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>
26 struct anon_vma_chain;
29 struct writeback_control;
32 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
33 extern unsigned long max_mapnr;
35 static inline void set_max_mapnr(unsigned long limit)
40 static inline void set_max_mapnr(unsigned long limit) { }
43 extern unsigned long totalram_pages;
44 extern void * high_memory;
45 extern int page_cluster;
48 extern int sysctl_legacy_va_layout;
50 #define sysctl_legacy_va_layout 0
54 #include <asm/pgtable.h>
55 #include <asm/processor.h>
58 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
62 * To prevent common memory management code establishing
63 * a zero page mapping on a read fault.
64 * This macro should be defined within <asm/pgtable.h>.
65 * s390 does this to prevent multiplexing of hardware bits
66 * related to the physical page in case of virtualization.
68 #ifndef mm_forbids_zeropage
69 #define mm_forbids_zeropage(X) (0)
72 extern unsigned long sysctl_user_reserve_kbytes;
73 extern unsigned long sysctl_admin_reserve_kbytes;
75 extern int sysctl_overcommit_memory;
76 extern int sysctl_overcommit_ratio;
77 extern unsigned long sysctl_overcommit_kbytes;
79 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
81 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
84 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
86 /* to align the pointer to the (next) page boundary */
87 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
89 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
90 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
93 * Linux kernel virtual memory manager primitives.
94 * The idea being to have a "virtual" mm in the same way
95 * we have a virtual fs - giving a cleaner interface to the
96 * mm details, and allowing different kinds of memory mappings
97 * (from shared memory to executable loading to arbitrary
101 extern struct kmem_cache *vm_area_cachep;
104 extern struct rb_root nommu_region_tree;
105 extern struct rw_semaphore nommu_region_sem;
107 extern unsigned int kobjsize(const void *objp);
111 * vm_flags in vm_area_struct, see mm_types.h.
113 #define VM_NONE 0x00000000
115 #define VM_READ 0x00000001 /* currently active flags */
116 #define VM_WRITE 0x00000002
117 #define VM_EXEC 0x00000004
118 #define VM_SHARED 0x00000008
120 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
121 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
122 #define VM_MAYWRITE 0x00000020
123 #define VM_MAYEXEC 0x00000040
124 #define VM_MAYSHARE 0x00000080
126 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
127 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
128 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
129 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
130 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
132 #define VM_LOCKED 0x00002000
133 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
135 /* Used by sys_madvise() */
136 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
137 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
139 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
140 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
141 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
142 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
143 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
144 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
145 #define VM_ARCH_2 0x02000000
146 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
148 #ifdef CONFIG_MEM_SOFT_DIRTY
149 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
151 # define VM_SOFTDIRTY 0
154 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
155 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
156 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
157 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
159 #if defined(CONFIG_X86)
160 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
161 #elif defined(CONFIG_PPC)
162 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
163 #elif defined(CONFIG_PARISC)
164 # define VM_GROWSUP VM_ARCH_1
165 #elif defined(CONFIG_METAG)
166 # define VM_GROWSUP VM_ARCH_1
167 #elif defined(CONFIG_IA64)
168 # define VM_GROWSUP VM_ARCH_1
169 #elif !defined(CONFIG_MMU)
170 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
173 #if defined(CONFIG_X86)
174 /* MPX specific bounds table or bounds directory */
175 # define VM_MPX VM_ARCH_2
179 # define VM_GROWSUP VM_NONE
182 /* Bits set in the VMA until the stack is in its final location */
183 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
185 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
186 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
189 #ifdef CONFIG_STACK_GROWSUP
190 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
192 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
196 * Special vmas that are non-mergable, non-mlock()able.
197 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
199 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
201 /* This mask defines which mm->def_flags a process can inherit its parent */
202 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
205 * mapping from the currently active vm_flags protection bits (the
206 * low four bits) to a page protection mask..
208 extern pgprot_t protection_map[16];
210 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
211 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
212 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
213 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
214 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
215 #define FAULT_FLAG_TRIED 0x20 /* Second try */
216 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
219 * vm_fault is filled by the the pagefault handler and passed to the vma's
220 * ->fault function. The vma's ->fault is responsible for returning a bitmask
221 * of VM_FAULT_xxx flags that give details about how the fault was handled.
223 * pgoff should be used in favour of virtual_address, if possible.
226 unsigned int flags; /* FAULT_FLAG_xxx flags */
227 pgoff_t pgoff; /* Logical page offset based on vma */
228 void __user *virtual_address; /* Faulting virtual address */
230 struct page *cow_page; /* Handler may choose to COW */
231 struct page *page; /* ->fault handlers should return a
232 * page here, unless VM_FAULT_NOPAGE
233 * is set (which is also implied by
236 /* for ->map_pages() only */
237 pgoff_t max_pgoff; /* map pages for offset from pgoff till
238 * max_pgoff inclusive */
239 pte_t *pte; /* pte entry associated with ->pgoff */
243 * These are the virtual MM functions - opening of an area, closing and
244 * unmapping it (needed to keep files on disk up-to-date etc), pointer
245 * to the functions called when a no-page or a wp-page exception occurs.
247 struct vm_operations_struct {
248 void (*open)(struct vm_area_struct * area);
249 void (*close)(struct vm_area_struct * area);
250 int (*mremap)(struct vm_area_struct * area);
251 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
252 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
253 pmd_t *, unsigned int flags);
254 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
256 /* notification that a previously read-only page is about to become
257 * writable, if an error is returned it will cause a SIGBUS */
258 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
260 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
261 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
263 /* called by access_process_vm when get_user_pages() fails, typically
264 * for use by special VMAs that can switch between memory and hardware
266 int (*access)(struct vm_area_struct *vma, unsigned long addr,
267 void *buf, int len, int write);
269 /* Called by the /proc/PID/maps code to ask the vma whether it
270 * has a special name. Returning non-NULL will also cause this
271 * vma to be dumped unconditionally. */
272 const char *(*name)(struct vm_area_struct *vma);
276 * set_policy() op must add a reference to any non-NULL @new mempolicy
277 * to hold the policy upon return. Caller should pass NULL @new to
278 * remove a policy and fall back to surrounding context--i.e. do not
279 * install a MPOL_DEFAULT policy, nor the task or system default
282 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
285 * get_policy() op must add reference [mpol_get()] to any policy at
286 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
287 * in mm/mempolicy.c will do this automatically.
288 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
289 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
290 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
291 * must return NULL--i.e., do not "fallback" to task or system default
294 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
298 * Called by vm_normal_page() for special PTEs to find the
299 * page for @addr. This is useful if the default behavior
300 * (using pte_page()) would not find the correct page.
302 struct page *(*find_special_page)(struct vm_area_struct *vma,
309 #define page_private(page) ((page)->private)
310 #define set_page_private(page, v) ((page)->private = (v))
313 * FIXME: take this include out, include page-flags.h in
314 * files which need it (119 of them)
316 #include <linux/page-flags.h>
317 #include <linux/huge_mm.h>
320 * Methods to modify the page usage count.
322 * What counts for a page usage:
323 * - cache mapping (page->mapping)
324 * - private data (page->private)
325 * - page mapped in a task's page tables, each mapping
326 * is counted separately
328 * Also, many kernel routines increase the page count before a critical
329 * routine so they can be sure the page doesn't go away from under them.
333 * Drop a ref, return true if the refcount fell to zero (the page has no users)
335 static inline int put_page_testzero(struct page *page)
337 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
338 return atomic_dec_and_test(&page->_count);
342 * Try to grab a ref unless the page has a refcount of zero, return false if
344 * This can be called when MMU is off so it must not access
345 * any of the virtual mappings.
347 static inline int get_page_unless_zero(struct page *page)
349 return atomic_inc_not_zero(&page->_count);
352 extern int page_is_ram(unsigned long pfn);
360 int region_intersects(resource_size_t offset, size_t size, const char *type);
362 /* Support for virtually mapped pages */
363 struct page *vmalloc_to_page(const void *addr);
364 unsigned long vmalloc_to_pfn(const void *addr);
367 * Determine if an address is within the vmalloc range
369 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
370 * is no special casing required.
372 static inline int is_vmalloc_addr(const void *x)
375 unsigned long addr = (unsigned long)x;
377 return addr >= VMALLOC_START && addr < VMALLOC_END;
383 extern int is_vmalloc_or_module_addr(const void *x);
385 static inline int is_vmalloc_or_module_addr(const void *x)
391 extern void kvfree(const void *addr);
393 static inline void compound_lock(struct page *page)
395 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
396 VM_BUG_ON_PAGE(PageSlab(page), page);
397 bit_spin_lock(PG_compound_lock, &page->flags);
401 static inline void compound_unlock(struct page *page)
403 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
404 VM_BUG_ON_PAGE(PageSlab(page), page);
405 bit_spin_unlock(PG_compound_lock, &page->flags);
409 static inline unsigned long compound_lock_irqsave(struct page *page)
411 unsigned long uninitialized_var(flags);
412 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
413 local_irq_save(flags);
419 static inline void compound_unlock_irqrestore(struct page *page,
422 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
423 compound_unlock(page);
424 local_irq_restore(flags);
428 static inline struct page *compound_head_by_tail(struct page *tail)
430 struct page *head = tail->first_page;
433 * page->first_page may be a dangling pointer to an old
434 * compound page, so recheck that it is still a tail
435 * page before returning.
438 if (likely(PageTail(tail)))
444 * Since either compound page could be dismantled asynchronously in THP
445 * or we access asynchronously arbitrary positioned struct page, there
446 * would be tail flag race. To handle this race, we should call
447 * smp_rmb() before checking tail flag. compound_head_by_tail() did it.
449 static inline struct page *compound_head(struct page *page)
451 if (unlikely(PageTail(page)))
452 return compound_head_by_tail(page);
457 * If we access compound page synchronously such as access to
458 * allocated page, there is no need to handle tail flag race, so we can
459 * check tail flag directly without any synchronization primitive.
461 static inline struct page *compound_head_fast(struct page *page)
463 if (unlikely(PageTail(page)))
464 return page->first_page;
469 * The atomic page->_mapcount, starts from -1: so that transitions
470 * both from it and to it can be tracked, using atomic_inc_and_test
471 * and atomic_add_negative(-1).
473 static inline void page_mapcount_reset(struct page *page)
475 atomic_set(&(page)->_mapcount, -1);
478 static inline int page_mapcount(struct page *page)
480 VM_BUG_ON_PAGE(PageSlab(page), page);
481 return atomic_read(&page->_mapcount) + 1;
484 static inline int page_count(struct page *page)
486 return atomic_read(&compound_head(page)->_count);
489 static inline bool __compound_tail_refcounted(struct page *page)
491 return PageAnon(page) && !PageSlab(page) && !PageHeadHuge(page);
495 * This takes a head page as parameter and tells if the
496 * tail page reference counting can be skipped.
498 * For this to be safe, PageSlab and PageHeadHuge must remain true on
499 * any given page where they return true here, until all tail pins
500 * have been released.
502 static inline bool compound_tail_refcounted(struct page *page)
504 VM_BUG_ON_PAGE(!PageHead(page), page);
505 return __compound_tail_refcounted(page);
508 static inline void get_huge_page_tail(struct page *page)
511 * __split_huge_page_refcount() cannot run from under us.
513 VM_BUG_ON_PAGE(!PageTail(page), page);
514 VM_BUG_ON_PAGE(page_mapcount(page) < 0, page);
515 VM_BUG_ON_PAGE(atomic_read(&page->_count) != 0, page);
516 if (compound_tail_refcounted(page->first_page))
517 atomic_inc(&page->_mapcount);
520 extern bool __get_page_tail(struct page *page);
522 static inline void get_page(struct page *page)
524 if (unlikely(PageTail(page)))
525 if (likely(__get_page_tail(page)))
528 * Getting a normal page or the head of a compound page
529 * requires to already have an elevated page->_count.
531 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
532 atomic_inc(&page->_count);
535 static inline struct page *virt_to_head_page(const void *x)
537 struct page *page = virt_to_page(x);
540 * We don't need to worry about synchronization of tail flag
541 * when we call virt_to_head_page() since it is only called for
542 * already allocated page and this page won't be freed until
543 * this virt_to_head_page() is finished. So use _fast variant.
545 return compound_head_fast(page);
549 * Setup the page count before being freed into the page allocator for
550 * the first time (boot or memory hotplug)
552 static inline void init_page_count(struct page *page)
554 atomic_set(&page->_count, 1);
557 void put_page(struct page *page);
558 void put_pages_list(struct list_head *pages);
560 void split_page(struct page *page, unsigned int order);
561 int split_free_page(struct page *page);
564 * Compound pages have a destructor function. Provide a
565 * prototype for that function and accessor functions.
566 * These are _only_ valid on the head of a PG_compound page.
569 static inline void set_compound_page_dtor(struct page *page,
570 compound_page_dtor *dtor)
572 page[1].compound_dtor = dtor;
575 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
577 return page[1].compound_dtor;
580 static inline int compound_order(struct page *page)
584 return page[1].compound_order;
587 static inline void set_compound_order(struct page *page, unsigned long order)
589 page[1].compound_order = order;
594 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
595 * servicing faults for write access. In the normal case, do always want
596 * pte_mkwrite. But get_user_pages can cause write faults for mappings
597 * that do not have writing enabled, when used by access_process_vm.
599 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
601 if (likely(vma->vm_flags & VM_WRITE))
602 pte = pte_mkwrite(pte);
606 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
607 struct page *page, pte_t *pte, bool write, bool anon);
611 * Multiple processes may "see" the same page. E.g. for untouched
612 * mappings of /dev/null, all processes see the same page full of
613 * zeroes, and text pages of executables and shared libraries have
614 * only one copy in memory, at most, normally.
616 * For the non-reserved pages, page_count(page) denotes a reference count.
617 * page_count() == 0 means the page is free. page->lru is then used for
618 * freelist management in the buddy allocator.
619 * page_count() > 0 means the page has been allocated.
621 * Pages are allocated by the slab allocator in order to provide memory
622 * to kmalloc and kmem_cache_alloc. In this case, the management of the
623 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
624 * unless a particular usage is carefully commented. (the responsibility of
625 * freeing the kmalloc memory is the caller's, of course).
627 * A page may be used by anyone else who does a __get_free_page().
628 * In this case, page_count still tracks the references, and should only
629 * be used through the normal accessor functions. The top bits of page->flags
630 * and page->virtual store page management information, but all other fields
631 * are unused and could be used privately, carefully. The management of this
632 * page is the responsibility of the one who allocated it, and those who have
633 * subsequently been given references to it.
635 * The other pages (we may call them "pagecache pages") are completely
636 * managed by the Linux memory manager: I/O, buffers, swapping etc.
637 * The following discussion applies only to them.
639 * A pagecache page contains an opaque `private' member, which belongs to the
640 * page's address_space. Usually, this is the address of a circular list of
641 * the page's disk buffers. PG_private must be set to tell the VM to call
642 * into the filesystem to release these pages.
644 * A page may belong to an inode's memory mapping. In this case, page->mapping
645 * is the pointer to the inode, and page->index is the file offset of the page,
646 * in units of PAGE_CACHE_SIZE.
648 * If pagecache pages are not associated with an inode, they are said to be
649 * anonymous pages. These may become associated with the swapcache, and in that
650 * case PG_swapcache is set, and page->private is an offset into the swapcache.
652 * In either case (swapcache or inode backed), the pagecache itself holds one
653 * reference to the page. Setting PG_private should also increment the
654 * refcount. The each user mapping also has a reference to the page.
656 * The pagecache pages are stored in a per-mapping radix tree, which is
657 * rooted at mapping->page_tree, and indexed by offset.
658 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
659 * lists, we instead now tag pages as dirty/writeback in the radix tree.
661 * All pagecache pages may be subject to I/O:
662 * - inode pages may need to be read from disk,
663 * - inode pages which have been modified and are MAP_SHARED may need
664 * to be written back to the inode on disk,
665 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
666 * modified may need to be swapped out to swap space and (later) to be read
671 * The zone field is never updated after free_area_init_core()
672 * sets it, so none of the operations on it need to be atomic.
675 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
676 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
677 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
678 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
679 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
682 * Define the bit shifts to access each section. For non-existent
683 * sections we define the shift as 0; that plus a 0 mask ensures
684 * the compiler will optimise away reference to them.
686 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
687 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
688 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
689 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
691 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
692 #ifdef NODE_NOT_IN_PAGE_FLAGS
693 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
694 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
695 SECTIONS_PGOFF : ZONES_PGOFF)
697 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
698 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
699 NODES_PGOFF : ZONES_PGOFF)
702 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
704 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
705 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
708 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
709 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
710 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
711 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
712 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
714 static inline enum zone_type page_zonenum(const struct page *page)
716 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
719 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
720 #define SECTION_IN_PAGE_FLAGS
724 * The identification function is mainly used by the buddy allocator for
725 * determining if two pages could be buddies. We are not really identifying
726 * the zone since we could be using the section number id if we do not have
727 * node id available in page flags.
728 * We only guarantee that it will return the same value for two combinable
731 static inline int page_zone_id(struct page *page)
733 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
736 static inline int zone_to_nid(struct zone *zone)
745 #ifdef NODE_NOT_IN_PAGE_FLAGS
746 extern int page_to_nid(const struct page *page);
748 static inline int page_to_nid(const struct page *page)
750 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
754 #ifdef CONFIG_NUMA_BALANCING
755 static inline int cpu_pid_to_cpupid(int cpu, int pid)
757 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
760 static inline int cpupid_to_pid(int cpupid)
762 return cpupid & LAST__PID_MASK;
765 static inline int cpupid_to_cpu(int cpupid)
767 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
770 static inline int cpupid_to_nid(int cpupid)
772 return cpu_to_node(cpupid_to_cpu(cpupid));
775 static inline bool cpupid_pid_unset(int cpupid)
777 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
780 static inline bool cpupid_cpu_unset(int cpupid)
782 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
785 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
787 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
790 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
791 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
792 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
794 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
797 static inline int page_cpupid_last(struct page *page)
799 return page->_last_cpupid;
801 static inline void page_cpupid_reset_last(struct page *page)
803 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
806 static inline int page_cpupid_last(struct page *page)
808 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
811 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
813 static inline void page_cpupid_reset_last(struct page *page)
815 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
817 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
818 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
820 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
821 #else /* !CONFIG_NUMA_BALANCING */
822 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
824 return page_to_nid(page); /* XXX */
827 static inline int page_cpupid_last(struct page *page)
829 return page_to_nid(page); /* XXX */
832 static inline int cpupid_to_nid(int cpupid)
837 static inline int cpupid_to_pid(int cpupid)
842 static inline int cpupid_to_cpu(int cpupid)
847 static inline int cpu_pid_to_cpupid(int nid, int pid)
852 static inline bool cpupid_pid_unset(int cpupid)
857 static inline void page_cpupid_reset_last(struct page *page)
861 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
865 #endif /* CONFIG_NUMA_BALANCING */
867 static inline struct zone *page_zone(const struct page *page)
869 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
872 #ifdef SECTION_IN_PAGE_FLAGS
873 static inline void set_page_section(struct page *page, unsigned long section)
875 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
876 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
879 static inline unsigned long page_to_section(const struct page *page)
881 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
885 static inline void set_page_zone(struct page *page, enum zone_type zone)
887 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
888 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
891 static inline void set_page_node(struct page *page, unsigned long node)
893 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
894 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
897 static inline void set_page_links(struct page *page, enum zone_type zone,
898 unsigned long node, unsigned long pfn)
900 set_page_zone(page, zone);
901 set_page_node(page, node);
902 #ifdef SECTION_IN_PAGE_FLAGS
903 set_page_section(page, pfn_to_section_nr(pfn));
908 * Some inline functions in vmstat.h depend on page_zone()
910 #include <linux/vmstat.h>
912 static __always_inline void *lowmem_page_address(const struct page *page)
914 return __va(PFN_PHYS(page_to_pfn(page)));
917 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
918 #define HASHED_PAGE_VIRTUAL
921 #if defined(WANT_PAGE_VIRTUAL)
922 static inline void *page_address(const struct page *page)
924 return page->virtual;
926 static inline void set_page_address(struct page *page, void *address)
928 page->virtual = address;
930 #define page_address_init() do { } while(0)
933 #if defined(HASHED_PAGE_VIRTUAL)
934 void *page_address(const struct page *page);
935 void set_page_address(struct page *page, void *virtual);
936 void page_address_init(void);
939 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
940 #define page_address(page) lowmem_page_address(page)
941 #define set_page_address(page, address) do { } while(0)
942 #define page_address_init() do { } while(0)
945 extern void *page_rmapping(struct page *page);
946 extern struct anon_vma *page_anon_vma(struct page *page);
947 extern struct address_space *page_mapping(struct page *page);
949 extern struct address_space *__page_file_mapping(struct page *);
952 struct address_space *page_file_mapping(struct page *page)
954 if (unlikely(PageSwapCache(page)))
955 return __page_file_mapping(page);
957 return page->mapping;
961 * Return the pagecache index of the passed page. Regular pagecache pages
962 * use ->index whereas swapcache pages use ->private
964 static inline pgoff_t page_index(struct page *page)
966 if (unlikely(PageSwapCache(page)))
967 return page_private(page);
971 extern pgoff_t __page_file_index(struct page *page);
974 * Return the file index of the page. Regular pagecache pages use ->index
975 * whereas swapcache pages use swp_offset(->private)
977 static inline pgoff_t page_file_index(struct page *page)
979 if (unlikely(PageSwapCache(page)))
980 return __page_file_index(page);
986 * Return true if this page is mapped into pagetables.
988 static inline int page_mapped(struct page *page)
990 return atomic_read(&(page)->_mapcount) >= 0;
994 * Return true only if the page has been allocated with
995 * ALLOC_NO_WATERMARKS and the low watermark was not
996 * met implying that the system is under some pressure.
998 static inline bool page_is_pfmemalloc(struct page *page)
1001 * Page index cannot be this large so this must be
1002 * a pfmemalloc page.
1004 return page->index == -1UL;
1008 * Only to be called by the page allocator on a freshly allocated
1011 static inline void set_page_pfmemalloc(struct page *page)
1016 static inline void clear_page_pfmemalloc(struct page *page)
1022 * Different kinds of faults, as returned by handle_mm_fault().
1023 * Used to decide whether a process gets delivered SIGBUS or
1024 * just gets major/minor fault counters bumped up.
1027 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
1029 #define VM_FAULT_OOM 0x0001
1030 #define VM_FAULT_SIGBUS 0x0002
1031 #define VM_FAULT_MAJOR 0x0004
1032 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1033 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1034 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1035 #define VM_FAULT_SIGSEGV 0x0040
1037 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1038 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1039 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1040 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1042 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1044 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1045 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1048 /* Encode hstate index for a hwpoisoned large page */
1049 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1050 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1053 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1055 extern void pagefault_out_of_memory(void);
1057 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1060 * Flags passed to show_mem() and show_free_areas() to suppress output in
1063 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1065 extern void show_free_areas(unsigned int flags);
1066 extern bool skip_free_areas_node(unsigned int flags, int nid);
1068 int shmem_zero_setup(struct vm_area_struct *);
1070 bool shmem_mapping(struct address_space *mapping);
1072 static inline bool shmem_mapping(struct address_space *mapping)
1078 extern int can_do_mlock(void);
1079 extern int user_shm_lock(size_t, struct user_struct *);
1080 extern void user_shm_unlock(size_t, struct user_struct *);
1083 * Parameter block passed down to zap_pte_range in exceptional cases.
1085 struct zap_details {
1086 struct address_space *check_mapping; /* Check page->mapping if set */
1087 pgoff_t first_index; /* Lowest page->index to unmap */
1088 pgoff_t last_index; /* Highest page->index to unmap */
1091 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1094 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1095 unsigned long size);
1096 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1097 unsigned long size, struct zap_details *);
1098 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1099 unsigned long start, unsigned long end);
1102 * mm_walk - callbacks for walk_page_range
1103 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1104 * this handler is required to be able to handle
1105 * pmd_trans_huge() pmds. They may simply choose to
1106 * split_huge_page() instead of handling it explicitly.
1107 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1108 * @pte_hole: if set, called for each hole at all levels
1109 * @hugetlb_entry: if set, called for each hugetlb entry
1110 * @test_walk: caller specific callback function to determine whether
1111 * we walk over the current vma or not. A positive returned
1112 * value means "do page table walk over the current vma,"
1113 * and a negative one means "abort current page table walk
1114 * right now." 0 means "skip the current vma."
1115 * @mm: mm_struct representing the target process of page table walk
1116 * @vma: vma currently walked (NULL if walking outside vmas)
1117 * @private: private data for callbacks' usage
1119 * (see the comment on walk_page_range() for more details)
1122 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1123 unsigned long next, struct mm_walk *walk);
1124 int (*pte_entry)(pte_t *pte, unsigned long addr,
1125 unsigned long next, struct mm_walk *walk);
1126 int (*pte_hole)(unsigned long addr, unsigned long next,
1127 struct mm_walk *walk);
1128 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1129 unsigned long addr, unsigned long next,
1130 struct mm_walk *walk);
1131 int (*test_walk)(unsigned long addr, unsigned long next,
1132 struct mm_walk *walk);
1133 struct mm_struct *mm;
1134 struct vm_area_struct *vma;
1138 int walk_page_range(unsigned long addr, unsigned long end,
1139 struct mm_walk *walk);
1140 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1141 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1142 unsigned long end, unsigned long floor, unsigned long ceiling);
1143 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1144 struct vm_area_struct *vma);
1145 void unmap_mapping_range(struct address_space *mapping,
1146 loff_t const holebegin, loff_t const holelen, int even_cows);
1147 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1148 unsigned long *pfn);
1149 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1150 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1151 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1152 void *buf, int len, int write);
1154 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1155 loff_t const holebegin, loff_t const holelen)
1157 unmap_mapping_range(mapping, holebegin, holelen, 0);
1160 extern void truncate_pagecache(struct inode *inode, loff_t new);
1161 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1162 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1163 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1164 int truncate_inode_page(struct address_space *mapping, struct page *page);
1165 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1166 int invalidate_inode_page(struct page *page);
1169 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1170 unsigned long address, unsigned int flags);
1171 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1172 unsigned long address, unsigned int fault_flags);
1174 static inline int handle_mm_fault(struct mm_struct *mm,
1175 struct vm_area_struct *vma, unsigned long address,
1178 /* should never happen if there's no MMU */
1180 return VM_FAULT_SIGBUS;
1182 static inline int fixup_user_fault(struct task_struct *tsk,
1183 struct mm_struct *mm, unsigned long address,
1184 unsigned int fault_flags)
1186 /* should never happen if there's no MMU */
1192 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1193 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1194 void *buf, int len, int write);
1196 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1197 unsigned long start, unsigned long nr_pages,
1198 unsigned int foll_flags, struct page **pages,
1199 struct vm_area_struct **vmas, int *nonblocking);
1200 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1201 unsigned long start, unsigned long nr_pages,
1202 int write, int force, struct page **pages,
1203 struct vm_area_struct **vmas);
1204 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1205 unsigned long start, unsigned long nr_pages,
1206 int write, int force, struct page **pages,
1208 long __get_user_pages_unlocked(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 unsigned int gup_flags);
1212 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1213 unsigned long start, unsigned long nr_pages,
1214 int write, int force, struct page **pages);
1215 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1216 struct page **pages);
1218 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1219 struct page **pages);
1220 int get_kernel_page(unsigned long start, int write, struct page **pages);
1221 struct page *get_dump_page(unsigned long addr);
1223 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1224 extern void do_invalidatepage(struct page *page, unsigned int offset,
1225 unsigned int length);
1227 int __set_page_dirty_nobuffers(struct page *page);
1228 int __set_page_dirty_no_writeback(struct page *page);
1229 int redirty_page_for_writepage(struct writeback_control *wbc,
1231 void account_page_dirtied(struct page *page, struct address_space *mapping,
1232 struct mem_cgroup *memcg);
1233 void account_page_cleaned(struct page *page, struct address_space *mapping,
1234 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1235 int set_page_dirty(struct page *page);
1236 int set_page_dirty_lock(struct page *page);
1237 void cancel_dirty_page(struct page *page);
1238 int clear_page_dirty_for_io(struct page *page);
1240 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1242 /* Is the vma a continuation of the stack vma above it? */
1243 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1245 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1248 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1250 return !vma->vm_ops;
1253 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1256 return (vma->vm_flags & VM_GROWSDOWN) &&
1257 (vma->vm_start == addr) &&
1258 !vma_growsdown(vma->vm_prev, addr);
1261 /* Is the vma a continuation of the stack vma below it? */
1262 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1264 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1267 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1270 return (vma->vm_flags & VM_GROWSUP) &&
1271 (vma->vm_end == addr) &&
1272 !vma_growsup(vma->vm_next, addr);
1275 extern struct task_struct *task_of_stack(struct task_struct *task,
1276 struct vm_area_struct *vma, bool in_group);
1278 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1279 unsigned long old_addr, struct vm_area_struct *new_vma,
1280 unsigned long new_addr, unsigned long len,
1281 bool need_rmap_locks);
1282 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1283 unsigned long end, pgprot_t newprot,
1284 int dirty_accountable, int prot_numa);
1285 extern int mprotect_fixup(struct vm_area_struct *vma,
1286 struct vm_area_struct **pprev, unsigned long start,
1287 unsigned long end, unsigned long newflags);
1290 * doesn't attempt to fault and will return short.
1292 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1293 struct page **pages);
1295 * per-process(per-mm_struct) statistics.
1297 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1299 long val = atomic_long_read(&mm->rss_stat.count[member]);
1301 #ifdef SPLIT_RSS_COUNTING
1303 * counter is updated in asynchronous manner and may go to minus.
1304 * But it's never be expected number for users.
1309 return (unsigned long)val;
1312 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1314 atomic_long_add(value, &mm->rss_stat.count[member]);
1317 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1319 atomic_long_inc(&mm->rss_stat.count[member]);
1322 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1324 atomic_long_dec(&mm->rss_stat.count[member]);
1327 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1329 return get_mm_counter(mm, MM_FILEPAGES) +
1330 get_mm_counter(mm, MM_ANONPAGES);
1333 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1335 return max(mm->hiwater_rss, get_mm_rss(mm));
1338 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1340 return max(mm->hiwater_vm, mm->total_vm);
1343 static inline void update_hiwater_rss(struct mm_struct *mm)
1345 unsigned long _rss = get_mm_rss(mm);
1347 if ((mm)->hiwater_rss < _rss)
1348 (mm)->hiwater_rss = _rss;
1351 static inline void update_hiwater_vm(struct mm_struct *mm)
1353 if (mm->hiwater_vm < mm->total_vm)
1354 mm->hiwater_vm = mm->total_vm;
1357 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1359 mm->hiwater_rss = get_mm_rss(mm);
1362 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1363 struct mm_struct *mm)
1365 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1367 if (*maxrss < hiwater_rss)
1368 *maxrss = hiwater_rss;
1371 #if defined(SPLIT_RSS_COUNTING)
1372 void sync_mm_rss(struct mm_struct *mm);
1374 static inline void sync_mm_rss(struct mm_struct *mm)
1379 int vma_wants_writenotify(struct vm_area_struct *vma);
1381 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1383 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1387 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1391 #ifdef __PAGETABLE_PUD_FOLDED
1392 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1393 unsigned long address)
1398 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1401 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1402 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1403 unsigned long address)
1408 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1410 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1415 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1416 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1419 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1421 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1423 atomic_long_set(&mm->nr_pmds, 0);
1426 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1428 return atomic_long_read(&mm->nr_pmds);
1431 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1433 atomic_long_inc(&mm->nr_pmds);
1436 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1438 atomic_long_dec(&mm->nr_pmds);
1442 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1443 pmd_t *pmd, unsigned long address);
1444 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1447 * The following ifdef needed to get the 4level-fixup.h header to work.
1448 * Remove it when 4level-fixup.h has been removed.
1450 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1451 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1453 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1454 NULL: pud_offset(pgd, address);
1457 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1459 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1460 NULL: pmd_offset(pud, address);
1462 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1464 #if USE_SPLIT_PTE_PTLOCKS
1465 #if ALLOC_SPLIT_PTLOCKS
1466 void __init ptlock_cache_init(void);
1467 extern bool ptlock_alloc(struct page *page);
1468 extern void ptlock_free(struct page *page);
1470 static inline spinlock_t *ptlock_ptr(struct page *page)
1474 #else /* ALLOC_SPLIT_PTLOCKS */
1475 static inline void ptlock_cache_init(void)
1479 static inline bool ptlock_alloc(struct page *page)
1484 static inline void ptlock_free(struct page *page)
1488 static inline spinlock_t *ptlock_ptr(struct page *page)
1492 #endif /* ALLOC_SPLIT_PTLOCKS */
1494 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1496 return ptlock_ptr(pmd_page(*pmd));
1499 static inline bool ptlock_init(struct page *page)
1502 * prep_new_page() initialize page->private (and therefore page->ptl)
1503 * with 0. Make sure nobody took it in use in between.
1505 * It can happen if arch try to use slab for page table allocation:
1506 * slab code uses page->slab_cache and page->first_page (for tail
1507 * pages), which share storage with page->ptl.
1509 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1510 if (!ptlock_alloc(page))
1512 spin_lock_init(ptlock_ptr(page));
1516 /* Reset page->mapping so free_pages_check won't complain. */
1517 static inline void pte_lock_deinit(struct page *page)
1519 page->mapping = NULL;
1523 #else /* !USE_SPLIT_PTE_PTLOCKS */
1525 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1527 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1529 return &mm->page_table_lock;
1531 static inline void ptlock_cache_init(void) {}
1532 static inline bool ptlock_init(struct page *page) { return true; }
1533 static inline void pte_lock_deinit(struct page *page) {}
1534 #endif /* USE_SPLIT_PTE_PTLOCKS */
1536 static inline void pgtable_init(void)
1538 ptlock_cache_init();
1539 pgtable_cache_init();
1542 static inline bool pgtable_page_ctor(struct page *page)
1544 inc_zone_page_state(page, NR_PAGETABLE);
1545 return ptlock_init(page);
1548 static inline void pgtable_page_dtor(struct page *page)
1550 pte_lock_deinit(page);
1551 dec_zone_page_state(page, NR_PAGETABLE);
1554 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1556 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1557 pte_t *__pte = pte_offset_map(pmd, address); \
1563 #define pte_unmap_unlock(pte, ptl) do { \
1568 #define pte_alloc_map(mm, vma, pmd, address) \
1569 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1571 NULL: pte_offset_map(pmd, address))
1573 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1574 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1576 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1578 #define pte_alloc_kernel(pmd, address) \
1579 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1580 NULL: pte_offset_kernel(pmd, address))
1582 #if USE_SPLIT_PMD_PTLOCKS
1584 static struct page *pmd_to_page(pmd_t *pmd)
1586 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1587 return virt_to_page((void *)((unsigned long) pmd & mask));
1590 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1592 return ptlock_ptr(pmd_to_page(pmd));
1595 static inline bool pgtable_pmd_page_ctor(struct page *page)
1597 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1598 page->pmd_huge_pte = NULL;
1600 return ptlock_init(page);
1603 static inline void pgtable_pmd_page_dtor(struct page *page)
1605 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1606 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1611 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1615 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1617 return &mm->page_table_lock;
1620 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1621 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1623 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1627 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1629 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1634 extern void free_area_init(unsigned long * zones_size);
1635 extern void free_area_init_node(int nid, unsigned long * zones_size,
1636 unsigned long zone_start_pfn, unsigned long *zholes_size);
1637 extern void free_initmem(void);
1640 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1641 * into the buddy system. The freed pages will be poisoned with pattern
1642 * "poison" if it's within range [0, UCHAR_MAX].
1643 * Return pages freed into the buddy system.
1645 extern unsigned long free_reserved_area(void *start, void *end,
1646 int poison, char *s);
1648 #ifdef CONFIG_HIGHMEM
1650 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1651 * and totalram_pages.
1653 extern void free_highmem_page(struct page *page);
1656 extern void adjust_managed_page_count(struct page *page, long count);
1657 extern void mem_init_print_info(const char *str);
1659 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1661 /* Free the reserved page into the buddy system, so it gets managed. */
1662 static inline void __free_reserved_page(struct page *page)
1664 ClearPageReserved(page);
1665 init_page_count(page);
1669 static inline void free_reserved_page(struct page *page)
1671 __free_reserved_page(page);
1672 adjust_managed_page_count(page, 1);
1675 static inline void mark_page_reserved(struct page *page)
1677 SetPageReserved(page);
1678 adjust_managed_page_count(page, -1);
1682 * Default method to free all the __init memory into the buddy system.
1683 * The freed pages will be poisoned with pattern "poison" if it's within
1684 * range [0, UCHAR_MAX].
1685 * Return pages freed into the buddy system.
1687 static inline unsigned long free_initmem_default(int poison)
1689 extern char __init_begin[], __init_end[];
1691 return free_reserved_area(&__init_begin, &__init_end,
1692 poison, "unused kernel");
1695 static inline unsigned long get_num_physpages(void)
1698 unsigned long phys_pages = 0;
1700 for_each_online_node(nid)
1701 phys_pages += node_present_pages(nid);
1706 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1708 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1709 * zones, allocate the backing mem_map and account for memory holes in a more
1710 * architecture independent manner. This is a substitute for creating the
1711 * zone_sizes[] and zholes_size[] arrays and passing them to
1712 * free_area_init_node()
1714 * An architecture is expected to register range of page frames backed by
1715 * physical memory with memblock_add[_node]() before calling
1716 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1717 * usage, an architecture is expected to do something like
1719 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1721 * for_each_valid_physical_page_range()
1722 * memblock_add_node(base, size, nid)
1723 * free_area_init_nodes(max_zone_pfns);
1725 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1726 * registered physical page range. Similarly
1727 * sparse_memory_present_with_active_regions() calls memory_present() for
1728 * each range when SPARSEMEM is enabled.
1730 * See mm/page_alloc.c for more information on each function exposed by
1731 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1733 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1734 unsigned long node_map_pfn_alignment(void);
1735 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1736 unsigned long end_pfn);
1737 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1738 unsigned long end_pfn);
1739 extern void get_pfn_range_for_nid(unsigned int nid,
1740 unsigned long *start_pfn, unsigned long *end_pfn);
1741 extern unsigned long find_min_pfn_with_active_regions(void);
1742 extern void free_bootmem_with_active_regions(int nid,
1743 unsigned long max_low_pfn);
1744 extern void sparse_memory_present_with_active_regions(int nid);
1746 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1748 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1749 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1750 static inline int __early_pfn_to_nid(unsigned long pfn,
1751 struct mminit_pfnnid_cache *state)
1756 /* please see mm/page_alloc.c */
1757 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1758 /* there is a per-arch backend function. */
1759 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1760 struct mminit_pfnnid_cache *state);
1763 extern void set_dma_reserve(unsigned long new_dma_reserve);
1764 extern void memmap_init_zone(unsigned long, int, unsigned long,
1765 unsigned long, enum memmap_context);
1766 extern void setup_per_zone_wmarks(void);
1767 extern int __meminit init_per_zone_wmark_min(void);
1768 extern void mem_init(void);
1769 extern void __init mmap_init(void);
1770 extern void show_mem(unsigned int flags);
1771 extern void si_meminfo(struct sysinfo * val);
1772 extern void si_meminfo_node(struct sysinfo *val, int nid);
1774 extern __printf(3, 4)
1775 void warn_alloc_failed(gfp_t gfp_mask, int order, const char *fmt, ...);
1777 extern void setup_per_cpu_pageset(void);
1779 extern void zone_pcp_update(struct zone *zone);
1780 extern void zone_pcp_reset(struct zone *zone);
1783 extern int min_free_kbytes;
1786 extern atomic_long_t mmap_pages_allocated;
1787 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1789 /* interval_tree.c */
1790 void vma_interval_tree_insert(struct vm_area_struct *node,
1791 struct rb_root *root);
1792 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1793 struct vm_area_struct *prev,
1794 struct rb_root *root);
1795 void vma_interval_tree_remove(struct vm_area_struct *node,
1796 struct rb_root *root);
1797 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1798 unsigned long start, unsigned long last);
1799 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1800 unsigned long start, unsigned long last);
1802 #define vma_interval_tree_foreach(vma, root, start, last) \
1803 for (vma = vma_interval_tree_iter_first(root, start, last); \
1804 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1806 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1807 struct rb_root *root);
1808 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1809 struct rb_root *root);
1810 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1811 struct rb_root *root, unsigned long start, unsigned long last);
1812 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1813 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1814 #ifdef CONFIG_DEBUG_VM_RB
1815 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1818 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1819 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1820 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1823 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1824 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1825 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1826 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1827 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1828 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1829 struct mempolicy *, struct vm_userfaultfd_ctx);
1830 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1831 extern int split_vma(struct mm_struct *,
1832 struct vm_area_struct *, unsigned long addr, int new_below);
1833 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1834 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1835 struct rb_node **, struct rb_node *);
1836 extern void unlink_file_vma(struct vm_area_struct *);
1837 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1838 unsigned long addr, unsigned long len, pgoff_t pgoff,
1839 bool *need_rmap_locks);
1840 extern void exit_mmap(struct mm_struct *);
1842 static inline int check_data_rlimit(unsigned long rlim,
1844 unsigned long start,
1845 unsigned long end_data,
1846 unsigned long start_data)
1848 if (rlim < RLIM_INFINITY) {
1849 if (((new - start) + (end_data - start_data)) > rlim)
1856 extern int mm_take_all_locks(struct mm_struct *mm);
1857 extern void mm_drop_all_locks(struct mm_struct *mm);
1859 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1860 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1862 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1863 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1864 unsigned long addr, unsigned long len,
1865 unsigned long flags,
1866 const struct vm_special_mapping *spec);
1867 /* This is an obsolete alternative to _install_special_mapping. */
1868 extern int install_special_mapping(struct mm_struct *mm,
1869 unsigned long addr, unsigned long len,
1870 unsigned long flags, struct page **pages);
1872 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1874 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1875 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1876 extern unsigned long do_mmap_pgoff(struct file *file, unsigned long addr,
1877 unsigned long len, unsigned long prot, unsigned long flags,
1878 unsigned long pgoff, unsigned long *populate);
1879 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1882 extern int __mm_populate(unsigned long addr, unsigned long len,
1884 static inline void mm_populate(unsigned long addr, unsigned long len)
1887 (void) __mm_populate(addr, len, 1);
1890 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1893 /* These take the mm semaphore themselves */
1894 extern unsigned long vm_brk(unsigned long, unsigned long);
1895 extern int vm_munmap(unsigned long, size_t);
1896 extern unsigned long vm_mmap(struct file *, unsigned long,
1897 unsigned long, unsigned long,
1898 unsigned long, unsigned long);
1900 struct vm_unmapped_area_info {
1901 #define VM_UNMAPPED_AREA_TOPDOWN 1
1902 unsigned long flags;
1903 unsigned long length;
1904 unsigned long low_limit;
1905 unsigned long high_limit;
1906 unsigned long align_mask;
1907 unsigned long align_offset;
1910 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1911 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1914 * Search for an unmapped address range.
1916 * We are looking for a range that:
1917 * - does not intersect with any VMA;
1918 * - is contained within the [low_limit, high_limit) interval;
1919 * - is at least the desired size.
1920 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1922 static inline unsigned long
1923 vm_unmapped_area(struct vm_unmapped_area_info *info)
1925 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
1926 return unmapped_area_topdown(info);
1928 return unmapped_area(info);
1932 extern void truncate_inode_pages(struct address_space *, loff_t);
1933 extern void truncate_inode_pages_range(struct address_space *,
1934 loff_t lstart, loff_t lend);
1935 extern void truncate_inode_pages_final(struct address_space *);
1937 /* generic vm_area_ops exported for stackable file systems */
1938 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1939 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1940 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1942 /* mm/page-writeback.c */
1943 int write_one_page(struct page *page, int wait);
1944 void task_dirty_inc(struct task_struct *tsk);
1947 #define VM_MAX_READAHEAD 128 /* kbytes */
1948 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1950 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1951 pgoff_t offset, unsigned long nr_to_read);
1953 void page_cache_sync_readahead(struct address_space *mapping,
1954 struct file_ra_state *ra,
1957 unsigned long size);
1959 void page_cache_async_readahead(struct address_space *mapping,
1960 struct file_ra_state *ra,
1964 unsigned long size);
1966 unsigned long max_sane_readahead(unsigned long nr);
1968 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1969 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1971 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1972 extern int expand_downwards(struct vm_area_struct *vma,
1973 unsigned long address);
1975 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1977 #define expand_upwards(vma, address) (0)
1980 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
1981 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
1982 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
1983 struct vm_area_struct **pprev);
1985 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
1986 NULL if none. Assume start_addr < end_addr. */
1987 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
1989 struct vm_area_struct * vma = find_vma(mm,start_addr);
1991 if (vma && end_addr <= vma->vm_start)
1996 static inline unsigned long vma_pages(struct vm_area_struct *vma)
1998 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2001 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2002 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2003 unsigned long vm_start, unsigned long vm_end)
2005 struct vm_area_struct *vma = find_vma(mm, vm_start);
2007 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2014 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2015 void vma_set_page_prot(struct vm_area_struct *vma);
2017 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2021 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2023 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2027 #ifdef CONFIG_NUMA_BALANCING
2028 unsigned long change_prot_numa(struct vm_area_struct *vma,
2029 unsigned long start, unsigned long end);
2032 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2033 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2034 unsigned long pfn, unsigned long size, pgprot_t);
2035 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2036 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2038 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2040 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2043 struct page *follow_page_mask(struct vm_area_struct *vma,
2044 unsigned long address, unsigned int foll_flags,
2045 unsigned int *page_mask);
2047 static inline struct page *follow_page(struct vm_area_struct *vma,
2048 unsigned long address, unsigned int foll_flags)
2050 unsigned int unused_page_mask;
2051 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2054 #define FOLL_WRITE 0x01 /* check pte is writable */
2055 #define FOLL_TOUCH 0x02 /* mark page accessed */
2056 #define FOLL_GET 0x04 /* do get_page on page */
2057 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2058 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2059 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2060 * and return without waiting upon it */
2061 #define FOLL_POPULATE 0x40 /* fault in page */
2062 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2063 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2064 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2065 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2066 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2068 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2070 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2071 unsigned long size, pte_fn_t fn, void *data);
2073 #ifdef CONFIG_PROC_FS
2074 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2076 static inline void vm_stat_account(struct mm_struct *mm,
2077 unsigned long flags, struct file *file, long pages)
2079 mm->total_vm += pages;
2081 #endif /* CONFIG_PROC_FS */
2083 #ifdef CONFIG_DEBUG_PAGEALLOC
2084 extern bool _debug_pagealloc_enabled;
2085 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2087 static inline bool debug_pagealloc_enabled(void)
2089 return _debug_pagealloc_enabled;
2093 kernel_map_pages(struct page *page, int numpages, int enable)
2095 if (!debug_pagealloc_enabled())
2098 __kernel_map_pages(page, numpages, enable);
2100 #ifdef CONFIG_HIBERNATION
2101 extern bool kernel_page_present(struct page *page);
2102 #endif /* CONFIG_HIBERNATION */
2105 kernel_map_pages(struct page *page, int numpages, int enable) {}
2106 #ifdef CONFIG_HIBERNATION
2107 static inline bool kernel_page_present(struct page *page) { return true; }
2108 #endif /* CONFIG_HIBERNATION */
2111 #ifdef __HAVE_ARCH_GATE_AREA
2112 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2113 extern int in_gate_area_no_mm(unsigned long addr);
2114 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2116 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2120 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2121 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2125 #endif /* __HAVE_ARCH_GATE_AREA */
2127 #ifdef CONFIG_SYSCTL
2128 extern int sysctl_drop_caches;
2129 int drop_caches_sysctl_handler(struct ctl_table *, int,
2130 void __user *, size_t *, loff_t *);
2133 void drop_slab(void);
2134 void drop_slab_node(int nid);
2137 #define randomize_va_space 0
2139 extern int randomize_va_space;
2142 const char * arch_vma_name(struct vm_area_struct *vma);
2143 void print_vma_addr(char *prefix, unsigned long rip);
2145 void sparse_mem_maps_populate_node(struct page **map_map,
2146 unsigned long pnum_begin,
2147 unsigned long pnum_end,
2148 unsigned long map_count,
2151 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2152 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2153 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2154 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2155 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2156 void *vmemmap_alloc_block(unsigned long size, int node);
2157 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2158 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2159 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2161 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2162 void vmemmap_populate_print_last(void);
2163 #ifdef CONFIG_MEMORY_HOTPLUG
2164 void vmemmap_free(unsigned long start, unsigned long end);
2166 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2167 unsigned long size);
2170 MF_COUNT_INCREASED = 1 << 0,
2171 MF_ACTION_REQUIRED = 1 << 1,
2172 MF_MUST_KILL = 1 << 2,
2173 MF_SOFT_OFFLINE = 1 << 3,
2175 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2176 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2177 extern int unpoison_memory(unsigned long pfn);
2178 extern int get_hwpoison_page(struct page *page);
2179 extern void put_hwpoison_page(struct page *page);
2180 extern int sysctl_memory_failure_early_kill;
2181 extern int sysctl_memory_failure_recovery;
2182 extern void shake_page(struct page *p, int access);
2183 extern atomic_long_t num_poisoned_pages;
2184 extern int soft_offline_page(struct page *page, int flags);
2188 * Error handlers for various types of pages.
2191 MF_IGNORED, /* Error: cannot be handled */
2192 MF_FAILED, /* Error: handling failed */
2193 MF_DELAYED, /* Will be handled later */
2194 MF_RECOVERED, /* Successfully recovered */
2197 enum mf_action_page_type {
2199 MF_MSG_KERNEL_HIGH_ORDER,
2201 MF_MSG_DIFFERENT_COMPOUND,
2202 MF_MSG_POISONED_HUGE,
2205 MF_MSG_UNMAP_FAILED,
2206 MF_MSG_DIRTY_SWAPCACHE,
2207 MF_MSG_CLEAN_SWAPCACHE,
2208 MF_MSG_DIRTY_MLOCKED_LRU,
2209 MF_MSG_CLEAN_MLOCKED_LRU,
2210 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2211 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2214 MF_MSG_TRUNCATED_LRU,
2220 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2221 extern void clear_huge_page(struct page *page,
2223 unsigned int pages_per_huge_page);
2224 extern void copy_user_huge_page(struct page *dst, struct page *src,
2225 unsigned long addr, struct vm_area_struct *vma,
2226 unsigned int pages_per_huge_page);
2227 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2229 extern struct page_ext_operations debug_guardpage_ops;
2230 extern struct page_ext_operations page_poisoning_ops;
2232 #ifdef CONFIG_DEBUG_PAGEALLOC
2233 extern unsigned int _debug_guardpage_minorder;
2234 extern bool _debug_guardpage_enabled;
2236 static inline unsigned int debug_guardpage_minorder(void)
2238 return _debug_guardpage_minorder;
2241 static inline bool debug_guardpage_enabled(void)
2243 return _debug_guardpage_enabled;
2246 static inline bool page_is_guard(struct page *page)
2248 struct page_ext *page_ext;
2250 if (!debug_guardpage_enabled())
2253 page_ext = lookup_page_ext(page);
2254 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2257 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2258 static inline bool debug_guardpage_enabled(void) { return false; }
2259 static inline bool page_is_guard(struct page *page) { return false; }
2260 #endif /* CONFIG_DEBUG_PAGEALLOC */
2262 #if MAX_NUMNODES > 1
2263 void __init setup_nr_node_ids(void);
2265 static inline void setup_nr_node_ids(void) {}
2268 #endif /* __KERNEL__ */
2269 #endif /* _LINUX_MM_H */