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>
25 #include <linux/page_ref.h>
29 struct anon_vma_chain;
32 struct writeback_control;
35 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
36 extern unsigned long max_mapnr;
38 static inline void set_max_mapnr(unsigned long limit)
43 static inline void set_max_mapnr(unsigned long limit) { }
46 extern unsigned long totalram_pages;
47 extern void * high_memory;
48 extern int page_cluster;
51 extern int sysctl_legacy_va_layout;
53 #define sysctl_legacy_va_layout 0
56 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
57 extern const int mmap_rnd_bits_min;
58 extern const int mmap_rnd_bits_max;
59 extern int mmap_rnd_bits __read_mostly;
61 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
62 extern const int mmap_rnd_compat_bits_min;
63 extern const int mmap_rnd_compat_bits_max;
64 extern int mmap_rnd_compat_bits __read_mostly;
68 #include <asm/pgtable.h>
69 #include <asm/processor.h>
72 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
76 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
80 * To prevent common memory management code establishing
81 * a zero page mapping on a read fault.
82 * This macro should be defined within <asm/pgtable.h>.
83 * s390 does this to prevent multiplexing of hardware bits
84 * related to the physical page in case of virtualization.
86 #ifndef mm_forbids_zeropage
87 #define mm_forbids_zeropage(X) (0)
91 * Default maximum number of active map areas, this limits the number of vmas
92 * per mm struct. Users can overwrite this number by sysctl but there is a
95 * When a program's coredump is generated as ELF format, a section is created
96 * per a vma. In ELF, the number of sections is represented in unsigned short.
97 * This means the number of sections should be smaller than 65535 at coredump.
98 * Because the kernel adds some informative sections to a image of program at
99 * generating coredump, we need some margin. The number of extra sections is
100 * 1-3 now and depends on arch. We use "5" as safe margin, here.
102 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
103 * not a hard limit any more. Although some userspace tools can be surprised by
106 #define MAPCOUNT_ELF_CORE_MARGIN (5)
107 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
109 extern int sysctl_max_map_count;
111 extern unsigned long sysctl_user_reserve_kbytes;
112 extern unsigned long sysctl_admin_reserve_kbytes;
114 extern int sysctl_overcommit_memory;
115 extern int sysctl_overcommit_ratio;
116 extern unsigned long sysctl_overcommit_kbytes;
118 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
120 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
123 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
125 /* to align the pointer to the (next) page boundary */
126 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
128 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
129 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
132 * Linux kernel virtual memory manager primitives.
133 * The idea being to have a "virtual" mm in the same way
134 * we have a virtual fs - giving a cleaner interface to the
135 * mm details, and allowing different kinds of memory mappings
136 * (from shared memory to executable loading to arbitrary
140 extern struct kmem_cache *vm_area_cachep;
143 extern struct rb_root nommu_region_tree;
144 extern struct rw_semaphore nommu_region_sem;
146 extern unsigned int kobjsize(const void *objp);
150 * vm_flags in vm_area_struct, see mm_types.h.
151 * When changing, update also include/trace/events/mmflags.h
153 #define VM_NONE 0x00000000
155 #define VM_READ 0x00000001 /* currently active flags */
156 #define VM_WRITE 0x00000002
157 #define VM_EXEC 0x00000004
158 #define VM_SHARED 0x00000008
160 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
161 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
162 #define VM_MAYWRITE 0x00000020
163 #define VM_MAYEXEC 0x00000040
164 #define VM_MAYSHARE 0x00000080
166 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
167 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
168 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
169 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
170 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
172 #define VM_LOCKED 0x00002000
173 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
175 /* Used by sys_madvise() */
176 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
177 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
179 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
180 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
181 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
182 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
183 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
184 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
185 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
186 #define VM_ARCH_2 0x02000000
187 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
189 #ifdef CONFIG_MEM_SOFT_DIRTY
190 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
192 # define VM_SOFTDIRTY 0
195 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
196 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
197 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
198 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
200 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
201 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
202 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
203 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
204 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
205 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
206 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
207 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
208 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
209 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
211 #if defined(CONFIG_X86)
212 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
213 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
214 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
215 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
216 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
217 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
218 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
220 #elif defined(CONFIG_PPC)
221 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
222 #elif defined(CONFIG_PARISC)
223 # define VM_GROWSUP VM_ARCH_1
224 #elif defined(CONFIG_METAG)
225 # define VM_GROWSUP VM_ARCH_1
226 #elif defined(CONFIG_IA64)
227 # define VM_GROWSUP VM_ARCH_1
228 #elif !defined(CONFIG_MMU)
229 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
232 #if defined(CONFIG_X86)
233 /* MPX specific bounds table or bounds directory */
234 # define VM_MPX VM_ARCH_2
238 # define VM_GROWSUP VM_NONE
241 /* Bits set in the VMA until the stack is in its final location */
242 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
244 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
245 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
248 #ifdef CONFIG_STACK_GROWSUP
249 #define VM_STACK VM_GROWSUP
251 #define VM_STACK VM_GROWSDOWN
254 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
257 * Special vmas that are non-mergable, non-mlock()able.
258 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
260 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
262 /* This mask defines which mm->def_flags a process can inherit its parent */
263 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
265 /* This mask is used to clear all the VMA flags used by mlock */
266 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
269 * mapping from the currently active vm_flags protection bits (the
270 * low four bits) to a page protection mask..
272 extern pgprot_t protection_map[16];
274 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
275 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
276 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
277 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
278 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
279 #define FAULT_FLAG_TRIED 0x20 /* Second try */
280 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
281 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
282 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
285 * vm_fault is filled by the the pagefault handler and passed to the vma's
286 * ->fault function. The vma's ->fault is responsible for returning a bitmask
287 * of VM_FAULT_xxx flags that give details about how the fault was handled.
289 * MM layer fills up gfp_mask for page allocations but fault handler might
290 * alter it if its implementation requires a different allocation context.
292 * pgoff should be used in favour of virtual_address, if possible.
295 unsigned int flags; /* FAULT_FLAG_xxx flags */
296 gfp_t gfp_mask; /* gfp mask to be used for allocations */
297 pgoff_t pgoff; /* Logical page offset based on vma */
298 void __user *virtual_address; /* Faulting virtual address */
300 struct page *cow_page; /* Handler may choose to COW */
301 struct page *page; /* ->fault handlers should return a
302 * page here, unless VM_FAULT_NOPAGE
303 * is set (which is also implied by
306 /* for ->map_pages() only */
307 pgoff_t max_pgoff; /* map pages for offset from pgoff till
308 * max_pgoff inclusive */
309 pte_t *pte; /* pte entry associated with ->pgoff */
313 * These are the virtual MM functions - opening of an area, closing and
314 * unmapping it (needed to keep files on disk up-to-date etc), pointer
315 * to the functions called when a no-page or a wp-page exception occurs.
317 struct vm_operations_struct {
318 void (*open)(struct vm_area_struct * area);
319 void (*close)(struct vm_area_struct * area);
320 int (*mremap)(struct vm_area_struct * area);
321 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
322 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
323 pmd_t *, unsigned int flags);
324 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
326 /* notification that a previously read-only page is about to become
327 * writable, if an error is returned it will cause a SIGBUS */
328 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
330 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
331 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
333 /* called by access_process_vm when get_user_pages() fails, typically
334 * for use by special VMAs that can switch between memory and hardware
336 int (*access)(struct vm_area_struct *vma, unsigned long addr,
337 void *buf, int len, int write);
339 /* Called by the /proc/PID/maps code to ask the vma whether it
340 * has a special name. Returning non-NULL will also cause this
341 * vma to be dumped unconditionally. */
342 const char *(*name)(struct vm_area_struct *vma);
346 * set_policy() op must add a reference to any non-NULL @new mempolicy
347 * to hold the policy upon return. Caller should pass NULL @new to
348 * remove a policy and fall back to surrounding context--i.e. do not
349 * install a MPOL_DEFAULT policy, nor the task or system default
352 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
355 * get_policy() op must add reference [mpol_get()] to any policy at
356 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
357 * in mm/mempolicy.c will do this automatically.
358 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
359 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
360 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
361 * must return NULL--i.e., do not "fallback" to task or system default
364 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
368 * Called by vm_normal_page() for special PTEs to find the
369 * page for @addr. This is useful if the default behavior
370 * (using pte_page()) would not find the correct page.
372 struct page *(*find_special_page)(struct vm_area_struct *vma,
379 #define page_private(page) ((page)->private)
380 #define set_page_private(page, v) ((page)->private = (v))
382 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
383 static inline int pmd_devmap(pmd_t pmd)
390 * FIXME: take this include out, include page-flags.h in
391 * files which need it (119 of them)
393 #include <linux/page-flags.h>
394 #include <linux/huge_mm.h>
397 * Methods to modify the page usage count.
399 * What counts for a page usage:
400 * - cache mapping (page->mapping)
401 * - private data (page->private)
402 * - page mapped in a task's page tables, each mapping
403 * is counted separately
405 * Also, many kernel routines increase the page count before a critical
406 * routine so they can be sure the page doesn't go away from under them.
410 * Drop a ref, return true if the refcount fell to zero (the page has no users)
412 static inline int put_page_testzero(struct page *page)
414 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
415 return page_ref_dec_and_test(page);
419 * Try to grab a ref unless the page has a refcount of zero, return false if
421 * This can be called when MMU is off so it must not access
422 * any of the virtual mappings.
424 static inline int get_page_unless_zero(struct page *page)
426 return page_ref_add_unless(page, 1, 0);
429 extern int page_is_ram(unsigned long pfn);
437 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
440 /* Support for virtually mapped pages */
441 struct page *vmalloc_to_page(const void *addr);
442 unsigned long vmalloc_to_pfn(const void *addr);
445 * Determine if an address is within the vmalloc range
447 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
448 * is no special casing required.
450 static inline int is_vmalloc_addr(const void *x)
453 unsigned long addr = (unsigned long)x;
455 return addr >= VMALLOC_START && addr < VMALLOC_END;
461 extern int is_vmalloc_or_module_addr(const void *x);
463 static inline int is_vmalloc_or_module_addr(const void *x)
469 extern void kvfree(const void *addr);
471 static inline atomic_t *compound_mapcount_ptr(struct page *page)
473 return &page[1].compound_mapcount;
476 static inline int compound_mapcount(struct page *page)
478 if (!PageCompound(page))
480 page = compound_head(page);
481 return atomic_read(compound_mapcount_ptr(page)) + 1;
485 * The atomic page->_mapcount, starts from -1: so that transitions
486 * both from it and to it can be tracked, using atomic_inc_and_test
487 * and atomic_add_negative(-1).
489 static inline void page_mapcount_reset(struct page *page)
491 atomic_set(&(page)->_mapcount, -1);
494 int __page_mapcount(struct page *page);
496 static inline int page_mapcount(struct page *page)
498 VM_BUG_ON_PAGE(PageSlab(page), page);
500 if (unlikely(PageCompound(page)))
501 return __page_mapcount(page);
502 return atomic_read(&page->_mapcount) + 1;
505 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
506 int total_mapcount(struct page *page);
508 static inline int total_mapcount(struct page *page)
510 return page_mapcount(page);
514 static inline struct page *virt_to_head_page(const void *x)
516 struct page *page = virt_to_page(x);
518 return compound_head(page);
521 void __put_page(struct page *page);
523 void put_pages_list(struct list_head *pages);
525 void split_page(struct page *page, unsigned int order);
526 int split_free_page(struct page *page);
529 * Compound pages have a destructor function. Provide a
530 * prototype for that function and accessor functions.
531 * These are _only_ valid on the head of a compound page.
533 typedef void compound_page_dtor(struct page *);
535 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
536 enum compound_dtor_id {
539 #ifdef CONFIG_HUGETLB_PAGE
542 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
547 extern compound_page_dtor * const compound_page_dtors[];
549 static inline void set_compound_page_dtor(struct page *page,
550 enum compound_dtor_id compound_dtor)
552 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
553 page[1].compound_dtor = compound_dtor;
556 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
558 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
559 return compound_page_dtors[page[1].compound_dtor];
562 static inline unsigned int compound_order(struct page *page)
566 return page[1].compound_order;
569 static inline void set_compound_order(struct page *page, unsigned int order)
571 page[1].compound_order = order;
574 void free_compound_page(struct page *page);
578 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
579 * servicing faults for write access. In the normal case, do always want
580 * pte_mkwrite. But get_user_pages can cause write faults for mappings
581 * that do not have writing enabled, when used by access_process_vm.
583 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
585 if (likely(vma->vm_flags & VM_WRITE))
586 pte = pte_mkwrite(pte);
590 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
591 struct page *page, pte_t *pte, bool write, bool anon);
595 * Multiple processes may "see" the same page. E.g. for untouched
596 * mappings of /dev/null, all processes see the same page full of
597 * zeroes, and text pages of executables and shared libraries have
598 * only one copy in memory, at most, normally.
600 * For the non-reserved pages, page_count(page) denotes a reference count.
601 * page_count() == 0 means the page is free. page->lru is then used for
602 * freelist management in the buddy allocator.
603 * page_count() > 0 means the page has been allocated.
605 * Pages are allocated by the slab allocator in order to provide memory
606 * to kmalloc and kmem_cache_alloc. In this case, the management of the
607 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
608 * unless a particular usage is carefully commented. (the responsibility of
609 * freeing the kmalloc memory is the caller's, of course).
611 * A page may be used by anyone else who does a __get_free_page().
612 * In this case, page_count still tracks the references, and should only
613 * be used through the normal accessor functions. The top bits of page->flags
614 * and page->virtual store page management information, but all other fields
615 * are unused and could be used privately, carefully. The management of this
616 * page is the responsibility of the one who allocated it, and those who have
617 * subsequently been given references to it.
619 * The other pages (we may call them "pagecache pages") are completely
620 * managed by the Linux memory manager: I/O, buffers, swapping etc.
621 * The following discussion applies only to them.
623 * A pagecache page contains an opaque `private' member, which belongs to the
624 * page's address_space. Usually, this is the address of a circular list of
625 * the page's disk buffers. PG_private must be set to tell the VM to call
626 * into the filesystem to release these pages.
628 * A page may belong to an inode's memory mapping. In this case, page->mapping
629 * is the pointer to the inode, and page->index is the file offset of the page,
630 * in units of PAGE_SIZE.
632 * If pagecache pages are not associated with an inode, they are said to be
633 * anonymous pages. These may become associated with the swapcache, and in that
634 * case PG_swapcache is set, and page->private is an offset into the swapcache.
636 * In either case (swapcache or inode backed), the pagecache itself holds one
637 * reference to the page. Setting PG_private should also increment the
638 * refcount. The each user mapping also has a reference to the page.
640 * The pagecache pages are stored in a per-mapping radix tree, which is
641 * rooted at mapping->page_tree, and indexed by offset.
642 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
643 * lists, we instead now tag pages as dirty/writeback in the radix tree.
645 * All pagecache pages may be subject to I/O:
646 * - inode pages may need to be read from disk,
647 * - inode pages which have been modified and are MAP_SHARED may need
648 * to be written back to the inode on disk,
649 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
650 * modified may need to be swapped out to swap space and (later) to be read
655 * The zone field is never updated after free_area_init_core()
656 * sets it, so none of the operations on it need to be atomic.
659 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
660 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
661 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
662 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
663 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
666 * Define the bit shifts to access each section. For non-existent
667 * sections we define the shift as 0; that plus a 0 mask ensures
668 * the compiler will optimise away reference to them.
670 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
671 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
672 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
673 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
675 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
676 #ifdef NODE_NOT_IN_PAGE_FLAGS
677 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
678 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
679 SECTIONS_PGOFF : ZONES_PGOFF)
681 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
682 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
683 NODES_PGOFF : ZONES_PGOFF)
686 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
688 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
689 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
692 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
693 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
694 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
695 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
696 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
698 static inline enum zone_type page_zonenum(const struct page *page)
700 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
703 #ifdef CONFIG_ZONE_DEVICE
704 void get_zone_device_page(struct page *page);
705 void put_zone_device_page(struct page *page);
706 static inline bool is_zone_device_page(const struct page *page)
708 return page_zonenum(page) == ZONE_DEVICE;
711 static inline void get_zone_device_page(struct page *page)
714 static inline void put_zone_device_page(struct page *page)
717 static inline bool is_zone_device_page(const struct page *page)
723 static inline void get_page(struct page *page)
725 page = compound_head(page);
727 * Getting a normal page or the head of a compound page
728 * requires to already have an elevated page->_count.
730 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
733 if (unlikely(is_zone_device_page(page)))
734 get_zone_device_page(page);
737 static inline void put_page(struct page *page)
739 page = compound_head(page);
741 if (put_page_testzero(page))
744 if (unlikely(is_zone_device_page(page)))
745 put_zone_device_page(page);
748 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
749 #define SECTION_IN_PAGE_FLAGS
753 * The identification function is mainly used by the buddy allocator for
754 * determining if two pages could be buddies. We are not really identifying
755 * the zone since we could be using the section number id if we do not have
756 * node id available in page flags.
757 * We only guarantee that it will return the same value for two combinable
760 static inline int page_zone_id(struct page *page)
762 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
765 static inline int zone_to_nid(struct zone *zone)
774 #ifdef NODE_NOT_IN_PAGE_FLAGS
775 extern int page_to_nid(const struct page *page);
777 static inline int page_to_nid(const struct page *page)
779 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
783 #ifdef CONFIG_NUMA_BALANCING
784 static inline int cpu_pid_to_cpupid(int cpu, int pid)
786 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
789 static inline int cpupid_to_pid(int cpupid)
791 return cpupid & LAST__PID_MASK;
794 static inline int cpupid_to_cpu(int cpupid)
796 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
799 static inline int cpupid_to_nid(int cpupid)
801 return cpu_to_node(cpupid_to_cpu(cpupid));
804 static inline bool cpupid_pid_unset(int cpupid)
806 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
809 static inline bool cpupid_cpu_unset(int cpupid)
811 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
814 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
816 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
819 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
820 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
821 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
823 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
826 static inline int page_cpupid_last(struct page *page)
828 return page->_last_cpupid;
830 static inline void page_cpupid_reset_last(struct page *page)
832 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
835 static inline int page_cpupid_last(struct page *page)
837 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
840 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
842 static inline void page_cpupid_reset_last(struct page *page)
844 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
846 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
847 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
849 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
850 #else /* !CONFIG_NUMA_BALANCING */
851 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
853 return page_to_nid(page); /* XXX */
856 static inline int page_cpupid_last(struct page *page)
858 return page_to_nid(page); /* XXX */
861 static inline int cpupid_to_nid(int cpupid)
866 static inline int cpupid_to_pid(int cpupid)
871 static inline int cpupid_to_cpu(int cpupid)
876 static inline int cpu_pid_to_cpupid(int nid, int pid)
881 static inline bool cpupid_pid_unset(int cpupid)
886 static inline void page_cpupid_reset_last(struct page *page)
890 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
894 #endif /* CONFIG_NUMA_BALANCING */
896 static inline struct zone *page_zone(const struct page *page)
898 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
901 #ifdef SECTION_IN_PAGE_FLAGS
902 static inline void set_page_section(struct page *page, unsigned long section)
904 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
905 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
908 static inline unsigned long page_to_section(const struct page *page)
910 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
914 static inline void set_page_zone(struct page *page, enum zone_type zone)
916 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
917 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
920 static inline void set_page_node(struct page *page, unsigned long node)
922 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
923 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
926 static inline void set_page_links(struct page *page, enum zone_type zone,
927 unsigned long node, unsigned long pfn)
929 set_page_zone(page, zone);
930 set_page_node(page, node);
931 #ifdef SECTION_IN_PAGE_FLAGS
932 set_page_section(page, pfn_to_section_nr(pfn));
937 static inline struct mem_cgroup *page_memcg(struct page *page)
939 return page->mem_cgroup;
942 static inline struct mem_cgroup *page_memcg(struct page *page)
949 * Some inline functions in vmstat.h depend on page_zone()
951 #include <linux/vmstat.h>
953 static __always_inline void *lowmem_page_address(const struct page *page)
955 return page_to_virt(page);
958 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
959 #define HASHED_PAGE_VIRTUAL
962 #if defined(WANT_PAGE_VIRTUAL)
963 static inline void *page_address(const struct page *page)
965 return page->virtual;
967 static inline void set_page_address(struct page *page, void *address)
969 page->virtual = address;
971 #define page_address_init() do { } while(0)
974 #if defined(HASHED_PAGE_VIRTUAL)
975 void *page_address(const struct page *page);
976 void set_page_address(struct page *page, void *virtual);
977 void page_address_init(void);
980 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
981 #define page_address(page) lowmem_page_address(page)
982 #define set_page_address(page, address) do { } while(0)
983 #define page_address_init() do { } while(0)
986 extern void *page_rmapping(struct page *page);
987 extern struct anon_vma *page_anon_vma(struct page *page);
988 extern struct address_space *page_mapping(struct page *page);
990 extern struct address_space *__page_file_mapping(struct page *);
993 struct address_space *page_file_mapping(struct page *page)
995 if (unlikely(PageSwapCache(page)))
996 return __page_file_mapping(page);
998 return page->mapping;
1002 * Return the pagecache index of the passed page. Regular pagecache pages
1003 * use ->index whereas swapcache pages use ->private
1005 static inline pgoff_t page_index(struct page *page)
1007 if (unlikely(PageSwapCache(page)))
1008 return page_private(page);
1012 extern pgoff_t __page_file_index(struct page *page);
1015 * Return the file index of the page. Regular pagecache pages use ->index
1016 * whereas swapcache pages use swp_offset(->private)
1018 static inline pgoff_t page_file_index(struct page *page)
1020 if (unlikely(PageSwapCache(page)))
1021 return __page_file_index(page);
1027 * Return true if this page is mapped into pagetables.
1028 * For compound page it returns true if any subpage of compound page is mapped.
1030 static inline bool page_mapped(struct page *page)
1033 if (likely(!PageCompound(page)))
1034 return atomic_read(&page->_mapcount) >= 0;
1035 page = compound_head(page);
1036 if (atomic_read(compound_mapcount_ptr(page)) >= 0)
1038 for (i = 0; i < hpage_nr_pages(page); i++) {
1039 if (atomic_read(&page[i]._mapcount) >= 0)
1046 * Return true only if the page has been allocated with
1047 * ALLOC_NO_WATERMARKS and the low watermark was not
1048 * met implying that the system is under some pressure.
1050 static inline bool page_is_pfmemalloc(struct page *page)
1053 * Page index cannot be this large so this must be
1054 * a pfmemalloc page.
1056 return page->index == -1UL;
1060 * Only to be called by the page allocator on a freshly allocated
1063 static inline void set_page_pfmemalloc(struct page *page)
1068 static inline void clear_page_pfmemalloc(struct page *page)
1074 * Different kinds of faults, as returned by handle_mm_fault().
1075 * Used to decide whether a process gets delivered SIGBUS or
1076 * just gets major/minor fault counters bumped up.
1079 #define VM_FAULT_OOM 0x0001
1080 #define VM_FAULT_SIGBUS 0x0002
1081 #define VM_FAULT_MAJOR 0x0004
1082 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1083 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1084 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1085 #define VM_FAULT_SIGSEGV 0x0040
1087 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1088 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1089 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1090 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1092 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1094 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1095 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1098 /* Encode hstate index for a hwpoisoned large page */
1099 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1100 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1103 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1105 extern void pagefault_out_of_memory(void);
1107 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1110 * Flags passed to show_mem() and show_free_areas() to suppress output in
1113 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1115 extern void show_free_areas(unsigned int flags);
1116 extern bool skip_free_areas_node(unsigned int flags, int nid);
1118 int shmem_zero_setup(struct vm_area_struct *);
1120 bool shmem_mapping(struct address_space *mapping);
1122 static inline bool shmem_mapping(struct address_space *mapping)
1128 extern bool can_do_mlock(void);
1129 extern int user_shm_lock(size_t, struct user_struct *);
1130 extern void user_shm_unlock(size_t, struct user_struct *);
1133 * Parameter block passed down to zap_pte_range in exceptional cases.
1135 struct zap_details {
1136 struct address_space *check_mapping; /* Check page->mapping if set */
1137 pgoff_t first_index; /* Lowest page->index to unmap */
1138 pgoff_t last_index; /* Highest page->index to unmap */
1139 bool ignore_dirty; /* Ignore dirty pages */
1140 bool check_swap_entries; /* Check also swap entries */
1143 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1146 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1147 unsigned long size);
1148 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1149 unsigned long size, struct zap_details *);
1150 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1151 unsigned long start, unsigned long end);
1154 * mm_walk - callbacks for walk_page_range
1155 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1156 * this handler is required to be able to handle
1157 * pmd_trans_huge() pmds. They may simply choose to
1158 * split_huge_page() instead of handling it explicitly.
1159 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1160 * @pte_hole: if set, called for each hole at all levels
1161 * @hugetlb_entry: if set, called for each hugetlb entry
1162 * @test_walk: caller specific callback function to determine whether
1163 * we walk over the current vma or not. A positive returned
1164 * value means "do page table walk over the current vma,"
1165 * and a negative one means "abort current page table walk
1166 * right now." 0 means "skip the current vma."
1167 * @mm: mm_struct representing the target process of page table walk
1168 * @vma: vma currently walked (NULL if walking outside vmas)
1169 * @private: private data for callbacks' usage
1171 * (see the comment on walk_page_range() for more details)
1174 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1175 unsigned long next, struct mm_walk *walk);
1176 int (*pte_entry)(pte_t *pte, unsigned long addr,
1177 unsigned long next, struct mm_walk *walk);
1178 int (*pte_hole)(unsigned long addr, unsigned long next,
1179 struct mm_walk *walk);
1180 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1181 unsigned long addr, unsigned long next,
1182 struct mm_walk *walk);
1183 int (*test_walk)(unsigned long addr, unsigned long next,
1184 struct mm_walk *walk);
1185 struct mm_struct *mm;
1186 struct vm_area_struct *vma;
1190 int walk_page_range(unsigned long addr, unsigned long end,
1191 struct mm_walk *walk);
1192 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1193 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1194 unsigned long end, unsigned long floor, unsigned long ceiling);
1195 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1196 struct vm_area_struct *vma);
1197 void unmap_mapping_range(struct address_space *mapping,
1198 loff_t const holebegin, loff_t const holelen, int even_cows);
1199 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1200 unsigned long *pfn);
1201 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1202 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1203 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1204 void *buf, int len, int write);
1206 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1207 loff_t const holebegin, loff_t const holelen)
1209 unmap_mapping_range(mapping, holebegin, holelen, 0);
1212 extern void truncate_pagecache(struct inode *inode, loff_t new);
1213 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1214 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1215 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1216 int truncate_inode_page(struct address_space *mapping, struct page *page);
1217 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1218 int invalidate_inode_page(struct page *page);
1221 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1222 unsigned long address, unsigned int flags);
1223 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1224 unsigned long address, unsigned int fault_flags,
1227 static inline int handle_mm_fault(struct mm_struct *mm,
1228 struct vm_area_struct *vma, unsigned long address,
1231 /* should never happen if there's no MMU */
1233 return VM_FAULT_SIGBUS;
1235 static inline int fixup_user_fault(struct task_struct *tsk,
1236 struct mm_struct *mm, unsigned long address,
1237 unsigned int fault_flags, bool *unlocked)
1239 /* should never happen if there's no MMU */
1245 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1246 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1247 void *buf, int len, int write);
1249 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1250 unsigned long start, unsigned long nr_pages,
1251 unsigned int foll_flags, struct page **pages,
1252 struct vm_area_struct **vmas, int *nonblocking);
1253 long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm,
1254 unsigned long start, unsigned long nr_pages,
1255 int write, int force, struct page **pages,
1256 struct vm_area_struct **vmas);
1257 long get_user_pages6(unsigned long start, unsigned long nr_pages,
1258 int write, int force, struct page **pages,
1259 struct vm_area_struct **vmas);
1260 long get_user_pages_locked6(unsigned long start, unsigned long nr_pages,
1261 int write, int force, struct page **pages, int *locked);
1262 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1263 unsigned long start, unsigned long nr_pages,
1264 int write, int force, struct page **pages,
1265 unsigned int gup_flags);
1266 long get_user_pages_unlocked5(unsigned long start, unsigned long nr_pages,
1267 int write, int force, struct page **pages);
1268 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1269 struct page **pages);
1271 /* suppress warnings from use in EXPORT_SYMBOL() */
1272 #ifndef __DISABLE_GUP_DEPRECATED
1273 #define __gup_deprecated __deprecated
1275 #define __gup_deprecated
1278 * These macros provide backward-compatibility with the old
1279 * get_user_pages() variants which took tsk/mm. These
1280 * functions/macros provide both compile-time __deprecated so we
1281 * can catch old-style use and not break the build. The actual
1282 * functions also have WARN_ON()s to let us know at runtime if
1283 * the get_user_pages() should have been the "remote" variant.
1285 * These are hideous, but temporary.
1287 * If you run into one of these __deprecated warnings, look
1288 * at how you are calling get_user_pages(). If you are calling
1289 * it with current/current->mm as the first two arguments,
1290 * simply remove those arguments. The behavior will be the same
1291 * as it is now. If you are calling it on another task, use
1292 * get_user_pages_remote() instead.
1294 * Any questions? Ask Dave Hansen <dave@sr71.net>
1298 get_user_pages8(struct task_struct *tsk, struct mm_struct *mm,
1299 unsigned long start, unsigned long nr_pages,
1300 int write, int force, struct page **pages,
1301 struct vm_area_struct **vmas);
1302 #define GUP_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, get_user_pages, ...) \
1304 #define get_user_pages(...) GUP_MACRO(__VA_ARGS__, \
1305 get_user_pages8, x, \
1306 get_user_pages6, x, x, x, x, x)(__VA_ARGS__)
1309 long get_user_pages_locked8(struct task_struct *tsk, struct mm_struct *mm,
1310 unsigned long start, unsigned long nr_pages,
1311 int write, int force, struct page **pages,
1313 #define GUPL_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, get_user_pages_locked, ...) \
1314 get_user_pages_locked
1315 #define get_user_pages_locked(...) GUPL_MACRO(__VA_ARGS__, \
1316 get_user_pages_locked8, x, \
1317 get_user_pages_locked6, x, x, x, x)(__VA_ARGS__)
1320 long get_user_pages_unlocked7(struct task_struct *tsk, struct mm_struct *mm,
1321 unsigned long start, unsigned long nr_pages,
1322 int write, int force, struct page **pages);
1323 #define GUPU_MACRO(_1, _2, _3, _4, _5, _6, _7, get_user_pages_unlocked, ...) \
1324 get_user_pages_unlocked
1325 #define get_user_pages_unlocked(...) GUPU_MACRO(__VA_ARGS__, \
1326 get_user_pages_unlocked7, x, \
1327 get_user_pages_unlocked5, x, x, x, x)(__VA_ARGS__)
1329 /* Container for pinned pfns / pages */
1330 struct frame_vector {
1331 unsigned int nr_allocated; /* Number of frames we have space for */
1332 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1333 bool got_ref; /* Did we pin pages by getting page ref? */
1334 bool is_pfns; /* Does array contain pages or pfns? */
1335 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1336 * pfns_vector_pages() or pfns_vector_pfns()
1340 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1341 void frame_vector_destroy(struct frame_vector *vec);
1342 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1343 bool write, bool force, struct frame_vector *vec);
1344 void put_vaddr_frames(struct frame_vector *vec);
1345 int frame_vector_to_pages(struct frame_vector *vec);
1346 void frame_vector_to_pfns(struct frame_vector *vec);
1348 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1350 return vec->nr_frames;
1353 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1356 int err = frame_vector_to_pages(vec);
1359 return ERR_PTR(err);
1361 return (struct page **)(vec->ptrs);
1364 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1367 frame_vector_to_pfns(vec);
1368 return (unsigned long *)(vec->ptrs);
1372 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1373 struct page **pages);
1374 int get_kernel_page(unsigned long start, int write, struct page **pages);
1375 struct page *get_dump_page(unsigned long addr);
1377 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1378 extern void do_invalidatepage(struct page *page, unsigned int offset,
1379 unsigned int length);
1381 int __set_page_dirty_nobuffers(struct page *page);
1382 int __set_page_dirty_no_writeback(struct page *page);
1383 int redirty_page_for_writepage(struct writeback_control *wbc,
1385 void account_page_dirtied(struct page *page, struct address_space *mapping);
1386 void account_page_cleaned(struct page *page, struct address_space *mapping,
1387 struct bdi_writeback *wb);
1388 int set_page_dirty(struct page *page);
1389 int set_page_dirty_lock(struct page *page);
1390 void cancel_dirty_page(struct page *page);
1391 int clear_page_dirty_for_io(struct page *page);
1393 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1395 /* Is the vma a continuation of the stack vma above it? */
1396 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1398 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1401 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1403 return !vma->vm_ops;
1406 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1409 return (vma->vm_flags & VM_GROWSDOWN) &&
1410 (vma->vm_start == addr) &&
1411 !vma_growsdown(vma->vm_prev, addr);
1414 /* Is the vma a continuation of the stack vma below it? */
1415 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1417 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1420 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1423 return (vma->vm_flags & VM_GROWSUP) &&
1424 (vma->vm_end == addr) &&
1425 !vma_growsup(vma->vm_next, addr);
1428 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1430 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1431 unsigned long old_addr, struct vm_area_struct *new_vma,
1432 unsigned long new_addr, unsigned long len,
1433 bool need_rmap_locks);
1434 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1435 unsigned long end, pgprot_t newprot,
1436 int dirty_accountable, int prot_numa);
1437 extern int mprotect_fixup(struct vm_area_struct *vma,
1438 struct vm_area_struct **pprev, unsigned long start,
1439 unsigned long end, unsigned long newflags);
1442 * doesn't attempt to fault and will return short.
1444 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1445 struct page **pages);
1447 * per-process(per-mm_struct) statistics.
1449 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1451 long val = atomic_long_read(&mm->rss_stat.count[member]);
1453 #ifdef SPLIT_RSS_COUNTING
1455 * counter is updated in asynchronous manner and may go to minus.
1456 * But it's never be expected number for users.
1461 return (unsigned long)val;
1464 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1466 atomic_long_add(value, &mm->rss_stat.count[member]);
1469 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1471 atomic_long_inc(&mm->rss_stat.count[member]);
1474 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1476 atomic_long_dec(&mm->rss_stat.count[member]);
1479 /* Optimized variant when page is already known not to be PageAnon */
1480 static inline int mm_counter_file(struct page *page)
1482 if (PageSwapBacked(page))
1483 return MM_SHMEMPAGES;
1484 return MM_FILEPAGES;
1487 static inline int mm_counter(struct page *page)
1490 return MM_ANONPAGES;
1491 return mm_counter_file(page);
1494 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1496 return get_mm_counter(mm, MM_FILEPAGES) +
1497 get_mm_counter(mm, MM_ANONPAGES) +
1498 get_mm_counter(mm, MM_SHMEMPAGES);
1501 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1503 return max(mm->hiwater_rss, get_mm_rss(mm));
1506 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1508 return max(mm->hiwater_vm, mm->total_vm);
1511 static inline void update_hiwater_rss(struct mm_struct *mm)
1513 unsigned long _rss = get_mm_rss(mm);
1515 if ((mm)->hiwater_rss < _rss)
1516 (mm)->hiwater_rss = _rss;
1519 static inline void update_hiwater_vm(struct mm_struct *mm)
1521 if (mm->hiwater_vm < mm->total_vm)
1522 mm->hiwater_vm = mm->total_vm;
1525 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1527 mm->hiwater_rss = get_mm_rss(mm);
1530 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1531 struct mm_struct *mm)
1533 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1535 if (*maxrss < hiwater_rss)
1536 *maxrss = hiwater_rss;
1539 #if defined(SPLIT_RSS_COUNTING)
1540 void sync_mm_rss(struct mm_struct *mm);
1542 static inline void sync_mm_rss(struct mm_struct *mm)
1547 #ifndef __HAVE_ARCH_PTE_DEVMAP
1548 static inline int pte_devmap(pte_t pte)
1554 int vma_wants_writenotify(struct vm_area_struct *vma);
1556 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1558 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1562 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1566 #ifdef __PAGETABLE_PUD_FOLDED
1567 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1568 unsigned long address)
1573 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1576 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1577 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1578 unsigned long address)
1583 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1585 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1590 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1591 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1594 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1596 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1598 atomic_long_set(&mm->nr_pmds, 0);
1601 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1603 return atomic_long_read(&mm->nr_pmds);
1606 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1608 atomic_long_inc(&mm->nr_pmds);
1611 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1613 atomic_long_dec(&mm->nr_pmds);
1617 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1618 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1621 * The following ifdef needed to get the 4level-fixup.h header to work.
1622 * Remove it when 4level-fixup.h has been removed.
1624 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1625 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1627 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1628 NULL: pud_offset(pgd, address);
1631 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1633 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1634 NULL: pmd_offset(pud, address);
1636 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1638 #if USE_SPLIT_PTE_PTLOCKS
1639 #if ALLOC_SPLIT_PTLOCKS
1640 void __init ptlock_cache_init(void);
1641 extern bool ptlock_alloc(struct page *page);
1642 extern void ptlock_free(struct page *page);
1644 static inline spinlock_t *ptlock_ptr(struct page *page)
1648 #else /* ALLOC_SPLIT_PTLOCKS */
1649 static inline void ptlock_cache_init(void)
1653 static inline bool ptlock_alloc(struct page *page)
1658 static inline void ptlock_free(struct page *page)
1662 static inline spinlock_t *ptlock_ptr(struct page *page)
1666 #endif /* ALLOC_SPLIT_PTLOCKS */
1668 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1670 return ptlock_ptr(pmd_page(*pmd));
1673 static inline bool ptlock_init(struct page *page)
1676 * prep_new_page() initialize page->private (and therefore page->ptl)
1677 * with 0. Make sure nobody took it in use in between.
1679 * It can happen if arch try to use slab for page table allocation:
1680 * slab code uses page->slab_cache, which share storage with page->ptl.
1682 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1683 if (!ptlock_alloc(page))
1685 spin_lock_init(ptlock_ptr(page));
1689 /* Reset page->mapping so free_pages_check won't complain. */
1690 static inline void pte_lock_deinit(struct page *page)
1692 page->mapping = NULL;
1696 #else /* !USE_SPLIT_PTE_PTLOCKS */
1698 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1700 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1702 return &mm->page_table_lock;
1704 static inline void ptlock_cache_init(void) {}
1705 static inline bool ptlock_init(struct page *page) { return true; }
1706 static inline void pte_lock_deinit(struct page *page) {}
1707 #endif /* USE_SPLIT_PTE_PTLOCKS */
1709 static inline void pgtable_init(void)
1711 ptlock_cache_init();
1712 pgtable_cache_init();
1715 static inline bool pgtable_page_ctor(struct page *page)
1717 if (!ptlock_init(page))
1719 inc_zone_page_state(page, NR_PAGETABLE);
1723 static inline void pgtable_page_dtor(struct page *page)
1725 pte_lock_deinit(page);
1726 dec_zone_page_state(page, NR_PAGETABLE);
1729 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1731 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1732 pte_t *__pte = pte_offset_map(pmd, address); \
1738 #define pte_unmap_unlock(pte, ptl) do { \
1743 #define pte_alloc(mm, pmd, address) \
1744 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1746 #define pte_alloc_map(mm, pmd, address) \
1747 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1749 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1750 (pte_alloc(mm, pmd, address) ? \
1751 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1753 #define pte_alloc_kernel(pmd, address) \
1754 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1755 NULL: pte_offset_kernel(pmd, address))
1757 #if USE_SPLIT_PMD_PTLOCKS
1759 static struct page *pmd_to_page(pmd_t *pmd)
1761 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1762 return virt_to_page((void *)((unsigned long) pmd & mask));
1765 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1767 return ptlock_ptr(pmd_to_page(pmd));
1770 static inline bool pgtable_pmd_page_ctor(struct page *page)
1772 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1773 page->pmd_huge_pte = NULL;
1775 return ptlock_init(page);
1778 static inline void pgtable_pmd_page_dtor(struct page *page)
1780 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1781 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1786 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1790 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1792 return &mm->page_table_lock;
1795 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1796 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1798 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1802 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1804 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1809 extern void free_area_init(unsigned long * zones_size);
1810 extern void free_area_init_node(int nid, unsigned long * zones_size,
1811 unsigned long zone_start_pfn, unsigned long *zholes_size);
1812 extern void free_initmem(void);
1815 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1816 * into the buddy system. The freed pages will be poisoned with pattern
1817 * "poison" if it's within range [0, UCHAR_MAX].
1818 * Return pages freed into the buddy system.
1820 extern unsigned long free_reserved_area(void *start, void *end,
1821 int poison, char *s);
1823 #ifdef CONFIG_HIGHMEM
1825 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1826 * and totalram_pages.
1828 extern void free_highmem_page(struct page *page);
1831 extern void adjust_managed_page_count(struct page *page, long count);
1832 extern void mem_init_print_info(const char *str);
1834 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1836 /* Free the reserved page into the buddy system, so it gets managed. */
1837 static inline void __free_reserved_page(struct page *page)
1839 ClearPageReserved(page);
1840 init_page_count(page);
1844 static inline void free_reserved_page(struct page *page)
1846 __free_reserved_page(page);
1847 adjust_managed_page_count(page, 1);
1850 static inline void mark_page_reserved(struct page *page)
1852 SetPageReserved(page);
1853 adjust_managed_page_count(page, -1);
1857 * Default method to free all the __init memory into the buddy system.
1858 * The freed pages will be poisoned with pattern "poison" if it's within
1859 * range [0, UCHAR_MAX].
1860 * Return pages freed into the buddy system.
1862 static inline unsigned long free_initmem_default(int poison)
1864 extern char __init_begin[], __init_end[];
1866 return free_reserved_area(&__init_begin, &__init_end,
1867 poison, "unused kernel");
1870 static inline unsigned long get_num_physpages(void)
1873 unsigned long phys_pages = 0;
1875 for_each_online_node(nid)
1876 phys_pages += node_present_pages(nid);
1881 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1883 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1884 * zones, allocate the backing mem_map and account for memory holes in a more
1885 * architecture independent manner. This is a substitute for creating the
1886 * zone_sizes[] and zholes_size[] arrays and passing them to
1887 * free_area_init_node()
1889 * An architecture is expected to register range of page frames backed by
1890 * physical memory with memblock_add[_node]() before calling
1891 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1892 * usage, an architecture is expected to do something like
1894 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1896 * for_each_valid_physical_page_range()
1897 * memblock_add_node(base, size, nid)
1898 * free_area_init_nodes(max_zone_pfns);
1900 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1901 * registered physical page range. Similarly
1902 * sparse_memory_present_with_active_regions() calls memory_present() for
1903 * each range when SPARSEMEM is enabled.
1905 * See mm/page_alloc.c for more information on each function exposed by
1906 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1908 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1909 unsigned long node_map_pfn_alignment(void);
1910 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1911 unsigned long end_pfn);
1912 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1913 unsigned long end_pfn);
1914 extern void get_pfn_range_for_nid(unsigned int nid,
1915 unsigned long *start_pfn, unsigned long *end_pfn);
1916 extern unsigned long find_min_pfn_with_active_regions(void);
1917 extern void free_bootmem_with_active_regions(int nid,
1918 unsigned long max_low_pfn);
1919 extern void sparse_memory_present_with_active_regions(int nid);
1921 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1923 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1924 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1925 static inline int __early_pfn_to_nid(unsigned long pfn,
1926 struct mminit_pfnnid_cache *state)
1931 /* please see mm/page_alloc.c */
1932 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1933 /* there is a per-arch backend function. */
1934 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1935 struct mminit_pfnnid_cache *state);
1938 extern void set_dma_reserve(unsigned long new_dma_reserve);
1939 extern void memmap_init_zone(unsigned long, int, unsigned long,
1940 unsigned long, enum memmap_context);
1941 extern void setup_per_zone_wmarks(void);
1942 extern int __meminit init_per_zone_wmark_min(void);
1943 extern void mem_init(void);
1944 extern void __init mmap_init(void);
1945 extern void show_mem(unsigned int flags);
1946 extern long si_mem_available(void);
1947 extern void si_meminfo(struct sysinfo * val);
1948 extern void si_meminfo_node(struct sysinfo *val, int nid);
1950 extern __printf(3, 4)
1951 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1952 const char *fmt, ...);
1954 extern void setup_per_cpu_pageset(void);
1956 extern void zone_pcp_update(struct zone *zone);
1957 extern void zone_pcp_reset(struct zone *zone);
1960 extern int min_free_kbytes;
1961 extern int watermark_scale_factor;
1964 extern atomic_long_t mmap_pages_allocated;
1965 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1967 /* interval_tree.c */
1968 void vma_interval_tree_insert(struct vm_area_struct *node,
1969 struct rb_root *root);
1970 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1971 struct vm_area_struct *prev,
1972 struct rb_root *root);
1973 void vma_interval_tree_remove(struct vm_area_struct *node,
1974 struct rb_root *root);
1975 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1976 unsigned long start, unsigned long last);
1977 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1978 unsigned long start, unsigned long last);
1980 #define vma_interval_tree_foreach(vma, root, start, last) \
1981 for (vma = vma_interval_tree_iter_first(root, start, last); \
1982 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1984 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1985 struct rb_root *root);
1986 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1987 struct rb_root *root);
1988 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1989 struct rb_root *root, unsigned long start, unsigned long last);
1990 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1991 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1992 #ifdef CONFIG_DEBUG_VM_RB
1993 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1996 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1997 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1998 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2001 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
2002 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
2003 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
2004 extern struct vm_area_struct *vma_merge(struct mm_struct *,
2005 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
2006 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
2007 struct mempolicy *, struct vm_userfaultfd_ctx);
2008 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
2009 extern int split_vma(struct mm_struct *,
2010 struct vm_area_struct *, unsigned long addr, int new_below);
2011 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
2012 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
2013 struct rb_node **, struct rb_node *);
2014 extern void unlink_file_vma(struct vm_area_struct *);
2015 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
2016 unsigned long addr, unsigned long len, pgoff_t pgoff,
2017 bool *need_rmap_locks);
2018 extern void exit_mmap(struct mm_struct *);
2020 static inline int check_data_rlimit(unsigned long rlim,
2022 unsigned long start,
2023 unsigned long end_data,
2024 unsigned long start_data)
2026 if (rlim < RLIM_INFINITY) {
2027 if (((new - start) + (end_data - start_data)) > rlim)
2034 extern int mm_take_all_locks(struct mm_struct *mm);
2035 extern void mm_drop_all_locks(struct mm_struct *mm);
2037 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
2038 extern struct file *get_mm_exe_file(struct mm_struct *mm);
2040 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
2041 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
2043 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
2044 unsigned long addr, unsigned long len,
2045 unsigned long flags,
2046 const struct vm_special_mapping *spec);
2047 /* This is an obsolete alternative to _install_special_mapping. */
2048 extern int install_special_mapping(struct mm_struct *mm,
2049 unsigned long addr, unsigned long len,
2050 unsigned long flags, struct page **pages);
2052 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
2054 extern unsigned long mmap_region(struct file *file, unsigned long addr,
2055 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
2056 extern unsigned long do_mmap(struct file *file, unsigned long addr,
2057 unsigned long len, unsigned long prot, unsigned long flags,
2058 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
2059 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
2061 static inline unsigned long
2062 do_mmap_pgoff(struct file *file, unsigned long addr,
2063 unsigned long len, unsigned long prot, unsigned long flags,
2064 unsigned long pgoff, unsigned long *populate)
2066 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
2070 extern int __mm_populate(unsigned long addr, unsigned long len,
2072 static inline void mm_populate(unsigned long addr, unsigned long len)
2075 (void) __mm_populate(addr, len, 1);
2078 static inline void mm_populate(unsigned long addr, unsigned long len) {}
2081 /* These take the mm semaphore themselves */
2082 extern unsigned long vm_brk(unsigned long, unsigned long);
2083 extern int vm_munmap(unsigned long, size_t);
2084 extern unsigned long vm_mmap(struct file *, unsigned long,
2085 unsigned long, unsigned long,
2086 unsigned long, unsigned long);
2088 struct vm_unmapped_area_info {
2089 #define VM_UNMAPPED_AREA_TOPDOWN 1
2090 unsigned long flags;
2091 unsigned long length;
2092 unsigned long low_limit;
2093 unsigned long high_limit;
2094 unsigned long align_mask;
2095 unsigned long align_offset;
2098 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2099 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2102 * Search for an unmapped address range.
2104 * We are looking for a range that:
2105 * - does not intersect with any VMA;
2106 * - is contained within the [low_limit, high_limit) interval;
2107 * - is at least the desired size.
2108 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2110 static inline unsigned long
2111 vm_unmapped_area(struct vm_unmapped_area_info *info)
2113 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2114 return unmapped_area_topdown(info);
2116 return unmapped_area(info);
2120 extern void truncate_inode_pages(struct address_space *, loff_t);
2121 extern void truncate_inode_pages_range(struct address_space *,
2122 loff_t lstart, loff_t lend);
2123 extern void truncate_inode_pages_final(struct address_space *);
2125 /* generic vm_area_ops exported for stackable file systems */
2126 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2127 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2128 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2130 /* mm/page-writeback.c */
2131 int write_one_page(struct page *page, int wait);
2132 void task_dirty_inc(struct task_struct *tsk);
2135 #define VM_MAX_READAHEAD 128 /* kbytes */
2136 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2138 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2139 pgoff_t offset, unsigned long nr_to_read);
2141 void page_cache_sync_readahead(struct address_space *mapping,
2142 struct file_ra_state *ra,
2145 unsigned long size);
2147 void page_cache_async_readahead(struct address_space *mapping,
2148 struct file_ra_state *ra,
2152 unsigned long size);
2154 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2155 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2157 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2158 extern int expand_downwards(struct vm_area_struct *vma,
2159 unsigned long address);
2161 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2163 #define expand_upwards(vma, address) (0)
2166 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2167 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2168 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2169 struct vm_area_struct **pprev);
2171 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2172 NULL if none. Assume start_addr < end_addr. */
2173 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2175 struct vm_area_struct * vma = find_vma(mm,start_addr);
2177 if (vma && end_addr <= vma->vm_start)
2182 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2184 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2187 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2188 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2189 unsigned long vm_start, unsigned long vm_end)
2191 struct vm_area_struct *vma = find_vma(mm, vm_start);
2193 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2200 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2201 void vma_set_page_prot(struct vm_area_struct *vma);
2203 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2207 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2209 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2213 #ifdef CONFIG_NUMA_BALANCING
2214 unsigned long change_prot_numa(struct vm_area_struct *vma,
2215 unsigned long start, unsigned long end);
2218 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2219 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2220 unsigned long pfn, unsigned long size, pgprot_t);
2221 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2222 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2224 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2225 unsigned long pfn, pgprot_t pgprot);
2226 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2228 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2231 struct page *follow_page_mask(struct vm_area_struct *vma,
2232 unsigned long address, unsigned int foll_flags,
2233 unsigned int *page_mask);
2235 static inline struct page *follow_page(struct vm_area_struct *vma,
2236 unsigned long address, unsigned int foll_flags)
2238 unsigned int unused_page_mask;
2239 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2242 #define FOLL_WRITE 0x01 /* check pte is writable */
2243 #define FOLL_TOUCH 0x02 /* mark page accessed */
2244 #define FOLL_GET 0x04 /* do get_page on page */
2245 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2246 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2247 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2248 * and return without waiting upon it */
2249 #define FOLL_POPULATE 0x40 /* fault in page */
2250 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2251 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2252 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2253 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2254 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2255 #define FOLL_MLOCK 0x1000 /* lock present pages */
2256 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2258 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2260 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2261 unsigned long size, pte_fn_t fn, void *data);
2264 #ifdef CONFIG_PAGE_POISONING
2265 extern bool page_poisoning_enabled(void);
2266 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2267 extern bool page_is_poisoned(struct page *page);
2269 static inline bool page_poisoning_enabled(void) { return false; }
2270 static inline void kernel_poison_pages(struct page *page, int numpages,
2272 static inline bool page_is_poisoned(struct page *page) { return false; }
2275 #ifdef CONFIG_DEBUG_PAGEALLOC
2276 extern bool _debug_pagealloc_enabled;
2277 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2279 static inline bool debug_pagealloc_enabled(void)
2281 return _debug_pagealloc_enabled;
2285 kernel_map_pages(struct page *page, int numpages, int enable)
2287 if (!debug_pagealloc_enabled())
2290 __kernel_map_pages(page, numpages, enable);
2292 #ifdef CONFIG_HIBERNATION
2293 extern bool kernel_page_present(struct page *page);
2294 #endif /* CONFIG_HIBERNATION */
2295 #else /* CONFIG_DEBUG_PAGEALLOC */
2297 kernel_map_pages(struct page *page, int numpages, int enable) {}
2298 #ifdef CONFIG_HIBERNATION
2299 static inline bool kernel_page_present(struct page *page) { return true; }
2300 #endif /* CONFIG_HIBERNATION */
2301 static inline bool debug_pagealloc_enabled(void)
2305 #endif /* CONFIG_DEBUG_PAGEALLOC */
2307 #ifdef __HAVE_ARCH_GATE_AREA
2308 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2309 extern int in_gate_area_no_mm(unsigned long addr);
2310 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2312 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2316 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2317 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2321 #endif /* __HAVE_ARCH_GATE_AREA */
2323 #ifdef CONFIG_SYSCTL
2324 extern int sysctl_drop_caches;
2325 int drop_caches_sysctl_handler(struct ctl_table *, int,
2326 void __user *, size_t *, loff_t *);
2329 void drop_slab(void);
2330 void drop_slab_node(int nid);
2333 #define randomize_va_space 0
2335 extern int randomize_va_space;
2338 const char * arch_vma_name(struct vm_area_struct *vma);
2339 void print_vma_addr(char *prefix, unsigned long rip);
2341 void sparse_mem_maps_populate_node(struct page **map_map,
2342 unsigned long pnum_begin,
2343 unsigned long pnum_end,
2344 unsigned long map_count,
2347 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2348 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2349 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2350 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2351 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2352 void *vmemmap_alloc_block(unsigned long size, int node);
2354 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2355 struct vmem_altmap *altmap);
2356 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2358 return __vmemmap_alloc_block_buf(size, node, NULL);
2361 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2362 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2364 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2365 void vmemmap_populate_print_last(void);
2366 #ifdef CONFIG_MEMORY_HOTPLUG
2367 void vmemmap_free(unsigned long start, unsigned long end);
2369 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2370 unsigned long size);
2373 MF_COUNT_INCREASED = 1 << 0,
2374 MF_ACTION_REQUIRED = 1 << 1,
2375 MF_MUST_KILL = 1 << 2,
2376 MF_SOFT_OFFLINE = 1 << 3,
2378 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2379 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2380 extern int unpoison_memory(unsigned long pfn);
2381 extern int get_hwpoison_page(struct page *page);
2382 #define put_hwpoison_page(page) put_page(page)
2383 extern int sysctl_memory_failure_early_kill;
2384 extern int sysctl_memory_failure_recovery;
2385 extern void shake_page(struct page *p, int access);
2386 extern atomic_long_t num_poisoned_pages;
2387 extern int soft_offline_page(struct page *page, int flags);
2391 * Error handlers for various types of pages.
2394 MF_IGNORED, /* Error: cannot be handled */
2395 MF_FAILED, /* Error: handling failed */
2396 MF_DELAYED, /* Will be handled later */
2397 MF_RECOVERED, /* Successfully recovered */
2400 enum mf_action_page_type {
2402 MF_MSG_KERNEL_HIGH_ORDER,
2404 MF_MSG_DIFFERENT_COMPOUND,
2405 MF_MSG_POISONED_HUGE,
2408 MF_MSG_UNMAP_FAILED,
2409 MF_MSG_DIRTY_SWAPCACHE,
2410 MF_MSG_CLEAN_SWAPCACHE,
2411 MF_MSG_DIRTY_MLOCKED_LRU,
2412 MF_MSG_CLEAN_MLOCKED_LRU,
2413 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2414 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2417 MF_MSG_TRUNCATED_LRU,
2423 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2424 extern void clear_huge_page(struct page *page,
2426 unsigned int pages_per_huge_page);
2427 extern void copy_user_huge_page(struct page *dst, struct page *src,
2428 unsigned long addr, struct vm_area_struct *vma,
2429 unsigned int pages_per_huge_page);
2430 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2432 extern struct page_ext_operations debug_guardpage_ops;
2433 extern struct page_ext_operations page_poisoning_ops;
2435 #ifdef CONFIG_DEBUG_PAGEALLOC
2436 extern unsigned int _debug_guardpage_minorder;
2437 extern bool _debug_guardpage_enabled;
2439 static inline unsigned int debug_guardpage_minorder(void)
2441 return _debug_guardpage_minorder;
2444 static inline bool debug_guardpage_enabled(void)
2446 return _debug_guardpage_enabled;
2449 static inline bool page_is_guard(struct page *page)
2451 struct page_ext *page_ext;
2453 if (!debug_guardpage_enabled())
2456 page_ext = lookup_page_ext(page);
2457 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2460 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2461 static inline bool debug_guardpage_enabled(void) { return false; }
2462 static inline bool page_is_guard(struct page *page) { return false; }
2463 #endif /* CONFIG_DEBUG_PAGEALLOC */
2465 #if MAX_NUMNODES > 1
2466 void __init setup_nr_node_ids(void);
2468 static inline void setup_nr_node_ids(void) {}
2471 #endif /* __KERNEL__ */
2472 #endif /* _LINUX_MM_H */