1 /* SPDX-License-Identifier: GPL-2.0 */
5 #include <linux/errno.h>
9 #include <linux/mmdebug.h>
10 #include <linux/gfp.h>
11 #include <linux/bug.h>
12 #include <linux/list.h>
13 #include <linux/mmzone.h>
14 #include <linux/rbtree.h>
15 #include <linux/atomic.h>
16 #include <linux/debug_locks.h>
17 #include <linux/mm_types.h>
18 #include <linux/range.h>
19 #include <linux/pfn.h>
20 #include <linux/percpu-refcount.h>
21 #include <linux/bit_spinlock.h>
22 #include <linux/shrinker.h>
23 #include <linux/resource.h>
24 #include <linux/page_ext.h>
25 #include <linux/err.h>
26 #include <linux/page_ref.h>
27 #include <linux/memremap.h>
31 struct anon_vma_chain
;
34 struct writeback_control
;
37 void init_mm_internals(void);
39 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
40 extern unsigned long max_mapnr
;
42 static inline void set_max_mapnr(unsigned long limit
)
47 static inline void set_max_mapnr(unsigned long limit
) { }
50 extern unsigned long totalram_pages
;
51 extern void * high_memory
;
52 extern int page_cluster
;
55 extern int sysctl_legacy_va_layout
;
57 #define sysctl_legacy_va_layout 0
60 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
61 extern const int mmap_rnd_bits_min
;
62 extern const int mmap_rnd_bits_max
;
63 extern int mmap_rnd_bits __read_mostly
;
65 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
66 extern const int mmap_rnd_compat_bits_min
;
67 extern const int mmap_rnd_compat_bits_max
;
68 extern int mmap_rnd_compat_bits __read_mostly
;
72 #include <asm/pgtable.h>
73 #include <asm/processor.h>
76 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
80 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
84 #define lm_alias(x) __va(__pa_symbol(x))
88 * To prevent common memory management code establishing
89 * a zero page mapping on a read fault.
90 * This macro should be defined within <asm/pgtable.h>.
91 * s390 does this to prevent multiplexing of hardware bits
92 * related to the physical page in case of virtualization.
94 #ifndef mm_forbids_zeropage
95 #define mm_forbids_zeropage(X) (0)
99 * On some architectures it is expensive to call memset() for small sizes.
100 * Those architectures should provide their own implementation of "struct page"
101 * zeroing by defining this macro in <asm/pgtable.h>.
103 #ifndef mm_zero_struct_page
104 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
108 * Default maximum number of active map areas, this limits the number of vmas
109 * per mm struct. Users can overwrite this number by sysctl but there is a
112 * When a program's coredump is generated as ELF format, a section is created
113 * per a vma. In ELF, the number of sections is represented in unsigned short.
114 * This means the number of sections should be smaller than 65535 at coredump.
115 * Because the kernel adds some informative sections to a image of program at
116 * generating coredump, we need some margin. The number of extra sections is
117 * 1-3 now and depends on arch. We use "5" as safe margin, here.
119 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
120 * not a hard limit any more. Although some userspace tools can be surprised by
123 #define MAPCOUNT_ELF_CORE_MARGIN (5)
124 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
126 extern int sysctl_max_map_count
;
128 extern unsigned long sysctl_user_reserve_kbytes
;
129 extern unsigned long sysctl_admin_reserve_kbytes
;
131 extern int sysctl_overcommit_memory
;
132 extern int sysctl_overcommit_ratio
;
133 extern unsigned long sysctl_overcommit_kbytes
;
135 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
137 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
140 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
142 /* to align the pointer to the (next) page boundary */
143 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
145 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
146 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
149 * Linux kernel virtual memory manager primitives.
150 * The idea being to have a "virtual" mm in the same way
151 * we have a virtual fs - giving a cleaner interface to the
152 * mm details, and allowing different kinds of memory mappings
153 * (from shared memory to executable loading to arbitrary
157 struct vm_area_struct
*vm_area_alloc(struct mm_struct
*);
158 struct vm_area_struct
*vm_area_dup(struct vm_area_struct
*);
159 void vm_area_free(struct vm_area_struct
*);
162 extern struct rb_root nommu_region_tree
;
163 extern struct rw_semaphore nommu_region_sem
;
165 extern unsigned int kobjsize(const void *objp
);
169 * vm_flags in vm_area_struct, see mm_types.h.
170 * When changing, update also include/trace/events/mmflags.h
172 #define VM_NONE 0x00000000
174 #define VM_READ 0x00000001 /* currently active flags */
175 #define VM_WRITE 0x00000002
176 #define VM_EXEC 0x00000004
177 #define VM_SHARED 0x00000008
179 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
180 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
181 #define VM_MAYWRITE 0x00000020
182 #define VM_MAYEXEC 0x00000040
183 #define VM_MAYSHARE 0x00000080
185 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
186 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
187 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
188 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
189 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
191 #define VM_LOCKED 0x00002000
192 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
194 /* Used by sys_madvise() */
195 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
196 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
198 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
199 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
200 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
201 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
202 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
203 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
204 #define VM_SYNC 0x00800000 /* Synchronous page faults */
205 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
206 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
207 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
209 #ifdef CONFIG_MEM_SOFT_DIRTY
210 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
212 # define VM_SOFTDIRTY 0
215 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
216 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
217 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
218 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
220 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
221 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
222 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
223 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
224 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
225 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
226 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
227 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
228 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
229 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
230 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
231 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
233 #if defined(CONFIG_X86)
234 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
235 #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS)
236 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
237 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
238 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1
239 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
240 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
242 #elif defined(CONFIG_PPC)
243 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
244 #elif defined(CONFIG_PARISC)
245 # define VM_GROWSUP VM_ARCH_1
246 #elif defined(CONFIG_METAG)
247 # define VM_GROWSUP VM_ARCH_1
248 #elif defined(CONFIG_IA64)
249 # define VM_GROWSUP VM_ARCH_1
250 #elif !defined(CONFIG_MMU)
251 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
254 #if defined(CONFIG_X86_INTEL_MPX)
255 /* MPX specific bounds table or bounds directory */
256 # define VM_MPX VM_HIGH_ARCH_4
258 # define VM_MPX VM_NONE
262 # define VM_GROWSUP VM_NONE
265 /* Bits set in the VMA until the stack is in its final location */
266 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
268 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
269 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
272 #ifdef CONFIG_STACK_GROWSUP
273 #define VM_STACK VM_GROWSUP
275 #define VM_STACK VM_GROWSDOWN
278 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
281 * Special vmas that are non-mergable, non-mlock()able.
282 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
284 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
286 /* This mask defines which mm->def_flags a process can inherit its parent */
287 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
289 /* This mask is used to clear all the VMA flags used by mlock */
290 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
293 * mapping from the currently active vm_flags protection bits (the
294 * low four bits) to a page protection mask..
296 extern pgprot_t protection_map
[16];
298 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
299 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
300 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
301 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
302 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
303 #define FAULT_FLAG_TRIED 0x20 /* Second try */
304 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
305 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
306 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
308 #define FAULT_FLAG_TRACE \
309 { FAULT_FLAG_WRITE, "WRITE" }, \
310 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
311 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
312 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
313 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
314 { FAULT_FLAG_TRIED, "TRIED" }, \
315 { FAULT_FLAG_USER, "USER" }, \
316 { FAULT_FLAG_REMOTE, "REMOTE" }, \
317 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
320 * vm_fault is filled by the the pagefault handler and passed to the vma's
321 * ->fault function. The vma's ->fault is responsible for returning a bitmask
322 * of VM_FAULT_xxx flags that give details about how the fault was handled.
324 * MM layer fills up gfp_mask for page allocations but fault handler might
325 * alter it if its implementation requires a different allocation context.
327 * pgoff should be used in favour of virtual_address, if possible.
330 struct vm_area_struct
*vma
; /* Target VMA */
331 unsigned int flags
; /* FAULT_FLAG_xxx flags */
332 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
333 pgoff_t pgoff
; /* Logical page offset based on vma */
334 unsigned long address
; /* Faulting virtual address */
335 pmd_t
*pmd
; /* Pointer to pmd entry matching
337 pud_t
*pud
; /* Pointer to pud entry matching
340 pte_t orig_pte
; /* Value of PTE at the time of fault */
342 struct page
*cow_page
; /* Page handler may use for COW fault */
343 struct mem_cgroup
*memcg
; /* Cgroup cow_page belongs to */
344 struct page
*page
; /* ->fault handlers should return a
345 * page here, unless VM_FAULT_NOPAGE
346 * is set (which is also implied by
349 /* These three entries are valid only while holding ptl lock */
350 pte_t
*pte
; /* Pointer to pte entry matching
351 * the 'address'. NULL if the page
352 * table hasn't been allocated.
354 spinlock_t
*ptl
; /* Page table lock.
355 * Protects pte page table if 'pte'
356 * is not NULL, otherwise pmd.
358 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
359 * vm_ops->map_pages() calls
360 * alloc_set_pte() from atomic context.
361 * do_fault_around() pre-allocates
362 * page table to avoid allocation from
367 /* page entry size for vm->huge_fault() */
368 enum page_entry_size
{
375 * These are the virtual MM functions - opening of an area, closing and
376 * unmapping it (needed to keep files on disk up-to-date etc), pointer
377 * to the functions called when a no-page or a wp-page exception occurs.
379 struct vm_operations_struct
{
380 void (*open
)(struct vm_area_struct
* area
);
381 void (*close
)(struct vm_area_struct
* area
);
382 int (*split
)(struct vm_area_struct
* area
, unsigned long addr
);
383 int (*mremap
)(struct vm_area_struct
* area
);
384 int (*fault
)(struct vm_fault
*vmf
);
385 int (*huge_fault
)(struct vm_fault
*vmf
, enum page_entry_size pe_size
);
386 void (*map_pages
)(struct vm_fault
*vmf
,
387 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
389 /* notification that a previously read-only page is about to become
390 * writable, if an error is returned it will cause a SIGBUS */
391 int (*page_mkwrite
)(struct vm_fault
*vmf
);
393 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
394 int (*pfn_mkwrite
)(struct vm_fault
*vmf
);
396 /* called by access_process_vm when get_user_pages() fails, typically
397 * for use by special VMAs that can switch between memory and hardware
399 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
400 void *buf
, int len
, int write
);
402 /* Called by the /proc/PID/maps code to ask the vma whether it
403 * has a special name. Returning non-NULL will also cause this
404 * vma to be dumped unconditionally. */
405 const char *(*name
)(struct vm_area_struct
*vma
);
409 * set_policy() op must add a reference to any non-NULL @new mempolicy
410 * to hold the policy upon return. Caller should pass NULL @new to
411 * remove a policy and fall back to surrounding context--i.e. do not
412 * install a MPOL_DEFAULT policy, nor the task or system default
415 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
418 * get_policy() op must add reference [mpol_get()] to any policy at
419 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
420 * in mm/mempolicy.c will do this automatically.
421 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
422 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
423 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
424 * must return NULL--i.e., do not "fallback" to task or system default
427 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
431 * Called by vm_normal_page() for special PTEs to find the
432 * page for @addr. This is useful if the default behavior
433 * (using pte_page()) would not find the correct page.
435 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
442 #define page_private(page) ((page)->private)
443 #define set_page_private(page, v) ((page)->private = (v))
445 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
446 static inline int pmd_devmap(pmd_t pmd
)
450 static inline int pud_devmap(pud_t pud
)
454 static inline int pgd_devmap(pgd_t pgd
)
461 * FIXME: take this include out, include page-flags.h in
462 * files which need it (119 of them)
464 #include <linux/page-flags.h>
465 #include <linux/huge_mm.h>
468 * Methods to modify the page usage count.
470 * What counts for a page usage:
471 * - cache mapping (page->mapping)
472 * - private data (page->private)
473 * - page mapped in a task's page tables, each mapping
474 * is counted separately
476 * Also, many kernel routines increase the page count before a critical
477 * routine so they can be sure the page doesn't go away from under them.
481 * Drop a ref, return true if the refcount fell to zero (the page has no users)
483 static inline int put_page_testzero(struct page
*page
)
485 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
486 return page_ref_dec_and_test(page
);
490 * Try to grab a ref unless the page has a refcount of zero, return false if
492 * This can be called when MMU is off so it must not access
493 * any of the virtual mappings.
495 static inline int get_page_unless_zero(struct page
*page
)
497 return page_ref_add_unless(page
, 1, 0);
500 extern int page_is_ram(unsigned long pfn
);
508 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
511 /* Support for virtually mapped pages */
512 struct page
*vmalloc_to_page(const void *addr
);
513 unsigned long vmalloc_to_pfn(const void *addr
);
516 * Determine if an address is within the vmalloc range
518 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
519 * is no special casing required.
521 static inline bool is_vmalloc_addr(const void *x
)
524 unsigned long addr
= (unsigned long)x
;
526 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
532 extern int is_vmalloc_or_module_addr(const void *x
);
534 static inline int is_vmalloc_or_module_addr(const void *x
)
540 extern void *kvmalloc_node(size_t size
, gfp_t flags
, int node
);
541 static inline void *kvmalloc(size_t size
, gfp_t flags
)
543 return kvmalloc_node(size
, flags
, NUMA_NO_NODE
);
545 static inline void *kvzalloc_node(size_t size
, gfp_t flags
, int node
)
547 return kvmalloc_node(size
, flags
| __GFP_ZERO
, node
);
549 static inline void *kvzalloc(size_t size
, gfp_t flags
)
551 return kvmalloc(size
, flags
| __GFP_ZERO
);
554 static inline void *kvmalloc_array(size_t n
, size_t size
, gfp_t flags
)
556 if (size
!= 0 && n
> SIZE_MAX
/ size
)
559 return kvmalloc(n
* size
, flags
);
562 extern void kvfree(const void *addr
);
564 static inline int compound_mapcount(struct page
*page
)
566 VM_BUG_ON_PAGE(!PageCompound(page
), page
);
567 page
= compound_head(page
);
568 return atomic_read(compound_mapcount_ptr(page
)) + 1;
572 * The atomic page->_mapcount, starts from -1: so that transitions
573 * both from it and to it can be tracked, using atomic_inc_and_test
574 * and atomic_add_negative(-1).
576 static inline void page_mapcount_reset(struct page
*page
)
578 atomic_set(&(page
)->_mapcount
, -1);
581 int __page_mapcount(struct page
*page
);
583 static inline int page_mapcount(struct page
*page
)
585 VM_BUG_ON_PAGE(PageSlab(page
), page
);
587 if (unlikely(PageCompound(page
)))
588 return __page_mapcount(page
);
589 return atomic_read(&page
->_mapcount
) + 1;
592 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
593 int total_mapcount(struct page
*page
);
594 int page_trans_huge_mapcount(struct page
*page
, int *total_mapcount
);
596 static inline int total_mapcount(struct page
*page
)
598 return page_mapcount(page
);
600 static inline int page_trans_huge_mapcount(struct page
*page
,
603 int mapcount
= page_mapcount(page
);
605 *total_mapcount
= mapcount
;
610 static inline struct page
*virt_to_head_page(const void *x
)
612 struct page
*page
= virt_to_page(x
);
614 return compound_head(page
);
617 void __put_page(struct page
*page
);
619 void put_pages_list(struct list_head
*pages
);
621 void split_page(struct page
*page
, unsigned int order
);
624 * Compound pages have a destructor function. Provide a
625 * prototype for that function and accessor functions.
626 * These are _only_ valid on the head of a compound page.
628 typedef void compound_page_dtor(struct page
*);
630 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
631 enum compound_dtor_id
{
634 #ifdef CONFIG_HUGETLB_PAGE
637 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
642 extern compound_page_dtor
* const compound_page_dtors
[];
644 static inline void set_compound_page_dtor(struct page
*page
,
645 enum compound_dtor_id compound_dtor
)
647 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
648 page
[1].compound_dtor
= compound_dtor
;
651 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
653 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
654 return compound_page_dtors
[page
[1].compound_dtor
];
657 static inline unsigned int compound_order(struct page
*page
)
661 return page
[1].compound_order
;
664 static inline void set_compound_order(struct page
*page
, unsigned int order
)
666 page
[1].compound_order
= order
;
669 void free_compound_page(struct page
*page
);
673 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
674 * servicing faults for write access. In the normal case, do always want
675 * pte_mkwrite. But get_user_pages can cause write faults for mappings
676 * that do not have writing enabled, when used by access_process_vm.
678 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
680 if (likely(vma
->vm_flags
& VM_WRITE
))
681 pte
= pte_mkwrite(pte
);
685 int alloc_set_pte(struct vm_fault
*vmf
, struct mem_cgroup
*memcg
,
687 int finish_fault(struct vm_fault
*vmf
);
688 int finish_mkwrite_fault(struct vm_fault
*vmf
);
692 * Multiple processes may "see" the same page. E.g. for untouched
693 * mappings of /dev/null, all processes see the same page full of
694 * zeroes, and text pages of executables and shared libraries have
695 * only one copy in memory, at most, normally.
697 * For the non-reserved pages, page_count(page) denotes a reference count.
698 * page_count() == 0 means the page is free. page->lru is then used for
699 * freelist management in the buddy allocator.
700 * page_count() > 0 means the page has been allocated.
702 * Pages are allocated by the slab allocator in order to provide memory
703 * to kmalloc and kmem_cache_alloc. In this case, the management of the
704 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
705 * unless a particular usage is carefully commented. (the responsibility of
706 * freeing the kmalloc memory is the caller's, of course).
708 * A page may be used by anyone else who does a __get_free_page().
709 * In this case, page_count still tracks the references, and should only
710 * be used through the normal accessor functions. The top bits of page->flags
711 * and page->virtual store page management information, but all other fields
712 * are unused and could be used privately, carefully. The management of this
713 * page is the responsibility of the one who allocated it, and those who have
714 * subsequently been given references to it.
716 * The other pages (we may call them "pagecache pages") are completely
717 * managed by the Linux memory manager: I/O, buffers, swapping etc.
718 * The following discussion applies only to them.
720 * A pagecache page contains an opaque `private' member, which belongs to the
721 * page's address_space. Usually, this is the address of a circular list of
722 * the page's disk buffers. PG_private must be set to tell the VM to call
723 * into the filesystem to release these pages.
725 * A page may belong to an inode's memory mapping. In this case, page->mapping
726 * is the pointer to the inode, and page->index is the file offset of the page,
727 * in units of PAGE_SIZE.
729 * If pagecache pages are not associated with an inode, they are said to be
730 * anonymous pages. These may become associated with the swapcache, and in that
731 * case PG_swapcache is set, and page->private is an offset into the swapcache.
733 * In either case (swapcache or inode backed), the pagecache itself holds one
734 * reference to the page. Setting PG_private should also increment the
735 * refcount. The each user mapping also has a reference to the page.
737 * The pagecache pages are stored in a per-mapping radix tree, which is
738 * rooted at mapping->page_tree, and indexed by offset.
739 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
740 * lists, we instead now tag pages as dirty/writeback in the radix tree.
742 * All pagecache pages may be subject to I/O:
743 * - inode pages may need to be read from disk,
744 * - inode pages which have been modified and are MAP_SHARED may need
745 * to be written back to the inode on disk,
746 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
747 * modified may need to be swapped out to swap space and (later) to be read
752 * The zone field is never updated after free_area_init_core()
753 * sets it, so none of the operations on it need to be atomic.
756 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
757 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
758 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
759 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
760 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
763 * Define the bit shifts to access each section. For non-existent
764 * sections we define the shift as 0; that plus a 0 mask ensures
765 * the compiler will optimise away reference to them.
767 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
768 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
769 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
770 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
772 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
773 #ifdef NODE_NOT_IN_PAGE_FLAGS
774 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
775 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
776 SECTIONS_PGOFF : ZONES_PGOFF)
778 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
779 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
780 NODES_PGOFF : ZONES_PGOFF)
783 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
785 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
786 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
789 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
790 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
791 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
792 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
793 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
795 static inline enum zone_type
page_zonenum(const struct page
*page
)
797 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
800 #ifdef CONFIG_ZONE_DEVICE
801 static inline bool is_zone_device_page(const struct page
*page
)
803 return page_zonenum(page
) == ZONE_DEVICE
;
806 static inline bool is_zone_device_page(const struct page
*page
)
812 #if defined(CONFIG_DEVICE_PRIVATE) || defined(CONFIG_DEVICE_PUBLIC)
813 void put_zone_device_private_or_public_page(struct page
*page
);
814 DECLARE_STATIC_KEY_FALSE(device_private_key
);
815 #define IS_HMM_ENABLED static_branch_unlikely(&device_private_key)
816 static inline bool is_device_private_page(const struct page
*page
);
817 static inline bool is_device_public_page(const struct page
*page
);
818 #else /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
819 static inline void put_zone_device_private_or_public_page(struct page
*page
)
822 #define IS_HMM_ENABLED 0
823 static inline bool is_device_private_page(const struct page
*page
)
827 static inline bool is_device_public_page(const struct page
*page
)
831 #endif /* CONFIG_DEVICE_PRIVATE || CONFIG_DEVICE_PUBLIC */
834 /* 127: arbitrary random number, small enough to assemble well */
835 #define page_ref_zero_or_close_to_overflow(page) \
836 ((unsigned int) page_ref_count(page) + 127u <= 127u)
838 static inline void get_page(struct page
*page
)
840 page
= compound_head(page
);
842 * Getting a normal page or the head of a compound page
843 * requires to already have an elevated page->_refcount.
845 VM_BUG_ON_PAGE(page_ref_zero_or_close_to_overflow(page
), page
);
849 static inline __must_check
bool try_get_page(struct page
*page
)
851 page
= compound_head(page
);
852 if (WARN_ON_ONCE(page_ref_count(page
) <= 0))
858 static inline void put_page(struct page
*page
)
860 page
= compound_head(page
);
863 * For private device pages we need to catch refcount transition from
864 * 2 to 1, when refcount reach one it means the private device page is
865 * free and we need to inform the device driver through callback. See
866 * include/linux/memremap.h and HMM for details.
868 if (IS_HMM_ENABLED
&& unlikely(is_device_private_page(page
) ||
869 unlikely(is_device_public_page(page
)))) {
870 put_zone_device_private_or_public_page(page
);
874 if (put_page_testzero(page
))
878 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
879 #define SECTION_IN_PAGE_FLAGS
883 * The identification function is mainly used by the buddy allocator for
884 * determining if two pages could be buddies. We are not really identifying
885 * the zone since we could be using the section number id if we do not have
886 * node id available in page flags.
887 * We only guarantee that it will return the same value for two combinable
890 static inline int page_zone_id(struct page
*page
)
892 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
895 static inline int zone_to_nid(struct zone
*zone
)
904 #ifdef NODE_NOT_IN_PAGE_FLAGS
905 extern int page_to_nid(const struct page
*page
);
907 static inline int page_to_nid(const struct page
*page
)
909 return (page
->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
913 #ifdef CONFIG_NUMA_BALANCING
914 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
916 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
919 static inline int cpupid_to_pid(int cpupid
)
921 return cpupid
& LAST__PID_MASK
;
924 static inline int cpupid_to_cpu(int cpupid
)
926 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
929 static inline int cpupid_to_nid(int cpupid
)
931 return cpu_to_node(cpupid_to_cpu(cpupid
));
934 static inline bool cpupid_pid_unset(int cpupid
)
936 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
939 static inline bool cpupid_cpu_unset(int cpupid
)
941 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
944 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
946 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
949 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
950 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
951 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
953 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
956 static inline int page_cpupid_last(struct page
*page
)
958 return page
->_last_cpupid
;
960 static inline void page_cpupid_reset_last(struct page
*page
)
962 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
965 static inline int page_cpupid_last(struct page
*page
)
967 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
970 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
972 static inline void page_cpupid_reset_last(struct page
*page
)
974 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
976 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
977 #else /* !CONFIG_NUMA_BALANCING */
978 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
980 return page_to_nid(page
); /* XXX */
983 static inline int page_cpupid_last(struct page
*page
)
985 return page_to_nid(page
); /* XXX */
988 static inline int cpupid_to_nid(int cpupid
)
993 static inline int cpupid_to_pid(int cpupid
)
998 static inline int cpupid_to_cpu(int cpupid
)
1003 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
1008 static inline bool cpupid_pid_unset(int cpupid
)
1013 static inline void page_cpupid_reset_last(struct page
*page
)
1017 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
1021 #endif /* CONFIG_NUMA_BALANCING */
1023 static inline struct zone
*page_zone(const struct page
*page
)
1025 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
1028 static inline pg_data_t
*page_pgdat(const struct page
*page
)
1030 return NODE_DATA(page_to_nid(page
));
1033 #ifdef SECTION_IN_PAGE_FLAGS
1034 static inline void set_page_section(struct page
*page
, unsigned long section
)
1036 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
1037 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
1040 static inline unsigned long page_to_section(const struct page
*page
)
1042 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
1046 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
1048 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
1049 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
1052 static inline void set_page_node(struct page
*page
, unsigned long node
)
1054 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
1055 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
1058 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
1059 unsigned long node
, unsigned long pfn
)
1061 set_page_zone(page
, zone
);
1062 set_page_node(page
, node
);
1063 #ifdef SECTION_IN_PAGE_FLAGS
1064 set_page_section(page
, pfn_to_section_nr(pfn
));
1069 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
1071 return page
->mem_cgroup
;
1073 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1075 WARN_ON_ONCE(!rcu_read_lock_held());
1076 return READ_ONCE(page
->mem_cgroup
);
1079 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
1083 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1085 WARN_ON_ONCE(!rcu_read_lock_held());
1091 * Some inline functions in vmstat.h depend on page_zone()
1093 #include <linux/vmstat.h>
1095 static __always_inline
void *lowmem_page_address(const struct page
*page
)
1097 return page_to_virt(page
);
1100 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1101 #define HASHED_PAGE_VIRTUAL
1104 #if defined(WANT_PAGE_VIRTUAL)
1105 static inline void *page_address(const struct page
*page
)
1107 return page
->virtual;
1109 static inline void set_page_address(struct page
*page
, void *address
)
1111 page
->virtual = address
;
1113 #define page_address_init() do { } while(0)
1116 #if defined(HASHED_PAGE_VIRTUAL)
1117 void *page_address(const struct page
*page
);
1118 void set_page_address(struct page
*page
, void *virtual);
1119 void page_address_init(void);
1122 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1123 #define page_address(page) lowmem_page_address(page)
1124 #define set_page_address(page, address) do { } while(0)
1125 #define page_address_init() do { } while(0)
1128 extern void *page_rmapping(struct page
*page
);
1129 extern struct anon_vma
*page_anon_vma(struct page
*page
);
1130 extern struct address_space
*page_mapping(struct page
*page
);
1132 extern struct address_space
*__page_file_mapping(struct page
*);
1135 struct address_space
*page_file_mapping(struct page
*page
)
1137 if (unlikely(PageSwapCache(page
)))
1138 return __page_file_mapping(page
);
1140 return page
->mapping
;
1143 extern pgoff_t
__page_file_index(struct page
*page
);
1146 * Return the pagecache index of the passed page. Regular pagecache pages
1147 * use ->index whereas swapcache pages use swp_offset(->private)
1149 static inline pgoff_t
page_index(struct page
*page
)
1151 if (unlikely(PageSwapCache(page
)))
1152 return __page_file_index(page
);
1156 bool page_mapped(struct page
*page
);
1157 struct address_space
*page_mapping(struct page
*page
);
1160 * Return true only if the page has been allocated with
1161 * ALLOC_NO_WATERMARKS and the low watermark was not
1162 * met implying that the system is under some pressure.
1164 static inline bool page_is_pfmemalloc(struct page
*page
)
1167 * Page index cannot be this large so this must be
1168 * a pfmemalloc page.
1170 return page
->index
== -1UL;
1174 * Only to be called by the page allocator on a freshly allocated
1177 static inline void set_page_pfmemalloc(struct page
*page
)
1182 static inline void clear_page_pfmemalloc(struct page
*page
)
1188 * Different kinds of faults, as returned by handle_mm_fault().
1189 * Used to decide whether a process gets delivered SIGBUS or
1190 * just gets major/minor fault counters bumped up.
1193 #define VM_FAULT_OOM 0x0001
1194 #define VM_FAULT_SIGBUS 0x0002
1195 #define VM_FAULT_MAJOR 0x0004
1196 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1197 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1198 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1199 #define VM_FAULT_SIGSEGV 0x0040
1201 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1202 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1203 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1204 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1205 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1206 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1207 * and needs fsync() to complete (for
1208 * synchronous page faults in DAX) */
1210 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1211 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1214 #define VM_FAULT_RESULT_TRACE \
1215 { VM_FAULT_OOM, "OOM" }, \
1216 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1217 { VM_FAULT_MAJOR, "MAJOR" }, \
1218 { VM_FAULT_WRITE, "WRITE" }, \
1219 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1220 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1221 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1222 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1223 { VM_FAULT_LOCKED, "LOCKED" }, \
1224 { VM_FAULT_RETRY, "RETRY" }, \
1225 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1226 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1227 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1229 /* Encode hstate index for a hwpoisoned large page */
1230 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1231 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1234 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1236 extern void pagefault_out_of_memory(void);
1238 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1241 * Flags passed to show_mem() and show_free_areas() to suppress output in
1244 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1246 extern void show_free_areas(unsigned int flags
, nodemask_t
*nodemask
);
1248 extern bool can_do_mlock(void);
1249 extern int user_shm_lock(size_t, struct user_struct
*);
1250 extern void user_shm_unlock(size_t, struct user_struct
*);
1253 * Parameter block passed down to zap_pte_range in exceptional cases.
1255 struct zap_details
{
1256 struct address_space
*check_mapping
; /* Check page->mapping if set */
1257 pgoff_t first_index
; /* Lowest page->index to unmap */
1258 pgoff_t last_index
; /* Highest page->index to unmap */
1261 struct page
*_vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1262 pte_t pte
, bool with_public_device
);
1263 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1265 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
1268 int zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1269 unsigned long size
);
1270 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1271 unsigned long size
);
1272 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1273 unsigned long start
, unsigned long end
);
1276 * mm_walk - callbacks for walk_page_range
1277 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1278 * this handler should only handle pud_trans_huge() puds.
1279 * the pmd_entry or pte_entry callbacks will be used for
1281 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1282 * this handler is required to be able to handle
1283 * pmd_trans_huge() pmds. They may simply choose to
1284 * split_huge_page() instead of handling it explicitly.
1285 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1286 * @pte_hole: if set, called for each hole at all levels
1287 * @hugetlb_entry: if set, called for each hugetlb entry
1288 * @test_walk: caller specific callback function to determine whether
1289 * we walk over the current vma or not. Returning 0
1290 * value means "do page table walk over the current vma,"
1291 * and a negative one means "abort current page table walk
1292 * right now." 1 means "skip the current vma."
1293 * @mm: mm_struct representing the target process of page table walk
1294 * @vma: vma currently walked (NULL if walking outside vmas)
1295 * @private: private data for callbacks' usage
1297 * (see the comment on walk_page_range() for more details)
1300 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
1301 unsigned long next
, struct mm_walk
*walk
);
1302 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1303 unsigned long next
, struct mm_walk
*walk
);
1304 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1305 unsigned long next
, struct mm_walk
*walk
);
1306 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1307 struct mm_walk
*walk
);
1308 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1309 unsigned long addr
, unsigned long next
,
1310 struct mm_walk
*walk
);
1311 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1312 struct mm_walk
*walk
);
1313 struct mm_struct
*mm
;
1314 struct vm_area_struct
*vma
;
1318 int walk_page_range(unsigned long addr
, unsigned long end
,
1319 struct mm_walk
*walk
);
1320 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1321 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1322 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1323 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1324 struct vm_area_struct
*vma
);
1325 void unmap_mapping_range(struct address_space
*mapping
,
1326 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1327 int follow_pte_pmd(struct mm_struct
*mm
, unsigned long address
,
1328 unsigned long *start
, unsigned long *end
,
1329 pte_t
**ptepp
, pmd_t
**pmdpp
, spinlock_t
**ptlp
);
1330 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1331 unsigned long *pfn
);
1332 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1333 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1334 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1335 void *buf
, int len
, int write
);
1337 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1338 loff_t
const holebegin
, loff_t
const holelen
)
1340 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1343 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1344 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1345 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1346 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1347 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1348 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1349 int invalidate_inode_page(struct page
*page
);
1352 extern int handle_mm_fault(struct vm_area_struct
*vma
, unsigned long address
,
1353 unsigned int flags
);
1354 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1355 unsigned long address
, unsigned int fault_flags
,
1358 static inline int handle_mm_fault(struct vm_area_struct
*vma
,
1359 unsigned long address
, unsigned int flags
)
1361 /* should never happen if there's no MMU */
1363 return VM_FAULT_SIGBUS
;
1365 static inline int fixup_user_fault(struct task_struct
*tsk
,
1366 struct mm_struct
*mm
, unsigned long address
,
1367 unsigned int fault_flags
, bool *unlocked
)
1369 /* should never happen if there's no MMU */
1375 extern void vma_do_file_update_time(struct vm_area_struct
*, const char[], int);
1376 extern struct file
*vma_do_pr_or_file(struct vm_area_struct
*, const char[],
1378 extern void vma_do_get_file(struct vm_area_struct
*, const char[], int);
1379 extern void vma_do_fput(struct vm_area_struct
*, const char[], int);
1381 #define vma_file_update_time(vma) vma_do_file_update_time(vma, __func__, \
1383 #define vma_pr_or_file(vma) vma_do_pr_or_file(vma, __func__, \
1385 #define vma_get_file(vma) vma_do_get_file(vma, __func__, __LINE__)
1386 #define vma_fput(vma) vma_do_fput(vma, __func__, __LINE__)
1389 extern struct file
*vmr_do_pr_or_file(struct vm_region
*, const char[], int);
1390 extern void vmr_do_fput(struct vm_region
*, const char[], int);
1392 #define vmr_pr_or_file(region) vmr_do_pr_or_file(region, __func__, \
1394 #define vmr_fput(region) vmr_do_fput(region, __func__, __LINE__)
1395 #endif /* !CONFIG_MMU */
1397 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
, void *buf
, int len
,
1398 unsigned int gup_flags
);
1399 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1400 void *buf
, int len
, unsigned int gup_flags
);
1401 extern int __access_remote_vm(struct task_struct
*tsk
, struct mm_struct
*mm
,
1402 unsigned long addr
, void *buf
, int len
, unsigned int gup_flags
);
1404 long get_user_pages_remote(struct task_struct
*tsk
, struct mm_struct
*mm
,
1405 unsigned long start
, unsigned long nr_pages
,
1406 unsigned int gup_flags
, struct page
**pages
,
1407 struct vm_area_struct
**vmas
, int *locked
);
1408 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
1409 unsigned int gup_flags
, struct page
**pages
,
1410 struct vm_area_struct
**vmas
);
1411 long get_user_pages_locked(unsigned long start
, unsigned long nr_pages
,
1412 unsigned int gup_flags
, struct page
**pages
, int *locked
);
1413 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
1414 struct page
**pages
, unsigned int gup_flags
);
1415 #ifdef CONFIG_FS_DAX
1416 long get_user_pages_longterm(unsigned long start
, unsigned long nr_pages
,
1417 unsigned int gup_flags
, struct page
**pages
,
1418 struct vm_area_struct
**vmas
);
1420 static inline long get_user_pages_longterm(unsigned long start
,
1421 unsigned long nr_pages
, unsigned int gup_flags
,
1422 struct page
**pages
, struct vm_area_struct
**vmas
)
1424 return get_user_pages(start
, nr_pages
, gup_flags
, pages
, vmas
);
1426 #endif /* CONFIG_FS_DAX */
1428 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1429 struct page
**pages
);
1431 /* Container for pinned pfns / pages */
1432 struct frame_vector
{
1433 unsigned int nr_allocated
; /* Number of frames we have space for */
1434 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1435 bool got_ref
; /* Did we pin pages by getting page ref? */
1436 bool is_pfns
; /* Does array contain pages or pfns? */
1437 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1438 * pfns_vector_pages() or pfns_vector_pfns()
1442 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1443 void frame_vector_destroy(struct frame_vector
*vec
);
1444 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1445 unsigned int gup_flags
, struct frame_vector
*vec
);
1446 void put_vaddr_frames(struct frame_vector
*vec
);
1447 int frame_vector_to_pages(struct frame_vector
*vec
);
1448 void frame_vector_to_pfns(struct frame_vector
*vec
);
1450 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1452 return vec
->nr_frames
;
1455 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1458 int err
= frame_vector_to_pages(vec
);
1461 return ERR_PTR(err
);
1463 return (struct page
**)(vec
->ptrs
);
1466 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1469 frame_vector_to_pfns(vec
);
1470 return (unsigned long *)(vec
->ptrs
);
1474 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1475 struct page
**pages
);
1476 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1477 struct page
*get_dump_page(unsigned long addr
);
1479 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1480 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1481 unsigned int length
);
1483 int __set_page_dirty_nobuffers(struct page
*page
);
1484 int __set_page_dirty_no_writeback(struct page
*page
);
1485 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1487 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1488 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1489 struct bdi_writeback
*wb
);
1490 int set_page_dirty(struct page
*page
);
1491 int set_page_dirty_lock(struct page
*page
);
1492 void __cancel_dirty_page(struct page
*page
);
1493 static inline void cancel_dirty_page(struct page
*page
)
1495 /* Avoid atomic ops, locking, etc. when not actually needed. */
1496 if (PageDirty(page
))
1497 __cancel_dirty_page(page
);
1499 int clear_page_dirty_for_io(struct page
*page
);
1501 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1503 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1505 return !vma
->vm_ops
;
1510 * The vma_is_shmem is not inline because it is used only by slow
1511 * paths in userfault.
1513 bool vma_is_shmem(struct vm_area_struct
*vma
);
1515 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
1518 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
1520 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1521 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1522 unsigned long new_addr
, unsigned long len
,
1523 bool need_rmap_locks
);
1524 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1525 unsigned long end
, pgprot_t newprot
,
1526 int dirty_accountable
, int prot_numa
);
1527 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1528 struct vm_area_struct
**pprev
, unsigned long start
,
1529 unsigned long end
, unsigned long newflags
);
1532 * doesn't attempt to fault and will return short.
1534 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1535 struct page
**pages
);
1537 * per-process(per-mm_struct) statistics.
1539 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1541 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1543 #ifdef SPLIT_RSS_COUNTING
1545 * counter is updated in asynchronous manner and may go to minus.
1546 * But it's never be expected number for users.
1551 return (unsigned long)val
;
1554 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1556 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1559 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1561 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1564 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1566 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1569 /* Optimized variant when page is already known not to be PageAnon */
1570 static inline int mm_counter_file(struct page
*page
)
1572 if (PageSwapBacked(page
))
1573 return MM_SHMEMPAGES
;
1574 return MM_FILEPAGES
;
1577 static inline int mm_counter(struct page
*page
)
1580 return MM_ANONPAGES
;
1581 return mm_counter_file(page
);
1584 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1586 return get_mm_counter(mm
, MM_FILEPAGES
) +
1587 get_mm_counter(mm
, MM_ANONPAGES
) +
1588 get_mm_counter(mm
, MM_SHMEMPAGES
);
1591 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1593 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1596 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1598 return max(mm
->hiwater_vm
, mm
->total_vm
);
1601 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1603 unsigned long _rss
= get_mm_rss(mm
);
1605 if ((mm
)->hiwater_rss
< _rss
)
1606 (mm
)->hiwater_rss
= _rss
;
1609 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1611 if (mm
->hiwater_vm
< mm
->total_vm
)
1612 mm
->hiwater_vm
= mm
->total_vm
;
1615 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1617 mm
->hiwater_rss
= get_mm_rss(mm
);
1620 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1621 struct mm_struct
*mm
)
1623 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1625 if (*maxrss
< hiwater_rss
)
1626 *maxrss
= hiwater_rss
;
1629 #if defined(SPLIT_RSS_COUNTING)
1630 void sync_mm_rss(struct mm_struct
*mm
);
1632 static inline void sync_mm_rss(struct mm_struct
*mm
)
1637 #ifndef __HAVE_ARCH_PTE_DEVMAP
1638 static inline int pte_devmap(pte_t pte
)
1644 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
1646 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1648 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1652 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1656 #ifdef __PAGETABLE_P4D_FOLDED
1657 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1658 unsigned long address
)
1663 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1666 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1667 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1668 unsigned long address
)
1672 static inline void mm_inc_nr_puds(struct mm_struct
*mm
) {}
1673 static inline void mm_dec_nr_puds(struct mm_struct
*mm
) {}
1676 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
1678 static inline void mm_inc_nr_puds(struct mm_struct
*mm
)
1680 if (mm_pud_folded(mm
))
1682 atomic_long_add(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
1685 static inline void mm_dec_nr_puds(struct mm_struct
*mm
)
1687 if (mm_pud_folded(mm
))
1689 atomic_long_sub(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
1693 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1694 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1695 unsigned long address
)
1700 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1701 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1704 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1706 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1708 if (mm_pmd_folded(mm
))
1710 atomic_long_add(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
1713 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1715 if (mm_pmd_folded(mm
))
1717 atomic_long_sub(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
1722 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
)
1724 atomic_long_set(&mm
->pgtables_bytes
, 0);
1727 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
1729 return atomic_long_read(&mm
->pgtables_bytes
);
1732 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
)
1734 atomic_long_add(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
1737 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
)
1739 atomic_long_sub(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
1743 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
) {}
1744 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
1749 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
) {}
1750 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
) {}
1753 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
1754 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1757 * The following ifdef needed to get the 4level-fixup.h header to work.
1758 * Remove it when 4level-fixup.h has been removed.
1760 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1762 #ifndef __ARCH_HAS_5LEVEL_HACK
1763 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1764 unsigned long address
)
1766 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
1767 NULL
: p4d_offset(pgd
, address
);
1770 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1771 unsigned long address
)
1773 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
1774 NULL
: pud_offset(p4d
, address
);
1776 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1778 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1780 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1781 NULL
: pmd_offset(pud
, address
);
1783 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1785 #if USE_SPLIT_PTE_PTLOCKS
1786 #if ALLOC_SPLIT_PTLOCKS
1787 void __init
ptlock_cache_init(void);
1788 extern bool ptlock_alloc(struct page
*page
);
1789 extern void ptlock_free(struct page
*page
);
1791 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1795 #else /* ALLOC_SPLIT_PTLOCKS */
1796 static inline void ptlock_cache_init(void)
1800 static inline bool ptlock_alloc(struct page
*page
)
1805 static inline void ptlock_free(struct page
*page
)
1809 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1813 #endif /* ALLOC_SPLIT_PTLOCKS */
1815 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1817 return ptlock_ptr(pmd_page(*pmd
));
1820 static inline bool ptlock_init(struct page
*page
)
1823 * prep_new_page() initialize page->private (and therefore page->ptl)
1824 * with 0. Make sure nobody took it in use in between.
1826 * It can happen if arch try to use slab for page table allocation:
1827 * slab code uses page->slab_cache, which share storage with page->ptl.
1829 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1830 if (!ptlock_alloc(page
))
1832 spin_lock_init(ptlock_ptr(page
));
1836 /* Reset page->mapping so free_pages_check won't complain. */
1837 static inline void pte_lock_deinit(struct page
*page
)
1839 page
->mapping
= NULL
;
1843 #else /* !USE_SPLIT_PTE_PTLOCKS */
1845 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1847 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1849 return &mm
->page_table_lock
;
1851 static inline void ptlock_cache_init(void) {}
1852 static inline bool ptlock_init(struct page
*page
) { return true; }
1853 static inline void pte_lock_deinit(struct page
*page
) {}
1854 #endif /* USE_SPLIT_PTE_PTLOCKS */
1856 static inline void pgtable_init(void)
1858 ptlock_cache_init();
1859 pgtable_cache_init();
1862 static inline bool pgtable_page_ctor(struct page
*page
)
1864 if (!ptlock_init(page
))
1866 inc_zone_page_state(page
, NR_PAGETABLE
);
1870 static inline void pgtable_page_dtor(struct page
*page
)
1872 pte_lock_deinit(page
);
1873 dec_zone_page_state(page
, NR_PAGETABLE
);
1876 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1878 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1879 pte_t *__pte = pte_offset_map(pmd, address); \
1885 #define pte_unmap_unlock(pte, ptl) do { \
1890 #define pte_alloc(mm, pmd, address) \
1891 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1893 #define pte_alloc_map(mm, pmd, address) \
1894 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1896 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1897 (pte_alloc(mm, pmd, address) ? \
1898 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1900 #define pte_alloc_kernel(pmd, address) \
1901 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1902 NULL: pte_offset_kernel(pmd, address))
1904 #if USE_SPLIT_PMD_PTLOCKS
1906 static struct page
*pmd_to_page(pmd_t
*pmd
)
1908 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1909 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1912 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1914 return ptlock_ptr(pmd_to_page(pmd
));
1917 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1919 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1920 page
->pmd_huge_pte
= NULL
;
1922 return ptlock_init(page
);
1925 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1927 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1928 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
1933 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
1937 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1939 return &mm
->page_table_lock
;
1942 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
1943 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
1945 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
1949 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
1951 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
1957 * No scalability reason to split PUD locks yet, but follow the same pattern
1958 * as the PMD locks to make it easier if we decide to. The VM should not be
1959 * considered ready to switch to split PUD locks yet; there may be places
1960 * which need to be converted from page_table_lock.
1962 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
1964 return &mm
->page_table_lock
;
1967 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
1969 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
1975 extern void __init
pagecache_init(void);
1976 extern void free_area_init(unsigned long * zones_size
);
1977 extern void free_area_init_node(int nid
, unsigned long * zones_size
,
1978 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
1979 extern void free_initmem(void);
1982 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
1983 * into the buddy system. The freed pages will be poisoned with pattern
1984 * "poison" if it's within range [0, UCHAR_MAX].
1985 * Return pages freed into the buddy system.
1987 extern unsigned long free_reserved_area(void *start
, void *end
,
1988 int poison
, char *s
);
1990 #ifdef CONFIG_HIGHMEM
1992 * Free a highmem page into the buddy system, adjusting totalhigh_pages
1993 * and totalram_pages.
1995 extern void free_highmem_page(struct page
*page
);
1998 extern void adjust_managed_page_count(struct page
*page
, long count
);
1999 extern void mem_init_print_info(const char *str
);
2001 extern void reserve_bootmem_region(phys_addr_t start
, phys_addr_t end
);
2003 /* Free the reserved page into the buddy system, so it gets managed. */
2004 static inline void __free_reserved_page(struct page
*page
)
2006 ClearPageReserved(page
);
2007 init_page_count(page
);
2011 static inline void free_reserved_page(struct page
*page
)
2013 __free_reserved_page(page
);
2014 adjust_managed_page_count(page
, 1);
2017 static inline void mark_page_reserved(struct page
*page
)
2019 SetPageReserved(page
);
2020 adjust_managed_page_count(page
, -1);
2024 * Default method to free all the __init memory into the buddy system.
2025 * The freed pages will be poisoned with pattern "poison" if it's within
2026 * range [0, UCHAR_MAX].
2027 * Return pages freed into the buddy system.
2029 static inline unsigned long free_initmem_default(int poison
)
2031 extern char __init_begin
[], __init_end
[];
2033 return free_reserved_area(&__init_begin
, &__init_end
,
2034 poison
, "unused kernel");
2037 static inline unsigned long get_num_physpages(void)
2040 unsigned long phys_pages
= 0;
2042 for_each_online_node(nid
)
2043 phys_pages
+= node_present_pages(nid
);
2048 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2050 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2051 * zones, allocate the backing mem_map and account for memory holes in a more
2052 * architecture independent manner. This is a substitute for creating the
2053 * zone_sizes[] and zholes_size[] arrays and passing them to
2054 * free_area_init_node()
2056 * An architecture is expected to register range of page frames backed by
2057 * physical memory with memblock_add[_node]() before calling
2058 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2059 * usage, an architecture is expected to do something like
2061 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2063 * for_each_valid_physical_page_range()
2064 * memblock_add_node(base, size, nid)
2065 * free_area_init_nodes(max_zone_pfns);
2067 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2068 * registered physical page range. Similarly
2069 * sparse_memory_present_with_active_regions() calls memory_present() for
2070 * each range when SPARSEMEM is enabled.
2072 * See mm/page_alloc.c for more information on each function exposed by
2073 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2075 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
2076 unsigned long node_map_pfn_alignment(void);
2077 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
2078 unsigned long end_pfn
);
2079 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
2080 unsigned long end_pfn
);
2081 extern void get_pfn_range_for_nid(unsigned int nid
,
2082 unsigned long *start_pfn
, unsigned long *end_pfn
);
2083 extern unsigned long find_min_pfn_with_active_regions(void);
2084 extern void free_bootmem_with_active_regions(int nid
,
2085 unsigned long max_low_pfn
);
2086 extern void sparse_memory_present_with_active_regions(int nid
);
2088 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2090 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2091 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2092 static inline int __early_pfn_to_nid(unsigned long pfn
,
2093 struct mminit_pfnnid_cache
*state
)
2098 /* please see mm/page_alloc.c */
2099 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
2100 /* there is a per-arch backend function. */
2101 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
2102 struct mminit_pfnnid_cache
*state
);
2105 #if defined(CONFIG_HAVE_MEMBLOCK) && !defined(CONFIG_FLAT_NODE_MEM_MAP)
2106 void zero_resv_unavail(void);
2108 static inline void zero_resv_unavail(void) {}
2111 extern void set_dma_reserve(unsigned long new_dma_reserve
);
2112 extern void memmap_init_zone(unsigned long, int, unsigned long,
2113 unsigned long, enum memmap_context
);
2114 extern void setup_per_zone_wmarks(void);
2115 extern int __meminit
init_per_zone_wmark_min(void);
2116 extern void mem_init(void);
2117 extern void __init
mmap_init(void);
2118 extern void show_mem(unsigned int flags
, nodemask_t
*nodemask
);
2119 extern long si_mem_available(void);
2120 extern void si_meminfo(struct sysinfo
* val
);
2121 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
2122 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2123 extern unsigned long arch_reserved_kernel_pages(void);
2126 extern __printf(3, 4)
2127 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
2129 extern void setup_per_cpu_pageset(void);
2131 extern void zone_pcp_update(struct zone
*zone
);
2132 extern void zone_pcp_reset(struct zone
*zone
);
2135 extern int min_free_kbytes
;
2136 extern int watermark_scale_factor
;
2139 extern atomic_long_t mmap_pages_allocated
;
2140 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
2142 /* interval_tree.c */
2143 void vma_interval_tree_insert(struct vm_area_struct
*node
,
2144 struct rb_root_cached
*root
);
2145 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
2146 struct vm_area_struct
*prev
,
2147 struct rb_root_cached
*root
);
2148 void vma_interval_tree_remove(struct vm_area_struct
*node
,
2149 struct rb_root_cached
*root
);
2150 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2151 unsigned long start
, unsigned long last
);
2152 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
2153 unsigned long start
, unsigned long last
);
2155 #define vma_interval_tree_foreach(vma, root, start, last) \
2156 for (vma = vma_interval_tree_iter_first(root, start, last); \
2157 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2159 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
2160 struct rb_root_cached
*root
);
2161 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
2162 struct rb_root_cached
*root
);
2163 struct anon_vma_chain
*
2164 anon_vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2165 unsigned long start
, unsigned long last
);
2166 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
2167 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
2168 #ifdef CONFIG_DEBUG_VM_RB
2169 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
2172 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2173 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2174 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2177 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
2178 extern int __vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2179 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
,
2180 struct vm_area_struct
*expand
);
2181 static inline int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2182 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
)
2184 return __vma_adjust(vma
, start
, end
, pgoff
, insert
, NULL
);
2186 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
2187 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
2188 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
2189 struct mempolicy
*, struct vm_userfaultfd_ctx
);
2190 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
2191 extern int __split_vma(struct mm_struct
*, struct vm_area_struct
*,
2192 unsigned long addr
, int new_below
);
2193 extern int split_vma(struct mm_struct
*, struct vm_area_struct
*,
2194 unsigned long addr
, int new_below
);
2195 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
2196 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
2197 struct rb_node
**, struct rb_node
*);
2198 extern void unlink_file_vma(struct vm_area_struct
*);
2199 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
2200 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
2201 bool *need_rmap_locks
);
2202 extern void exit_mmap(struct mm_struct
*);
2204 static inline int check_data_rlimit(unsigned long rlim
,
2206 unsigned long start
,
2207 unsigned long end_data
,
2208 unsigned long start_data
)
2210 if (rlim
< RLIM_INFINITY
) {
2211 if (((new - start
) + (end_data
- start_data
)) > rlim
)
2218 extern int mm_take_all_locks(struct mm_struct
*mm
);
2219 extern void mm_drop_all_locks(struct mm_struct
*mm
);
2221 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2222 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2223 extern struct file
*get_task_exe_file(struct task_struct
*task
);
2225 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2226 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2228 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
2229 const struct vm_special_mapping
*sm
);
2230 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2231 unsigned long addr
, unsigned long len
,
2232 unsigned long flags
,
2233 const struct vm_special_mapping
*spec
);
2234 /* This is an obsolete alternative to _install_special_mapping. */
2235 extern int install_special_mapping(struct mm_struct
*mm
,
2236 unsigned long addr
, unsigned long len
,
2237 unsigned long flags
, struct page
**pages
);
2239 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
2241 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
2242 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
2243 struct list_head
*uf
);
2244 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
2245 unsigned long len
, unsigned long prot
, unsigned long flags
,
2246 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
,
2247 struct list_head
*uf
);
2248 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
2249 struct list_head
*uf
);
2251 static inline unsigned long
2252 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
2253 unsigned long len
, unsigned long prot
, unsigned long flags
,
2254 unsigned long pgoff
, unsigned long *populate
,
2255 struct list_head
*uf
)
2257 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
, uf
);
2261 extern int __mm_populate(unsigned long addr
, unsigned long len
,
2263 static inline void mm_populate(unsigned long addr
, unsigned long len
)
2266 (void) __mm_populate(addr
, len
, 1);
2269 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
2272 /* These take the mm semaphore themselves */
2273 extern int __must_check
vm_brk(unsigned long, unsigned long);
2274 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
2275 extern int vm_munmap(unsigned long, size_t);
2276 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
2277 unsigned long, unsigned long,
2278 unsigned long, unsigned long);
2280 struct vm_unmapped_area_info
{
2281 #define VM_UNMAPPED_AREA_TOPDOWN 1
2282 unsigned long flags
;
2283 unsigned long length
;
2284 unsigned long low_limit
;
2285 unsigned long high_limit
;
2286 unsigned long align_mask
;
2287 unsigned long align_offset
;
2290 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
2291 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
2294 * Search for an unmapped address range.
2296 * We are looking for a range that:
2297 * - does not intersect with any VMA;
2298 * - is contained within the [low_limit, high_limit) interval;
2299 * - is at least the desired size.
2300 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2302 static inline unsigned long
2303 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
2305 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
2306 return unmapped_area_topdown(info
);
2308 return unmapped_area(info
);
2312 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2313 extern void truncate_inode_pages_range(struct address_space
*,
2314 loff_t lstart
, loff_t lend
);
2315 extern void truncate_inode_pages_final(struct address_space
*);
2317 /* generic vm_area_ops exported for stackable file systems */
2318 extern int filemap_fault(struct vm_fault
*vmf
);
2319 extern void filemap_map_pages(struct vm_fault
*vmf
,
2320 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
2321 extern int filemap_page_mkwrite(struct vm_fault
*vmf
);
2323 /* mm/page-writeback.c */
2324 int __must_check
write_one_page(struct page
*page
);
2325 void task_dirty_inc(struct task_struct
*tsk
);
2328 #define VM_MAX_READAHEAD 128 /* kbytes */
2329 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2331 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2332 pgoff_t offset
, unsigned long nr_to_read
);
2334 void page_cache_sync_readahead(struct address_space
*mapping
,
2335 struct file_ra_state
*ra
,
2338 unsigned long size
);
2340 void page_cache_async_readahead(struct address_space
*mapping
,
2341 struct file_ra_state
*ra
,
2345 unsigned long size
);
2347 extern unsigned long stack_guard_gap
;
2348 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2349 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2351 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2352 extern int expand_downwards(struct vm_area_struct
*vma
,
2353 unsigned long address
);
2355 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2357 #define expand_upwards(vma, address) (0)
2360 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2361 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2362 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2363 struct vm_area_struct
**pprev
);
2365 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2366 NULL if none. Assume start_addr < end_addr. */
2367 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2369 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2371 if (vma
&& end_addr
<= vma
->vm_start
)
2376 static inline unsigned long vm_start_gap(struct vm_area_struct
*vma
)
2378 unsigned long vm_start
= vma
->vm_start
;
2380 if (vma
->vm_flags
& VM_GROWSDOWN
) {
2381 vm_start
-= stack_guard_gap
;
2382 if (vm_start
> vma
->vm_start
)
2388 static inline unsigned long vm_end_gap(struct vm_area_struct
*vma
)
2390 unsigned long vm_end
= vma
->vm_end
;
2392 if (vma
->vm_flags
& VM_GROWSUP
) {
2393 vm_end
+= stack_guard_gap
;
2394 if (vm_end
< vma
->vm_end
)
2395 vm_end
= -PAGE_SIZE
;
2400 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2402 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2405 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2406 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2407 unsigned long vm_start
, unsigned long vm_end
)
2409 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2411 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2417 static inline bool range_in_vma(struct vm_area_struct
*vma
,
2418 unsigned long start
, unsigned long end
)
2420 return (vma
&& vma
->vm_start
<= start
&& end
<= vma
->vm_end
);
2424 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2425 void vma_set_page_prot(struct vm_area_struct
*vma
);
2427 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2431 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2433 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2437 #ifdef CONFIG_NUMA_BALANCING
2438 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2439 unsigned long start
, unsigned long end
);
2442 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2443 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2444 unsigned long pfn
, unsigned long size
, pgprot_t
);
2445 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2446 int vm_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2448 int vm_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2449 unsigned long pfn
, pgprot_t pgprot
);
2450 int vm_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2452 int vm_insert_mixed_mkwrite(struct vm_area_struct
*vma
, unsigned long addr
,
2454 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2457 struct page
*follow_page_mask(struct vm_area_struct
*vma
,
2458 unsigned long address
, unsigned int foll_flags
,
2459 unsigned int *page_mask
);
2461 static inline struct page
*follow_page(struct vm_area_struct
*vma
,
2462 unsigned long address
, unsigned int foll_flags
)
2464 unsigned int unused_page_mask
;
2465 return follow_page_mask(vma
, address
, foll_flags
, &unused_page_mask
);
2468 #define FOLL_WRITE 0x01 /* check pte is writable */
2469 #define FOLL_TOUCH 0x02 /* mark page accessed */
2470 #define FOLL_GET 0x04 /* do get_page on page */
2471 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2472 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2473 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2474 * and return without waiting upon it */
2475 #define FOLL_POPULATE 0x40 /* fault in page */
2476 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2477 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2478 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2479 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2480 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2481 #define FOLL_MLOCK 0x1000 /* lock present pages */
2482 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2483 #define FOLL_COW 0x4000 /* internal GUP flag */
2484 #define FOLL_ANON 0x8000 /* don't do file mappings */
2486 static inline int vm_fault_to_errno(int vm_fault
, int foll_flags
)
2488 if (vm_fault
& VM_FAULT_OOM
)
2490 if (vm_fault
& (VM_FAULT_HWPOISON
| VM_FAULT_HWPOISON_LARGE
))
2491 return (foll_flags
& FOLL_HWPOISON
) ? -EHWPOISON
: -EFAULT
;
2492 if (vm_fault
& (VM_FAULT_SIGBUS
| VM_FAULT_SIGSEGV
))
2497 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2499 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2500 unsigned long size
, pte_fn_t fn
, void *data
);
2503 #ifdef CONFIG_PAGE_POISONING
2504 extern bool page_poisoning_enabled(void);
2505 extern void kernel_poison_pages(struct page
*page
, int numpages
, int enable
);
2506 extern bool page_is_poisoned(struct page
*page
);
2508 static inline bool page_poisoning_enabled(void) { return false; }
2509 static inline void kernel_poison_pages(struct page
*page
, int numpages
,
2511 static inline bool page_is_poisoned(struct page
*page
) { return false; }
2514 #ifdef CONFIG_DEBUG_PAGEALLOC
2515 extern bool _debug_pagealloc_enabled
;
2516 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2518 static inline bool debug_pagealloc_enabled(void)
2520 return _debug_pagealloc_enabled
;
2524 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2526 if (!debug_pagealloc_enabled())
2529 __kernel_map_pages(page
, numpages
, enable
);
2531 #ifdef CONFIG_HIBERNATION
2532 extern bool kernel_page_present(struct page
*page
);
2533 #endif /* CONFIG_HIBERNATION */
2534 #else /* CONFIG_DEBUG_PAGEALLOC */
2536 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2537 #ifdef CONFIG_HIBERNATION
2538 static inline bool kernel_page_present(struct page
*page
) { return true; }
2539 #endif /* CONFIG_HIBERNATION */
2540 static inline bool debug_pagealloc_enabled(void)
2544 #endif /* CONFIG_DEBUG_PAGEALLOC */
2546 #ifdef __HAVE_ARCH_GATE_AREA
2547 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2548 extern int in_gate_area_no_mm(unsigned long addr
);
2549 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2551 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2555 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2556 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2560 #endif /* __HAVE_ARCH_GATE_AREA */
2562 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
2564 #ifdef CONFIG_SYSCTL
2565 extern int sysctl_drop_caches
;
2566 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2567 void __user
*, size_t *, loff_t
*);
2570 void drop_slab(void);
2571 void drop_slab_node(int nid
);
2574 #define randomize_va_space 0
2576 extern int randomize_va_space
;
2579 const char * arch_vma_name(struct vm_area_struct
*vma
);
2580 void print_vma_addr(char *prefix
, unsigned long rip
);
2582 void sparse_mem_maps_populate_node(struct page
**map_map
,
2583 unsigned long pnum_begin
,
2584 unsigned long pnum_end
,
2585 unsigned long map_count
,
2588 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
);
2589 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2590 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2591 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
2592 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2593 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2594 void *vmemmap_alloc_block(unsigned long size
, int node
);
2596 void *__vmemmap_alloc_block_buf(unsigned long size
, int node
,
2597 struct vmem_altmap
*altmap
);
2598 static inline void *vmemmap_alloc_block_buf(unsigned long size
, int node
)
2600 return __vmemmap_alloc_block_buf(size
, node
, NULL
);
2603 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2604 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2606 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
);
2607 void vmemmap_populate_print_last(void);
2608 #ifdef CONFIG_MEMORY_HOTPLUG
2609 void vmemmap_free(unsigned long start
, unsigned long end
);
2611 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2612 unsigned long nr_pages
);
2615 MF_COUNT_INCREASED
= 1 << 0,
2616 MF_ACTION_REQUIRED
= 1 << 1,
2617 MF_MUST_KILL
= 1 << 2,
2618 MF_SOFT_OFFLINE
= 1 << 3,
2620 extern int memory_failure(unsigned long pfn
, int trapno
, int flags
);
2621 extern void memory_failure_queue(unsigned long pfn
, int trapno
, int flags
);
2622 extern int unpoison_memory(unsigned long pfn
);
2623 extern int get_hwpoison_page(struct page
*page
);
2624 #define put_hwpoison_page(page) put_page(page)
2625 extern int sysctl_memory_failure_early_kill
;
2626 extern int sysctl_memory_failure_recovery
;
2627 extern void shake_page(struct page
*p
, int access
);
2628 extern atomic_long_t num_poisoned_pages
;
2629 extern int soft_offline_page(struct page
*page
, int flags
);
2633 * Error handlers for various types of pages.
2636 MF_IGNORED
, /* Error: cannot be handled */
2637 MF_FAILED
, /* Error: handling failed */
2638 MF_DELAYED
, /* Will be handled later */
2639 MF_RECOVERED
, /* Successfully recovered */
2642 enum mf_action_page_type
{
2644 MF_MSG_KERNEL_HIGH_ORDER
,
2646 MF_MSG_DIFFERENT_COMPOUND
,
2647 MF_MSG_POISONED_HUGE
,
2650 MF_MSG_NON_PMD_HUGE
,
2651 MF_MSG_UNMAP_FAILED
,
2652 MF_MSG_DIRTY_SWAPCACHE
,
2653 MF_MSG_CLEAN_SWAPCACHE
,
2654 MF_MSG_DIRTY_MLOCKED_LRU
,
2655 MF_MSG_CLEAN_MLOCKED_LRU
,
2656 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2657 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2660 MF_MSG_TRUNCATED_LRU
,
2666 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2667 extern void clear_huge_page(struct page
*page
,
2668 unsigned long addr_hint
,
2669 unsigned int pages_per_huge_page
);
2670 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2671 unsigned long addr
, struct vm_area_struct
*vma
,
2672 unsigned int pages_per_huge_page
);
2673 extern long copy_huge_page_from_user(struct page
*dst_page
,
2674 const void __user
*usr_src
,
2675 unsigned int pages_per_huge_page
,
2676 bool allow_pagefault
);
2677 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2679 extern struct page_ext_operations debug_guardpage_ops
;
2681 #ifdef CONFIG_DEBUG_PAGEALLOC
2682 extern unsigned int _debug_guardpage_minorder
;
2683 extern bool _debug_guardpage_enabled
;
2685 static inline unsigned int debug_guardpage_minorder(void)
2687 return _debug_guardpage_minorder
;
2690 static inline bool debug_guardpage_enabled(void)
2692 return _debug_guardpage_enabled
;
2695 static inline bool page_is_guard(struct page
*page
)
2697 struct page_ext
*page_ext
;
2699 if (!debug_guardpage_enabled())
2702 page_ext
= lookup_page_ext(page
);
2703 if (unlikely(!page_ext
))
2706 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2709 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2710 static inline bool debug_guardpage_enabled(void) { return false; }
2711 static inline bool page_is_guard(struct page
*page
) { return false; }
2712 #endif /* CONFIG_DEBUG_PAGEALLOC */
2714 #if MAX_NUMNODES > 1
2715 void __init
setup_nr_node_ids(void);
2717 static inline void setup_nr_node_ids(void) {}
2720 #endif /* __KERNEL__ */
2721 #endif /* _LINUX_MM_H */