page(struct page *page);
567 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
568 * servicing faults for write access. In the normal case, do always want
569 * pte_mkwrite. But get_user_pages can cause write faults for mappings
570 * that do not have writing enabled, when used by access_process_vm.
572 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
574 if (likely(vma->vm_flags & VM_WRITE))
575 pte = pte_mkwrite(pte);
579 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
580 struct page *page, pte_t *pte, bool write, bool anon);
584 * Multiple processes may "see" the same page. E.g. for untouched
585 * mappings of /dev/null, all processes see the same page full of
586 * zeroes, and text pages of executables and shared libraries have
587 * only one copy in memory, at most, normally.
589 * For the non-reserved pages, page_count(page) denotes a reference count.
590 * page_count() == 0 means the page is free. page->lru is then used for
591 * freelist management in the buddy allocator.
592 * page_count() > 0 means the page has been allocated.
594 * Pages are allocated by the slab allocator in order to provide memory
595 * to kmalloc and kmem_cache_alloc. In this case, the management of the
596 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
597 * unless a particular usage is carefully commented. (the responsibility of
598 * freeing the kmalloc memory is the caller's, of course).
600 * A page may be used by anyone else who does a __get_free_page().
601 * In this case, page_count still tracks the references, and should only
602 * be used through the normal accessor functions. The top bits of page->flags
603 * and page->virtual store page management information, but all other fields
604 * are unused and could be used privately, carefully. The management of this
605 * page is the responsibility of the one who allocated it, and those who have
606 * subsequently been given references to it.
608 * The other pages (we may call them "pagecache pages") are completely
609 * managed by the Linux memory manager: I/O, buffers, swapping etc.
610 * The following discussion applies only to them.
612 * A pagecache page contains an opaque `private' member, which belongs to the
613 * page's address_space. Usually, this is the address of a circular list of
614 * the page's disk buffers. PG_private must be set to tell the VM to call
615 * into the filesystem to release these pages.
617 * A page may belong to an inode's memory mapping. In this case, page->mapping
618 * is the pointer to the inode, and page->index is the file offset of the page,
619 * in units of PAGE_CACHE_SIZE.
621 * If pagecache pages are not associated with an inode, they are said to be
622 * anonymous pages. These may become associated with the swapcache, and in that
623 * case PG_swapcache is set, and page->private is an offset into the swapcache.
625 * In either case (swapcache or inode backed), the pagecache itself holds one
626 * reference to the page. Setting PG_private should also increment the
627 * refcount. The each user mapping also has a reference to the page.
629 * The pagecache pages are stored in a per-mapping radix tree, which is
630 * rooted at mapping->page_tree, and indexed by offset.
631 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
632 * lists, we instead now tag pages as dirty/writeback in the radix tree.
634 * All pagecache pages may be subject to I/O:
635 * - inode pages may need to be read from disk,
636 * - inode pages which have been modified and are MAP_SHARED may need
637 * to be written back to the inode on disk,
638 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
639 * modified may need to be swapped out to swap space and (later) to be read
644 * The zone field is never updated after free_area_init_core()
645 * sets it, so none of the operations on it need to be atomic.
648 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
649 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
650 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
651 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
652 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
655 * Define the bit shifts to access each section. For non-existent
656 * sections we define the shift as 0; that plus a 0 mask ensures
657 * the compiler will optimise away reference to them.
659 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
660 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
661 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
662 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
664 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
665 #ifdef NODE_NOT_IN_PAGE_FLAGS
666 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
667 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
668 SECTIONS_PGOFF : ZONES_PGOFF)
670 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
671 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
672 NODES_PGOFF : ZONES_PGOFF)
675 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
677 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
678 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
681 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
682 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
683 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
684 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
685 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
687 static inline enum zone_type page_zonenum(const struct page *page)
689 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
692 #ifdef CONFIG_ZONE_DEVICE
693 void get_zone_device_page(struct page *page);
694 void put_zone_device_page(struct page *page);
695 static inline bool is_zone_device_page(const struct page *page)
697 return page_zonenum(page) == ZONE_DEVICE;
700 static inline void get_zone_device_page(struct page *page)
703 static inline void put_zone_device_page(struct page *page)
706 static inline bool is_zone_device_page(const struct page *page)
712 static inline void get_page(struct page *page)
714 page = compound_head(page);
716 * Getting a normal page or the head of a compound page
717 * requires to already have an elevated page->_count.
719 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
720 atomic_inc(&page->_count);
722 if (unlikely(is_zone_device_page(page)))
723 get_zone_device_page(page);
726 static inline void put_page(struct page *page)
728 page = compound_head(page);
730 if (put_page_testzero(page))
733 if (unlikely(is_zone_device_page(page)))
734 put_zone_device_page(page);
737 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
738 #define SECTION_IN_PAGE_FLAGS
742 * The identification function is mainly used by the buddy allocator for
743 * determining if two pages could be buddies. We are not really identifying
744 * the zone since we could be using the section number id if we do not have
745 * node id available in page flags.
746 * We only guarantee that it will return the same value for two combinable
749 static inline int page_zone_id(struct page *page)
751 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
754 static inline int zone_to_nid(struct zone *zone)
763 #ifdef NODE_NOT_IN_PAGE_FLAGS
764 extern int page_to_nid(const struct page *page);
766 static inline int page_to_nid(const struct page *page)
768 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
772 #ifdef CONFIG_NUMA_BALANCING
773 static inline int cpu_pid_to_cpupid(int cpu, int pid)
775 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
778 static inline int cpupid_to_pid(int cpupid)
780 return cpupid & LAST__PID_MASK;
783 static inline int cpupid_to_cpu(int cpupid)
785 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
788 static inline int cpupid_to_nid(int cpupid)
790 return cpu_to_node(cpupid_to_cpu(cpupid));
793 static inline bool cpupid_pid_unset(int cpupid)
795 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
798 static inline bool cpupid_cpu_unset(int cpupid)
800 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
803 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
805 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
808 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
809 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
810 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
812 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
815 static inline int page_cpupid_last(struct page *page)
817 return page->_last_cpupid;
819 static inline void page_cpupid_reset_last(struct page *page)
821 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
824 static inline int page_cpupid_last(struct page *page)
826 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
829 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
831 static inline void page_cpupid_reset_last(struct page *page)
833 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
835 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
836 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
838 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
839 #else /* !CONFIG_NUMA_BALANCING */
840 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
842 return page_to_nid(page); /* XXX */
845 static inline int page_cpupid_last(struct page *page)
847 return page_to_nid(page); /* XXX */
850 static inline int cpupid_to_nid(int cpupid)
855 static inline int cpupid_to_pid(int cpupid)
860 static inline int cpupid_to_cpu(int cpupid)
865 static inline int cpu_pid_to_cpupid(int nid, int pid)
870 static inline bool cpupid_pid_unset(int cpupid)
875 static inline void page_cpupid_reset_last(struct page *page)
879 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
883 #endif /* CONFIG_NUMA_BALANCING */
885 static inline struct zone *page_zone(const struct page *page)
887 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
890 #ifdef SECTION_IN_PAGE_FLAGS
891 static inline void set_page_section(struct page *page, unsigned long section)
893 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
894 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
897 static inline unsigned long page_to_section(const struct page *page)
899 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
903 static inline void set_page_zone(struct page *page, enum zone_type zone)
905 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
906 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
909 static inline void set_page_node(struct page *page, unsigned long node)
911 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
912 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
915 static inline void set_page_links(struct page *page, enum zone_type zone,
916 unsigned long node, unsigned long pfn)
918 set_page_zone(page, zone);
919 set_page_node(page, node);
920 #ifdef SECTION_IN_PAGE_FLAGS
921 set_page_section(page, pfn_to_section_nr(pfn));
926 static inline struct mem_cgroup *page_memcg(struct page *page)
928 return page->mem_cgroup;
931 static inline struct mem_cgroup *page_memcg(struct page *page)
938 * Some inline functions in vmstat.h depend on page_zone()
940 #include <linux/vmstat.h>
942 static __always_inline void *lowmem_page_address(const struct page *page)
944 return __va(PFN_PHYS(page_to_pfn(page)));
947 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
948 #define HASHED_PAGE_VIRTUAL
951 #if defined(WANT_PAGE_VIRTUAL)
952 static inline void *page_address(const struct page *page)
954 return page->virtual;
956 static inline void set_page_address(struct page *page, void *address)
958 page->virtual = address;
960 #define page_address_init() do { } while(0)
963 #if defined(HASHED_PAGE_VIRTUAL)
964 void *page_address(const struct page *page);
965 void set_page_address(struct page *page, void *virtual);
966 void page_address_init(void);
969 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
970 #define page_address(page) lowmem_page_address(page)
971 #define set_page_address(page, address) do { } while(0)
972 #define page_address_init() do { } while(0)
975 extern void *page_rmapping(struct page *page);
976 extern struct anon_vma *page_anon_vma(struct page *page);
977 extern struct address_space *page_mapping(struct page *page);
979 extern struct address_space *__page_file_mapping(struct page *);
982 struct address_space *page_file_mapping(struct page *page)
984 if (unlikely(PageSwapCache(page)))
985 return __page_file_mapping(page);
987 return page->mapping;
991 * Return the pagecache index of the passed page. Regular pagecache pages
992 * use ->index whereas swapcache pages use ->private
994 static inline pgoff_t page_index(struct page *page)
996 if (unlikely(PageSwapCache(page)))
997 return page_private(page);
1001 extern pgoff_t __page_file_index(struct page *page);
1004 * Return the file index of the page. Regular pagecache pages use ->index
1005 * whereas swapcache pages use swp_offset(->private)
1007 static inline pgoff_t page_file_index(struct page *page)
1009 if (unlikely(PageSwapCache(page)))
1010 return __page_file_index(page);
1016 * Return true if this page is mapped into pagetables.
1017 * For compound page it returns true if any subpage of compound page is mapped.
1019 static inline bool page_mapped(struct page *page)
1022 if (likely(!PageCompound(page)))
1023 return atomic_read(&page->_mapcount) >= 0;
1024 page = compound_head(page);
1025 if (atomic_read(compound_mapcount_ptr(page)) >= 0)
1027 for (i = 0; i < hpage_nr_pages(page); i++) {
1028 if (atomic_read(&page[i]._mapcount) >= 0)
1035 * Return true only if the page has been allocated with
1036 * ALLOC_NO_WATERMARKS and the low watermark was not
1037 * met implying that the system is under some pressure.
1039 static inline bool page_is_pfmemalloc(struct page *page)
1042 * Page index cannot be this large so this must be
1043 * a pfmemalloc page.
1045 return page->index == -1UL;
1049 * Only to be called by the page allocator on a freshly allocated
1052 static inline void set_page_pfmemalloc(struct page *page)
1057 static inline void clear_page_pfmemalloc(struct page *page)
1063 * Different kinds of faults, as returned by handle_mm_fault().
1064 * Used to decide whether a process gets delivered SIGBUS or
1065 * just gets major/minor fault counters bumped up.
1068 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1070 #define VM_FAULT_OOM 0x0001
1071 #define VM_FAULT_SIGBUS 0x0002
1072 #define VM_FAULT_MAJOR 0x0004
1073 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1074 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1075 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1076 #define VM_FAULT_SIGSEGV 0x0040
1078 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1079 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1080 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1081 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1083 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1085 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1086 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1089 /* Encode hstate index for a hwpoisoned large page */
1090 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1091 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1094 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1096 extern void pagefault_out_of_memory(void);
1098 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1101 * Flags passed to show_mem() and show_free_areas() to suppress output in
1104 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1106 extern void show_free_areas(unsigned int flags);
1107 extern bool skip_free_areas_node(unsigned int flags, int nid);
1109 int shmem_zero_setup(struct vm_area_struct *);
1111 bool shmem_mapping(struct address_space *mapping);
1113 static inline bool shmem_mapping(struct address_space *mapping)
1119 extern bool can_do_mlock(void);
1120 extern int user_shm_lock(size_t, struct user_struct *);
1121 extern void user_shm_unlock(size_t, struct user_struct *);
1124 * Parameter block passed down to zap_pte_range in exceptional cases.
1126 struct zap_details {
1127 struct address_space *check_mapping; /* Check page->mapping if set */
1128 pgoff_t first_index; /* Lowest page->index to unmap */
1129 pgoff_t last_index; /* Highest page->index to unmap */
1132 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1135 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1136 unsigned long size);
1137 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1138 unsigned long size, struct zap_details *);
1139 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1140 unsigned long start, unsigned long end);
1143 * mm_walk - callbacks for walk_page_range
1144 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1145 * this handler is required to be able to handle
1146 * pmd_trans_huge() pmds. They may simply choose to
1147 * split_huge_page() instead of handling it explicitly.
1148 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1149 * @pte_hole: if set, called for each hole at all levels
1150 * @hugetlb_entry: if set, called for each hugetlb entry
1151 * @test_walk: caller specific callback function to determine whether
1152 * we walk over the current vma or not. A positive returned
1153 * value means "do page table walk over the current vma,"
1154 * and a negative one means "abort current page table walk
1155 * right now." 0 means "skip the current vma."
1156 * @mm: mm_struct representing the target process of page table walk
1157 * @vma: vma currently walked (NULL if walking outside vmas)
1158 * @private: private data for callbacks' usage
1160 * (see the comment on walk_page_range() for more details)
1163 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1164 unsigned long next, struct mm_walk *walk);
1165 int (*pte_entry)(pte_t *pte, unsigned long addr,
1166 unsigned long next, struct mm_walk *walk);
1167 int (*pte_hole)(unsigned long addr, unsigned long next,
1168 struct mm_walk *walk);
1169 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1170 unsigned long addr, unsigned long next,
1171 struct mm_walk *walk);
1172 int (*test_walk)(unsigned long addr, unsigned long next,
1173 struct mm_walk *walk);
1174 struct mm_struct *mm;
1175 struct vm_area_struct *vma;
1179 int walk_page_range(unsigned long addr, unsigned long end,
1180 struct mm_walk *walk);
1181 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1182 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1183 unsigned long end, unsigned long floor, unsigned long ceiling);
1184 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1185 struct vm_area_struct *vma);
1186 void unmap_mapping_range(struct address_space *mapping,
1187 loff_t const holebegin, loff_t const holelen, int even_cows);
1188 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1189 unsigned long *pfn);
1190 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1191 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1192 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1193 void *buf, int len, int write);
1195 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1196 loff_t const holebegin, loff_t const holelen)
1198 unmap_mapping_range(mapping, holebegin, holelen, 0);
1201 extern void truncate_pagecache(struct inode *inode, loff_t new);
1202 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1203 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1204 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1205 int truncate_inode_page(struct address_space *mapping, struct page *page);
1206 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1207 int invalidate_inode_page(struct page *page);
1210 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1211 unsigned long address, unsigned int flags);
1212 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1213 unsigned long address, unsigned int fault_flags,
1216 static inline int handle_mm_fault(struct mm_struct *mm,
1217 struct vm_area_struct *vma, unsigned long address,
1220 /* should never happen if there's no MMU */
1222 return VM_FAULT_SIGBUS;
1224 static inline int fixup_user_fault(struct task_struct *tsk,
1225 struct mm_struct *mm, unsigned long address,
1226 unsigned int fault_flags, bool *unlocked)
1228 /* should never happen if there's no MMU */
1234 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1235 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1236 void *buf, int len, int write);
1238 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1239 unsigned long start, unsigned long nr_pages,
1240 unsigned int foll_flags, struct page **pages,
1241 struct vm_area_struct **vmas, int *nonblocking);
1242 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1243 unsigned long start, unsigned long nr_pages,
1244 int write, int force, struct page **pages,
1245 struct vm_area_struct **vmas);
1246 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1247 unsigned long start, unsigned long nr_pages,
1248 int write, int force, struct page **pages,
1250 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1251 unsigned long start, unsigned long nr_pages,
1252 int write, int force, struct page **pages,
1253 unsigned int gup_flags);
1254 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1255 unsigned long start, unsigned long nr_pages,
1256 int write, int force, struct page **pages);
1257 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1258 struct page **pages);
1260 /* Container for pinned pfns / pages */
1261 struct frame_vector {
1262 unsigned int nr_allocated; /* Number of frames we have space for */
1263 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1264 bool got_ref; /* Did we pin pages by getting page ref? */
1265 bool is_pfns; /* Does array contain pages or pfns? */
1266 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1267 * pfns_vector_pages() or pfns_vector_pfns()
1271 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1272 void frame_vector_destroy(struct frame_vector *vec);
1273 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1274 bool write, bool force, struct frame_vector *vec);
1275 void put_vaddr_frames(struct frame_vector *vec);
1276 int frame_vector_to_pages(struct frame_vector *vec);
1277 void frame_vector_to_pfns(struct frame_vector *vec);
1279 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1281 return vec->nr_frames;
1284 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1287 int err = frame_vector_to_pages(vec);
1290 return ERR_PTR(err);
1292 return (struct page **)(vec->ptrs);
1295 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1298 frame_vector_to_pfns(vec);
1299 return (unsigned long *)(vec->ptrs);
1303 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1304 struct page **pages);
1305 int get_kernel_page(unsigned long start, int write, struct page **pages);
1306 struct page *get_dump_page(unsigned long addr);
1308 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1309 extern void do_invalidatepage(struct page *page, unsigned int offset,
1310 unsigned int length);
1312 int __set_page_dirty_nobuffers(struct page *page);
1313 int __set_page_dirty_no_writeback(struct page *page);
1314 int redirty_page_for_writepage(struct writeback_control *wbc,
1316 void account_page_dirtied(struct page *page, struct address_space *mapping);
1317 void account_page_cleaned(struct page *page, struct address_space *mapping,
1318 struct bdi_writeback *wb);
1319 int set_page_dirty(struct page *page);
1320 int set_page_dirty_lock(struct page *page);
1321 void cancel_dirty_page(struct page *page);
1322 int clear_page_dirty_for_io(struct page *page);
1324 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1326 /* Is the vma a continuation of the stack vma above it? */
1327 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1329 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1332 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1334 return !vma->vm_ops;
1337 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1340 return (vma->vm_flags & VM_GROWSDOWN) &&
1341 (vma->vm_start == addr) &&
1342 !vma_growsdown(vma->vm_prev, addr);
1345 /* Is the vma a continuation of the stack vma below it? */
1346 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1348 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1351 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1354 return (vma->vm_flags & VM_GROWSUP) &&
1355 (vma->vm_end == addr) &&
1356 !vma_growsup(vma->vm_next, addr);
1359 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1361 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1362 unsigned long old_addr, struct vm_area_struct *new_vma,
1363 unsigned long new_addr, unsigned long len,
1364 bool need_rmap_locks);
1365 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1366 unsigned long end, pgprot_t newprot,
1367 int dirty_accountable, int prot_numa);
1368 extern int mprotect_fixup(struct vm_area_struct *vma,
1369 struct vm_area_struct **pprev, unsigned long start,
1370 unsigned long end, unsigned long newflags);
1373 * doesn't attempt to fault and will return short.
1375 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1376 struct page **pages);
1378 * per-process(per-mm_struct) statistics.
1380 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1382 long val = atomic_long_read(&mm->rss_stat.count[member]);
1384 #ifdef SPLIT_RSS_COUNTING
1386 * counter is updated in asynchronous manner and may go to minus.
1387 * But it's never be expected number for users.
1392 return (unsigned long)val;
1395 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1397 atomic_long_add(value, &mm->rss_stat.count[member]);
1400 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1402 atomic_long_inc(&mm->rss_stat.count[member]);
1405 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1407 atomic_long_dec(&mm->rss_stat.count[member]);
1410 /* Optimized variant when page is already known not to be PageAnon */
1411 static inline int mm_counter_file(struct page *page)
1413 if (PageSwapBacked(page))
1414 return MM_SHMEMPAGES;
1415 return MM_FILEPAGES;
1418 static inline int mm_counter(struct page *page)
1421 return MM_ANONPAGES;
1422 return mm_counter_file(page);
1425 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1427 return get_mm_counter(mm, MM_FILEPAGES) +
1428 get_mm_counter(mm, MM_ANONPAGES) +
1429 get_mm_counter(mm, MM_SHMEMPAGES);
1432 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1434 return max(mm->hiwater_rss, get_mm_rss(mm));
1437 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1439 return max(mm->hiwater_vm, mm->total_vm);
1442 static inline void update_hiwater_rss(struct mm_struct *mm)
1444 unsigned long _rss = get_mm_rss(mm);
1446 if ((mm)->hiwater_rss < _rss)
1447 (mm)->hiwater_rss = _rss;
1450 static inline void update_hiwater_vm(struct mm_struct *mm)
1452 if (mm->hiwater_vm < mm->total_vm)
1453 mm->hiwater_vm = mm->total_vm;
1456 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1458 mm->hiwater_rss = get_mm_rss(mm);
1461 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1462 struct mm_struct *mm)
1464 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1466 if (*maxrss < hiwater_rss)
1467 *maxrss = hiwater_rss;
1470 #if defined(SPLIT_RSS_COUNTING)
1471 void sync_mm_rss(struct mm_struct *mm);
1473 static inline void sync_mm_rss(struct mm_struct *mm)
1478 #ifndef __HAVE_ARCH_PTE_DEVMAP
1479 static inline int pte_devmap(pte_t pte)
1485 int vma_wants_writenotify(struct vm_area_struct *vma);
1487 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1489 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1493 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1497 #ifdef __PAGETABLE_PUD_FOLDED
1498 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1499 unsigned long address)
1504 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1507 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1508 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1509 unsigned long address)
1514 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1516 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1521 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1522 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1525 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1527 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1529 atomic_long_set(&mm->nr_pmds, 0);
1532 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1534 return atomic_long_read(&mm->nr_pmds);
1537 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1539 atomic_long_inc(&mm->nr_pmds);
1542 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1544 atomic_long_dec(&mm->nr_pmds);
1548 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1549 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1552 * The following ifdef needed to get the 4level-fixup.h header to work.
1553 * Remove it when 4level-fixup.h has been removed.
1555 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1556 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1558 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1559 NULL: pud_offset(pgd, address);
1562 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1564 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1565 NULL: pmd_offset(pud, address);
1567 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1569 #if USE_SPLIT_PTE_PTLOCKS
1570 #if ALLOC_SPLIT_PTLOCKS
1571 void __init ptlock_cache_init(void);
1572 extern bool ptlock_alloc(struct page *page);
1573 extern void ptlock_free(struct page *page);
1575 static inline spinlock_t *ptlock_ptr(struct page *page)
1579 #else /* ALLOC_SPLIT_PTLOCKS */
1580 static inline void ptlock_cache_init(void)
1584 static inline bool ptlock_alloc(struct page *page)
1589 static inline void ptlock_free(struct page *page)
1593 static inline spinlock_t *ptlock_ptr(struct page *page)
1597 #endif /* ALLOC_SPLIT_PTLOCKS */
1599 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1601 return ptlock_ptr(pmd_page(*pmd));
1604 static inline bool ptlock_init(struct page *page)
1607 * prep_new_page() initialize page->private (and therefore page->ptl)
1608 * with 0. Make sure nobody took it in use in between.
1610 * It can happen if arch try to use slab for page table allocation:
1611 * slab code uses page->slab_cache, which share storage with page->ptl.
1613 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1614 if (!ptlock_alloc(page))
1616 spin_lock_init(ptlock_ptr(page));
1620 /* Reset page->mapping so free_pages_check won't complain. */
1621 static inline void pte_lock_deinit(struct page *page)
1623 page->mapping = NULL;
1627 #else /* !USE_SPLIT_PTE_PTLOCKS */
1629 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1631 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1633 return &mm->page_table_lock;
1635 static inline void ptlock_cache_init(void) {}
1636 static inline bool ptlock_init(struct page *page) { return true; }
1637 static inline void pte_lock_deinit(struct page *page) {}
1638 #endif /* USE_SPLIT_PTE_PTLOCKS */
1640 static inline void pgtable_init(void)
1642 ptlock_cache_init();
1643 pgtable_cache_init();
1646 static inline bool pgtable_page_ctor(struct page *page)
1648 if (!ptlock_init(page))
1650 inc_zone_page_state(page, NR_PAGETABLE);
1654 static inline void pgtable_page_dtor(struct page *page)
1656 pte_lock_deinit(page);
1657 dec_zone_page_state(page, NR_PAGETABLE);
1660 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1662 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1663 pte_t *__pte = pte_offset_map(pmd, address); \
1669 #define pte_unmap_unlock(pte, ptl) do { \
1674 #define pte_alloc(mm, pmd, address) \
1675 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1677 #define pte_alloc_map(mm, pmd, address) \
1678 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1680 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1681 (pte_alloc(mm, pmd, address) ? \
1682 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1684 #define pte_alloc_kernel(pmd, address) \
1685 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1686 NULL: pte_offset_kernel(pmd, address))
1688 #if USE_SPLIT_PMD_PTLOCKS
1690 static struct page *pmd_to_page(pmd_t *pmd)
1692 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1693 return virt_to_page((void *)((unsigned long) pmd & mask));
1696 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1698 return ptlock_ptr(pmd_to_page(pmd));
1701 static inline bool pgtable_pmd_page_ctor(struct page *page)
1703 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1704 page->pmd_huge_pte = NULL;
1706 return ptlock_init(page);
1709 static inline void pgtable_pmd_page_dtor(struct page *page)
1711 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1712 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1717 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1721 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1723 return &mm->page_table_lock;
1726 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1727 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1729 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1733 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1735 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1740 extern void free_area_init(unsigned long * zones_size);
1741 extern void free_area_init_node(int nid, unsigned long * zones_size,
1742 unsigned long zone_start_pfn, unsigned long *zholes_size);
1743 extern void free_initmem(void);
1746 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1747 * into the buddy system. The freed pages will be poisoned with pattern
1748 * "poison" if it's within range [0, UCHAR_MAX].
1749 * Return pages freed into the buddy system.
1751 extern unsigned long free_reserved_area(void *start, void *end,
1752 int poison, char *s);
1754 #ifdef CONFIG_HIGHMEM
1756 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1757 * and totalram_pages.
1759 extern void free_highmem_page(struct page *page);
1762 extern void adjust_managed_page_count(struct page *page, long count);
1763 extern void mem_init_print_info(const char *str);
1765 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1767 /* Free the reserved page into the buddy system, so it gets managed. */
1768 static inline void __free_reserved_page(struct page *page)
1770 ClearPageReserved(page);
1771 init_page_count(page);
1775 static inline void free_reserved_page(struct page *page)
1777 __free_reserved_page(page);
1778 adjust_managed_page_count(page, 1);
1781 static inline void mark_page_reserved(struct page *page)
1783 SetPageReserved(page);
1784 adjust_managed_page_count(page, -1);
1788 * Default method to free all the __init memory into the buddy system.
1789 * The freed pages will be poisoned with pattern "poison" if it's within
1790 * range [0, UCHAR_MAX].
1791 * Return pages freed into the buddy system.
1793 static inline unsigned long free_initmem_default(int poison)
1795 extern char __init_begin[], __init_end[];
1797 return free_reserved_area(&__init_begin, &__init_end,
1798 poison, "unused kernel");
1801 static inline unsigned long get_num_physpages(void)
1804 unsigned long phys_pages = 0;
1806 for_each_online_node(nid)
1807 phys_pages += node_present_pages(nid);
1812 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1814 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1815 * zones, allocate the backing mem_map and account for memory holes in a more
1816 * architecture independent manner. This is a substitute for creating the
1817 * zone_sizes[] and zholes_size[] arrays and passing them to
1818 * free_area_init_node()
1820 * An architecture is expected to register range of page frames backed by
1821 * physical memory with memblock_add[_node]() before calling
1822 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1823 * usage, an architecture is expected to do something like
1825 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1827 * for_each_valid_physical_page_range()
1828 * memblock_add_node(base, size, nid)
1829 * free_area_init_nodes(max_zone_pfns);
1831 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1832 * registered physical page range. Similarly
1833 * sparse_memory_present_with_active_regions() calls memory_present() for
1834 * each range when SPARSEMEM is enabled.
1836 * See mm/page_alloc.c for more information on each function exposed by
1837 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1839 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1840 unsigned long node_map_pfn_alignment(void);
1841 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1842 unsigned long end_pfn);
1843 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1844 unsigned long end_pfn);
1845 extern void get_pfn_range_for_nid(unsigned int nid,
1846 unsigned long *start_pfn, unsigned long *end_pfn);
1847 extern unsigned long find_min_pfn_with_active_regions(void);
1848 extern void free_bootmem_with_active_regions(int nid,
1849 unsigned long max_low_pfn);
1850 extern void sparse_memory_present_with_active_regions(int nid);
1852 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1854 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1855 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1856 static inline int __early_pfn_to_nid(unsigned long pfn,
1857 struct mminit_pfnnid_cache *state)
1862 /* please see mm/page_alloc.c */
1863 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1864 /* there is a per-arch backend function. */
1865 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1866 struct mminit_pfnnid_cache *state);
1869 extern void set_dma_reserve(unsigned long new_dma_reserve);
1870 extern void memmap_init_zone(unsigned long, int, unsigned long,
1871 unsigned long, enum memmap_context);
1872 extern void setup_per_zone_wmarks(void);
1873 extern int __meminit init_per_zone_wmark_min(void);
1874 extern void mem_init(void);
1875 extern void __init mmap_init(void);
1876 extern void show_mem(unsigned int flags);
1877 extern long si_mem_available(void);
1878 extern void si_meminfo(struct sysinfo * val);
1879 extern void si_meminfo_node(struct sysinfo *val, int nid);
1881 extern __printf(3, 4)
1882 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1883 const char *fmt, ...);
1885 extern void setup_per_cpu_pageset(void);
1887 extern void zone_pcp_update(struct zone *zone);
1888 extern void zone_pcp_reset(struct zone *zone);
1891 extern int min_free_kbytes;
1892 extern int watermark_scale_factor;
1895 extern atomic_long_t mmap_pages_allocated;
1896 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1898 /* interval_tree.c */
1899 void vma_interval_tree_insert(struct vm_area_struct *node,
1900 struct rb_root *root);
1901 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1902 struct vm_area_struct *prev,
1903 struct rb_root *root);
1904 void vma_interval_tree_remove(struct vm_area_struct *node,
1905 struct rb_root *root);
1906 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1907 unsigned long start, unsigned long last);
1908 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1909 unsigned long start, unsigned long last);
1911 #define vma_interval_tree_foreach(vma, root, start, last) \
1912 for (vma = vma_interval_tree_iter_first(root, start, last); \
1913 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1915 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1916 struct rb_root *root);
1917 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1918 struct rb_root *root);
1919 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1920 struct rb_root *root, unsigned long start, unsigned long last);
1921 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1922 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1923 #ifdef CONFIG_DEBUG_VM_RB
1924 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1927 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1928 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1929 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1932 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1933 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1934 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1935 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1936 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1937 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1938 struct mempolicy *, struct vm_userfaultfd_ctx);
1939 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1940 extern int split_vma(struct mm_struct *,
1941 struct vm_area_struct *, unsigned long addr, int new_below);
1942 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1943 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1944 struct rb_node **, struct rb_node *);
1945 extern void unlink_file_vma(struct vm_area_struct *);
1946 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1947 unsigned long addr, unsigned long len, pgoff_t pgoff,
1948 bool *need_rmap_locks);
1949 extern void exit_mmap(struct mm_struct *);
1951 static inline int check_data_rlimit(unsigned long rlim,
1953 unsigned long start,
1954 unsigned long end_data,
1955 unsigned long start_data)
1957 if (rlim < RLIM_INFINITY) {
1958 if (((new - start) + (end_data - start_data)) > rlim)
1965 extern int mm_take_all_locks(struct mm_struct *mm);
1966 extern void mm_drop_all_locks(struct mm_struct *mm);
1968 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1969 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1971 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
1972 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
1974 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1975 unsigned long addr, unsigned long len,
1976 unsigned long flags,
1977 const struct vm_special_mapping *spec);
1978 /* This is an obsolete alternative to _install_special_mapping. */
1979 extern int install_special_mapping(struct mm_struct *mm,
1980 unsigned long addr, unsigned long len,
1981 unsigned long flags, struct page **pages);
1983 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1985 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1986 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1987 extern unsigned long do_mmap(struct file *file, unsigned long addr,
1988 unsigned long len, unsigned long prot, unsigned long flags,
1989 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
1990 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1992 static inline unsigned long
1993 do_mmap_pgoff(struct file *file, unsigned long addr,
1994 unsigned long len, unsigned long prot, unsigned long flags,
1995 unsigned long pgoff, unsigned long *populate)
1997 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2001 extern int __mm_populate(unsigned long addr, unsigned long len,
2003 static inline void mm_populate(unsigned long addr, unsigned long len)
2006 (void) __mm_populate(addr, len, 1);
2009 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2012 /* These take the mm semaphore themselves */
2013 extern unsigned long vm_brk(unsigned long, unsigned long);
2014 extern int vm_munmap(unsigned long, size_t);
2015 extern unsigned long vm_mmap(struct file *, unsigned long,
2016 unsigned long, unsigned long,
2017 unsigned long, unsigned long);
2019 struct vm_unmapped_area_info {
2020 #define VM_UNMAPPED_AREA_TOPDOWN 1
2021 unsigned long flags;
2022 unsigned long length;
2023 unsigned long low_limit;
2024 unsigned long high_limit;
2025 unsigned long align_mask;
2026 unsigned long align_offset;
2029 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2030 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2033 * Search for an unmapped address range.
2035 * We are looking for a range that:
2036 * - does not intersect with any VMA;
2037 * - is contained within the [low_limit, high_limit) interval;
2038 * - is at least the desired size.
2039 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2041 static inline unsigned long
2042 vm_unmapped_area(struct vm_unmapped_area_info *info)
2044 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2045 return unmapped_area_topdown(info);
2047 return unmapped_area(info);
2051 extern void truncate_inode_pages(struct address_space *, loff_t);
2052 extern void truncate_inode_pages_range(struct address_space *,
2053 loff_t lstart, loff_t lend);
2054 extern void truncate_inode_pages_final(struct address_space *);
2056 /* generic vm_area_ops exported for stackable file systems */
2057 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2058 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2059 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2061 /* mm/page-writeback.c */
2062 int write_one_page(struct page *page, int wait);
2063 void task_dirty_inc(struct task_struct *tsk);
2066 #define VM_MAX_READAHEAD 128 /* kbytes */
2067 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2069 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2070 pgoff_t offset, unsigned long nr_to_read);
2072 void page_cache_sync_readahead(struct address_space *mapping,
2073 struct file_ra_state *ra,
2076 unsigned long size);
2078 void page_cache_async_readahead(struct address_space *mapping,
2079 struct file_ra_state *ra,
2083 unsigned long size);
2085 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2086 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2088 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2089 extern int expand_downwards(struct vm_area_struct *vma,
2090 unsigned long address);
2092 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2094 #define expand_upwards(vma, address) (0)
2097 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2098 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2099 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2100 struct vm_area_struct **pprev);
2102 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2103 NULL if none. Assume start_addr < end_addr. */
2104 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2106 struct vm_area_struct * vma = find_vma(mm,start_addr);
2108 if (vma && end_addr <= vma->vm_start)
2113 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2115 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2118 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2119 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2120 unsigned long vm_start, unsigned long vm_end)
2122 struct vm_area_struct *vma = find_vma(mm, vm_start);
2124 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2131 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2132 void vma_set_page_prot(struct vm_area_struct *vma);
2134 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2138 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2140 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2144 #ifdef CONFIG_NUMA_BALANCING
2145 unsigned long change_prot_numa(struct vm_area_struct *vma,
2146 unsigned long start, unsigned long end);
2149 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2150 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2151 unsigned long pfn, unsigned long size, pgprot_t);
2152 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2153 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2155 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2156 unsigned long pfn, pgprot_t pgprot);
2157 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2159 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2162 struct page *follow_page_mask(struct vm_area_struct *vma,
2163 unsigned long address, unsigned int foll_flags,
2164 unsigned int *page_mask);
2166 static inline struct page *follow_page(struct vm_area_struct *vma,
2167 unsigned long address, unsigned int foll_flags)
2169 unsigned int unused_page_mask;
2170 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2173 #define FOLL_WRITE 0x01 /* check pte is writable */
2174 #define FOLL_TOUCH 0x02 /* mark page accessed */
2175 #define FOLL_GET 0x04 /* do get_page on page */
2176 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2177 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2178 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2179 * and return without waiting upon it */
2180 #define FOLL_POPULATE 0x40 /* fault in page */
2181 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2182 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2183 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2184 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2185 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2186 #define FOLL_MLOCK 0x1000 /* lock present pages */
2188 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2190 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2191 unsigned long size, pte_fn_t fn, void *data);
2194 #ifdef CONFIG_PAGE_POISONING
2195 extern bool page_poisoning_enabled(void);
2196 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2197 extern bool page_is_poisoned(struct page *page);
2199 static inline bool page_poisoning_enabled(void) { return false; }
2200 static inline void kernel_poison_pages(struct page *page, int numpages,
2202 static inline bool page_is_poisoned(struct page *page) { return false; }
2205 #ifdef CONFIG_DEBUG_PAGEALLOC
2206 extern bool _debug_pagealloc_enabled;
2207 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2209 static inline bool debug_pagealloc_enabled(void)
2211 return _debug_pagealloc_enabled;
2215 kernel_map_pages(struct page *page, int numpages, int enable)
2217 if (!debug_pagealloc_enabled())
2220 __kernel_map_pages(page, numpages, enable);
2222 #ifdef CONFIG_HIBERNATION
2223 extern bool kernel_page_present(struct page *page);
2224 #endif /* CONFIG_HIBERNATION */
2225 #else /* CONFIG_DEBUG_PAGEALLOC */
2227 kernel_map_pages(struct page *page, int numpages, int enable) {}
2228 #ifdef CONFIG_HIBERNATION
2229 static inline bool kernel_page_present(struct page *page) { return true; }
2230 #endif /* CONFIG_HIBERNATION */
2231 static inline bool debug_pagealloc_enabled(void)
2235 #endif /* CONFIG_DEBUG_PAGEALLOC */
2237 #ifdef __HAVE_ARCH_GATE_AREA
2238 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2239 extern int in_gate_area_no_mm(unsigned long addr);
2240 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2242 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2246 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2247 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2251 #endif /* __HAVE_ARCH_GATE_AREA */
2253 #ifdef CONFIG_SYSCTL
2254 extern int sysctl_drop_caches;
2255 int drop_caches_sysctl_handler(struct ctl_table *, int,
2256 void __user *, size_t *, loff_t *);
2259 void drop_slab(void);
2260 void drop_slab_node(int nid);
2263 #define randomize_va_space 0
2265 extern int randomize_va_space;
2268 const char * arch_vma_name(struct vm_area_struct *vma);
2269 void print_vma_addr(char *prefix, unsigned long rip);
2271 void sparse_mem_maps_populate_node(struct page **map_map,
2272 unsigned long pnum_begin,
2273 unsigned long pnum_end,
2274 unsigned long map_count,
2277 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2278 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2279 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2280 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2281 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2282 void *vmemmap_alloc_block(unsigned long size, int node);
2284 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2285 struct vmem_altmap *altmap);
2286 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2288 return __vmemmap_alloc_block_buf(size, node, NULL);
2291 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2292 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2294 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2295 void vmemmap_populate_print_last(void);
2296 #ifdef CONFIG_MEMORY_HOTPLUG
2297 void vmemmap_free(unsigned long start, unsigned long end);
2299 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2300 unsigned long size);
2303 MF_COUNT_INCREASED = 1 << 0,
2304 MF_ACTION_REQUIRED = 1 << 1,
2305 MF_MUST_KILL = 1 << 2,
2306 MF_SOFT_OFFLINE = 1 << 3,
2308 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2309 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2310 extern int unpoison_memory(unsigned long pfn);
2311 extern int get_hwpoison_page(struct page *page);
2312 #define put_hwpoison_page(page) put_page(page)
2313 extern int sysctl_memory_failure_early_kill;
2314 extern int sysctl_memory_failure_recovery;
2315 extern void shake_page(struct page *p, int access);
2316 extern atomic_long_t num_poisoned_pages;
2317 extern int soft_offline_page(struct page *page, int flags);
2321 * Error handlers for various types of pages.
2324 MF_IGNORED, /* Error: cannot be handled */
2325 MF_FAILED, /* Error: handling failed */
2326 MF_DELAYED, /* Will be handled later */
2327 MF_RECOVERED, /* Successfully recovered */
2330 enum mf_action_page_type {
2332 MF_MSG_KERNEL_HIGH_ORDER,
2334 MF_MSG_DIFFERENT_COMPOUND,
2335 MF_MSG_POISONED_HUGE,
2338 MF_MSG_UNMAP_FAILED,
2339 MF_MSG_DIRTY_SWAPCACHE,
2340 MF_MSG_CLEAN_SWAPCACHE,
2341 MF_MSG_DIRTY_MLOCKED_LRU,
2342 MF_MSG_CLEAN_MLOCKED_LRU,
2343 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2344 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2347 MF_MSG_TRUNCATED_LRU,
2353 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2354 extern void clear_huge_page(struct page *page,
2356 unsigned int pages_per_huge_page);
2357 extern void copy_user_huge_page(struct page *dst, struct page *src,
2358 unsigned long addr, struct vm_area_struct *vma,
2359 unsigned int pages_per_huge_page);
2360 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2362 extern struct page_ext_operations debug_guardpage_ops;
2363 extern struct page_ext_operations page_poisoning_ops;
2365 #ifdef CONFIG_DEBUG_PAGEALLOC
2366 extern unsigned int _debug_guardpage_minorder;
2367 extern bool _debug_guardpage_enabled;
2369 static inline unsigned int debug_guardpage_minorder(void)
2371 return _debug_guardpage_minorder;
2374 static inline bool debug_guardpage_enabled(void)
2376 return _debug_guardpage_enabled;
2379 static inline bool page_is_guard(struct page *page)
2381 struct page_ext *page_ext;
2383 if (!debug_guardpage_enabled())
2386 page_ext = lookup_page_ext(page);
2387 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2390 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2391 static inline bool debug_guardpage_enabled(void) { return false; }
2392 static inline bool page_is_guard(struct page *page) { return false; }
2393 #endif /* CONFIG_DEBUG_PAGEALLOC */
2395 #if MAX_NUMNODES > 1
2396 void __init setup_nr_node_ids(void);
2398 static inline void setup_nr_node_ids(void) {}
2401 #endif /* __KERNEL__ */
2402 #endif /* _LINUX_MM_H */