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 * To prevent common memory management code establishing
77 * a zero page mapping on a read fault.
78 * This macro should be defined within <asm/pgtable.h>.
79 * s390 does this to prevent multiplexing of hardware bits
80 * related to the physical page in case of virtualization.
82 #ifndef mm_forbids_zeropage
83 #define mm_forbids_zeropage(X) (0)
87 * Default maximum number of active map areas, this limits the number of vmas
88 * per mm struct. Users can overwrite this number by sysctl but there is a
91 * When a program's coredump is generated as ELF format, a section is created
92 * per a vma. In ELF, the number of sections is represented in unsigned short.
93 * This means the number of sections should be smaller than 65535 at coredump.
94 * Because the kernel adds some informative sections to a image of program at
95 * generating coredump, we need some margin. The number of extra sections is
96 * 1-3 now and depends on arch. We use "5" as safe margin, here.
98 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
99 * not a hard limit any more. Although some userspace tools can be surprised by
102 #define MAPCOUNT_ELF_CORE_MARGIN (5)
103 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
105 extern int sysctl_max_map_count;
107 extern unsigned long sysctl_user_reserve_kbytes;
108 extern unsigned long sysctl_admin_reserve_kbytes;
110 extern int sysctl_overcommit_memory;
111 extern int sysctl_overcommit_ratio;
112 extern unsigned long sysctl_overcommit_kbytes;
114 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
116 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
119 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
121 /* to align the pointer to the (next) page boundary */
122 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
124 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
125 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
128 * Linux kernel virtual memory manager primitives.
129 * The idea being to have a "virtual" mm in the same way
130 * we have a virtual fs - giving a cleaner interface to the
131 * mm details, and allowing different kinds of memory mappings
132 * (from shared memory to executable loading to arbitrary
136 extern struct kmem_cache *vm_area_cachep;
139 extern struct rb_root nommu_region_tree;
140 extern struct rw_semaphore nommu_region_sem;
142 extern unsigned int kobjsize(const void *objp);
146 * vm_flags in vm_area_struct, see mm_types.h.
147 * When changing, update also include/trace/events/mmflags.h
149 #define VM_NONE 0x00000000
151 #define VM_READ 0x00000001 /* currently active flags */
152 #define VM_WRITE 0x00000002
153 #define VM_EXEC 0x00000004
154 #define VM_SHARED 0x00000008
156 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
157 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
158 #define VM_MAYWRITE 0x00000020
159 #define VM_MAYEXEC 0x00000040
160 #define VM_MAYSHARE 0x00000080
162 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
163 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
164 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
165 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
166 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
168 #define VM_LOCKED 0x00002000
169 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
171 /* Used by sys_madvise() */
172 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
173 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
175 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
176 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
177 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
178 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
179 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
180 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
181 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
182 #define VM_ARCH_2 0x02000000
183 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
185 #ifdef CONFIG_MEM_SOFT_DIRTY
186 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
188 # define VM_SOFTDIRTY 0
191 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
192 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
193 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
194 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
196 #if defined(CONFIG_X86)
197 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
198 #elif defined(CONFIG_PPC)
199 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
200 #elif defined(CONFIG_PARISC)
201 # define VM_GROWSUP VM_ARCH_1
202 #elif defined(CONFIG_METAG)
203 # define VM_GROWSUP VM_ARCH_1
204 #elif defined(CONFIG_IA64)
205 # define VM_GROWSUP VM_ARCH_1
206 #elif !defined(CONFIG_MMU)
207 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
210 #if defined(CONFIG_X86)
211 /* MPX specific bounds table or bounds directory */
212 # define VM_MPX VM_ARCH_2
216 # define VM_GROWSUP VM_NONE
219 /* Bits set in the VMA until the stack is in its final location */
220 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
222 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
223 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
226 #ifdef CONFIG_STACK_GROWSUP
227 #define VM_STACK VM_GROWSUP
229 #define VM_STACK VM_GROWSDOWN
232 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
235 * Special vmas that are non-mergable, non-mlock()able.
236 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
238 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
240 /* This mask defines which mm->def_flags a process can inherit its parent */
241 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
243 /* This mask is used to clear all the VMA flags used by mlock */
244 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
247 * mapping from the currently active vm_flags protection bits (the
248 * low four bits) to a page protection mask..
250 extern pgprot_t protection_map[16];
252 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
253 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
254 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
255 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
256 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
257 #define FAULT_FLAG_TRIED 0x20 /* Second try */
258 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
261 * vm_fault is filled by the the pagefault handler and passed to the vma's
262 * ->fault function. The vma's ->fault is responsible for returning a bitmask
263 * of VM_FAULT_xxx flags that give details about how the fault was handled.
265 * MM layer fills up gfp_mask for page allocations but fault handler might
266 * alter it if its implementation requires a different allocation context.
268 * pgoff should be used in favour of virtual_address, if possible.
271 unsigned int flags; /* FAULT_FLAG_xxx flags */
272 gfp_t gfp_mask; /* gfp mask to be used for allocations */
273 pgoff_t pgoff; /* Logical page offset based on vma */
274 void __user *virtual_address; /* Faulting virtual address */
276 struct page *cow_page; /* Handler may choose to COW */
277 struct page *page; /* ->fault handlers should return a
278 * page here, unless VM_FAULT_NOPAGE
279 * is set (which is also implied by
282 /* for ->map_pages() only */
283 pgoff_t max_pgoff; /* map pages for offset from pgoff till
284 * max_pgoff inclusive */
285 pte_t *pte; /* pte entry associated with ->pgoff */
289 * These are the virtual MM functions - opening of an area, closing and
290 * unmapping it (needed to keep files on disk up-to-date etc), pointer
291 * to the functions called when a no-page or a wp-page exception occurs.
293 struct vm_operations_struct {
294 void (*open)(struct vm_area_struct * area);
295 void (*close)(struct vm_area_struct * area);
296 int (*mremap)(struct vm_area_struct * area);
297 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
298 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
299 pmd_t *, unsigned int flags);
300 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
302 /* notification that a previously read-only page is about to become
303 * writable, if an error is returned it will cause a SIGBUS */
304 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
306 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
307 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
309 /* called by access_process_vm when get_user_pages() fails, typically
310 * for use by special VMAs that can switch between memory and hardware
312 int (*access)(struct vm_area_struct *vma, unsigned long addr,
313 void *buf, int len, int write);
315 /* Called by the /proc/PID/maps code to ask the vma whether it
316 * has a special name. Returning non-NULL will also cause this
317 * vma to be dumped unconditionally. */
318 const char *(*name)(struct vm_area_struct *vma);
322 * set_policy() op must add a reference to any non-NULL @new mempolicy
323 * to hold the policy upon return. Caller should pass NULL @new to
324 * remove a policy and fall back to surrounding context--i.e. do not
325 * install a MPOL_DEFAULT policy, nor the task or system default
328 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
331 * get_policy() op must add reference [mpol_get()] to any policy at
332 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
333 * in mm/mempolicy.c will do this automatically.
334 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
335 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
336 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
337 * must return NULL--i.e., do not "fallback" to task or system default
340 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
344 * Called by vm_normal_page() for special PTEs to find the
345 * page for @addr. This is useful if the default behavior
346 * (using pte_page()) would not find the correct page.
348 struct page *(*find_special_page)(struct vm_area_struct *vma,
355 #define page_private(page) ((page)->private)
356 #define set_page_private(page, v) ((page)->private = (v))
358 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
359 static inline int pmd_devmap(pmd_t pmd)
366 * FIXME: take this include out, include page-flags.h in
367 * files which need it (119 of them)
369 #include <linux/page-flags.h>
370 #include <linux/huge_mm.h>
373 * Methods to modify the page usage count.
375 * What counts for a page usage:
376 * - cache mapping (page->mapping)
377 * - private data (page->private)
378 * - page mapped in a task's page tables, each mapping
379 * is counted separately
381 * Also, many kernel routines increase the page count before a critical
382 * routine so they can be sure the page doesn't go away from under them.
386 * Drop a ref, return true if the refcount fell to zero (the page has no users)
388 static inline int put_page_testzero(struct page *page)
390 VM_BUG_ON_PAGE(page_ref_count(page) == 0, page);
391 return page_ref_dec_and_test(page);
395 * Try to grab a ref unless the page has a refcount of zero, return false if
397 * This can be called when MMU is off so it must not access
398 * any of the virtual mappings.
400 static inline int get_page_unless_zero(struct page *page)
402 return page_ref_add_unless(page, 1, 0);
405 extern int page_is_ram(unsigned long pfn);
413 int region_intersects(resource_size_t offset, size_t size, unsigned long flags,
416 /* Support for virtually mapped pages */
417 struct page *vmalloc_to_page(const void *addr);
418 unsigned long vmalloc_to_pfn(const void *addr);
421 * Determine if an address is within the vmalloc range
423 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
424 * is no special casing required.
426 static inline int is_vmalloc_addr(const void *x)
429 unsigned long addr = (unsigned long)x;
431 return addr >= VMALLOC_START && addr < VMALLOC_END;
437 extern int is_vmalloc_or_module_addr(const void *x);
439 static inline int is_vmalloc_or_module_addr(const void *x)
445 extern void kvfree(const void *addr);
447 static inline atomic_t *compound_mapcount_ptr(struct page *page)
449 return &page[1].compound_mapcount;
452 static inline int compound_mapcount(struct page *page)
454 if (!PageCompound(page))
456 page = compound_head(page);
457 return atomic_read(compound_mapcount_ptr(page)) + 1;
461 * The atomic page->_mapcount, starts from -1: so that transitions
462 * both from it and to it can be tracked, using atomic_inc_and_test
463 * and atomic_add_negative(-1).
465 static inline void page_mapcount_reset(struct page *page)
467 atomic_set(&(page)->_mapcount, -1);
470 int __page_mapcount(struct page *page);
472 static inline int page_mapcount(struct page *page)
474 VM_BUG_ON_PAGE(PageSlab(page), page);
476 if (unlikely(PageCompound(page)))
477 return __page_mapcount(page);
478 return atomic_read(&page->_mapcount) + 1;
481 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
482 int total_mapcount(struct page *page);
484 static inline int total_mapcount(struct page *page)
486 return page_mapcount(page);
490 static inline struct page *virt_to_head_page(const void *x)
492 struct page *page = virt_to_page(x);
494 return compound_head(page);
497 void __put_page(struct page *page);
499 void put_pages_list(struct list_head *pages);
501 void split_page(struct page *page, unsigned int order);
502 int split_free_page(struct page *page);
505 * Compound pages have a destructor function. Provide a
506 * prototype for that function and accessor functions.
507 * These are _only_ valid on the head of a compound page.
509 typedef void compound_page_dtor(struct page *);
511 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
512 enum compound_dtor_id {
515 #ifdef CONFIG_HUGETLB_PAGE
518 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
523 extern compound_page_dtor * const compound_page_dtors[];
525 static inline void set_compound_page_dtor(struct page *page,
526 enum compound_dtor_id compound_dtor)
528 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
529 page[1].compound_dtor = compound_dtor;
532 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
534 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
535 return compound_page_dtors[page[1].compound_dtor];
538 static inline unsigned int compound_order(struct page *page)
542 return page[1].compound_order;
545 static inline void set_compound_order(struct page *page, unsigned int order)
547 page[1].compound_order = order;
550 void free_compound_page(struct page *page);
554 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
555 * servicing faults for write access. In the normal case, do always want
556 * pte_mkwrite. But get_user_pages can cause write faults for mappings
557 * that do not have writing enabled, when used by access_process_vm.
559 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
561 if (likely(vma->vm_flags & VM_WRITE))
562 pte = pte_mkwrite(pte);
566 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
567 struct page *page, pte_t *pte, bool write, bool anon);
571 * Multiple processes may "see" the same page. E.g. for untouched
572 * mappings of /dev/null, all processes see the same page full of
573 * zeroes, and text pages of executables and shared libraries have
574 * only one copy in memory, at most, normally.
576 * For the non-reserved pages, page_count(page) denotes a reference count.
577 * page_count() == 0 means the page is free. page->lru is then used for
578 * freelist management in the buddy allocator.
579 * page_count() > 0 means the page has been allocated.
581 * Pages are allocated by the slab allocator in order to provide memory
582 * to kmalloc and kmem_cache_alloc. In this case, the management of the
583 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
584 * unless a particular usage is carefully commented. (the responsibility of
585 * freeing the kmalloc memory is the caller's, of course).
587 * A page may be used by anyone else who does a __get_free_page().
588 * In this case, page_count still tracks the references, and should only
589 * be used through the normal accessor functions. The top bits of page->flags
590 * and page->virtual store page management information, but all other fields
591 * are unused and could be used privately, carefully. The management of this
592 * page is the responsibility of the one who allocated it, and those who have
593 * subsequently been given references to it.
595 * The other pages (we may call them "pagecache pages") are completely
596 * managed by the Linux memory manager: I/O, buffers, swapping etc.
597 * The following discussion applies only to them.
599 * A pagecache page contains an opaque `private' member, which belongs to the
600 * page's address_space. Usually, this is the address of a circular list of
601 * the page's disk buffers. PG_private must be set to tell the VM to call
602 * into the filesystem to release these pages.
604 * A page may belong to an inode's memory mapping. In this case, page->mapping
605 * is the pointer to the inode, and page->index is the file offset of the page,
606 * in units of PAGE_CACHE_SIZE.
608 * If pagecache pages are not associated with an inode, they are said to be
609 * anonymous pages. These may become associated with the swapcache, and in that
610 * case PG_swapcache is set, and page->private is an offset into the swapcache.
612 * In either case (swapcache or inode backed), the pagecache itself holds one
613 * reference to the page. Setting PG_private should also increment the
614 * refcount. The each user mapping also has a reference to the page.
616 * The pagecache pages are stored in a per-mapping radix tree, which is
617 * rooted at mapping->page_tree, and indexed by offset.
618 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
619 * lists, we instead now tag pages as dirty/writeback in the radix tree.
621 * All pagecache pages may be subject to I/O:
622 * - inode pages may need to be read from disk,
623 * - inode pages which have been modified and are MAP_SHARED may need
624 * to be written back to the inode on disk,
625 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
626 * modified may need to be swapped out to swap space and (later) to be read
631 * The zone field is never updated after free_area_init_core()
632 * sets it, so none of the operations on it need to be atomic.
635 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
636 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
637 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
638 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
639 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
642 * Define the bit shifts to access each section. For non-existent
643 * sections we define the shift as 0; that plus a 0 mask ensures
644 * the compiler will optimise away reference to them.
646 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
647 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
648 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
649 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
651 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
652 #ifdef NODE_NOT_IN_PAGE_FLAGS
653 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
654 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
655 SECTIONS_PGOFF : ZONES_PGOFF)
657 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
658 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
659 NODES_PGOFF : ZONES_PGOFF)
662 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
664 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
665 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
668 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
669 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
670 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
671 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
672 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
674 static inline enum zone_type page_zonenum(const struct page *page)
676 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
679 #ifdef CONFIG_ZONE_DEVICE
680 void get_zone_device_page(struct page *page);
681 void put_zone_device_page(struct page *page);
682 static inline bool is_zone_device_page(const struct page *page)
684 return page_zonenum(page) == ZONE_DEVICE;
687 static inline void get_zone_device_page(struct page *page)
690 static inline void put_zone_device_page(struct page *page)
693 static inline bool is_zone_device_page(const struct page *page)
699 static inline void get_page(struct page *page)
701 page = compound_head(page);
703 * Getting a normal page or the head of a compound page
704 * requires to already have an elevated page->_count.
706 VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page);
709 if (unlikely(is_zone_device_page(page)))
710 get_zone_device_page(page);
713 static inline void put_page(struct page *page)
715 page = compound_head(page);
717 if (put_page_testzero(page))
720 if (unlikely(is_zone_device_page(page)))
721 put_zone_device_page(page);
724 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
725 #define SECTION_IN_PAGE_FLAGS
729 * The identification function is mainly used by the buddy allocator for
730 * determining if two pages could be buddies. We are not really identifying
731 * the zone since we could be using the section number id if we do not have
732 * node id available in page flags.
733 * We only guarantee that it will return the same value for two combinable
736 static inline int page_zone_id(struct page *page)
738 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
741 static inline int zone_to_nid(struct zone *zone)
750 #ifdef NODE_NOT_IN_PAGE_FLAGS
751 extern int page_to_nid(const struct page *page);
753 static inline int page_to_nid(const struct page *page)
755 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
759 #ifdef CONFIG_NUMA_BALANCING
760 static inline int cpu_pid_to_cpupid(int cpu, int pid)
762 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
765 static inline int cpupid_to_pid(int cpupid)
767 return cpupid & LAST__PID_MASK;
770 static inline int cpupid_to_cpu(int cpupid)
772 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
775 static inline int cpupid_to_nid(int cpupid)
777 return cpu_to_node(cpupid_to_cpu(cpupid));
780 static inline bool cpupid_pid_unset(int cpupid)
782 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
785 static inline bool cpupid_cpu_unset(int cpupid)
787 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
790 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
792 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
795 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
796 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
797 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
799 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
802 static inline int page_cpupid_last(struct page *page)
804 return page->_last_cpupid;
806 static inline void page_cpupid_reset_last(struct page *page)
808 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
811 static inline int page_cpupid_last(struct page *page)
813 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
816 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
818 static inline void page_cpupid_reset_last(struct page *page)
820 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
822 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
823 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
825 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
826 #else /* !CONFIG_NUMA_BALANCING */
827 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
829 return page_to_nid(page); /* XXX */
832 static inline int page_cpupid_last(struct page *page)
834 return page_to_nid(page); /* XXX */
837 static inline int cpupid_to_nid(int cpupid)
842 static inline int cpupid_to_pid(int cpupid)
847 static inline int cpupid_to_cpu(int cpupid)
852 static inline int cpu_pid_to_cpupid(int nid, int pid)
857 static inline bool cpupid_pid_unset(int cpupid)
862 static inline void page_cpupid_reset_last(struct page *page)
866 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
870 #endif /* CONFIG_NUMA_BALANCING */
872 static inline struct zone *page_zone(const struct page *page)
874 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
877 #ifdef SECTION_IN_PAGE_FLAGS
878 static inline void set_page_section(struct page *page, unsigned long section)
880 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
881 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
884 static inline unsigned long page_to_section(const struct page *page)
886 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
890 static inline void set_page_zone(struct page *page, enum zone_type zone)
892 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
893 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
896 static inline void set_page_node(struct page *page, unsigned long node)
898 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
899 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
902 static inline void set_page_links(struct page *page, enum zone_type zone,
903 unsigned long node, unsigned long pfn)
905 set_page_zone(page, zone);
906 set_page_node(page, node);
907 #ifdef SECTION_IN_PAGE_FLAGS
908 set_page_section(page, pfn_to_section_nr(pfn));
913 static inline struct mem_cgroup *page_memcg(struct page *page)
915 return page->mem_cgroup;
918 static inline struct mem_cgroup *page_memcg(struct page *page)
925 * Some inline functions in vmstat.h depend on page_zone()
927 #include <linux/vmstat.h>
929 static __always_inline void *lowmem_page_address(const struct page *page)
931 return __va(PFN_PHYS(page_to_pfn(page)));
934 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
935 #define HASHED_PAGE_VIRTUAL
938 #if defined(WANT_PAGE_VIRTUAL)
939 static inline void *page_address(const struct page *page)
941 return page->virtual;
943 static inline void set_page_address(struct page *page, void *address)
945 page->virtual = address;
947 #define page_address_init() do { } while(0)
950 #if defined(HASHED_PAGE_VIRTUAL)
951 void *page_address(const struct page *page);
952 void set_page_address(struct page *page, void *virtual);
953 void page_address_init(void);
956 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
957 #define page_address(page) lowmem_page_address(page)
958 #define set_page_address(page, address) do { } while(0)
959 #define page_address_init() do { } while(0)
962 extern void *page_rmapping(struct page *page);
963 extern struct anon_vma *page_anon_vma(struct page *page);
964 extern struct address_space *page_mapping(struct page *page);
966 extern struct address_space *__page_file_mapping(struct page *);
969 struct address_space *page_file_mapping(struct page *page)
971 if (unlikely(PageSwapCache(page)))
972 return __page_file_mapping(page);
974 return page->mapping;
978 * Return the pagecache index of the passed page. Regular pagecache pages
979 * use ->index whereas swapcache pages use ->private
981 static inline pgoff_t page_index(struct page *page)
983 if (unlikely(PageSwapCache(page)))
984 return page_private(page);
988 extern pgoff_t __page_file_index(struct page *page);
991 * Return the file index of the page. Regular pagecache pages use ->index
992 * whereas swapcache pages use swp_offset(->private)
994 static inline pgoff_t page_file_index(struct page *page)
996 if (unlikely(PageSwapCache(page)))
997 return __page_file_index(page);
1003 * Return true if this page is mapped into pagetables.
1004 * For compound page it returns true if any subpage of compound page is mapped.
1006 static inline bool page_mapped(struct page *page)
1009 if (likely(!PageCompound(page)))
1010 return atomic_read(&page->_mapcount) >= 0;
1011 page = compound_head(page);
1012 if (atomic_read(compound_mapcount_ptr(page)) >= 0)
1014 for (i = 0; i < hpage_nr_pages(page); i++) {
1015 if (atomic_read(&page[i]._mapcount) >= 0)
1022 * Return true only if the page has been allocated with
1023 * ALLOC_NO_WATERMARKS and the low watermark was not
1024 * met implying that the system is under some pressure.
1026 static inline bool page_is_pfmemalloc(struct page *page)
1029 * Page index cannot be this large so this must be
1030 * a pfmemalloc page.
1032 return page->index == -1UL;
1036 * Only to be called by the page allocator on a freshly allocated
1039 static inline void set_page_pfmemalloc(struct page *page)
1044 static inline void clear_page_pfmemalloc(struct page *page)
1050 * Different kinds of faults, as returned by handle_mm_fault().
1051 * Used to decide whether a process gets delivered SIGBUS or
1052 * just gets major/minor fault counters bumped up.
1055 #define VM_FAULT_OOM 0x0001
1056 #define VM_FAULT_SIGBUS 0x0002
1057 #define VM_FAULT_MAJOR 0x0004
1058 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1059 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1060 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1061 #define VM_FAULT_SIGSEGV 0x0040
1063 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1064 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1065 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1066 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1068 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1070 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1071 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1074 /* Encode hstate index for a hwpoisoned large page */
1075 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1076 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1079 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1081 extern void pagefault_out_of_memory(void);
1083 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1086 * Flags passed to show_mem() and show_free_areas() to suppress output in
1089 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1091 extern void show_free_areas(unsigned int flags);
1092 extern bool skip_free_areas_node(unsigned int flags, int nid);
1094 int shmem_zero_setup(struct vm_area_struct *);
1096 bool shmem_mapping(struct address_space *mapping);
1098 static inline bool shmem_mapping(struct address_space *mapping)
1104 extern bool can_do_mlock(void);
1105 extern int user_shm_lock(size_t, struct user_struct *);
1106 extern void user_shm_unlock(size_t, struct user_struct *);
1109 * Parameter block passed down to zap_pte_range in exceptional cases.
1111 struct zap_details {
1112 struct address_space *check_mapping; /* Check page->mapping if set */
1113 pgoff_t first_index; /* Lowest page->index to unmap */
1114 pgoff_t last_index; /* Highest page->index to unmap */
1117 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1120 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1121 unsigned long size);
1122 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1123 unsigned long size, struct zap_details *);
1124 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1125 unsigned long start, unsigned long end);
1128 * mm_walk - callbacks for walk_page_range
1129 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1130 * this handler is required to be able to handle
1131 * pmd_trans_huge() pmds. They may simply choose to
1132 * split_huge_page() instead of handling it explicitly.
1133 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1134 * @pte_hole: if set, called for each hole at all levels
1135 * @hugetlb_entry: if set, called for each hugetlb entry
1136 * @test_walk: caller specific callback function to determine whether
1137 * we walk over the current vma or not. A positive returned
1138 * value means "do page table walk over the current vma,"
1139 * and a negative one means "abort current page table walk
1140 * right now." 0 means "skip the current vma."
1141 * @mm: mm_struct representing the target process of page table walk
1142 * @vma: vma currently walked (NULL if walking outside vmas)
1143 * @private: private data for callbacks' usage
1145 * (see the comment on walk_page_range() for more details)
1148 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1149 unsigned long next, struct mm_walk *walk);
1150 int (*pte_entry)(pte_t *pte, unsigned long addr,
1151 unsigned long next, struct mm_walk *walk);
1152 int (*pte_hole)(unsigned long addr, unsigned long next,
1153 struct mm_walk *walk);
1154 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1155 unsigned long addr, unsigned long next,
1156 struct mm_walk *walk);
1157 int (*test_walk)(unsigned long addr, unsigned long next,
1158 struct mm_walk *walk);
1159 struct mm_struct *mm;
1160 struct vm_area_struct *vma;
1164 int walk_page_range(unsigned long addr, unsigned long end,
1165 struct mm_walk *walk);
1166 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1167 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1168 unsigned long end, unsigned long floor, unsigned long ceiling);
1169 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1170 struct vm_area_struct *vma);
1171 void unmap_mapping_range(struct address_space *mapping,
1172 loff_t const holebegin, loff_t const holelen, int even_cows);
1173 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1174 unsigned long *pfn);
1175 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1176 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1177 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1178 void *buf, int len, int write);
1180 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1181 loff_t const holebegin, loff_t const holelen)
1183 unmap_mapping_range(mapping, holebegin, holelen, 0);
1186 extern void truncate_pagecache(struct inode *inode, loff_t new);
1187 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1188 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1189 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1190 int truncate_inode_page(struct address_space *mapping, struct page *page);
1191 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1192 int invalidate_inode_page(struct page *page);
1195 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1196 unsigned long address, unsigned int flags);
1197 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1198 unsigned long address, unsigned int fault_flags,
1201 static inline int handle_mm_fault(struct mm_struct *mm,
1202 struct vm_area_struct *vma, unsigned long address,
1205 /* should never happen if there's no MMU */
1207 return VM_FAULT_SIGBUS;
1209 static inline int fixup_user_fault(struct task_struct *tsk,
1210 struct mm_struct *mm, unsigned long address,
1211 unsigned int fault_flags, bool *unlocked)
1213 /* should never happen if there's no MMU */
1219 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1220 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1221 void *buf, int len, int write);
1223 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1224 unsigned long start, unsigned long nr_pages,
1225 unsigned int foll_flags, struct page **pages,
1226 struct vm_area_struct **vmas, int *nonblocking);
1227 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1228 unsigned long start, unsigned long nr_pages,
1229 int write, int force, struct page **pages,
1230 struct vm_area_struct **vmas);
1231 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1232 unsigned long start, unsigned long nr_pages,
1233 int write, int force, struct page **pages,
1235 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1236 unsigned long start, unsigned long nr_pages,
1237 int write, int force, struct page **pages,
1238 unsigned int gup_flags);
1239 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1240 unsigned long start, unsigned long nr_pages,
1241 int write, int force, struct page **pages);
1242 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1243 struct page **pages);
1245 /* Container for pinned pfns / pages */
1246 struct frame_vector {
1247 unsigned int nr_allocated; /* Number of frames we have space for */
1248 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1249 bool got_ref; /* Did we pin pages by getting page ref? */
1250 bool is_pfns; /* Does array contain pages or pfns? */
1251 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1252 * pfns_vector_pages() or pfns_vector_pfns()
1256 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1257 void frame_vector_destroy(struct frame_vector *vec);
1258 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1259 bool write, bool force, struct frame_vector *vec);
1260 void put_vaddr_frames(struct frame_vector *vec);
1261 int frame_vector_to_pages(struct frame_vector *vec);
1262 void frame_vector_to_pfns(struct frame_vector *vec);
1264 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1266 return vec->nr_frames;
1269 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1272 int err = frame_vector_to_pages(vec);
1275 return ERR_PTR(err);
1277 return (struct page **)(vec->ptrs);
1280 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1283 frame_vector_to_pfns(vec);
1284 return (unsigned long *)(vec->ptrs);
1288 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1289 struct page **pages);
1290 int get_kernel_page(unsigned long start, int write, struct page **pages);
1291 struct page *get_dump_page(unsigned long addr);
1293 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1294 extern void do_invalidatepage(struct page *page, unsigned int offset,
1295 unsigned int length);
1297 int __set_page_dirty_nobuffers(struct page *page);
1298 int __set_page_dirty_no_writeback(struct page *page);
1299 int redirty_page_for_writepage(struct writeback_control *wbc,
1301 void account_page_dirtied(struct page *page, struct address_space *mapping);
1302 void account_page_cleaned(struct page *page, struct address_space *mapping,
1303 struct bdi_writeback *wb);
1304 int set_page_dirty(struct page *page);
1305 int set_page_dirty_lock(struct page *page);
1306 void cancel_dirty_page(struct page *page);
1307 int clear_page_dirty_for_io(struct page *page);
1309 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1311 /* Is the vma a continuation of the stack vma above it? */
1312 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1314 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1317 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1319 return !vma->vm_ops;
1322 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1325 return (vma->vm_flags & VM_GROWSDOWN) &&
1326 (vma->vm_start == addr) &&
1327 !vma_growsdown(vma->vm_prev, addr);
1330 /* Is the vma a continuation of the stack vma below it? */
1331 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1333 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1336 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1339 return (vma->vm_flags & VM_GROWSUP) &&
1340 (vma->vm_end == addr) &&
1341 !vma_growsup(vma->vm_next, addr);
1344 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1346 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1347 unsigned long old_addr, struct vm_area_struct *new_vma,
1348 unsigned long new_addr, unsigned long len,
1349 bool need_rmap_locks);
1350 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1351 unsigned long end, pgprot_t newprot,
1352 int dirty_accountable, int prot_numa);
1353 extern int mprotect_fixup(struct vm_area_struct *vma,
1354 struct vm_area_struct **pprev, unsigned long start,
1355 unsigned long end, unsigned long newflags);
1358 * doesn't attempt to fault and will return short.
1360 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1361 struct page **pages);
1363 * per-process(per-mm_struct) statistics.
1365 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1367 long val = atomic_long_read(&mm->rss_stat.count[member]);
1369 #ifdef SPLIT_RSS_COUNTING
1371 * counter is updated in asynchronous manner and may go to minus.
1372 * But it's never be expected number for users.
1377 return (unsigned long)val;
1380 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1382 atomic_long_add(value, &mm->rss_stat.count[member]);
1385 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1387 atomic_long_inc(&mm->rss_stat.count[member]);
1390 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1392 atomic_long_dec(&mm->rss_stat.count[member]);
1395 /* Optimized variant when page is already known not to be PageAnon */
1396 static inline int mm_counter_file(struct page *page)
1398 if (PageSwapBacked(page))
1399 return MM_SHMEMPAGES;
1400 return MM_FILEPAGES;
1403 static inline int mm_counter(struct page *page)
1406 return MM_ANONPAGES;
1407 return mm_counter_file(page);
1410 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1412 return get_mm_counter(mm, MM_FILEPAGES) +
1413 get_mm_counter(mm, MM_ANONPAGES) +
1414 get_mm_counter(mm, MM_SHMEMPAGES);
1417 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1419 return max(mm->hiwater_rss, get_mm_rss(mm));
1422 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1424 return max(mm->hiwater_vm, mm->total_vm);
1427 static inline void update_hiwater_rss(struct mm_struct *mm)
1429 unsigned long _rss = get_mm_rss(mm);
1431 if ((mm)->hiwater_rss < _rss)
1432 (mm)->hiwater_rss = _rss;
1435 static inline void update_hiwater_vm(struct mm_struct *mm)
1437 if (mm->hiwater_vm < mm->total_vm)
1438 mm->hiwater_vm = mm->total_vm;
1441 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1443 mm->hiwater_rss = get_mm_rss(mm);
1446 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1447 struct mm_struct *mm)
1449 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1451 if (*maxrss < hiwater_rss)
1452 *maxrss = hiwater_rss;
1455 #if defined(SPLIT_RSS_COUNTING)
1456 void sync_mm_rss(struct mm_struct *mm);
1458 static inline void sync_mm_rss(struct mm_struct *mm)
1463 #ifndef __HAVE_ARCH_PTE_DEVMAP
1464 static inline int pte_devmap(pte_t pte)
1470 int vma_wants_writenotify(struct vm_area_struct *vma);
1472 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1474 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1478 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1482 #ifdef __PAGETABLE_PUD_FOLDED
1483 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1484 unsigned long address)
1489 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1492 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1493 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1494 unsigned long address)
1499 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1501 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1506 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1507 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1510 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1512 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1514 atomic_long_set(&mm->nr_pmds, 0);
1517 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1519 return atomic_long_read(&mm->nr_pmds);
1522 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1524 atomic_long_inc(&mm->nr_pmds);
1527 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1529 atomic_long_dec(&mm->nr_pmds);
1533 int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address);
1534 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1537 * The following ifdef needed to get the 4level-fixup.h header to work.
1538 * Remove it when 4level-fixup.h has been removed.
1540 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1541 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1543 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1544 NULL: pud_offset(pgd, address);
1547 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1549 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1550 NULL: pmd_offset(pud, address);
1552 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1554 #if USE_SPLIT_PTE_PTLOCKS
1555 #if ALLOC_SPLIT_PTLOCKS
1556 void __init ptlock_cache_init(void);
1557 extern bool ptlock_alloc(struct page *page);
1558 extern void ptlock_free(struct page *page);
1560 static inline spinlock_t *ptlock_ptr(struct page *page)
1564 #else /* ALLOC_SPLIT_PTLOCKS */
1565 static inline void ptlock_cache_init(void)
1569 static inline bool ptlock_alloc(struct page *page)
1574 static inline void ptlock_free(struct page *page)
1578 static inline spinlock_t *ptlock_ptr(struct page *page)
1582 #endif /* ALLOC_SPLIT_PTLOCKS */
1584 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1586 return ptlock_ptr(pmd_page(*pmd));
1589 static inline bool ptlock_init(struct page *page)
1592 * prep_new_page() initialize page->private (and therefore page->ptl)
1593 * with 0. Make sure nobody took it in use in between.
1595 * It can happen if arch try to use slab for page table allocation:
1596 * slab code uses page->slab_cache, which share storage with page->ptl.
1598 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1599 if (!ptlock_alloc(page))
1601 spin_lock_init(ptlock_ptr(page));
1605 /* Reset page->mapping so free_pages_check won't complain. */
1606 static inline void pte_lock_deinit(struct page *page)
1608 page->mapping = NULL;
1612 #else /* !USE_SPLIT_PTE_PTLOCKS */
1614 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1616 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1618 return &mm->page_table_lock;
1620 static inline void ptlock_cache_init(void) {}
1621 static inline bool ptlock_init(struct page *page) { return true; }
1622 static inline void pte_lock_deinit(struct page *page) {}
1623 #endif /* USE_SPLIT_PTE_PTLOCKS */
1625 static inline void pgtable_init(void)
1627 ptlock_cache_init();
1628 pgtable_cache_init();
1631 static inline bool pgtable_page_ctor(struct page *page)
1633 if (!ptlock_init(page))
1635 inc_zone_page_state(page, NR_PAGETABLE);
1639 static inline void pgtable_page_dtor(struct page *page)
1641 pte_lock_deinit(page);
1642 dec_zone_page_state(page, NR_PAGETABLE);
1645 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1647 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1648 pte_t *__pte = pte_offset_map(pmd, address); \
1654 #define pte_unmap_unlock(pte, ptl) do { \
1659 #define pte_alloc(mm, pmd, address) \
1660 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1662 #define pte_alloc_map(mm, pmd, address) \
1663 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1665 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1666 (pte_alloc(mm, pmd, address) ? \
1667 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1669 #define pte_alloc_kernel(pmd, address) \
1670 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1671 NULL: pte_offset_kernel(pmd, address))
1673 #if USE_SPLIT_PMD_PTLOCKS
1675 static struct page *pmd_to_page(pmd_t *pmd)
1677 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1678 return virt_to_page((void *)((unsigned long) pmd & mask));
1681 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1683 return ptlock_ptr(pmd_to_page(pmd));
1686 static inline bool pgtable_pmd_page_ctor(struct page *page)
1688 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1689 page->pmd_huge_pte = NULL;
1691 return ptlock_init(page);
1694 static inline void pgtable_pmd_page_dtor(struct page *page)
1696 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1697 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1702 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1706 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1708 return &mm->page_table_lock;
1711 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1712 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1714 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1718 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1720 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1725 extern void free_area_init(unsigned long * zones_size);
1726 extern void free_area_init_node(int nid, unsigned long * zones_size,
1727 unsigned long zone_start_pfn, unsigned long *zholes_size);
1728 extern void free_initmem(void);
1731 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1732 * into the buddy system. The freed pages will be poisoned with pattern
1733 * "poison" if it's within range [0, UCHAR_MAX].
1734 * Return pages freed into the buddy system.
1736 extern unsigned long free_reserved_area(void *start, void *end,
1737 int poison, char *s);
1739 #ifdef CONFIG_HIGHMEM
1741 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1742 * and totalram_pages.
1744 extern void free_highmem_page(struct page *page);
1747 extern void adjust_managed_page_count(struct page *page, long count);
1748 extern void mem_init_print_info(const char *str);
1750 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1752 /* Free the reserved page into the buddy system, so it gets managed. */
1753 static inline void __free_reserved_page(struct page *page)
1755 ClearPageReserved(page);
1756 init_page_count(page);
1760 static inline void free_reserved_page(struct page *page)
1762 __free_reserved_page(page);
1763 adjust_managed_page_count(page, 1);
1766 static inline void mark_page_reserved(struct page *page)
1768 SetPageReserved(page);
1769 adjust_managed_page_count(page, -1);
1773 * Default method to free all the __init memory into the buddy system.
1774 * The freed pages will be poisoned with pattern "poison" if it's within
1775 * range [0, UCHAR_MAX].
1776 * Return pages freed into the buddy system.
1778 static inline unsigned long free_initmem_default(int poison)
1780 extern char __init_begin[], __init_end[];
1782 return free_reserved_area(&__init_begin, &__init_end,
1783 poison, "unused kernel");
1786 static inline unsigned long get_num_physpages(void)
1789 unsigned long phys_pages = 0;
1791 for_each_online_node(nid)
1792 phys_pages += node_present_pages(nid);
1797 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1799 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1800 * zones, allocate the backing mem_map and account for memory holes in a more
1801 * architecture independent manner. This is a substitute for creating the
1802 * zone_sizes[] and zholes_size[] arrays and passing them to
1803 * free_area_init_node()
1805 * An architecture is expected to register range of page frames backed by
1806 * physical memory with memblock_add[_node]() before calling
1807 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1808 * usage, an architecture is expected to do something like
1810 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1812 * for_each_valid_physical_page_range()
1813 * memblock_add_node(base, size, nid)
1814 * free_area_init_nodes(max_zone_pfns);
1816 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1817 * registered physical page range. Similarly
1818 * sparse_memory_present_with_active_regions() calls memory_present() for
1819 * each range when SPARSEMEM is enabled.
1821 * See mm/page_alloc.c for more information on each function exposed by
1822 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1824 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1825 unsigned long node_map_pfn_alignment(void);
1826 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1827 unsigned long end_pfn);
1828 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1829 unsigned long end_pfn);
1830 extern void get_pfn_range_for_nid(unsigned int nid,
1831 unsigned long *start_pfn, unsigned long *end_pfn);
1832 extern unsigned long find_min_pfn_with_active_regions(void);
1833 extern void free_bootmem_with_active_regions(int nid,
1834 unsigned long max_low_pfn);
1835 extern void sparse_memory_present_with_active_regions(int nid);
1837 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1839 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1840 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1841 static inline int __early_pfn_to_nid(unsigned long pfn,
1842 struct mminit_pfnnid_cache *state)
1847 /* please see mm/page_alloc.c */
1848 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1849 /* there is a per-arch backend function. */
1850 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1851 struct mminit_pfnnid_cache *state);
1854 extern void set_dma_reserve(unsigned long new_dma_reserve);
1855 extern void memmap_init_zone(unsigned long, int, unsigned long,
1856 unsigned long, enum memmap_context);
1857 extern void setup_per_zone_wmarks(void);
1858 extern int __meminit init_per_zone_wmark_min(void);
1859 extern void mem_init(void);
1860 extern void __init mmap_init(void);
1861 extern void show_mem(unsigned int flags);
1862 extern long si_mem_available(void);
1863 extern void si_meminfo(struct sysinfo * val);
1864 extern void si_meminfo_node(struct sysinfo *val, int nid);
1866 extern __printf(3, 4)
1867 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1868 const char *fmt, ...);
1870 extern void setup_per_cpu_pageset(void);
1872 extern void zone_pcp_update(struct zone *zone);
1873 extern void zone_pcp_reset(struct zone *zone);
1876 extern int min_free_kbytes;
1877 extern int watermark_scale_factor;
1880 extern atomic_long_t mmap_pages_allocated;
1881 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1883 /* interval_tree.c */
1884 void vma_interval_tree_insert(struct vm_area_struct *node,
1885 struct rb_root *root);
1886 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1887 struct vm_area_struct *prev,
1888 struct rb_root *root);
1889 void vma_interval_tree_remove(struct vm_area_struct *node,
1890 struct rb_root *root);
1891 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1892 unsigned long start, unsigned long last);
1893 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1894 unsigned long start, unsigned long last);
1896 #define vma_interval_tree_foreach(vma, root, start, last) \
1897 for (vma = vma_interval_tree_iter_first(root, start, last); \
1898 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1900 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1901 struct rb_root *root);
1902 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1903 struct rb_root *root);
1904 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1905 struct rb_root *root, unsigned long start, unsigned long last);
1906 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1907 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1908 #ifdef CONFIG_DEBUG_VM_RB
1909 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1912 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1913 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1914 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1917 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1918 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1919 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1920 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1921 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1922 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1923 struct mempolicy *, struct vm_userfaultfd_ctx);
1924 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1925 extern int split_vma(struct mm_struct *,
1926 struct vm_area_struct *, unsigned long addr, int new_below);
1927 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1928 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1929 struct rb_node **, struct rb_node *);
1930 extern void unlink_file_vma(struct vm_area_struct *);
1931 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1932 unsigned long addr, unsigned long len, pgoff_t pgoff,
1933 bool *need_rmap_locks);
1934 extern void exit_mmap(struct mm_struct *);
1936 static inline int check_data_rlimit(unsigned long rlim,
1938 unsigned long start,
1939 unsigned long end_data,
1940 unsigned long start_data)
1942 if (rlim < RLIM_INFINITY) {
1943 if (((new - start) + (end_data - start_data)) > rlim)
1950 extern int mm_take_all_locks(struct mm_struct *mm);
1951 extern void mm_drop_all_locks(struct mm_struct *mm);
1953 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1954 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1956 extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages);
1957 extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages);
1959 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1960 unsigned long addr, unsigned long len,
1961 unsigned long flags,
1962 const struct vm_special_mapping *spec);
1963 /* This is an obsolete alternative to _install_special_mapping. */
1964 extern int install_special_mapping(struct mm_struct *mm,
1965 unsigned long addr, unsigned long len,
1966 unsigned long flags, struct page **pages);
1968 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1970 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1971 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1972 extern unsigned long do_mmap(struct file *file, unsigned long addr,
1973 unsigned long len, unsigned long prot, unsigned long flags,
1974 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
1975 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1977 static inline unsigned long
1978 do_mmap_pgoff(struct file *file, unsigned long addr,
1979 unsigned long len, unsigned long prot, unsigned long flags,
1980 unsigned long pgoff, unsigned long *populate)
1982 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
1986 extern int __mm_populate(unsigned long addr, unsigned long len,
1988 static inline void mm_populate(unsigned long addr, unsigned long len)
1991 (void) __mm_populate(addr, len, 1);
1994 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1997 /* These take the mm semaphore themselves */
1998 extern unsigned long vm_brk(unsigned long, unsigned long);
1999 extern int vm_munmap(unsigned long, size_t);
2000 extern unsigned long vm_mmap(struct file *, unsigned long,
2001 unsigned long, unsigned long,
2002 unsigned long, unsigned long);
2004 struct vm_unmapped_area_info {
2005 #define VM_UNMAPPED_AREA_TOPDOWN 1
2006 unsigned long flags;
2007 unsigned long length;
2008 unsigned long low_limit;
2009 unsigned long high_limit;
2010 unsigned long align_mask;
2011 unsigned long align_offset;
2014 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
2015 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
2018 * Search for an unmapped address range.
2020 * We are looking for a range that:
2021 * - does not intersect with any VMA;
2022 * - is contained within the [low_limit, high_limit) interval;
2023 * - is at least the desired size.
2024 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2026 static inline unsigned long
2027 vm_unmapped_area(struct vm_unmapped_area_info *info)
2029 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
2030 return unmapped_area_topdown(info);
2032 return unmapped_area(info);
2036 extern void truncate_inode_pages(struct address_space *, loff_t);
2037 extern void truncate_inode_pages_range(struct address_space *,
2038 loff_t lstart, loff_t lend);
2039 extern void truncate_inode_pages_final(struct address_space *);
2041 /* generic vm_area_ops exported for stackable file systems */
2042 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2043 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2044 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2046 /* mm/page-writeback.c */
2047 int write_one_page(struct page *page, int wait);
2048 void task_dirty_inc(struct task_struct *tsk);
2051 #define VM_MAX_READAHEAD 128 /* kbytes */
2052 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2054 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2055 pgoff_t offset, unsigned long nr_to_read);
2057 void page_cache_sync_readahead(struct address_space *mapping,
2058 struct file_ra_state *ra,
2061 unsigned long size);
2063 void page_cache_async_readahead(struct address_space *mapping,
2064 struct file_ra_state *ra,
2068 unsigned long size);
2070 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2071 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2073 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2074 extern int expand_downwards(struct vm_area_struct *vma,
2075 unsigned long address);
2077 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2079 #define expand_upwards(vma, address) (0)
2082 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2083 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2084 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2085 struct vm_area_struct **pprev);
2087 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2088 NULL if none. Assume start_addr < end_addr. */
2089 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2091 struct vm_area_struct * vma = find_vma(mm,start_addr);
2093 if (vma && end_addr <= vma->vm_start)
2098 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2100 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2103 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2104 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2105 unsigned long vm_start, unsigned long vm_end)
2107 struct vm_area_struct *vma = find_vma(mm, vm_start);
2109 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2116 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2117 void vma_set_page_prot(struct vm_area_struct *vma);
2119 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2123 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2125 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2129 #ifdef CONFIG_NUMA_BALANCING
2130 unsigned long change_prot_numa(struct vm_area_struct *vma,
2131 unsigned long start, unsigned long end);
2134 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2135 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2136 unsigned long pfn, unsigned long size, pgprot_t);
2137 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2138 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2140 int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr,
2141 unsigned long pfn, pgprot_t pgprot);
2142 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2144 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2147 struct page *follow_page_mask(struct vm_area_struct *vma,
2148 unsigned long address, unsigned int foll_flags,
2149 unsigned int *page_mask);
2151 static inline struct page *follow_page(struct vm_area_struct *vma,
2152 unsigned long address, unsigned int foll_flags)
2154 unsigned int unused_page_mask;
2155 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2158 #define FOLL_WRITE 0x01 /* check pte is writable */
2159 #define FOLL_TOUCH 0x02 /* mark page accessed */
2160 #define FOLL_GET 0x04 /* do get_page on page */
2161 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2162 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2163 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2164 * and return without waiting upon it */
2165 #define FOLL_POPULATE 0x40 /* fault in page */
2166 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2167 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2168 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2169 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2170 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2171 #define FOLL_MLOCK 0x1000 /* lock present pages */
2173 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2175 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2176 unsigned long size, pte_fn_t fn, void *data);
2179 #ifdef CONFIG_PAGE_POISONING
2180 extern bool page_poisoning_enabled(void);
2181 extern void kernel_poison_pages(struct page *page, int numpages, int enable);
2182 extern bool page_is_poisoned(struct page *page);
2184 static inline bool page_poisoning_enabled(void) { return false; }
2185 static inline void kernel_poison_pages(struct page *page, int numpages,
2187 static inline bool page_is_poisoned(struct page *page) { return false; }
2190 #ifdef CONFIG_DEBUG_PAGEALLOC
2191 extern bool _debug_pagealloc_enabled;
2192 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2194 static inline bool debug_pagealloc_enabled(void)
2196 return _debug_pagealloc_enabled;
2200 kernel_map_pages(struct page *page, int numpages, int enable)
2202 if (!debug_pagealloc_enabled())
2205 __kernel_map_pages(page, numpages, enable);
2207 #ifdef CONFIG_HIBERNATION
2208 extern bool kernel_page_present(struct page *page);
2209 #endif /* CONFIG_HIBERNATION */
2210 #else /* CONFIG_DEBUG_PAGEALLOC */
2212 kernel_map_pages(struct page *page, int numpages, int enable) {}
2213 #ifdef CONFIG_HIBERNATION
2214 static inline bool kernel_page_present(struct page *page) { return true; }
2215 #endif /* CONFIG_HIBERNATION */
2216 static inline bool debug_pagealloc_enabled(void)
2220 #endif /* CONFIG_DEBUG_PAGEALLOC */
2222 #ifdef __HAVE_ARCH_GATE_AREA
2223 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2224 extern int in_gate_area_no_mm(unsigned long addr);
2225 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2227 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2231 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2232 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2236 #endif /* __HAVE_ARCH_GATE_AREA */
2238 #ifdef CONFIG_SYSCTL
2239 extern int sysctl_drop_caches;
2240 int drop_caches_sysctl_handler(struct ctl_table *, int,
2241 void __user *, size_t *, loff_t *);
2244 void drop_slab(void);
2245 void drop_slab_node(int nid);
2248 #define randomize_va_space 0
2250 extern int randomize_va_space;
2253 const char * arch_vma_name(struct vm_area_struct *vma);
2254 void print_vma_addr(char *prefix, unsigned long rip);
2256 void sparse_mem_maps_populate_node(struct page **map_map,
2257 unsigned long pnum_begin,
2258 unsigned long pnum_end,
2259 unsigned long map_count,
2262 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2263 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2264 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2265 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2266 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2267 void *vmemmap_alloc_block(unsigned long size, int node);
2269 void *__vmemmap_alloc_block_buf(unsigned long size, int node,
2270 struct vmem_altmap *altmap);
2271 static inline void *vmemmap_alloc_block_buf(unsigned long size, int node)
2273 return __vmemmap_alloc_block_buf(size, node, NULL);
2276 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2277 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2279 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2280 void vmemmap_populate_print_last(void);
2281 #ifdef CONFIG_MEMORY_HOTPLUG
2282 void vmemmap_free(unsigned long start, unsigned long end);
2284 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2285 unsigned long size);
2288 MF_COUNT_INCREASED = 1 << 0,
2289 MF_ACTION_REQUIRED = 1 << 1,
2290 MF_MUST_KILL = 1 << 2,
2291 MF_SOFT_OFFLINE = 1 << 3,
2293 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2294 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2295 extern int unpoison_memory(unsigned long pfn);
2296 extern int get_hwpoison_page(struct page *page);
2297 #define put_hwpoison_page(page) put_page(page)
2298 extern int sysctl_memory_failure_early_kill;
2299 extern int sysctl_memory_failure_recovery;
2300 extern void shake_page(struct page *p, int access);
2301 extern atomic_long_t num_poisoned_pages;
2302 extern int soft_offline_page(struct page *page, int flags);
2306 * Error handlers for various types of pages.
2309 MF_IGNORED, /* Error: cannot be handled */
2310 MF_FAILED, /* Error: handling failed */
2311 MF_DELAYED, /* Will be handled later */
2312 MF_RECOVERED, /* Successfully recovered */
2315 enum mf_action_page_type {
2317 MF_MSG_KERNEL_HIGH_ORDER,
2319 MF_MSG_DIFFERENT_COMPOUND,
2320 MF_MSG_POISONED_HUGE,
2323 MF_MSG_UNMAP_FAILED,
2324 MF_MSG_DIRTY_SWAPCACHE,
2325 MF_MSG_CLEAN_SWAPCACHE,
2326 MF_MSG_DIRTY_MLOCKED_LRU,
2327 MF_MSG_CLEAN_MLOCKED_LRU,
2328 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2329 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2332 MF_MSG_TRUNCATED_LRU,
2338 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2339 extern void clear_huge_page(struct page *page,
2341 unsigned int pages_per_huge_page);
2342 extern void copy_user_huge_page(struct page *dst, struct page *src,
2343 unsigned long addr, struct vm_area_struct *vma,
2344 unsigned int pages_per_huge_page);
2345 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2347 extern struct page_ext_operations debug_guardpage_ops;
2348 extern struct page_ext_operations page_poisoning_ops;
2350 #ifdef CONFIG_DEBUG_PAGEALLOC
2351 extern unsigned int _debug_guardpage_minorder;
2352 extern bool _debug_guardpage_enabled;
2354 static inline unsigned int debug_guardpage_minorder(void)
2356 return _debug_guardpage_minorder;
2359 static inline bool debug_guardpage_enabled(void)
2361 return _debug_guardpage_enabled;
2364 static inline bool page_is_guard(struct page *page)
2366 struct page_ext *page_ext;
2368 if (!debug_guardpage_enabled())
2371 page_ext = lookup_page_ext(page);
2372 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2375 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2376 static inline bool debug_guardpage_enabled(void) { return false; }
2377 static inline bool page_is_guard(struct page *page) { return false; }
2378 #endif /* CONFIG_DEBUG_PAGEALLOC */
2380 #if MAX_NUMNODES > 1
2381 void __init setup_nr_node_ids(void);
2383 static inline void setup_nr_node_ids(void) {}
2386 #endif /* __KERNEL__ */
2387 #endif /* _LINUX_MM_H */