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
55 #include <asm/pgtable.h>
56 #include <asm/processor.h>
59 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
63 * To prevent common memory management code establishing
64 * a zero page mapping on a read fault.
65 * This macro should be defined within <asm/pgtable.h>.
66 * s390 does this to prevent multiplexing of hardware bits
67 * related to the physical page in case of virtualization.
69 #ifndef mm_forbids_zeropage
70 #define mm_forbids_zeropage(X) (0)
73 extern unsigned long sysctl_user_reserve_kbytes;
74 extern unsigned long sysctl_admin_reserve_kbytes;
76 extern int sysctl_overcommit_memory;
77 extern int sysctl_overcommit_ratio;
78 extern unsigned long sysctl_overcommit_kbytes;
80 extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *,
82 extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
85 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
87 /* to align the pointer to the (next) page boundary */
88 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
90 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
91 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
94 * Linux kernel virtual memory manager primitives.
95 * The idea being to have a "virtual" mm in the same way
96 * we have a virtual fs - giving a cleaner interface to the
97 * mm details, and allowing different kinds of memory mappings
98 * (from shared memory to executable loading to arbitrary
102 extern struct kmem_cache *vm_area_cachep;
105 extern struct rb_root nommu_region_tree;
106 extern struct rw_semaphore nommu_region_sem;
108 extern unsigned int kobjsize(const void *objp);
112 * vm_flags in vm_area_struct, see mm_types.h.
114 #define VM_NONE 0x00000000
116 #define VM_READ 0x00000001 /* currently active flags */
117 #define VM_WRITE 0x00000002
118 #define VM_EXEC 0x00000004
119 #define VM_SHARED 0x00000008
121 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
122 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
123 #define VM_MAYWRITE 0x00000020
124 #define VM_MAYEXEC 0x00000040
125 #define VM_MAYSHARE 0x00000080
127 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
128 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
129 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
130 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
131 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
133 #define VM_LOCKED 0x00002000
134 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
136 /* Used by sys_madvise() */
137 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
138 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
140 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
141 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
142 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
143 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
144 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
145 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
146 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
147 #define VM_ARCH_2 0x02000000
148 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
150 #ifdef CONFIG_MEM_SOFT_DIRTY
151 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
153 # define VM_SOFTDIRTY 0
156 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
157 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
158 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
159 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
161 #if defined(CONFIG_X86)
162 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
163 #elif defined(CONFIG_PPC)
164 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
165 #elif defined(CONFIG_PARISC)
166 # define VM_GROWSUP VM_ARCH_1
167 #elif defined(CONFIG_METAG)
168 # define VM_GROWSUP VM_ARCH_1
169 #elif defined(CONFIG_IA64)
170 # define VM_GROWSUP VM_ARCH_1
171 #elif !defined(CONFIG_MMU)
172 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
175 #if defined(CONFIG_X86)
176 /* MPX specific bounds table or bounds directory */
177 # define VM_MPX VM_ARCH_2
181 # define VM_GROWSUP VM_NONE
184 /* Bits set in the VMA until the stack is in its final location */
185 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
187 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
188 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
191 #ifdef CONFIG_STACK_GROWSUP
192 #define VM_STACK_FLAGS (VM_GROWSUP | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
194 #define VM_STACK_FLAGS (VM_GROWSDOWN | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
198 * Special vmas that are non-mergable, non-mlock()able.
199 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
201 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
203 /* This mask defines which mm->def_flags a process can inherit its parent */
204 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
206 /* This mask is used to clear all the VMA flags used by mlock */
207 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
210 * mapping from the currently active vm_flags protection bits (the
211 * low four bits) to a page protection mask..
213 extern pgprot_t protection_map[16];
215 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
216 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
217 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
218 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
219 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
220 #define FAULT_FLAG_TRIED 0x20 /* Second try */
221 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
224 * vm_fault is filled by the the pagefault handler and passed to the vma's
225 * ->fault function. The vma's ->fault is responsible for returning a bitmask
226 * of VM_FAULT_xxx flags that give details about how the fault was handled.
228 * pgoff should be used in favour of virtual_address, if possible.
231 unsigned int flags; /* FAULT_FLAG_xxx flags */
232 pgoff_t pgoff; /* Logical page offset based on vma */
233 void __user *virtual_address; /* Faulting virtual address */
235 struct page *cow_page; /* Handler may choose to COW */
236 struct page *page; /* ->fault handlers should return a
237 * page here, unless VM_FAULT_NOPAGE
238 * is set (which is also implied by
241 /* for ->map_pages() only */
242 pgoff_t max_pgoff; /* map pages for offset from pgoff till
243 * max_pgoff inclusive */
244 pte_t *pte; /* pte entry associated with ->pgoff */
248 * These are the virtual MM functions - opening of an area, closing and
249 * unmapping it (needed to keep files on disk up-to-date etc), pointer
250 * to the functions called when a no-page or a wp-page exception occurs.
252 struct vm_operations_struct {
253 void (*open)(struct vm_area_struct * area);
254 void (*close)(struct vm_area_struct * area);
255 int (*mremap)(struct vm_area_struct * area);
256 int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf);
257 int (*pmd_fault)(struct vm_area_struct *, unsigned long address,
258 pmd_t *, unsigned int flags);
259 void (*map_pages)(struct vm_area_struct *vma, struct vm_fault *vmf);
261 /* notification that a previously read-only page is about to become
262 * writable, if an error is returned it will cause a SIGBUS */
263 int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
265 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
266 int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf);
268 /* called by access_process_vm when get_user_pages() fails, typically
269 * for use by special VMAs that can switch between memory and hardware
271 int (*access)(struct vm_area_struct *vma, unsigned long addr,
272 void *buf, int len, int write);
274 /* Called by the /proc/PID/maps code to ask the vma whether it
275 * has a special name. Returning non-NULL will also cause this
276 * vma to be dumped unconditionally. */
277 const char *(*name)(struct vm_area_struct *vma);
281 * set_policy() op must add a reference to any non-NULL @new mempolicy
282 * to hold the policy upon return. Caller should pass NULL @new to
283 * remove a policy and fall back to surrounding context--i.e. do not
284 * install a MPOL_DEFAULT policy, nor the task or system default
287 int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new);
290 * get_policy() op must add reference [mpol_get()] to any policy at
291 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
292 * in mm/mempolicy.c will do this automatically.
293 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
294 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
295 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
296 * must return NULL--i.e., do not "fallback" to task or system default
299 struct mempolicy *(*get_policy)(struct vm_area_struct *vma,
303 * Called by vm_normal_page() for special PTEs to find the
304 * page for @addr. This is useful if the default behavior
305 * (using pte_page()) would not find the correct page.
307 struct page *(*find_special_page)(struct vm_area_struct *vma,
314 #define page_private(page) ((page)->private)
315 #define set_page_private(page, v) ((page)->private = (v))
318 * FIXME: take this include out, include page-flags.h in
319 * files which need it (119 of them)
321 #include <linux/page-flags.h>
322 #include <linux/huge_mm.h>
325 * Methods to modify the page usage count.
327 * What counts for a page usage:
328 * - cache mapping (page->mapping)
329 * - private data (page->private)
330 * - page mapped in a task's page tables, each mapping
331 * is counted separately
333 * Also, many kernel routines increase the page count before a critical
334 * routine so they can be sure the page doesn't go away from under them.
338 * Drop a ref, return true if the refcount fell to zero (the page has no users)
340 static inline int put_page_testzero(struct page *page)
342 VM_BUG_ON_PAGE(atomic_read(&page->_count) == 0, page);
343 return atomic_dec_and_test(&page->_count);
347 * Try to grab a ref unless the page has a refcount of zero, return false if
349 * This can be called when MMU is off so it must not access
350 * any of the virtual mappings.
352 static inline int get_page_unless_zero(struct page *page)
354 return atomic_inc_not_zero(&page->_count);
357 extern int page_is_ram(unsigned long pfn);
365 int region_intersects(resource_size_t offset, size_t size, const char *type);
367 /* Support for virtually mapped pages */
368 struct page *vmalloc_to_page(const void *addr);
369 unsigned long vmalloc_to_pfn(const void *addr);
372 * Determine if an address is within the vmalloc range
374 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
375 * is no special casing required.
377 static inline int is_vmalloc_addr(const void *x)
380 unsigned long addr = (unsigned long)x;
382 return addr >= VMALLOC_START && addr < VMALLOC_END;
388 extern int is_vmalloc_or_module_addr(const void *x);
390 static inline int is_vmalloc_or_module_addr(const void *x)
396 extern void kvfree(const void *addr);
398 static inline atomic_t *compound_mapcount_ptr(struct page *page)
400 return &page[1].compound_mapcount;
403 static inline int compound_mapcount(struct page *page)
405 if (!PageCompound(page))
407 page = compound_head(page);
408 return atomic_read(compound_mapcount_ptr(page)) + 1;
412 * The atomic page->_mapcount, starts from -1: so that transitions
413 * both from it and to it can be tracked, using atomic_inc_and_test
414 * and atomic_add_negative(-1).
416 static inline void page_mapcount_reset(struct page *page)
418 atomic_set(&(page)->_mapcount, -1);
421 static inline int page_mapcount(struct page *page)
424 VM_BUG_ON_PAGE(PageSlab(page), page);
426 ret = atomic_read(&page->_mapcount) + 1;
427 if (PageCompound(page)) {
428 page = compound_head(page);
429 ret += atomic_read(compound_mapcount_ptr(page)) + 1;
430 if (PageDoubleMap(page))
436 static inline int page_count(struct page *page)
438 return atomic_read(&compound_head(page)->_count);
441 static inline void get_page(struct page *page)
443 page = compound_head(page);
445 * Getting a normal page or the head of a compound page
446 * requires to already have an elevated page->_count.
448 VM_BUG_ON_PAGE(atomic_read(&page->_count) <= 0, page);
449 atomic_inc(&page->_count);
452 static inline struct page *virt_to_head_page(const void *x)
454 struct page *page = virt_to_page(x);
456 return compound_head(page);
460 * Setup the page count before being freed into the page allocator for
461 * the first time (boot or memory hotplug)
463 static inline void init_page_count(struct page *page)
465 atomic_set(&page->_count, 1);
468 void __put_page(struct page *page);
470 static inline void put_page(struct page *page)
472 page = compound_head(page);
473 if (put_page_testzero(page))
477 void put_pages_list(struct list_head *pages);
479 void split_page(struct page *page, unsigned int order);
480 int split_free_page(struct page *page);
483 * Compound pages have a destructor function. Provide a
484 * prototype for that function and accessor functions.
485 * These are _only_ valid on the head of a compound page.
487 typedef void compound_page_dtor(struct page *);
489 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
490 enum compound_dtor_id {
493 #ifdef CONFIG_HUGETLB_PAGE
496 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
501 extern compound_page_dtor * const compound_page_dtors[];
503 static inline void set_compound_page_dtor(struct page *page,
504 enum compound_dtor_id compound_dtor)
506 VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page);
507 page[1].compound_dtor = compound_dtor;
510 static inline compound_page_dtor *get_compound_page_dtor(struct page *page)
512 VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page);
513 return compound_page_dtors[page[1].compound_dtor];
516 static inline unsigned int compound_order(struct page *page)
520 return page[1].compound_order;
523 static inline void set_compound_order(struct page *page, unsigned int order)
525 page[1].compound_order = order;
528 void free_compound_page(struct page *page);
532 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
533 * servicing faults for write access. In the normal case, do always want
534 * pte_mkwrite. But get_user_pages can cause write faults for mappings
535 * that do not have writing enabled, when used by access_process_vm.
537 static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma)
539 if (likely(vma->vm_flags & VM_WRITE))
540 pte = pte_mkwrite(pte);
544 void do_set_pte(struct vm_area_struct *vma, unsigned long address,
545 struct page *page, pte_t *pte, bool write, bool anon);
549 * Multiple processes may "see" the same page. E.g. for untouched
550 * mappings of /dev/null, all processes see the same page full of
551 * zeroes, and text pages of executables and shared libraries have
552 * only one copy in memory, at most, normally.
554 * For the non-reserved pages, page_count(page) denotes a reference count.
555 * page_count() == 0 means the page is free. page->lru is then used for
556 * freelist management in the buddy allocator.
557 * page_count() > 0 means the page has been allocated.
559 * Pages are allocated by the slab allocator in order to provide memory
560 * to kmalloc and kmem_cache_alloc. In this case, the management of the
561 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
562 * unless a particular usage is carefully commented. (the responsibility of
563 * freeing the kmalloc memory is the caller's, of course).
565 * A page may be used by anyone else who does a __get_free_page().
566 * In this case, page_count still tracks the references, and should only
567 * be used through the normal accessor functions. The top bits of page->flags
568 * and page->virtual store page management information, but all other fields
569 * are unused and could be used privately, carefully. The management of this
570 * page is the responsibility of the one who allocated it, and those who have
571 * subsequently been given references to it.
573 * The other pages (we may call them "pagecache pages") are completely
574 * managed by the Linux memory manager: I/O, buffers, swapping etc.
575 * The following discussion applies only to them.
577 * A pagecache page contains an opaque `private' member, which belongs to the
578 * page's address_space. Usually, this is the address of a circular list of
579 * the page's disk buffers. PG_private must be set to tell the VM to call
580 * into the filesystem to release these pages.
582 * A page may belong to an inode's memory mapping. In this case, page->mapping
583 * is the pointer to the inode, and page->index is the file offset of the page,
584 * in units of PAGE_CACHE_SIZE.
586 * If pagecache pages are not associated with an inode, they are said to be
587 * anonymous pages. These may become associated with the swapcache, and in that
588 * case PG_swapcache is set, and page->private is an offset into the swapcache.
590 * In either case (swapcache or inode backed), the pagecache itself holds one
591 * reference to the page. Setting PG_private should also increment the
592 * refcount. The each user mapping also has a reference to the page.
594 * The pagecache pages are stored in a per-mapping radix tree, which is
595 * rooted at mapping->page_tree, and indexed by offset.
596 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
597 * lists, we instead now tag pages as dirty/writeback in the radix tree.
599 * All pagecache pages may be subject to I/O:
600 * - inode pages may need to be read from disk,
601 * - inode pages which have been modified and are MAP_SHARED may need
602 * to be written back to the inode on disk,
603 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
604 * modified may need to be swapped out to swap space and (later) to be read
609 * The zone field is never updated after free_area_init_core()
610 * sets it, so none of the operations on it need to be atomic.
613 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
614 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
615 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
616 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
617 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
620 * Define the bit shifts to access each section. For non-existent
621 * sections we define the shift as 0; that plus a 0 mask ensures
622 * the compiler will optimise away reference to them.
624 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
625 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
626 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
627 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
629 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
630 #ifdef NODE_NOT_IN_PAGE_FLAGS
631 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
632 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
633 SECTIONS_PGOFF : ZONES_PGOFF)
635 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
636 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
637 NODES_PGOFF : ZONES_PGOFF)
640 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
642 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
643 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
646 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
647 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
648 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
649 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
650 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
652 static inline enum zone_type page_zonenum(const struct page *page)
654 return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK;
657 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
658 #define SECTION_IN_PAGE_FLAGS
662 * The identification function is mainly used by the buddy allocator for
663 * determining if two pages could be buddies. We are not really identifying
664 * the zone since we could be using the section number id if we do not have
665 * node id available in page flags.
666 * We only guarantee that it will return the same value for two combinable
669 static inline int page_zone_id(struct page *page)
671 return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK;
674 static inline int zone_to_nid(struct zone *zone)
683 #ifdef NODE_NOT_IN_PAGE_FLAGS
684 extern int page_to_nid(const struct page *page);
686 static inline int page_to_nid(const struct page *page)
688 return (page->flags >> NODES_PGSHIFT) & NODES_MASK;
692 #ifdef CONFIG_NUMA_BALANCING
693 static inline int cpu_pid_to_cpupid(int cpu, int pid)
695 return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK);
698 static inline int cpupid_to_pid(int cpupid)
700 return cpupid & LAST__PID_MASK;
703 static inline int cpupid_to_cpu(int cpupid)
705 return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK;
708 static inline int cpupid_to_nid(int cpupid)
710 return cpu_to_node(cpupid_to_cpu(cpupid));
713 static inline bool cpupid_pid_unset(int cpupid)
715 return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK);
718 static inline bool cpupid_cpu_unset(int cpupid)
720 return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK);
723 static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid)
725 return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid);
728 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
729 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
730 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
732 return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK);
735 static inline int page_cpupid_last(struct page *page)
737 return page->_last_cpupid;
739 static inline void page_cpupid_reset_last(struct page *page)
741 page->_last_cpupid = -1 & LAST_CPUPID_MASK;
744 static inline int page_cpupid_last(struct page *page)
746 return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK;
749 extern int page_cpupid_xchg_last(struct page *page, int cpupid);
751 static inline void page_cpupid_reset_last(struct page *page)
753 int cpupid = (1 << LAST_CPUPID_SHIFT) - 1;
755 page->flags &= ~(LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT);
756 page->flags |= (cpupid & LAST_CPUPID_MASK) << LAST_CPUPID_PGSHIFT;
758 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
759 #else /* !CONFIG_NUMA_BALANCING */
760 static inline int page_cpupid_xchg_last(struct page *page, int cpupid)
762 return page_to_nid(page); /* XXX */
765 static inline int page_cpupid_last(struct page *page)
767 return page_to_nid(page); /* XXX */
770 static inline int cpupid_to_nid(int cpupid)
775 static inline int cpupid_to_pid(int cpupid)
780 static inline int cpupid_to_cpu(int cpupid)
785 static inline int cpu_pid_to_cpupid(int nid, int pid)
790 static inline bool cpupid_pid_unset(int cpupid)
795 static inline void page_cpupid_reset_last(struct page *page)
799 static inline bool cpupid_match_pid(struct task_struct *task, int cpupid)
803 #endif /* CONFIG_NUMA_BALANCING */
805 static inline struct zone *page_zone(const struct page *page)
807 return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)];
810 #ifdef SECTION_IN_PAGE_FLAGS
811 static inline void set_page_section(struct page *page, unsigned long section)
813 page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT);
814 page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT;
817 static inline unsigned long page_to_section(const struct page *page)
819 return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK;
823 static inline void set_page_zone(struct page *page, enum zone_type zone)
825 page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT);
826 page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT;
829 static inline void set_page_node(struct page *page, unsigned long node)
831 page->flags &= ~(NODES_MASK << NODES_PGSHIFT);
832 page->flags |= (node & NODES_MASK) << NODES_PGSHIFT;
835 static inline void set_page_links(struct page *page, enum zone_type zone,
836 unsigned long node, unsigned long pfn)
838 set_page_zone(page, zone);
839 set_page_node(page, node);
840 #ifdef SECTION_IN_PAGE_FLAGS
841 set_page_section(page, pfn_to_section_nr(pfn));
846 static inline struct mem_cgroup *page_memcg(struct page *page)
848 return page->mem_cgroup;
851 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
853 page->mem_cgroup = memcg;
856 static inline struct mem_cgroup *page_memcg(struct page *page)
861 static inline void set_page_memcg(struct page *page, struct mem_cgroup *memcg)
867 * Some inline functions in vmstat.h depend on page_zone()
869 #include <linux/vmstat.h>
871 static __always_inline void *lowmem_page_address(const struct page *page)
873 return __va(PFN_PHYS(page_to_pfn(page)));
876 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
877 #define HASHED_PAGE_VIRTUAL
880 #if defined(WANT_PAGE_VIRTUAL)
881 static inline void *page_address(const struct page *page)
883 return page->virtual;
885 static inline void set_page_address(struct page *page, void *address)
887 page->virtual = address;
889 #define page_address_init() do { } while(0)
892 #if defined(HASHED_PAGE_VIRTUAL)
893 void *page_address(const struct page *page);
894 void set_page_address(struct page *page, void *virtual);
895 void page_address_init(void);
898 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
899 #define page_address(page) lowmem_page_address(page)
900 #define set_page_address(page, address) do { } while(0)
901 #define page_address_init() do { } while(0)
904 extern void *page_rmapping(struct page *page);
905 extern struct anon_vma *page_anon_vma(struct page *page);
906 extern struct address_space *page_mapping(struct page *page);
908 extern struct address_space *__page_file_mapping(struct page *);
911 struct address_space *page_file_mapping(struct page *page)
913 if (unlikely(PageSwapCache(page)))
914 return __page_file_mapping(page);
916 return page->mapping;
920 * Return the pagecache index of the passed page. Regular pagecache pages
921 * use ->index whereas swapcache pages use ->private
923 static inline pgoff_t page_index(struct page *page)
925 if (unlikely(PageSwapCache(page)))
926 return page_private(page);
930 extern pgoff_t __page_file_index(struct page *page);
933 * Return the file index of the page. Regular pagecache pages use ->index
934 * whereas swapcache pages use swp_offset(->private)
936 static inline pgoff_t page_file_index(struct page *page)
938 if (unlikely(PageSwapCache(page)))
939 return __page_file_index(page);
945 * Return true if this page is mapped into pagetables.
946 * For compound page it returns true if any subpage of compound page is mapped.
948 static inline bool page_mapped(struct page *page)
951 if (likely(!PageCompound(page)))
952 return atomic_read(&page->_mapcount) >= 0;
953 page = compound_head(page);
954 if (atomic_read(compound_mapcount_ptr(page)) >= 0)
956 for (i = 0; i < hpage_nr_pages(page); i++) {
957 if (atomic_read(&page[i]._mapcount) >= 0)
964 * Return true only if the page has been allocated with
965 * ALLOC_NO_WATERMARKS and the low watermark was not
966 * met implying that the system is under some pressure.
968 static inline bool page_is_pfmemalloc(struct page *page)
971 * Page index cannot be this large so this must be
974 return page->index == -1UL;
978 * Only to be called by the page allocator on a freshly allocated
981 static inline void set_page_pfmemalloc(struct page *page)
986 static inline void clear_page_pfmemalloc(struct page *page)
992 * Different kinds of faults, as returned by handle_mm_fault().
993 * Used to decide whether a process gets delivered SIGBUS or
994 * just gets major/minor fault counters bumped up.
997 #define VM_FAULT_MINOR 0 /* For backwards compat. Remove me quickly. */
999 #define VM_FAULT_OOM 0x0001
1000 #define VM_FAULT_SIGBUS 0x0002
1001 #define VM_FAULT_MAJOR 0x0004
1002 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1003 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1004 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1005 #define VM_FAULT_SIGSEGV 0x0040
1007 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1008 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1009 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1010 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1012 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1014 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1015 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1018 /* Encode hstate index for a hwpoisoned large page */
1019 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1020 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1023 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1025 extern void pagefault_out_of_memory(void);
1027 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1030 * Flags passed to show_mem() and show_free_areas() to suppress output in
1033 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1035 extern void show_free_areas(unsigned int flags);
1036 extern bool skip_free_areas_node(unsigned int flags, int nid);
1038 int shmem_zero_setup(struct vm_area_struct *);
1040 bool shmem_mapping(struct address_space *mapping);
1042 static inline bool shmem_mapping(struct address_space *mapping)
1048 extern int can_do_mlock(void);
1049 extern int user_shm_lock(size_t, struct user_struct *);
1050 extern void user_shm_unlock(size_t, struct user_struct *);
1053 * Parameter block passed down to zap_pte_range in exceptional cases.
1055 struct zap_details {
1056 struct address_space *check_mapping; /* Check page->mapping if set */
1057 pgoff_t first_index; /* Lowest page->index to unmap */
1058 pgoff_t last_index; /* Highest page->index to unmap */
1061 struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr,
1064 int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address,
1065 unsigned long size);
1066 void zap_page_range(struct vm_area_struct *vma, unsigned long address,
1067 unsigned long size, struct zap_details *);
1068 void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma,
1069 unsigned long start, unsigned long end);
1072 * mm_walk - callbacks for walk_page_range
1073 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1074 * this handler is required to be able to handle
1075 * pmd_trans_huge() pmds. They may simply choose to
1076 * split_huge_page() instead of handling it explicitly.
1077 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1078 * @pte_hole: if set, called for each hole at all levels
1079 * @hugetlb_entry: if set, called for each hugetlb entry
1080 * @test_walk: caller specific callback function to determine whether
1081 * we walk over the current vma or not. A positive returned
1082 * value means "do page table walk over the current vma,"
1083 * and a negative one means "abort current page table walk
1084 * right now." 0 means "skip the current vma."
1085 * @mm: mm_struct representing the target process of page table walk
1086 * @vma: vma currently walked (NULL if walking outside vmas)
1087 * @private: private data for callbacks' usage
1089 * (see the comment on walk_page_range() for more details)
1092 int (*pmd_entry)(pmd_t *pmd, unsigned long addr,
1093 unsigned long next, struct mm_walk *walk);
1094 int (*pte_entry)(pte_t *pte, unsigned long addr,
1095 unsigned long next, struct mm_walk *walk);
1096 int (*pte_hole)(unsigned long addr, unsigned long next,
1097 struct mm_walk *walk);
1098 int (*hugetlb_entry)(pte_t *pte, unsigned long hmask,
1099 unsigned long addr, unsigned long next,
1100 struct mm_walk *walk);
1101 int (*test_walk)(unsigned long addr, unsigned long next,
1102 struct mm_walk *walk);
1103 struct mm_struct *mm;
1104 struct vm_area_struct *vma;
1108 int walk_page_range(unsigned long addr, unsigned long end,
1109 struct mm_walk *walk);
1110 int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk);
1111 void free_pgd_range(struct mmu_gather *tlb, unsigned long addr,
1112 unsigned long end, unsigned long floor, unsigned long ceiling);
1113 int copy_page_range(struct mm_struct *dst, struct mm_struct *src,
1114 struct vm_area_struct *vma);
1115 void unmap_mapping_range(struct address_space *mapping,
1116 loff_t const holebegin, loff_t const holelen, int even_cows);
1117 int follow_pfn(struct vm_area_struct *vma, unsigned long address,
1118 unsigned long *pfn);
1119 int follow_phys(struct vm_area_struct *vma, unsigned long address,
1120 unsigned int flags, unsigned long *prot, resource_size_t *phys);
1121 int generic_access_phys(struct vm_area_struct *vma, unsigned long addr,
1122 void *buf, int len, int write);
1124 static inline void unmap_shared_mapping_range(struct address_space *mapping,
1125 loff_t const holebegin, loff_t const holelen)
1127 unmap_mapping_range(mapping, holebegin, holelen, 0);
1130 extern void truncate_pagecache(struct inode *inode, loff_t new);
1131 extern void truncate_setsize(struct inode *inode, loff_t newsize);
1132 void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to);
1133 void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end);
1134 int truncate_inode_page(struct address_space *mapping, struct page *page);
1135 int generic_error_remove_page(struct address_space *mapping, struct page *page);
1136 int invalidate_inode_page(struct page *page);
1139 extern int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
1140 unsigned long address, unsigned int flags);
1141 extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm,
1142 unsigned long address, unsigned int fault_flags);
1144 static inline int handle_mm_fault(struct mm_struct *mm,
1145 struct vm_area_struct *vma, unsigned long address,
1148 /* should never happen if there's no MMU */
1150 return VM_FAULT_SIGBUS;
1152 static inline int fixup_user_fault(struct task_struct *tsk,
1153 struct mm_struct *mm, unsigned long address,
1154 unsigned int fault_flags)
1156 /* should never happen if there's no MMU */
1162 extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, int write);
1163 extern int access_remote_vm(struct mm_struct *mm, unsigned long addr,
1164 void *buf, int len, int write);
1166 long __get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1167 unsigned long start, unsigned long nr_pages,
1168 unsigned int foll_flags, struct page **pages,
1169 struct vm_area_struct **vmas, int *nonblocking);
1170 long get_user_pages(struct task_struct *tsk, struct mm_struct *mm,
1171 unsigned long start, unsigned long nr_pages,
1172 int write, int force, struct page **pages,
1173 struct vm_area_struct **vmas);
1174 long get_user_pages_locked(struct task_struct *tsk, struct mm_struct *mm,
1175 unsigned long start, unsigned long nr_pages,
1176 int write, int force, struct page **pages,
1178 long __get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1179 unsigned long start, unsigned long nr_pages,
1180 int write, int force, struct page **pages,
1181 unsigned int gup_flags);
1182 long get_user_pages_unlocked(struct task_struct *tsk, struct mm_struct *mm,
1183 unsigned long start, unsigned long nr_pages,
1184 int write, int force, struct page **pages);
1185 int get_user_pages_fast(unsigned long start, int nr_pages, int write,
1186 struct page **pages);
1188 /* Container for pinned pfns / pages */
1189 struct frame_vector {
1190 unsigned int nr_allocated; /* Number of frames we have space for */
1191 unsigned int nr_frames; /* Number of frames stored in ptrs array */
1192 bool got_ref; /* Did we pin pages by getting page ref? */
1193 bool is_pfns; /* Does array contain pages or pfns? */
1194 void *ptrs[0]; /* Array of pinned pfns / pages. Use
1195 * pfns_vector_pages() or pfns_vector_pfns()
1199 struct frame_vector *frame_vector_create(unsigned int nr_frames);
1200 void frame_vector_destroy(struct frame_vector *vec);
1201 int get_vaddr_frames(unsigned long start, unsigned int nr_pfns,
1202 bool write, bool force, struct frame_vector *vec);
1203 void put_vaddr_frames(struct frame_vector *vec);
1204 int frame_vector_to_pages(struct frame_vector *vec);
1205 void frame_vector_to_pfns(struct frame_vector *vec);
1207 static inline unsigned int frame_vector_count(struct frame_vector *vec)
1209 return vec->nr_frames;
1212 static inline struct page **frame_vector_pages(struct frame_vector *vec)
1215 int err = frame_vector_to_pages(vec);
1218 return ERR_PTR(err);
1220 return (struct page **)(vec->ptrs);
1223 static inline unsigned long *frame_vector_pfns(struct frame_vector *vec)
1226 frame_vector_to_pfns(vec);
1227 return (unsigned long *)(vec->ptrs);
1231 int get_kernel_pages(const struct kvec *iov, int nr_pages, int write,
1232 struct page **pages);
1233 int get_kernel_page(unsigned long start, int write, struct page **pages);
1234 struct page *get_dump_page(unsigned long addr);
1236 extern int try_to_release_page(struct page * page, gfp_t gfp_mask);
1237 extern void do_invalidatepage(struct page *page, unsigned int offset,
1238 unsigned int length);
1240 int __set_page_dirty_nobuffers(struct page *page);
1241 int __set_page_dirty_no_writeback(struct page *page);
1242 int redirty_page_for_writepage(struct writeback_control *wbc,
1244 void account_page_dirtied(struct page *page, struct address_space *mapping,
1245 struct mem_cgroup *memcg);
1246 void account_page_cleaned(struct page *page, struct address_space *mapping,
1247 struct mem_cgroup *memcg, struct bdi_writeback *wb);
1248 int set_page_dirty(struct page *page);
1249 int set_page_dirty_lock(struct page *page);
1250 void cancel_dirty_page(struct page *page);
1251 int clear_page_dirty_for_io(struct page *page);
1253 int get_cmdline(struct task_struct *task, char *buffer, int buflen);
1255 /* Is the vma a continuation of the stack vma above it? */
1256 static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr)
1258 return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN);
1261 static inline bool vma_is_anonymous(struct vm_area_struct *vma)
1263 return !vma->vm_ops;
1266 static inline int stack_guard_page_start(struct vm_area_struct *vma,
1269 return (vma->vm_flags & VM_GROWSDOWN) &&
1270 (vma->vm_start == addr) &&
1271 !vma_growsdown(vma->vm_prev, addr);
1274 /* Is the vma a continuation of the stack vma below it? */
1275 static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr)
1277 return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP);
1280 static inline int stack_guard_page_end(struct vm_area_struct *vma,
1283 return (vma->vm_flags & VM_GROWSUP) &&
1284 (vma->vm_end == addr) &&
1285 !vma_growsup(vma->vm_next, addr);
1288 extern struct task_struct *task_of_stack(struct task_struct *task,
1289 struct vm_area_struct *vma, bool in_group);
1291 extern unsigned long move_page_tables(struct vm_area_struct *vma,
1292 unsigned long old_addr, struct vm_area_struct *new_vma,
1293 unsigned long new_addr, unsigned long len,
1294 bool need_rmap_locks);
1295 extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start,
1296 unsigned long end, pgprot_t newprot,
1297 int dirty_accountable, int prot_numa);
1298 extern int mprotect_fixup(struct vm_area_struct *vma,
1299 struct vm_area_struct **pprev, unsigned long start,
1300 unsigned long end, unsigned long newflags);
1303 * doesn't attempt to fault and will return short.
1305 int __get_user_pages_fast(unsigned long start, int nr_pages, int write,
1306 struct page **pages);
1308 * per-process(per-mm_struct) statistics.
1310 static inline unsigned long get_mm_counter(struct mm_struct *mm, int member)
1312 long val = atomic_long_read(&mm->rss_stat.count[member]);
1314 #ifdef SPLIT_RSS_COUNTING
1316 * counter is updated in asynchronous manner and may go to minus.
1317 * But it's never be expected number for users.
1322 return (unsigned long)val;
1325 static inline void add_mm_counter(struct mm_struct *mm, int member, long value)
1327 atomic_long_add(value, &mm->rss_stat.count[member]);
1330 static inline void inc_mm_counter(struct mm_struct *mm, int member)
1332 atomic_long_inc(&mm->rss_stat.count[member]);
1335 static inline void dec_mm_counter(struct mm_struct *mm, int member)
1337 atomic_long_dec(&mm->rss_stat.count[member]);
1340 static inline unsigned long get_mm_rss(struct mm_struct *mm)
1342 return get_mm_counter(mm, MM_FILEPAGES) +
1343 get_mm_counter(mm, MM_ANONPAGES);
1346 static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm)
1348 return max(mm->hiwater_rss, get_mm_rss(mm));
1351 static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm)
1353 return max(mm->hiwater_vm, mm->total_vm);
1356 static inline void update_hiwater_rss(struct mm_struct *mm)
1358 unsigned long _rss = get_mm_rss(mm);
1360 if ((mm)->hiwater_rss < _rss)
1361 (mm)->hiwater_rss = _rss;
1364 static inline void update_hiwater_vm(struct mm_struct *mm)
1366 if (mm->hiwater_vm < mm->total_vm)
1367 mm->hiwater_vm = mm->total_vm;
1370 static inline void reset_mm_hiwater_rss(struct mm_struct *mm)
1372 mm->hiwater_rss = get_mm_rss(mm);
1375 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss,
1376 struct mm_struct *mm)
1378 unsigned long hiwater_rss = get_mm_hiwater_rss(mm);
1380 if (*maxrss < hiwater_rss)
1381 *maxrss = hiwater_rss;
1384 #if defined(SPLIT_RSS_COUNTING)
1385 void sync_mm_rss(struct mm_struct *mm);
1387 static inline void sync_mm_rss(struct mm_struct *mm)
1392 int vma_wants_writenotify(struct vm_area_struct *vma);
1394 extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr,
1396 static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr,
1400 __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl));
1404 #ifdef __PAGETABLE_PUD_FOLDED
1405 static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd,
1406 unsigned long address)
1411 int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address);
1414 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1415 static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud,
1416 unsigned long address)
1421 static inline void mm_nr_pmds_init(struct mm_struct *mm) {}
1423 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1428 static inline void mm_inc_nr_pmds(struct mm_struct *mm) {}
1429 static inline void mm_dec_nr_pmds(struct mm_struct *mm) {}
1432 int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address);
1434 static inline void mm_nr_pmds_init(struct mm_struct *mm)
1436 atomic_long_set(&mm->nr_pmds, 0);
1439 static inline unsigned long mm_nr_pmds(struct mm_struct *mm)
1441 return atomic_long_read(&mm->nr_pmds);
1444 static inline void mm_inc_nr_pmds(struct mm_struct *mm)
1446 atomic_long_inc(&mm->nr_pmds);
1449 static inline void mm_dec_nr_pmds(struct mm_struct *mm)
1451 atomic_long_dec(&mm->nr_pmds);
1455 int __pte_alloc(struct mm_struct *mm, struct vm_area_struct *vma,
1456 pmd_t *pmd, unsigned long address);
1457 int __pte_alloc_kernel(pmd_t *pmd, unsigned long address);
1460 * The following ifdef needed to get the 4level-fixup.h header to work.
1461 * Remove it when 4level-fixup.h has been removed.
1463 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1464 static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address)
1466 return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))?
1467 NULL: pud_offset(pgd, address);
1470 static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address)
1472 return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))?
1473 NULL: pmd_offset(pud, address);
1475 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1477 #if USE_SPLIT_PTE_PTLOCKS
1478 #if ALLOC_SPLIT_PTLOCKS
1479 void __init ptlock_cache_init(void);
1480 extern bool ptlock_alloc(struct page *page);
1481 extern void ptlock_free(struct page *page);
1483 static inline spinlock_t *ptlock_ptr(struct page *page)
1487 #else /* ALLOC_SPLIT_PTLOCKS */
1488 static inline void ptlock_cache_init(void)
1492 static inline bool ptlock_alloc(struct page *page)
1497 static inline void ptlock_free(struct page *page)
1501 static inline spinlock_t *ptlock_ptr(struct page *page)
1505 #endif /* ALLOC_SPLIT_PTLOCKS */
1507 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1509 return ptlock_ptr(pmd_page(*pmd));
1512 static inline bool ptlock_init(struct page *page)
1515 * prep_new_page() initialize page->private (and therefore page->ptl)
1516 * with 0. Make sure nobody took it in use in between.
1518 * It can happen if arch try to use slab for page table allocation:
1519 * slab code uses page->slab_cache, which share storage with page->ptl.
1521 VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page);
1522 if (!ptlock_alloc(page))
1524 spin_lock_init(ptlock_ptr(page));
1528 /* Reset page->mapping so free_pages_check won't complain. */
1529 static inline void pte_lock_deinit(struct page *page)
1531 page->mapping = NULL;
1535 #else /* !USE_SPLIT_PTE_PTLOCKS */
1537 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1539 static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd)
1541 return &mm->page_table_lock;
1543 static inline void ptlock_cache_init(void) {}
1544 static inline bool ptlock_init(struct page *page) { return true; }
1545 static inline void pte_lock_deinit(struct page *page) {}
1546 #endif /* USE_SPLIT_PTE_PTLOCKS */
1548 static inline void pgtable_init(void)
1550 ptlock_cache_init();
1551 pgtable_cache_init();
1554 static inline bool pgtable_page_ctor(struct page *page)
1556 if (!ptlock_init(page))
1558 inc_zone_page_state(page, NR_PAGETABLE);
1562 static inline void pgtable_page_dtor(struct page *page)
1564 pte_lock_deinit(page);
1565 dec_zone_page_state(page, NR_PAGETABLE);
1568 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1570 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1571 pte_t *__pte = pte_offset_map(pmd, address); \
1577 #define pte_unmap_unlock(pte, ptl) do { \
1582 #define pte_alloc_map(mm, vma, pmd, address) \
1583 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, vma, \
1585 NULL: pte_offset_map(pmd, address))
1587 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1588 ((unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, NULL, \
1590 NULL: pte_offset_map_lock(mm, pmd, address, ptlp))
1592 #define pte_alloc_kernel(pmd, address) \
1593 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1594 NULL: pte_offset_kernel(pmd, address))
1596 #if USE_SPLIT_PMD_PTLOCKS
1598 static struct page *pmd_to_page(pmd_t *pmd)
1600 unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1);
1601 return virt_to_page((void *)((unsigned long) pmd & mask));
1604 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1606 return ptlock_ptr(pmd_to_page(pmd));
1609 static inline bool pgtable_pmd_page_ctor(struct page *page)
1611 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1612 page->pmd_huge_pte = NULL;
1614 return ptlock_init(page);
1617 static inline void pgtable_pmd_page_dtor(struct page *page)
1619 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1620 VM_BUG_ON_PAGE(page->pmd_huge_pte, page);
1625 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1629 static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd)
1631 return &mm->page_table_lock;
1634 static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; }
1635 static inline void pgtable_pmd_page_dtor(struct page *page) {}
1637 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1641 static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd)
1643 spinlock_t *ptl = pmd_lockptr(mm, pmd);
1648 extern void free_area_init(unsigned long * zones_size);
1649 extern void free_area_init_node(int nid, unsigned long * zones_size,
1650 unsigned long zone_start_pfn, unsigned long *zholes_size);
1651 extern void free_initmem(void);
1654 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1655 * into the buddy system. The freed pages will be poisoned with pattern
1656 * "poison" if it's within range [0, UCHAR_MAX].
1657 * Return pages freed into the buddy system.
1659 extern unsigned long free_reserved_area(void *start, void *end,
1660 int poison, char *s);
1662 #ifdef CONFIG_HIGHMEM
1664 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1665 * and totalram_pages.
1667 extern void free_highmem_page(struct page *page);
1670 extern void adjust_managed_page_count(struct page *page, long count);
1671 extern void mem_init_print_info(const char *str);
1673 extern void reserve_bootmem_region(unsigned long start, unsigned long end);
1675 /* Free the reserved page into the buddy system, so it gets managed. */
1676 static inline void __free_reserved_page(struct page *page)
1678 ClearPageReserved(page);
1679 init_page_count(page);
1683 static inline void free_reserved_page(struct page *page)
1685 __free_reserved_page(page);
1686 adjust_managed_page_count(page, 1);
1689 static inline void mark_page_reserved(struct page *page)
1691 SetPageReserved(page);
1692 adjust_managed_page_count(page, -1);
1696 * Default method to free all the __init memory into the buddy system.
1697 * The freed pages will be poisoned with pattern "poison" if it's within
1698 * range [0, UCHAR_MAX].
1699 * Return pages freed into the buddy system.
1701 static inline unsigned long free_initmem_default(int poison)
1703 extern char __init_begin[], __init_end[];
1705 return free_reserved_area(&__init_begin, &__init_end,
1706 poison, "unused kernel");
1709 static inline unsigned long get_num_physpages(void)
1712 unsigned long phys_pages = 0;
1714 for_each_online_node(nid)
1715 phys_pages += node_present_pages(nid);
1720 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1722 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1723 * zones, allocate the backing mem_map and account for memory holes in a more
1724 * architecture independent manner. This is a substitute for creating the
1725 * zone_sizes[] and zholes_size[] arrays and passing them to
1726 * free_area_init_node()
1728 * An architecture is expected to register range of page frames backed by
1729 * physical memory with memblock_add[_node]() before calling
1730 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1731 * usage, an architecture is expected to do something like
1733 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1735 * for_each_valid_physical_page_range()
1736 * memblock_add_node(base, size, nid)
1737 * free_area_init_nodes(max_zone_pfns);
1739 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1740 * registered physical page range. Similarly
1741 * sparse_memory_present_with_active_regions() calls memory_present() for
1742 * each range when SPARSEMEM is enabled.
1744 * See mm/page_alloc.c for more information on each function exposed by
1745 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1747 extern void free_area_init_nodes(unsigned long *max_zone_pfn);
1748 unsigned long node_map_pfn_alignment(void);
1749 unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn,
1750 unsigned long end_pfn);
1751 extern unsigned long absent_pages_in_range(unsigned long start_pfn,
1752 unsigned long end_pfn);
1753 extern void get_pfn_range_for_nid(unsigned int nid,
1754 unsigned long *start_pfn, unsigned long *end_pfn);
1755 extern unsigned long find_min_pfn_with_active_regions(void);
1756 extern void free_bootmem_with_active_regions(int nid,
1757 unsigned long max_low_pfn);
1758 extern void sparse_memory_present_with_active_regions(int nid);
1760 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1762 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1763 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1764 static inline int __early_pfn_to_nid(unsigned long pfn,
1765 struct mminit_pfnnid_cache *state)
1770 /* please see mm/page_alloc.c */
1771 extern int __meminit early_pfn_to_nid(unsigned long pfn);
1772 /* there is a per-arch backend function. */
1773 extern int __meminit __early_pfn_to_nid(unsigned long pfn,
1774 struct mminit_pfnnid_cache *state);
1777 extern void set_dma_reserve(unsigned long new_dma_reserve);
1778 extern void memmap_init_zone(unsigned long, int, unsigned long,
1779 unsigned long, enum memmap_context);
1780 extern void setup_per_zone_wmarks(void);
1781 extern int __meminit init_per_zone_wmark_min(void);
1782 extern void mem_init(void);
1783 extern void __init mmap_init(void);
1784 extern void show_mem(unsigned int flags);
1785 extern void si_meminfo(struct sysinfo * val);
1786 extern void si_meminfo_node(struct sysinfo *val, int nid);
1788 extern __printf(3, 4)
1789 void warn_alloc_failed(gfp_t gfp_mask, unsigned int order,
1790 const char *fmt, ...);
1792 extern void setup_per_cpu_pageset(void);
1794 extern void zone_pcp_update(struct zone *zone);
1795 extern void zone_pcp_reset(struct zone *zone);
1798 extern int min_free_kbytes;
1801 extern atomic_long_t mmap_pages_allocated;
1802 extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t);
1804 /* interval_tree.c */
1805 void vma_interval_tree_insert(struct vm_area_struct *node,
1806 struct rb_root *root);
1807 void vma_interval_tree_insert_after(struct vm_area_struct *node,
1808 struct vm_area_struct *prev,
1809 struct rb_root *root);
1810 void vma_interval_tree_remove(struct vm_area_struct *node,
1811 struct rb_root *root);
1812 struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root,
1813 unsigned long start, unsigned long last);
1814 struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node,
1815 unsigned long start, unsigned long last);
1817 #define vma_interval_tree_foreach(vma, root, start, last) \
1818 for (vma = vma_interval_tree_iter_first(root, start, last); \
1819 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1821 void anon_vma_interval_tree_insert(struct anon_vma_chain *node,
1822 struct rb_root *root);
1823 void anon_vma_interval_tree_remove(struct anon_vma_chain *node,
1824 struct rb_root *root);
1825 struct anon_vma_chain *anon_vma_interval_tree_iter_first(
1826 struct rb_root *root, unsigned long start, unsigned long last);
1827 struct anon_vma_chain *anon_vma_interval_tree_iter_next(
1828 struct anon_vma_chain *node, unsigned long start, unsigned long last);
1829 #ifdef CONFIG_DEBUG_VM_RB
1830 void anon_vma_interval_tree_verify(struct anon_vma_chain *node);
1833 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1834 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1835 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1838 extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin);
1839 extern int vma_adjust(struct vm_area_struct *vma, unsigned long start,
1840 unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert);
1841 extern struct vm_area_struct *vma_merge(struct mm_struct *,
1842 struct vm_area_struct *prev, unsigned long addr, unsigned long end,
1843 unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t,
1844 struct mempolicy *, struct vm_userfaultfd_ctx);
1845 extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *);
1846 extern int split_vma(struct mm_struct *,
1847 struct vm_area_struct *, unsigned long addr, int new_below);
1848 extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *);
1849 extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *,
1850 struct rb_node **, struct rb_node *);
1851 extern void unlink_file_vma(struct vm_area_struct *);
1852 extern struct vm_area_struct *copy_vma(struct vm_area_struct **,
1853 unsigned long addr, unsigned long len, pgoff_t pgoff,
1854 bool *need_rmap_locks);
1855 extern void exit_mmap(struct mm_struct *);
1857 static inline int check_data_rlimit(unsigned long rlim,
1859 unsigned long start,
1860 unsigned long end_data,
1861 unsigned long start_data)
1863 if (rlim < RLIM_INFINITY) {
1864 if (((new - start) + (end_data - start_data)) > rlim)
1871 extern int mm_take_all_locks(struct mm_struct *mm);
1872 extern void mm_drop_all_locks(struct mm_struct *mm);
1874 extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file);
1875 extern struct file *get_mm_exe_file(struct mm_struct *mm);
1877 extern int may_expand_vm(struct mm_struct *mm, unsigned long npages);
1878 extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm,
1879 unsigned long addr, unsigned long len,
1880 unsigned long flags,
1881 const struct vm_special_mapping *spec);
1882 /* This is an obsolete alternative to _install_special_mapping. */
1883 extern int install_special_mapping(struct mm_struct *mm,
1884 unsigned long addr, unsigned long len,
1885 unsigned long flags, struct page **pages);
1887 extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long);
1889 extern unsigned long mmap_region(struct file *file, unsigned long addr,
1890 unsigned long len, vm_flags_t vm_flags, unsigned long pgoff);
1891 extern unsigned long do_mmap(struct file *file, unsigned long addr,
1892 unsigned long len, unsigned long prot, unsigned long flags,
1893 vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate);
1894 extern int do_munmap(struct mm_struct *, unsigned long, size_t);
1896 static inline unsigned long
1897 do_mmap_pgoff(struct file *file, unsigned long addr,
1898 unsigned long len, unsigned long prot, unsigned long flags,
1899 unsigned long pgoff, unsigned long *populate)
1901 return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate);
1905 extern int __mm_populate(unsigned long addr, unsigned long len,
1907 static inline void mm_populate(unsigned long addr, unsigned long len)
1910 (void) __mm_populate(addr, len, 1);
1913 static inline void mm_populate(unsigned long addr, unsigned long len) {}
1916 /* These take the mm semaphore themselves */
1917 extern unsigned long vm_brk(unsigned long, unsigned long);
1918 extern int vm_munmap(unsigned long, size_t);
1919 extern unsigned long vm_mmap(struct file *, unsigned long,
1920 unsigned long, unsigned long,
1921 unsigned long, unsigned long);
1923 struct vm_unmapped_area_info {
1924 #define VM_UNMAPPED_AREA_TOPDOWN 1
1925 unsigned long flags;
1926 unsigned long length;
1927 unsigned long low_limit;
1928 unsigned long high_limit;
1929 unsigned long align_mask;
1930 unsigned long align_offset;
1933 extern unsigned long unmapped_area(struct vm_unmapped_area_info *info);
1934 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info);
1937 * Search for an unmapped address range.
1939 * We are looking for a range that:
1940 * - does not intersect with any VMA;
1941 * - is contained within the [low_limit, high_limit) interval;
1942 * - is at least the desired size.
1943 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
1945 static inline unsigned long
1946 vm_unmapped_area(struct vm_unmapped_area_info *info)
1948 if (info->flags & VM_UNMAPPED_AREA_TOPDOWN)
1949 return unmapped_area_topdown(info);
1951 return unmapped_area(info);
1955 extern void truncate_inode_pages(struct address_space *, loff_t);
1956 extern void truncate_inode_pages_range(struct address_space *,
1957 loff_t lstart, loff_t lend);
1958 extern void truncate_inode_pages_final(struct address_space *);
1960 /* generic vm_area_ops exported for stackable file systems */
1961 extern int filemap_fault(struct vm_area_struct *, struct vm_fault *);
1962 extern void filemap_map_pages(struct vm_area_struct *vma, struct vm_fault *vmf);
1963 extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1965 /* mm/page-writeback.c */
1966 int write_one_page(struct page *page, int wait);
1967 void task_dirty_inc(struct task_struct *tsk);
1970 #define VM_MAX_READAHEAD 128 /* kbytes */
1971 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
1973 int force_page_cache_readahead(struct address_space *mapping, struct file *filp,
1974 pgoff_t offset, unsigned long nr_to_read);
1976 void page_cache_sync_readahead(struct address_space *mapping,
1977 struct file_ra_state *ra,
1980 unsigned long size);
1982 void page_cache_async_readahead(struct address_space *mapping,
1983 struct file_ra_state *ra,
1987 unsigned long size);
1989 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
1990 extern int expand_stack(struct vm_area_struct *vma, unsigned long address);
1992 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
1993 extern int expand_downwards(struct vm_area_struct *vma,
1994 unsigned long address);
1996 extern int expand_upwards(struct vm_area_struct *vma, unsigned long address);
1998 #define expand_upwards(vma, address) (0)
2001 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2002 extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr);
2003 extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr,
2004 struct vm_area_struct **pprev);
2006 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2007 NULL if none. Assume start_addr < end_addr. */
2008 static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr)
2010 struct vm_area_struct * vma = find_vma(mm,start_addr);
2012 if (vma && end_addr <= vma->vm_start)
2017 static inline unsigned long vma_pages(struct vm_area_struct *vma)
2019 return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT;
2022 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2023 static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm,
2024 unsigned long vm_start, unsigned long vm_end)
2026 struct vm_area_struct *vma = find_vma(mm, vm_start);
2028 if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end))
2035 pgprot_t vm_get_page_prot(unsigned long vm_flags);
2036 void vma_set_page_prot(struct vm_area_struct *vma);
2038 static inline pgprot_t vm_get_page_prot(unsigned long vm_flags)
2042 static inline void vma_set_page_prot(struct vm_area_struct *vma)
2044 vma->vm_page_prot = vm_get_page_prot(vma->vm_flags);
2048 #ifdef CONFIG_NUMA_BALANCING
2049 unsigned long change_prot_numa(struct vm_area_struct *vma,
2050 unsigned long start, unsigned long end);
2053 struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr);
2054 int remap_pfn_range(struct vm_area_struct *, unsigned long addr,
2055 unsigned long pfn, unsigned long size, pgprot_t);
2056 int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *);
2057 int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr,
2059 int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr,
2061 int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len);
2064 struct page *follow_page_mask(struct vm_area_struct *vma,
2065 unsigned long address, unsigned int foll_flags,
2066 unsigned int *page_mask);
2068 static inline struct page *follow_page(struct vm_area_struct *vma,
2069 unsigned long address, unsigned int foll_flags)
2071 unsigned int unused_page_mask;
2072 return follow_page_mask(vma, address, foll_flags, &unused_page_mask);
2075 #define FOLL_WRITE 0x01 /* check pte is writable */
2076 #define FOLL_TOUCH 0x02 /* mark page accessed */
2077 #define FOLL_GET 0x04 /* do get_page on page */
2078 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2079 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2080 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2081 * and return without waiting upon it */
2082 #define FOLL_POPULATE 0x40 /* fault in page */
2083 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2084 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2085 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2086 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2087 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2088 #define FOLL_MLOCK 0x1000 /* lock present pages */
2090 typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr,
2092 extern int apply_to_page_range(struct mm_struct *mm, unsigned long address,
2093 unsigned long size, pte_fn_t fn, void *data);
2095 #ifdef CONFIG_PROC_FS
2096 void vm_stat_account(struct mm_struct *, unsigned long, struct file *, long);
2098 static inline void vm_stat_account(struct mm_struct *mm,
2099 unsigned long flags, struct file *file, long pages)
2101 mm->total_vm += pages;
2103 #endif /* CONFIG_PROC_FS */
2105 #ifdef CONFIG_DEBUG_PAGEALLOC
2106 extern bool _debug_pagealloc_enabled;
2107 extern void __kernel_map_pages(struct page *page, int numpages, int enable);
2109 static inline bool debug_pagealloc_enabled(void)
2111 return _debug_pagealloc_enabled;
2115 kernel_map_pages(struct page *page, int numpages, int enable)
2117 if (!debug_pagealloc_enabled())
2120 __kernel_map_pages(page, numpages, enable);
2122 #ifdef CONFIG_HIBERNATION
2123 extern bool kernel_page_present(struct page *page);
2124 #endif /* CONFIG_HIBERNATION */
2127 kernel_map_pages(struct page *page, int numpages, int enable) {}
2128 #ifdef CONFIG_HIBERNATION
2129 static inline bool kernel_page_present(struct page *page) { return true; }
2130 #endif /* CONFIG_HIBERNATION */
2133 #ifdef __HAVE_ARCH_GATE_AREA
2134 extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm);
2135 extern int in_gate_area_no_mm(unsigned long addr);
2136 extern int in_gate_area(struct mm_struct *mm, unsigned long addr);
2138 static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm)
2142 static inline int in_gate_area_no_mm(unsigned long addr) { return 0; }
2143 static inline int in_gate_area(struct mm_struct *mm, unsigned long addr)
2147 #endif /* __HAVE_ARCH_GATE_AREA */
2149 #ifdef CONFIG_SYSCTL
2150 extern int sysctl_drop_caches;
2151 int drop_caches_sysctl_handler(struct ctl_table *, int,
2152 void __user *, size_t *, loff_t *);
2155 void drop_slab(void);
2156 void drop_slab_node(int nid);
2159 #define randomize_va_space 0
2161 extern int randomize_va_space;
2164 const char * arch_vma_name(struct vm_area_struct *vma);
2165 void print_vma_addr(char *prefix, unsigned long rip);
2167 void sparse_mem_maps_populate_node(struct page **map_map,
2168 unsigned long pnum_begin,
2169 unsigned long pnum_end,
2170 unsigned long map_count,
2173 struct page *sparse_mem_map_populate(unsigned long pnum, int nid);
2174 pgd_t *vmemmap_pgd_populate(unsigned long addr, int node);
2175 pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node);
2176 pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node);
2177 pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node);
2178 void *vmemmap_alloc_block(unsigned long size, int node);
2179 void *vmemmap_alloc_block_buf(unsigned long size, int node);
2180 void vmemmap_verify(pte_t *, int, unsigned long, unsigned long);
2181 int vmemmap_populate_basepages(unsigned long start, unsigned long end,
2183 int vmemmap_populate(unsigned long start, unsigned long end, int node);
2184 void vmemmap_populate_print_last(void);
2185 #ifdef CONFIG_MEMORY_HOTPLUG
2186 void vmemmap_free(unsigned long start, unsigned long end);
2188 void register_page_bootmem_memmap(unsigned long section_nr, struct page *map,
2189 unsigned long size);
2192 MF_COUNT_INCREASED = 1 << 0,
2193 MF_ACTION_REQUIRED = 1 << 1,
2194 MF_MUST_KILL = 1 << 2,
2195 MF_SOFT_OFFLINE = 1 << 3,
2197 extern int memory_failure(unsigned long pfn, int trapno, int flags);
2198 extern void memory_failure_queue(unsigned long pfn, int trapno, int flags);
2199 extern int unpoison_memory(unsigned long pfn);
2200 extern int get_hwpoison_page(struct page *page);
2201 extern void put_hwpoison_page(struct page *page);
2202 extern int sysctl_memory_failure_early_kill;
2203 extern int sysctl_memory_failure_recovery;
2204 extern void shake_page(struct page *p, int access);
2205 extern atomic_long_t num_poisoned_pages;
2206 extern int soft_offline_page(struct page *page, int flags);
2210 * Error handlers for various types of pages.
2213 MF_IGNORED, /* Error: cannot be handled */
2214 MF_FAILED, /* Error: handling failed */
2215 MF_DELAYED, /* Will be handled later */
2216 MF_RECOVERED, /* Successfully recovered */
2219 enum mf_action_page_type {
2221 MF_MSG_KERNEL_HIGH_ORDER,
2223 MF_MSG_DIFFERENT_COMPOUND,
2224 MF_MSG_POISONED_HUGE,
2227 MF_MSG_UNMAP_FAILED,
2228 MF_MSG_DIRTY_SWAPCACHE,
2229 MF_MSG_CLEAN_SWAPCACHE,
2230 MF_MSG_DIRTY_MLOCKED_LRU,
2231 MF_MSG_CLEAN_MLOCKED_LRU,
2232 MF_MSG_DIRTY_UNEVICTABLE_LRU,
2233 MF_MSG_CLEAN_UNEVICTABLE_LRU,
2236 MF_MSG_TRUNCATED_LRU,
2242 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2243 extern void clear_huge_page(struct page *page,
2245 unsigned int pages_per_huge_page);
2246 extern void copy_user_huge_page(struct page *dst, struct page *src,
2247 unsigned long addr, struct vm_area_struct *vma,
2248 unsigned int pages_per_huge_page);
2249 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2251 extern struct page_ext_operations debug_guardpage_ops;
2252 extern struct page_ext_operations page_poisoning_ops;
2254 #ifdef CONFIG_DEBUG_PAGEALLOC
2255 extern unsigned int _debug_guardpage_minorder;
2256 extern bool _debug_guardpage_enabled;
2258 static inline unsigned int debug_guardpage_minorder(void)
2260 return _debug_guardpage_minorder;
2263 static inline bool debug_guardpage_enabled(void)
2265 return _debug_guardpage_enabled;
2268 static inline bool page_is_guard(struct page *page)
2270 struct page_ext *page_ext;
2272 if (!debug_guardpage_enabled())
2275 page_ext = lookup_page_ext(page);
2276 return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags);
2279 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2280 static inline bool debug_guardpage_enabled(void) { return false; }
2281 static inline bool page_is_guard(struct page *page) { return false; }
2282 #endif /* CONFIG_DEBUG_PAGEALLOC */
2284 #if MAX_NUMNODES > 1
2285 void __init setup_nr_node_ids(void);
2287 static inline void setup_nr_node_ids(void) {}
2290 #endif /* __KERNEL__ */
2291 #endif /* _LINUX_MM_H */