4 #include <linux/errno.h>
8 #include <linux/mmdebug.h>
10 #include <linux/bug.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/bit_spinlock.h>
20 #include <linux/shrinker.h>
21 #include <linux/resource.h>
22 #include <linux/page_ext.h>
23 #include <linux/err.h>
27 struct anon_vma_chain;
30 struct writeback_control;
33 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
34 extern unsigned long max_mapnr;
36 static inline void set_max_mapnr(unsigned long limit)
41 static inline void set_max_mapnr(unsigned long limit) { }
44 extern unsigned long totalram_pages;
45 extern void * high_memory;
46 extern int page_cluster;
49 extern int sysctl_legacy_va_layout;
51 #define sysctl_legacy_va_layout 0
54 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
55 extern const int mmap_rnd_bits_min;
56 extern const int mmap_rnd_bits_max;
57 extern int mmap_rnd_bits __read_mostly;
59 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
60 extern const int mmap_rnd_compat_bits_min;
61 extern const int mmap_rnd_compat_bits_max;
62 extern int mmap_rnd_compat_bits __read_mostly;
66 #include <asm/pgtable.h>
67 #include <asm/processor.h>
70 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
74 * To prevent common memory management code establishing
75 * a zero page mapping on a read fault.
76 * This macro should be defined within <asm/pgtable.h>.
77 * s390 does this to prevent multiplexing of hardware bits
78 * related to the physical page in case of virtualization.
80 #ifndef mm_forbids_zeropage
81 #define mm_forbids_zeropage(X) (0)
84 extern unsigned long sysctl_user_reserve_kbytes;
85 extern unsigned long sysctl_admin_reserve_kbytes;
87 extern int sysctl_overcommit_memory;
88 extern int sysctl_overcommit_ratio;
89 extern unsigned long sysctl_overcommit_kbytes;
91 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
93 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
96 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
98 /* to align the pointer to the (next) page boundary */
99 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
101 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
102 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
105 * Linux kernel virtual memory manager primitives.
106 * The idea being to have a "virtual" mm in the same way
107 * we have a virtual fs - giving a cleaner interface to the
108 * mm details, and allowing different kinds of memory mappings
109 * (from shared memory to executable loading to arbitrary
113 extern struct kmem_cache *vm_area_cachep;
116 extern struct rb_root nommu_region_tree;
117 extern struct rw_semaphore nommu_region_sem;
119 extern unsigned int kobjsize(const void *objp);
123 * vm_flags in vm_area_struct, see mm_types.h.
125 #define VM_NONE 0x00000000
127 #define VM_READ 0x00000001 /* currently active flags */
128 #define VM_WRITE 0x00000002
129 #define VM_EXEC 0x00000004
130 #define VM_SHARED 0x00000008
132 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
133 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
134 #define VM_MAYWRITE 0x00000020
135 #define VM_MAYEXEC 0x00000040
136 #define VM_MAYSHARE 0x00000080
138 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
139 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
140 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
141 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
142 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
144 #define VM_LOCKED 0x00002000
145 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
147 /* Used by sys_madvise() */
148 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
149 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
151 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
152 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
153 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
154 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
155 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
156 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
157 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
158 #define VM_ARCH_2 0x02000000
159 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
161 #ifdef CONFIG_MEM_SOFT_DIRTY
162 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
164 # define VM_SOFTDIRTY 0
167 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
168 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
169 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
170 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
172 #if defined(CONFIG_X86)
173 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
174 #elif defined(CONFIG_PPC)
175 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
176 #elif defined(CONFIG_PARISC)
177 # define VM_GROWSUP VM_ARCH_1
178 #elif defined(CONFIG_METAG)
179 # define VM_GROWSUP VM_ARCH_1
180 #elif defined(CONFIG_IA64)
181 # define VM_GROWSUP VM_ARCH_1
182 #elif !defined(CONFIG_MMU)
183 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
186 #if defined(CONFIG_X86)
187 /* MPX specific bounds table or bounds directory */
188 # define VM_MPX VM_ARCH_2
192 # define VM_GROWSUP VM_NONE
195 /* Bits set in the VMA until the stack is in its final location */
196 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
198 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
199 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
202 #ifdef CONFIG_STACK_GROWSUP
203 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
205 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
209 * Special vmas that are non-mergable, non-mlock()able.
210 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
212 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
214 /* This mask defines which mm->def_flags a process can inherit its parent */
215 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
217 /* This mask is used to clear all the VMA flags used by mlock */
218 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
221 * mapping from the currently active vm_flags protection bits (the
222 * low four bits) to a page protection mask..
224 extern pgprot_t protection_map[16];
226 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
227 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
228 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
229 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
230 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
231 #define FAULT_FLAG_TRIED 0x20 /* Second try */
232 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
235 * vm_fault is filled by the the pagefault handler and passed to the vma's
236 * ->fault function. The vma's ->fault is responsible for returning a bitmask
237 * of VM_FAULT_xxx flags that give details about how the fault was handled.
239 * pgoff should be used in favour of virtual_address, if possible.
242 unsigned int flags; /* FAULT_FLAG_xxx flags */
243 pgoff_t pgoff; /* Logical page offset based on vma */
244 void __user *virtual_address; /* Faulting virtual address */
246 struct page *cow_page; /* Handler may choose to COW */
247 struct page *page; /* ->fault handlers should return a
248 * page here, unless VM_FAULT_NOPAGE
249 * is set (which is also implied by
252 /* for ->map_pages() only */
253 pgoff_t max_pgoff; /* map pages for offset from pgoff till
254 * max_pgoff inclusive */
255 pte_t *pte; /* pte entry associated with ->pgoff */
259 * These are the virtual MM functions - opening of an area, closing and
260 * unmapping it (needed to keep files on disk up-to-date etc), pointer
261 * to the functions called when a no-page or a wp-page exception occurs.
263 struct vm_operations_struct {
264 void (*open)(struct vm_area_struct * area);
265 void (*close)(struct vm_area_struct * area);
266 int (*mremap)(struct vm_area_struct * area);
267 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
268 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
269 pmd_t *, unsigned int flags);
270 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
272 /* notification that a previously read-only page is about to become
273 * writable, if an error is returned it will cause a SIGBUS */
274 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
276 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
277 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
279 /* called by access_process_vm when get_user_pages() fails, typically
280 * for use by special VMAs that can switch between memory and hardware
282 int (*access)(struct vm_area_struct *vma, unsigned long addr,
283 void *buf, int len, int write);
285 /* Called by the /proc/PID/maps code to ask the vma whether it
286 * has a special name. Returning non-NULL will also cause this
287 * vma to be dumped unconditionally. */
288 const char *(*name)(struct vm_area_struct *vma);
292 * set_policy() op must add a reference to any non-NULL @new mempolicy
293 * to hold the policy upon return. Caller should pass NULL @new to
294 * remove a policy and fall back to surrounding context--i.e. do not
295 * install a MPOL_DEFAULT policy, nor the task or system default
298 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
301 * get_policy() op must add reference [mpol_get()] to any policy at
302 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
303 * in mm/mempolicy.c will do this automatically.
304 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
305 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
306 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
307 * must return NULL--i.e., do not "fallback" to task or system default
310 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
314 * Called by vm_normal_page() for special PTEs to find the
315 * page for @addr. This is useful if the default behavior
316 * (using pte_page()) would not find the correct page.
318 struct page *(*find_special_page)(struct vm_area_struct *vma,
325 #define page_private(page) ((page)->private)
326 #define set_page_private(page, v) ((page)->private = (v))
329 * FIXME: take this include out, include page-flags.h in
330 * files which need it (119 of them)
332 #include <linux/page-flags.h>
333 #include <linux/huge_mm.h>
336 * Methods to modify the page usage count.
338 * What counts for a page usage:
339 * - cache mapping (page->mapping)
340 * - private data (page->private)
341 * - page mapped in a task's page tables, each mapping
342 * is counted separately
344 * Also, many kernel routines increase the page count before a critical
345 * routine so they can be sure the page doesn't go away from under them.
349 * Drop a ref, return true if the refcount fell to zero (the page has no users)
351 static inline int put_page_testzero(struct page *page)
353 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
354 return atomic_dec_and_test(&page->_count);
358 * Try to grab a ref unless the page has a refcount of zero, return false if
360 * This can be called when MMU is off so it must not access
361 * any of the virtual mappings.
363 static inline int get_page_unless_zero(struct page *page)
365 return atomic_inc_not_zero(&page->_count);
368 extern int page_is_ram(unsigned long pfn);
376 int region_intersects(resource_size_t offset, size_t size, const char *type);
378 /* Support for virtually mapped pages */
379 struct page *vmalloc_to_page(const void *addr);
380 unsigned long vmalloc_to_pfn(const void *addr);
383 * Determine if an address is within the vmalloc range
385 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
386 * is no special casing required.
388 static inline int is_vmalloc_addr(const void *x)
391 unsigned long addr = (unsigned long)x;
393 return addr >= VMALLOC_START && addr < VMALLOC_END;
399 extern int is_vmalloc_or_module_addr(const void *x);
401 static inline int is_vmalloc_or_module_addr(const void *x)
407 extern void kvfree(const void *addr);
409 static inline void compound_lock(struct page *page)
411 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
412 VM_BUG_ON_PAGE(PageSlab(page), page);
413 bit_spin_lock(PG_compound_lock, &page->flags);
417 static inline void compound_unlock(struct page *page)
419 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
420 VM_BUG_ON_PAGE(PageSlab(page), page);
421 bit_spin_unlock(PG_compound_lock, &page->flags);
425 static inline unsigned long compound_lock_irqsave(struct page *page)
427 unsigned long uninitialized_var(flags);
428 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
429 local_irq_save(flags);
435 static inline void compound_unlock_irqrestore(struct page *page,
438 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
439 compound_unlock(page);
440 local_irq_restore(flags);
445 * The atomic page->_mapcount, starts from -1: so that transitions
446 * both from it and to it can be tracked, using atomic_inc_and_test
447 * and atomic_add_negative(-1).
449 static inline void page_mapcount_reset(struct page *page)
451 atomic_set(&(page)->_mapcount, -1);
454 static inline int page_mapcount(struct page *page)
456 VM_BUG_ON_PAGE(PageSlab(page), page);
457 return atomic_read(&page->_mapcount) + 1;
460 static inline int page_count(struct page *page)
462 return atomic_read(&compound_head(page)->_count);
465 static inline bool __compound_tail_refcounted(struct page *page)
467 return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
471 * This takes a head page as parameter and tells if the
472 * tail page reference counting can be skipped.
474 * For this to be safe, PageSlab and PageHeadHuge must remain true on
475 * any given page where they return true here, until all tail pins
476 * have been released.
478 static inline bool compound_tail_refcounted(struct page *page)
480 VM_BUG_ON_PAGE(!PageHead(page), page);
481 return __compound_tail_refcounted(page);
484 static inline void get_huge_page_tail(struct page *page)
487 * __split_huge_page_refcount() cannot run from under us.
489 VM_BUG_ON_PAGE(!PageTail(page), page);
490 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
491 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
492 if (compound_tail_refcounted(compound_head(page)))
493 atomic_inc(&page->_mapcount);
496 extern bool __get_page_tail(struct page *page);
498 static inline void get_page(struct page *page)
500 if (unlikely(PageTail(page)))
501 if (likely(__get_page_tail(page)))
504 * Getting a normal page or the head of a compound page
505 * requires to already have an elevated page->_count.
507 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
508 atomic_inc(&page->_count);
511 static inline struct page *virt_to_head_page(const void *x)
513 struct page *page = virt_to_page(x);
515 return compound_head(page);
519 * Setup the page count before being freed into the page allocator for
520 * the first time (boot or memory hotplug)
522 static inline void init_page_count(struct page *page)
524 atomic_set(&page->_count, 1);
527 void put_page(struct page *page);
528 void put_pages_list(struct list_head *pages);
530 void split_page(struct page *page, unsigned int order);
531 int split_free_page(struct page *page);
534 * Compound pages have a destructor function. Provide a
535 * prototype for that function and accessor functions.
536 * These are _only_ valid on the head of a compound page.
538 typedef void compound_page_dtor(struct page *);
540 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
541 enum compound_dtor_id {
544 #ifdef CONFIG_HUGETLB_PAGE
549 extern compound_page_dtor * const compound_page_dtors[];
551 static inline void set_compound_page_dtor(struct page *page,
552 enum compound_dtor_id compound_dtor)
554 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
555 page[1].compound_dtor = compound_dtor;
558 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
560 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
561 return compound_page_dtors[page[1].compound_dtor];
564 static inline unsigned int compound_order(struct page *page)
568 return page[1].compound_order;
571 static inline void set_compound_order(struct page *page, unsigned int order)
573 page[1].compound_order = order;
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_CACHE_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 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
704 #define SECTION_IN_PAGE_FLAGS
708 * The identification function is mainly used by the buddy allocator for
709 * determining if two pages could be buddies. We are not really identifying
710 * the zone since we could be using the section number id if we do not have
711 * node id available in page flags.
712 * We only guarantee that it will return the same value for two combinable
715 static inline int page_zone_id(struct page *page)
717 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
720 static inline int zone_to_nid(struct zone *zone)
729 #ifdef NODE_NOT_IN_PAGE_FLAGS
730 extern int page_to_nid(const struct page *page);
732 static inline int page_to_nid(const struct page *page)
734 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
738 #ifdef CONFIG_NUMA_BALANCING
739 static inline int cpu_pid_to_cpupid(int cpu, int pid)
741 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
744 static inline int cpupid_to_pid(int cpupid)
746 return cpupid & LAST__PID_MASK;
749 static inline int cpupid_to_cpu(int cpupid)
751 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
754 static inline int cpupid_to_nid(int cpupid)
756 return cpu_to_node(cpupid_to_cpu(cpupid));
759 static inline bool cpupid_pid_unset(int cpupid)
761 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
764 static inline bool cpupid_cpu_unset(int cpupid)
766 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
769 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
771 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
774 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
775 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
776 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
778 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
781 static inline int page_cpupid_last(struct page *page)
783 return page->_last_cpupid;
785 static inline void page_cpupid_reset_last(struct page *page)
787 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
790 static inline int page_cpupid_last(struct page *page)
792 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
795 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
797 static inline void page_cpupid_reset_last(struct page *page)
799 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
801 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
802 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
804 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
805 #else /* !CONFIG_NUMA_BALANCING */
806 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
808 return page_to_nid(page); /* XXX */
811 static inline int page_cpupid_last(struct page *page)
813 return page_to_nid(page); /* XXX */
816 static inline int cpupid_to_nid(int cpupid)
821 static inline int cpupid_to_pid(int cpupid)
826 static inline int cpupid_to_cpu(int cpupid)
831 static inline int cpu_pid_to_cpupid(int nid, int pid)
836 static inline bool cpupid_pid_unset(int cpupid)
841 static inline void page_cpupid_reset_last(struct page *page)
845 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
849 #endif /* CONFIG_NUMA_BALANCING */
851 static inline struct zone *page_zone(const struct page *page)
853 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
856 #ifdef SECTION_IN_PAGE_FLAGS
857 static inline void set_page_section(struct page *page, unsigned long section)
859 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
860 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
863 static inline unsigned long page_to_section(const struct page *page)
865 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
869 static inline void set_page_zone(struct page *page, enum zone_type zone)
871 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
872 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
875 static inline void set_page_node(struct page *page, unsigned long node)
877 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
878 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
881 static inline void set_page_links(struct page *page, enum zone_type zone,
882 unsigned long node, unsigned long pfn)
884 set_page_zone(page, zone);
885 set_page_node(page, node);
886 #ifdef SECTION_IN_PAGE_FLAGS
887 set_page_section(page, pfn_to_section_nr(pfn));
892 static inline struct mem_cgroup *page_memcg(struct page *page)
894 return page->mem_cgroup;
897 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
899 page->mem_cgroup = memcg;
902 static inline struct mem_cgroup *page_memcg(struct page *page)
907 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
913 * Some inline functions in vmstat.h depend on page_zone()
915 #include <linux/vmstat.h>
917 static __always_inline void *lowmem_page_address(const struct page *page)
919 return __va(PFN_PHYS(page_to_pfn(page)));
922 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
923 #define HASHED_PAGE_VIRTUAL
926 #if defined(WANT_PAGE_VIRTUAL)
927 static inline void *page_address(const struct page *page)
929 return page->virtual;
931 static inline void set_page_address(struct page *page, void *address)
933 page->virtual = address;
935 #define page_address_init() do { } while(0)
938 #if defined(HASHED_PAGE_VIRTUAL)
939 void *page_address(const struct page *page);
940 void set_page_address(struct page *page, void *virtual);
941 void page_address_init(void);
944 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
945 #define page_address(page) lowmem_page_address(page)
946 #define set_page_address(page, address) do { } while(0)
947 #define page_address_init() do { } while(0)
950 extern void *page_rmapping(struct page *page);
951 extern struct anon_vma *page_anon_vma(struct page *page);
952 extern struct address_space *page_mapping(struct page *page);
954 extern struct address_space *__page_file_mapping(struct page *);
957 struct address_space *page_file_mapping(struct page *page)
959 if (unlikely(PageSwapCache(page)))
960 return __page_file_mapping(page);
962 return page->mapping;
966 * Return the pagecache index of the passed page. Regular pagecache pages
967 * use ->index whereas swapcache pages use ->private
969 static inline pgoff_t page_index(struct page *page)
971 if (unlikely(PageSwapCache(page)))
972 return page_private(page);
976 extern pgoff_t __page_file_index(struct page *page);
979 * Return the file index of the page. Regular pagecache pages use ->index
980 * whereas swapcache pages use swp_offset(->private)
982 static inline pgoff_t page_file_index(struct page *page)
984 if (unlikely(PageSwapCache(page)))
985 return __page_file_index(page);
991 * Return true if this page is mapped into pagetables.
993 static inline int page_mapped(struct page *page)
995 return atomic_read(&(page)->_mapcount) >= 0;
999 * Return true only if the page has been allocated with
1000 * ALLOC_NO_WATERMARKS and the low watermark was not
1001 * met implying that the system is under some pressure.
1003 static inline bool page_is_pfmemalloc(struct page *page)
1006 * Page index cannot be this large so this must be
1007 * a pfmemalloc page.
1009 return page->index == -1UL;
1013 * Only to be called by the page allocator on a freshly allocated
1016 static inline void set_page_pfmemalloc(struct page *page)
1021 static inline void clear_page_pfmemalloc(struct page *page)
1027 * Different kinds of faults, as returned by handle_mm_fault().
1028 * Used to decide whether a process gets delivered SIGBUS or
1029 * just gets major/minor fault counters bumped up.
1032 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1034 #define VM_FAULT_OOM 0x0001
1035 #define VM_FAULT_SIGBUS 0x0002
1036 #define VM_FAULT_MAJOR 0x0004
1037 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1038 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1039 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1040 #define VM_FAULT_SIGSEGV 0x0040
1042 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1043 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1044 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1045 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1047 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1049 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1050 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1053 /* Encode hstate index for a hwpoisoned large page */
1054 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1055 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1058 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1060 extern void pagefault_out_of_memory(void);
1062 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1065 * Flags passed to show_mem() and show_free_areas() to suppress output in
1068 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1070 extern void show_free_areas(unsigned int flags);
1071 extern bool skip_free_areas_node(unsigned int flags, int nid);
1073 void shmem_set_file(struct vm_area_struct *vma, struct file *file);
1074 int shmem_zero_setup(struct vm_area_struct *);
1076 bool shmem_mapping(struct address_space *mapping);
1078 static inline bool shmem_mapping(struct address_space *mapping)
1084 extern int can_do_mlock(void);
1085 extern int user_shm_lock(size_t, struct user_struct *);
1086 extern void user_shm_unlock(size_t, struct user_struct *);
1089 * Parameter block passed down to zap_pte_range in exceptional cases.
1091 struct zap_details {
1092 struct address_space *check_mapping; /* Check page->mapping if set */
1093 pgoff_t first_index; /* Lowest page->index to unmap */
1094 pgoff_t last_index; /* Highest page->index to unmap */
1097 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1099 struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr,
1102 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1103 unsigned long size);
1104 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1105 unsigned long size, struct zap_details *);
1106 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1107 unsigned long start, unsigned long end);
1110 * mm_walk - callbacks for walk_page_range
1111 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1112 * this handler is required to be able to handle
1113 * pmd_trans_huge() pmds. They may simply choose to
1114 * split_huge_page() instead of handling it explicitly.
1115 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1116 * @pte_hole: if set, called for each hole at all levels
1117 * @hugetlb_entry: if set, called for each hugetlb entry
1118 * @test_walk: caller specific callback function to determine whether
1119 * we walk over the current vma or not. A positive returned
1120 * value means "do page table walk over the current vma,"
1121 * and a negative one means "abort current page table walk
1122 * right now." 0 means "skip the current vma."
1123 * @mm: mm_struct representing the target process of page table walk
1124 * @vma: vma currently walked (NULL if walking outside vmas)
1125 * @private: private data for callbacks' usage
1127 * (see the comment on walk_page_range() for more details)
1130 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1131 unsigned long next, struct mm_walk *walk);
1132 int (*pte_entry)(pte_t *pte, unsigned long addr,
1133 unsigned long next, struct mm_walk *walk);
1134 int (*pte_hole)(unsigned long addr, unsigned long next,
1135 struct mm_walk *walk);
1136 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1137 unsigned long addr, unsigned long next,
1138 struct mm_walk *walk);
1139 int (*test_walk)(unsigned long addr, unsigned long next,
1140 struct mm_walk *walk);
1141 struct mm_struct *mm;
1142 struct vm_area_struct *vma;
1146 int walk_page_range(unsigned long addr, unsigned long end,
1147 struct mm_walk *walk);
1148 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1149 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1150 unsigned long end, unsigned long floor, unsigned long ceiling);
1151 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1152 struct vm_area_struct *vma);
1153 void unmap_mapping_range(struct address_space *mapping,
1154 loff_t const holebegin, loff_t const holelen, int even_cows);
1155 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1156 unsigned long *pfn);
1157 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1158 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1159 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1160 void *buf, int len, int write);
1162 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1163 loff_t const holebegin, loff_t const holelen)
1165 unmap_mapping_range(mapping, holebegin, holelen, 0);
1168 extern void truncate_pagecache(struct inode *inode, loff_t new);
1169 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1170 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1171 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1172 int truncate_inode_page(struct address_space *mapping, struct page *page);
1173 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1174 int invalidate_inode_page(struct page *page);
1177 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1178 unsigned long address, unsigned int flags);
1179 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1180 unsigned long address, unsigned int fault_flags);
1182 static inline int handle_mm_fault(struct mm_struct *mm,
1183 struct vm_area_struct *vma, unsigned long address,
1186 /* should never happen if there's no MMU */
1188 return VM_FAULT_SIGBUS;
1190 static inline int fixup_user_fault(struct task_struct *tsk,
1191 struct mm_struct *mm, unsigned long address,
1192 unsigned int fault_flags)
1194 /* should never happen if there's no MMU */
1200 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1201 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1202 void *buf, int len, int write);
1204 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1205 unsigned long start, unsigned long nr_pages,
1206 unsigned int foll_flags, struct page **pages,
1207 struct vm_area_struct **vmas, int *nonblocking);
1208 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1209 unsigned long start, unsigned long nr_pages,
1210 int write, int force, struct page **pages,
1211 struct vm_area_struct **vmas);
1212 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1213 unsigned long start, unsigned long nr_pages,
1214 int write, int force, struct page **pages,
1216 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1217 unsigned long start, unsigned long nr_pages,
1218 int write, int force, struct page **pages,
1219 unsigned int gup_flags);
1220 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1221 unsigned long start, unsigned long nr_pages,
1222 int write, int force, struct page **pages);
1223 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1224 struct page **pages);
1226 /* Container for pinned pfns / pages */
1227 struct frame_vector {
1228 unsigned int nr_allocated; /* Number of frames we have space for */
1229 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1230 bool got_ref; /* Did we pin pages by getting page ref? */
1231 bool is_pfns; /* Does array contain pages or pfns? */
1232 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1233 * pfns_vector_pages() or pfns_vector_pfns()
1237 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1238 void frame_vector_destroy(struct frame_vector *vec);
1239 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1240 bool write, bool force, struct frame_vector *vec);
1241 void put_vaddr_frames(struct frame_vector *vec);
1242 int frame_vector_to_pages(struct frame_vector *vec);
1243 void frame_vector_to_pfns(struct frame_vector *vec);
1245 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1247 return vec->nr_frames;
1250 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1253 int err = frame_vector_to_pages(vec);
1256 return ERR_PTR(err);
1258 return (struct page **)(vec->ptrs);
1261 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1264 frame_vector_to_pfns(vec);
1265 return (unsigned long *)(vec->ptrs);
1269 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1270 struct page **pages);
1271 int get_kernel_page(unsigned long start, int write, struct page **pages);
1272 struct page *get_dump_page(unsigned long addr);
1274 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1275 extern void do_invalidatepage(struct page *page, unsigned int offset,
1276 unsigned int length);
1278 int __set_page_dirty_nobuffers(struct page *page);
1279 int __set_page_dirty_no_writeback(struct page *page);
1280 int redirty_page_for_writepage(struct writeback_control *wbc,
1282 void account_page_dirtied(struct page *page, struct address_space *mapping,
1283 struct mem_cgroup *memcg);
1284 void account_page_cleaned(struct page *page, struct address_space *mapping,
1285 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1286 int set_page_dirty(struct page *page);
1287 int set_page_dirty_lock(struct page *page);
1288 void cancel_dirty_page(struct page *page);
1289 int clear_page_dirty_for_io(struct page *page);
1291 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1293 /* Is the vma a continuation of the stack vma above it? */
1294 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1296 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1299 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1301 return !vma->vm_ops;
1304 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1307 return (vma->vm_flags & VM_GROWSDOWN) &&
1308 (vma->vm_start == addr) &&
1309 !vma_growsdown(vma->vm_prev, addr);
1312 /* Is the vma a continuation of the stack vma below it? */
1313 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1315 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1318 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1321 return (vma->vm_flags & VM_GROWSUP) &&
1322 (vma->vm_end == addr) &&
1323 !vma_growsup(vma->vm_next, addr);
1326 int vma_is_stack_for_task(struct vm_area_struct *vma, struct task_struct *t);
1328 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1329 unsigned long old_addr, struct vm_area_struct *new_vma,
1330 unsigned long new_addr, unsigned long len,
1331 bool need_rmap_locks);
1332 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1333 unsigned long end, pgprot_t newprot,
1334 int dirty_accountable, int prot_numa);
1335 extern int mprotect_fixup(struct vm_area_struct *vma,
1336 struct vm_area_struct **pprev, unsigned long start,
1337 unsigned long end, unsigned long newflags);
1340 * doesn't attempt to fault and will return short.
1342 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1343 struct page **pages);
1345 * per-process(per-mm_struct) statistics.
1347 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1349 long val = atomic_long_read(&mm->rss_stat.count[member]);
1351 #ifdef SPLIT_RSS_COUNTING
1353 * counter is updated in asynchronous manner and may go to minus.
1354 * But it's never be expected number for users.
1359 return (unsigned long)val;
1362 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1364 atomic_long_add(value, &mm->rss_stat.count[member]);
1367 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1369 atomic_long_inc(&mm->rss_stat.count[member]);
1372 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1374 atomic_long_dec(&mm->rss_stat.count[member]);
1377 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1379 return get_mm_counter(mm, MM_FILEPAGES) +
1380 get_mm_counter(mm, MM_ANONPAGES);
1383 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1385 return max(mm->hiwater_rss, get_mm_rss(mm));
1388 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1390 return max(mm->hiwater_vm, mm->total_vm);
1393 static inline void update_hiwater_rss(struct mm_struct *mm)
1395 unsigned long _rss = get_mm_rss(mm);
1397 if ((mm)->hiwater_rss < _rss)
1398 (mm)->hiwater_rss = _rss;
1401 static inline void update_hiwater_vm(struct mm_struct *mm)
1403 if (mm->hiwater_vm < mm->total_vm)
1404 mm->hiwater_vm = mm->total_vm;
1407 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1409 mm->hiwater_rss = get_mm_rss(mm);
1412 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1413 struct mm_struct *mm)
1415 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1417 if (*maxrss < hiwater_rss)
1418 *maxrss = hiwater_rss;
1421 #if defined(SPLIT_RSS_COUNTING)
1422 void sync_mm_rss(struct mm_struct *mm);
1424 static inline void sync_mm_rss(struct mm_struct *mm)
1429 int vma_wants_writenotify(struct vm_area_struct *vma);
1431 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1433 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1437 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1441 #ifdef __PAGETABLE_PUD_FOLDED
1442 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1443 unsigned long address)
1448 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1451 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1452 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1453 unsigned long address)
1458 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1460 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1465 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1466 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1469 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1471 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1473 atomic_long_set(&mm->nr_pmds, 0);
1476 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1478 return atomic_long_read(&mm->nr_pmds);
1481 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1483 atomic_long_inc(&mm->nr_pmds);
1486 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1488 atomic_long_dec(&mm->nr_pmds);
1492 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1493 pmd_t *pmd, unsigned long address);
1494 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1497 * The following ifdef needed to get the 4level-fixup.h header to work.
1498 * Remove it when 4level-fixup.h has been removed.
1500 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1501 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1503 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1504 NULL: pud_offset(pgd, address);
1507 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1509 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1510 NULL: pmd_offset(pud, address);
1512 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1514 #if USE_SPLIT_PTE_PTLOCKS
1515 #if ALLOC_SPLIT_PTLOCKS
1516 void __init ptlock_cache_init(void);
1517 extern bool ptlock_alloc(struct page *page);
1518 extern void ptlock_free(struct page *page);
1520 static inline spinlock_t *ptlock_ptr(struct page *page)
1524 #else /* ALLOC_SPLIT_PTLOCKS */
1525 static inline void ptlock_cache_init(void)
1529 static inline bool ptlock_alloc(struct page *page)
1534 static inline void ptlock_free(struct page *page)
1538 static inline spinlock_t *ptlock_ptr(struct page *page)
1542 #endif /* ALLOC_SPLIT_PTLOCKS */
1544 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1546 return ptlock_ptr(pmd_page(*pmd));
1549 static inline bool ptlock_init(struct page *page)
1552 * prep_new_page() initialize page->private (and therefore page->ptl)
1553 * with 0. Make sure nobody took it in use in between.
1555 * It can happen if arch try to use slab for page table allocation:
1556 * slab code uses page->slab_cache, which share storage with page->ptl.
1558 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1559 if (!ptlock_alloc(page))
1561 spin_lock_init(ptlock_ptr(page));
1565 /* Reset page->mapping so free_pages_check won't complain. */
1566 static inline void pte_lock_deinit(struct page *page)
1568 page->mapping = NULL;
1572 #else /* !USE_SPLIT_PTE_PTLOCKS */
1574 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1576 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1578 return &mm->page_table_lock;
1580 static inline void ptlock_cache_init(void) {}
1581 static inline bool ptlock_init(struct page *page) { return true; }
1582 static inline void pte_lock_deinit(struct page *page) {}
1583 #endif /* USE_SPLIT_PTE_PTLOCKS */
1585 static inline void pgtable_init(void)
1587 ptlock_cache_init();
1588 pgtable_cache_init();
1591 static inline bool pgtable_page_ctor(struct page *page)
1593 if (!ptlock_init(page))
1595 inc_zone_page_state(page, NR_PAGETABLE);
1599 static inline void pgtable_page_dtor(struct page *page)
1601 pte_lock_deinit(page);
1602 dec_zone_page_state(page, NR_PAGETABLE);
1605 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1607 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1608 pte_t *__pte = pte_offset_map(pmd, address); \
1614 #define pte_unmap_unlock(pte, ptl) do { \
1619 #define pte_alloc_map(mm, vma, pmd, address) \
1620 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1622 NULL: pte_offset_map(pmd, address))
1624 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1625 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1627 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1629 #define pte_alloc_kernel(pmd, address) \
1630 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1631 NULL: pte_offset_kernel(pmd, address))
1633 #if USE_SPLIT_PMD_PTLOCKS
1635 static struct page *pmd_to_page(pmd_t *pmd)
1637 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1638 return virt_to_page((void *)((unsigned long) pmd & mask));
1641 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1643 return ptlock_ptr(pmd_to_page(pmd));
1646 static inline bool pgtable_pmd_page_ctor(struct page *page)
1648 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1649 page->pmd_huge_pte = NULL;
1651 return ptlock_init(page);
1654 static inline void pgtable_pmd_page_dtor(struct page *page)
1656 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1657 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1662 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1666 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1668 return &mm->page_table_lock;
1671 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1672 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1674 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1678 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1680 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1685 extern void free_area_init(unsigned long * zones_size);
1686 extern void free_area_init_node(int nid, unsigned long * zones_size,
1687 unsigned long zone_start_pfn, unsigned long *zholes_size);
1688 extern void free_initmem(void);
1691 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1692 * into the buddy system. The freed pages will be poisoned with pattern
1693 * "poison" if it's within range [0, UCHAR_MAX].
1694 * Return pages freed into the buddy system.
1696 extern unsigned long free_reserved_area(void *start, void *end,
1697 int poison, char *s);
1699 #ifdef CONFIG_HIGHMEM
1701 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1702 * and totalram_pages.
1704 extern void free_highmem_page(struct page *page);
1707 extern void adjust_managed_page_count(struct page *page, long count);
1708 extern void mem_init_print_info(const char *str);
1710 extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end);
1712 /* Free the reserved page into the buddy system, so it gets managed. */
1713 static inline void __free_reserved_page(struct page *page)
1715 ClearPageReserved(page);
1716 init_page_count(page);
1720 static inline void free_reserved_page(struct page *page)
1722 __free_reserved_page(page);
1723 adjust_managed_page_count(page, 1);
1726 static inline void mark_page_reserved(struct page *page)
1728 SetPageReserved(page);
1729 adjust_managed_page_count(page, -1);
1733 * Default method to free all the __init memory into the buddy system.
1734 * The freed pages will be poisoned with pattern "poison" if it's within
1735 * range [0, UCHAR_MAX].
1736 * Return pages freed into the buddy system.
1738 static inline unsigned long free_initmem_default(int poison)
1740 extern char __init_begin[], __init_end[];
1742 return free_reserved_area(&__init_begin, &__init_end,
1743 poison, "unused kernel");
1746 static inline unsigned long get_num_physpages(void)
1749 unsigned long phys_pages = 0;
1751 for_each_online_node(nid)
1752 phys_pages += node_present_pages(nid);
1757 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1759 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1760 * zones, allocate the backing mem_map and account for memory holes in a more
1761 * architecture independent manner. This is a substitute for creating the
1762 * zone_sizes[] and zholes_size[] arrays and passing them to
1763 * free_area_init_node()
1765 * An architecture is expected to register range of page frames backed by
1766 * physical memory with memblock_add[_node]() before calling
1767 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1768 * usage, an architecture is expected to do something like
1770 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1772 * for_each_valid_physical_page_range()
1773 * memblock_add_node(base, size, nid)
1774 * free_area_init_nodes(max_zone_pfns);
1776 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1777 * registered physical page range. Similarly
1778 * sparse_memory_present_with_active_regions() calls memory_present() for
1779 * each range when SPARSEMEM is enabled.
1781 * See mm/page_alloc.c for more information on each function exposed by
1782 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1784 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1785 unsigned long node_map_pfn_alignment(void);
1786 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1787 unsigned long end_pfn);
1788 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1789 unsigned long end_pfn);
1790 extern void get_pfn_range_for_nid(unsigned int nid,
1791 unsigned long *start_pfn, unsigned long *end_pfn);
1792 extern unsigned long find_min_pfn_with_active_regions(void);
1793 extern void free_bootmem_with_active_regions(int nid,
1794 unsigned long max_low_pfn);
1795 extern void sparse_memory_present_with_active_regions(int nid);
1797 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1799 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1800 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1801 static inline int __early_pfn_to_nid(unsigned long pfn,
1802 struct mminit_pfnnid_cache *state)
1807 /* please see mm/page_alloc.c */
1808 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1809 /* there is a per-arch backend function. */
1810 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1811 struct mminit_pfnnid_cache *state);
1814 extern void set_dma_reserve(unsigned long new_dma_reserve);
1815 extern void memmap_init_zone(unsigned long, int, unsigned long,
1816 unsigned long, enum memmap_context);
1817 extern void setup_per_zone_wmarks(void);
1818 extern int __meminit init_per_zone_wmark_min(void);
1819 extern void mem_init(void);
1820 extern void __init mmap_init(void);
1821 extern void show_mem(unsigned int flags);
1822 extern void si_meminfo(struct sysinfo * val);
1823 extern void si_meminfo_node(struct sysinfo *val, int nid);
1825 extern __printf(3, 4)
1826 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1827 const char *fmt, ...);
1829 extern void setup_per_cpu_pageset(void);
1831 extern void zone_pcp_update(struct zone *zone);
1832 extern void zone_pcp_reset(struct zone *zone);
1835 extern int min_free_kbytes;
1838 extern atomic_long_t mmap_pages_allocated;
1839 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1841 /* interval_tree.c */
1842 void vma_interval_tree_insert(struct vm_area_struct *node,
1843 struct rb_root *root);
1844 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1845 struct vm_area_struct *prev,
1846 struct rb_root *root);
1847 void vma_interval_tree_remove(struct vm_area_struct *node,
1848 struct rb_root *root);
1849 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1850 unsigned long start, unsigned long last);
1851 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1852 unsigned long start, unsigned long last);
1854 #define vma_interval_tree_foreach(vma, root, start, last) \
1855 for (vma = vma_interval_tree_iter_first(root, start, last); \
1856 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1858 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1859 struct rb_root *root);
1860 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1861 struct rb_root *root);
1862 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1863 struct rb_root *root, unsigned long start, unsigned long last);
1864 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1865 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1866 #ifdef CONFIG_DEBUG_VM_RB
1867 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1870 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1871 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1872 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1875 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1876 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1877 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1878 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1879 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1880 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1881 struct mempolicy *, struct vm_userfaultfd_ctx, const char __user *);
1882 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1883 extern int split_vma(struct mm_struct *,
1884 struct vm_area_struct *, unsigned long addr, int new_below);
1885 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1886 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1887 struct rb_node **, struct rb_node *);
1888 extern void unlink_file_vma(struct vm_area_struct *);
1889 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1890 unsigned long addr, unsigned long len, pgoff_t pgoff,
1891 bool *need_rmap_locks);
1892 extern void exit_mmap(struct mm_struct *);
1894 static inline int check_data_rlimit(unsigned long rlim,
1896 unsigned long start,
1897 unsigned long end_data,
1898 unsigned long start_data)
1900 if (rlim < RLIM_INFINITY) {
1901 if (((new - start) + (end_data - start_data)) > rlim)
1908 extern int mm_take_all_locks(struct mm_struct *mm);
1909 extern void mm_drop_all_locks(struct mm_struct *mm);
1911 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1912 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1913 extern struct file *get_task_exe_file(struct task_struct *task);
1915 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1916 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1917 unsigned long addr, unsigned long len,
1918 unsigned long flags,
1919 const struct vm_special_mapping *spec);
1920 /* This is an obsolete alternative to _install_special_mapping. */
1921 extern int install_special_mapping(struct mm_struct *mm,
1922 unsigned long addr, unsigned long len,
1923 unsigned long flags, struct page **pages);
1925 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1927 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1928 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1929 extern unsigned long do_mmap(struct file *file, unsigned long addr,
1930 unsigned long len, unsigned long prot, unsigned long flags,
1931 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
1932 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1934 static inline unsigned long
1935 do_mmap_pgoff(struct file *file, unsigned long addr,
1936 unsigned long len, unsigned long prot, unsigned long flags,
1937 unsigned long pgoff, unsigned long *populate)
1939 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
1943 extern int __mm_populate(unsigned long addr, unsigned long len,
1945 static inline void mm_populate(unsigned long addr, unsigned long len)
1948 (void) __mm_populate(addr, len, 1);
1951 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1954 /* These take the mm semaphore themselves */
1955 extern unsigned long vm_brk(unsigned long, unsigned long);
1956 extern int vm_munmap(unsigned long, size_t);
1957 extern unsigned long vm_mmap(struct file *, unsigned long,
1958 unsigned long, unsigned long,
1959 unsigned long, unsigned long);
1961 struct vm_unmapped_area_info {
1962 #define VM_UNMAPPED_AREA_TOPDOWN 1
1963 unsigned long flags;
1964 unsigned long length;
1965 unsigned long low_limit;
1966 unsigned long high_limit;
1967 unsigned long align_mask;
1968 unsigned long align_offset;
1971 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1972 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1975 * Search for an unmapped address range.
1977 * We are looking for a range that:
1978 * - does not intersect with any VMA;
1979 * - is contained within the [low_limit, high_limit) interval;
1980 * - is at least the desired size.
1981 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1983 static inline unsigned long
1984 vm_unmapped_area(struct vm_unmapped_area_info *info)
1986 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
1987 return unmapped_area_topdown(info);
1989 return unmapped_area(info);
1993 extern void truncate_inode_pages(struct address_space *, loff_t);
1994 extern void truncate_inode_pages_range(struct address_space *,
1995 loff_t lstart, loff_t lend);
1996 extern void truncate_inode_pages_final(struct address_space *);
1998 /* generic vm_area_ops exported for stackable file systems */
1999 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
2000 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
2001 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
2003 /* mm/page-writeback.c */
2004 int write_one_page(struct page *page, int wait);
2005 void task_dirty_inc(struct task_struct *tsk);
2008 #define VM_MAX_READAHEAD 128 /* kbytes */
2009 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2011 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
2012 pgoff_t offset, unsigned long nr_to_read);
2014 void page_cache_sync_readahead(struct address_space *mapping,
2015 struct file_ra_state *ra,
2018 unsigned long size);
2020 void page_cache_async_readahead(struct address_space *mapping,
2021 struct file_ra_state *ra,
2025 unsigned long size);
2027 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2028 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
2030 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2031 extern int expand_downwards(struct vm_area_struct *vma,
2032 unsigned long address);
2034 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
2036 #define expand_upwards(vma, address) (0)
2039 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2040 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2041 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2042 struct vm_area_struct **pprev);
2044 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2045 NULL if none. Assume start_addr < end_addr. */
2046 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2048 struct vm_area_struct * vma = find_vma(mm,start_addr);
2050 if (vma && end_addr <= vma->vm_start)
2055 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2057 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2060 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2061 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2062 unsigned long vm_start, unsigned long vm_end)
2064 struct vm_area_struct *vma = find_vma(mm, vm_start);
2066 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2073 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2074 void vma_set_page_prot(struct vm_area_struct *vma);
2076 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2080 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2082 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2086 #ifdef CONFIG_NUMA_BALANCING
2087 unsigned long change_prot_numa(struct vm_area_struct *vma,
2088 unsigned long start, unsigned long end);
2091 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2092 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2093 unsigned long pfn, unsigned long size, pgprot_t);
2094 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2095 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2097 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2099 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2102 struct page *follow_page_mask(struct vm_area_struct *vma,
2103 unsigned long address, unsigned int foll_flags,
2104 unsigned int *page_mask);
2106 static inline struct page *follow_page(struct vm_area_struct *vma,
2107 unsigned long address, unsigned int foll_flags)
2109 unsigned int unused_page_mask;
2110 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2113 #define FOLL_WRITE 0x01 /* check pte is writable */
2114 #define FOLL_TOUCH 0x02 /* mark page accessed */
2115 #define FOLL_GET 0x04 /* do get_page on page */
2116 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2117 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2118 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2119 * and return without waiting upon it */
2120 #define FOLL_POPULATE 0x40 /* fault in page */
2121 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2122 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2123 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2124 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2125 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2126 #define FOLL_MLOCK 0x1000 /* lock present pages */
2127 #define FOLL_COW 0x4000 /* internal GUP flag */
2129 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2131 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2132 unsigned long size, pte_fn_t fn, void *data);
2134 #ifdef CONFIG_PROC_FS
2135 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2137 static inline void vm_stat_account(struct mm_struct *mm,
2138 unsigned long flags, struct file *file, long pages)
2140 mm->total_vm += pages;
2142 #endif /* CONFIG_PROC_FS */
2144 #ifdef CONFIG_DEBUG_PAGEALLOC
2145 extern bool _debug_pagealloc_enabled;
2146 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2148 static inline bool debug_pagealloc_enabled(void)
2150 return _debug_pagealloc_enabled;
2154 kernel_map_pages(struct page *page, int numpages, int enable)
2156 if (!debug_pagealloc_enabled())
2159 __kernel_map_pages(page, numpages, enable);
2161 #ifdef CONFIG_HIBERNATION
2162 extern bool kernel_page_present(struct page *page);
2163 #endif /* CONFIG_HIBERNATION */
2166 kernel_map_pages(struct page *page, int numpages, int enable) {}
2167 #ifdef CONFIG_HIBERNATION
2168 static inline bool kernel_page_present(struct page *page) { return true; }
2169 #endif /* CONFIG_HIBERNATION */
2172 #ifdef __HAVE_ARCH_GATE_AREA
2173 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2174 extern int in_gate_area_no_mm(unsigned long addr);
2175 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2177 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2181 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2182 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2186 #endif /* __HAVE_ARCH_GATE_AREA */
2188 #ifdef CONFIG_SYSCTL
2189 extern int sysctl_drop_caches;
2190 int drop_caches_sysctl_handler(struct ctl_table *, int,
2191 void __user *, size_t *, loff_t *);
2194 void drop_slab(void);
2195 void drop_slab_node(int nid);
2198 #define randomize_va_space 0
2200 extern int randomize_va_space;
2203 const char * arch_vma_name(struct vm_area_struct *vma);
2204 void print_vma_addr(char *prefix, unsigned long rip);
2206 void sparse_mem_maps_populate_node(struct page **map_map,
2207 unsigned long pnum_begin,
2208 unsigned long pnum_end,
2209 unsigned long map_count,
2212 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2213 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2214 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2215 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2216 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2217 void *vmemmap_alloc_block(unsigned long size, int node);
2218 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2219 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2220 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2222 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2223 void vmemmap_populate_print_last(void);
2224 #ifdef CONFIG_MEMORY_HOTPLUG
2225 void vmemmap_free(unsigned long start, unsigned long end);
2227 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2228 unsigned long size);
2231 MF_COUNT_INCREASED = 1 << 0,
2232 MF_ACTION_REQUIRED = 1 << 1,
2233 MF_MUST_KILL = 1 << 2,
2234 MF_SOFT_OFFLINE = 1 << 3,
2236 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2237 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2238 extern int unpoison_memory(unsigned long pfn);
2239 extern int get_hwpoison_page(struct page *page);
2240 extern void put_hwpoison_page(struct page *page);
2241 extern int sysctl_memory_failure_early_kill;
2242 extern int sysctl_memory_failure_recovery;
2243 extern void shake_page(struct page *p, int access);
2244 extern atomic_long_t num_poisoned_pages;
2245 extern int soft_offline_page(struct page *page, int flags);
2249 * Error handlers for various types of pages.
2252 MF_IGNORED, /* Error: cannot be handled */
2253 MF_FAILED, /* Error: handling failed */
2254 MF_DELAYED, /* Will be handled later */
2255 MF_RECOVERED, /* Successfully recovered */
2258 enum mf_action_page_type {
2260 MF_MSG_KERNEL_HIGH_ORDER,
2262 MF_MSG_DIFFERENT_COMPOUND,
2263 MF_MSG_POISONED_HUGE,
2266 MF_MSG_UNMAP_FAILED,
2267 MF_MSG_DIRTY_SWAPCACHE,
2268 MF_MSG_CLEAN_SWAPCACHE,
2269 MF_MSG_DIRTY_MLOCKED_LRU,
2270 MF_MSG_CLEAN_MLOCKED_LRU,
2271 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2272 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2275 MF_MSG_TRUNCATED_LRU,
2281 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2282 extern void clear_huge_page(struct page *page,
2284 unsigned int pages_per_huge_page);
2285 extern void copy_user_huge_page(struct page *dst, struct page *src,
2286 unsigned long addr, struct vm_area_struct *vma,
2287 unsigned int pages_per_huge_page);
2288 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2290 extern struct page_ext_operations debug_guardpage_ops;
2291 extern struct page_ext_operations page_poisoning_ops;
2293 #ifdef CONFIG_DEBUG_PAGEALLOC
2294 extern unsigned int _debug_guardpage_minorder;
2295 extern bool _debug_guardpage_enabled;
2297 static inline unsigned int debug_guardpage_minorder(void)
2299 return _debug_guardpage_minorder;
2302 static inline bool debug_guardpage_enabled(void)
2304 return _debug_guardpage_enabled;
2307 static inline bool page_is_guard(struct page *page)
2309 struct page_ext *page_ext;
2311 if (!debug_guardpage_enabled())
2314 page_ext = lookup_page_ext(page);
2315 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2318 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2319 static inline bool debug_guardpage_enabled(void) { return false; }
2320 static inline bool page_is_guard(struct page *page) { return false; }
2321 #endif /* CONFIG_DEBUG_PAGEALLOC */
2323 #if MAX_NUMNODES > 1
2324 void __init setup_nr_node_ids(void);
2326 static inline void setup_nr_node_ids(void) {}
2329 #endif /* __KERNEL__ */
2330 #endif /* _LINUX_MM_H */