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 * To prevent common memory management code establishing
77 * a zero page mapping on a read fault.
78 * This macro should be defined within <asm/pgtable.h>.
79 * s390 does this to prevent multiplexing of hardware bits
80 * related to the physical page in case of virtualization.
82 #ifndef mm_forbids_zeropage
83 #define mm_forbids_zeropage(X) (0)
87 * Default maximum number of active map areas, this limits the number of vmas
88 * per mm struct. Users can overwrite this number by sysctl but there is a
91 * When a program's coredump is generated as ELF format, a section is created
92 * per a vma. In ELF, the number of sections is represented in unsigned short.
93 * This means the number of sections should be smaller than 65535 at coredump.
94 * Because the kernel adds some informative sections to a image of program at
95 * generating coredump, we need some margin. The number of extra sections is
96 * 1-3 now and depends on arch. We use "5" as safe margin, here.
98 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
99 * not a hard limit any more. Although some userspace tools can be surprised by
102 #define MAPCOUNT_ELF_CORE_MARGIN (5)
103 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
105 extern int sysctl_max_map_count
;
107 extern unsigned long sysctl_user_reserve_kbytes
;
108 extern unsigned long sysctl_admin_reserve_kbytes
;
110 extern int sysctl_overcommit_memory
;
111 extern int sysctl_overcommit_ratio
;
112 extern unsigned long sysctl_overcommit_kbytes
;
114 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
116 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
119 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
121 /* to align the pointer to the (next) page boundary */
122 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
124 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
125 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)addr, PAGE_SIZE)
128 * Linux kernel virtual memory manager primitives.
129 * The idea being to have a "virtual" mm in the same way
130 * we have a virtual fs - giving a cleaner interface to the
131 * mm details, and allowing different kinds of memory mappings
132 * (from shared memory to executable loading to arbitrary
136 extern struct kmem_cache
*vm_area_cachep
;
139 extern struct rb_root nommu_region_tree
;
140 extern struct rw_semaphore nommu_region_sem
;
142 extern unsigned int kobjsize(const void *objp
);
146 * vm_flags in vm_area_struct, see mm_types.h.
147 * When changing, update also include/trace/events/mmflags.h
149 #define VM_NONE 0x00000000
151 #define VM_READ 0x00000001 /* currently active flags */
152 #define VM_WRITE 0x00000002
153 #define VM_EXEC 0x00000004
154 #define VM_SHARED 0x00000008
156 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
157 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
158 #define VM_MAYWRITE 0x00000020
159 #define VM_MAYEXEC 0x00000040
160 #define VM_MAYSHARE 0x00000080
162 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
163 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
164 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
165 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
166 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
168 #define VM_LOCKED 0x00002000
169 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
171 /* Used by sys_madvise() */
172 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
173 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
175 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
176 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
177 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
178 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
179 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
180 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
181 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
182 #define VM_ARCH_2 0x02000000
183 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
185 #ifdef CONFIG_MEM_SOFT_DIRTY
186 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
188 # define VM_SOFTDIRTY 0
191 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
192 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
193 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
194 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
196 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
197 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
198 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
199 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
200 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
201 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
202 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
203 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
204 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
205 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
207 #if defined(CONFIG_X86)
208 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
209 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
210 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
211 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
212 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
213 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
214 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
216 #elif defined(CONFIG_PPC)
217 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
218 #elif defined(CONFIG_PARISC)
219 # define VM_GROWSUP VM_ARCH_1
220 #elif defined(CONFIG_METAG)
221 # define VM_GROWSUP VM_ARCH_1
222 #elif defined(CONFIG_IA64)
223 # define VM_GROWSUP VM_ARCH_1
224 #elif !defined(CONFIG_MMU)
225 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
228 #if defined(CONFIG_X86)
229 /* MPX specific bounds table or bounds directory */
230 # define VM_MPX VM_ARCH_2
234 # define VM_GROWSUP VM_NONE
237 /* Bits set in the VMA until the stack is in its final location */
238 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
240 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
241 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
244 #ifdef CONFIG_STACK_GROWSUP
245 #define VM_STACK VM_GROWSUP
247 #define VM_STACK VM_GROWSDOWN
250 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
253 * Special vmas that are non-mergable, non-mlock()able.
254 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
256 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
258 /* This mask defines which mm->def_flags a process can inherit its parent */
259 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
261 /* This mask is used to clear all the VMA flags used by mlock */
262 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
265 * mapping from the currently active vm_flags protection bits (the
266 * low four bits) to a page protection mask..
268 extern pgprot_t protection_map
[16];
270 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
271 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
272 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
273 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
274 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
275 #define FAULT_FLAG_TRIED 0x20 /* Second try */
276 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
277 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
278 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
281 * vm_fault is filled by the the pagefault handler and passed to the vma's
282 * ->fault function. The vma's ->fault is responsible for returning a bitmask
283 * of VM_FAULT_xxx flags that give details about how the fault was handled.
285 * MM layer fills up gfp_mask for page allocations but fault handler might
286 * alter it if its implementation requires a different allocation context.
288 * pgoff should be used in favour of virtual_address, if possible.
291 unsigned int flags
; /* FAULT_FLAG_xxx flags */
292 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
293 pgoff_t pgoff
; /* Logical page offset based on vma */
294 void __user
*virtual_address
; /* Faulting virtual address */
296 struct page
*cow_page
; /* Handler may choose to COW */
297 struct page
*page
; /* ->fault handlers should return a
298 * page here, unless VM_FAULT_NOPAGE
299 * is set (which is also implied by
302 /* for ->map_pages() only */
303 pgoff_t max_pgoff
; /* map pages for offset from pgoff till
304 * max_pgoff inclusive */
305 pte_t
*pte
; /* pte entry associated with ->pgoff */
309 * These are the virtual MM functions - opening of an area, closing and
310 * unmapping it (needed to keep files on disk up-to-date etc), pointer
311 * to the functions called when a no-page or a wp-page exception occurs.
313 struct vm_operations_struct
{
314 void (*open
)(struct vm_area_struct
* area
);
315 void (*close
)(struct vm_area_struct
* area
);
316 int (*mremap
)(struct vm_area_struct
* area
);
317 int (*fault
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
318 int (*pmd_fault
)(struct vm_area_struct
*, unsigned long address
,
319 pmd_t
*, unsigned int flags
);
320 void (*map_pages
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
322 /* notification that a previously read-only page is about to become
323 * writable, if an error is returned it will cause a SIGBUS */
324 int (*page_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
326 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
327 int (*pfn_mkwrite
)(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
329 /* called by access_process_vm when get_user_pages() fails, typically
330 * for use by special VMAs that can switch between memory and hardware
332 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
333 void *buf
, int len
, int write
);
335 /* Called by the /proc/PID/maps code to ask the vma whether it
336 * has a special name. Returning non-NULL will also cause this
337 * vma to be dumped unconditionally. */
338 const char *(*name
)(struct vm_area_struct
*vma
);
342 * set_policy() op must add a reference to any non-NULL @new mempolicy
343 * to hold the policy upon return. Caller should pass NULL @new to
344 * remove a policy and fall back to surrounding context--i.e. do not
345 * install a MPOL_DEFAULT policy, nor the task or system default
348 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
351 * get_policy() op must add reference [mpol_get()] to any policy at
352 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
353 * in mm/mempolicy.c will do this automatically.
354 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
355 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
356 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
357 * must return NULL--i.e., do not "fallback" to task or system default
360 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
364 * Called by vm_normal_page() for special PTEs to find the
365 * page for @addr. This is useful if the default behavior
366 * (using pte_page()) would not find the correct page.
368 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
375 #define page_private(page) ((page)->private)
376 #define set_page_private(page, v) ((page)->private = (v))
378 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
379 static inline int pmd_devmap(pmd_t pmd
)
386 * FIXME: take this include out, include page-flags.h in
387 * files which need it (119 of them)
389 #include <linux/page-flags.h>
390 #include <linux/huge_mm.h>
393 * Methods to modify the page usage count.
395 * What counts for a page usage:
396 * - cache mapping (page->mapping)
397 * - private data (page->private)
398 * - page mapped in a task's page tables, each mapping
399 * is counted separately
401 * Also, many kernel routines increase the page count before a critical
402 * routine so they can be sure the page doesn't go away from under them.
406 * Drop a ref, return true if the refcount fell to zero (the page has no users)
408 static inline int put_page_testzero(struct page
*page
)
410 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
411 return page_ref_dec_and_test(page
);
415 * Try to grab a ref unless the page has a refcount of zero, return false if
417 * This can be called when MMU is off so it must not access
418 * any of the virtual mappings.
420 static inline int get_page_unless_zero(struct page
*page
)
422 return page_ref_add_unless(page
, 1, 0);
425 extern int page_is_ram(unsigned long pfn
);
433 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
436 /* Support for virtually mapped pages */
437 struct page
*vmalloc_to_page(const void *addr
);
438 unsigned long vmalloc_to_pfn(const void *addr
);
441 * Determine if an address is within the vmalloc range
443 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
444 * is no special casing required.
446 static inline int is_vmalloc_addr(const void *x
)
449 unsigned long addr
= (unsigned long)x
;
451 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
457 extern int is_vmalloc_or_module_addr(const void *x
);
459 static inline int is_vmalloc_or_module_addr(const void *x
)
465 extern void kvfree(const void *addr
);
467 static inline atomic_t
*compound_mapcount_ptr(struct page
*page
)
469 return &page
[1].compound_mapcount
;
472 static inline int compound_mapcount(struct page
*page
)
474 if (!PageCompound(page
))
476 page
= compound_head(page
);
477 return atomic_read(compound_mapcount_ptr(page
)) + 1;
481 * The atomic page->_mapcount, starts from -1: so that transitions
482 * both from it and to it can be tracked, using atomic_inc_and_test
483 * and atomic_add_negative(-1).
485 static inline void page_mapcount_reset(struct page
*page
)
487 atomic_set(&(page
)->_mapcount
, -1);
490 int __page_mapcount(struct page
*page
);
492 static inline int page_mapcount(struct page
*page
)
494 VM_BUG_ON_PAGE(PageSlab(page
), page
);
496 if (unlikely(PageCompound(page
)))
497 return __page_mapcount(page
);
498 return atomic_read(&page
->_mapcount
) + 1;
501 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
502 int total_mapcount(struct page
*page
);
504 static inline int total_mapcount(struct page
*page
)
506 return page_mapcount(page
);
510 static inline struct page
*virt_to_head_page(const void *x
)
512 struct page
*page
= virt_to_page(x
);
514 return compound_head(page
);
517 void __put_page(struct page
*page
);
519 void put_pages_list(struct list_head
*pages
);
521 void split_page(struct page
*page
, unsigned int order
);
522 int split_free_page(struct page
*page
);
525 * Compound pages have a destructor function. Provide a
526 * prototype for that function and accessor functions.
527 * These are _only_ valid on the head of a compound page.
529 typedef void compound_page_dtor(struct page
*);
531 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
532 enum compound_dtor_id
{
535 #ifdef CONFIG_HUGETLB_PAGE
538 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
543 extern compound_page_dtor
* const compound_page_dtors
[];
545 static inline void set_compound_page_dtor(struct page
*page
,
546 enum compound_dtor_id compound_dtor
)
548 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
549 page
[1].compound_dtor
= compound_dtor
;
552 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
554 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
555 return compound_page_dtors
[page
[1].compound_dtor
];
558 static inline unsigned int compound_order(struct page
*page
)
562 return page
[1].compound_order
;
565 static inline void set_compound_order(struct page
*page
, unsigned int order
)
567 page
[1].compound_order
= order
;
570 void free_compound_page(struct page
*page
);
574 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
575 * servicing faults for write access. In the normal case, do always want
576 * pte_mkwrite. But get_user_pages can cause write faults for mappings
577 * that do not have writing enabled, when used by access_process_vm.
579 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
581 if (likely(vma
->vm_flags
& VM_WRITE
))
582 pte
= pte_mkwrite(pte
);
586 void do_set_pte(struct vm_area_struct
*vma
, unsigned long address
,
587 struct page
*page
, pte_t
*pte
, bool write
, bool anon
);
591 * Multiple processes may "see" the same page. E.g. for untouched
592 * mappings of /dev/null, all processes see the same page full of
593 * zeroes, and text pages of executables and shared libraries have
594 * only one copy in memory, at most, normally.
596 * For the non-reserved pages, page_count(page) denotes a reference count.
597 * page_count() == 0 means the page is free. page->lru is then used for
598 * freelist management in the buddy allocator.
599 * page_count() > 0 means the page has been allocated.
601 * Pages are allocated by the slab allocator in order to provide memory
602 * to kmalloc and kmem_cache_alloc. In this case, the management of the
603 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
604 * unless a particular usage is carefully commented. (the responsibility of
605 * freeing the kmalloc memory is the caller's, of course).
607 * A page may be used by anyone else who does a __get_free_page().
608 * In this case, page_count still tracks the references, and should only
609 * be used through the normal accessor functions. The top bits of page->flags
610 * and page->virtual store page management information, but all other fields
611 * are unused and could be used privately, carefully. The management of this
612 * page is the responsibility of the one who allocated it, and those who have
613 * subsequently been given references to it.
615 * The other pages (we may call them "pagecache pages") are completely
616 * managed by the Linux memory manager: I/O, buffers, swapping etc.
617 * The following discussion applies only to them.
619 * A pagecache page contains an opaque `private' member, which belongs to the
620 * page's address_space. Usually, this is the address of a circular list of
621 * the page's disk buffers. PG_private must be set to tell the VM to call
622 * into the filesystem to release these pages.
624 * A page may belong to an inode's memory mapping. In this case, page->mapping
625 * is the pointer to the inode, and page->index is the file offset of the page,
626 * in units of PAGE_SIZE.
628 * If pagecache pages are not associated with an inode, they are said to be
629 * anonymous pages. These may become associated with the swapcache, and in that
630 * case PG_swapcache is set, and page->private is an offset into the swapcache.
632 * In either case (swapcache or inode backed), the pagecache itself holds one
633 * reference to the page. Setting PG_private should also increment the
634 * refcount. The each user mapping also has a reference to the page.
636 * The pagecache pages are stored in a per-mapping radix tree, which is
637 * rooted at mapping->page_tree, and indexed by offset.
638 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
639 * lists, we instead now tag pages as dirty/writeback in the radix tree.
641 * All pagecache pages may be subject to I/O:
642 * - inode pages may need to be read from disk,
643 * - inode pages which have been modified and are MAP_SHARED may need
644 * to be written back to the inode on disk,
645 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
646 * modified may need to be swapped out to swap space and (later) to be read
651 * The zone field is never updated after free_area_init_core()
652 * sets it, so none of the operations on it need to be atomic.
655 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
656 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
657 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
658 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
659 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
662 * Define the bit shifts to access each section. For non-existent
663 * sections we define the shift as 0; that plus a 0 mask ensures
664 * the compiler will optimise away reference to them.
666 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
667 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
668 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
669 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
671 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
672 #ifdef NODE_NOT_IN_PAGE_FLAGS
673 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
674 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
675 SECTIONS_PGOFF : ZONES_PGOFF)
677 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
678 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
679 NODES_PGOFF : ZONES_PGOFF)
682 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
684 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
685 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
688 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
689 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
690 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
691 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
692 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
694 static inline enum zone_type
page_zonenum(const struct page
*page
)
696 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
699 #ifdef CONFIG_ZONE_DEVICE
700 void get_zone_device_page(struct page
*page
);
701 void put_zone_device_page(struct page
*page
);
702 static inline bool is_zone_device_page(const struct page
*page
)
704 return page_zonenum(page
) == ZONE_DEVICE
;
707 static inline void get_zone_device_page(struct page
*page
)
710 static inline void put_zone_device_page(struct page
*page
)
713 static inline bool is_zone_device_page(const struct page
*page
)
719 static inline void get_page(struct page
*page
)
721 page
= compound_head(page
);
723 * Getting a normal page or the head of a compound page
724 * requires to already have an elevated page->_count.
726 VM_BUG_ON_PAGE(page_ref_count(page
) <= 0, page
);
729 if (unlikely(is_zone_device_page(page
)))
730 get_zone_device_page(page
);
733 static inline void put_page(struct page
*page
)
735 page
= compound_head(page
);
737 if (put_page_testzero(page
))
740 if (unlikely(is_zone_device_page(page
)))
741 put_zone_device_page(page
);
744 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
745 #define SECTION_IN_PAGE_FLAGS
749 * The identification function is mainly used by the buddy allocator for
750 * determining if two pages could be buddies. We are not really identifying
751 * the zone since we could be using the section number id if we do not have
752 * node id available in page flags.
753 * We only guarantee that it will return the same value for two combinable
756 static inline int page_zone_id(struct page
*page
)
758 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
761 static inline int zone_to_nid(struct zone
*zone
)
770 #ifdef NODE_NOT_IN_PAGE_FLAGS
771 extern int page_to_nid(const struct page
*page
);
773 static inline int page_to_nid(const struct page
*page
)
775 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
779 #ifdef CONFIG_NUMA_BALANCING
780 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
782 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
785 static inline int cpupid_to_pid(int cpupid
)
787 return cpupid
& LAST__PID_MASK
;
790 static inline int cpupid_to_cpu(int cpupid
)
792 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
795 static inline int cpupid_to_nid(int cpupid
)
797 return cpu_to_node(cpupid_to_cpu(cpupid
));
800 static inline bool cpupid_pid_unset(int cpupid
)
802 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
805 static inline bool cpupid_cpu_unset(int cpupid
)
807 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
810 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
812 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
815 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
816 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
817 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
819 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
822 static inline int page_cpupid_last(struct page
*page
)
824 return page
->_last_cpupid
;
826 static inline void page_cpupid_reset_last(struct page
*page
)
828 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
831 static inline int page_cpupid_last(struct page
*page
)
833 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
836 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
838 static inline void page_cpupid_reset_last(struct page
*page
)
840 int cpupid
= (1 << LAST_CPUPID_SHIFT
) - 1;
842 page
->flags
&= ~(LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
);
843 page
->flags
|= (cpupid
& LAST_CPUPID_MASK
) << LAST_CPUPID_PGSHIFT
;
845 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
846 #else /* !CONFIG_NUMA_BALANCING */
847 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
849 return page_to_nid(page
); /* XXX */
852 static inline int page_cpupid_last(struct page
*page
)
854 return page_to_nid(page
); /* XXX */
857 static inline int cpupid_to_nid(int cpupid
)
862 static inline int cpupid_to_pid(int cpupid
)
867 static inline int cpupid_to_cpu(int cpupid
)
872 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
877 static inline bool cpupid_pid_unset(int cpupid
)
882 static inline void page_cpupid_reset_last(struct page
*page
)
886 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
890 #endif /* CONFIG_NUMA_BALANCING */
892 static inline struct zone
*page_zone(const struct page
*page
)
894 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
897 #ifdef SECTION_IN_PAGE_FLAGS
898 static inline void set_page_section(struct page
*page
, unsigned long section
)
900 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
901 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
904 static inline unsigned long page_to_section(const struct page
*page
)
906 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
910 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
912 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
913 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
916 static inline void set_page_node(struct page
*page
, unsigned long node
)
918 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
919 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
922 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
923 unsigned long node
, unsigned long pfn
)
925 set_page_zone(page
, zone
);
926 set_page_node(page
, node
);
927 #ifdef SECTION_IN_PAGE_FLAGS
928 set_page_section(page
, pfn_to_section_nr(pfn
));
933 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
935 return page
->mem_cgroup
;
938 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
945 * Some inline functions in vmstat.h depend on page_zone()
947 #include <linux/vmstat.h>
949 static __always_inline
void *lowmem_page_address(const struct page
*page
)
951 return __va(PFN_PHYS(page_to_pfn(page
)));
954 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
955 #define HASHED_PAGE_VIRTUAL
958 #if defined(WANT_PAGE_VIRTUAL)
959 static inline void *page_address(const struct page
*page
)
961 return page
->virtual;
963 static inline void set_page_address(struct page
*page
, void *address
)
965 page
->virtual = address
;
967 #define page_address_init() do { } while(0)
970 #if defined(HASHED_PAGE_VIRTUAL)
971 void *page_address(const struct page
*page
);
972 void set_page_address(struct page
*page
, void *virtual);
973 void page_address_init(void);
976 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
977 #define page_address(page) lowmem_page_address(page)
978 #define set_page_address(page, address) do { } while(0)
979 #define page_address_init() do { } while(0)
982 extern void *page_rmapping(struct page
*page
);
983 extern struct anon_vma
*page_anon_vma(struct page
*page
);
984 extern struct address_space
*page_mapping(struct page
*page
);
986 extern struct address_space
*__page_file_mapping(struct page
*);
989 struct address_space
*page_file_mapping(struct page
*page
)
991 if (unlikely(PageSwapCache(page
)))
992 return __page_file_mapping(page
);
994 return page
->mapping
;
998 * Return the pagecache index of the passed page. Regular pagecache pages
999 * use ->index whereas swapcache pages use ->private
1001 static inline pgoff_t
page_index(struct page
*page
)
1003 if (unlikely(PageSwapCache(page
)))
1004 return page_private(page
);
1008 extern pgoff_t
__page_file_index(struct page
*page
);
1011 * Return the file index of the page. Regular pagecache pages use ->index
1012 * whereas swapcache pages use swp_offset(->private)
1014 static inline pgoff_t
page_file_index(struct page
*page
)
1016 if (unlikely(PageSwapCache(page
)))
1017 return __page_file_index(page
);
1023 * Return true if this page is mapped into pagetables.
1024 * For compound page it returns true if any subpage of compound page is mapped.
1026 static inline bool page_mapped(struct page
*page
)
1029 if (likely(!PageCompound(page
)))
1030 return atomic_read(&page
->_mapcount
) >= 0;
1031 page
= compound_head(page
);
1032 if (atomic_read(compound_mapcount_ptr(page
)) >= 0)
1034 for (i
= 0; i
< hpage_nr_pages(page
); i
++) {
1035 if (atomic_read(&page
[i
]._mapcount
) >= 0)
1042 * Return true only if the page has been allocated with
1043 * ALLOC_NO_WATERMARKS and the low watermark was not
1044 * met implying that the system is under some pressure.
1046 static inline bool page_is_pfmemalloc(struct page
*page
)
1049 * Page index cannot be this large so this must be
1050 * a pfmemalloc page.
1052 return page
->index
== -1UL;
1056 * Only to be called by the page allocator on a freshly allocated
1059 static inline void set_page_pfmemalloc(struct page
*page
)
1064 static inline void clear_page_pfmemalloc(struct page
*page
)
1070 * Different kinds of faults, as returned by handle_mm_fault().
1071 * Used to decide whether a process gets delivered SIGBUS or
1072 * just gets major/minor fault counters bumped up.
1075 #define VM_FAULT_OOM 0x0001
1076 #define VM_FAULT_SIGBUS 0x0002
1077 #define VM_FAULT_MAJOR 0x0004
1078 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1079 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1080 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1081 #define VM_FAULT_SIGSEGV 0x0040
1083 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1084 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1085 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1086 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1088 #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */
1090 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1091 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1094 /* Encode hstate index for a hwpoisoned large page */
1095 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1096 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1099 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1101 extern void pagefault_out_of_memory(void);
1103 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1106 * Flags passed to show_mem() and show_free_areas() to suppress output in
1109 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1111 extern void show_free_areas(unsigned int flags
);
1112 extern bool skip_free_areas_node(unsigned int flags
, int nid
);
1114 int shmem_zero_setup(struct vm_area_struct
*);
1116 bool shmem_mapping(struct address_space
*mapping
);
1118 static inline bool shmem_mapping(struct address_space
*mapping
)
1124 extern bool can_do_mlock(void);
1125 extern int user_shm_lock(size_t, struct user_struct
*);
1126 extern void user_shm_unlock(size_t, struct user_struct
*);
1129 * Parameter block passed down to zap_pte_range in exceptional cases.
1131 struct zap_details
{
1132 struct address_space
*check_mapping
; /* Check page->mapping if set */
1133 pgoff_t first_index
; /* Lowest page->index to unmap */
1134 pgoff_t last_index
; /* Highest page->index to unmap */
1135 bool ignore_dirty
; /* Ignore dirty pages */
1136 bool check_swap_entries
; /* Check also swap entries */
1139 struct page
*vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1142 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1143 unsigned long size
);
1144 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1145 unsigned long size
, struct zap_details
*);
1146 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1147 unsigned long start
, unsigned long end
);
1150 * mm_walk - callbacks for walk_page_range
1151 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1152 * this handler is required to be able to handle
1153 * pmd_trans_huge() pmds. They may simply choose to
1154 * split_huge_page() instead of handling it explicitly.
1155 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1156 * @pte_hole: if set, called for each hole at all levels
1157 * @hugetlb_entry: if set, called for each hugetlb entry
1158 * @test_walk: caller specific callback function to determine whether
1159 * we walk over the current vma or not. A positive returned
1160 * value means "do page table walk over the current vma,"
1161 * and a negative one means "abort current page table walk
1162 * right now." 0 means "skip the current vma."
1163 * @mm: mm_struct representing the target process of page table walk
1164 * @vma: vma currently walked (NULL if walking outside vmas)
1165 * @private: private data for callbacks' usage
1167 * (see the comment on walk_page_range() for more details)
1170 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1171 unsigned long next
, struct mm_walk
*walk
);
1172 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1173 unsigned long next
, struct mm_walk
*walk
);
1174 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1175 struct mm_walk
*walk
);
1176 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1177 unsigned long addr
, unsigned long next
,
1178 struct mm_walk
*walk
);
1179 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1180 struct mm_walk
*walk
);
1181 struct mm_struct
*mm
;
1182 struct vm_area_struct
*vma
;
1186 int walk_page_range(unsigned long addr
, unsigned long end
,
1187 struct mm_walk
*walk
);
1188 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1189 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1190 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1191 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1192 struct vm_area_struct
*vma
);
1193 void unmap_mapping_range(struct address_space
*mapping
,
1194 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1195 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1196 unsigned long *pfn
);
1197 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1198 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1199 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1200 void *buf
, int len
, int write
);
1202 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1203 loff_t
const holebegin
, loff_t
const holelen
)
1205 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1208 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1209 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1210 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1211 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1212 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1213 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1214 int invalidate_inode_page(struct page
*page
);
1217 extern int handle_mm_fault(struct mm_struct
*mm
, struct vm_area_struct
*vma
,
1218 unsigned long address
, unsigned int flags
);
1219 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1220 unsigned long address
, unsigned int fault_flags
,
1223 static inline int handle_mm_fault(struct mm_struct
*mm
,
1224 struct vm_area_struct
*vma
, unsigned long address
,
1227 /* should never happen if there's no MMU */
1229 return VM_FAULT_SIGBUS
;
1231 static inline int fixup_user_fault(struct task_struct
*tsk
,
1232 struct mm_struct
*mm
, unsigned long address
,
1233 unsigned int fault_flags
, bool *unlocked
)
1235 /* should never happen if there's no MMU */
1241 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
, int write
);
1242 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1243 void *buf
, int len
, int write
);
1245 long __get_user_pages(struct task_struct
*tsk
, struct mm_struct
*mm
,
1246 unsigned long start
, unsigned long nr_pages
,
1247 unsigned int foll_flags
, struct page
**pages
,
1248 struct vm_area_struct
**vmas
, int *nonblocking
);
1249 long get_user_pages_remote(struct task_struct
*tsk
, struct mm_struct
*mm
,
1250 unsigned long start
, unsigned long nr_pages
,
1251 int write
, int force
, struct page
**pages
,
1252 struct vm_area_struct
**vmas
);
1253 long get_user_pages6(unsigned long start
, unsigned long nr_pages
,
1254 int write
, int force
, struct page
**pages
,
1255 struct vm_area_struct
**vmas
);
1256 long get_user_pages_locked6(unsigned long start
, unsigned long nr_pages
,
1257 int write
, int force
, struct page
**pages
, int *locked
);
1258 long __get_user_pages_unlocked(struct task_struct
*tsk
, struct mm_struct
*mm
,
1259 unsigned long start
, unsigned long nr_pages
,
1260 int write
, int force
, struct page
**pages
,
1261 unsigned int gup_flags
);
1262 long get_user_pages_unlocked5(unsigned long start
, unsigned long nr_pages
,
1263 int write
, int force
, struct page
**pages
);
1264 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1265 struct page
**pages
);
1267 /* suppress warnings from use in EXPORT_SYMBOL() */
1268 #ifndef __DISABLE_GUP_DEPRECATED
1269 #define __gup_deprecated __deprecated
1271 #define __gup_deprecated
1274 * These macros provide backward-compatibility with the old
1275 * get_user_pages() variants which took tsk/mm. These
1276 * functions/macros provide both compile-time __deprecated so we
1277 * can catch old-style use and not break the build. The actual
1278 * functions also have WARN_ON()s to let us know at runtime if
1279 * the get_user_pages() should have been the "remote" variant.
1281 * These are hideous, but temporary.
1283 * If you run into one of these __deprecated warnings, look
1284 * at how you are calling get_user_pages(). If you are calling
1285 * it with current/current->mm as the first two arguments,
1286 * simply remove those arguments. The behavior will be the same
1287 * as it is now. If you are calling it on another task, use
1288 * get_user_pages_remote() instead.
1290 * Any questions? Ask Dave Hansen <dave@sr71.net>
1294 get_user_pages8(struct task_struct
*tsk
, struct mm_struct
*mm
,
1295 unsigned long start
, unsigned long nr_pages
,
1296 int write
, int force
, struct page
**pages
,
1297 struct vm_area_struct
**vmas
);
1298 #define GUP_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, get_user_pages, ...) \
1300 #define get_user_pages(...) GUP_MACRO(__VA_ARGS__, \
1301 get_user_pages8, x, \
1302 get_user_pages6, x, x, x, x, x)(__VA_ARGS__)
1305 long get_user_pages_locked8(struct task_struct
*tsk
, struct mm_struct
*mm
,
1306 unsigned long start
, unsigned long nr_pages
,
1307 int write
, int force
, struct page
**pages
,
1309 #define GUPL_MACRO(_1, _2, _3, _4, _5, _6, _7, _8, get_user_pages_locked, ...) \
1310 get_user_pages_locked
1311 #define get_user_pages_locked(...) GUPL_MACRO(__VA_ARGS__, \
1312 get_user_pages_locked8, x, \
1313 get_user_pages_locked6, x, x, x, x)(__VA_ARGS__)
1316 long get_user_pages_unlocked7(struct task_struct
*tsk
, struct mm_struct
*mm
,
1317 unsigned long start
, unsigned long nr_pages
,
1318 int write
, int force
, struct page
**pages
);
1319 #define GUPU_MACRO(_1, _2, _3, _4, _5, _6, _7, get_user_pages_unlocked, ...) \
1320 get_user_pages_unlocked
1321 #define get_user_pages_unlocked(...) GUPU_MACRO(__VA_ARGS__, \
1322 get_user_pages_unlocked7, x, \
1323 get_user_pages_unlocked5, x, x, x, x)(__VA_ARGS__)
1325 /* Container for pinned pfns / pages */
1326 struct frame_vector
{
1327 unsigned int nr_allocated
; /* Number of frames we have space for */
1328 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1329 bool got_ref
; /* Did we pin pages by getting page ref? */
1330 bool is_pfns
; /* Does array contain pages or pfns? */
1331 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1332 * pfns_vector_pages() or pfns_vector_pfns()
1336 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1337 void frame_vector_destroy(struct frame_vector
*vec
);
1338 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1339 bool write
, bool force
, struct frame_vector
*vec
);
1340 void put_vaddr_frames(struct frame_vector
*vec
);
1341 int frame_vector_to_pages(struct frame_vector
*vec
);
1342 void frame_vector_to_pfns(struct frame_vector
*vec
);
1344 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1346 return vec
->nr_frames
;
1349 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1352 int err
= frame_vector_to_pages(vec
);
1355 return ERR_PTR(err
);
1357 return (struct page
**)(vec
->ptrs
);
1360 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1363 frame_vector_to_pfns(vec
);
1364 return (unsigned long *)(vec
->ptrs
);
1368 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1369 struct page
**pages
);
1370 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1371 struct page
*get_dump_page(unsigned long addr
);
1373 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1374 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1375 unsigned int length
);
1377 int __set_page_dirty_nobuffers(struct page
*page
);
1378 int __set_page_dirty_no_writeback(struct page
*page
);
1379 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1381 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1382 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1383 struct bdi_writeback
*wb
);
1384 int set_page_dirty(struct page
*page
);
1385 int set_page_dirty_lock(struct page
*page
);
1386 void cancel_dirty_page(struct page
*page
);
1387 int clear_page_dirty_for_io(struct page
*page
);
1389 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1391 /* Is the vma a continuation of the stack vma above it? */
1392 static inline int vma_growsdown(struct vm_area_struct
*vma
, unsigned long addr
)
1394 return vma
&& (vma
->vm_end
== addr
) && (vma
->vm_flags
& VM_GROWSDOWN
);
1397 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1399 return !vma
->vm_ops
;
1402 static inline int stack_guard_page_start(struct vm_area_struct
*vma
,
1405 return (vma
->vm_flags
& VM_GROWSDOWN
) &&
1406 (vma
->vm_start
== addr
) &&
1407 !vma_growsdown(vma
->vm_prev
, addr
);
1410 /* Is the vma a continuation of the stack vma below it? */
1411 static inline int vma_growsup(struct vm_area_struct
*vma
, unsigned long addr
)
1413 return vma
&& (vma
->vm_start
== addr
) && (vma
->vm_flags
& VM_GROWSUP
);
1416 static inline int stack_guard_page_end(struct vm_area_struct
*vma
,
1419 return (vma
->vm_flags
& VM_GROWSUP
) &&
1420 (vma
->vm_end
== addr
) &&
1421 !vma_growsup(vma
->vm_next
, addr
);
1424 int vma_is_stack_for_task(struct vm_area_struct
*vma
, struct task_struct
*t
);
1426 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1427 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1428 unsigned long new_addr
, unsigned long len
,
1429 bool need_rmap_locks
);
1430 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1431 unsigned long end
, pgprot_t newprot
,
1432 int dirty_accountable
, int prot_numa
);
1433 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1434 struct vm_area_struct
**pprev
, unsigned long start
,
1435 unsigned long end
, unsigned long newflags
);
1438 * doesn't attempt to fault and will return short.
1440 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1441 struct page
**pages
);
1443 * per-process(per-mm_struct) statistics.
1445 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1447 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1449 #ifdef SPLIT_RSS_COUNTING
1451 * counter is updated in asynchronous manner and may go to minus.
1452 * But it's never be expected number for users.
1457 return (unsigned long)val
;
1460 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1462 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1465 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1467 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1470 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1472 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1475 /* Optimized variant when page is already known not to be PageAnon */
1476 static inline int mm_counter_file(struct page
*page
)
1478 if (PageSwapBacked(page
))
1479 return MM_SHMEMPAGES
;
1480 return MM_FILEPAGES
;
1483 static inline int mm_counter(struct page
*page
)
1486 return MM_ANONPAGES
;
1487 return mm_counter_file(page
);
1490 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1492 return get_mm_counter(mm
, MM_FILEPAGES
) +
1493 get_mm_counter(mm
, MM_ANONPAGES
) +
1494 get_mm_counter(mm
, MM_SHMEMPAGES
);
1497 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1499 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1502 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1504 return max(mm
->hiwater_vm
, mm
->total_vm
);
1507 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1509 unsigned long _rss
= get_mm_rss(mm
);
1511 if ((mm
)->hiwater_rss
< _rss
)
1512 (mm
)->hiwater_rss
= _rss
;
1515 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1517 if (mm
->hiwater_vm
< mm
->total_vm
)
1518 mm
->hiwater_vm
= mm
->total_vm
;
1521 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1523 mm
->hiwater_rss
= get_mm_rss(mm
);
1526 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1527 struct mm_struct
*mm
)
1529 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1531 if (*maxrss
< hiwater_rss
)
1532 *maxrss
= hiwater_rss
;
1535 #if defined(SPLIT_RSS_COUNTING)
1536 void sync_mm_rss(struct mm_struct
*mm
);
1538 static inline void sync_mm_rss(struct mm_struct
*mm
)
1543 #ifndef __HAVE_ARCH_PTE_DEVMAP
1544 static inline int pte_devmap(pte_t pte
)
1550 int vma_wants_writenotify(struct vm_area_struct
*vma
);
1552 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1554 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1558 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1562 #ifdef __PAGETABLE_PUD_FOLDED
1563 static inline int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1564 unsigned long address
)
1569 int __pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1572 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1573 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1574 unsigned long address
)
1579 static inline void mm_nr_pmds_init(struct mm_struct
*mm
) {}
1581 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1586 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1587 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1590 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1592 static inline void mm_nr_pmds_init(struct mm_struct
*mm
)
1594 atomic_long_set(&mm
->nr_pmds
, 0);
1597 static inline unsigned long mm_nr_pmds(struct mm_struct
*mm
)
1599 return atomic_long_read(&mm
->nr_pmds
);
1602 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1604 atomic_long_inc(&mm
->nr_pmds
);
1607 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1609 atomic_long_dec(&mm
->nr_pmds
);
1613 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
1614 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1617 * The following ifdef needed to get the 4level-fixup.h header to work.
1618 * Remove it when 4level-fixup.h has been removed.
1620 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1621 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
)
1623 return (unlikely(pgd_none(*pgd
)) && __pud_alloc(mm
, pgd
, address
))?
1624 NULL
: pud_offset(pgd
, address
);
1627 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1629 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1630 NULL
: pmd_offset(pud
, address
);
1632 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1634 #if USE_SPLIT_PTE_PTLOCKS
1635 #if ALLOC_SPLIT_PTLOCKS
1636 void __init
ptlock_cache_init(void);
1637 extern bool ptlock_alloc(struct page
*page
);
1638 extern void ptlock_free(struct page
*page
);
1640 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1644 #else /* ALLOC_SPLIT_PTLOCKS */
1645 static inline void ptlock_cache_init(void)
1649 static inline bool ptlock_alloc(struct page
*page
)
1654 static inline void ptlock_free(struct page
*page
)
1658 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1662 #endif /* ALLOC_SPLIT_PTLOCKS */
1664 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1666 return ptlock_ptr(pmd_page(*pmd
));
1669 static inline bool ptlock_init(struct page
*page
)
1672 * prep_new_page() initialize page->private (and therefore page->ptl)
1673 * with 0. Make sure nobody took it in use in between.
1675 * It can happen if arch try to use slab for page table allocation:
1676 * slab code uses page->slab_cache, which share storage with page->ptl.
1678 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1679 if (!ptlock_alloc(page
))
1681 spin_lock_init(ptlock_ptr(page
));
1685 /* Reset page->mapping so free_pages_check won't complain. */
1686 static inline void pte_lock_deinit(struct page
*page
)
1688 page
->mapping
= NULL
;
1692 #else /* !USE_SPLIT_PTE_PTLOCKS */
1694 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1696 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1698 return &mm
->page_table_lock
;
1700 static inline void ptlock_cache_init(void) {}
1701 static inline bool ptlock_init(struct page
*page
) { return true; }
1702 static inline void pte_lock_deinit(struct page
*page
) {}
1703 #endif /* USE_SPLIT_PTE_PTLOCKS */
1705 static inline void pgtable_init(void)
1707 ptlock_cache_init();
1708 pgtable_cache_init();
1711 static inline bool pgtable_page_ctor(struct page
*page
)
1713 if (!ptlock_init(page
))
1715 inc_zone_page_state(page
, NR_PAGETABLE
);
1719 static inline void pgtable_page_dtor(struct page
*page
)
1721 pte_lock_deinit(page
);
1722 dec_zone_page_state(page
, NR_PAGETABLE
);
1725 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1727 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1728 pte_t *__pte = pte_offset_map(pmd, address); \
1734 #define pte_unmap_unlock(pte, ptl) do { \
1739 #define pte_alloc(mm, pmd, address) \
1740 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1742 #define pte_alloc_map(mm, pmd, address) \
1743 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1745 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1746 (pte_alloc(mm, pmd, address) ? \
1747 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1749 #define pte_alloc_kernel(pmd, address) \
1750 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1751 NULL: pte_offset_kernel(pmd, address))
1753 #if USE_SPLIT_PMD_PTLOCKS
1755 static struct page
*pmd_to_page(pmd_t
*pmd
)
1757 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1758 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1761 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1763 return ptlock_ptr(pmd_to_page(pmd
));
1766 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1768 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1769 page
->pmd_huge_pte
= NULL
;
1771 return ptlock_init(page
);
1774 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1776 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1777 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1782 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1786 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1788 return &mm
->page_table_lock
;
1791 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1792 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1794 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1798 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1800 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1805 extern void free_area_init(unsigned long * zones_size
);
1806 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1807 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1808 extern void free_initmem(void);
1811 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1812 * into the buddy system. The freed pages will be poisoned with pattern
1813 * "poison" if it's within range [0, UCHAR_MAX].
1814 * Return pages freed into the buddy system.
1816 extern unsigned long free_reserved_area(void *start
, void *end
,
1817 int poison
, char *s
);
1819 #ifdef CONFIG_HIGHMEM
1821 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1822 * and totalram_pages.
1824 extern void free_highmem_page(struct page
*page
);
1827 extern void adjust_managed_page_count(struct page
*page
, long count
);
1828 extern void mem_init_print_info(const char *str
);
1830 extern void reserve_bootmem_region(unsigned long start
, unsigned long end
);
1832 /* Free the reserved page into the buddy system, so it gets managed. */
1833 static inline void __free_reserved_page(struct page
*page
)
1835 ClearPageReserved(page
);
1836 init_page_count(page
);
1840 static inline void free_reserved_page(struct page
*page
)
1842 __free_reserved_page(page
);
1843 adjust_managed_page_count(page
, 1);
1846 static inline void mark_page_reserved(struct page
*page
)
1848 SetPageReserved(page
);
1849 adjust_managed_page_count(page
, -1);
1853 * Default method to free all the __init memory into the buddy system.
1854 * The freed pages will be poisoned with pattern "poison" if it's within
1855 * range [0, UCHAR_MAX].
1856 * Return pages freed into the buddy system.
1858 static inline unsigned long free_initmem_default(int poison
)
1860 extern char __init_begin
[], __init_end
[];
1862 return free_reserved_area(&__init_begin
, &__init_end
,
1863 poison
, "unused kernel");
1866 static inline unsigned long get_num_physpages(void)
1869 unsigned long phys_pages
= 0;
1871 for_each_online_node(nid
)
1872 phys_pages
+= node_present_pages(nid
);
1877 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
1879 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
1880 * zones, allocate the backing mem_map and account for memory holes in a more
1881 * architecture independent manner. This is a substitute for creating the
1882 * zone_sizes[] and zholes_size[] arrays and passing them to
1883 * free_area_init_node()
1885 * An architecture is expected to register range of page frames backed by
1886 * physical memory with memblock_add[_node]() before calling
1887 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
1888 * usage, an architecture is expected to do something like
1890 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
1892 * for_each_valid_physical_page_range()
1893 * memblock_add_node(base, size, nid)
1894 * free_area_init_nodes(max_zone_pfns);
1896 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
1897 * registered physical page range. Similarly
1898 * sparse_memory_present_with_active_regions() calls memory_present() for
1899 * each range when SPARSEMEM is enabled.
1901 * See mm/page_alloc.c for more information on each function exposed by
1902 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
1904 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
1905 unsigned long node_map_pfn_alignment(void);
1906 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
1907 unsigned long end_pfn
);
1908 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
1909 unsigned long end_pfn
);
1910 extern void get_pfn_range_for_nid(unsigned int nid
,
1911 unsigned long *start_pfn
, unsigned long *end_pfn
);
1912 extern unsigned long find_min_pfn_with_active_regions(void);
1913 extern void free_bootmem_with_active_regions(int nid
,
1914 unsigned long max_low_pfn
);
1915 extern void sparse_memory_present_with_active_regions(int nid
);
1917 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
1919 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
1920 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
1921 static inline int __early_pfn_to_nid(unsigned long pfn
,
1922 struct mminit_pfnnid_cache
*state
)
1927 /* please see mm/page_alloc.c */
1928 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
1929 /* there is a per-arch backend function. */
1930 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
1931 struct mminit_pfnnid_cache
*state
);
1934 extern void set_dma_reserve(unsigned long new_dma_reserve
);
1935 extern void memmap_init_zone(unsigned long, int, unsigned long,
1936 unsigned long, enum memmap_context
);
1937 extern void setup_per_zone_wmarks(void);
1938 extern int __meminit
init_per_zone_wmark_min(void);
1939 extern void mem_init(void);
1940 extern void __init
mmap_init(void);
1941 extern void show_mem(unsigned int flags
);
1942 extern long si_mem_available(void);
1943 extern void si_meminfo(struct sysinfo
* val
);
1944 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
1946 extern __printf(3, 4)
1947 void warn_alloc_failed(gfp_t gfp_mask
, unsigned int order
,
1948 const char *fmt
, ...);
1950 extern void setup_per_cpu_pageset(void);
1952 extern void zone_pcp_update(struct zone
*zone
);
1953 extern void zone_pcp_reset(struct zone
*zone
);
1956 extern int min_free_kbytes
;
1957 extern int watermark_scale_factor
;
1960 extern atomic_long_t mmap_pages_allocated
;
1961 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
1963 /* interval_tree.c */
1964 void vma_interval_tree_insert(struct vm_area_struct
*node
,
1965 struct rb_root
*root
);
1966 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
1967 struct vm_area_struct
*prev
,
1968 struct rb_root
*root
);
1969 void vma_interval_tree_remove(struct vm_area_struct
*node
,
1970 struct rb_root
*root
);
1971 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root
*root
,
1972 unsigned long start
, unsigned long last
);
1973 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
1974 unsigned long start
, unsigned long last
);
1976 #define vma_interval_tree_foreach(vma, root, start, last) \
1977 for (vma = vma_interval_tree_iter_first(root, start, last); \
1978 vma; vma = vma_interval_tree_iter_next(vma, start, last))
1980 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
1981 struct rb_root
*root
);
1982 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
1983 struct rb_root
*root
);
1984 struct anon_vma_chain
*anon_vma_interval_tree_iter_first(
1985 struct rb_root
*root
, unsigned long start
, unsigned long last
);
1986 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
1987 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
1988 #ifdef CONFIG_DEBUG_VM_RB
1989 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
1992 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
1993 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
1994 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
1997 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
1998 extern int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
1999 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
);
2000 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
2001 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
2002 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
2003 struct mempolicy
*, struct vm_userfaultfd_ctx
);
2004 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
2005 extern int split_vma(struct mm_struct
*,
2006 struct vm_area_struct
*, unsigned long addr
, int new_below
);
2007 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
2008 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
2009 struct rb_node
**, struct rb_node
*);
2010 extern void unlink_file_vma(struct vm_area_struct
*);
2011 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
2012 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
2013 bool *need_rmap_locks
);
2014 extern void exit_mmap(struct mm_struct
*);
2016 static inline int check_data_rlimit(unsigned long rlim
,
2018 unsigned long start
,
2019 unsigned long end_data
,
2020 unsigned long start_data
)
2022 if (rlim
< RLIM_INFINITY
) {
2023 if (((new - start
) + (end_data
- start_data
)) > rlim
)
2030 extern int mm_take_all_locks(struct mm_struct
*mm
);
2031 extern void mm_drop_all_locks(struct mm_struct
*mm
);
2033 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2034 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2036 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2037 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2039 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2040 unsigned long addr
, unsigned long len
,
2041 unsigned long flags
,
2042 const struct vm_special_mapping
*spec
);
2043 /* This is an obsolete alternative to _install_special_mapping. */
2044 extern int install_special_mapping(struct mm_struct
*mm
,
2045 unsigned long addr
, unsigned long len
,
2046 unsigned long flags
, struct page
**pages
);
2048 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
2050 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
2051 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
);
2052 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
2053 unsigned long len
, unsigned long prot
, unsigned long flags
,
2054 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
);
2055 extern int do_munmap(struct mm_struct
*, unsigned long, size_t);
2057 static inline unsigned long
2058 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
2059 unsigned long len
, unsigned long prot
, unsigned long flags
,
2060 unsigned long pgoff
, unsigned long *populate
)
2062 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
);
2066 extern int __mm_populate(unsigned long addr
, unsigned long len
,
2068 static inline void mm_populate(unsigned long addr
, unsigned long len
)
2071 (void) __mm_populate(addr
, len
, 1);
2074 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
2077 /* These take the mm semaphore themselves */
2078 extern unsigned long vm_brk(unsigned long, unsigned long);
2079 extern int vm_munmap(unsigned long, size_t);
2080 extern unsigned long vm_mmap(struct file
*, unsigned long,
2081 unsigned long, unsigned long,
2082 unsigned long, unsigned long);
2084 struct vm_unmapped_area_info
{
2085 #define VM_UNMAPPED_AREA_TOPDOWN 1
2086 unsigned long flags
;
2087 unsigned long length
;
2088 unsigned long low_limit
;
2089 unsigned long high_limit
;
2090 unsigned long align_mask
;
2091 unsigned long align_offset
;
2094 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
2095 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
2098 * Search for an unmapped address range.
2100 * We are looking for a range that:
2101 * - does not intersect with any VMA;
2102 * - is contained within the [low_limit, high_limit) interval;
2103 * - is at least the desired size.
2104 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2106 static inline unsigned long
2107 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
2109 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
2110 return unmapped_area_topdown(info
);
2112 return unmapped_area(info
);
2116 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2117 extern void truncate_inode_pages_range(struct address_space
*,
2118 loff_t lstart
, loff_t lend
);
2119 extern void truncate_inode_pages_final(struct address_space
*);
2121 /* generic vm_area_ops exported for stackable file systems */
2122 extern int filemap_fault(struct vm_area_struct
*, struct vm_fault
*);
2123 extern void filemap_map_pages(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
2124 extern int filemap_page_mkwrite(struct vm_area_struct
*vma
, struct vm_fault
*vmf
);
2126 /* mm/page-writeback.c */
2127 int write_one_page(struct page
*page
, int wait
);
2128 void task_dirty_inc(struct task_struct
*tsk
);
2131 #define VM_MAX_READAHEAD 128 /* kbytes */
2132 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2134 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2135 pgoff_t offset
, unsigned long nr_to_read
);
2137 void page_cache_sync_readahead(struct address_space
*mapping
,
2138 struct file_ra_state
*ra
,
2141 unsigned long size
);
2143 void page_cache_async_readahead(struct address_space
*mapping
,
2144 struct file_ra_state
*ra
,
2148 unsigned long size
);
2150 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2151 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2153 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2154 extern int expand_downwards(struct vm_area_struct
*vma
,
2155 unsigned long address
);
2157 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2159 #define expand_upwards(vma, address) (0)
2162 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2163 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2164 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2165 struct vm_area_struct
**pprev
);
2167 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2168 NULL if none. Assume start_addr < end_addr. */
2169 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2171 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2173 if (vma
&& end_addr
<= vma
->vm_start
)
2178 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2180 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2183 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2184 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2185 unsigned long vm_start
, unsigned long vm_end
)
2187 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2189 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2196 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2197 void vma_set_page_prot(struct vm_area_struct
*vma
);
2199 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2203 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2205 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2209 #ifdef CONFIG_NUMA_BALANCING
2210 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2211 unsigned long start
, unsigned long end
);
2214 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2215 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2216 unsigned long pfn
, unsigned long size
, pgprot_t
);
2217 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2218 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2220 int vm_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2221 unsigned long pfn
, pgprot_t pgprot
);
2222 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2224 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2227 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2228 unsigned long address
, unsigned int foll_flags
,
2229 unsigned int *page_mask
);
2231 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2232 unsigned long address
, unsigned int foll_flags
)
2234 unsigned int unused_page_mask
;
2235 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2238 #define FOLL_WRITE 0x01 /* check pte is writable */
2239 #define FOLL_TOUCH 0x02 /* mark page accessed */
2240 #define FOLL_GET 0x04 /* do get_page on page */
2241 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2242 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2243 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2244 * and return without waiting upon it */
2245 #define FOLL_POPULATE 0x40 /* fault in page */
2246 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2247 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2248 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2249 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2250 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2251 #define FOLL_MLOCK 0x1000 /* lock present pages */
2252 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2254 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2256 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2257 unsigned long size
, pte_fn_t fn
, void *data
);
2260 #ifdef CONFIG_PAGE_POISONING
2261 extern bool page_poisoning_enabled(void);
2262 extern void kernel_poison_pages(struct page
*page
, int numpages
, int enable
);
2263 extern bool page_is_poisoned(struct page
*page
);
2265 static inline bool page_poisoning_enabled(void) { return false; }
2266 static inline void kernel_poison_pages(struct page
*page
, int numpages
,
2268 static inline bool page_is_poisoned(struct page
*page
) { return false; }
2271 #ifdef CONFIG_DEBUG_PAGEALLOC
2272 extern bool _debug_pagealloc_enabled
;
2273 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2275 static inline bool debug_pagealloc_enabled(void)
2277 return _debug_pagealloc_enabled
;
2281 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2283 if (!debug_pagealloc_enabled())
2286 __kernel_map_pages(page
, numpages
, enable
);
2288 #ifdef CONFIG_HIBERNATION
2289 extern bool kernel_page_present(struct page
*page
);
2290 #endif /* CONFIG_HIBERNATION */
2291 #else /* CONFIG_DEBUG_PAGEALLOC */
2293 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2294 #ifdef CONFIG_HIBERNATION
2295 static inline bool kernel_page_present(struct page
*page
) { return true; }
2296 #endif /* CONFIG_HIBERNATION */
2297 static inline bool debug_pagealloc_enabled(void)
2301 #endif /* CONFIG_DEBUG_PAGEALLOC */
2303 #ifdef __HAVE_ARCH_GATE_AREA
2304 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2305 extern int in_gate_area_no_mm(unsigned long addr
);
2306 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2308 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2312 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2313 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2317 #endif /* __HAVE_ARCH_GATE_AREA */
2319 #ifdef CONFIG_SYSCTL
2320 extern int sysctl_drop_caches
;
2321 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2322 void __user
*, size_t *, loff_t
*);
2325 void drop_slab(void);
2326 void drop_slab_node(int nid
);
2329 #define randomize_va_space 0
2331 extern int randomize_va_space
;
2334 const char * arch_vma_name(struct vm_area_struct
*vma
);
2335 void print_vma_addr(char *prefix
, unsigned long rip
);
2337 void sparse_mem_maps_populate_node(struct page
**map_map
,
2338 unsigned long pnum_begin
,
2339 unsigned long pnum_end
,
2340 unsigned long map_count
,
2343 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2344 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2345 pud_t
*vmemmap_pud_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2346 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2347 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2348 void *vmemmap_alloc_block(unsigned long size
, int node
);
2350 void *__vmemmap_alloc_block_buf(unsigned long size
, int node
,
2351 struct vmem_altmap
*altmap
);
2352 static inline void *vmemmap_alloc_block_buf(unsigned long size
, int node
)
2354 return __vmemmap_alloc_block_buf(size
, node
, NULL
);
2357 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2358 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2360 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2361 void vmemmap_populate_print_last(void);
2362 #ifdef CONFIG_MEMORY_HOTPLUG
2363 void vmemmap_free(unsigned long start
, unsigned long end
);
2365 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2366 unsigned long size
);
2369 MF_COUNT_INCREASED
= 1 << 0,
2370 MF_ACTION_REQUIRED
= 1 << 1,
2371 MF_MUST_KILL
= 1 << 2,
2372 MF_SOFT_OFFLINE
= 1 << 3,
2374 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2375 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2376 extern int unpoison_memory(unsigned long pfn
);
2377 extern int get_hwpoison_page(struct page
*page
);
2378 #define put_hwpoison_page(page) put_page(page)
2379 extern int sysctl_memory_failure_early_kill
;
2380 extern int sysctl_memory_failure_recovery
;
2381 extern void shake_page(struct page
*p
, int access
);
2382 extern atomic_long_t num_poisoned_pages
;
2383 extern int soft_offline_page(struct page
*page
, int flags
);
2387 * Error handlers for various types of pages.
2390 MF_IGNORED
, /* Error: cannot be handled */
2391 MF_FAILED
, /* Error: handling failed */
2392 MF_DELAYED
, /* Will be handled later */
2393 MF_RECOVERED
, /* Successfully recovered */
2396 enum mf_action_page_type
{
2398 MF_MSG_KERNEL_HIGH_ORDER
,
2400 MF_MSG_DIFFERENT_COMPOUND
,
2401 MF_MSG_POISONED_HUGE
,
2404 MF_MSG_UNMAP_FAILED
,
2405 MF_MSG_DIRTY_SWAPCACHE
,
2406 MF_MSG_CLEAN_SWAPCACHE
,
2407 MF_MSG_DIRTY_MLOCKED_LRU
,
2408 MF_MSG_CLEAN_MLOCKED_LRU
,
2409 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2410 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2413 MF_MSG_TRUNCATED_LRU
,
2419 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2420 extern void clear_huge_page(struct page
*page
,
2422 unsigned int pages_per_huge_page
);
2423 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2424 unsigned long addr
, struct vm_area_struct
*vma
,
2425 unsigned int pages_per_huge_page
);
2426 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2428 extern struct page_ext_operations debug_guardpage_ops
;
2429 extern struct page_ext_operations page_poisoning_ops
;
2431 #ifdef CONFIG_DEBUG_PAGEALLOC
2432 extern unsigned int _debug_guardpage_minorder
;
2433 extern bool _debug_guardpage_enabled
;
2435 static inline unsigned int debug_guardpage_minorder(void)
2437 return _debug_guardpage_minorder
;
2440 static inline bool debug_guardpage_enabled(void)
2442 return _debug_guardpage_enabled
;
2445 static inline bool page_is_guard(struct page
*page
)
2447 struct page_ext
*page_ext
;
2449 if (!debug_guardpage_enabled())
2452 page_ext
= lookup_page_ext(page
);
2453 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2456 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2457 static inline bool debug_guardpage_enabled(void) { return false; }
2458 static inline bool page_is_guard(struct page
*page
) { return false; }
2459 #endif /* CONFIG_DEBUG_PAGEALLOC */
2461 #if MAX_NUMNODES > 1
2462 void __init
setup_nr_node_ids(void);
2464 static inline void setup_nr_node_ids(void) {}
2467 #endif /* __KERNEL__ */
2468 #endif /* _LINUX_MM_H */