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 #define lm_alias(x) __va(__pa_symbol(x))
84 * To prevent common memory management code establishing
85 * a zero page mapping on a read fault.
86 * This macro should be defined within <asm/pgtable.h>.
87 * s390 does this to prevent multiplexing of hardware bits
88 * related to the physical page in case of virtualization.
90 #ifndef mm_forbids_zeropage
91 #define mm_forbids_zeropage(X) (0)
95 * Default maximum number of active map areas, this limits the number of vmas
96 * per mm struct. Users can overwrite this number by sysctl but there is a
99 * When a program's coredump is generated as ELF format, a section is created
100 * per a vma. In ELF, the number of sections is represented in unsigned short.
101 * This means the number of sections should be smaller than 65535 at coredump.
102 * Because the kernel adds some informative sections to a image of program at
103 * generating coredump, we need some margin. The number of extra sections is
104 * 1-3 now and depends on arch. We use "5" as safe margin, here.
106 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
107 * not a hard limit any more. Although some userspace tools can be surprised by
110 #define MAPCOUNT_ELF_CORE_MARGIN (5)
111 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
113 extern int sysctl_max_map_count
;
115 extern unsigned long sysctl_user_reserve_kbytes
;
116 extern unsigned long sysctl_admin_reserve_kbytes
;
118 extern int sysctl_overcommit_memory
;
119 extern int sysctl_overcommit_ratio
;
120 extern unsigned long sysctl_overcommit_kbytes
;
122 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
124 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
127 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
129 /* to align the pointer to the (next) page boundary */
130 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
132 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
133 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
136 * Linux kernel virtual memory manager primitives.
137 * The idea being to have a "virtual" mm in the same way
138 * we have a virtual fs - giving a cleaner interface to the
139 * mm details, and allowing different kinds of memory mappings
140 * (from shared memory to executable loading to arbitrary
144 extern struct kmem_cache
*vm_area_cachep
;
147 extern struct rb_root nommu_region_tree
;
148 extern struct rw_semaphore nommu_region_sem
;
150 extern unsigned int kobjsize(const void *objp
);
154 * vm_flags in vm_area_struct, see mm_types.h.
155 * When changing, update also include/trace/events/mmflags.h
157 #define VM_NONE 0x00000000
159 #define VM_READ 0x00000001 /* currently active flags */
160 #define VM_WRITE 0x00000002
161 #define VM_EXEC 0x00000004
162 #define VM_SHARED 0x00000008
164 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
165 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
166 #define VM_MAYWRITE 0x00000020
167 #define VM_MAYEXEC 0x00000040
168 #define VM_MAYSHARE 0x00000080
170 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
171 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
172 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
173 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
174 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
176 #define VM_LOCKED 0x00002000
177 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
179 /* Used by sys_madvise() */
180 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
181 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
183 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
184 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
185 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
186 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
187 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
188 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
189 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
190 #define VM_ARCH_2 0x02000000
191 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
193 #ifdef CONFIG_MEM_SOFT_DIRTY
194 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
196 # define VM_SOFTDIRTY 0
199 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
200 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
201 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
202 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
204 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
205 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
206 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
207 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
208 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
209 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
210 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
211 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
212 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
213 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
215 #if defined(CONFIG_X86)
216 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
217 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
218 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
219 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
220 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
221 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
222 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
224 #elif defined(CONFIG_PPC)
225 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
226 #elif defined(CONFIG_PARISC)
227 # define VM_GROWSUP VM_ARCH_1
228 #elif defined(CONFIG_METAG)
229 # define VM_GROWSUP VM_ARCH_1
230 #elif defined(CONFIG_IA64)
231 # define VM_GROWSUP VM_ARCH_1
232 #elif !defined(CONFIG_MMU)
233 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
236 #if defined(CONFIG_X86)
237 /* MPX specific bounds table or bounds directory */
238 # define VM_MPX VM_ARCH_2
242 # define VM_GROWSUP VM_NONE
245 /* Bits set in the VMA until the stack is in its final location */
246 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
248 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
249 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
252 #ifdef CONFIG_STACK_GROWSUP
253 #define VM_STACK VM_GROWSUP
255 #define VM_STACK VM_GROWSDOWN
258 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
261 * Special vmas that are non-mergable, non-mlock()able.
262 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
264 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
266 /* This mask defines which mm->def_flags a process can inherit its parent */
267 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
269 /* This mask is used to clear all the VMA flags used by mlock */
270 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
273 * mapping from the currently active vm_flags protection bits (the
274 * low four bits) to a page protection mask..
276 extern pgprot_t protection_map
[16];
278 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
279 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
280 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
281 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
282 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
283 #define FAULT_FLAG_TRIED 0x20 /* Second try */
284 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
285 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
286 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
288 #define FAULT_FLAG_TRACE \
289 { FAULT_FLAG_WRITE, "WRITE" }, \
290 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
291 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
292 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
293 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
294 { FAULT_FLAG_TRIED, "TRIED" }, \
295 { FAULT_FLAG_USER, "USER" }, \
296 { FAULT_FLAG_REMOTE, "REMOTE" }, \
297 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
300 * vm_fault is filled by the the pagefault handler and passed to the vma's
301 * ->fault function. The vma's ->fault is responsible for returning a bitmask
302 * of VM_FAULT_xxx flags that give details about how the fault was handled.
304 * MM layer fills up gfp_mask for page allocations but fault handler might
305 * alter it if its implementation requires a different allocation context.
307 * pgoff should be used in favour of virtual_address, if possible.
310 struct vm_area_struct
*vma
; /* Target VMA */
311 unsigned int flags
; /* FAULT_FLAG_xxx flags */
312 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
313 pgoff_t pgoff
; /* Logical page offset based on vma */
314 unsigned long address
; /* Faulting virtual address */
315 pmd_t
*pmd
; /* Pointer to pmd entry matching
317 pud_t
*pud
; /* Pointer to pud entry matching
320 pte_t orig_pte
; /* Value of PTE at the time of fault */
322 struct page
*cow_page
; /* Page handler may use for COW fault */
323 struct mem_cgroup
*memcg
; /* Cgroup cow_page belongs to */
324 struct page
*page
; /* ->fault handlers should return a
325 * page here, unless VM_FAULT_NOPAGE
326 * is set (which is also implied by
329 /* These three entries are valid only while holding ptl lock */
330 pte_t
*pte
; /* Pointer to pte entry matching
331 * the 'address'. NULL if the page
332 * table hasn't been allocated.
334 spinlock_t
*ptl
; /* Page table lock.
335 * Protects pte page table if 'pte'
336 * is not NULL, otherwise pmd.
338 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
339 * vm_ops->map_pages() calls
340 * alloc_set_pte() from atomic context.
341 * do_fault_around() pre-allocates
342 * page table to avoid allocation from
347 /* page entry size for vm->huge_fault() */
348 enum page_entry_size
{
355 * These are the virtual MM functions - opening of an area, closing and
356 * unmapping it (needed to keep files on disk up-to-date etc), pointer
357 * to the functions called when a no-page or a wp-page exception occurs.
359 struct vm_operations_struct
{
360 void (*open
)(struct vm_area_struct
* area
);
361 void (*close
)(struct vm_area_struct
* area
);
362 int (*mremap
)(struct vm_area_struct
* area
);
363 int (*fault
)(struct vm_fault
*vmf
);
364 int (*huge_fault
)(struct vm_fault
*vmf
, enum page_entry_size pe_size
);
365 void (*map_pages
)(struct vm_fault
*vmf
,
366 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
368 /* notification that a previously read-only page is about to become
369 * writable, if an error is returned it will cause a SIGBUS */
370 int (*page_mkwrite
)(struct vm_fault
*vmf
);
372 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
373 int (*pfn_mkwrite
)(struct vm_fault
*vmf
);
375 /* called by access_process_vm when get_user_pages() fails, typically
376 * for use by special VMAs that can switch between memory and hardware
378 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
379 void *buf
, int len
, int write
);
381 /* Called by the /proc/PID/maps code to ask the vma whether it
382 * has a special name. Returning non-NULL will also cause this
383 * vma to be dumped unconditionally. */
384 const char *(*name
)(struct vm_area_struct
*vma
);
388 * set_policy() op must add a reference to any non-NULL @new mempolicy
389 * to hold the policy upon return. Caller should pass NULL @new to
390 * remove a policy and fall back to surrounding context--i.e. do not
391 * install a MPOL_DEFAULT policy, nor the task or system default
394 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
397 * get_policy() op must add reference [mpol_get()] to any policy at
398 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
399 * in mm/mempolicy.c will do this automatically.
400 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
401 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
402 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
403 * must return NULL--i.e., do not "fallback" to task or system default
406 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
410 * Called by vm_normal_page() for special PTEs to find the
411 * page for @addr. This is useful if the default behavior
412 * (using pte_page()) would not find the correct page.
414 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
421 #define page_private(page) ((page)->private)
422 #define set_page_private(page, v) ((page)->private = (v))
424 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
425 static inline int pmd_devmap(pmd_t pmd
)
429 static inline int pud_devmap(pud_t pud
)
436 * FIXME: take this include out, include page-flags.h in
437 * files which need it (119 of them)
439 #include <linux/page-flags.h>
440 #include <linux/huge_mm.h>
443 * Methods to modify the page usage count.
445 * What counts for a page usage:
446 * - cache mapping (page->mapping)
447 * - private data (page->private)
448 * - page mapped in a task's page tables, each mapping
449 * is counted separately
451 * Also, many kernel routines increase the page count before a critical
452 * routine so they can be sure the page doesn't go away from under them.
456 * Drop a ref, return true if the refcount fell to zero (the page has no users)
458 static inline int put_page_testzero(struct page
*page
)
460 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
461 return page_ref_dec_and_test(page
);
465 * Try to grab a ref unless the page has a refcount of zero, return false if
467 * This can be called when MMU is off so it must not access
468 * any of the virtual mappings.
470 static inline int get_page_unless_zero(struct page
*page
)
472 return page_ref_add_unless(page
, 1, 0);
475 extern int page_is_ram(unsigned long pfn
);
483 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
486 /* Support for virtually mapped pages */
487 struct page
*vmalloc_to_page(const void *addr
);
488 unsigned long vmalloc_to_pfn(const void *addr
);
491 * Determine if an address is within the vmalloc range
493 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
494 * is no special casing required.
496 static inline bool is_vmalloc_addr(const void *x
)
499 unsigned long addr
= (unsigned long)x
;
501 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
507 extern int is_vmalloc_or_module_addr(const void *x
);
509 static inline int is_vmalloc_or_module_addr(const void *x
)
515 extern void kvfree(const void *addr
);
517 static inline atomic_t
*compound_mapcount_ptr(struct page
*page
)
519 return &page
[1].compound_mapcount
;
522 static inline int compound_mapcount(struct page
*page
)
524 VM_BUG_ON_PAGE(!PageCompound(page
), page
);
525 page
= compound_head(page
);
526 return atomic_read(compound_mapcount_ptr(page
)) + 1;
530 * The atomic page->_mapcount, starts from -1: so that transitions
531 * both from it and to it can be tracked, using atomic_inc_and_test
532 * and atomic_add_negative(-1).
534 static inline void page_mapcount_reset(struct page
*page
)
536 atomic_set(&(page
)->_mapcount
, -1);
539 int __page_mapcount(struct page
*page
);
541 static inline int page_mapcount(struct page
*page
)
543 VM_BUG_ON_PAGE(PageSlab(page
), page
);
545 if (unlikely(PageCompound(page
)))
546 return __page_mapcount(page
);
547 return atomic_read(&page
->_mapcount
) + 1;
550 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
551 int total_mapcount(struct page
*page
);
552 int page_trans_huge_mapcount(struct page
*page
, int *total_mapcount
);
554 static inline int total_mapcount(struct page
*page
)
556 return page_mapcount(page
);
558 static inline int page_trans_huge_mapcount(struct page
*page
,
561 int mapcount
= page_mapcount(page
);
563 *total_mapcount
= mapcount
;
568 static inline struct page
*virt_to_head_page(const void *x
)
570 struct page
*page
= virt_to_page(x
);
572 return compound_head(page
);
575 void __put_page(struct page
*page
);
577 void put_pages_list(struct list_head
*pages
);
579 void split_page(struct page
*page
, unsigned int order
);
582 * Compound pages have a destructor function. Provide a
583 * prototype for that function and accessor functions.
584 * These are _only_ valid on the head of a compound page.
586 typedef void compound_page_dtor(struct page
*);
588 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
589 enum compound_dtor_id
{
592 #ifdef CONFIG_HUGETLB_PAGE
595 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
600 extern compound_page_dtor
* const compound_page_dtors
[];
602 static inline void set_compound_page_dtor(struct page
*page
,
603 enum compound_dtor_id compound_dtor
)
605 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
606 page
[1].compound_dtor
= compound_dtor
;
609 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
611 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
612 return compound_page_dtors
[page
[1].compound_dtor
];
615 static inline unsigned int compound_order(struct page
*page
)
619 return page
[1].compound_order
;
622 static inline void set_compound_order(struct page
*page
, unsigned int order
)
624 page
[1].compound_order
= order
;
627 void free_compound_page(struct page
*page
);
631 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
632 * servicing faults for write access. In the normal case, do always want
633 * pte_mkwrite. But get_user_pages can cause write faults for mappings
634 * that do not have writing enabled, when used by access_process_vm.
636 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
638 if (likely(vma
->vm_flags
& VM_WRITE
))
639 pte
= pte_mkwrite(pte
);
643 int alloc_set_pte(struct vm_fault
*vmf
, struct mem_cgroup
*memcg
,
645 int finish_fault(struct vm_fault
*vmf
);
646 int finish_mkwrite_fault(struct vm_fault
*vmf
);
650 * Multiple processes may "see" the same page. E.g. for untouched
651 * mappings of /dev/null, all processes see the same page full of
652 * zeroes, and text pages of executables and shared libraries have
653 * only one copy in memory, at most, normally.
655 * For the non-reserved pages, page_count(page) denotes a reference count.
656 * page_count() == 0 means the page is free. page->lru is then used for
657 * freelist management in the buddy allocator.
658 * page_count() > 0 means the page has been allocated.
660 * Pages are allocated by the slab allocator in order to provide memory
661 * to kmalloc and kmem_cache_alloc. In this case, the management of the
662 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
663 * unless a particular usage is carefully commented. (the responsibility of
664 * freeing the kmalloc memory is the caller's, of course).
666 * A page may be used by anyone else who does a __get_free_page().
667 * In this case, page_count still tracks the references, and should only
668 * be used through the normal accessor functions. The top bits of page->flags
669 * and page->virtual store page management information, but all other fields
670 * are unused and could be used privately, carefully. The management of this
671 * page is the responsibility of the one who allocated it, and those who have
672 * subsequently been given references to it.
674 * The other pages (we may call them "pagecache pages") are completely
675 * managed by the Linux memory manager: I/O, buffers, swapping etc.
676 * The following discussion applies only to them.
678 * A pagecache page contains an opaque `private' member, which belongs to the
679 * page's address_space. Usually, this is the address of a circular list of
680 * the page's disk buffers. PG_private must be set to tell the VM to call
681 * into the filesystem to release these pages.
683 * A page may belong to an inode's memory mapping. In this case, page->mapping
684 * is the pointer to the inode, and page->index is the file offset of the page,
685 * in units of PAGE_SIZE.
687 * If pagecache pages are not associated with an inode, they are said to be
688 * anonymous pages. These may become associated with the swapcache, and in that
689 * case PG_swapcache is set, and page->private is an offset into the swapcache.
691 * In either case (swapcache or inode backed), the pagecache itself holds one
692 * reference to the page. Setting PG_private should also increment the
693 * refcount. The each user mapping also has a reference to the page.
695 * The pagecache pages are stored in a per-mapping radix tree, which is
696 * rooted at mapping->page_tree, and indexed by offset.
697 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
698 * lists, we instead now tag pages as dirty/writeback in the radix tree.
700 * All pagecache pages may be subject to I/O:
701 * - inode pages may need to be read from disk,
702 * - inode pages which have been modified and are MAP_SHARED may need
703 * to be written back to the inode on disk,
704 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
705 * modified may need to be swapped out to swap space and (later) to be read
710 * The zone field is never updated after free_area_init_core()
711 * sets it, so none of the operations on it need to be atomic.
714 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
715 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
716 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
717 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
718 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
721 * Define the bit shifts to access each section. For non-existent
722 * sections we define the shift as 0; that plus a 0 mask ensures
723 * the compiler will optimise away reference to them.
725 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
726 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
727 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
728 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
730 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
731 #ifdef NODE_NOT_IN_PAGE_FLAGS
732 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
733 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
734 SECTIONS_PGOFF : ZONES_PGOFF)
736 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
737 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
738 NODES_PGOFF : ZONES_PGOFF)
741 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
743 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
744 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
747 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
748 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
749 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
750 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
751 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
753 static inline enum zone_type
page_zonenum(const struct page
*page
)
755 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
758 #ifdef CONFIG_ZONE_DEVICE
759 void get_zone_device_page(struct page
*page
);
760 void put_zone_device_page(struct page
*page
);
761 static inline bool is_zone_device_page(const struct page
*page
)
763 return page_zonenum(page
) == ZONE_DEVICE
;
766 static inline void get_zone_device_page(struct page
*page
)
769 static inline void put_zone_device_page(struct page
*page
)
772 static inline bool is_zone_device_page(const struct page
*page
)
778 static inline void get_page(struct page
*page
)
780 page
= compound_head(page
);
782 * Getting a normal page or the head of a compound page
783 * requires to already have an elevated page->_refcount.
785 VM_BUG_ON_PAGE(page_ref_count(page
) <= 0, page
);
788 if (unlikely(is_zone_device_page(page
)))
789 get_zone_device_page(page
);
792 static inline void put_page(struct page
*page
)
794 page
= compound_head(page
);
796 if (put_page_testzero(page
))
799 if (unlikely(is_zone_device_page(page
)))
800 put_zone_device_page(page
);
803 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
804 #define SECTION_IN_PAGE_FLAGS
808 * The identification function is mainly used by the buddy allocator for
809 * determining if two pages could be buddies. We are not really identifying
810 * the zone since we could be using the section number id if we do not have
811 * node id available in page flags.
812 * We only guarantee that it will return the same value for two combinable
815 static inline int page_zone_id(struct page
*page
)
817 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
820 static inline int zone_to_nid(struct zone
*zone
)
829 #ifdef NODE_NOT_IN_PAGE_FLAGS
830 extern int page_to_nid(const struct page
*page
);
832 static inline int page_to_nid(const struct page
*page
)
834 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
838 #ifdef CONFIG_NUMA_BALANCING
839 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
841 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
844 static inline int cpupid_to_pid(int cpupid
)
846 return cpupid
& LAST__PID_MASK
;
849 static inline int cpupid_to_cpu(int cpupid
)
851 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
854 static inline int cpupid_to_nid(int cpupid
)
856 return cpu_to_node(cpupid_to_cpu(cpupid
));
859 static inline bool cpupid_pid_unset(int cpupid
)
861 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
864 static inline bool cpupid_cpu_unset(int cpupid
)
866 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
869 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
871 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
874 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
875 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
876 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
878 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
881 static inline int page_cpupid_last(struct page
*page
)
883 return page
->_last_cpupid
;
885 static inline void page_cpupid_reset_last(struct page
*page
)
887 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
890 static inline int page_cpupid_last(struct page
*page
)
892 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
895 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
897 static inline void page_cpupid_reset_last(struct page
*page
)
899 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
901 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
902 #else /* !CONFIG_NUMA_BALANCING */
903 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
905 return page_to_nid(page
); /* XXX */
908 static inline int page_cpupid_last(struct page
*page
)
910 return page_to_nid(page
); /* XXX */
913 static inline int cpupid_to_nid(int cpupid
)
918 static inline int cpupid_to_pid(int cpupid
)
923 static inline int cpupid_to_cpu(int cpupid
)
928 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
933 static inline bool cpupid_pid_unset(int cpupid
)
938 static inline void page_cpupid_reset_last(struct page
*page
)
942 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
946 #endif /* CONFIG_NUMA_BALANCING */
948 static inline struct zone
*page_zone(const struct page
*page
)
950 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
953 static inline pg_data_t
*page_pgdat(const struct page
*page
)
955 return NODE_DATA(page_to_nid(page
));
958 #ifdef SECTION_IN_PAGE_FLAGS
959 static inline void set_page_section(struct page
*page
, unsigned long section
)
961 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
962 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
965 static inline unsigned long page_to_section(const struct page
*page
)
967 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
971 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
973 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
974 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
977 static inline void set_page_node(struct page
*page
, unsigned long node
)
979 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
980 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
983 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
984 unsigned long node
, unsigned long pfn
)
986 set_page_zone(page
, zone
);
987 set_page_node(page
, node
);
988 #ifdef SECTION_IN_PAGE_FLAGS
989 set_page_section(page
, pfn_to_section_nr(pfn
));
994 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
996 return page
->mem_cgroup
;
998 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1000 WARN_ON_ONCE(!rcu_read_lock_held());
1001 return READ_ONCE(page
->mem_cgroup
);
1004 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
1008 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1010 WARN_ON_ONCE(!rcu_read_lock_held());
1016 * Some inline functions in vmstat.h depend on page_zone()
1018 #include <linux/vmstat.h>
1020 static __always_inline
void *lowmem_page_address(const struct page
*page
)
1022 return page_to_virt(page
);
1025 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1026 #define HASHED_PAGE_VIRTUAL
1029 #if defined(WANT_PAGE_VIRTUAL)
1030 static inline void *page_address(const struct page
*page
)
1032 return page
->virtual;
1034 static inline void set_page_address(struct page
*page
, void *address
)
1036 page
->virtual = address
;
1038 #define page_address_init() do { } while(0)
1041 #if defined(HASHED_PAGE_VIRTUAL)
1042 void *page_address(const struct page
*page
);
1043 void set_page_address(struct page
*page
, void *virtual);
1044 void page_address_init(void);
1047 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1048 #define page_address(page) lowmem_page_address(page)
1049 #define set_page_address(page, address) do { } while(0)
1050 #define page_address_init() do { } while(0)
1053 extern void *page_rmapping(struct page
*page
);
1054 extern struct anon_vma
*page_anon_vma(struct page
*page
);
1055 extern struct address_space
*page_mapping(struct page
*page
);
1057 extern struct address_space
*__page_file_mapping(struct page
*);
1060 struct address_space
*page_file_mapping(struct page
*page
)
1062 if (unlikely(PageSwapCache(page
)))
1063 return __page_file_mapping(page
);
1065 return page
->mapping
;
1068 extern pgoff_t
__page_file_index(struct page
*page
);
1071 * Return the pagecache index of the passed page. Regular pagecache pages
1072 * use ->index whereas swapcache pages use swp_offset(->private)
1074 static inline pgoff_t
page_index(struct page
*page
)
1076 if (unlikely(PageSwapCache(page
)))
1077 return __page_file_index(page
);
1081 bool page_mapped(struct page
*page
);
1082 struct address_space
*page_mapping(struct page
*page
);
1085 * Return true only if the page has been allocated with
1086 * ALLOC_NO_WATERMARKS and the low watermark was not
1087 * met implying that the system is under some pressure.
1089 static inline bool page_is_pfmemalloc(struct page
*page
)
1092 * Page index cannot be this large so this must be
1093 * a pfmemalloc page.
1095 return page
->index
== -1UL;
1099 * Only to be called by the page allocator on a freshly allocated
1102 static inline void set_page_pfmemalloc(struct page
*page
)
1107 static inline void clear_page_pfmemalloc(struct page
*page
)
1113 * Different kinds of faults, as returned by handle_mm_fault().
1114 * Used to decide whether a process gets delivered SIGBUS or
1115 * just gets major/minor fault counters bumped up.
1118 #define VM_FAULT_OOM 0x0001
1119 #define VM_FAULT_SIGBUS 0x0002
1120 #define VM_FAULT_MAJOR 0x0004
1121 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1122 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1123 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1124 #define VM_FAULT_SIGSEGV 0x0040
1126 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1127 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1128 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1129 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1130 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1132 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1134 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1135 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1138 #define VM_FAULT_RESULT_TRACE \
1139 { VM_FAULT_OOM, "OOM" }, \
1140 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1141 { VM_FAULT_MAJOR, "MAJOR" }, \
1142 { VM_FAULT_WRITE, "WRITE" }, \
1143 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1144 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1145 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1146 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1147 { VM_FAULT_LOCKED, "LOCKED" }, \
1148 { VM_FAULT_RETRY, "RETRY" }, \
1149 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1150 { VM_FAULT_DONE_COW, "DONE_COW" }
1152 /* Encode hstate index for a hwpoisoned large page */
1153 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1154 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1157 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1159 extern void pagefault_out_of_memory(void);
1161 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1164 * Flags passed to show_mem() and show_free_areas() to suppress output in
1167 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1169 extern void show_free_areas(unsigned int flags
, nodemask_t
*nodemask
);
1171 extern bool can_do_mlock(void);
1172 extern int user_shm_lock(size_t, struct user_struct
*);
1173 extern void user_shm_unlock(size_t, struct user_struct
*);
1176 * Parameter block passed down to zap_pte_range in exceptional cases.
1178 struct zap_details
{
1179 struct address_space
*check_mapping
; /* Check page->mapping if set */
1180 pgoff_t first_index
; /* Lowest page->index to unmap */
1181 pgoff_t last_index
; /* Highest page->index to unmap */
1184 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1186 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
1189 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1190 unsigned long size
);
1191 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1192 unsigned long size
);
1193 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1194 unsigned long start
, unsigned long end
);
1197 * mm_walk - callbacks for walk_page_range
1198 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1199 * this handler should only handle pud_trans_huge() puds.
1200 * the pmd_entry or pte_entry callbacks will be used for
1202 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1203 * this handler is required to be able to handle
1204 * pmd_trans_huge() pmds. They may simply choose to
1205 * split_huge_page() instead of handling it explicitly.
1206 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1207 * @pte_hole: if set, called for each hole at all levels
1208 * @hugetlb_entry: if set, called for each hugetlb entry
1209 * @test_walk: caller specific callback function to determine whether
1210 * we walk over the current vma or not. Returning 0
1211 * value means "do page table walk over the current vma,"
1212 * and a negative one means "abort current page table walk
1213 * right now." 1 means "skip the current vma."
1214 * @mm: mm_struct representing the target process of page table walk
1215 * @vma: vma currently walked (NULL if walking outside vmas)
1216 * @private: private data for callbacks' usage
1218 * (see the comment on walk_page_range() for more details)
1221 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
1222 unsigned long next
, struct mm_walk
*walk
);
1223 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1224 unsigned long next
, struct mm_walk
*walk
);
1225 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1226 unsigned long next
, struct mm_walk
*walk
);
1227 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1228 struct mm_walk
*walk
);
1229 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1230 unsigned long addr
, unsigned long next
,
1231 struct mm_walk
*walk
);
1232 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1233 struct mm_walk
*walk
);
1234 struct mm_struct
*mm
;
1235 struct vm_area_struct
*vma
;
1239 int walk_page_range(unsigned long addr
, unsigned long end
,
1240 struct mm_walk
*walk
);
1241 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1242 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1243 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1244 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1245 struct vm_area_struct
*vma
);
1246 void unmap_mapping_range(struct address_space
*mapping
,
1247 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1248 int follow_pte_pmd(struct mm_struct
*mm
, unsigned long address
,
1249 pte_t
**ptepp
, pmd_t
**pmdpp
, spinlock_t
**ptlp
);
1250 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1251 unsigned long *pfn
);
1252 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1253 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1254 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1255 void *buf
, int len
, int write
);
1257 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1258 loff_t
const holebegin
, loff_t
const holelen
)
1260 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1263 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1264 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1265 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1266 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1267 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1268 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1269 int invalidate_inode_page(struct page
*page
);
1272 extern int handle_mm_fault(struct vm_area_struct
*vma
, unsigned long address
,
1273 unsigned int flags
);
1274 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1275 unsigned long address
, unsigned int fault_flags
,
1278 static inline int handle_mm_fault(struct vm_area_struct
*vma
,
1279 unsigned long address
, unsigned int flags
)
1281 /* should never happen if there's no MMU */
1283 return VM_FAULT_SIGBUS
;
1285 static inline int fixup_user_fault(struct task_struct
*tsk
,
1286 struct mm_struct
*mm
, unsigned long address
,
1287 unsigned int fault_flags
, bool *unlocked
)
1289 /* should never happen if there's no MMU */
1295 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
,
1296 unsigned int gup_flags
);
1297 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1298 void *buf
, int len
, unsigned int gup_flags
);
1299 extern int __access_remote_vm(struct task_struct
*tsk
, struct mm_struct
*mm
,
1300 unsigned long addr
, void *buf
, int len
, unsigned int gup_flags
);
1302 long get_user_pages_remote(struct task_struct
*tsk
, struct mm_struct
*mm
,
1303 unsigned long start
, unsigned long nr_pages
,
1304 unsigned int gup_flags
, struct page
**pages
,
1305 struct vm_area_struct
**vmas
, int *locked
);
1306 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
1307 unsigned int gup_flags
, struct page
**pages
,
1308 struct vm_area_struct
**vmas
);
1309 long get_user_pages_locked(unsigned long start
, unsigned long nr_pages
,
1310 unsigned int gup_flags
, struct page
**pages
, int *locked
);
1311 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
1312 struct page
**pages
, unsigned int gup_flags
);
1313 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1314 struct page
**pages
);
1316 /* Container for pinned pfns / pages */
1317 struct frame_vector
{
1318 unsigned int nr_allocated
; /* Number of frames we have space for */
1319 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1320 bool got_ref
; /* Did we pin pages by getting page ref? */
1321 bool is_pfns
; /* Does array contain pages or pfns? */
1322 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1323 * pfns_vector_pages() or pfns_vector_pfns()
1327 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1328 void frame_vector_destroy(struct frame_vector
*vec
);
1329 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1330 unsigned int gup_flags
, struct frame_vector
*vec
);
1331 void put_vaddr_frames(struct frame_vector
*vec
);
1332 int frame_vector_to_pages(struct frame_vector
*vec
);
1333 void frame_vector_to_pfns(struct frame_vector
*vec
);
1335 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1337 return vec
->nr_frames
;
1340 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1343 int err
= frame_vector_to_pages(vec
);
1346 return ERR_PTR(err
);
1348 return (struct page
**)(vec
->ptrs
);
1351 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1354 frame_vector_to_pfns(vec
);
1355 return (unsigned long *)(vec
->ptrs
);
1359 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1360 struct page
**pages
);
1361 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1362 struct page
*get_dump_page(unsigned long addr
);
1364 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1365 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1366 unsigned int length
);
1368 int __set_page_dirty_nobuffers(struct page
*page
);
1369 int __set_page_dirty_no_writeback(struct page
*page
);
1370 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1372 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1373 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1374 struct bdi_writeback
*wb
);
1375 int set_page_dirty(struct page
*page
);
1376 int set_page_dirty_lock(struct page
*page
);
1377 void cancel_dirty_page(struct page
*page
);
1378 int clear_page_dirty_for_io(struct page
*page
);
1380 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1382 /* Is the vma a continuation of the stack vma above it? */
1383 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1385 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1388 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1390 return !vma
->vm_ops
;
1395 * The vma_is_shmem is not inline because it is used only by slow
1396 * paths in userfault.
1398 bool vma_is_shmem(struct vm_area_struct
*vma
);
1400 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
1403 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1406 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1407 (vma
->vm_start
== addr
) &&
1408 !vma_growsdown(vma
->vm_prev
, addr
);
1411 /* Is the vma a continuation of the stack vma below it? */
1412 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1414 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1417 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1420 return (vma
->vm_flags
& VM_GROWSUP
) &&
1421 (vma
->vm_end
== addr
) &&
1422 !vma_growsup(vma
->vm_next
, addr
);
1425 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
1427 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1428 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1429 unsigned long new_addr
, unsigned long len
,
1430 bool need_rmap_locks
);
1431 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1432 unsigned long end
, pgprot_t newprot
,
1433 int dirty_accountable
, int prot_numa
);
1434 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1435 struct vm_area_struct
**pprev
, unsigned long start
,
1436 unsigned long end
, unsigned long newflags
);
1439 * doesn't attempt to fault and will return short.
1441 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1442 struct page
**pages
);
1444 * per-process(per-mm_struct) statistics.
1446 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1448 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1450 #ifdef SPLIT_RSS_COUNTING
1452 * counter is updated in asynchronous manner and may go to minus.
1453 * But it's never be expected number for users.
1458 return (unsigned long)val
;
1461 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1463 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1466 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1468 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1471 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1473 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1476 /* Optimized variant when page is already known not to be PageAnon */
1477 static inline int mm_counter_file(struct page
*page
)
1479 if (PageSwapBacked(page
))
1480 return MM_SHMEMPAGES
;
1481 return MM_FILEPAGES
;
1484 static inline int mm_counter(struct page
*page
)
1487 return MM_ANONPAGES
;
1488 return mm_counter_file(page
);
1491 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1493 return get_mm_counter(mm
, MM_FILEPAGES
) +
1494 get_mm_counter(mm
, MM_ANONPAGES
) +
1495 get_mm_counter(mm
, MM_SHMEMPAGES
);
1498 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1500 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1503 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1505 return max(mm
->hiwater_vm
, mm
->total_vm
);
1508 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1510 unsigned long _rss
= get_mm_rss(mm
);
1512 if ((mm
)->hiwater_rss
< _rss
)
1513 (mm
)->hiwater_rss
= _rss
;
1516 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1518 if (mm
->hiwater_vm
< mm
->total_vm
)
1519 mm
->hiwater_vm
= mm
->total_vm
;
1522 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1524 mm
->hiwater_rss
= get_mm_rss(mm
);
1527 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1528 struct mm_struct
*mm
)
1530 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1532 if (*maxrss
< hiwater_rss
)
1533 *maxrss
= hiwater_rss
;
1536 #if defined(SPLIT_RSS_COUNTING)
1537 void sync_mm_rss(struct mm_struct
*mm
);
1539 static inline void sync_mm_rss(struct mm_struct
*mm
)
1544 #ifndef __HAVE_ARCH_PTE_DEVMAP
1545 static inline int pte_devmap(pte_t pte
)
1551 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
1553 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1555 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1559 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1563 #ifdef __PAGETABLE_P4D_FOLDED
1564 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1565 unsigned long address
)
1570 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1573 #ifdef __PAGETABLE_PUD_FOLDED
1574 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1575 unsigned long address
)
1580 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
1583 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1584 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1585 unsigned long address
)
1590 static inline void mm_nr_pmds_init(struct mm_struct
*mm
) {}
1592 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1597 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1598 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1601 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1603 static inline void mm_nr_pmds_init(struct mm_struct
*mm
)
1605 atomic_long_set(&mm
->nr_pmds
, 0);
1608 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1610 return atomic_long_read(&mm
->nr_pmds
);
1613 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1615 atomic_long_inc(&mm
->nr_pmds
);
1618 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1620 atomic_long_dec(&mm
->nr_pmds
);
1624 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
1625 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1628 * The following ifdef needed to get the 4level-fixup.h header to work.
1629 * Remove it when 4level-fixup.h has been removed.
1631 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1633 #ifndef __ARCH_HAS_5LEVEL_HACK
1634 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1635 unsigned long address
)
1637 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
1638 NULL
: p4d_offset(pgd
, address
);
1641 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1642 unsigned long address
)
1644 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
1645 NULL
: pud_offset(p4d
, address
);
1647 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1649 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1651 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1652 NULL
: pmd_offset(pud
, address
);
1654 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1656 #if USE_SPLIT_PTE_PTLOCKS
1657 #if ALLOC_SPLIT_PTLOCKS
1658 void __init
ptlock_cache_init(void);
1659 extern bool ptlock_alloc(struct page
*page
);
1660 extern void ptlock_free(struct page
*page
);
1662 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1666 #else /* ALLOC_SPLIT_PTLOCKS */
1667 static inline void ptlock_cache_init(void)
1671 static inline bool ptlock_alloc(struct page
*page
)
1676 static inline void ptlock_free(struct page
*page
)
1680 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1684 #endif /* ALLOC_SPLIT_PTLOCKS */
1686 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1688 return ptlock_ptr(pmd_page(*pmd
));
1691 static inline bool ptlock_init(struct page
*page
)
1694 * prep_new_page() initialize page->private (and therefore page->ptl)
1695 * with 0. Make sure nobody took it in use in between.
1697 * It can happen if arch try to use slab for page table allocation:
1698 * slab code uses page->slab_cache, which share storage with page->ptl.
1700 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1701 if (!ptlock_alloc(page
))
1703 spin_lock_init(ptlock_ptr(page
));
1707 /* Reset page->mapping so free_pages_check won't complain. */
1708 static inline void pte_lock_deinit(struct page
*page
)
1710 page
->mapping
= NULL
;
1714 #else /* !USE_SPLIT_PTE_PTLOCKS */
1716 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1718 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1720 return &mm
->page_table_lock
;
1722 static inline void ptlock_cache_init(void) {}
1723 static inline bool ptlock_init(struct page
*page
) { return true; }
1724 static inline void pte_lock_deinit(struct page
*page
) {}
1725 #endif /* USE_SPLIT_PTE_PTLOCKS */
1727 static inline void pgtable_init(void)
1729 ptlock_cache_init();
1730 pgtable_cache_init();
1733 static inline bool pgtable_page_ctor(struct page
*page
)
1735 if (!ptlock_init(page
))
1737 inc_zone_page_state(page
, NR_PAGETABLE
);
1741 static inline void pgtable_page_dtor(struct page
*page
)
1743 pte_lock_deinit(page
);
1744 dec_zone_page_state(page
, NR_PAGETABLE
);
1747 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1749 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1750 pte_t *__pte = pte_offset_map(pmd, address); \
1756 #define pte_unmap_unlock(pte, ptl) do { \
1761 #define pte_alloc(mm, pmd, address) \
1762 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1764 #define pte_alloc_map(mm, pmd, address) \
1765 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1767 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1768 (pte_alloc(mm, pmd, address) ? \
1769 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1771 #define pte_alloc_kernel(pmd, address) \
1772 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1773 NULL: pte_offset_kernel(pmd, address))
1775 #if USE_SPLIT_PMD_PTLOCKS
1777 static struct page
*pmd_to_page(pmd_t
*pmd
)
1779 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1780 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1783 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1785 return ptlock_ptr(pmd_to_page(pmd
));
1788 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1790 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1791 page
->pmd_huge_pte
= NULL
;
1793 return ptlock_init(page
);
1796 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1798 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1799 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1804 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1808 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1810 return &mm
->page_table_lock
;
1813 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1814 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1816 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1820 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1822 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1828 * No scalability reason to split PUD locks yet, but follow the same pattern
1829 * as the PMD locks to make it easier if we decide to. The VM should not be
1830 * considered ready to switch to split PUD locks yet; there may be places
1831 * which need to be converted from page_table_lock.
1833 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
1835 return &mm
->page_table_lock
;
1838 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
1840 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
1846 extern void __init
pagecache_init(void);
1847 extern void free_area_init(unsigned long * zones_size
);
1848 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1849 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1850 extern void free_initmem(void);
1853 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1854 * into the buddy system. The freed pages will be poisoned with pattern
1855 * "poison" if it's within range [0, UCHAR_MAX].
1856 * Return pages freed into the buddy system.
1858 extern unsigned long free_reserved_area(void *start
, void *end
,
1859 int poison
, char *s
);
1861 #ifdef CONFIG_HIGHMEM
1863 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1864 * and totalram_pages.
1866 extern void free_highmem_page(struct page
*page
);
1869 extern void adjust_managed_page_count(struct page
*page
, long count
);
1870 extern void mem_init_print_info(const char *str
);
1872 extern void reserve_bootmem_region(phys_addr_t start
, phys_addr_t end
);
1874 /* Free the reserved page into the buddy system, so it gets managed. */
1875 static inline void __free_reserved_page(struct page
*page
)
1877 ClearPageReserved(page
);
1878 init_page_count(page
);
1882 static inline void free_reserved_page(struct page
*page
)
1884 __free_reserved_page(page
);
1885 adjust_managed_page_count(page
, 1);
1888 static inline void mark_page_reserved(struct page
*page
)
1890 SetPageReserved(page
);
1891 adjust_managed_page_count(page
, -1);
1895 * Default method to free all the __init memory into the buddy system.
1896 * The freed pages will be poisoned with pattern "poison" if it's within
1897 * range [0, UCHAR_MAX].
1898 * Return pages freed into the buddy system.
1900 static inline unsigned long free_initmem_default(int poison
)
1902 extern char __init_begin
[], __init_end
[];
1904 return free_reserved_area(&__init_begin
, &__init_end
,
1905 poison
, "unused kernel");
1908 static inline unsigned long get_num_physpages(void)
1911 unsigned long phys_pages
= 0;
1913 for_each_online_node(nid
)
1914 phys_pages
+= node_present_pages(nid
);
1919 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1921 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1922 * zones, allocate the backing mem_map and account for memory holes in a more
1923 * architecture independent manner. This is a substitute for creating the
1924 * zone_sizes[] and zholes_size[] arrays and passing them to
1925 * free_area_init_node()
1927 * An architecture is expected to register range of page frames backed by
1928 * physical memory with memblock_add[_node]() before calling
1929 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1930 * usage, an architecture is expected to do something like
1932 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1934 * for_each_valid_physical_page_range()
1935 * memblock_add_node(base, size, nid)
1936 * free_area_init_nodes(max_zone_pfns);
1938 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1939 * registered physical page range. Similarly
1940 * sparse_memory_present_with_active_regions() calls memory_present() for
1941 * each range when SPARSEMEM is enabled.
1943 * See mm/page_alloc.c for more information on each function exposed by
1944 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1946 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1947 unsigned long node_map_pfn_alignment(void);
1948 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1949 unsigned long end_pfn
);
1950 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1951 unsigned long end_pfn
);
1952 extern void get_pfn_range_for_nid(unsigned int nid
,
1953 unsigned long *start_pfn
, unsigned long *end_pfn
);
1954 extern unsigned long find_min_pfn_with_active_regions(void);
1955 extern void free_bootmem_with_active_regions(int nid
,
1956 unsigned long max_low_pfn
);
1957 extern void sparse_memory_present_with_active_regions(int nid
);
1959 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1961 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1962 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1963 static inline int __early_pfn_to_nid(unsigned long pfn
,
1964 struct mminit_pfnnid_cache
*state
)
1969 /* please see mm/page_alloc.c */
1970 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1971 /* there is a per-arch backend function. */
1972 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
1973 struct mminit_pfnnid_cache
*state
);
1976 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1977 extern void memmap_init_zone(unsigned long, int, unsigned long,
1978 unsigned long, enum memmap_context
);
1979 extern void setup_per_zone_wmarks(void);
1980 extern int __meminit
init_per_zone_wmark_min(void);
1981 extern void mem_init(void);
1982 extern void __init
mmap_init(void);
1983 extern void show_mem(unsigned int flags
, nodemask_t
*nodemask
);
1984 extern long si_mem_available(void);
1985 extern void si_meminfo(struct sysinfo
* val
);
1986 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1987 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
1988 extern unsigned long arch_reserved_kernel_pages(void);
1991 extern __printf(3, 4)
1992 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
1994 extern void setup_per_cpu_pageset(void);
1996 extern void zone_pcp_update(struct zone
*zone
);
1997 extern void zone_pcp_reset(struct zone
*zone
);
2000 extern int min_free_kbytes
;
2001 extern int watermark_scale_factor
;
2004 extern atomic_long_t mmap_pages_allocated
;
2005 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
2007 /* interval_tree.c */
2008 void vma_interval_tree_insert(struct vm_area_struct
*node
,
2009 struct rb_root
*root
);
2010 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
2011 struct vm_area_struct
*prev
,
2012 struct rb_root
*root
);
2013 void vma_interval_tree_remove(struct vm_area_struct
*node
,
2014 struct rb_root
*root
);
2015 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
2016 unsigned long start
, unsigned long last
);
2017 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
2018 unsigned long start
, unsigned long last
);
2020 #define vma_interval_tree_foreach(vma, root, start, last) \
2021 for (vma = vma_interval_tree_iter_first(root, start, last); \
2022 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2024 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
2025 struct rb_root
*root
);
2026 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
2027 struct rb_root
*root
);
2028 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
2029 struct rb_root
*root
, unsigned long start
, unsigned long last
);
2030 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
2031 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
2032 #ifdef CONFIG_DEBUG_VM_RB
2033 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
2036 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2037 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2038 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2041 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
2042 extern int __vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2043 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
,
2044 struct vm_area_struct
*expand
);
2045 static inline int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2046 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
)
2048 return __vma_adjust(vma
, start
, end
, pgoff
, insert
, NULL
);
2050 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
2051 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
2052 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
2053 struct mempolicy
*, struct vm_userfaultfd_ctx
);
2054 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
2055 extern int __split_vma(struct mm_struct
*, struct vm_area_struct
*,
2056 unsigned long addr
, int new_below
);
2057 extern int split_vma(struct mm_struct
*, struct vm_area_struct
*,
2058 unsigned long addr
, int new_below
);
2059 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
2060 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
2061 struct rb_node
**, struct rb_node
*);
2062 extern void unlink_file_vma(struct vm_area_struct
*);
2063 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
2064 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
2065 bool *need_rmap_locks
);
2066 extern void exit_mmap(struct mm_struct
*);
2068 static inline int check_data_rlimit(unsigned long rlim
,
2070 unsigned long start
,
2071 unsigned long end_data
,
2072 unsigned long start_data
)
2074 if (rlim
< RLIM_INFINITY
) {
2075 if (((new - start
) + (end_data
- start_data
)) > rlim
)
2082 extern int mm_take_all_locks(struct mm_struct
*mm
);
2083 extern void mm_drop_all_locks(struct mm_struct
*mm
);
2085 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2086 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2087 extern struct file
*get_task_exe_file(struct task_struct
*task
);
2089 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2090 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2092 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
2093 const struct vm_special_mapping
*sm
);
2094 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2095 unsigned long addr
, unsigned long len
,
2096 unsigned long flags
,
2097 const struct vm_special_mapping
*spec
);
2098 /* This is an obsolete alternative to _install_special_mapping. */
2099 extern int install_special_mapping(struct mm_struct
*mm
,
2100 unsigned long addr
, unsigned long len
,
2101 unsigned long flags
, struct page
**pages
);
2103 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
2105 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
2106 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
2107 struct list_head
*uf
);
2108 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
2109 unsigned long len
, unsigned long prot
, unsigned long flags
,
2110 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
,
2111 struct list_head
*uf
);
2112 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
2113 struct list_head
*uf
);
2115 static inline unsigned long
2116 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
2117 unsigned long len
, unsigned long prot
, unsigned long flags
,
2118 unsigned long pgoff
, unsigned long *populate
,
2119 struct list_head
*uf
)
2121 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
, uf
);
2125 extern int __mm_populate(unsigned long addr
, unsigned long len
,
2127 static inline void mm_populate(unsigned long addr
, unsigned long len
)
2130 (void) __mm_populate(addr
, len
, 1);
2133 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
2136 /* These take the mm semaphore themselves */
2137 extern int __must_check
vm_brk(unsigned long, unsigned long);
2138 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
2139 extern int vm_munmap(unsigned long, size_t);
2140 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
2141 unsigned long, unsigned long,
2142 unsigned long, unsigned long);
2144 struct vm_unmapped_area_info
{
2145 #define VM_UNMAPPED_AREA_TOPDOWN 1
2146 unsigned long flags
;
2147 unsigned long length
;
2148 unsigned long low_limit
;
2149 unsigned long high_limit
;
2150 unsigned long align_mask
;
2151 unsigned long align_offset
;
2154 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
2155 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
2158 * Search for an unmapped address range.
2160 * We are looking for a range that:
2161 * - does not intersect with any VMA;
2162 * - is contained within the [low_limit, high_limit) interval;
2163 * - is at least the desired size.
2164 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2166 static inline unsigned long
2167 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
2169 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
2170 return unmapped_area_topdown(info
);
2172 return unmapped_area(info
);
2176 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2177 extern void truncate_inode_pages_range(struct address_space
*,
2178 loff_t lstart
, loff_t lend
);
2179 extern void truncate_inode_pages_final(struct address_space
*);
2181 /* generic vm_area_ops exported for stackable file systems */
2182 extern int filemap_fault(struct vm_fault
*vmf
);
2183 extern void filemap_map_pages(struct vm_fault
*vmf
,
2184 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
2185 extern int filemap_page_mkwrite(struct vm_fault
*vmf
);
2187 /* mm/page-writeback.c */
2188 int write_one_page(struct page
*page
, int wait
);
2189 void task_dirty_inc(struct task_struct
*tsk
);
2192 #define VM_MAX_READAHEAD 128 /* kbytes */
2193 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2195 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2196 pgoff_t offset
, unsigned long nr_to_read
);
2198 void page_cache_sync_readahead(struct address_space
*mapping
,
2199 struct file_ra_state
*ra
,
2202 unsigned long size
);
2204 void page_cache_async_readahead(struct address_space
*mapping
,
2205 struct file_ra_state
*ra
,
2209 unsigned long size
);
2211 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2212 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2214 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2215 extern int expand_downwards(struct vm_area_struct
*vma
,
2216 unsigned long address
);
2218 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2220 #define expand_upwards(vma, address) (0)
2223 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2224 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2225 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2226 struct vm_area_struct
**pprev
);
2228 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2229 NULL if none. Assume start_addr < end_addr. */
2230 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2232 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2234 if (vma
&& end_addr
<= vma
->vm_start
)
2239 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2241 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2244 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2245 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2246 unsigned long vm_start
, unsigned long vm_end
)
2248 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2250 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2257 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2258 void vma_set_page_prot(struct vm_area_struct
*vma
);
2260 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2264 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2266 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2270 #ifdef CONFIG_NUMA_BALANCING
2271 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2272 unsigned long start
, unsigned long end
);
2275 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2276 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2277 unsigned long pfn
, unsigned long size
, pgprot_t
);
2278 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2279 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2281 int vm_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2282 unsigned long pfn
, pgprot_t pgprot
);
2283 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2285 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2288 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2289 unsigned long address
, unsigned int foll_flags
,
2290 unsigned int *page_mask
);
2292 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2293 unsigned long address
, unsigned int foll_flags
)
2295 unsigned int unused_page_mask
;
2296 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2299 #define FOLL_WRITE 0x01 /* check pte is writable */
2300 #define FOLL_TOUCH 0x02 /* mark page accessed */
2301 #define FOLL_GET 0x04 /* do get_page on page */
2302 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2303 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2304 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2305 * and return without waiting upon it */
2306 #define FOLL_POPULATE 0x40 /* fault in page */
2307 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2308 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2309 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2310 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2311 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2312 #define FOLL_MLOCK 0x1000 /* lock present pages */
2313 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2314 #define FOLL_COW 0x4000 /* internal GUP flag */
2316 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2318 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2319 unsigned long size
, pte_fn_t fn
, void *data
);
2322 #ifdef CONFIG_PAGE_POISONING
2323 extern bool page_poisoning_enabled(void);
2324 extern void kernel_poison_pages(struct page
*page
, int numpages
, int enable
);
2325 extern bool page_is_poisoned(struct page
*page
);
2327 static inline bool page_poisoning_enabled(void) { return false; }
2328 static inline void kernel_poison_pages(struct page
*page
, int numpages
,
2330 static inline bool page_is_poisoned(struct page
*page
) { return false; }
2333 #ifdef CONFIG_DEBUG_PAGEALLOC
2334 extern bool _debug_pagealloc_enabled
;
2335 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2337 static inline bool debug_pagealloc_enabled(void)
2339 return _debug_pagealloc_enabled
;
2343 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2345 if (!debug_pagealloc_enabled())
2348 __kernel_map_pages(page
, numpages
, enable
);
2350 #ifdef CONFIG_HIBERNATION
2351 extern bool kernel_page_present(struct page
*page
);
2352 #endif /* CONFIG_HIBERNATION */
2353 #else /* CONFIG_DEBUG_PAGEALLOC */
2355 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2356 #ifdef CONFIG_HIBERNATION
2357 static inline bool kernel_page_present(struct page
*page
) { return true; }
2358 #endif /* CONFIG_HIBERNATION */
2359 static inline bool debug_pagealloc_enabled(void)
2363 #endif /* CONFIG_DEBUG_PAGEALLOC */
2365 #ifdef __HAVE_ARCH_GATE_AREA
2366 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2367 extern int in_gate_area_no_mm(unsigned long addr
);
2368 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2370 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2374 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2375 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2379 #endif /* __HAVE_ARCH_GATE_AREA */
2381 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
2383 #ifdef CONFIG_SYSCTL
2384 extern int sysctl_drop_caches
;
2385 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2386 void __user
*, size_t *, loff_t
*);
2389 void drop_slab(void);
2390 void drop_slab_node(int nid
);
2393 #define randomize_va_space 0
2395 extern int randomize_va_space
;
2398 const char * arch_vma_name(struct vm_area_struct
*vma
);
2399 void print_vma_addr(char *prefix
, unsigned long rip
);
2401 void sparse_mem_maps_populate_node(struct page
**map_map
,
2402 unsigned long pnum_begin
,
2403 unsigned long pnum_end
,
2404 unsigned long map_count
,
2407 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2408 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2409 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2410 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
2411 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2412 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2413 void *vmemmap_alloc_block(unsigned long size
, int node
);
2415 void *__vmemmap_alloc_block_buf(unsigned long size
, int node
,
2416 struct vmem_altmap
*altmap
);
2417 static inline void *vmemmap_alloc_block_buf(unsigned long size
, int node
)
2419 return __vmemmap_alloc_block_buf(size
, node
, NULL
);
2422 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2423 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2425 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2426 void vmemmap_populate_print_last(void);
2427 #ifdef CONFIG_MEMORY_HOTPLUG
2428 void vmemmap_free(unsigned long start
, unsigned long end
);
2430 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2431 unsigned long size
);
2434 MF_COUNT_INCREASED
= 1 << 0,
2435 MF_ACTION_REQUIRED
= 1 << 1,
2436 MF_MUST_KILL
= 1 << 2,
2437 MF_SOFT_OFFLINE
= 1 << 3,
2439 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2440 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2441 extern int unpoison_memory(unsigned long pfn
);
2442 extern int get_hwpoison_page(struct page
*page
);
2443 #define put_hwpoison_page(page) put_page(page)
2444 extern int sysctl_memory_failure_early_kill
;
2445 extern int sysctl_memory_failure_recovery
;
2446 extern void shake_page(struct page
*p
, int access
);
2447 extern atomic_long_t num_poisoned_pages
;
2448 extern int soft_offline_page(struct page
*page
, int flags
);
2452 * Error handlers for various types of pages.
2455 MF_IGNORED
, /* Error: cannot be handled */
2456 MF_FAILED
, /* Error: handling failed */
2457 MF_DELAYED
, /* Will be handled later */
2458 MF_RECOVERED
, /* Successfully recovered */
2461 enum mf_action_page_type
{
2463 MF_MSG_KERNEL_HIGH_ORDER
,
2465 MF_MSG_DIFFERENT_COMPOUND
,
2466 MF_MSG_POISONED_HUGE
,
2469 MF_MSG_UNMAP_FAILED
,
2470 MF_MSG_DIRTY_SWAPCACHE
,
2471 MF_MSG_CLEAN_SWAPCACHE
,
2472 MF_MSG_DIRTY_MLOCKED_LRU
,
2473 MF_MSG_CLEAN_MLOCKED_LRU
,
2474 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2475 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2478 MF_MSG_TRUNCATED_LRU
,
2484 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2485 extern void clear_huge_page(struct page
*page
,
2487 unsigned int pages_per_huge_page
);
2488 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2489 unsigned long addr
, struct vm_area_struct
*vma
,
2490 unsigned int pages_per_huge_page
);
2491 extern long copy_huge_page_from_user(struct page
*dst_page
,
2492 const void __user
*usr_src
,
2493 unsigned int pages_per_huge_page
,
2494 bool allow_pagefault
);
2495 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2497 extern struct page_ext_operations debug_guardpage_ops
;
2498 extern struct page_ext_operations page_poisoning_ops
;
2500 #ifdef CONFIG_DEBUG_PAGEALLOC
2501 extern unsigned int _debug_guardpage_minorder
;
2502 extern bool _debug_guardpage_enabled
;
2504 static inline unsigned int debug_guardpage_minorder(void)
2506 return _debug_guardpage_minorder
;
2509 static inline bool debug_guardpage_enabled(void)
2511 return _debug_guardpage_enabled
;
2514 static inline bool page_is_guard(struct page
*page
)
2516 struct page_ext
*page_ext
;
2518 if (!debug_guardpage_enabled())
2521 page_ext
= lookup_page_ext(page
);
2522 if (unlikely(!page_ext
))
2525 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2528 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2529 static inline bool debug_guardpage_enabled(void) { return false; }
2530 static inline bool page_is_guard(struct page
*page
) { return false; }
2531 #endif /* CONFIG_DEBUG_PAGEALLOC */
2533 #if MAX_NUMNODES > 1
2534 void __init
setup_nr_node_ids(void);
2536 static inline void setup_nr_node_ids(void) {}
2539 #endif /* __KERNEL__ */
2540 #endif /* _LINUX_MM_H */