4 #include <linux/errno.h>
8 #include <linux/mmdebug.h>
10 #include <linux/bug.h>
11 #include <linux/list.h>
12 #include <linux/mmzone.h>
13 #include <linux/rbtree.h>
14 #include <linux/atomic.h>
15 #include <linux/debug_locks.h>
16 #include <linux/mm_types.h>
17 #include <linux/range.h>
18 #include <linux/pfn.h>
19 #include <linux/percpu-refcount.h>
20 #include <linux/bit_spinlock.h>
21 #include <linux/shrinker.h>
22 #include <linux/resource.h>
23 #include <linux/page_ext.h>
24 #include <linux/err.h>
25 #include <linux/page_ref.h>
29 struct anon_vma_chain
;
32 struct writeback_control
;
35 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
36 extern unsigned long max_mapnr
;
38 static inline void set_max_mapnr(unsigned long limit
)
43 static inline void set_max_mapnr(unsigned long limit
) { }
46 extern unsigned long totalram_pages
;
47 extern void * high_memory
;
48 extern int page_cluster
;
51 extern int sysctl_legacy_va_layout
;
53 #define sysctl_legacy_va_layout 0
56 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
57 extern const int mmap_rnd_bits_min
;
58 extern const int mmap_rnd_bits_max
;
59 extern int mmap_rnd_bits __read_mostly
;
61 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
62 extern const int mmap_rnd_compat_bits_min
;
63 extern const int mmap_rnd_compat_bits_max
;
64 extern int mmap_rnd_compat_bits __read_mostly
;
68 #include <asm/pgtable.h>
69 #include <asm/processor.h>
72 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
76 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
80 * To prevent common memory management code establishing
81 * a zero page mapping on a read fault.
82 * This macro should be defined within <asm/pgtable.h>.
83 * s390 does this to prevent multiplexing of hardware bits
84 * related to the physical page in case of virtualization.
86 #ifndef mm_forbids_zeropage
87 #define mm_forbids_zeropage(X) (0)
91 * Default maximum number of active map areas, this limits the number of vmas
92 * per mm struct. Users can overwrite this number by sysctl but there is a
95 * When a program's coredump is generated as ELF format, a section is created
96 * per a vma. In ELF, the number of sections is represented in unsigned short.
97 * This means the number of sections should be smaller than 65535 at coredump.
98 * Because the kernel adds some informative sections to a image of program at
99 * generating coredump, we need some margin. The number of extra sections is
100 * 1-3 now and depends on arch. We use "5" as safe margin, here.
102 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
103 * not a hard limit any more. Although some userspace tools can be surprised by
106 #define MAPCOUNT_ELF_CORE_MARGIN (5)
107 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
109 extern int sysctl_max_map_count
;
111 extern unsigned long sysctl_user_reserve_kbytes
;
112 extern unsigned long sysctl_admin_reserve_kbytes
;
114 extern int sysctl_overcommit_memory
;
115 extern int sysctl_overcommit_ratio
;
116 extern unsigned long sysctl_overcommit_kbytes
;
118 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
120 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
123 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
125 /* to align the pointer to the (next) page boundary */
126 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
128 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
129 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
132 * Linux kernel virtual memory manager primitives.
133 * The idea being to have a "virtual" mm in the same way
134 * we have a virtual fs - giving a cleaner interface to the
135 * mm details, and allowing different kinds of memory mappings
136 * (from shared memory to executable loading to arbitrary
140 extern struct kmem_cache
*vm_area_cachep
;
143 extern struct rb_root nommu_region_tree
;
144 extern struct rw_semaphore nommu_region_sem
;
146 extern unsigned int kobjsize(const void *objp
);
150 * vm_flags in vm_area_struct, see mm_types.h.
151 * When changing, update also include/trace/events/mmflags.h
153 #define VM_NONE 0x00000000
155 #define VM_READ 0x00000001 /* currently active flags */
156 #define VM_WRITE 0x00000002
157 #define VM_EXEC 0x00000004
158 #define VM_SHARED 0x00000008
160 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
161 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
162 #define VM_MAYWRITE 0x00000020
163 #define VM_MAYEXEC 0x00000040
164 #define VM_MAYSHARE 0x00000080
166 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
167 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
168 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
169 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
170 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
172 #define VM_LOCKED 0x00002000
173 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
175 /* Used by sys_madvise() */
176 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
177 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
179 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
180 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
181 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
182 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
183 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
184 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
185 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
186 #define VM_ARCH_2 0x02000000
187 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
189 #ifdef CONFIG_MEM_SOFT_DIRTY
190 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
192 # define VM_SOFTDIRTY 0
195 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
196 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
197 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
198 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
200 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
201 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
202 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
203 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
204 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
205 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
206 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
207 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
208 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
209 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
211 #if defined(CONFIG_X86)
212 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
213 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
214 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
215 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
216 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
217 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
218 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
220 #elif defined(CONFIG_PPC)
221 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
222 #elif defined(CONFIG_PARISC)
223 # define VM_GROWSUP VM_ARCH_1
224 #elif defined(CONFIG_METAG)
225 # define VM_GROWSUP VM_ARCH_1
226 #elif defined(CONFIG_IA64)
227 # define VM_GROWSUP VM_ARCH_1
228 #elif !defined(CONFIG_MMU)
229 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
232 #if defined(CONFIG_X86)
233 /* MPX specific bounds table or bounds directory */
234 # define VM_MPX VM_ARCH_2
238 # define VM_GROWSUP VM_NONE
241 /* Bits set in the VMA until the stack is in its final location */
242 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
244 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
245 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
248 #ifdef CONFIG_STACK_GROWSUP
249 #define VM_STACK VM_GROWSUP
251 #define VM_STACK VM_GROWSDOWN
254 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
257 * Special vmas that are non-mergable, non-mlock()able.
258 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
260 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
262 /* This mask defines which mm->def_flags a process can inherit its parent */
263 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
265 /* This mask is used to clear all the VMA flags used by mlock */
266 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
269 * mapping from the currently active vm_flags protection bits (the
270 * low four bits) to a page protection mask..
272 extern pgprot_t protection_map
[16];
274 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
275 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
276 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
277 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
278 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
279 #define FAULT_FLAG_TRIED 0x20 /* Second try */
280 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
281 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
282 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
285 * vm_fault is filled by the the pagefault handler and passed to the vma's
286 * ->fault function. The vma's ->fault is responsible for returning a bitmask
287 * of VM_FAULT_xxx flags that give details about how the fault was handled.
289 * MM layer fills up gfp_mask for page allocations but fault handler might
290 * alter it if its implementation requires a different allocation context.
292 * pgoff should be used in favour of virtual_address, if possible.
295 struct vm_area_struct
*vma
; /* Target VMA */
296 unsigned int flags
; /* FAULT_FLAG_xxx flags */
297 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
298 pgoff_t pgoff
; /* Logical page offset based on vma */
299 unsigned long address
; /* Faulting virtual address */
300 pmd_t
*pmd
; /* Pointer to pmd entry matching
302 pte_t orig_pte
; /* Value of PTE at the time of fault */
304 struct page
*cow_page
; /* Page handler may use for COW fault */
305 struct mem_cgroup
*memcg
; /* Cgroup cow_page belongs to */
306 struct page
*page
; /* ->fault handlers should return a
307 * page here, unless VM_FAULT_NOPAGE
308 * is set (which is also implied by
311 void *entry
; /* ->fault handler can alternatively
312 * return locked DAX entry. In that
313 * case handler should return
314 * VM_FAULT_DAX_LOCKED and fill in
317 /* These three entries are valid only while holding ptl lock */
318 pte_t
*pte
; /* Pointer to pte entry matching
319 * the 'address'. NULL if the page
320 * table hasn't been allocated.
322 spinlock_t
*ptl
; /* Page table lock.
323 * Protects pte page table if 'pte'
324 * is not NULL, otherwise pmd.
326 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
327 * vm_ops->map_pages() calls
328 * alloc_set_pte() from atomic context.
329 * do_fault_around() pre-allocates
330 * page table to avoid allocation from
336 * These are the virtual MM functions - opening of an area, closing and
337 * unmapping it (needed to keep files on disk up-to-date etc), pointer
338 * to the functions called when a no-page or a wp-page exception occurs.
340 struct vm_operations_struct
{
341 void (*open
)(struct vm_area_struct
* area
);
342 void (*close
)(struct vm_area_struct
* area
);
343 int (*mremap
)(struct vm_area_struct
* area
);
344 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
345 int (*pmd_fault
)(struct vm_area_struct
*, unsigned long address
,
346 pmd_t
*, unsigned int flags
);
347 void (*map_pages
)(struct vm_fault
*vmf
,
348 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
350 /* notification that a previously read-only page is about to become
351 * writable, if an error is returned it will cause a SIGBUS */
352 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
354 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
355 int (*pfn_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
357 /* called by access_process_vm when get_user_pages() fails, typically
358 * for use by special VMAs that can switch between memory and hardware
360 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
361 void *buf
, int len
, int write
);
363 /* Called by the /proc/PID/maps code to ask the vma whether it
364 * has a special name. Returning non-NULL will also cause this
365 * vma to be dumped unconditionally. */
366 const char *(*name
)(struct vm_area_struct
*vma
);
370 * set_policy() op must add a reference to any non-NULL @new mempolicy
371 * to hold the policy upon return. Caller should pass NULL @new to
372 * remove a policy and fall back to surrounding context--i.e. do not
373 * install a MPOL_DEFAULT policy, nor the task or system default
376 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
379 * get_policy() op must add reference [mpol_get()] to any policy at
380 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
381 * in mm/mempolicy.c will do this automatically.
382 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
383 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
384 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
385 * must return NULL--i.e., do not "fallback" to task or system default
388 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
392 * Called by vm_normal_page() for special PTEs to find the
393 * page for @addr. This is useful if the default behavior
394 * (using pte_page()) would not find the correct page.
396 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
403 #define page_private(page) ((page)->private)
404 #define set_page_private(page, v) ((page)->private = (v))
406 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
407 static inline int pmd_devmap(pmd_t pmd
)
414 * FIXME: take this include out, include page-flags.h in
415 * files which need it (119 of them)
417 #include <linux/page-flags.h>
418 #include <linux/huge_mm.h>
421 * Methods to modify the page usage count.
423 * What counts for a page usage:
424 * - cache mapping (page->mapping)
425 * - private data (page->private)
426 * - page mapped in a task's page tables, each mapping
427 * is counted separately
429 * Also, many kernel routines increase the page count before a critical
430 * routine so they can be sure the page doesn't go away from under them.
434 * Drop a ref, return true if the refcount fell to zero (the page has no users)
436 static inline int put_page_testzero(struct page
*page
)
438 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
439 return page_ref_dec_and_test(page
);
443 * Try to grab a ref unless the page has a refcount of zero, return false if
445 * This can be called when MMU is off so it must not access
446 * any of the virtual mappings.
448 static inline int get_page_unless_zero(struct page
*page
)
450 return page_ref_add_unless(page
, 1, 0);
453 extern int page_is_ram(unsigned long pfn
);
461 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
464 /* Support for virtually mapped pages */
465 struct page
*vmalloc_to_page(const void *addr
);
466 unsigned long vmalloc_to_pfn(const void *addr
);
469 * Determine if an address is within the vmalloc range
471 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
472 * is no special casing required.
474 static inline bool is_vmalloc_addr(const void *x
)
477 unsigned long addr
= (unsigned long)x
;
479 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
485 extern int is_vmalloc_or_module_addr(const void *x
);
487 static inline int is_vmalloc_or_module_addr(const void *x
)
493 extern void kvfree(const void *addr
);
495 static inline atomic_t
*compound_mapcount_ptr(struct page
*page
)
497 return &page
[1].compound_mapcount
;
500 static inline int compound_mapcount(struct page
*page
)
502 VM_BUG_ON_PAGE(!PageCompound(page
), page
);
503 page
= compound_head(page
);
504 return atomic_read(compound_mapcount_ptr(page
)) + 1;
508 * The atomic page->_mapcount, starts from -1: so that transitions
509 * both from it and to it can be tracked, using atomic_inc_and_test
510 * and atomic_add_negative(-1).
512 static inline void page_mapcount_reset(struct page
*page
)
514 atomic_set(&(page
)->_mapcount
, -1);
517 int __page_mapcount(struct page
*page
);
519 static inline int page_mapcount(struct page
*page
)
521 VM_BUG_ON_PAGE(PageSlab(page
), page
);
523 if (unlikely(PageCompound(page
)))
524 return __page_mapcount(page
);
525 return atomic_read(&page
->_mapcount
) + 1;
528 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
529 int total_mapcount(struct page
*page
);
530 int page_trans_huge_mapcount(struct page
*page
, int *total_mapcount
);
532 static inline int total_mapcount(struct page
*page
)
534 return page_mapcount(page
);
536 static inline int page_trans_huge_mapcount(struct page
*page
,
539 int mapcount
= page_mapcount(page
);
541 *total_mapcount
= mapcount
;
546 static inline struct page
*virt_to_head_page(const void *x
)
548 struct page
*page
= virt_to_page(x
);
550 return compound_head(page
);
553 void __put_page(struct page
*page
);
555 void put_pages_list(struct list_head
*pages
);
557 void split_page(struct page
*page
, unsigned int order
);
560 * Compound pages have a destructor function. Provide a
561 * prototype for that function and accessor functions.
562 * These are _only_ valid on the head of a compound page.
564 typedef void compound_page_dtor(struct page
*);
566 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
567 enum compound_dtor_id
{
570 #ifdef CONFIG_HUGETLB_PAGE
573 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
578 extern compound_page_dtor
* const compound_page_dtors
[];
580 static inline void set_compound_page_dtor(struct page
*page
,
581 enum compound_dtor_id compound_dtor
)
583 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
584 page
[1].compound_dtor
= compound_dtor
;
587 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
589 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
590 return compound_page_dtors
[page
[1].compound_dtor
];
593 static inline unsigned int compound_order(struct page
*page
)
597 return page
[1].compound_order
;
600 static inline void set_compound_order(struct page
*page
, unsigned int order
)
602 page
[1].compound_order
= order
;
605 void free_compound_page(struct page
*page
);
609 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
610 * servicing faults for write access. In the normal case, do always want
611 * pte_mkwrite. But get_user_pages can cause write faults for mappings
612 * that do not have writing enabled, when used by access_process_vm.
614 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
616 if (likely(vma
->vm_flags
& VM_WRITE
))
617 pte
= pte_mkwrite(pte
);
621 int alloc_set_pte(struct vm_fault
*vmf
, struct mem_cgroup
*memcg
,
626 * Multiple processes may "see" the same page. E.g. for untouched
627 * mappings of /dev/null, all processes see the same page full of
628 * zeroes, and text pages of executables and shared libraries have
629 * only one copy in memory, at most, normally.
631 * For the non-reserved pages, page_count(page) denotes a reference count.
632 * page_count() == 0 means the page is free. page->lru is then used for
633 * freelist management in the buddy allocator.
634 * page_count() > 0 means the page has been allocated.
636 * Pages are allocated by the slab allocator in order to provide memory
637 * to kmalloc and kmem_cache_alloc. In this case, the management of the
638 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
639 * unless a particular usage is carefully commented. (the responsibility of
640 * freeing the kmalloc memory is the caller's, of course).
642 * A page may be used by anyone else who does a __get_free_page().
643 * In this case, page_count still tracks the references, and should only
644 * be used through the normal accessor functions. The top bits of page->flags
645 * and page->virtual store page management information, but all other fields
646 * are unused and could be used privately, carefully. The management of this
647 * page is the responsibility of the one who allocated it, and those who have
648 * subsequently been given references to it.
650 * The other pages (we may call them "pagecache pages") are completely
651 * managed by the Linux memory manager: I/O, buffers, swapping etc.
652 * The following discussion applies only to them.
654 * A pagecache page contains an opaque `private' member, which belongs to the
655 * page's address_space. Usually, this is the address of a circular list of
656 * the page's disk buffers. PG_private must be set to tell the VM to call
657 * into the filesystem to release these pages.
659 * A page may belong to an inode's memory mapping. In this case, page->mapping
660 * is the pointer to the inode, and page->index is the file offset of the page,
661 * in units of PAGE_SIZE.
663 * If pagecache pages are not associated with an inode, they are said to be
664 * anonymous pages. These may become associated with the swapcache, and in that
665 * case PG_swapcache is set, and page->private is an offset into the swapcache.
667 * In either case (swapcache or inode backed), the pagecache itself holds one
668 * reference to the page. Setting PG_private should also increment the
669 * refcount. The each user mapping also has a reference to the page.
671 * The pagecache pages are stored in a per-mapping radix tree, which is
672 * rooted at mapping->page_tree, and indexed by offset.
673 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
674 * lists, we instead now tag pages as dirty/writeback in the radix tree.
676 * All pagecache pages may be subject to I/O:
677 * - inode pages may need to be read from disk,
678 * - inode pages which have been modified and are MAP_SHARED may need
679 * to be written back to the inode on disk,
680 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
681 * modified may need to be swapped out to swap space and (later) to be read
686 * The zone field is never updated after free_area_init_core()
687 * sets it, so none of the operations on it need to be atomic.
690 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
691 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
692 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
693 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
694 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
697 * Define the bit shifts to access each section. For non-existent
698 * sections we define the shift as 0; that plus a 0 mask ensures
699 * the compiler will optimise away reference to them.
701 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
702 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
703 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
704 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
706 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
707 #ifdef NODE_NOT_IN_PAGE_FLAGS
708 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
709 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
710 SECTIONS_PGOFF : ZONES_PGOFF)
712 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
713 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
714 NODES_PGOFF : ZONES_PGOFF)
717 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
719 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
720 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
723 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
724 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
725 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
726 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
727 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
729 static inline enum zone_type
page_zonenum(const struct page
*page
)
731 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
734 #ifdef CONFIG_ZONE_DEVICE
735 void get_zone_device_page(struct page
*page
);
736 void put_zone_device_page(struct page
*page
);
737 static inline bool is_zone_device_page(const struct page
*page
)
739 return page_zonenum(page
) == ZONE_DEVICE
;
742 static inline void get_zone_device_page(struct page
*page
)
745 static inline void put_zone_device_page(struct page
*page
)
748 static inline bool is_zone_device_page(const struct page
*page
)
754 static inline void get_page(struct page
*page
)
756 page
= compound_head(page
);
758 * Getting a normal page or the head of a compound page
759 * requires to already have an elevated page->_refcount.
761 VM_BUG_ON_PAGE(page_ref_count(page
) <= 0, page
);
764 if (unlikely(is_zone_device_page(page
)))
765 get_zone_device_page(page
);
768 static inline void put_page(struct page
*page
)
770 page
= compound_head(page
);
772 if (put_page_testzero(page
))
775 if (unlikely(is_zone_device_page(page
)))
776 put_zone_device_page(page
);
779 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
780 #define SECTION_IN_PAGE_FLAGS
784 * The identification function is mainly used by the buddy allocator for
785 * determining if two pages could be buddies. We are not really identifying
786 * the zone since we could be using the section number id if we do not have
787 * node id available in page flags.
788 * We only guarantee that it will return the same value for two combinable
791 static inline int page_zone_id(struct page
*page
)
793 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
796 static inline int zone_to_nid(struct zone
*zone
)
805 #ifdef NODE_NOT_IN_PAGE_FLAGS
806 extern int page_to_nid(const struct page
*page
);
808 static inline int page_to_nid(const struct page
*page
)
810 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
814 #ifdef CONFIG_NUMA_BALANCING
815 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
817 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
820 static inline int cpupid_to_pid(int cpupid
)
822 return cpupid
& LAST__PID_MASK
;
825 static inline int cpupid_to_cpu(int cpupid
)
827 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
830 static inline int cpupid_to_nid(int cpupid
)
832 return cpu_to_node(cpupid_to_cpu(cpupid
));
835 static inline bool cpupid_pid_unset(int cpupid
)
837 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
840 static inline bool cpupid_cpu_unset(int cpupid
)
842 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
845 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
847 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
850 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
851 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
852 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
854 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
857 static inline int page_cpupid_last(struct page
*page
)
859 return page
->_last_cpupid
;
861 static inline void page_cpupid_reset_last(struct page
*page
)
863 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
866 static inline int page_cpupid_last(struct page
*page
)
868 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
871 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
873 static inline void page_cpupid_reset_last(struct page
*page
)
875 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
877 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
878 #else /* !CONFIG_NUMA_BALANCING */
879 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
881 return page_to_nid(page
); /* XXX */
884 static inline int page_cpupid_last(struct page
*page
)
886 return page_to_nid(page
); /* XXX */
889 static inline int cpupid_to_nid(int cpupid
)
894 static inline int cpupid_to_pid(int cpupid
)
899 static inline int cpupid_to_cpu(int cpupid
)
904 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
909 static inline bool cpupid_pid_unset(int cpupid
)
914 static inline void page_cpupid_reset_last(struct page
*page
)
918 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
922 #endif /* CONFIG_NUMA_BALANCING */
924 static inline struct zone
*page_zone(const struct page
*page
)
926 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
929 static inline pg_data_t
*page_pgdat(const struct page
*page
)
931 return NODE_DATA(page_to_nid(page
));
934 #ifdef SECTION_IN_PAGE_FLAGS
935 static inline void set_page_section(struct page
*page
, unsigned long section
)
937 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
938 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
941 static inline unsigned long page_to_section(const struct page
*page
)
943 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
947 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
949 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
950 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
953 static inline void set_page_node(struct page
*page
, unsigned long node
)
955 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
956 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
959 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
960 unsigned long node
, unsigned long pfn
)
962 set_page_zone(page
, zone
);
963 set_page_node(page
, node
);
964 #ifdef SECTION_IN_PAGE_FLAGS
965 set_page_section(page
, pfn_to_section_nr(pfn
));
970 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
972 return page
->mem_cgroup
;
974 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
976 WARN_ON_ONCE(!rcu_read_lock_held());
977 return READ_ONCE(page
->mem_cgroup
);
980 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
984 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
986 WARN_ON_ONCE(!rcu_read_lock_held());
992 * Some inline functions in vmstat.h depend on page_zone()
994 #include <linux/vmstat.h>
996 static __always_inline
void *lowmem_page_address(const struct page
*page
)
998 return page_to_virt(page
);
1001 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1002 #define HASHED_PAGE_VIRTUAL
1005 #if defined(WANT_PAGE_VIRTUAL)
1006 static inline void *page_address(const struct page
*page
)
1008 return page
->virtual;
1010 static inline void set_page_address(struct page
*page
, void *address
)
1012 page
->virtual = address
;
1014 #define page_address_init() do { } while(0)
1017 #if defined(HASHED_PAGE_VIRTUAL)
1018 void *page_address(const struct page
*page
);
1019 void set_page_address(struct page
*page
, void *virtual);
1020 void page_address_init(void);
1023 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1024 #define page_address(page) lowmem_page_address(page)
1025 #define set_page_address(page, address) do { } while(0)
1026 #define page_address_init() do { } while(0)
1029 extern void *page_rmapping(struct page
*page
);
1030 extern struct anon_vma
*page_anon_vma(struct page
*page
);
1031 extern struct address_space
*page_mapping(struct page
*page
);
1033 extern struct address_space
*__page_file_mapping(struct page
*);
1036 struct address_space
*page_file_mapping(struct page
*page
)
1038 if (unlikely(PageSwapCache(page
)))
1039 return __page_file_mapping(page
);
1041 return page
->mapping
;
1044 extern pgoff_t
__page_file_index(struct page
*page
);
1047 * Return the pagecache index of the passed page. Regular pagecache pages
1048 * use ->index whereas swapcache pages use swp_offset(->private)
1050 static inline pgoff_t
page_index(struct page
*page
)
1052 if (unlikely(PageSwapCache(page
)))
1053 return __page_file_index(page
);
1057 bool page_mapped(struct page
*page
);
1058 struct address_space
*page_mapping(struct page
*page
);
1061 * Return true only if the page has been allocated with
1062 * ALLOC_NO_WATERMARKS and the low watermark was not
1063 * met implying that the system is under some pressure.
1065 static inline bool page_is_pfmemalloc(struct page
*page
)
1068 * Page index cannot be this large so this must be
1069 * a pfmemalloc page.
1071 return page
->index
== -1UL;
1075 * Only to be called by the page allocator on a freshly allocated
1078 static inline void set_page_pfmemalloc(struct page
*page
)
1083 static inline void clear_page_pfmemalloc(struct page
*page
)
1089 * Different kinds of faults, as returned by handle_mm_fault().
1090 * Used to decide whether a process gets delivered SIGBUS or
1091 * just gets major/minor fault counters bumped up.
1094 #define VM_FAULT_OOM 0x0001
1095 #define VM_FAULT_SIGBUS 0x0002
1096 #define VM_FAULT_MAJOR 0x0004
1097 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1098 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1099 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1100 #define VM_FAULT_SIGSEGV 0x0040
1102 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1103 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1104 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1105 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1106 #define VM_FAULT_DAX_LOCKED 0x1000 /* ->fault has locked DAX entry */
1107 #define VM_FAULT_DONE_COW 0x2000 /* ->fault has fully handled COW */
1109 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1111 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1112 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1115 /* Encode hstate index for a hwpoisoned large page */
1116 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1117 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1120 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1122 extern void pagefault_out_of_memory(void);
1124 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1127 * Flags passed to show_mem() and show_free_areas() to suppress output in
1130 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1132 extern void show_free_areas(unsigned int flags
);
1133 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
1135 int shmem_zero_setup(struct vm_area_struct
*);
1137 bool shmem_mapping(struct address_space
*mapping
);
1139 static inline bool shmem_mapping(struct address_space
*mapping
)
1145 extern bool can_do_mlock(void);
1146 extern int user_shm_lock(size_t, struct user_struct
*);
1147 extern void user_shm_unlock(size_t, struct user_struct
*);
1150 * Parameter block passed down to zap_pte_range in exceptional cases.
1152 struct zap_details
{
1153 struct address_space
*check_mapping
; /* Check page->mapping if set */
1154 pgoff_t first_index
; /* Lowest page->index to unmap */
1155 pgoff_t last_index
; /* Highest page->index to unmap */
1156 bool ignore_dirty
; /* Ignore dirty pages */
1157 bool check_swap_entries
; /* Check also swap entries */
1160 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1162 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
1165 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1166 unsigned long size
);
1167 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1168 unsigned long size
, struct zap_details
*);
1169 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1170 unsigned long start
, unsigned long end
);
1173 * mm_walk - callbacks for walk_page_range
1174 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1175 * this handler is required to be able to handle
1176 * pmd_trans_huge() pmds. They may simply choose to
1177 * split_huge_page() instead of handling it explicitly.
1178 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1179 * @pte_hole: if set, called for each hole at all levels
1180 * @hugetlb_entry: if set, called for each hugetlb entry
1181 * @test_walk: caller specific callback function to determine whether
1182 * we walk over the current vma or not. Returning 0
1183 * value means "do page table walk over the current vma,"
1184 * and a negative one means "abort current page table walk
1185 * right now." 1 means "skip the current vma."
1186 * @mm: mm_struct representing the target process of page table walk
1187 * @vma: vma currently walked (NULL if walking outside vmas)
1188 * @private: private data for callbacks' usage
1190 * (see the comment on walk_page_range() for more details)
1193 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1194 unsigned long next
, struct mm_walk
*walk
);
1195 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1196 unsigned long next
, struct mm_walk
*walk
);
1197 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1198 struct mm_walk
*walk
);
1199 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1200 unsigned long addr
, unsigned long next
,
1201 struct mm_walk
*walk
);
1202 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1203 struct mm_walk
*walk
);
1204 struct mm_struct
*mm
;
1205 struct vm_area_struct
*vma
;
1209 int walk_page_range(unsigned long addr
, unsigned long end
,
1210 struct mm_walk
*walk
);
1211 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1212 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1213 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1214 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1215 struct vm_area_struct
*vma
);
1216 void unmap_mapping_range(struct address_space
*mapping
,
1217 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1218 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1219 unsigned long *pfn
);
1220 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1221 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1222 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1223 void *buf
, int len
, int write
);
1225 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1226 loff_t
const holebegin
, loff_t
const holelen
)
1228 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1231 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1232 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1233 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1234 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1235 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1236 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1237 int invalidate_inode_page(struct page
*page
);
1240 extern int handle_mm_fault(struct vm_area_struct
*vma
, unsigned long address
,
1241 unsigned int flags
);
1242 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1243 unsigned long address
, unsigned int fault_flags
,
1246 static inline int handle_mm_fault(struct vm_area_struct
*vma
,
1247 unsigned long address
, unsigned int flags
)
1249 /* should never happen if there's no MMU */
1251 return VM_FAULT_SIGBUS
;
1253 static inline int fixup_user_fault(struct task_struct
*tsk
,
1254 struct mm_struct
*mm
, unsigned long address
,
1255 unsigned int fault_flags
, bool *unlocked
)
1257 /* should never happen if there's no MMU */
1263 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
,
1264 unsigned int gup_flags
);
1265 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1266 void *buf
, int len
, unsigned int gup_flags
);
1268 long get_user_pages_remote(struct task_struct
*tsk
, struct mm_struct
*mm
,
1269 unsigned long start
, unsigned long nr_pages
,
1270 unsigned int gup_flags
, struct page
**pages
,
1271 struct vm_area_struct
**vmas
, int *locked
);
1272 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
1273 unsigned int gup_flags
, struct page
**pages
,
1274 struct vm_area_struct
**vmas
);
1275 long get_user_pages_locked(unsigned long start
, unsigned long nr_pages
,
1276 unsigned int gup_flags
, struct page
**pages
, int *locked
);
1277 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
1278 struct page
**pages
, unsigned int gup_flags
);
1279 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1280 struct page
**pages
);
1282 /* Container for pinned pfns / pages */
1283 struct frame_vector
{
1284 unsigned int nr_allocated
; /* Number of frames we have space for */
1285 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1286 bool got_ref
; /* Did we pin pages by getting page ref? */
1287 bool is_pfns
; /* Does array contain pages or pfns? */
1288 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1289 * pfns_vector_pages() or pfns_vector_pfns()
1293 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1294 void frame_vector_destroy(struct frame_vector
*vec
);
1295 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1296 unsigned int gup_flags
, struct frame_vector
*vec
);
1297 void put_vaddr_frames(struct frame_vector
*vec
);
1298 int frame_vector_to_pages(struct frame_vector
*vec
);
1299 void frame_vector_to_pfns(struct frame_vector
*vec
);
1301 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1303 return vec
->nr_frames
;
1306 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1309 int err
= frame_vector_to_pages(vec
);
1312 return ERR_PTR(err
);
1314 return (struct page
**)(vec
->ptrs
);
1317 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1320 frame_vector_to_pfns(vec
);
1321 return (unsigned long *)(vec
->ptrs
);
1325 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1326 struct page
**pages
);
1327 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1328 struct page
*get_dump_page(unsigned long addr
);
1330 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1331 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1332 unsigned int length
);
1334 int __set_page_dirty_nobuffers(struct page
*page
);
1335 int __set_page_dirty_no_writeback(struct page
*page
);
1336 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1338 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1339 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1340 struct bdi_writeback
*wb
);
1341 int set_page_dirty(struct page
*page
);
1342 int set_page_dirty_lock(struct page
*page
);
1343 void cancel_dirty_page(struct page
*page
);
1344 int clear_page_dirty_for_io(struct page
*page
);
1346 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1348 /* Is the vma a continuation of the stack vma above it? */
1349 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1351 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1354 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1356 return !vma
->vm_ops
;
1359 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1362 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1363 (vma
->vm_start
== addr
) &&
1364 !vma_growsdown(vma
->vm_prev
, addr
);
1367 /* Is the vma a continuation of the stack vma below it? */
1368 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1370 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1373 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1376 return (vma
->vm_flags
& VM_GROWSUP
) &&
1377 (vma
->vm_end
== addr
) &&
1378 !vma_growsup(vma
->vm_next
, addr
);
1381 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
1383 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1384 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1385 unsigned long new_addr
, unsigned long len
,
1386 bool need_rmap_locks
);
1387 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1388 unsigned long end
, pgprot_t newprot
,
1389 int dirty_accountable
, int prot_numa
);
1390 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1391 struct vm_area_struct
**pprev
, unsigned long start
,
1392 unsigned long end
, unsigned long newflags
);
1395 * doesn't attempt to fault and will return short.
1397 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1398 struct page
**pages
);
1400 * per-process(per-mm_struct) statistics.
1402 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1404 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1406 #ifdef SPLIT_RSS_COUNTING
1408 * counter is updated in asynchronous manner and may go to minus.
1409 * But it's never be expected number for users.
1414 return (unsigned long)val
;
1417 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1419 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1422 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1424 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1427 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1429 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1432 /* Optimized variant when page is already known not to be PageAnon */
1433 static inline int mm_counter_file(struct page
*page
)
1435 if (PageSwapBacked(page
))
1436 return MM_SHMEMPAGES
;
1437 return MM_FILEPAGES
;
1440 static inline int mm_counter(struct page
*page
)
1443 return MM_ANONPAGES
;
1444 return mm_counter_file(page
);
1447 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1449 return get_mm_counter(mm
, MM_FILEPAGES
) +
1450 get_mm_counter(mm
, MM_ANONPAGES
) +
1451 get_mm_counter(mm
, MM_SHMEMPAGES
);
1454 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1456 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1459 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1461 return max(mm
->hiwater_vm
, mm
->total_vm
);
1464 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1466 unsigned long _rss
= get_mm_rss(mm
);
1468 if ((mm
)->hiwater_rss
< _rss
)
1469 (mm
)->hiwater_rss
= _rss
;
1472 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1474 if (mm
->hiwater_vm
< mm
->total_vm
)
1475 mm
->hiwater_vm
= mm
->total_vm
;
1478 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1480 mm
->hiwater_rss
= get_mm_rss(mm
);
1483 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1484 struct mm_struct
*mm
)
1486 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1488 if (*maxrss
< hiwater_rss
)
1489 *maxrss
= hiwater_rss
;
1492 #if defined(SPLIT_RSS_COUNTING)
1493 void sync_mm_rss(struct mm_struct
*mm
);
1495 static inline void sync_mm_rss(struct mm_struct
*mm
)
1500 #ifndef __HAVE_ARCH_PTE_DEVMAP
1501 static inline int pte_devmap(pte_t pte
)
1507 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
1509 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1511 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1515 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1519 #ifdef __PAGETABLE_PUD_FOLDED
1520 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1521 unsigned long address
)
1526 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1529 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1530 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1531 unsigned long address
)
1536 static inline void mm_nr_pmds_init(struct mm_struct
*mm
) {}
1538 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1543 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1544 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1547 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1549 static inline void mm_nr_pmds_init(struct mm_struct
*mm
)
1551 atomic_long_set(&mm
->nr_pmds
, 0);
1554 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1556 return atomic_long_read(&mm
->nr_pmds
);
1559 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1561 atomic_long_inc(&mm
->nr_pmds
);
1564 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1566 atomic_long_dec(&mm
->nr_pmds
);
1570 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
1571 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1574 * The following ifdef needed to get the 4level-fixup.h header to work.
1575 * Remove it when 4level-fixup.h has been removed.
1577 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1578 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1580 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1581 NULL
: pud_offset(pgd
, address
);
1584 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1586 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1587 NULL
: pmd_offset(pud
, address
);
1589 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1591 #if USE_SPLIT_PTE_PTLOCKS
1592 #if ALLOC_SPLIT_PTLOCKS
1593 void __init
ptlock_cache_init(void);
1594 extern bool ptlock_alloc(struct page
*page
);
1595 extern void ptlock_free(struct page
*page
);
1597 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1601 #else /* ALLOC_SPLIT_PTLOCKS */
1602 static inline void ptlock_cache_init(void)
1606 static inline bool ptlock_alloc(struct page
*page
)
1611 static inline void ptlock_free(struct page
*page
)
1615 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1619 #endif /* ALLOC_SPLIT_PTLOCKS */
1621 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1623 return ptlock_ptr(pmd_page(*pmd
));
1626 static inline bool ptlock_init(struct page
*page
)
1629 * prep_new_page() initialize page->private (and therefore page->ptl)
1630 * with 0. Make sure nobody took it in use in between.
1632 * It can happen if arch try to use slab for page table allocation:
1633 * slab code uses page->slab_cache, which share storage with page->ptl.
1635 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1636 if (!ptlock_alloc(page
))
1638 spin_lock_init(ptlock_ptr(page
));
1642 /* Reset page->mapping so free_pages_check won't complain. */
1643 static inline void pte_lock_deinit(struct page
*page
)
1645 page
->mapping
= NULL
;
1649 #else /* !USE_SPLIT_PTE_PTLOCKS */
1651 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1653 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1655 return &mm
->page_table_lock
;
1657 static inline void ptlock_cache_init(void) {}
1658 static inline bool ptlock_init(struct page
*page
) { return true; }
1659 static inline void pte_lock_deinit(struct page
*page
) {}
1660 #endif /* USE_SPLIT_PTE_PTLOCKS */
1662 static inline void pgtable_init(void)
1664 ptlock_cache_init();
1665 pgtable_cache_init();
1668 static inline bool pgtable_page_ctor(struct page
*page
)
1670 if (!ptlock_init(page
))
1672 inc_zone_page_state(page
, NR_PAGETABLE
);
1676 static inline void pgtable_page_dtor(struct page
*page
)
1678 pte_lock_deinit(page
);
1679 dec_zone_page_state(page
, NR_PAGETABLE
);
1682 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1684 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1685 pte_t *__pte = pte_offset_map(pmd, address); \
1691 #define pte_unmap_unlock(pte, ptl) do { \
1696 #define pte_alloc(mm, pmd, address) \
1697 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1699 #define pte_alloc_map(mm, pmd, address) \
1700 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1702 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1703 (pte_alloc(mm, pmd, address) ? \
1704 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1706 #define pte_alloc_kernel(pmd, address) \
1707 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1708 NULL: pte_offset_kernel(pmd, address))
1710 #if USE_SPLIT_PMD_PTLOCKS
1712 static struct page
*pmd_to_page(pmd_t
*pmd
)
1714 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1715 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1718 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1720 return ptlock_ptr(pmd_to_page(pmd
));
1723 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1725 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1726 page
->pmd_huge_pte
= NULL
;
1728 return ptlock_init(page
);
1731 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1733 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1734 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1739 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1743 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1745 return &mm
->page_table_lock
;
1748 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1749 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1751 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1755 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1757 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1762 extern void free_area_init(unsigned long * zones_size
);
1763 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1764 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1765 extern void free_initmem(void);
1768 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1769 * into the buddy system. The freed pages will be poisoned with pattern
1770 * "poison" if it's within range [0, UCHAR_MAX].
1771 * Return pages freed into the buddy system.
1773 extern unsigned long free_reserved_area(void *start
, void *end
,
1774 int poison
, char *s
);
1776 #ifdef CONFIG_HIGHMEM
1778 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1779 * and totalram_pages.
1781 extern void free_highmem_page(struct page
*page
);
1784 extern void adjust_managed_page_count(struct page
*page
, long count
);
1785 extern void mem_init_print_info(const char *str
);
1787 extern void reserve_bootmem_region(phys_addr_t start
, phys_addr_t end
);
1789 /* Free the reserved page into the buddy system, so it gets managed. */
1790 static inline void __free_reserved_page(struct page
*page
)
1792 ClearPageReserved(page
);
1793 init_page_count(page
);
1797 static inline void free_reserved_page(struct page
*page
)
1799 __free_reserved_page(page
);
1800 adjust_managed_page_count(page
, 1);
1803 static inline void mark_page_reserved(struct page
*page
)
1805 SetPageReserved(page
);
1806 adjust_managed_page_count(page
, -1);
1810 * Default method to free all the __init memory into the buddy system.
1811 * The freed pages will be poisoned with pattern "poison" if it's within
1812 * range [0, UCHAR_MAX].
1813 * Return pages freed into the buddy system.
1815 static inline unsigned long free_initmem_default(int poison
)
1817 extern char __init_begin
[], __init_end
[];
1819 return free_reserved_area(&__init_begin
, &__init_end
,
1820 poison
, "unused kernel");
1823 static inline unsigned long get_num_physpages(void)
1826 unsigned long phys_pages
= 0;
1828 for_each_online_node(nid
)
1829 phys_pages
+= node_present_pages(nid
);
1834 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1836 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1837 * zones, allocate the backing mem_map and account for memory holes in a more
1838 * architecture independent manner. This is a substitute for creating the
1839 * zone_sizes[] and zholes_size[] arrays and passing them to
1840 * free_area_init_node()
1842 * An architecture is expected to register range of page frames backed by
1843 * physical memory with memblock_add[_node]() before calling
1844 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1845 * usage, an architecture is expected to do something like
1847 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1849 * for_each_valid_physical_page_range()
1850 * memblock_add_node(base, size, nid)
1851 * free_area_init_nodes(max_zone_pfns);
1853 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1854 * registered physical page range. Similarly
1855 * sparse_memory_present_with_active_regions() calls memory_present() for
1856 * each range when SPARSEMEM is enabled.
1858 * See mm/page_alloc.c for more information on each function exposed by
1859 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1861 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1862 unsigned long node_map_pfn_alignment(void);
1863 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1864 unsigned long end_pfn
);
1865 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1866 unsigned long end_pfn
);
1867 extern void get_pfn_range_for_nid(unsigned int nid
,
1868 unsigned long *start_pfn
, unsigned long *end_pfn
);
1869 extern unsigned long find_min_pfn_with_active_regions(void);
1870 extern void free_bootmem_with_active_regions(int nid
,
1871 unsigned long max_low_pfn
);
1872 extern void sparse_memory_present_with_active_regions(int nid
);
1874 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1876 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1877 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1878 static inline int __early_pfn_to_nid(unsigned long pfn
,
1879 struct mminit_pfnnid_cache
*state
)
1884 /* please see mm/page_alloc.c */
1885 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1886 /* there is a per-arch backend function. */
1887 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
1888 struct mminit_pfnnid_cache
*state
);
1891 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1892 extern void memmap_init_zone(unsigned long, int, unsigned long,
1893 unsigned long, enum memmap_context
);
1894 extern void setup_per_zone_wmarks(void);
1895 extern int __meminit
init_per_zone_wmark_min(void);
1896 extern void mem_init(void);
1897 extern void __init
mmap_init(void);
1898 extern void show_mem(unsigned int flags
);
1899 extern long si_mem_available(void);
1900 extern void si_meminfo(struct sysinfo
* val
);
1901 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1902 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
1903 extern unsigned long arch_reserved_kernel_pages(void);
1906 extern __printf(2, 3)
1907 void warn_alloc(gfp_t gfp_mask
, const char *fmt
, ...);
1909 extern void setup_per_cpu_pageset(void);
1911 extern void zone_pcp_update(struct zone
*zone
);
1912 extern void zone_pcp_reset(struct zone
*zone
);
1915 extern int min_free_kbytes
;
1916 extern int watermark_scale_factor
;
1919 extern atomic_long_t mmap_pages_allocated
;
1920 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1922 /* interval_tree.c */
1923 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1924 struct rb_root
*root
);
1925 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1926 struct vm_area_struct
*prev
,
1927 struct rb_root
*root
);
1928 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1929 struct rb_root
*root
);
1930 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1931 unsigned long start
, unsigned long last
);
1932 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1933 unsigned long start
, unsigned long last
);
1935 #define vma_interval_tree_foreach(vma, root, start, last) \
1936 for (vma = vma_interval_tree_iter_first(root, start, last); \
1937 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1939 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1940 struct rb_root
*root
);
1941 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1942 struct rb_root
*root
);
1943 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1944 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1945 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1946 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1947 #ifdef CONFIG_DEBUG_VM_RB
1948 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1951 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1952 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1953 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1956 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1957 extern int __vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1958 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
,
1959 struct vm_area_struct
*expand
);
1960 static inline int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1961 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
)
1963 return __vma_adjust(vma
, start
, end
, pgoff
, insert
, NULL
);
1965 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
1966 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
1967 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
1968 struct mempolicy
*, struct vm_userfaultfd_ctx
);
1969 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
1970 extern int split_vma(struct mm_struct
*,
1971 struct vm_area_struct
*, unsigned long addr
, int new_below
);
1972 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
1973 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
1974 struct rb_node
**, struct rb_node
*);
1975 extern void unlink_file_vma(struct vm_area_struct
*);
1976 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
1977 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
1978 bool *need_rmap_locks
);
1979 extern void exit_mmap(struct mm_struct
*);
1981 static inline int check_data_rlimit(unsigned long rlim
,
1983 unsigned long start
,
1984 unsigned long end_data
,
1985 unsigned long start_data
)
1987 if (rlim
< RLIM_INFINITY
) {
1988 if (((new - start
) + (end_data
- start_data
)) > rlim
)
1995 extern int mm_take_all_locks(struct mm_struct
*mm
);
1996 extern void mm_drop_all_locks(struct mm_struct
*mm
);
1998 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
1999 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2000 extern struct file
*get_task_exe_file(struct task_struct
*task
);
2002 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2003 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2005 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
2006 const struct vm_special_mapping
*sm
);
2007 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2008 unsigned long addr
, unsigned long len
,
2009 unsigned long flags
,
2010 const struct vm_special_mapping
*spec
);
2011 /* This is an obsolete alternative to _install_special_mapping. */
2012 extern int install_special_mapping(struct mm_struct
*mm
,
2013 unsigned long addr
, unsigned long len
,
2014 unsigned long flags
, struct page
**pages
);
2016 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
2018 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
2019 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
2020 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
2021 unsigned long len
, unsigned long prot
, unsigned long flags
,
2022 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
);
2023 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
2025 static inline unsigned long
2026 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
2027 unsigned long len
, unsigned long prot
, unsigned long flags
,
2028 unsigned long pgoff
, unsigned long *populate
)
2030 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
);
2034 extern int __mm_populate(unsigned long addr
, unsigned long len
,
2036 static inline void mm_populate(unsigned long addr
, unsigned long len
)
2039 (void) __mm_populate(addr
, len
, 1);
2042 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
2045 /* These take the mm semaphore themselves */
2046 extern int __must_check
vm_brk(unsigned long, unsigned long);
2047 extern int vm_munmap(unsigned long, size_t);
2048 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
2049 unsigned long, unsigned long,
2050 unsigned long, unsigned long);
2052 struct vm_unmapped_area_info
{
2053 #define VM_UNMAPPED_AREA_TOPDOWN 1
2054 unsigned long flags
;
2055 unsigned long length
;
2056 unsigned long low_limit
;
2057 unsigned long high_limit
;
2058 unsigned long align_mask
;
2059 unsigned long align_offset
;
2062 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
2063 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
2066 * Search for an unmapped address range.
2068 * We are looking for a range that:
2069 * - does not intersect with any VMA;
2070 * - is contained within the [low_limit, high_limit) interval;
2071 * - is at least the desired size.
2072 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2074 static inline unsigned long
2075 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
2077 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
2078 return unmapped_area_topdown(info
);
2080 return unmapped_area(info
);
2084 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2085 extern void truncate_inode_pages_range(struct address_space
*,
2086 loff_t lstart
, loff_t lend
);
2087 extern void truncate_inode_pages_final(struct address_space
*);
2089 /* generic vm_area_ops exported for stackable file systems */
2090 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
2091 extern void filemap_map_pages(struct vm_fault
*vmf
,
2092 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
2093 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
2095 /* mm/page-writeback.c */
2096 int write_one_page(struct page
*page
, int wait
);
2097 void task_dirty_inc(struct task_struct
*tsk
);
2100 #define VM_MAX_READAHEAD 128 /* kbytes */
2101 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2103 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2104 pgoff_t offset
, unsigned long nr_to_read
);
2106 void page_cache_sync_readahead(struct address_space
*mapping
,
2107 struct file_ra_state
*ra
,
2110 unsigned long size
);
2112 void page_cache_async_readahead(struct address_space
*mapping
,
2113 struct file_ra_state
*ra
,
2117 unsigned long size
);
2119 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2120 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2122 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2123 extern int expand_downwards(struct vm_area_struct
*vma
,
2124 unsigned long address
);
2126 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2128 #define expand_upwards(vma, address) (0)
2131 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2132 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2133 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2134 struct vm_area_struct
**pprev
);
2136 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2137 NULL if none. Assume start_addr < end_addr. */
2138 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2140 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2142 if (vma
&& end_addr
<= vma
->vm_start
)
2147 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2149 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2152 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2153 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2154 unsigned long vm_start
, unsigned long vm_end
)
2156 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2158 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2165 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2166 void vma_set_page_prot(struct vm_area_struct
*vma
);
2168 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2172 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2174 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2178 #ifdef CONFIG_NUMA_BALANCING
2179 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2180 unsigned long start
, unsigned long end
);
2183 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2184 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2185 unsigned long pfn
, unsigned long size
, pgprot_t
);
2186 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2187 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2189 int vm_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2190 unsigned long pfn
, pgprot_t pgprot
);
2191 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2193 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2196 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2197 unsigned long address
, unsigned int foll_flags
,
2198 unsigned int *page_mask
);
2200 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2201 unsigned long address
, unsigned int foll_flags
)
2203 unsigned int unused_page_mask
;
2204 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2207 #define FOLL_WRITE 0x01 /* check pte is writable */
2208 #define FOLL_TOUCH 0x02 /* mark page accessed */
2209 #define FOLL_GET 0x04 /* do get_page on page */
2210 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2211 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2212 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2213 * and return without waiting upon it */
2214 #define FOLL_POPULATE 0x40 /* fault in page */
2215 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2216 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2217 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2218 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2219 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2220 #define FOLL_MLOCK 0x1000 /* lock present pages */
2221 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2222 #define FOLL_COW 0x4000 /* internal GUP flag */
2224 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2226 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2227 unsigned long size
, pte_fn_t fn
, void *data
);
2230 #ifdef CONFIG_PAGE_POISONING
2231 extern bool page_poisoning_enabled(void);
2232 extern void kernel_poison_pages(struct page
*page
, int numpages
, int enable
);
2233 extern bool page_is_poisoned(struct page
*page
);
2235 static inline bool page_poisoning_enabled(void) { return false; }
2236 static inline void kernel_poison_pages(struct page
*page
, int numpages
,
2238 static inline bool page_is_poisoned(struct page
*page
) { return false; }
2241 #ifdef CONFIG_DEBUG_PAGEALLOC
2242 extern bool _debug_pagealloc_enabled
;
2243 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2245 static inline bool debug_pagealloc_enabled(void)
2247 return _debug_pagealloc_enabled
;
2251 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2253 if (!debug_pagealloc_enabled())
2256 __kernel_map_pages(page
, numpages
, enable
);
2258 #ifdef CONFIG_HIBERNATION
2259 extern bool kernel_page_present(struct page
*page
);
2260 #endif /* CONFIG_HIBERNATION */
2261 #else /* CONFIG_DEBUG_PAGEALLOC */
2263 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2264 #ifdef CONFIG_HIBERNATION
2265 static inline bool kernel_page_present(struct page
*page
) { return true; }
2266 #endif /* CONFIG_HIBERNATION */
2267 static inline bool debug_pagealloc_enabled(void)
2271 #endif /* CONFIG_DEBUG_PAGEALLOC */
2273 #ifdef __HAVE_ARCH_GATE_AREA
2274 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2275 extern int in_gate_area_no_mm(unsigned long addr
);
2276 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2278 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2282 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2283 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2287 #endif /* __HAVE_ARCH_GATE_AREA */
2289 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
2291 #ifdef CONFIG_SYSCTL
2292 extern int sysctl_drop_caches
;
2293 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2294 void __user
*, size_t *, loff_t
*);
2297 void drop_slab(void);
2298 void drop_slab_node(int nid
);
2301 #define randomize_va_space 0
2303 extern int randomize_va_space
;
2306 const char * arch_vma_name(struct vm_area_struct
*vma
);
2307 void print_vma_addr(char *prefix
, unsigned long rip
);
2309 void sparse_mem_maps_populate_node(struct page
**map_map
,
2310 unsigned long pnum_begin
,
2311 unsigned long pnum_end
,
2312 unsigned long map_count
,
2315 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2316 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2317 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2318 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2319 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2320 void *vmemmap_alloc_block(unsigned long size
, int node
);
2322 void *__vmemmap_alloc_block_buf(unsigned long size
, int node
,
2323 struct vmem_altmap
*altmap
);
2324 static inline void *vmemmap_alloc_block_buf(unsigned long size
, int node
)
2326 return __vmemmap_alloc_block_buf(size
, node
, NULL
);
2329 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2330 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2332 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2333 void vmemmap_populate_print_last(void);
2334 #ifdef CONFIG_MEMORY_HOTPLUG
2335 void vmemmap_free(unsigned long start
, unsigned long end
);
2337 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2338 unsigned long size
);
2341 MF_COUNT_INCREASED
= 1 << 0,
2342 MF_ACTION_REQUIRED
= 1 << 1,
2343 MF_MUST_KILL
= 1 << 2,
2344 MF_SOFT_OFFLINE
= 1 << 3,
2346 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2347 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2348 extern int unpoison_memory(unsigned long pfn
);
2349 extern int get_hwpoison_page(struct page
*page
);
2350 #define put_hwpoison_page(page) put_page(page)
2351 extern int sysctl_memory_failure_early_kill
;
2352 extern int sysctl_memory_failure_recovery
;
2353 extern void shake_page(struct page
*p
, int access
);
2354 extern atomic_long_t num_poisoned_pages
;
2355 extern int soft_offline_page(struct page
*page
, int flags
);
2359 * Error handlers for various types of pages.
2362 MF_IGNORED
, /* Error: cannot be handled */
2363 MF_FAILED
, /* Error: handling failed */
2364 MF_DELAYED
, /* Will be handled later */
2365 MF_RECOVERED
, /* Successfully recovered */
2368 enum mf_action_page_type
{
2370 MF_MSG_KERNEL_HIGH_ORDER
,
2372 MF_MSG_DIFFERENT_COMPOUND
,
2373 MF_MSG_POISONED_HUGE
,
2376 MF_MSG_UNMAP_FAILED
,
2377 MF_MSG_DIRTY_SWAPCACHE
,
2378 MF_MSG_CLEAN_SWAPCACHE
,
2379 MF_MSG_DIRTY_MLOCKED_LRU
,
2380 MF_MSG_CLEAN_MLOCKED_LRU
,
2381 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2382 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2385 MF_MSG_TRUNCATED_LRU
,
2391 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2392 extern void clear_huge_page(struct page
*page
,
2394 unsigned int pages_per_huge_page
);
2395 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2396 unsigned long addr
, struct vm_area_struct
*vma
,
2397 unsigned int pages_per_huge_page
);
2398 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2400 extern struct page_ext_operations debug_guardpage_ops
;
2401 extern struct page_ext_operations page_poisoning_ops
;
2403 #ifdef CONFIG_DEBUG_PAGEALLOC
2404 extern unsigned int _debug_guardpage_minorder
;
2405 extern bool _debug_guardpage_enabled
;
2407 static inline unsigned int debug_guardpage_minorder(void)
2409 return _debug_guardpage_minorder
;
2412 static inline bool debug_guardpage_enabled(void)
2414 return _debug_guardpage_enabled
;
2417 static inline bool page_is_guard(struct page
*page
)
2419 struct page_ext
*page_ext
;
2421 if (!debug_guardpage_enabled())
2424 page_ext
= lookup_page_ext(page
);
2425 if (unlikely(!page_ext
))
2428 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2431 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2432 static inline bool debug_guardpage_enabled(void) { return false; }
2433 static inline bool page_is_guard(struct page
*page
) { return false; }
2434 #endif /* CONFIG_DEBUG_PAGEALLOC */
2436 #if MAX_NUMNODES > 1
2437 void __init
setup_nr_node_ids(void);
2439 static inline void setup_nr_node_ids(void) {}
2442 #endif /* __KERNEL__ */
2443 #endif /* _LINUX_MM_H */