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 void init_mm_internals(void);
37 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
38 extern unsigned long max_mapnr
;
40 static inline void set_max_mapnr(unsigned long limit
)
45 static inline void set_max_mapnr(unsigned long limit
) { }
48 extern unsigned long totalram_pages
;
49 extern void * high_memory
;
50 extern int page_cluster
;
53 extern int sysctl_legacy_va_layout
;
55 #define sysctl_legacy_va_layout 0
58 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
59 extern const int mmap_rnd_bits_min
;
60 extern const int mmap_rnd_bits_max
;
61 extern int mmap_rnd_bits __read_mostly
;
63 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
64 extern const int mmap_rnd_compat_bits_min
;
65 extern const int mmap_rnd_compat_bits_max
;
66 extern int mmap_rnd_compat_bits __read_mostly
;
70 #include <asm/pgtable.h>
71 #include <asm/processor.h>
74 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
78 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
82 #define lm_alias(x) __va(__pa_symbol(x))
86 * To prevent common memory management code establishing
87 * a zero page mapping on a read fault.
88 * This macro should be defined within <asm/pgtable.h>.
89 * s390 does this to prevent multiplexing of hardware bits
90 * related to the physical page in case of virtualization.
92 #ifndef mm_forbids_zeropage
93 #define mm_forbids_zeropage(X) (0)
97 * Default maximum number of active map areas, this limits the number of vmas
98 * per mm struct. Users can overwrite this number by sysctl but there is a
101 * When a program's coredump is generated as ELF format, a section is created
102 * per a vma. In ELF, the number of sections is represented in unsigned short.
103 * This means the number of sections should be smaller than 65535 at coredump.
104 * Because the kernel adds some informative sections to a image of program at
105 * generating coredump, we need some margin. The number of extra sections is
106 * 1-3 now and depends on arch. We use "5" as safe margin, here.
108 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
109 * not a hard limit any more. Although some userspace tools can be surprised by
112 #define MAPCOUNT_ELF_CORE_MARGIN (5)
113 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
115 extern int sysctl_max_map_count
;
117 extern unsigned long sysctl_user_reserve_kbytes
;
118 extern unsigned long sysctl_admin_reserve_kbytes
;
120 extern int sysctl_overcommit_memory
;
121 extern int sysctl_overcommit_ratio
;
122 extern unsigned long sysctl_overcommit_kbytes
;
124 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
126 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
129 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
131 /* to align the pointer to the (next) page boundary */
132 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
134 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
135 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
138 * Linux kernel virtual memory manager primitives.
139 * The idea being to have a "virtual" mm in the same way
140 * we have a virtual fs - giving a cleaner interface to the
141 * mm details, and allowing different kinds of memory mappings
142 * (from shared memory to executable loading to arbitrary
146 extern struct kmem_cache
*vm_area_cachep
;
149 extern struct rb_root nommu_region_tree
;
150 extern struct rw_semaphore nommu_region_sem
;
152 extern unsigned int kobjsize(const void *objp
);
156 * vm_flags in vm_area_struct, see mm_types.h.
157 * When changing, update also include/trace/events/mmflags.h
159 #define VM_NONE 0x00000000
161 #define VM_READ 0x00000001 /* currently active flags */
162 #define VM_WRITE 0x00000002
163 #define VM_EXEC 0x00000004
164 #define VM_SHARED 0x00000008
166 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
167 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
168 #define VM_MAYWRITE 0x00000020
169 #define VM_MAYEXEC 0x00000040
170 #define VM_MAYSHARE 0x00000080
172 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
173 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
174 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
175 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
176 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
178 #define VM_LOCKED 0x00002000
179 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
181 /* Used by sys_madvise() */
182 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
183 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
185 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
186 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
187 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
188 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
189 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
190 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
191 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
192 #define VM_ARCH_2 0x02000000
193 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
195 #ifdef CONFIG_MEM_SOFT_DIRTY
196 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
198 # define VM_SOFTDIRTY 0
201 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
202 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
203 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
204 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
206 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
207 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
208 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
209 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
210 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
211 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
212 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
213 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
214 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
215 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
217 #if defined(CONFIG_X86)
218 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
219 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
220 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
221 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
222 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
223 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
224 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
226 #elif defined(CONFIG_PPC)
227 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
228 #elif defined(CONFIG_PARISC)
229 # define VM_GROWSUP VM_ARCH_1
230 #elif defined(CONFIG_METAG)
231 # define VM_GROWSUP VM_ARCH_1
232 #elif defined(CONFIG_IA64)
233 # define VM_GROWSUP VM_ARCH_1
234 #elif !defined(CONFIG_MMU)
235 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
238 #if defined(CONFIG_X86)
239 /* MPX specific bounds table or bounds directory */
240 # define VM_MPX VM_ARCH_2
244 # define VM_GROWSUP VM_NONE
247 /* Bits set in the VMA until the stack is in its final location */
248 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
250 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
251 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
254 #ifdef CONFIG_STACK_GROWSUP
255 #define VM_STACK VM_GROWSUP
257 #define VM_STACK VM_GROWSDOWN
260 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
263 * Special vmas that are non-mergable, non-mlock()able.
264 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
266 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
268 /* This mask defines which mm->def_flags a process can inherit its parent */
269 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
271 /* This mask is used to clear all the VMA flags used by mlock */
272 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
275 * mapping from the currently active vm_flags protection bits (the
276 * low four bits) to a page protection mask..
278 extern pgprot_t protection_map
[16];
280 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
281 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
282 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
283 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
284 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
285 #define FAULT_FLAG_TRIED 0x20 /* Second try */
286 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
287 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
288 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
290 #define FAULT_FLAG_TRACE \
291 { FAULT_FLAG_WRITE, "WRITE" }, \
292 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
293 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
294 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
295 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
296 { FAULT_FLAG_TRIED, "TRIED" }, \
297 { FAULT_FLAG_USER, "USER" }, \
298 { FAULT_FLAG_REMOTE, "REMOTE" }, \
299 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
302 * vm_fault is filled by the the pagefault handler and passed to the vma's
303 * ->fault function. The vma's ->fault is responsible for returning a bitmask
304 * of VM_FAULT_xxx flags that give details about how the fault was handled.
306 * MM layer fills up gfp_mask for page allocations but fault handler might
307 * alter it if its implementation requires a different allocation context.
309 * pgoff should be used in favour of virtual_address, if possible.
312 struct vm_area_struct
*vma
; /* Target VMA */
313 unsigned int flags
; /* FAULT_FLAG_xxx flags */
314 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
315 pgoff_t pgoff
; /* Logical page offset based on vma */
316 unsigned long address
; /* Faulting virtual address */
317 pmd_t
*pmd
; /* Pointer to pmd entry matching
319 pud_t
*pud
; /* Pointer to pud entry matching
322 pte_t orig_pte
; /* Value of PTE at the time of fault */
324 struct page
*cow_page
; /* Page handler may use for COW fault */
325 struct mem_cgroup
*memcg
; /* Cgroup cow_page belongs to */
326 struct page
*page
; /* ->fault handlers should return a
327 * page here, unless VM_FAULT_NOPAGE
328 * is set (which is also implied by
331 /* These three entries are valid only while holding ptl lock */
332 pte_t
*pte
; /* Pointer to pte entry matching
333 * the 'address'. NULL if the page
334 * table hasn't been allocated.
336 spinlock_t
*ptl
; /* Page table lock.
337 * Protects pte page table if 'pte'
338 * is not NULL, otherwise pmd.
340 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
341 * vm_ops->map_pages() calls
342 * alloc_set_pte() from atomic context.
343 * do_fault_around() pre-allocates
344 * page table to avoid allocation from
349 /* page entry size for vm->huge_fault() */
350 enum page_entry_size
{
357 * These are the virtual MM functions - opening of an area, closing and
358 * unmapping it (needed to keep files on disk up-to-date etc), pointer
359 * to the functions called when a no-page or a wp-page exception occurs.
361 struct vm_operations_struct
{
362 void (*open
)(struct vm_area_struct
* area
);
363 void (*close
)(struct vm_area_struct
* area
);
364 int (*mremap
)(struct vm_area_struct
* area
);
365 int (*fault
)(struct vm_fault
*vmf
);
366 int (*huge_fault
)(struct vm_fault
*vmf
, enum page_entry_size pe_size
);
367 void (*map_pages
)(struct vm_fault
*vmf
,
368 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
370 /* notification that a previously read-only page is about to become
371 * writable, if an error is returned it will cause a SIGBUS */
372 int (*page_mkwrite
)(struct vm_fault
*vmf
);
374 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
375 int (*pfn_mkwrite
)(struct vm_fault
*vmf
);
377 /* called by access_process_vm when get_user_pages() fails, typically
378 * for use by special VMAs that can switch between memory and hardware
380 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
381 void *buf
, int len
, int write
);
383 /* Called by the /proc/PID/maps code to ask the vma whether it
384 * has a special name. Returning non-NULL will also cause this
385 * vma to be dumped unconditionally. */
386 const char *(*name
)(struct vm_area_struct
*vma
);
390 * set_policy() op must add a reference to any non-NULL @new mempolicy
391 * to hold the policy upon return. Caller should pass NULL @new to
392 * remove a policy and fall back to surrounding context--i.e. do not
393 * install a MPOL_DEFAULT policy, nor the task or system default
396 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
399 * get_policy() op must add reference [mpol_get()] to any policy at
400 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
401 * in mm/mempolicy.c will do this automatically.
402 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
403 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
404 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
405 * must return NULL--i.e., do not "fallback" to task or system default
408 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
412 * Called by vm_normal_page() for special PTEs to find the
413 * page for @addr. This is useful if the default behavior
414 * (using pte_page()) would not find the correct page.
416 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
423 #define page_private(page) ((page)->private)
424 #define set_page_private(page, v) ((page)->private = (v))
426 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
427 static inline int pmd_devmap(pmd_t pmd
)
431 static inline int pud_devmap(pud_t pud
)
435 static inline int pgd_devmap(pgd_t pgd
)
442 * FIXME: take this include out, include page-flags.h in
443 * files which need it (119 of them)
445 #include <linux/page-flags.h>
446 #include <linux/huge_mm.h>
449 * Methods to modify the page usage count.
451 * What counts for a page usage:
452 * - cache mapping (page->mapping)
453 * - private data (page->private)
454 * - page mapped in a task's page tables, each mapping
455 * is counted separately
457 * Also, many kernel routines increase the page count before a critical
458 * routine so they can be sure the page doesn't go away from under them.
462 * Drop a ref, return true if the refcount fell to zero (the page has no users)
464 static inline int put_page_testzero(struct page
*page
)
466 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
467 return page_ref_dec_and_test(page
);
471 * Try to grab a ref unless the page has a refcount of zero, return false if
473 * This can be called when MMU is off so it must not access
474 * any of the virtual mappings.
476 static inline int get_page_unless_zero(struct page
*page
)
478 return page_ref_add_unless(page
, 1, 0);
481 extern int page_is_ram(unsigned long pfn
);
489 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
492 /* Support for virtually mapped pages */
493 struct page
*vmalloc_to_page(const void *addr
);
494 unsigned long vmalloc_to_pfn(const void *addr
);
497 * Determine if an address is within the vmalloc range
499 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
500 * is no special casing required.
502 static inline bool is_vmalloc_addr(const void *x
)
505 unsigned long addr
= (unsigned long)x
;
507 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
513 extern int is_vmalloc_or_module_addr(const void *x
);
515 static inline int is_vmalloc_or_module_addr(const void *x
)
521 extern void kvfree(const void *addr
);
523 static inline atomic_t
*compound_mapcount_ptr(struct page
*page
)
525 return &page
[1].compound_mapcount
;
528 static inline int compound_mapcount(struct page
*page
)
530 VM_BUG_ON_PAGE(!PageCompound(page
), page
);
531 page
= compound_head(page
);
532 return atomic_read(compound_mapcount_ptr(page
)) + 1;
536 * The atomic page->_mapcount, starts from -1: so that transitions
537 * both from it and to it can be tracked, using atomic_inc_and_test
538 * and atomic_add_negative(-1).
540 static inline void page_mapcount_reset(struct page
*page
)
542 atomic_set(&(page
)->_mapcount
, -1);
545 int __page_mapcount(struct page
*page
);
547 static inline int page_mapcount(struct page
*page
)
549 VM_BUG_ON_PAGE(PageSlab(page
), page
);
551 if (unlikely(PageCompound(page
)))
552 return __page_mapcount(page
);
553 return atomic_read(&page
->_mapcount
) + 1;
556 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
557 int total_mapcount(struct page
*page
);
558 int page_trans_huge_mapcount(struct page
*page
, int *total_mapcount
);
560 static inline int total_mapcount(struct page
*page
)
562 return page_mapcount(page
);
564 static inline int page_trans_huge_mapcount(struct page
*page
,
567 int mapcount
= page_mapcount(page
);
569 *total_mapcount
= mapcount
;
574 static inline struct page
*virt_to_head_page(const void *x
)
576 struct page
*page
= virt_to_page(x
);
578 return compound_head(page
);
581 void __put_page(struct page
*page
);
583 void put_pages_list(struct list_head
*pages
);
585 void split_page(struct page
*page
, unsigned int order
);
588 * Compound pages have a destructor function. Provide a
589 * prototype for that function and accessor functions.
590 * These are _only_ valid on the head of a compound page.
592 typedef void compound_page_dtor(struct page
*);
594 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
595 enum compound_dtor_id
{
598 #ifdef CONFIG_HUGETLB_PAGE
601 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
606 extern compound_page_dtor
* const compound_page_dtors
[];
608 static inline void set_compound_page_dtor(struct page
*page
,
609 enum compound_dtor_id compound_dtor
)
611 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
612 page
[1].compound_dtor
= compound_dtor
;
615 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
617 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
618 return compound_page_dtors
[page
[1].compound_dtor
];
621 static inline unsigned int compound_order(struct page
*page
)
625 return page
[1].compound_order
;
628 static inline void set_compound_order(struct page
*page
, unsigned int order
)
630 page
[1].compound_order
= order
;
633 void free_compound_page(struct page
*page
);
637 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
638 * servicing faults for write access. In the normal case, do always want
639 * pte_mkwrite. But get_user_pages can cause write faults for mappings
640 * that do not have writing enabled, when used by access_process_vm.
642 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
644 if (likely(vma
->vm_flags
& VM_WRITE
))
645 pte
= pte_mkwrite(pte
);
649 int alloc_set_pte(struct vm_fault
*vmf
, struct mem_cgroup
*memcg
,
651 int finish_fault(struct vm_fault
*vmf
);
652 int finish_mkwrite_fault(struct vm_fault
*vmf
);
656 * Multiple processes may "see" the same page. E.g. for untouched
657 * mappings of /dev/null, all processes see the same page full of
658 * zeroes, and text pages of executables and shared libraries have
659 * only one copy in memory, at most, normally.
661 * For the non-reserved pages, page_count(page) denotes a reference count.
662 * page_count() == 0 means the page is free. page->lru is then used for
663 * freelist management in the buddy allocator.
664 * page_count() > 0 means the page has been allocated.
666 * Pages are allocated by the slab allocator in order to provide memory
667 * to kmalloc and kmem_cache_alloc. In this case, the management of the
668 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
669 * unless a particular usage is carefully commented. (the responsibility of
670 * freeing the kmalloc memory is the caller's, of course).
672 * A page may be used by anyone else who does a __get_free_page().
673 * In this case, page_count still tracks the references, and should only
674 * be used through the normal accessor functions. The top bits of page->flags
675 * and page->virtual store page management information, but all other fields
676 * are unused and could be used privately, carefully. The management of this
677 * page is the responsibility of the one who allocated it, and those who have
678 * subsequently been given references to it.
680 * The other pages (we may call them "pagecache pages") are completely
681 * managed by the Linux memory manager: I/O, buffers, swapping etc.
682 * The following discussion applies only to them.
684 * A pagecache page contains an opaque `private' member, which belongs to the
685 * page's address_space. Usually, this is the address of a circular list of
686 * the page's disk buffers. PG_private must be set to tell the VM to call
687 * into the filesystem to release these pages.
689 * A page may belong to an inode's memory mapping. In this case, page->mapping
690 * is the pointer to the inode, and page->index is the file offset of the page,
691 * in units of PAGE_SIZE.
693 * If pagecache pages are not associated with an inode, they are said to be
694 * anonymous pages. These may become associated with the swapcache, and in that
695 * case PG_swapcache is set, and page->private is an offset into the swapcache.
697 * In either case (swapcache or inode backed), the pagecache itself holds one
698 * reference to the page. Setting PG_private should also increment the
699 * refcount. The each user mapping also has a reference to the page.
701 * The pagecache pages are stored in a per-mapping radix tree, which is
702 * rooted at mapping->page_tree, and indexed by offset.
703 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
704 * lists, we instead now tag pages as dirty/writeback in the radix tree.
706 * All pagecache pages may be subject to I/O:
707 * - inode pages may need to be read from disk,
708 * - inode pages which have been modified and are MAP_SHARED may need
709 * to be written back to the inode on disk,
710 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
711 * modified may need to be swapped out to swap space and (later) to be read
716 * The zone field is never updated after free_area_init_core()
717 * sets it, so none of the operations on it need to be atomic.
720 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
721 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
722 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
723 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
724 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
727 * Define the bit shifts to access each section. For non-existent
728 * sections we define the shift as 0; that plus a 0 mask ensures
729 * the compiler will optimise away reference to them.
731 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
732 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
733 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
734 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
736 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
737 #ifdef NODE_NOT_IN_PAGE_FLAGS
738 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
739 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
740 SECTIONS_PGOFF : ZONES_PGOFF)
742 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
743 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
744 NODES_PGOFF : ZONES_PGOFF)
747 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
749 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
750 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
753 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
754 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
755 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
756 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
757 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
759 static inline enum zone_type
page_zonenum(const struct page
*page
)
761 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
764 #ifdef CONFIG_ZONE_DEVICE
765 static inline bool is_zone_device_page(const struct page
*page
)
767 return page_zonenum(page
) == ZONE_DEVICE
;
770 static inline bool is_zone_device_page(const struct page
*page
)
776 static inline void get_page(struct page
*page
)
778 page
= compound_head(page
);
780 * Getting a normal page or the head of a compound page
781 * requires to already have an elevated page->_refcount.
783 VM_BUG_ON_PAGE(page_ref_count(page
) <= 0, page
);
787 static inline void put_page(struct page
*page
)
789 page
= compound_head(page
);
791 if (put_page_testzero(page
))
795 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
796 #define SECTION_IN_PAGE_FLAGS
800 * The identification function is mainly used by the buddy allocator for
801 * determining if two pages could be buddies. We are not really identifying
802 * the zone since we could be using the section number id if we do not have
803 * node id available in page flags.
804 * We only guarantee that it will return the same value for two combinable
807 static inline int page_zone_id(struct page
*page
)
809 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
812 static inline int zone_to_nid(struct zone
*zone
)
821 #ifdef NODE_NOT_IN_PAGE_FLAGS
822 extern int page_to_nid(const struct page
*page
);
824 static inline int page_to_nid(const struct page
*page
)
826 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
830 #ifdef CONFIG_NUMA_BALANCING
831 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
833 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
836 static inline int cpupid_to_pid(int cpupid
)
838 return cpupid
& LAST__PID_MASK
;
841 static inline int cpupid_to_cpu(int cpupid
)
843 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
846 static inline int cpupid_to_nid(int cpupid
)
848 return cpu_to_node(cpupid_to_cpu(cpupid
));
851 static inline bool cpupid_pid_unset(int cpupid
)
853 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
856 static inline bool cpupid_cpu_unset(int cpupid
)
858 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
861 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
863 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
866 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
867 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
868 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
870 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
873 static inline int page_cpupid_last(struct page
*page
)
875 return page
->_last_cpupid
;
877 static inline void page_cpupid_reset_last(struct page
*page
)
879 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
882 static inline int page_cpupid_last(struct page
*page
)
884 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
887 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
889 static inline void page_cpupid_reset_last(struct page
*page
)
891 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
893 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
894 #else /* !CONFIG_NUMA_BALANCING */
895 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
897 return page_to_nid(page
); /* XXX */
900 static inline int page_cpupid_last(struct page
*page
)
902 return page_to_nid(page
); /* XXX */
905 static inline int cpupid_to_nid(int cpupid
)
910 static inline int cpupid_to_pid(int cpupid
)
915 static inline int cpupid_to_cpu(int cpupid
)
920 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
925 static inline bool cpupid_pid_unset(int cpupid
)
930 static inline void page_cpupid_reset_last(struct page
*page
)
934 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
938 #endif /* CONFIG_NUMA_BALANCING */
940 static inline struct zone
*page_zone(const struct page
*page
)
942 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
945 static inline pg_data_t
*page_pgdat(const struct page
*page
)
947 return NODE_DATA(page_to_nid(page
));
950 #ifdef SECTION_IN_PAGE_FLAGS
951 static inline void set_page_section(struct page
*page
, unsigned long section
)
953 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
954 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
957 static inline unsigned long page_to_section(const struct page
*page
)
959 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
963 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
965 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
966 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
969 static inline void set_page_node(struct page
*page
, unsigned long node
)
971 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
972 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
975 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
976 unsigned long node
, unsigned long pfn
)
978 set_page_zone(page
, zone
);
979 set_page_node(page
, node
);
980 #ifdef SECTION_IN_PAGE_FLAGS
981 set_page_section(page
, pfn_to_section_nr(pfn
));
986 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
988 return page
->mem_cgroup
;
990 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
992 WARN_ON_ONCE(!rcu_read_lock_held());
993 return READ_ONCE(page
->mem_cgroup
);
996 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
1000 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1002 WARN_ON_ONCE(!rcu_read_lock_held());
1008 * Some inline functions in vmstat.h depend on page_zone()
1010 #include <linux/vmstat.h>
1012 static __always_inline
void *lowmem_page_address(const struct page
*page
)
1014 return page_to_virt(page
);
1017 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1018 #define HASHED_PAGE_VIRTUAL
1021 #if defined(WANT_PAGE_VIRTUAL)
1022 static inline void *page_address(const struct page
*page
)
1024 return page
->virtual;
1026 static inline void set_page_address(struct page
*page
, void *address
)
1028 page
->virtual = address
;
1030 #define page_address_init() do { } while(0)
1033 #if defined(HASHED_PAGE_VIRTUAL)
1034 void *page_address(const struct page
*page
);
1035 void set_page_address(struct page
*page
, void *virtual);
1036 void page_address_init(void);
1039 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1040 #define page_address(page) lowmem_page_address(page)
1041 #define set_page_address(page, address) do { } while(0)
1042 #define page_address_init() do { } while(0)
1045 extern void *page_rmapping(struct page
*page
);
1046 extern struct anon_vma
*page_anon_vma(struct page
*page
);
1047 extern struct address_space
*page_mapping(struct page
*page
);
1049 extern struct address_space
*__page_file_mapping(struct page
*);
1052 struct address_space
*page_file_mapping(struct page
*page
)
1054 if (unlikely(PageSwapCache(page
)))
1055 return __page_file_mapping(page
);
1057 return page
->mapping
;
1060 extern pgoff_t
__page_file_index(struct page
*page
);
1063 * Return the pagecache index of the passed page. Regular pagecache pages
1064 * use ->index whereas swapcache pages use swp_offset(->private)
1066 static inline pgoff_t
page_index(struct page
*page
)
1068 if (unlikely(PageSwapCache(page
)))
1069 return __page_file_index(page
);
1073 bool page_mapped(struct page
*page
);
1074 struct address_space
*page_mapping(struct page
*page
);
1077 * Return true only if the page has been allocated with
1078 * ALLOC_NO_WATERMARKS and the low watermark was not
1079 * met implying that the system is under some pressure.
1081 static inline bool page_is_pfmemalloc(struct page
*page
)
1084 * Page index cannot be this large so this must be
1085 * a pfmemalloc page.
1087 return page
->index
== -1UL;
1091 * Only to be called by the page allocator on a freshly allocated
1094 static inline void set_page_pfmemalloc(struct page
*page
)
1099 static inline void clear_page_pfmemalloc(struct page
*page
)
1105 * Different kinds of faults, as returned by handle_mm_fault().
1106 * Used to decide whether a process gets delivered SIGBUS or
1107 * just gets major/minor fault counters bumped up.
1110 #define VM_FAULT_OOM 0x0001
1111 #define VM_FAULT_SIGBUS 0x0002
1112 #define VM_FAULT_MAJOR 0x0004
1113 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1114 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1115 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1116 #define VM_FAULT_SIGSEGV 0x0040
1118 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1119 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1120 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1121 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1122 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1124 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1126 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1127 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1130 #define VM_FAULT_RESULT_TRACE \
1131 { VM_FAULT_OOM, "OOM" }, \
1132 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1133 { VM_FAULT_MAJOR, "MAJOR" }, \
1134 { VM_FAULT_WRITE, "WRITE" }, \
1135 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1136 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1137 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1138 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1139 { VM_FAULT_LOCKED, "LOCKED" }, \
1140 { VM_FAULT_RETRY, "RETRY" }, \
1141 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1142 { VM_FAULT_DONE_COW, "DONE_COW" }
1144 /* Encode hstate index for a hwpoisoned large page */
1145 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1146 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1149 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1151 extern void pagefault_out_of_memory(void);
1153 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1156 * Flags passed to show_mem() and show_free_areas() to suppress output in
1159 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1161 extern void show_free_areas(unsigned int flags
, nodemask_t
*nodemask
);
1163 extern bool can_do_mlock(void);
1164 extern int user_shm_lock(size_t, struct user_struct
*);
1165 extern void user_shm_unlock(size_t, struct user_struct
*);
1168 * Parameter block passed down to zap_pte_range in exceptional cases.
1170 struct zap_details
{
1171 struct address_space
*check_mapping
; /* Check page->mapping if set */
1172 pgoff_t first_index
; /* Lowest page->index to unmap */
1173 pgoff_t last_index
; /* Highest page->index to unmap */
1176 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1178 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
1181 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1182 unsigned long size
);
1183 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1184 unsigned long size
);
1185 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1186 unsigned long start
, unsigned long end
);
1189 * mm_walk - callbacks for walk_page_range
1190 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1191 * this handler should only handle pud_trans_huge() puds.
1192 * the pmd_entry or pte_entry callbacks will be used for
1194 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1195 * this handler is required to be able to handle
1196 * pmd_trans_huge() pmds. They may simply choose to
1197 * split_huge_page() instead of handling it explicitly.
1198 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1199 * @pte_hole: if set, called for each hole at all levels
1200 * @hugetlb_entry: if set, called for each hugetlb entry
1201 * @test_walk: caller specific callback function to determine whether
1202 * we walk over the current vma or not. Returning 0
1203 * value means "do page table walk over the current vma,"
1204 * and a negative one means "abort current page table walk
1205 * right now." 1 means "skip the current vma."
1206 * @mm: mm_struct representing the target process of page table walk
1207 * @vma: vma currently walked (NULL if walking outside vmas)
1208 * @private: private data for callbacks' usage
1210 * (see the comment on walk_page_range() for more details)
1213 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
1214 unsigned long next
, struct mm_walk
*walk
);
1215 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1216 unsigned long next
, struct mm_walk
*walk
);
1217 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1218 unsigned long next
, struct mm_walk
*walk
);
1219 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1220 struct mm_walk
*walk
);
1221 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1222 unsigned long addr
, unsigned long next
,
1223 struct mm_walk
*walk
);
1224 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1225 struct mm_walk
*walk
);
1226 struct mm_struct
*mm
;
1227 struct vm_area_struct
*vma
;
1231 int walk_page_range(unsigned long addr
, unsigned long end
,
1232 struct mm_walk
*walk
);
1233 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1234 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1235 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1236 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1237 struct vm_area_struct
*vma
);
1238 void unmap_mapping_range(struct address_space
*mapping
,
1239 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1240 int follow_pte_pmd(struct mm_struct
*mm
, unsigned long address
,
1241 pte_t
**ptepp
, pmd_t
**pmdpp
, spinlock_t
**ptlp
);
1242 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1243 unsigned long *pfn
);
1244 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1245 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1246 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1247 void *buf
, int len
, int write
);
1249 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1250 loff_t
const holebegin
, loff_t
const holelen
)
1252 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1255 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1256 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1257 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1258 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1259 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1260 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1261 int invalidate_inode_page(struct page
*page
);
1264 extern int handle_mm_fault(struct vm_area_struct
*vma
, unsigned long address
,
1265 unsigned int flags
);
1266 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1267 unsigned long address
, unsigned int fault_flags
,
1270 static inline int handle_mm_fault(struct vm_area_struct
*vma
,
1271 unsigned long address
, unsigned int flags
)
1273 /* should never happen if there's no MMU */
1275 return VM_FAULT_SIGBUS
;
1277 static inline int fixup_user_fault(struct task_struct
*tsk
,
1278 struct mm_struct
*mm
, unsigned long address
,
1279 unsigned int fault_flags
, bool *unlocked
)
1281 /* should never happen if there's no MMU */
1287 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
,
1288 unsigned int gup_flags
);
1289 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1290 void *buf
, int len
, unsigned int gup_flags
);
1291 extern int __access_remote_vm(struct task_struct
*tsk
, struct mm_struct
*mm
,
1292 unsigned long addr
, void *buf
, int len
, unsigned int gup_flags
);
1294 long get_user_pages_remote(struct task_struct
*tsk
, struct mm_struct
*mm
,
1295 unsigned long start
, unsigned long nr_pages
,
1296 unsigned int gup_flags
, struct page
**pages
,
1297 struct vm_area_struct
**vmas
, int *locked
);
1298 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
1299 unsigned int gup_flags
, struct page
**pages
,
1300 struct vm_area_struct
**vmas
);
1301 long get_user_pages_locked(unsigned long start
, unsigned long nr_pages
,
1302 unsigned int gup_flags
, struct page
**pages
, int *locked
);
1303 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
1304 struct page
**pages
, unsigned int gup_flags
);
1305 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1306 struct page
**pages
);
1308 /* Container for pinned pfns / pages */
1309 struct frame_vector
{
1310 unsigned int nr_allocated
; /* Number of frames we have space for */
1311 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1312 bool got_ref
; /* Did we pin pages by getting page ref? */
1313 bool is_pfns
; /* Does array contain pages or pfns? */
1314 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1315 * pfns_vector_pages() or pfns_vector_pfns()
1319 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1320 void frame_vector_destroy(struct frame_vector
*vec
);
1321 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1322 unsigned int gup_flags
, struct frame_vector
*vec
);
1323 void put_vaddr_frames(struct frame_vector
*vec
);
1324 int frame_vector_to_pages(struct frame_vector
*vec
);
1325 void frame_vector_to_pfns(struct frame_vector
*vec
);
1327 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1329 return vec
->nr_frames
;
1332 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1335 int err
= frame_vector_to_pages(vec
);
1338 return ERR_PTR(err
);
1340 return (struct page
**)(vec
->ptrs
);
1343 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1346 frame_vector_to_pfns(vec
);
1347 return (unsigned long *)(vec
->ptrs
);
1351 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1352 struct page
**pages
);
1353 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1354 struct page
*get_dump_page(unsigned long addr
);
1356 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1357 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1358 unsigned int length
);
1360 int __set_page_dirty_nobuffers(struct page
*page
);
1361 int __set_page_dirty_no_writeback(struct page
*page
);
1362 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1364 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1365 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1366 struct bdi_writeback
*wb
);
1367 int set_page_dirty(struct page
*page
);
1368 int set_page_dirty_lock(struct page
*page
);
1369 void cancel_dirty_page(struct page
*page
);
1370 int clear_page_dirty_for_io(struct page
*page
);
1372 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1374 /* Is the vma a continuation of the stack vma above it? */
1375 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1377 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1380 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1382 return !vma
->vm_ops
;
1387 * The vma_is_shmem is not inline because it is used only by slow
1388 * paths in userfault.
1390 bool vma_is_shmem(struct vm_area_struct
*vma
);
1392 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
1395 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1398 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1399 (vma
->vm_start
== addr
) &&
1400 !vma_growsdown(vma
->vm_prev
, addr
);
1403 /* Is the vma a continuation of the stack vma below it? */
1404 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1406 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1409 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1412 return (vma
->vm_flags
& VM_GROWSUP
) &&
1413 (vma
->vm_end
== addr
) &&
1414 !vma_growsup(vma
->vm_next
, addr
);
1417 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
1419 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1420 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1421 unsigned long new_addr
, unsigned long len
,
1422 bool need_rmap_locks
);
1423 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1424 unsigned long end
, pgprot_t newprot
,
1425 int dirty_accountable
, int prot_numa
);
1426 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1427 struct vm_area_struct
**pprev
, unsigned long start
,
1428 unsigned long end
, unsigned long newflags
);
1431 * doesn't attempt to fault and will return short.
1433 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1434 struct page
**pages
);
1436 * per-process(per-mm_struct) statistics.
1438 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1440 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1442 #ifdef SPLIT_RSS_COUNTING
1444 * counter is updated in asynchronous manner and may go to minus.
1445 * But it's never be expected number for users.
1450 return (unsigned long)val
;
1453 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1455 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1458 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1460 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1463 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1465 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1468 /* Optimized variant when page is already known not to be PageAnon */
1469 static inline int mm_counter_file(struct page
*page
)
1471 if (PageSwapBacked(page
))
1472 return MM_SHMEMPAGES
;
1473 return MM_FILEPAGES
;
1476 static inline int mm_counter(struct page
*page
)
1479 return MM_ANONPAGES
;
1480 return mm_counter_file(page
);
1483 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1485 return get_mm_counter(mm
, MM_FILEPAGES
) +
1486 get_mm_counter(mm
, MM_ANONPAGES
) +
1487 get_mm_counter(mm
, MM_SHMEMPAGES
);
1490 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1492 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1495 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1497 return max(mm
->hiwater_vm
, mm
->total_vm
);
1500 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1502 unsigned long _rss
= get_mm_rss(mm
);
1504 if ((mm
)->hiwater_rss
< _rss
)
1505 (mm
)->hiwater_rss
= _rss
;
1508 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1510 if (mm
->hiwater_vm
< mm
->total_vm
)
1511 mm
->hiwater_vm
= mm
->total_vm
;
1514 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1516 mm
->hiwater_rss
= get_mm_rss(mm
);
1519 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1520 struct mm_struct
*mm
)
1522 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1524 if (*maxrss
< hiwater_rss
)
1525 *maxrss
= hiwater_rss
;
1528 #if defined(SPLIT_RSS_COUNTING)
1529 void sync_mm_rss(struct mm_struct
*mm
);
1531 static inline void sync_mm_rss(struct mm_struct
*mm
)
1536 #ifndef __HAVE_ARCH_PTE_DEVMAP
1537 static inline int pte_devmap(pte_t pte
)
1543 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
1545 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1547 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1551 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1555 #ifdef __PAGETABLE_P4D_FOLDED
1556 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1557 unsigned long address
)
1562 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1565 #ifdef __PAGETABLE_PUD_FOLDED
1566 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1567 unsigned long address
)
1572 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
1575 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1576 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1577 unsigned long address
)
1582 static inline void mm_nr_pmds_init(struct mm_struct
*mm
) {}
1584 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1589 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1590 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1593 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1595 static inline void mm_nr_pmds_init(struct mm_struct
*mm
)
1597 atomic_long_set(&mm
->nr_pmds
, 0);
1600 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1602 return atomic_long_read(&mm
->nr_pmds
);
1605 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1607 atomic_long_inc(&mm
->nr_pmds
);
1610 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1612 atomic_long_dec(&mm
->nr_pmds
);
1616 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
1617 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1620 * The following ifdef needed to get the 4level-fixup.h header to work.
1621 * Remove it when 4level-fixup.h has been removed.
1623 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1625 #ifndef __ARCH_HAS_5LEVEL_HACK
1626 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1627 unsigned long address
)
1629 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
1630 NULL
: p4d_offset(pgd
, address
);
1633 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1634 unsigned long address
)
1636 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
1637 NULL
: pud_offset(p4d
, address
);
1639 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1641 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1643 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1644 NULL
: pmd_offset(pud
, address
);
1646 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1648 #if USE_SPLIT_PTE_PTLOCKS
1649 #if ALLOC_SPLIT_PTLOCKS
1650 void __init
ptlock_cache_init(void);
1651 extern bool ptlock_alloc(struct page
*page
);
1652 extern void ptlock_free(struct page
*page
);
1654 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1658 #else /* ALLOC_SPLIT_PTLOCKS */
1659 static inline void ptlock_cache_init(void)
1663 static inline bool ptlock_alloc(struct page
*page
)
1668 static inline void ptlock_free(struct page
*page
)
1672 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1676 #endif /* ALLOC_SPLIT_PTLOCKS */
1678 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1680 return ptlock_ptr(pmd_page(*pmd
));
1683 static inline bool ptlock_init(struct page
*page
)
1686 * prep_new_page() initialize page->private (and therefore page->ptl)
1687 * with 0. Make sure nobody took it in use in between.
1689 * It can happen if arch try to use slab for page table allocation:
1690 * slab code uses page->slab_cache, which share storage with page->ptl.
1692 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1693 if (!ptlock_alloc(page
))
1695 spin_lock_init(ptlock_ptr(page
));
1699 /* Reset page->mapping so free_pages_check won't complain. */
1700 static inline void pte_lock_deinit(struct page
*page
)
1702 page
->mapping
= NULL
;
1706 #else /* !USE_SPLIT_PTE_PTLOCKS */
1708 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1710 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1712 return &mm
->page_table_lock
;
1714 static inline void ptlock_cache_init(void) {}
1715 static inline bool ptlock_init(struct page
*page
) { return true; }
1716 static inline void pte_lock_deinit(struct page
*page
) {}
1717 #endif /* USE_SPLIT_PTE_PTLOCKS */
1719 static inline void pgtable_init(void)
1721 ptlock_cache_init();
1722 pgtable_cache_init();
1725 static inline bool pgtable_page_ctor(struct page
*page
)
1727 if (!ptlock_init(page
))
1729 inc_zone_page_state(page
, NR_PAGETABLE
);
1733 static inline void pgtable_page_dtor(struct page
*page
)
1735 pte_lock_deinit(page
);
1736 dec_zone_page_state(page
, NR_PAGETABLE
);
1739 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1741 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1742 pte_t *__pte = pte_offset_map(pmd, address); \
1748 #define pte_unmap_unlock(pte, ptl) do { \
1753 #define pte_alloc(mm, pmd, address) \
1754 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1756 #define pte_alloc_map(mm, pmd, address) \
1757 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1759 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1760 (pte_alloc(mm, pmd, address) ? \
1761 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1763 #define pte_alloc_kernel(pmd, address) \
1764 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1765 NULL: pte_offset_kernel(pmd, address))
1767 #if USE_SPLIT_PMD_PTLOCKS
1769 static struct page
*pmd_to_page(pmd_t
*pmd
)
1771 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1772 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1775 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1777 return ptlock_ptr(pmd_to_page(pmd
));
1780 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1782 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1783 page
->pmd_huge_pte
= NULL
;
1785 return ptlock_init(page
);
1788 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1790 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1791 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1796 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1800 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1802 return &mm
->page_table_lock
;
1805 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1806 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1808 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1812 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1814 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1820 * No scalability reason to split PUD locks yet, but follow the same pattern
1821 * as the PMD locks to make it easier if we decide to. The VM should not be
1822 * considered ready to switch to split PUD locks yet; there may be places
1823 * which need to be converted from page_table_lock.
1825 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
1827 return &mm
->page_table_lock
;
1830 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
1832 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
1838 extern void __init
pagecache_init(void);
1839 extern void free_area_init(unsigned long * zones_size
);
1840 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1841 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1842 extern void free_initmem(void);
1845 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1846 * into the buddy system. The freed pages will be poisoned with pattern
1847 * "poison" if it's within range [0, UCHAR_MAX].
1848 * Return pages freed into the buddy system.
1850 extern unsigned long free_reserved_area(void *start
, void *end
,
1851 int poison
, char *s
);
1853 #ifdef CONFIG_HIGHMEM
1855 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1856 * and totalram_pages.
1858 extern void free_highmem_page(struct page
*page
);
1861 extern void adjust_managed_page_count(struct page
*page
, long count
);
1862 extern void mem_init_print_info(const char *str
);
1864 extern void reserve_bootmem_region(phys_addr_t start
, phys_addr_t end
);
1866 /* Free the reserved page into the buddy system, so it gets managed. */
1867 static inline void __free_reserved_page(struct page
*page
)
1869 ClearPageReserved(page
);
1870 init_page_count(page
);
1874 static inline void free_reserved_page(struct page
*page
)
1876 __free_reserved_page(page
);
1877 adjust_managed_page_count(page
, 1);
1880 static inline void mark_page_reserved(struct page
*page
)
1882 SetPageReserved(page
);
1883 adjust_managed_page_count(page
, -1);
1887 * Default method to free all the __init memory into the buddy system.
1888 * The freed pages will be poisoned with pattern "poison" if it's within
1889 * range [0, UCHAR_MAX].
1890 * Return pages freed into the buddy system.
1892 static inline unsigned long free_initmem_default(int poison
)
1894 extern char __init_begin
[], __init_end
[];
1896 return free_reserved_area(&__init_begin
, &__init_end
,
1897 poison
, "unused kernel");
1900 static inline unsigned long get_num_physpages(void)
1903 unsigned long phys_pages
= 0;
1905 for_each_online_node(nid
)
1906 phys_pages
+= node_present_pages(nid
);
1911 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1913 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1914 * zones, allocate the backing mem_map and account for memory holes in a more
1915 * architecture independent manner. This is a substitute for creating the
1916 * zone_sizes[] and zholes_size[] arrays and passing them to
1917 * free_area_init_node()
1919 * An architecture is expected to register range of page frames backed by
1920 * physical memory with memblock_add[_node]() before calling
1921 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1922 * usage, an architecture is expected to do something like
1924 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1926 * for_each_valid_physical_page_range()
1927 * memblock_add_node(base, size, nid)
1928 * free_area_init_nodes(max_zone_pfns);
1930 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1931 * registered physical page range. Similarly
1932 * sparse_memory_present_with_active_regions() calls memory_present() for
1933 * each range when SPARSEMEM is enabled.
1935 * See mm/page_alloc.c for more information on each function exposed by
1936 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1938 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1939 unsigned long node_map_pfn_alignment(void);
1940 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1941 unsigned long end_pfn
);
1942 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1943 unsigned long end_pfn
);
1944 extern void get_pfn_range_for_nid(unsigned int nid
,
1945 unsigned long *start_pfn
, unsigned long *end_pfn
);
1946 extern unsigned long find_min_pfn_with_active_regions(void);
1947 extern void free_bootmem_with_active_regions(int nid
,
1948 unsigned long max_low_pfn
);
1949 extern void sparse_memory_present_with_active_regions(int nid
);
1951 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1953 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1954 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1955 static inline int __early_pfn_to_nid(unsigned long pfn
,
1956 struct mminit_pfnnid_cache
*state
)
1961 /* please see mm/page_alloc.c */
1962 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1963 /* there is a per-arch backend function. */
1964 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
1965 struct mminit_pfnnid_cache
*state
);
1968 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1969 extern void memmap_init_zone(unsigned long, int, unsigned long,
1970 unsigned long, enum memmap_context
);
1971 extern void setup_per_zone_wmarks(void);
1972 extern int __meminit
init_per_zone_wmark_min(void);
1973 extern void mem_init(void);
1974 extern void __init
mmap_init(void);
1975 extern void show_mem(unsigned int flags
, nodemask_t
*nodemask
);
1976 extern long si_mem_available(void);
1977 extern void si_meminfo(struct sysinfo
* val
);
1978 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1979 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
1980 extern unsigned long arch_reserved_kernel_pages(void);
1983 extern __printf(3, 4)
1984 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
1986 extern void setup_per_cpu_pageset(void);
1988 extern void zone_pcp_update(struct zone
*zone
);
1989 extern void zone_pcp_reset(struct zone
*zone
);
1992 extern int min_free_kbytes
;
1993 extern int watermark_scale_factor
;
1996 extern atomic_long_t mmap_pages_allocated
;
1997 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1999 /* interval_tree.c */
2000 void vma_interval_tree_insert(struct vm_area_struct
*node
,
2001 struct rb_root
*root
);
2002 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
2003 struct vm_area_struct
*prev
,
2004 struct rb_root
*root
);
2005 void vma_interval_tree_remove(struct vm_area_struct
*node
,
2006 struct rb_root
*root
);
2007 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
2008 unsigned long start
, unsigned long last
);
2009 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
2010 unsigned long start
, unsigned long last
);
2012 #define vma_interval_tree_foreach(vma, root, start, last) \
2013 for (vma = vma_interval_tree_iter_first(root, start, last); \
2014 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2016 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
2017 struct rb_root
*root
);
2018 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
2019 struct rb_root
*root
);
2020 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
2021 struct rb_root
*root
, unsigned long start
, unsigned long last
);
2022 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
2023 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
2024 #ifdef CONFIG_DEBUG_VM_RB
2025 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
2028 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2029 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2030 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2033 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
2034 extern int __vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2035 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
,
2036 struct vm_area_struct
*expand
);
2037 static inline int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2038 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
)
2040 return __vma_adjust(vma
, start
, end
, pgoff
, insert
, NULL
);
2042 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
2043 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
2044 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
2045 struct mempolicy
*, struct vm_userfaultfd_ctx
);
2046 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
2047 extern int __split_vma(struct mm_struct
*, struct vm_area_struct
*,
2048 unsigned long addr
, int new_below
);
2049 extern int split_vma(struct mm_struct
*, struct vm_area_struct
*,
2050 unsigned long addr
, int new_below
);
2051 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
2052 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
2053 struct rb_node
**, struct rb_node
*);
2054 extern void unlink_file_vma(struct vm_area_struct
*);
2055 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
2056 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
2057 bool *need_rmap_locks
);
2058 extern void exit_mmap(struct mm_struct
*);
2060 static inline int check_data_rlimit(unsigned long rlim
,
2062 unsigned long start
,
2063 unsigned long end_data
,
2064 unsigned long start_data
)
2066 if (rlim
< RLIM_INFINITY
) {
2067 if (((new - start
) + (end_data
- start_data
)) > rlim
)
2074 extern int mm_take_all_locks(struct mm_struct
*mm
);
2075 extern void mm_drop_all_locks(struct mm_struct
*mm
);
2077 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2078 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2079 extern struct file
*get_task_exe_file(struct task_struct
*task
);
2081 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2082 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2084 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
2085 const struct vm_special_mapping
*sm
);
2086 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2087 unsigned long addr
, unsigned long len
,
2088 unsigned long flags
,
2089 const struct vm_special_mapping
*spec
);
2090 /* This is an obsolete alternative to _install_special_mapping. */
2091 extern int install_special_mapping(struct mm_struct
*mm
,
2092 unsigned long addr
, unsigned long len
,
2093 unsigned long flags
, struct page
**pages
);
2095 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
2097 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
2098 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
2099 struct list_head
*uf
);
2100 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
2101 unsigned long len
, unsigned long prot
, unsigned long flags
,
2102 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
,
2103 struct list_head
*uf
);
2104 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
2105 struct list_head
*uf
);
2107 static inline unsigned long
2108 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
2109 unsigned long len
, unsigned long prot
, unsigned long flags
,
2110 unsigned long pgoff
, unsigned long *populate
,
2111 struct list_head
*uf
)
2113 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
, uf
);
2117 extern int __mm_populate(unsigned long addr
, unsigned long len
,
2119 static inline void mm_populate(unsigned long addr
, unsigned long len
)
2122 (void) __mm_populate(addr
, len
, 1);
2125 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
2128 /* These take the mm semaphore themselves */
2129 extern int __must_check
vm_brk(unsigned long, unsigned long);
2130 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
2131 extern int vm_munmap(unsigned long, size_t);
2132 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
2133 unsigned long, unsigned long,
2134 unsigned long, unsigned long);
2136 struct vm_unmapped_area_info
{
2137 #define VM_UNMAPPED_AREA_TOPDOWN 1
2138 unsigned long flags
;
2139 unsigned long length
;
2140 unsigned long low_limit
;
2141 unsigned long high_limit
;
2142 unsigned long align_mask
;
2143 unsigned long align_offset
;
2146 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
2147 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
2150 * Search for an unmapped address range.
2152 * We are looking for a range that:
2153 * - does not intersect with any VMA;
2154 * - is contained within the [low_limit, high_limit) interval;
2155 * - is at least the desired size.
2156 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2158 static inline unsigned long
2159 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
2161 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
2162 return unmapped_area_topdown(info
);
2164 return unmapped_area(info
);
2168 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2169 extern void truncate_inode_pages_range(struct address_space
*,
2170 loff_t lstart
, loff_t lend
);
2171 extern void truncate_inode_pages_final(struct address_space
*);
2173 /* generic vm_area_ops exported for stackable file systems */
2174 extern int filemap_fault(struct vm_fault
*vmf
);
2175 extern void filemap_map_pages(struct vm_fault
*vmf
,
2176 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
2177 extern int filemap_page_mkwrite(struct vm_fault
*vmf
);
2179 /* mm/page-writeback.c */
2180 int write_one_page(struct page
*page
, int wait
);
2181 void task_dirty_inc(struct task_struct
*tsk
);
2184 #define VM_MAX_READAHEAD 128 /* kbytes */
2185 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2187 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2188 pgoff_t offset
, unsigned long nr_to_read
);
2190 void page_cache_sync_readahead(struct address_space
*mapping
,
2191 struct file_ra_state
*ra
,
2194 unsigned long size
);
2196 void page_cache_async_readahead(struct address_space
*mapping
,
2197 struct file_ra_state
*ra
,
2201 unsigned long size
);
2203 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2204 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2206 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2207 extern int expand_downwards(struct vm_area_struct
*vma
,
2208 unsigned long address
);
2210 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2212 #define expand_upwards(vma, address) (0)
2215 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2216 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2217 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2218 struct vm_area_struct
**pprev
);
2220 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2221 NULL if none. Assume start_addr < end_addr. */
2222 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2224 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2226 if (vma
&& end_addr
<= vma
->vm_start
)
2231 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2233 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2236 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2237 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2238 unsigned long vm_start
, unsigned long vm_end
)
2240 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2242 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2249 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2250 void vma_set_page_prot(struct vm_area_struct
*vma
);
2252 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2256 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2258 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2262 #ifdef CONFIG_NUMA_BALANCING
2263 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2264 unsigned long start
, unsigned long end
);
2267 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2268 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2269 unsigned long pfn
, unsigned long size
, pgprot_t
);
2270 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2271 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2273 int vm_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2274 unsigned long pfn
, pgprot_t pgprot
);
2275 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2277 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2280 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2281 unsigned long address
, unsigned int foll_flags
,
2282 unsigned int *page_mask
);
2284 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2285 unsigned long address
, unsigned int foll_flags
)
2287 unsigned int unused_page_mask
;
2288 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2291 #define FOLL_WRITE 0x01 /* check pte is writable */
2292 #define FOLL_TOUCH 0x02 /* mark page accessed */
2293 #define FOLL_GET 0x04 /* do get_page on page */
2294 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2295 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2296 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2297 * and return without waiting upon it */
2298 #define FOLL_POPULATE 0x40 /* fault in page */
2299 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2300 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2301 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2302 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2303 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2304 #define FOLL_MLOCK 0x1000 /* lock present pages */
2305 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2306 #define FOLL_COW 0x4000 /* internal GUP flag */
2308 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2310 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2311 unsigned long size
, pte_fn_t fn
, void *data
);
2314 #ifdef CONFIG_PAGE_POISONING
2315 extern bool page_poisoning_enabled(void);
2316 extern void kernel_poison_pages(struct page
*page
, int numpages
, int enable
);
2317 extern bool page_is_poisoned(struct page
*page
);
2319 static inline bool page_poisoning_enabled(void) { return false; }
2320 static inline void kernel_poison_pages(struct page
*page
, int numpages
,
2322 static inline bool page_is_poisoned(struct page
*page
) { return false; }
2325 #ifdef CONFIG_DEBUG_PAGEALLOC
2326 extern bool _debug_pagealloc_enabled
;
2327 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2329 static inline bool debug_pagealloc_enabled(void)
2331 return _debug_pagealloc_enabled
;
2335 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2337 if (!debug_pagealloc_enabled())
2340 __kernel_map_pages(page
, numpages
, enable
);
2342 #ifdef CONFIG_HIBERNATION
2343 extern bool kernel_page_present(struct page
*page
);
2344 #endif /* CONFIG_HIBERNATION */
2345 #else /* CONFIG_DEBUG_PAGEALLOC */
2347 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2348 #ifdef CONFIG_HIBERNATION
2349 static inline bool kernel_page_present(struct page
*page
) { return true; }
2350 #endif /* CONFIG_HIBERNATION */
2351 static inline bool debug_pagealloc_enabled(void)
2355 #endif /* CONFIG_DEBUG_PAGEALLOC */
2357 #ifdef __HAVE_ARCH_GATE_AREA
2358 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2359 extern int in_gate_area_no_mm(unsigned long addr
);
2360 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2362 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2366 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2367 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2371 #endif /* __HAVE_ARCH_GATE_AREA */
2373 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
2375 #ifdef CONFIG_SYSCTL
2376 extern int sysctl_drop_caches
;
2377 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2378 void __user
*, size_t *, loff_t
*);
2381 void drop_slab(void);
2382 void drop_slab_node(int nid
);
2385 #define randomize_va_space 0
2387 extern int randomize_va_space
;
2390 const char * arch_vma_name(struct vm_area_struct
*vma
);
2391 void print_vma_addr(char *prefix
, unsigned long rip
);
2393 void sparse_mem_maps_populate_node(struct page
**map_map
,
2394 unsigned long pnum_begin
,
2395 unsigned long pnum_end
,
2396 unsigned long map_count
,
2399 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2400 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2401 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2402 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
2403 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2404 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2405 void *vmemmap_alloc_block(unsigned long size
, int node
);
2407 void *__vmemmap_alloc_block_buf(unsigned long size
, int node
,
2408 struct vmem_altmap
*altmap
);
2409 static inline void *vmemmap_alloc_block_buf(unsigned long size
, int node
)
2411 return __vmemmap_alloc_block_buf(size
, node
, NULL
);
2414 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2415 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2417 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2418 void vmemmap_populate_print_last(void);
2419 #ifdef CONFIG_MEMORY_HOTPLUG
2420 void vmemmap_free(unsigned long start
, unsigned long end
);
2422 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2423 unsigned long size
);
2426 MF_COUNT_INCREASED
= 1 << 0,
2427 MF_ACTION_REQUIRED
= 1 << 1,
2428 MF_MUST_KILL
= 1 << 2,
2429 MF_SOFT_OFFLINE
= 1 << 3,
2431 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2432 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2433 extern int unpoison_memory(unsigned long pfn
);
2434 extern int get_hwpoison_page(struct page
*page
);
2435 #define put_hwpoison_page(page) put_page(page)
2436 extern int sysctl_memory_failure_early_kill
;
2437 extern int sysctl_memory_failure_recovery
;
2438 extern void shake_page(struct page
*p
, int access
);
2439 extern atomic_long_t num_poisoned_pages
;
2440 extern int soft_offline_page(struct page
*page
, int flags
);
2444 * Error handlers for various types of pages.
2447 MF_IGNORED
, /* Error: cannot be handled */
2448 MF_FAILED
, /* Error: handling failed */
2449 MF_DELAYED
, /* Will be handled later */
2450 MF_RECOVERED
, /* Successfully recovered */
2453 enum mf_action_page_type
{
2455 MF_MSG_KERNEL_HIGH_ORDER
,
2457 MF_MSG_DIFFERENT_COMPOUND
,
2458 MF_MSG_POISONED_HUGE
,
2461 MF_MSG_UNMAP_FAILED
,
2462 MF_MSG_DIRTY_SWAPCACHE
,
2463 MF_MSG_CLEAN_SWAPCACHE
,
2464 MF_MSG_DIRTY_MLOCKED_LRU
,
2465 MF_MSG_CLEAN_MLOCKED_LRU
,
2466 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2467 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2470 MF_MSG_TRUNCATED_LRU
,
2476 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2477 extern void clear_huge_page(struct page
*page
,
2479 unsigned int pages_per_huge_page
);
2480 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2481 unsigned long addr
, struct vm_area_struct
*vma
,
2482 unsigned int pages_per_huge_page
);
2483 extern long copy_huge_page_from_user(struct page
*dst_page
,
2484 const void __user
*usr_src
,
2485 unsigned int pages_per_huge_page
,
2486 bool allow_pagefault
);
2487 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2489 extern struct page_ext_operations debug_guardpage_ops
;
2491 #ifdef CONFIG_DEBUG_PAGEALLOC
2492 extern unsigned int _debug_guardpage_minorder
;
2493 extern bool _debug_guardpage_enabled
;
2495 static inline unsigned int debug_guardpage_minorder(void)
2497 return _debug_guardpage_minorder
;
2500 static inline bool debug_guardpage_enabled(void)
2502 return _debug_guardpage_enabled
;
2505 static inline bool page_is_guard(struct page
*page
)
2507 struct page_ext
*page_ext
;
2509 if (!debug_guardpage_enabled())
2512 page_ext
= lookup_page_ext(page
);
2513 if (unlikely(!page_ext
))
2516 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2519 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2520 static inline bool debug_guardpage_enabled(void) { return false; }
2521 static inline bool page_is_guard(struct page
*page
) { return false; }
2522 #endif /* CONFIG_DEBUG_PAGEALLOC */
2524 #if MAX_NUMNODES > 1
2525 void __init
setup_nr_node_ids(void);
2527 static inline void setup_nr_node_ids(void) {}
2530 #endif /* __KERNEL__ */
2531 #endif /* _LINUX_MM_H */