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>
28 #include <linux/overflow.h>
32 struct anon_vma_chain
;
35 struct writeback_control
;
38 void init_mm_internals(void);
40 #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */
41 extern unsigned long max_mapnr
;
43 static inline void set_max_mapnr(unsigned long limit
)
48 static inline void set_max_mapnr(unsigned long limit
) { }
51 extern unsigned long totalram_pages
;
52 extern void * high_memory
;
53 extern int page_cluster
;
56 extern int sysctl_legacy_va_layout
;
58 #define sysctl_legacy_va_layout 0
61 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS
62 extern const int mmap_rnd_bits_min
;
63 extern const int mmap_rnd_bits_max
;
64 extern int mmap_rnd_bits __read_mostly
;
66 #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS
67 extern const int mmap_rnd_compat_bits_min
;
68 extern const int mmap_rnd_compat_bits_max
;
69 extern int mmap_rnd_compat_bits __read_mostly
;
73 #include <asm/pgtable.h>
74 #include <asm/processor.h>
77 #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0))
81 #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x)))
85 #define lm_alias(x) __va(__pa_symbol(x))
89 * To prevent common memory management code establishing
90 * a zero page mapping on a read fault.
91 * This macro should be defined within <asm/pgtable.h>.
92 * s390 does this to prevent multiplexing of hardware bits
93 * related to the physical page in case of virtualization.
95 #ifndef mm_forbids_zeropage
96 #define mm_forbids_zeropage(X) (0)
100 * On some architectures it is expensive to call memset() for small sizes.
101 * Those architectures should provide their own implementation of "struct page"
102 * zeroing by defining this macro in <asm/pgtable.h>.
104 #ifndef mm_zero_struct_page
105 #define mm_zero_struct_page(pp) ((void)memset((pp), 0, sizeof(struct page)))
109 * Default maximum number of active map areas, this limits the number of vmas
110 * per mm struct. Users can overwrite this number by sysctl but there is a
113 * When a program's coredump is generated as ELF format, a section is created
114 * per a vma. In ELF, the number of sections is represented in unsigned short.
115 * This means the number of sections should be smaller than 65535 at coredump.
116 * Because the kernel adds some informative sections to a image of program at
117 * generating coredump, we need some margin. The number of extra sections is
118 * 1-3 now and depends on arch. We use "5" as safe margin, here.
120 * ELF extended numbering allows more than 65535 sections, so 16-bit bound is
121 * not a hard limit any more. Although some userspace tools can be surprised by
124 #define MAPCOUNT_ELF_CORE_MARGIN (5)
125 #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN)
127 extern int sysctl_max_map_count
;
129 extern unsigned long sysctl_user_reserve_kbytes
;
130 extern unsigned long sysctl_admin_reserve_kbytes
;
132 extern int sysctl_overcommit_memory
;
133 extern int sysctl_overcommit_ratio
;
134 extern unsigned long sysctl_overcommit_kbytes
;
136 extern int overcommit_ratio_handler(struct ctl_table
*, int, void __user
*,
138 extern int overcommit_kbytes_handler(struct ctl_table
*, int, void __user
*,
141 #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n))
143 /* to align the pointer to the (next) page boundary */
144 #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE)
146 /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */
147 #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE)
150 * Linux kernel virtual memory manager primitives.
151 * The idea being to have a "virtual" mm in the same way
152 * we have a virtual fs - giving a cleaner interface to the
153 * mm details, and allowing different kinds of memory mappings
154 * (from shared memory to executable loading to arbitrary
158 struct vm_area_struct
*vm_area_alloc(struct mm_struct
*);
159 struct vm_area_struct
*vm_area_dup(struct vm_area_struct
*);
160 void vm_area_free(struct vm_area_struct
*);
163 extern struct rb_root nommu_region_tree
;
164 extern struct rw_semaphore nommu_region_sem
;
166 extern unsigned int kobjsize(const void *objp
);
170 * vm_flags in vm_area_struct, see mm_types.h.
171 * When changing, update also include/trace/events/mmflags.h
173 #define VM_NONE 0x00000000
175 #define VM_READ 0x00000001 /* currently active flags */
176 #define VM_WRITE 0x00000002
177 #define VM_EXEC 0x00000004
178 #define VM_SHARED 0x00000008
180 /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */
181 #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */
182 #define VM_MAYWRITE 0x00000020
183 #define VM_MAYEXEC 0x00000040
184 #define VM_MAYSHARE 0x00000080
186 #define VM_GROWSDOWN 0x00000100 /* general info on the segment */
187 #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */
188 #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */
189 #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */
190 #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */
192 #define VM_LOCKED 0x00002000
193 #define VM_IO 0x00004000 /* Memory mapped I/O or similar */
195 /* Used by sys_madvise() */
196 #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */
197 #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */
199 #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */
200 #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */
201 #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */
202 #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */
203 #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */
204 #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */
205 #define VM_SYNC 0x00800000 /* Synchronous page faults */
206 #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */
207 #define VM_WIPEONFORK 0x02000000 /* Wipe VMA contents in child. */
208 #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */
210 #ifdef CONFIG_MEM_SOFT_DIRTY
211 # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */
213 # define VM_SOFTDIRTY 0
216 #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */
217 #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */
218 #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */
219 #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */
221 #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS
222 #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */
223 #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */
224 #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */
225 #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */
226 #define VM_HIGH_ARCH_BIT_4 36 /* bit only usable on 64-bit architectures */
227 #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0)
228 #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1)
229 #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2)
230 #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3)
231 #define VM_HIGH_ARCH_4 BIT(VM_HIGH_ARCH_BIT_4)
232 #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */
234 #ifdef CONFIG_ARCH_HAS_PKEYS
235 # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0
236 # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */
237 # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 /* on x86 and 5-bit value on ppc64 */
238 # define VM_PKEY_BIT2 VM_HIGH_ARCH_2
239 # define VM_PKEY_BIT3 VM_HIGH_ARCH_3
241 # define VM_PKEY_BIT4 VM_HIGH_ARCH_4
243 # define VM_PKEY_BIT4 0
245 #endif /* CONFIG_ARCH_HAS_PKEYS */
247 #if defined(CONFIG_X86)
248 # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */
249 #elif defined(CONFIG_PPC)
250 # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */
251 #elif defined(CONFIG_PARISC)
252 # define VM_GROWSUP VM_ARCH_1
253 #elif defined(CONFIG_IA64)
254 # define VM_GROWSUP VM_ARCH_1
255 #elif defined(CONFIG_SPARC64)
256 # define VM_SPARC_ADI VM_ARCH_1 /* Uses ADI tag for access control */
257 # define VM_ARCH_CLEAR VM_SPARC_ADI
258 #elif !defined(CONFIG_MMU)
259 # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */
262 #if defined(CONFIG_X86_INTEL_MPX)
263 /* MPX specific bounds table or bounds directory */
264 # define VM_MPX VM_HIGH_ARCH_4
266 # define VM_MPX VM_NONE
270 # define VM_GROWSUP VM_NONE
273 /* Bits set in the VMA until the stack is in its final location */
274 #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ)
276 #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */
277 #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS
280 #ifdef CONFIG_STACK_GROWSUP
281 #define VM_STACK VM_GROWSUP
283 #define VM_STACK VM_GROWSDOWN
286 #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT)
289 * Special vmas that are non-mergable, non-mlock()able.
290 * Note: mm/huge_memory.c VM_NO_THP depends on this definition.
292 #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP)
294 /* This mask defines which mm->def_flags a process can inherit its parent */
295 #define VM_INIT_DEF_MASK VM_NOHUGEPAGE
297 /* This mask is used to clear all the VMA flags used by mlock */
298 #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT))
300 /* Arch-specific flags to clear when updating VM flags on protection change */
301 #ifndef VM_ARCH_CLEAR
302 # define VM_ARCH_CLEAR VM_NONE
304 #define VM_FLAGS_CLEAR (ARCH_VM_PKEY_FLAGS | VM_ARCH_CLEAR)
307 * mapping from the currently active vm_flags protection bits (the
308 * low four bits) to a page protection mask..
310 extern pgprot_t protection_map
[16];
312 #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */
313 #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */
314 #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */
315 #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */
316 #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */
317 #define FAULT_FLAG_TRIED 0x20 /* Second try */
318 #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */
319 #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */
320 #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */
322 #define FAULT_FLAG_TRACE \
323 { FAULT_FLAG_WRITE, "WRITE" }, \
324 { FAULT_FLAG_MKWRITE, "MKWRITE" }, \
325 { FAULT_FLAG_ALLOW_RETRY, "ALLOW_RETRY" }, \
326 { FAULT_FLAG_RETRY_NOWAIT, "RETRY_NOWAIT" }, \
327 { FAULT_FLAG_KILLABLE, "KILLABLE" }, \
328 { FAULT_FLAG_TRIED, "TRIED" }, \
329 { FAULT_FLAG_USER, "USER" }, \
330 { FAULT_FLAG_REMOTE, "REMOTE" }, \
331 { FAULT_FLAG_INSTRUCTION, "INSTRUCTION" }
334 * vm_fault is filled by the the pagefault handler and passed to the vma's
335 * ->fault function. The vma's ->fault is responsible for returning a bitmask
336 * of VM_FAULT_xxx flags that give details about how the fault was handled.
338 * MM layer fills up gfp_mask for page allocations but fault handler might
339 * alter it if its implementation requires a different allocation context.
341 * pgoff should be used in favour of virtual_address, if possible.
344 struct vm_area_struct
*vma
; /* Target VMA */
345 unsigned int flags
; /* FAULT_FLAG_xxx flags */
346 gfp_t gfp_mask
; /* gfp mask to be used for allocations */
347 pgoff_t pgoff
; /* Logical page offset based on vma */
348 unsigned long address
; /* Faulting virtual address */
349 pmd_t
*pmd
; /* Pointer to pmd entry matching
351 pud_t
*pud
; /* Pointer to pud entry matching
354 pte_t orig_pte
; /* Value of PTE at the time of fault */
356 struct page
*cow_page
; /* Page handler may use for COW fault */
357 struct mem_cgroup
*memcg
; /* Cgroup cow_page belongs to */
358 struct page
*page
; /* ->fault handlers should return a
359 * page here, unless VM_FAULT_NOPAGE
360 * is set (which is also implied by
363 /* These three entries are valid only while holding ptl lock */
364 pte_t
*pte
; /* Pointer to pte entry matching
365 * the 'address'. NULL if the page
366 * table hasn't been allocated.
368 spinlock_t
*ptl
; /* Page table lock.
369 * Protects pte page table if 'pte'
370 * is not NULL, otherwise pmd.
372 pgtable_t prealloc_pte
; /* Pre-allocated pte page table.
373 * vm_ops->map_pages() calls
374 * alloc_set_pte() from atomic context.
375 * do_fault_around() pre-allocates
376 * page table to avoid allocation from
381 /* page entry size for vm->huge_fault() */
382 enum page_entry_size
{
389 * These are the virtual MM functions - opening of an area, closing and
390 * unmapping it (needed to keep files on disk up-to-date etc), pointer
391 * to the functions called when a no-page or a wp-page exception occurs.
393 struct vm_operations_struct
{
394 void (*open
)(struct vm_area_struct
* area
);
395 void (*close
)(struct vm_area_struct
* area
);
396 int (*split
)(struct vm_area_struct
* area
, unsigned long addr
);
397 int (*mremap
)(struct vm_area_struct
* area
);
398 vm_fault_t (*fault
)(struct vm_fault
*vmf
);
399 vm_fault_t (*huge_fault
)(struct vm_fault
*vmf
,
400 enum page_entry_size pe_size
);
401 void (*map_pages
)(struct vm_fault
*vmf
,
402 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
403 unsigned long (*pagesize
)(struct vm_area_struct
* area
);
405 /* notification that a previously read-only page is about to become
406 * writable, if an error is returned it will cause a SIGBUS */
407 vm_fault_t (*page_mkwrite
)(struct vm_fault
*vmf
);
409 /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */
410 vm_fault_t (*pfn_mkwrite
)(struct vm_fault
*vmf
);
412 /* called by access_process_vm when get_user_pages() fails, typically
413 * for use by special VMAs that can switch between memory and hardware
415 int (*access
)(struct vm_area_struct
*vma
, unsigned long addr
,
416 void *buf
, int len
, int write
);
418 /* Called by the /proc/PID/maps code to ask the vma whether it
419 * has a special name. Returning non-NULL will also cause this
420 * vma to be dumped unconditionally. */
421 const char *(*name
)(struct vm_area_struct
*vma
);
425 * set_policy() op must add a reference to any non-NULL @new mempolicy
426 * to hold the policy upon return. Caller should pass NULL @new to
427 * remove a policy and fall back to surrounding context--i.e. do not
428 * install a MPOL_DEFAULT policy, nor the task or system default
431 int (*set_policy
)(struct vm_area_struct
*vma
, struct mempolicy
*new);
434 * get_policy() op must add reference [mpol_get()] to any policy at
435 * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure
436 * in mm/mempolicy.c will do this automatically.
437 * get_policy() must NOT add a ref if the policy at (vma,addr) is not
438 * marked as MPOL_SHARED. vma policies are protected by the mmap_sem.
439 * If no [shared/vma] mempolicy exists at the addr, get_policy() op
440 * must return NULL--i.e., do not "fallback" to task or system default
443 struct mempolicy
*(*get_policy
)(struct vm_area_struct
*vma
,
447 * Called by vm_normal_page() for special PTEs to find the
448 * page for @addr. This is useful if the default behavior
449 * (using pte_page()) would not find the correct page.
451 struct page
*(*find_special_page
)(struct vm_area_struct
*vma
,
455 static inline void vma_init(struct vm_area_struct
*vma
, struct mm_struct
*mm
)
457 static const struct vm_operations_struct dummy_vm_ops
= {};
459 memset(vma
, 0, sizeof(*vma
));
461 vma
->vm_ops
= &dummy_vm_ops
;
462 INIT_LIST_HEAD(&vma
->anon_vma_chain
);
465 static inline void vma_set_anonymous(struct vm_area_struct
*vma
)
470 /* flush_tlb_range() takes a vma, not a mm, and can care about flags */
471 #define TLB_FLUSH_VMA(mm,flags) { .vm_mm = (mm), .vm_flags = (flags) }
476 #define page_private(page) ((page)->private)
477 #define set_page_private(page, v) ((page)->private = (v))
479 #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE)
480 static inline int pmd_devmap(pmd_t pmd
)
484 static inline int pud_devmap(pud_t pud
)
488 static inline int pgd_devmap(pgd_t pgd
)
495 * FIXME: take this include out, include page-flags.h in
496 * files which need it (119 of them)
498 #include <linux/page-flags.h>
499 #include <linux/huge_mm.h>
502 * Methods to modify the page usage count.
504 * What counts for a page usage:
505 * - cache mapping (page->mapping)
506 * - private data (page->private)
507 * - page mapped in a task's page tables, each mapping
508 * is counted separately
510 * Also, many kernel routines increase the page count before a critical
511 * routine so they can be sure the page doesn't go away from under them.
515 * Drop a ref, return true if the refcount fell to zero (the page has no users)
517 static inline int put_page_testzero(struct page
*page
)
519 VM_BUG_ON_PAGE(page_ref_count(page
) == 0, page
);
520 return page_ref_dec_and_test(page
);
524 * Try to grab a ref unless the page has a refcount of zero, return false if
526 * This can be called when MMU is off so it must not access
527 * any of the virtual mappings.
529 static inline int get_page_unless_zero(struct page
*page
)
531 return page_ref_add_unless(page
, 1, 0);
534 extern int page_is_ram(unsigned long pfn
);
542 int region_intersects(resource_size_t offset
, size_t size
, unsigned long flags
,
545 /* Support for virtually mapped pages */
546 struct page
*vmalloc_to_page(const void *addr
);
547 unsigned long vmalloc_to_pfn(const void *addr
);
550 * Determine if an address is within the vmalloc range
552 * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there
553 * is no special casing required.
555 static inline bool is_vmalloc_addr(const void *x
)
558 unsigned long addr
= (unsigned long)x
;
560 return addr
>= VMALLOC_START
&& addr
< VMALLOC_END
;
566 extern int is_vmalloc_or_module_addr(const void *x
);
568 static inline int is_vmalloc_or_module_addr(const void *x
)
574 extern void *kvmalloc_node(size_t size
, gfp_t flags
, int node
);
575 static inline void *kvmalloc(size_t size
, gfp_t flags
)
577 return kvmalloc_node(size
, flags
, NUMA_NO_NODE
);
579 static inline void *kvzalloc_node(size_t size
, gfp_t flags
, int node
)
581 return kvmalloc_node(size
, flags
| __GFP_ZERO
, node
);
583 static inline void *kvzalloc(size_t size
, gfp_t flags
)
585 return kvmalloc(size
, flags
| __GFP_ZERO
);
588 static inline void *kvmalloc_array(size_t n
, size_t size
, gfp_t flags
)
592 if (unlikely(check_mul_overflow(n
, size
, &bytes
)))
595 return kvmalloc(bytes
, flags
);
598 static inline void *kvcalloc(size_t n
, size_t size
, gfp_t flags
)
600 return kvmalloc_array(n
, size
, flags
| __GFP_ZERO
);
603 extern void kvfree(const void *addr
);
605 static inline atomic_t
*compound_mapcount_ptr(struct page
*page
)
607 return &page
[1].compound_mapcount
;
610 static inline int compound_mapcount(struct page
*page
)
612 VM_BUG_ON_PAGE(!PageCompound(page
), page
);
613 page
= compound_head(page
);
614 return atomic_read(compound_mapcount_ptr(page
)) + 1;
618 * The atomic page->_mapcount, starts from -1: so that transitions
619 * both from it and to it can be tracked, using atomic_inc_and_test
620 * and atomic_add_negative(-1).
622 static inline void page_mapcount_reset(struct page
*page
)
624 atomic_set(&(page
)->_mapcount
, -1);
627 int __page_mapcount(struct page
*page
);
629 static inline int page_mapcount(struct page
*page
)
631 VM_BUG_ON_PAGE(PageSlab(page
), page
);
633 if (unlikely(PageCompound(page
)))
634 return __page_mapcount(page
);
635 return atomic_read(&page
->_mapcount
) + 1;
638 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
639 int total_mapcount(struct page
*page
);
640 int page_trans_huge_mapcount(struct page
*page
, int *total_mapcount
);
642 static inline int total_mapcount(struct page
*page
)
644 return page_mapcount(page
);
646 static inline int page_trans_huge_mapcount(struct page
*page
,
649 int mapcount
= page_mapcount(page
);
651 *total_mapcount
= mapcount
;
656 static inline struct page
*virt_to_head_page(const void *x
)
658 struct page
*page
= virt_to_page(x
);
660 return compound_head(page
);
663 void __put_page(struct page
*page
);
665 void put_pages_list(struct list_head
*pages
);
667 void split_page(struct page
*page
, unsigned int order
);
670 * Compound pages have a destructor function. Provide a
671 * prototype for that function and accessor functions.
672 * These are _only_ valid on the head of a compound page.
674 typedef void compound_page_dtor(struct page
*);
676 /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */
677 enum compound_dtor_id
{
680 #ifdef CONFIG_HUGETLB_PAGE
683 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
688 extern compound_page_dtor
* const compound_page_dtors
[];
690 static inline void set_compound_page_dtor(struct page
*page
,
691 enum compound_dtor_id compound_dtor
)
693 VM_BUG_ON_PAGE(compound_dtor
>= NR_COMPOUND_DTORS
, page
);
694 page
[1].compound_dtor
= compound_dtor
;
697 static inline compound_page_dtor
*get_compound_page_dtor(struct page
*page
)
699 VM_BUG_ON_PAGE(page
[1].compound_dtor
>= NR_COMPOUND_DTORS
, page
);
700 return compound_page_dtors
[page
[1].compound_dtor
];
703 static inline unsigned int compound_order(struct page
*page
)
707 return page
[1].compound_order
;
710 static inline void set_compound_order(struct page
*page
, unsigned int order
)
712 page
[1].compound_order
= order
;
715 void free_compound_page(struct page
*page
);
719 * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when
720 * servicing faults for write access. In the normal case, do always want
721 * pte_mkwrite. But get_user_pages can cause write faults for mappings
722 * that do not have writing enabled, when used by access_process_vm.
724 static inline pte_t
maybe_mkwrite(pte_t pte
, struct vm_area_struct
*vma
)
726 if (likely(vma
->vm_flags
& VM_WRITE
))
727 pte
= pte_mkwrite(pte
);
731 vm_fault_t
alloc_set_pte(struct vm_fault
*vmf
, struct mem_cgroup
*memcg
,
733 vm_fault_t
finish_fault(struct vm_fault
*vmf
);
734 vm_fault_t
finish_mkwrite_fault(struct vm_fault
*vmf
);
738 * Multiple processes may "see" the same page. E.g. for untouched
739 * mappings of /dev/null, all processes see the same page full of
740 * zeroes, and text pages of executables and shared libraries have
741 * only one copy in memory, at most, normally.
743 * For the non-reserved pages, page_count(page) denotes a reference count.
744 * page_count() == 0 means the page is free. page->lru is then used for
745 * freelist management in the buddy allocator.
746 * page_count() > 0 means the page has been allocated.
748 * Pages are allocated by the slab allocator in order to provide memory
749 * to kmalloc and kmem_cache_alloc. In this case, the management of the
750 * page, and the fields in 'struct page' are the responsibility of mm/slab.c
751 * unless a particular usage is carefully commented. (the responsibility of
752 * freeing the kmalloc memory is the caller's, of course).
754 * A page may be used by anyone else who does a __get_free_page().
755 * In this case, page_count still tracks the references, and should only
756 * be used through the normal accessor functions. The top bits of page->flags
757 * and page->virtual store page management information, but all other fields
758 * are unused and could be used privately, carefully. The management of this
759 * page is the responsibility of the one who allocated it, and those who have
760 * subsequently been given references to it.
762 * The other pages (we may call them "pagecache pages") are completely
763 * managed by the Linux memory manager: I/O, buffers, swapping etc.
764 * The following discussion applies only to them.
766 * A pagecache page contains an opaque `private' member, which belongs to the
767 * page's address_space. Usually, this is the address of a circular list of
768 * the page's disk buffers. PG_private must be set to tell the VM to call
769 * into the filesystem to release these pages.
771 * A page may belong to an inode's memory mapping. In this case, page->mapping
772 * is the pointer to the inode, and page->index is the file offset of the page,
773 * in units of PAGE_SIZE.
775 * If pagecache pages are not associated with an inode, they are said to be
776 * anonymous pages. These may become associated with the swapcache, and in that
777 * case PG_swapcache is set, and page->private is an offset into the swapcache.
779 * In either case (swapcache or inode backed), the pagecache itself holds one
780 * reference to the page. Setting PG_private should also increment the
781 * refcount. The each user mapping also has a reference to the page.
783 * The pagecache pages are stored in a per-mapping radix tree, which is
784 * rooted at mapping->i_pages, and indexed by offset.
785 * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space
786 * lists, we instead now tag pages as dirty/writeback in the radix tree.
788 * All pagecache pages may be subject to I/O:
789 * - inode pages may need to be read from disk,
790 * - inode pages which have been modified and are MAP_SHARED may need
791 * to be written back to the inode on disk,
792 * - anonymous pages (including MAP_PRIVATE file mappings) which have been
793 * modified may need to be swapped out to swap space and (later) to be read
798 * The zone field is never updated after free_area_init_core()
799 * sets it, so none of the operations on it need to be atomic.
802 /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */
803 #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH)
804 #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH)
805 #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH)
806 #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH)
809 * Define the bit shifts to access each section. For non-existent
810 * sections we define the shift as 0; that plus a 0 mask ensures
811 * the compiler will optimise away reference to them.
813 #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0))
814 #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0))
815 #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0))
816 #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0))
818 /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */
819 #ifdef NODE_NOT_IN_PAGE_FLAGS
820 #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT)
821 #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \
822 SECTIONS_PGOFF : ZONES_PGOFF)
824 #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT)
825 #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \
826 NODES_PGOFF : ZONES_PGOFF)
829 #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0))
831 #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
832 #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS
835 #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1)
836 #define NODES_MASK ((1UL << NODES_WIDTH) - 1)
837 #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1)
838 #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1)
839 #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1)
841 static inline enum zone_type
page_zonenum(const struct page
*page
)
843 return (page
->flags
>> ZONES_PGSHIFT
) & ZONES_MASK
;
846 #ifdef CONFIG_ZONE_DEVICE
847 static inline bool is_zone_device_page(const struct page
*page
)
849 return page_zonenum(page
) == ZONE_DEVICE
;
851 extern void memmap_init_zone_device(struct zone
*, unsigned long,
852 unsigned long, struct dev_pagemap
*);
854 static inline bool is_zone_device_page(const struct page
*page
)
860 #ifdef CONFIG_DEV_PAGEMAP_OPS
861 void dev_pagemap_get_ops(void);
862 void dev_pagemap_put_ops(void);
863 void __put_devmap_managed_page(struct page
*page
);
864 DECLARE_STATIC_KEY_FALSE(devmap_managed_key
);
865 static inline bool put_devmap_managed_page(struct page
*page
)
867 if (!static_branch_unlikely(&devmap_managed_key
))
869 if (!is_zone_device_page(page
))
871 switch (page
->pgmap
->type
) {
872 case MEMORY_DEVICE_PRIVATE
:
873 case MEMORY_DEVICE_PUBLIC
:
874 case MEMORY_DEVICE_FS_DAX
:
875 __put_devmap_managed_page(page
);
883 static inline bool is_device_private_page(const struct page
*page
)
885 return is_zone_device_page(page
) &&
886 page
->pgmap
->type
== MEMORY_DEVICE_PRIVATE
;
889 static inline bool is_device_public_page(const struct page
*page
)
891 return is_zone_device_page(page
) &&
892 page
->pgmap
->type
== MEMORY_DEVICE_PUBLIC
;
895 #ifdef CONFIG_PCI_P2PDMA
896 static inline bool is_pci_p2pdma_page(const struct page
*page
)
898 return is_zone_device_page(page
) &&
899 page
->pgmap
->type
== MEMORY_DEVICE_PCI_P2PDMA
;
901 #else /* CONFIG_PCI_P2PDMA */
902 static inline bool is_pci_p2pdma_page(const struct page
*page
)
906 #endif /* CONFIG_PCI_P2PDMA */
908 #else /* CONFIG_DEV_PAGEMAP_OPS */
909 static inline void dev_pagemap_get_ops(void)
913 static inline void dev_pagemap_put_ops(void)
917 static inline bool put_devmap_managed_page(struct page
*page
)
922 static inline bool is_device_private_page(const struct page
*page
)
927 static inline bool is_device_public_page(const struct page
*page
)
932 static inline bool is_pci_p2pdma_page(const struct page
*page
)
936 #endif /* CONFIG_DEV_PAGEMAP_OPS */
938 static inline void get_page(struct page
*page
)
940 page
= compound_head(page
);
942 * Getting a normal page or the head of a compound page
943 * requires to already have an elevated page->_refcount.
945 VM_BUG_ON_PAGE(page_ref_count(page
) <= 0, page
);
949 static inline void put_page(struct page
*page
)
951 page
= compound_head(page
);
954 * For devmap managed pages we need to catch refcount transition from
955 * 2 to 1, when refcount reach one it means the page is free and we
956 * need to inform the device driver through callback. See
957 * include/linux/memremap.h and HMM for details.
959 if (put_devmap_managed_page(page
))
962 if (put_page_testzero(page
))
966 #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP)
967 #define SECTION_IN_PAGE_FLAGS
971 * The identification function is mainly used by the buddy allocator for
972 * determining if two pages could be buddies. We are not really identifying
973 * the zone since we could be using the section number id if we do not have
974 * node id available in page flags.
975 * We only guarantee that it will return the same value for two combinable
978 static inline int page_zone_id(struct page
*page
)
980 return (page
->flags
>> ZONEID_PGSHIFT
) & ZONEID_MASK
;
983 #ifdef NODE_NOT_IN_PAGE_FLAGS
984 extern int page_to_nid(const struct page
*page
);
986 static inline int page_to_nid(const struct page
*page
)
988 struct page
*p
= (struct page
*)page
;
990 return (PF_POISONED_CHECK(p
)->flags
>> NODES_PGSHIFT
) & NODES_MASK
;
994 #ifdef CONFIG_NUMA_BALANCING
995 static inline int cpu_pid_to_cpupid(int cpu
, int pid
)
997 return ((cpu
& LAST__CPU_MASK
) << LAST__PID_SHIFT
) | (pid
& LAST__PID_MASK
);
1000 static inline int cpupid_to_pid(int cpupid
)
1002 return cpupid
& LAST__PID_MASK
;
1005 static inline int cpupid_to_cpu(int cpupid
)
1007 return (cpupid
>> LAST__PID_SHIFT
) & LAST__CPU_MASK
;
1010 static inline int cpupid_to_nid(int cpupid
)
1012 return cpu_to_node(cpupid_to_cpu(cpupid
));
1015 static inline bool cpupid_pid_unset(int cpupid
)
1017 return cpupid_to_pid(cpupid
) == (-1 & LAST__PID_MASK
);
1020 static inline bool cpupid_cpu_unset(int cpupid
)
1022 return cpupid_to_cpu(cpupid
) == (-1 & LAST__CPU_MASK
);
1025 static inline bool __cpupid_match_pid(pid_t task_pid
, int cpupid
)
1027 return (task_pid
& LAST__PID_MASK
) == cpupid_to_pid(cpupid
);
1030 #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid)
1031 #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS
1032 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1034 return xchg(&page
->_last_cpupid
, cpupid
& LAST_CPUPID_MASK
);
1037 static inline int page_cpupid_last(struct page
*page
)
1039 return page
->_last_cpupid
;
1041 static inline void page_cpupid_reset_last(struct page
*page
)
1043 page
->_last_cpupid
= -1 & LAST_CPUPID_MASK
;
1046 static inline int page_cpupid_last(struct page
*page
)
1048 return (page
->flags
>> LAST_CPUPID_PGSHIFT
) & LAST_CPUPID_MASK
;
1051 extern int page_cpupid_xchg_last(struct page
*page
, int cpupid
);
1053 static inline void page_cpupid_reset_last(struct page
*page
)
1055 page
->flags
|= LAST_CPUPID_MASK
<< LAST_CPUPID_PGSHIFT
;
1057 #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */
1058 #else /* !CONFIG_NUMA_BALANCING */
1059 static inline int page_cpupid_xchg_last(struct page
*page
, int cpupid
)
1061 return page_to_nid(page
); /* XXX */
1064 static inline int page_cpupid_last(struct page
*page
)
1066 return page_to_nid(page
); /* XXX */
1069 static inline int cpupid_to_nid(int cpupid
)
1074 static inline int cpupid_to_pid(int cpupid
)
1079 static inline int cpupid_to_cpu(int cpupid
)
1084 static inline int cpu_pid_to_cpupid(int nid
, int pid
)
1089 static inline bool cpupid_pid_unset(int cpupid
)
1094 static inline void page_cpupid_reset_last(struct page
*page
)
1098 static inline bool cpupid_match_pid(struct task_struct
*task
, int cpupid
)
1102 #endif /* CONFIG_NUMA_BALANCING */
1104 static inline struct zone
*page_zone(const struct page
*page
)
1106 return &NODE_DATA(page_to_nid(page
))->node_zones
[page_zonenum(page
)];
1109 static inline pg_data_t
*page_pgdat(const struct page
*page
)
1111 return NODE_DATA(page_to_nid(page
));
1114 #ifdef SECTION_IN_PAGE_FLAGS
1115 static inline void set_page_section(struct page
*page
, unsigned long section
)
1117 page
->flags
&= ~(SECTIONS_MASK
<< SECTIONS_PGSHIFT
);
1118 page
->flags
|= (section
& SECTIONS_MASK
) << SECTIONS_PGSHIFT
;
1121 static inline unsigned long page_to_section(const struct page
*page
)
1123 return (page
->flags
>> SECTIONS_PGSHIFT
) & SECTIONS_MASK
;
1127 static inline void set_page_zone(struct page
*page
, enum zone_type zone
)
1129 page
->flags
&= ~(ZONES_MASK
<< ZONES_PGSHIFT
);
1130 page
->flags
|= (zone
& ZONES_MASK
) << ZONES_PGSHIFT
;
1133 static inline void set_page_node(struct page
*page
, unsigned long node
)
1135 page
->flags
&= ~(NODES_MASK
<< NODES_PGSHIFT
);
1136 page
->flags
|= (node
& NODES_MASK
) << NODES_PGSHIFT
;
1139 static inline void set_page_links(struct page
*page
, enum zone_type zone
,
1140 unsigned long node
, unsigned long pfn
)
1142 set_page_zone(page
, zone
);
1143 set_page_node(page
, node
);
1144 #ifdef SECTION_IN_PAGE_FLAGS
1145 set_page_section(page
, pfn_to_section_nr(pfn
));
1150 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
1152 return page
->mem_cgroup
;
1154 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1156 WARN_ON_ONCE(!rcu_read_lock_held());
1157 return READ_ONCE(page
->mem_cgroup
);
1160 static inline struct mem_cgroup
*page_memcg(struct page
*page
)
1164 static inline struct mem_cgroup
*page_memcg_rcu(struct page
*page
)
1166 WARN_ON_ONCE(!rcu_read_lock_held());
1172 * Some inline functions in vmstat.h depend on page_zone()
1174 #include <linux/vmstat.h>
1176 static __always_inline
void *lowmem_page_address(const struct page
*page
)
1178 return page_to_virt(page
);
1181 #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL)
1182 #define HASHED_PAGE_VIRTUAL
1185 #if defined(WANT_PAGE_VIRTUAL)
1186 static inline void *page_address(const struct page
*page
)
1188 return page
->virtual;
1190 static inline void set_page_address(struct page
*page
, void *address
)
1192 page
->virtual = address
;
1194 #define page_address_init() do { } while(0)
1197 #if defined(HASHED_PAGE_VIRTUAL)
1198 void *page_address(const struct page
*page
);
1199 void set_page_address(struct page
*page
, void *virtual);
1200 void page_address_init(void);
1203 #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL)
1204 #define page_address(page) lowmem_page_address(page)
1205 #define set_page_address(page, address) do { } while(0)
1206 #define page_address_init() do { } while(0)
1209 extern void *page_rmapping(struct page
*page
);
1210 extern struct anon_vma
*page_anon_vma(struct page
*page
);
1211 extern struct address_space
*page_mapping(struct page
*page
);
1213 extern struct address_space
*__page_file_mapping(struct page
*);
1216 struct address_space
*page_file_mapping(struct page
*page
)
1218 if (unlikely(PageSwapCache(page
)))
1219 return __page_file_mapping(page
);
1221 return page
->mapping
;
1224 extern pgoff_t
__page_file_index(struct page
*page
);
1227 * Return the pagecache index of the passed page. Regular pagecache pages
1228 * use ->index whereas swapcache pages use swp_offset(->private)
1230 static inline pgoff_t
page_index(struct page
*page
)
1232 if (unlikely(PageSwapCache(page
)))
1233 return __page_file_index(page
);
1237 bool page_mapped(struct page
*page
);
1238 struct address_space
*page_mapping(struct page
*page
);
1239 struct address_space
*page_mapping_file(struct page
*page
);
1242 * Return true only if the page has been allocated with
1243 * ALLOC_NO_WATERMARKS and the low watermark was not
1244 * met implying that the system is under some pressure.
1246 static inline bool page_is_pfmemalloc(struct page
*page
)
1249 * Page index cannot be this large so this must be
1250 * a pfmemalloc page.
1252 return page
->index
== -1UL;
1256 * Only to be called by the page allocator on a freshly allocated
1259 static inline void set_page_pfmemalloc(struct page
*page
)
1264 static inline void clear_page_pfmemalloc(struct page
*page
)
1270 * Different kinds of faults, as returned by handle_mm_fault().
1271 * Used to decide whether a process gets delivered SIGBUS or
1272 * just gets major/minor fault counters bumped up.
1275 #define VM_FAULT_OOM 0x0001
1276 #define VM_FAULT_SIGBUS 0x0002
1277 #define VM_FAULT_MAJOR 0x0004
1278 #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */
1279 #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */
1280 #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */
1281 #define VM_FAULT_SIGSEGV 0x0040
1283 #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */
1284 #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */
1285 #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */
1286 #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */
1287 #define VM_FAULT_DONE_COW 0x1000 /* ->fault has fully handled COW */
1288 #define VM_FAULT_NEEDDSYNC 0x2000 /* ->fault did not modify page tables
1289 * and needs fsync() to complete (for
1290 * synchronous page faults in DAX) */
1292 #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \
1293 VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \
1296 #define VM_FAULT_RESULT_TRACE \
1297 { VM_FAULT_OOM, "OOM" }, \
1298 { VM_FAULT_SIGBUS, "SIGBUS" }, \
1299 { VM_FAULT_MAJOR, "MAJOR" }, \
1300 { VM_FAULT_WRITE, "WRITE" }, \
1301 { VM_FAULT_HWPOISON, "HWPOISON" }, \
1302 { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \
1303 { VM_FAULT_SIGSEGV, "SIGSEGV" }, \
1304 { VM_FAULT_NOPAGE, "NOPAGE" }, \
1305 { VM_FAULT_LOCKED, "LOCKED" }, \
1306 { VM_FAULT_RETRY, "RETRY" }, \
1307 { VM_FAULT_FALLBACK, "FALLBACK" }, \
1308 { VM_FAULT_DONE_COW, "DONE_COW" }, \
1309 { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" }
1311 /* Encode hstate index for a hwpoisoned large page */
1312 #define VM_FAULT_SET_HINDEX(x) ((x) << 12)
1313 #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf)
1316 * Can be called by the pagefault handler when it gets a VM_FAULT_OOM.
1318 extern void pagefault_out_of_memory(void);
1320 #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK)
1323 * Flags passed to show_mem() and show_free_areas() to suppress output in
1326 #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */
1328 extern void show_free_areas(unsigned int flags
, nodemask_t
*nodemask
);
1330 extern bool can_do_mlock(void);
1331 extern int user_shm_lock(size_t, struct user_struct
*);
1332 extern void user_shm_unlock(size_t, struct user_struct
*);
1335 * Parameter block passed down to zap_pte_range in exceptional cases.
1337 struct zap_details
{
1338 struct address_space
*check_mapping
; /* Check page->mapping if set */
1339 pgoff_t first_index
; /* Lowest page->index to unmap */
1340 pgoff_t last_index
; /* Highest page->index to unmap */
1343 struct page
*_vm_normal_page(struct vm_area_struct
*vma
, unsigned long addr
,
1344 pte_t pte
, bool with_public_device
);
1345 #define vm_normal_page(vma, addr, pte) _vm_normal_page(vma, addr, pte, false)
1347 struct page
*vm_normal_page_pmd(struct vm_area_struct
*vma
, unsigned long addr
,
1350 void zap_vma_ptes(struct vm_area_struct
*vma
, unsigned long address
,
1351 unsigned long size
);
1352 void zap_page_range(struct vm_area_struct
*vma
, unsigned long address
,
1353 unsigned long size
);
1354 void unmap_vmas(struct mmu_gather
*tlb
, struct vm_area_struct
*start_vma
,
1355 unsigned long start
, unsigned long end
);
1358 * mm_walk - callbacks for walk_page_range
1359 * @pud_entry: if set, called for each non-empty PUD (2nd-level) entry
1360 * this handler should only handle pud_trans_huge() puds.
1361 * the pmd_entry or pte_entry callbacks will be used for
1363 * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry
1364 * this handler is required to be able to handle
1365 * pmd_trans_huge() pmds. They may simply choose to
1366 * split_huge_page() instead of handling it explicitly.
1367 * @pte_entry: if set, called for each non-empty PTE (4th-level) entry
1368 * @pte_hole: if set, called for each hole at all levels
1369 * @hugetlb_entry: if set, called for each hugetlb entry
1370 * @test_walk: caller specific callback function to determine whether
1371 * we walk over the current vma or not. Returning 0
1372 * value means "do page table walk over the current vma,"
1373 * and a negative one means "abort current page table walk
1374 * right now." 1 means "skip the current vma."
1375 * @mm: mm_struct representing the target process of page table walk
1376 * @vma: vma currently walked (NULL if walking outside vmas)
1377 * @private: private data for callbacks' usage
1379 * (see the comment on walk_page_range() for more details)
1382 int (*pud_entry
)(pud_t
*pud
, unsigned long addr
,
1383 unsigned long next
, struct mm_walk
*walk
);
1384 int (*pmd_entry
)(pmd_t
*pmd
, unsigned long addr
,
1385 unsigned long next
, struct mm_walk
*walk
);
1386 int (*pte_entry
)(pte_t
*pte
, unsigned long addr
,
1387 unsigned long next
, struct mm_walk
*walk
);
1388 int (*pte_hole
)(unsigned long addr
, unsigned long next
,
1389 struct mm_walk
*walk
);
1390 int (*hugetlb_entry
)(pte_t
*pte
, unsigned long hmask
,
1391 unsigned long addr
, unsigned long next
,
1392 struct mm_walk
*walk
);
1393 int (*test_walk
)(unsigned long addr
, unsigned long next
,
1394 struct mm_walk
*walk
);
1395 struct mm_struct
*mm
;
1396 struct vm_area_struct
*vma
;
1400 int walk_page_range(unsigned long addr
, unsigned long end
,
1401 struct mm_walk
*walk
);
1402 int walk_page_vma(struct vm_area_struct
*vma
, struct mm_walk
*walk
);
1403 void free_pgd_range(struct mmu_gather
*tlb
, unsigned long addr
,
1404 unsigned long end
, unsigned long floor
, unsigned long ceiling
);
1405 int copy_page_range(struct mm_struct
*dst
, struct mm_struct
*src
,
1406 struct vm_area_struct
*vma
);
1407 int follow_pte_pmd(struct mm_struct
*mm
, unsigned long address
,
1408 unsigned long *start
, unsigned long *end
,
1409 pte_t
**ptepp
, pmd_t
**pmdpp
, spinlock_t
**ptlp
);
1410 int follow_pfn(struct vm_area_struct
*vma
, unsigned long address
,
1411 unsigned long *pfn
);
1412 int follow_phys(struct vm_area_struct
*vma
, unsigned long address
,
1413 unsigned int flags
, unsigned long *prot
, resource_size_t
*phys
);
1414 int generic_access_phys(struct vm_area_struct
*vma
, unsigned long addr
,
1415 void *buf
, int len
, int write
);
1417 extern void truncate_pagecache(struct inode
*inode
, loff_t
new);
1418 extern void truncate_setsize(struct inode
*inode
, loff_t newsize
);
1419 void pagecache_isize_extended(struct inode
*inode
, loff_t from
, loff_t to
);
1420 void truncate_pagecache_range(struct inode
*inode
, loff_t offset
, loff_t end
);
1421 int truncate_inode_page(struct address_space
*mapping
, struct page
*page
);
1422 int generic_error_remove_page(struct address_space
*mapping
, struct page
*page
);
1423 int invalidate_inode_page(struct page
*page
);
1426 extern vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
1427 unsigned long address
, unsigned int flags
);
1428 extern int fixup_user_fault(struct task_struct
*tsk
, struct mm_struct
*mm
,
1429 unsigned long address
, unsigned int fault_flags
,
1431 void unmap_mapping_pages(struct address_space
*mapping
,
1432 pgoff_t start
, pgoff_t nr
, bool even_cows
);
1433 void unmap_mapping_range(struct address_space
*mapping
,
1434 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
);
1436 static inline vm_fault_t
handle_mm_fault(struct vm_area_struct
*vma
,
1437 unsigned long address
, unsigned int flags
)
1439 /* should never happen if there's no MMU */
1441 return VM_FAULT_SIGBUS
;
1443 static inline int fixup_user_fault(struct task_struct
*tsk
,
1444 struct mm_struct
*mm
, unsigned long address
,
1445 unsigned int fault_flags
, bool *unlocked
)
1447 /* should never happen if there's no MMU */
1451 static inline void unmap_mapping_pages(struct address_space
*mapping
,
1452 pgoff_t start
, pgoff_t nr
, bool even_cows
) { }
1453 static inline void unmap_mapping_range(struct address_space
*mapping
,
1454 loff_t
const holebegin
, loff_t
const holelen
, int even_cows
) { }
1457 static inline void unmap_shared_mapping_range(struct address_space
*mapping
,
1458 loff_t
const holebegin
, loff_t
const holelen
)
1460 unmap_mapping_range(mapping
, holebegin
, holelen
, 0);
1463 extern int access_process_vm(struct task_struct
*tsk
, unsigned long addr
,
1464 void *buf
, int len
, unsigned int gup_flags
);
1465 extern int access_remote_vm(struct mm_struct
*mm
, unsigned long addr
,
1466 void *buf
, int len
, unsigned int gup_flags
);
1467 extern int __access_remote_vm(struct task_struct
*tsk
, struct mm_struct
*mm
,
1468 unsigned long addr
, void *buf
, int len
, unsigned int gup_flags
);
1470 long get_user_pages_remote(struct task_struct
*tsk
, struct mm_struct
*mm
,
1471 unsigned long start
, unsigned long nr_pages
,
1472 unsigned int gup_flags
, struct page
**pages
,
1473 struct vm_area_struct
**vmas
, int *locked
);
1474 long get_user_pages(unsigned long start
, unsigned long nr_pages
,
1475 unsigned int gup_flags
, struct page
**pages
,
1476 struct vm_area_struct
**vmas
);
1477 long get_user_pages_locked(unsigned long start
, unsigned long nr_pages
,
1478 unsigned int gup_flags
, struct page
**pages
, int *locked
);
1479 long get_user_pages_unlocked(unsigned long start
, unsigned long nr_pages
,
1480 struct page
**pages
, unsigned int gup_flags
);
1481 #ifdef CONFIG_FS_DAX
1482 long get_user_pages_longterm(unsigned long start
, unsigned long nr_pages
,
1483 unsigned int gup_flags
, struct page
**pages
,
1484 struct vm_area_struct
**vmas
);
1486 static inline long get_user_pages_longterm(unsigned long start
,
1487 unsigned long nr_pages
, unsigned int gup_flags
,
1488 struct page
**pages
, struct vm_area_struct
**vmas
)
1490 return get_user_pages(start
, nr_pages
, gup_flags
, pages
, vmas
);
1492 #endif /* CONFIG_FS_DAX */
1494 int get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1495 struct page
**pages
);
1497 /* Container for pinned pfns / pages */
1498 struct frame_vector
{
1499 unsigned int nr_allocated
; /* Number of frames we have space for */
1500 unsigned int nr_frames
; /* Number of frames stored in ptrs array */
1501 bool got_ref
; /* Did we pin pages by getting page ref? */
1502 bool is_pfns
; /* Does array contain pages or pfns? */
1503 void *ptrs
[0]; /* Array of pinned pfns / pages. Use
1504 * pfns_vector_pages() or pfns_vector_pfns()
1508 struct frame_vector
*frame_vector_create(unsigned int nr_frames
);
1509 void frame_vector_destroy(struct frame_vector
*vec
);
1510 int get_vaddr_frames(unsigned long start
, unsigned int nr_pfns
,
1511 unsigned int gup_flags
, struct frame_vector
*vec
);
1512 void put_vaddr_frames(struct frame_vector
*vec
);
1513 int frame_vector_to_pages(struct frame_vector
*vec
);
1514 void frame_vector_to_pfns(struct frame_vector
*vec
);
1516 static inline unsigned int frame_vector_count(struct frame_vector
*vec
)
1518 return vec
->nr_frames
;
1521 static inline struct page
**frame_vector_pages(struct frame_vector
*vec
)
1524 int err
= frame_vector_to_pages(vec
);
1527 return ERR_PTR(err
);
1529 return (struct page
**)(vec
->ptrs
);
1532 static inline unsigned long *frame_vector_pfns(struct frame_vector
*vec
)
1535 frame_vector_to_pfns(vec
);
1536 return (unsigned long *)(vec
->ptrs
);
1540 int get_kernel_pages(const struct kvec
*iov
, int nr_pages
, int write
,
1541 struct page
**pages
);
1542 int get_kernel_page(unsigned long start
, int write
, struct page
**pages
);
1543 struct page
*get_dump_page(unsigned long addr
);
1545 extern int try_to_release_page(struct page
* page
, gfp_t gfp_mask
);
1546 extern void do_invalidatepage(struct page
*page
, unsigned int offset
,
1547 unsigned int length
);
1549 void __set_page_dirty(struct page
*, struct address_space
*, int warn
);
1550 int __set_page_dirty_nobuffers(struct page
*page
);
1551 int __set_page_dirty_no_writeback(struct page
*page
);
1552 int redirty_page_for_writepage(struct writeback_control
*wbc
,
1554 void account_page_dirtied(struct page
*page
, struct address_space
*mapping
);
1555 void account_page_cleaned(struct page
*page
, struct address_space
*mapping
,
1556 struct bdi_writeback
*wb
);
1557 int set_page_dirty(struct page
*page
);
1558 int set_page_dirty_lock(struct page
*page
);
1559 void __cancel_dirty_page(struct page
*page
);
1560 static inline void cancel_dirty_page(struct page
*page
)
1562 /* Avoid atomic ops, locking, etc. when not actually needed. */
1563 if (PageDirty(page
))
1564 __cancel_dirty_page(page
);
1566 int clear_page_dirty_for_io(struct page
*page
);
1568 int get_cmdline(struct task_struct
*task
, char *buffer
, int buflen
);
1570 static inline bool vma_is_anonymous(struct vm_area_struct
*vma
)
1572 return !vma
->vm_ops
;
1577 * The vma_is_shmem is not inline because it is used only by slow
1578 * paths in userfault.
1580 bool vma_is_shmem(struct vm_area_struct
*vma
);
1582 static inline bool vma_is_shmem(struct vm_area_struct
*vma
) { return false; }
1585 int vma_is_stack_for_current(struct vm_area_struct
*vma
);
1587 extern unsigned long move_page_tables(struct vm_area_struct
*vma
,
1588 unsigned long old_addr
, struct vm_area_struct
*new_vma
,
1589 unsigned long new_addr
, unsigned long len
,
1590 bool need_rmap_locks
);
1591 extern unsigned long change_protection(struct vm_area_struct
*vma
, unsigned long start
,
1592 unsigned long end
, pgprot_t newprot
,
1593 int dirty_accountable
, int prot_numa
);
1594 extern int mprotect_fixup(struct vm_area_struct
*vma
,
1595 struct vm_area_struct
**pprev
, unsigned long start
,
1596 unsigned long end
, unsigned long newflags
);
1599 * doesn't attempt to fault and will return short.
1601 int __get_user_pages_fast(unsigned long start
, int nr_pages
, int write
,
1602 struct page
**pages
);
1604 * per-process(per-mm_struct) statistics.
1606 static inline unsigned long get_mm_counter(struct mm_struct
*mm
, int member
)
1608 long val
= atomic_long_read(&mm
->rss_stat
.count
[member
]);
1610 #ifdef SPLIT_RSS_COUNTING
1612 * counter is updated in asynchronous manner and may go to minus.
1613 * But it's never be expected number for users.
1618 return (unsigned long)val
;
1621 static inline void add_mm_counter(struct mm_struct
*mm
, int member
, long value
)
1623 atomic_long_add(value
, &mm
->rss_stat
.count
[member
]);
1626 static inline void inc_mm_counter(struct mm_struct
*mm
, int member
)
1628 atomic_long_inc(&mm
->rss_stat
.count
[member
]);
1631 static inline void dec_mm_counter(struct mm_struct
*mm
, int member
)
1633 atomic_long_dec(&mm
->rss_stat
.count
[member
]);
1636 /* Optimized variant when page is already known not to be PageAnon */
1637 static inline int mm_counter_file(struct page
*page
)
1639 if (PageSwapBacked(page
))
1640 return MM_SHMEMPAGES
;
1641 return MM_FILEPAGES
;
1644 static inline int mm_counter(struct page
*page
)
1647 return MM_ANONPAGES
;
1648 return mm_counter_file(page
);
1651 static inline unsigned long get_mm_rss(struct mm_struct
*mm
)
1653 return get_mm_counter(mm
, MM_FILEPAGES
) +
1654 get_mm_counter(mm
, MM_ANONPAGES
) +
1655 get_mm_counter(mm
, MM_SHMEMPAGES
);
1658 static inline unsigned long get_mm_hiwater_rss(struct mm_struct
*mm
)
1660 return max(mm
->hiwater_rss
, get_mm_rss(mm
));
1663 static inline unsigned long get_mm_hiwater_vm(struct mm_struct
*mm
)
1665 return max(mm
->hiwater_vm
, mm
->total_vm
);
1668 static inline void update_hiwater_rss(struct mm_struct
*mm
)
1670 unsigned long _rss
= get_mm_rss(mm
);
1672 if ((mm
)->hiwater_rss
< _rss
)
1673 (mm
)->hiwater_rss
= _rss
;
1676 static inline void update_hiwater_vm(struct mm_struct
*mm
)
1678 if (mm
->hiwater_vm
< mm
->total_vm
)
1679 mm
->hiwater_vm
= mm
->total_vm
;
1682 static inline void reset_mm_hiwater_rss(struct mm_struct
*mm
)
1684 mm
->hiwater_rss
= get_mm_rss(mm
);
1687 static inline void setmax_mm_hiwater_rss(unsigned long *maxrss
,
1688 struct mm_struct
*mm
)
1690 unsigned long hiwater_rss
= get_mm_hiwater_rss(mm
);
1692 if (*maxrss
< hiwater_rss
)
1693 *maxrss
= hiwater_rss
;
1696 #if defined(SPLIT_RSS_COUNTING)
1697 void sync_mm_rss(struct mm_struct
*mm
);
1699 static inline void sync_mm_rss(struct mm_struct
*mm
)
1704 #ifndef __HAVE_ARCH_PTE_DEVMAP
1705 static inline int pte_devmap(pte_t pte
)
1711 int vma_wants_writenotify(struct vm_area_struct
*vma
, pgprot_t vm_page_prot
);
1713 extern pte_t
*__get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1715 static inline pte_t
*get_locked_pte(struct mm_struct
*mm
, unsigned long addr
,
1719 __cond_lock(*ptl
, ptep
= __get_locked_pte(mm
, addr
, ptl
));
1723 #ifdef __PAGETABLE_P4D_FOLDED
1724 static inline int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1725 unsigned long address
)
1730 int __p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
, unsigned long address
);
1733 #if defined(__PAGETABLE_PUD_FOLDED) || !defined(CONFIG_MMU)
1734 static inline int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1735 unsigned long address
)
1739 static inline void mm_inc_nr_puds(struct mm_struct
*mm
) {}
1740 static inline void mm_dec_nr_puds(struct mm_struct
*mm
) {}
1743 int __pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
, unsigned long address
);
1745 static inline void mm_inc_nr_puds(struct mm_struct
*mm
)
1747 if (mm_pud_folded(mm
))
1749 atomic_long_add(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
1752 static inline void mm_dec_nr_puds(struct mm_struct
*mm
)
1754 if (mm_pud_folded(mm
))
1756 atomic_long_sub(PTRS_PER_PUD
* sizeof(pud_t
), &mm
->pgtables_bytes
);
1760 #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU)
1761 static inline int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
,
1762 unsigned long address
)
1767 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
) {}
1768 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
) {}
1771 int __pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
);
1773 static inline void mm_inc_nr_pmds(struct mm_struct
*mm
)
1775 if (mm_pmd_folded(mm
))
1777 atomic_long_add(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
1780 static inline void mm_dec_nr_pmds(struct mm_struct
*mm
)
1782 if (mm_pmd_folded(mm
))
1784 atomic_long_sub(PTRS_PER_PMD
* sizeof(pmd_t
), &mm
->pgtables_bytes
);
1789 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
)
1791 atomic_long_set(&mm
->pgtables_bytes
, 0);
1794 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
1796 return atomic_long_read(&mm
->pgtables_bytes
);
1799 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
)
1801 atomic_long_add(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
1804 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
)
1806 atomic_long_sub(PTRS_PER_PTE
* sizeof(pte_t
), &mm
->pgtables_bytes
);
1810 static inline void mm_pgtables_bytes_init(struct mm_struct
*mm
) {}
1811 static inline unsigned long mm_pgtables_bytes(const struct mm_struct
*mm
)
1816 static inline void mm_inc_nr_ptes(struct mm_struct
*mm
) {}
1817 static inline void mm_dec_nr_ptes(struct mm_struct
*mm
) {}
1820 int __pte_alloc(struct mm_struct
*mm
, pmd_t
*pmd
, unsigned long address
);
1821 int __pte_alloc_kernel(pmd_t
*pmd
, unsigned long address
);
1824 * The following ifdef needed to get the 4level-fixup.h header to work.
1825 * Remove it when 4level-fixup.h has been removed.
1827 #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK)
1829 #ifndef __ARCH_HAS_5LEVEL_HACK
1830 static inline p4d_t
*p4d_alloc(struct mm_struct
*mm
, pgd_t
*pgd
,
1831 unsigned long address
)
1833 return (unlikely(pgd_none(*pgd
)) && __p4d_alloc(mm
, pgd
, address
)) ?
1834 NULL
: p4d_offset(pgd
, address
);
1837 static inline pud_t
*pud_alloc(struct mm_struct
*mm
, p4d_t
*p4d
,
1838 unsigned long address
)
1840 return (unlikely(p4d_none(*p4d
)) && __pud_alloc(mm
, p4d
, address
)) ?
1841 NULL
: pud_offset(p4d
, address
);
1843 #endif /* !__ARCH_HAS_5LEVEL_HACK */
1845 static inline pmd_t
*pmd_alloc(struct mm_struct
*mm
, pud_t
*pud
, unsigned long address
)
1847 return (unlikely(pud_none(*pud
)) && __pmd_alloc(mm
, pud
, address
))?
1848 NULL
: pmd_offset(pud
, address
);
1850 #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */
1852 #if USE_SPLIT_PTE_PTLOCKS
1853 #if ALLOC_SPLIT_PTLOCKS
1854 void __init
ptlock_cache_init(void);
1855 extern bool ptlock_alloc(struct page
*page
);
1856 extern void ptlock_free(struct page
*page
);
1858 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1862 #else /* ALLOC_SPLIT_PTLOCKS */
1863 static inline void ptlock_cache_init(void)
1867 static inline bool ptlock_alloc(struct page
*page
)
1872 static inline void ptlock_free(struct page
*page
)
1876 static inline spinlock_t
*ptlock_ptr(struct page
*page
)
1880 #endif /* ALLOC_SPLIT_PTLOCKS */
1882 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1884 return ptlock_ptr(pmd_page(*pmd
));
1887 static inline bool ptlock_init(struct page
*page
)
1890 * prep_new_page() initialize page->private (and therefore page->ptl)
1891 * with 0. Make sure nobody took it in use in between.
1893 * It can happen if arch try to use slab for page table allocation:
1894 * slab code uses page->slab_cache, which share storage with page->ptl.
1896 VM_BUG_ON_PAGE(*(unsigned long *)&page
->ptl
, page
);
1897 if (!ptlock_alloc(page
))
1899 spin_lock_init(ptlock_ptr(page
));
1903 /* Reset page->mapping so free_pages_check won't complain. */
1904 static inline void pte_lock_deinit(struct page
*page
)
1906 page
->mapping
= NULL
;
1910 #else /* !USE_SPLIT_PTE_PTLOCKS */
1912 * We use mm->page_table_lock to guard all pagetable pages of the mm.
1914 static inline spinlock_t
*pte_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1916 return &mm
->page_table_lock
;
1918 static inline void ptlock_cache_init(void) {}
1919 static inline bool ptlock_init(struct page
*page
) { return true; }
1920 static inline void pte_lock_deinit(struct page
*page
) {}
1921 #endif /* USE_SPLIT_PTE_PTLOCKS */
1923 static inline void pgtable_init(void)
1925 ptlock_cache_init();
1926 pgtable_cache_init();
1929 static inline bool pgtable_page_ctor(struct page
*page
)
1931 if (!ptlock_init(page
))
1933 __SetPageTable(page
);
1934 inc_zone_page_state(page
, NR_PAGETABLE
);
1938 static inline void pgtable_page_dtor(struct page
*page
)
1940 pte_lock_deinit(page
);
1941 __ClearPageTable(page
);
1942 dec_zone_page_state(page
, NR_PAGETABLE
);
1945 #define pte_offset_map_lock(mm, pmd, address, ptlp) \
1947 spinlock_t *__ptl = pte_lockptr(mm, pmd); \
1948 pte_t *__pte = pte_offset_map(pmd, address); \
1954 #define pte_unmap_unlock(pte, ptl) do { \
1959 #define pte_alloc(mm, pmd, address) \
1960 (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address))
1962 #define pte_alloc_map(mm, pmd, address) \
1963 (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address))
1965 #define pte_alloc_map_lock(mm, pmd, address, ptlp) \
1966 (pte_alloc(mm, pmd, address) ? \
1967 NULL : pte_offset_map_lock(mm, pmd, address, ptlp))
1969 #define pte_alloc_kernel(pmd, address) \
1970 ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \
1971 NULL: pte_offset_kernel(pmd, address))
1973 #if USE_SPLIT_PMD_PTLOCKS
1975 static struct page
*pmd_to_page(pmd_t
*pmd
)
1977 unsigned long mask
= ~(PTRS_PER_PMD
* sizeof(pmd_t
) - 1);
1978 return virt_to_page((void *)((unsigned long) pmd
& mask
));
1981 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
1983 return ptlock_ptr(pmd_to_page(pmd
));
1986 static inline bool pgtable_pmd_page_ctor(struct page
*page
)
1988 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1989 page
->pmd_huge_pte
= NULL
;
1991 return ptlock_init(page
);
1994 static inline void pgtable_pmd_page_dtor(struct page
*page
)
1996 #ifdef CONFIG_TRANSPARENT_HUGEPAGE
1997 VM_BUG_ON_PAGE(page
->pmd_huge_pte
, page
);
2002 #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte)
2006 static inline spinlock_t
*pmd_lockptr(struct mm_struct
*mm
, pmd_t
*pmd
)
2008 return &mm
->page_table_lock
;
2011 static inline bool pgtable_pmd_page_ctor(struct page
*page
) { return true; }
2012 static inline void pgtable_pmd_page_dtor(struct page
*page
) {}
2014 #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte)
2018 static inline spinlock_t
*pmd_lock(struct mm_struct
*mm
, pmd_t
*pmd
)
2020 spinlock_t
*ptl
= pmd_lockptr(mm
, pmd
);
2026 * No scalability reason to split PUD locks yet, but follow the same pattern
2027 * as the PMD locks to make it easier if we decide to. The VM should not be
2028 * considered ready to switch to split PUD locks yet; there may be places
2029 * which need to be converted from page_table_lock.
2031 static inline spinlock_t
*pud_lockptr(struct mm_struct
*mm
, pud_t
*pud
)
2033 return &mm
->page_table_lock
;
2036 static inline spinlock_t
*pud_lock(struct mm_struct
*mm
, pud_t
*pud
)
2038 spinlock_t
*ptl
= pud_lockptr(mm
, pud
);
2044 extern void __init
pagecache_init(void);
2045 extern void free_area_init(unsigned long * zones_size
);
2046 extern void __init
free_area_init_node(int nid
, unsigned long * zones_size
,
2047 unsigned long zone_start_pfn
, unsigned long *zholes_size
);
2048 extern void free_initmem(void);
2051 * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK)
2052 * into the buddy system. The freed pages will be poisoned with pattern
2053 * "poison" if it's within range [0, UCHAR_MAX].
2054 * Return pages freed into the buddy system.
2056 extern unsigned long free_reserved_area(void *start
, void *end
,
2057 int poison
, char *s
);
2059 #ifdef CONFIG_HIGHMEM
2061 * Free a highmem page into the buddy system, adjusting totalhigh_pages
2062 * and totalram_pages.
2064 extern void free_highmem_page(struct page
*page
);
2067 extern void adjust_managed_page_count(struct page
*page
, long count
);
2068 extern void mem_init_print_info(const char *str
);
2070 extern void reserve_bootmem_region(phys_addr_t start
, phys_addr_t end
);
2072 /* Free the reserved page into the buddy system, so it gets managed. */
2073 static inline void __free_reserved_page(struct page
*page
)
2075 ClearPageReserved(page
);
2076 init_page_count(page
);
2080 static inline void free_reserved_page(struct page
*page
)
2082 __free_reserved_page(page
);
2083 adjust_managed_page_count(page
, 1);
2086 static inline void mark_page_reserved(struct page
*page
)
2088 SetPageReserved(page
);
2089 adjust_managed_page_count(page
, -1);
2093 * Default method to free all the __init memory into the buddy system.
2094 * The freed pages will be poisoned with pattern "poison" if it's within
2095 * range [0, UCHAR_MAX].
2096 * Return pages freed into the buddy system.
2098 static inline unsigned long free_initmem_default(int poison
)
2100 extern char __init_begin
[], __init_end
[];
2102 return free_reserved_area(&__init_begin
, &__init_end
,
2103 poison
, "unused kernel");
2106 static inline unsigned long get_num_physpages(void)
2109 unsigned long phys_pages
= 0;
2111 for_each_online_node(nid
)
2112 phys_pages
+= node_present_pages(nid
);
2117 #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP
2119 * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its
2120 * zones, allocate the backing mem_map and account for memory holes in a more
2121 * architecture independent manner. This is a substitute for creating the
2122 * zone_sizes[] and zholes_size[] arrays and passing them to
2123 * free_area_init_node()
2125 * An architecture is expected to register range of page frames backed by
2126 * physical memory with memblock_add[_node]() before calling
2127 * free_area_init_nodes() passing in the PFN each zone ends at. At a basic
2128 * usage, an architecture is expected to do something like
2130 * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn,
2132 * for_each_valid_physical_page_range()
2133 * memblock_add_node(base, size, nid)
2134 * free_area_init_nodes(max_zone_pfns);
2136 * free_bootmem_with_active_regions() calls free_bootmem_node() for each
2137 * registered physical page range. Similarly
2138 * sparse_memory_present_with_active_regions() calls memory_present() for
2139 * each range when SPARSEMEM is enabled.
2141 * See mm/page_alloc.c for more information on each function exposed by
2142 * CONFIG_HAVE_MEMBLOCK_NODE_MAP.
2144 extern void free_area_init_nodes(unsigned long *max_zone_pfn
);
2145 unsigned long node_map_pfn_alignment(void);
2146 unsigned long __absent_pages_in_range(int nid
, unsigned long start_pfn
,
2147 unsigned long end_pfn
);
2148 extern unsigned long absent_pages_in_range(unsigned long start_pfn
,
2149 unsigned long end_pfn
);
2150 extern void get_pfn_range_for_nid(unsigned int nid
,
2151 unsigned long *start_pfn
, unsigned long *end_pfn
);
2152 extern unsigned long find_min_pfn_with_active_regions(void);
2153 extern void free_bootmem_with_active_regions(int nid
,
2154 unsigned long max_low_pfn
);
2155 extern void sparse_memory_present_with_active_regions(int nid
);
2157 #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */
2159 #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \
2160 !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID)
2161 static inline int __early_pfn_to_nid(unsigned long pfn
,
2162 struct mminit_pfnnid_cache
*state
)
2167 /* please see mm/page_alloc.c */
2168 extern int __meminit
early_pfn_to_nid(unsigned long pfn
);
2169 /* there is a per-arch backend function. */
2170 extern int __meminit
__early_pfn_to_nid(unsigned long pfn
,
2171 struct mminit_pfnnid_cache
*state
);
2174 #if !defined(CONFIG_FLAT_NODE_MEM_MAP)
2175 void zero_resv_unavail(void);
2177 static inline void zero_resv_unavail(void) {}
2180 extern void set_dma_reserve(unsigned long new_dma_reserve
);
2181 extern void memmap_init_zone(unsigned long, int, unsigned long, unsigned long,
2182 enum memmap_context
, struct vmem_altmap
*);
2183 extern void setup_per_zone_wmarks(void);
2184 extern int __meminit
init_per_zone_wmark_min(void);
2185 extern void mem_init(void);
2186 extern void __init
mmap_init(void);
2187 extern void show_mem(unsigned int flags
, nodemask_t
*nodemask
);
2188 extern long si_mem_available(void);
2189 extern void si_meminfo(struct sysinfo
* val
);
2190 extern void si_meminfo_node(struct sysinfo
*val
, int nid
);
2191 #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES
2192 extern unsigned long arch_reserved_kernel_pages(void);
2195 extern __printf(3, 4)
2196 void warn_alloc(gfp_t gfp_mask
, nodemask_t
*nodemask
, const char *fmt
, ...);
2198 extern void setup_per_cpu_pageset(void);
2200 extern void zone_pcp_update(struct zone
*zone
);
2201 extern void zone_pcp_reset(struct zone
*zone
);
2204 extern int min_free_kbytes
;
2205 extern int watermark_scale_factor
;
2208 extern atomic_long_t mmap_pages_allocated
;
2209 extern int nommu_shrink_inode_mappings(struct inode
*, size_t, size_t);
2211 /* interval_tree.c */
2212 void vma_interval_tree_insert(struct vm_area_struct
*node
,
2213 struct rb_root_cached
*root
);
2214 void vma_interval_tree_insert_after(struct vm_area_struct
*node
,
2215 struct vm_area_struct
*prev
,
2216 struct rb_root_cached
*root
);
2217 void vma_interval_tree_remove(struct vm_area_struct
*node
,
2218 struct rb_root_cached
*root
);
2219 struct vm_area_struct
*vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2220 unsigned long start
, unsigned long last
);
2221 struct vm_area_struct
*vma_interval_tree_iter_next(struct vm_area_struct
*node
,
2222 unsigned long start
, unsigned long last
);
2224 #define vma_interval_tree_foreach(vma, root, start, last) \
2225 for (vma = vma_interval_tree_iter_first(root, start, last); \
2226 vma; vma = vma_interval_tree_iter_next(vma, start, last))
2228 void anon_vma_interval_tree_insert(struct anon_vma_chain
*node
,
2229 struct rb_root_cached
*root
);
2230 void anon_vma_interval_tree_remove(struct anon_vma_chain
*node
,
2231 struct rb_root_cached
*root
);
2232 struct anon_vma_chain
*
2233 anon_vma_interval_tree_iter_first(struct rb_root_cached
*root
,
2234 unsigned long start
, unsigned long last
);
2235 struct anon_vma_chain
*anon_vma_interval_tree_iter_next(
2236 struct anon_vma_chain
*node
, unsigned long start
, unsigned long last
);
2237 #ifdef CONFIG_DEBUG_VM_RB
2238 void anon_vma_interval_tree_verify(struct anon_vma_chain
*node
);
2241 #define anon_vma_interval_tree_foreach(avc, root, start, last) \
2242 for (avc = anon_vma_interval_tree_iter_first(root, start, last); \
2243 avc; avc = anon_vma_interval_tree_iter_next(avc, start, last))
2246 extern int __vm_enough_memory(struct mm_struct
*mm
, long pages
, int cap_sys_admin
);
2247 extern int __vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2248 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
,
2249 struct vm_area_struct
*expand
);
2250 static inline int vma_adjust(struct vm_area_struct
*vma
, unsigned long start
,
2251 unsigned long end
, pgoff_t pgoff
, struct vm_area_struct
*insert
)
2253 return __vma_adjust(vma
, start
, end
, pgoff
, insert
, NULL
);
2255 extern struct vm_area_struct
*vma_merge(struct mm_struct
*,
2256 struct vm_area_struct
*prev
, unsigned long addr
, unsigned long end
,
2257 unsigned long vm_flags
, struct anon_vma
*, struct file
*, pgoff_t
,
2258 struct mempolicy
*, struct vm_userfaultfd_ctx
);
2259 extern struct anon_vma
*find_mergeable_anon_vma(struct vm_area_struct
*);
2260 extern int __split_vma(struct mm_struct
*, struct vm_area_struct
*,
2261 unsigned long addr
, int new_below
);
2262 extern int split_vma(struct mm_struct
*, struct vm_area_struct
*,
2263 unsigned long addr
, int new_below
);
2264 extern int insert_vm_struct(struct mm_struct
*, struct vm_area_struct
*);
2265 extern void __vma_link_rb(struct mm_struct
*, struct vm_area_struct
*,
2266 struct rb_node
**, struct rb_node
*);
2267 extern void unlink_file_vma(struct vm_area_struct
*);
2268 extern struct vm_area_struct
*copy_vma(struct vm_area_struct
**,
2269 unsigned long addr
, unsigned long len
, pgoff_t pgoff
,
2270 bool *need_rmap_locks
);
2271 extern void exit_mmap(struct mm_struct
*);
2273 static inline int check_data_rlimit(unsigned long rlim
,
2275 unsigned long start
,
2276 unsigned long end_data
,
2277 unsigned long start_data
)
2279 if (rlim
< RLIM_INFINITY
) {
2280 if (((new - start
) + (end_data
- start_data
)) > rlim
)
2287 extern int mm_take_all_locks(struct mm_struct
*mm
);
2288 extern void mm_drop_all_locks(struct mm_struct
*mm
);
2290 extern void set_mm_exe_file(struct mm_struct
*mm
, struct file
*new_exe_file
);
2291 extern struct file
*get_mm_exe_file(struct mm_struct
*mm
);
2292 extern struct file
*get_task_exe_file(struct task_struct
*task
);
2294 extern bool may_expand_vm(struct mm_struct
*, vm_flags_t
, unsigned long npages
);
2295 extern void vm_stat_account(struct mm_struct
*, vm_flags_t
, long npages
);
2297 extern bool vma_is_special_mapping(const struct vm_area_struct
*vma
,
2298 const struct vm_special_mapping
*sm
);
2299 extern struct vm_area_struct
*_install_special_mapping(struct mm_struct
*mm
,
2300 unsigned long addr
, unsigned long len
,
2301 unsigned long flags
,
2302 const struct vm_special_mapping
*spec
);
2303 /* This is an obsolete alternative to _install_special_mapping. */
2304 extern int install_special_mapping(struct mm_struct
*mm
,
2305 unsigned long addr
, unsigned long len
,
2306 unsigned long flags
, struct page
**pages
);
2308 extern unsigned long get_unmapped_area(struct file
*, unsigned long, unsigned long, unsigned long, unsigned long);
2310 extern unsigned long mmap_region(struct file
*file
, unsigned long addr
,
2311 unsigned long len
, vm_flags_t vm_flags
, unsigned long pgoff
,
2312 struct list_head
*uf
);
2313 extern unsigned long do_mmap(struct file
*file
, unsigned long addr
,
2314 unsigned long len
, unsigned long prot
, unsigned long flags
,
2315 vm_flags_t vm_flags
, unsigned long pgoff
, unsigned long *populate
,
2316 struct list_head
*uf
);
2317 extern int __do_munmap(struct mm_struct
*, unsigned long, size_t,
2318 struct list_head
*uf
, bool downgrade
);
2319 extern int do_munmap(struct mm_struct
*, unsigned long, size_t,
2320 struct list_head
*uf
);
2322 static inline unsigned long
2323 do_mmap_pgoff(struct file
*file
, unsigned long addr
,
2324 unsigned long len
, unsigned long prot
, unsigned long flags
,
2325 unsigned long pgoff
, unsigned long *populate
,
2326 struct list_head
*uf
)
2328 return do_mmap(file
, addr
, len
, prot
, flags
, 0, pgoff
, populate
, uf
);
2332 extern int __mm_populate(unsigned long addr
, unsigned long len
,
2334 static inline void mm_populate(unsigned long addr
, unsigned long len
)
2337 (void) __mm_populate(addr
, len
, 1);
2340 static inline void mm_populate(unsigned long addr
, unsigned long len
) {}
2343 /* These take the mm semaphore themselves */
2344 extern int __must_check
vm_brk(unsigned long, unsigned long);
2345 extern int __must_check
vm_brk_flags(unsigned long, unsigned long, unsigned long);
2346 extern int vm_munmap(unsigned long, size_t);
2347 extern unsigned long __must_check
vm_mmap(struct file
*, unsigned long,
2348 unsigned long, unsigned long,
2349 unsigned long, unsigned long);
2351 struct vm_unmapped_area_info
{
2352 #define VM_UNMAPPED_AREA_TOPDOWN 1
2353 unsigned long flags
;
2354 unsigned long length
;
2355 unsigned long low_limit
;
2356 unsigned long high_limit
;
2357 unsigned long align_mask
;
2358 unsigned long align_offset
;
2361 extern unsigned long unmapped_area(struct vm_unmapped_area_info
*info
);
2362 extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info
*info
);
2365 * Search for an unmapped address range.
2367 * We are looking for a range that:
2368 * - does not intersect with any VMA;
2369 * - is contained within the [low_limit, high_limit) interval;
2370 * - is at least the desired size.
2371 * - satisfies (begin_addr & align_mask) == (align_offset & align_mask)
2373 static inline unsigned long
2374 vm_unmapped_area(struct vm_unmapped_area_info
*info
)
2376 if (info
->flags
& VM_UNMAPPED_AREA_TOPDOWN
)
2377 return unmapped_area_topdown(info
);
2379 return unmapped_area(info
);
2383 extern void truncate_inode_pages(struct address_space
*, loff_t
);
2384 extern void truncate_inode_pages_range(struct address_space
*,
2385 loff_t lstart
, loff_t lend
);
2386 extern void truncate_inode_pages_final(struct address_space
*);
2388 /* generic vm_area_ops exported for stackable file systems */
2389 extern vm_fault_t
filemap_fault(struct vm_fault
*vmf
);
2390 extern void filemap_map_pages(struct vm_fault
*vmf
,
2391 pgoff_t start_pgoff
, pgoff_t end_pgoff
);
2392 extern vm_fault_t
filemap_page_mkwrite(struct vm_fault
*vmf
);
2394 /* mm/page-writeback.c */
2395 int __must_check
write_one_page(struct page
*page
);
2396 void task_dirty_inc(struct task_struct
*tsk
);
2399 #define VM_MAX_READAHEAD 128 /* kbytes */
2400 #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */
2402 int force_page_cache_readahead(struct address_space
*mapping
, struct file
*filp
,
2403 pgoff_t offset
, unsigned long nr_to_read
);
2405 void page_cache_sync_readahead(struct address_space
*mapping
,
2406 struct file_ra_state
*ra
,
2409 unsigned long size
);
2411 void page_cache_async_readahead(struct address_space
*mapping
,
2412 struct file_ra_state
*ra
,
2416 unsigned long size
);
2418 extern unsigned long stack_guard_gap
;
2419 /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */
2420 extern int expand_stack(struct vm_area_struct
*vma
, unsigned long address
);
2422 /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */
2423 extern int expand_downwards(struct vm_area_struct
*vma
,
2424 unsigned long address
);
2426 extern int expand_upwards(struct vm_area_struct
*vma
, unsigned long address
);
2428 #define expand_upwards(vma, address) (0)
2431 /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */
2432 extern struct vm_area_struct
* find_vma(struct mm_struct
* mm
, unsigned long addr
);
2433 extern struct vm_area_struct
* find_vma_prev(struct mm_struct
* mm
, unsigned long addr
,
2434 struct vm_area_struct
**pprev
);
2436 /* Look up the first VMA which intersects the interval start_addr..end_addr-1,
2437 NULL if none. Assume start_addr < end_addr. */
2438 static inline struct vm_area_struct
* find_vma_intersection(struct mm_struct
* mm
, unsigned long start_addr
, unsigned long end_addr
)
2440 struct vm_area_struct
* vma
= find_vma(mm
,start_addr
);
2442 if (vma
&& end_addr
<= vma
->vm_start
)
2447 static inline unsigned long vm_start_gap(struct vm_area_struct
*vma
)
2449 unsigned long vm_start
= vma
->vm_start
;
2451 if (vma
->vm_flags
& VM_GROWSDOWN
) {
2452 vm_start
-= stack_guard_gap
;
2453 if (vm_start
> vma
->vm_start
)
2459 static inline unsigned long vm_end_gap(struct vm_area_struct
*vma
)
2461 unsigned long vm_end
= vma
->vm_end
;
2463 if (vma
->vm_flags
& VM_GROWSUP
) {
2464 vm_end
+= stack_guard_gap
;
2465 if (vm_end
< vma
->vm_end
)
2466 vm_end
= -PAGE_SIZE
;
2471 static inline unsigned long vma_pages(struct vm_area_struct
*vma
)
2473 return (vma
->vm_end
- vma
->vm_start
) >> PAGE_SHIFT
;
2476 /* Look up the first VMA which exactly match the interval vm_start ... vm_end */
2477 static inline struct vm_area_struct
*find_exact_vma(struct mm_struct
*mm
,
2478 unsigned long vm_start
, unsigned long vm_end
)
2480 struct vm_area_struct
*vma
= find_vma(mm
, vm_start
);
2482 if (vma
&& (vma
->vm_start
!= vm_start
|| vma
->vm_end
!= vm_end
))
2488 static inline bool range_in_vma(struct vm_area_struct
*vma
,
2489 unsigned long start
, unsigned long end
)
2491 return (vma
&& vma
->vm_start
<= start
&& end
<= vma
->vm_end
);
2495 pgprot_t
vm_get_page_prot(unsigned long vm_flags
);
2496 void vma_set_page_prot(struct vm_area_struct
*vma
);
2498 static inline pgprot_t
vm_get_page_prot(unsigned long vm_flags
)
2502 static inline void vma_set_page_prot(struct vm_area_struct
*vma
)
2504 vma
->vm_page_prot
= vm_get_page_prot(vma
->vm_flags
);
2508 #ifdef CONFIG_NUMA_BALANCING
2509 unsigned long change_prot_numa(struct vm_area_struct
*vma
,
2510 unsigned long start
, unsigned long end
);
2513 struct vm_area_struct
*find_extend_vma(struct mm_struct
*, unsigned long addr
);
2514 int remap_pfn_range(struct vm_area_struct
*, unsigned long addr
,
2515 unsigned long pfn
, unsigned long size
, pgprot_t
);
2516 int vm_insert_page(struct vm_area_struct
*, unsigned long addr
, struct page
*);
2517 vm_fault_t
vmf_insert_pfn(struct vm_area_struct
*vma
, unsigned long addr
,
2519 vm_fault_t
vmf_insert_pfn_prot(struct vm_area_struct
*vma
, unsigned long addr
,
2520 unsigned long pfn
, pgprot_t pgprot
);
2521 vm_fault_t
vmf_insert_mixed(struct vm_area_struct
*vma
, unsigned long addr
,
2523 vm_fault_t
vmf_insert_mixed_mkwrite(struct vm_area_struct
*vma
,
2524 unsigned long addr
, pfn_t pfn
);
2525 int vm_iomap_memory(struct vm_area_struct
*vma
, phys_addr_t start
, unsigned long len
);
2527 static inline vm_fault_t
vmf_insert_page(struct vm_area_struct
*vma
,
2528 unsigned long addr
, struct page
*page
)
2530 int err
= vm_insert_page(vma
, addr
, page
);
2533 return VM_FAULT_OOM
;
2534 if (err
< 0 && err
!= -EBUSY
)
2535 return VM_FAULT_SIGBUS
;
2537 return VM_FAULT_NOPAGE
;
2540 static inline vm_fault_t
vmf_error(int err
)
2543 return VM_FAULT_OOM
;
2544 return VM_FAULT_SIGBUS
;
2547 struct page
*follow_page(struct vm_area_struct
*vma
, unsigned long address
,
2548 unsigned int foll_flags
);
2550 #define FOLL_WRITE 0x01 /* check pte is writable */
2551 #define FOLL_TOUCH 0x02 /* mark page accessed */
2552 #define FOLL_GET 0x04 /* do get_page on page */
2553 #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */
2554 #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */
2555 #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO
2556 * and return without waiting upon it */
2557 #define FOLL_POPULATE 0x40 /* fault in page */
2558 #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */
2559 #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */
2560 #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */
2561 #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */
2562 #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */
2563 #define FOLL_MLOCK 0x1000 /* lock present pages */
2564 #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */
2565 #define FOLL_COW 0x4000 /* internal GUP flag */
2566 #define FOLL_ANON 0x8000 /* don't do file mappings */
2568 static inline int vm_fault_to_errno(vm_fault_t vm_fault
, int foll_flags
)
2570 if (vm_fault
& VM_FAULT_OOM
)
2572 if (vm_fault
& (VM_FAULT_HWPOISON
| VM_FAULT_HWPOISON_LARGE
))
2573 return (foll_flags
& FOLL_HWPOISON
) ? -EHWPOISON
: -EFAULT
;
2574 if (vm_fault
& (VM_FAULT_SIGBUS
| VM_FAULT_SIGSEGV
))
2579 typedef int (*pte_fn_t
)(pte_t
*pte
, pgtable_t token
, unsigned long addr
,
2581 extern int apply_to_page_range(struct mm_struct
*mm
, unsigned long address
,
2582 unsigned long size
, pte_fn_t fn
, void *data
);
2585 #ifdef CONFIG_PAGE_POISONING
2586 extern bool page_poisoning_enabled(void);
2587 extern void kernel_poison_pages(struct page
*page
, int numpages
, int enable
);
2589 static inline bool page_poisoning_enabled(void) { return false; }
2590 static inline void kernel_poison_pages(struct page
*page
, int numpages
,
2594 #ifdef CONFIG_DEBUG_PAGEALLOC
2595 extern bool _debug_pagealloc_enabled
;
2596 extern void __kernel_map_pages(struct page
*page
, int numpages
, int enable
);
2598 static inline bool debug_pagealloc_enabled(void)
2600 return _debug_pagealloc_enabled
;
2604 kernel_map_pages(struct page
*page
, int numpages
, int enable
)
2606 if (!debug_pagealloc_enabled())
2609 __kernel_map_pages(page
, numpages
, enable
);
2611 #ifdef CONFIG_HIBERNATION
2612 extern bool kernel_page_present(struct page
*page
);
2613 #endif /* CONFIG_HIBERNATION */
2614 #else /* CONFIG_DEBUG_PAGEALLOC */
2616 kernel_map_pages(struct page
*page
, int numpages
, int enable
) {}
2617 #ifdef CONFIG_HIBERNATION
2618 static inline bool kernel_page_present(struct page
*page
) { return true; }
2619 #endif /* CONFIG_HIBERNATION */
2620 static inline bool debug_pagealloc_enabled(void)
2624 #endif /* CONFIG_DEBUG_PAGEALLOC */
2626 #ifdef __HAVE_ARCH_GATE_AREA
2627 extern struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
);
2628 extern int in_gate_area_no_mm(unsigned long addr
);
2629 extern int in_gate_area(struct mm_struct
*mm
, unsigned long addr
);
2631 static inline struct vm_area_struct
*get_gate_vma(struct mm_struct
*mm
)
2635 static inline int in_gate_area_no_mm(unsigned long addr
) { return 0; }
2636 static inline int in_gate_area(struct mm_struct
*mm
, unsigned long addr
)
2640 #endif /* __HAVE_ARCH_GATE_AREA */
2642 extern bool process_shares_mm(struct task_struct
*p
, struct mm_struct
*mm
);
2644 #ifdef CONFIG_SYSCTL
2645 extern int sysctl_drop_caches
;
2646 int drop_caches_sysctl_handler(struct ctl_table
*, int,
2647 void __user
*, size_t *, loff_t
*);
2650 void drop_slab(void);
2651 void drop_slab_node(int nid
);
2654 #define randomize_va_space 0
2656 extern int randomize_va_space
;
2659 const char * arch_vma_name(struct vm_area_struct
*vma
);
2660 void print_vma_addr(char *prefix
, unsigned long rip
);
2662 void *sparse_buffer_alloc(unsigned long size
);
2663 struct page
*sparse_mem_map_populate(unsigned long pnum
, int nid
,
2664 struct vmem_altmap
*altmap
);
2665 pgd_t
*vmemmap_pgd_populate(unsigned long addr
, int node
);
2666 p4d_t
*vmemmap_p4d_populate(pgd_t
*pgd
, unsigned long addr
, int node
);
2667 pud_t
*vmemmap_pud_populate(p4d_t
*p4d
, unsigned long addr
, int node
);
2668 pmd_t
*vmemmap_pmd_populate(pud_t
*pud
, unsigned long addr
, int node
);
2669 pte_t
*vmemmap_pte_populate(pmd_t
*pmd
, unsigned long addr
, int node
);
2670 void *vmemmap_alloc_block(unsigned long size
, int node
);
2672 void *vmemmap_alloc_block_buf(unsigned long size
, int node
);
2673 void *altmap_alloc_block_buf(unsigned long size
, struct vmem_altmap
*altmap
);
2674 void vmemmap_verify(pte_t
*, int, unsigned long, unsigned long);
2675 int vmemmap_populate_basepages(unsigned long start
, unsigned long end
,
2677 int vmemmap_populate(unsigned long start
, unsigned long end
, int node
,
2678 struct vmem_altmap
*altmap
);
2679 void vmemmap_populate_print_last(void);
2680 #ifdef CONFIG_MEMORY_HOTPLUG
2681 void vmemmap_free(unsigned long start
, unsigned long end
,
2682 struct vmem_altmap
*altmap
);
2684 void register_page_bootmem_memmap(unsigned long section_nr
, struct page
*map
,
2685 unsigned long nr_pages
);
2688 MF_COUNT_INCREASED
= 1 << 0,
2689 MF_ACTION_REQUIRED
= 1 << 1,
2690 MF_MUST_KILL
= 1 << 2,
2691 MF_SOFT_OFFLINE
= 1 << 3,
2693 extern int memory_failure(unsigned long pfn
, int flags
);
2694 extern void memory_failure_queue(unsigned long pfn
, int flags
);
2695 extern int unpoison_memory(unsigned long pfn
);
2696 extern int get_hwpoison_page(struct page
*page
);
2697 #define put_hwpoison_page(page) put_page(page)
2698 extern int sysctl_memory_failure_early_kill
;
2699 extern int sysctl_memory_failure_recovery
;
2700 extern void shake_page(struct page
*p
, int access
);
2701 extern atomic_long_t num_poisoned_pages __read_mostly
;
2702 extern int soft_offline_page(struct page
*page
, int flags
);
2706 * Error handlers for various types of pages.
2709 MF_IGNORED
, /* Error: cannot be handled */
2710 MF_FAILED
, /* Error: handling failed */
2711 MF_DELAYED
, /* Will be handled later */
2712 MF_RECOVERED
, /* Successfully recovered */
2715 enum mf_action_page_type
{
2717 MF_MSG_KERNEL_HIGH_ORDER
,
2719 MF_MSG_DIFFERENT_COMPOUND
,
2720 MF_MSG_POISONED_HUGE
,
2723 MF_MSG_NON_PMD_HUGE
,
2724 MF_MSG_UNMAP_FAILED
,
2725 MF_MSG_DIRTY_SWAPCACHE
,
2726 MF_MSG_CLEAN_SWAPCACHE
,
2727 MF_MSG_DIRTY_MLOCKED_LRU
,
2728 MF_MSG_CLEAN_MLOCKED_LRU
,
2729 MF_MSG_DIRTY_UNEVICTABLE_LRU
,
2730 MF_MSG_CLEAN_UNEVICTABLE_LRU
,
2733 MF_MSG_TRUNCATED_LRU
,
2740 #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS)
2741 extern void clear_huge_page(struct page
*page
,
2742 unsigned long addr_hint
,
2743 unsigned int pages_per_huge_page
);
2744 extern void copy_user_huge_page(struct page
*dst
, struct page
*src
,
2745 unsigned long addr_hint
,
2746 struct vm_area_struct
*vma
,
2747 unsigned int pages_per_huge_page
);
2748 extern long copy_huge_page_from_user(struct page
*dst_page
,
2749 const void __user
*usr_src
,
2750 unsigned int pages_per_huge_page
,
2751 bool allow_pagefault
);
2752 #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */
2754 extern struct page_ext_operations debug_guardpage_ops
;
2756 #ifdef CONFIG_DEBUG_PAGEALLOC
2757 extern unsigned int _debug_guardpage_minorder
;
2758 extern bool _debug_guardpage_enabled
;
2760 static inline unsigned int debug_guardpage_minorder(void)
2762 return _debug_guardpage_minorder
;
2765 static inline bool debug_guardpage_enabled(void)
2767 return _debug_guardpage_enabled
;
2770 static inline bool page_is_guard(struct page
*page
)
2772 struct page_ext
*page_ext
;
2774 if (!debug_guardpage_enabled())
2777 page_ext
= lookup_page_ext(page
);
2778 if (unlikely(!page_ext
))
2781 return test_bit(PAGE_EXT_DEBUG_GUARD
, &page_ext
->flags
);
2784 static inline unsigned int debug_guardpage_minorder(void) { return 0; }
2785 static inline bool debug_guardpage_enabled(void) { return false; }
2786 static inline bool page_is_guard(struct page
*page
) { return false; }
2787 #endif /* CONFIG_DEBUG_PAGEALLOC */
2789 #if MAX_NUMNODES > 1
2790 void __init
setup_nr_node_ids(void);
2792 static inline void setup_nr_node_ids(void) {}
2795 #endif /* __KERNEL__ */
2796 #endif /* _LINUX_MM_H */