]>
Commit | Line | Data |
---|---|---|
1 | #ifndef _LINUX_MM_H | |
2 | #define _LINUX_MM_H | |
3 | ||
4 | #include <linux/errno.h> | |
5 | ||
6 | #ifdef __KERNEL__ | |
7 | ||
8 | #include <linux/mmdebug.h> | |
9 | #include <linux/gfp.h> | |
10 | #include <linux/bug.h> | |
11 | #include <linux/list.h> | |
12 | #include <linux/mmzone.h> | |
13 | #include <linux/rbtree.h> | |
14 | #include <linux/atomic.h> | |
15 | #include <linux/debug_locks.h> | |
16 | #include <linux/mm_types.h> | |
17 | #include <linux/range.h> | |
18 | #include <linux/pfn.h> | |
19 | #include <linux/percpu-refcount.h> | |
20 | #include <linux/bit_spinlock.h> | |
21 | #include <linux/shrinker.h> | |
22 | #include <linux/resource.h> | |
23 | #include <linux/page_ext.h> | |
24 | #include <linux/err.h> | |
25 | #include <linux/page_ref.h> | |
26 | ||
27 | struct mempolicy; | |
28 | struct anon_vma; | |
29 | struct anon_vma_chain; | |
30 | struct file_ra_state; | |
31 | struct user_struct; | |
32 | struct writeback_control; | |
33 | struct bdi_writeback; | |
34 | ||
35 | #ifndef CONFIG_NEED_MULTIPLE_NODES /* Don't use mapnrs, do it properly */ | |
36 | extern unsigned long max_mapnr; | |
37 | ||
38 | static inline void set_max_mapnr(unsigned long limit) | |
39 | { | |
40 | max_mapnr = limit; | |
41 | } | |
42 | #else | |
43 | static inline void set_max_mapnr(unsigned long limit) { } | |
44 | #endif | |
45 | ||
46 | extern unsigned long totalram_pages; | |
47 | extern void * high_memory; | |
48 | extern int page_cluster; | |
49 | ||
50 | #ifdef CONFIG_SYSCTL | |
51 | extern int sysctl_legacy_va_layout; | |
52 | #else | |
53 | #define sysctl_legacy_va_layout 0 | |
54 | #endif | |
55 | ||
56 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_BITS | |
57 | extern const int mmap_rnd_bits_min; | |
58 | extern const int mmap_rnd_bits_max; | |
59 | extern int mmap_rnd_bits __read_mostly; | |
60 | #endif | |
61 | #ifdef CONFIG_HAVE_ARCH_MMAP_RND_COMPAT_BITS | |
62 | extern const int mmap_rnd_compat_bits_min; | |
63 | extern const int mmap_rnd_compat_bits_max; | |
64 | extern int mmap_rnd_compat_bits __read_mostly; | |
65 | #endif | |
66 | ||
67 | #include <asm/page.h> | |
68 | #include <asm/pgtable.h> | |
69 | #include <asm/processor.h> | |
70 | ||
71 | #ifndef __pa_symbol | |
72 | #define __pa_symbol(x) __pa(RELOC_HIDE((unsigned long)(x), 0)) | |
73 | #endif | |
74 | ||
75 | #ifndef page_to_virt | |
76 | #define page_to_virt(x) __va(PFN_PHYS(page_to_pfn(x))) | |
77 | #endif | |
78 | ||
79 | /* | |
80 | * To prevent common memory management code establishing | |
81 | * a zero page mapping on a read fault. | |
82 | * This macro should be defined within <asm/pgtable.h>. | |
83 | * s390 does this to prevent multiplexing of hardware bits | |
84 | * related to the physical page in case of virtualization. | |
85 | */ | |
86 | #ifndef mm_forbids_zeropage | |
87 | #define mm_forbids_zeropage(X) (0) | |
88 | #endif | |
89 | ||
90 | /* | |
91 | * Default maximum number of active map areas, this limits the number of vmas | |
92 | * per mm struct. Users can overwrite this number by sysctl but there is a | |
93 | * problem. | |
94 | * | |
95 | * When a program's coredump is generated as ELF format, a section is created | |
96 | * per a vma. In ELF, the number of sections is represented in unsigned short. | |
97 | * This means the number of sections should be smaller than 65535 at coredump. | |
98 | * Because the kernel adds some informative sections to a image of program at | |
99 | * generating coredump, we need some margin. The number of extra sections is | |
100 | * 1-3 now and depends on arch. We use "5" as safe margin, here. | |
101 | * | |
102 | * ELF extended numbering allows more than 65535 sections, so 16-bit bound is | |
103 | * not a hard limit any more. Although some userspace tools can be surprised by | |
104 | * that. | |
105 | */ | |
106 | #define MAPCOUNT_ELF_CORE_MARGIN (5) | |
107 | #define DEFAULT_MAX_MAP_COUNT (USHRT_MAX - MAPCOUNT_ELF_CORE_MARGIN) | |
108 | ||
109 | extern int sysctl_max_map_count; | |
110 | ||
111 | extern unsigned long sysctl_user_reserve_kbytes; | |
112 | extern unsigned long sysctl_admin_reserve_kbytes; | |
113 | ||
114 | extern int sysctl_overcommit_memory; | |
115 | extern int sysctl_overcommit_ratio; | |
116 | extern unsigned long sysctl_overcommit_kbytes; | |
117 | ||
118 | extern int overcommit_ratio_handler(struct ctl_table *, int, void __user *, | |
119 | size_t *, loff_t *); | |
120 | extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *, | |
121 | size_t *, loff_t *); | |
122 | ||
123 | #define nth_page(page,n) pfn_to_page(page_to_pfn((page)) + (n)) | |
124 | ||
125 | /* to align the pointer to the (next) page boundary */ | |
126 | #define PAGE_ALIGN(addr) ALIGN(addr, PAGE_SIZE) | |
127 | ||
128 | /* test whether an address (unsigned long or pointer) is aligned to PAGE_SIZE */ | |
129 | #define PAGE_ALIGNED(addr) IS_ALIGNED((unsigned long)(addr), PAGE_SIZE) | |
130 | ||
131 | /* | |
132 | * Linux kernel virtual memory manager primitives. | |
133 | * The idea being to have a "virtual" mm in the same way | |
134 | * we have a virtual fs - giving a cleaner interface to the | |
135 | * mm details, and allowing different kinds of memory mappings | |
136 | * (from shared memory to executable loading to arbitrary | |
137 | * mmap() functions). | |
138 | */ | |
139 | ||
140 | extern struct kmem_cache *vm_area_cachep; | |
141 | ||
142 | #ifndef CONFIG_MMU | |
143 | extern struct rb_root nommu_region_tree; | |
144 | extern struct rw_semaphore nommu_region_sem; | |
145 | ||
146 | extern unsigned int kobjsize(const void *objp); | |
147 | #endif | |
148 | ||
149 | /* | |
150 | * vm_flags in vm_area_struct, see mm_types.h. | |
151 | * When changing, update also include/trace/events/mmflags.h | |
152 | */ | |
153 | #define VM_NONE 0x00000000 | |
154 | ||
155 | #define VM_READ 0x00000001 /* currently active flags */ | |
156 | #define VM_WRITE 0x00000002 | |
157 | #define VM_EXEC 0x00000004 | |
158 | #define VM_SHARED 0x00000008 | |
159 | ||
160 | /* mprotect() hardcodes VM_MAYREAD >> 4 == VM_READ, and so for r/w/x bits. */ | |
161 | #define VM_MAYREAD 0x00000010 /* limits for mprotect() etc */ | |
162 | #define VM_MAYWRITE 0x00000020 | |
163 | #define VM_MAYEXEC 0x00000040 | |
164 | #define VM_MAYSHARE 0x00000080 | |
165 | ||
166 | #define VM_GROWSDOWN 0x00000100 /* general info on the segment */ | |
167 | #define VM_UFFD_MISSING 0x00000200 /* missing pages tracking */ | |
168 | #define VM_PFNMAP 0x00000400 /* Page-ranges managed without "struct page", just pure PFN */ | |
169 | #define VM_DENYWRITE 0x00000800 /* ETXTBSY on write attempts.. */ | |
170 | #define VM_UFFD_WP 0x00001000 /* wrprotect pages tracking */ | |
171 | ||
172 | #define VM_LOCKED 0x00002000 | |
173 | #define VM_IO 0x00004000 /* Memory mapped I/O or similar */ | |
174 | ||
175 | /* Used by sys_madvise() */ | |
176 | #define VM_SEQ_READ 0x00008000 /* App will access data sequentially */ | |
177 | #define VM_RAND_READ 0x00010000 /* App will not benefit from clustered reads */ | |
178 | ||
179 | #define VM_DONTCOPY 0x00020000 /* Do not copy this vma on fork */ | |
180 | #define VM_DONTEXPAND 0x00040000 /* Cannot expand with mremap() */ | |
181 | #define VM_LOCKONFAULT 0x00080000 /* Lock the pages covered when they are faulted in */ | |
182 | #define VM_ACCOUNT 0x00100000 /* Is a VM accounted object */ | |
183 | #define VM_NORESERVE 0x00200000 /* should the VM suppress accounting */ | |
184 | #define VM_HUGETLB 0x00400000 /* Huge TLB Page VM */ | |
185 | #define VM_ARCH_1 0x01000000 /* Architecture-specific flag */ | |
186 | #define VM_ARCH_2 0x02000000 | |
187 | #define VM_DONTDUMP 0x04000000 /* Do not include in the core dump */ | |
188 | ||
189 | #ifdef CONFIG_MEM_SOFT_DIRTY | |
190 | # define VM_SOFTDIRTY 0x08000000 /* Not soft dirty clean area */ | |
191 | #else | |
192 | # define VM_SOFTDIRTY 0 | |
193 | #endif | |
194 | ||
195 | #define VM_MIXEDMAP 0x10000000 /* Can contain "struct page" and pure PFN pages */ | |
196 | #define VM_HUGEPAGE 0x20000000 /* MADV_HUGEPAGE marked this vma */ | |
197 | #define VM_NOHUGEPAGE 0x40000000 /* MADV_NOHUGEPAGE marked this vma */ | |
198 | #define VM_MERGEABLE 0x80000000 /* KSM may merge identical pages */ | |
199 | ||
200 | #ifdef CONFIG_ARCH_USES_HIGH_VMA_FLAGS | |
201 | #define VM_HIGH_ARCH_BIT_0 32 /* bit only usable on 64-bit architectures */ | |
202 | #define VM_HIGH_ARCH_BIT_1 33 /* bit only usable on 64-bit architectures */ | |
203 | #define VM_HIGH_ARCH_BIT_2 34 /* bit only usable on 64-bit architectures */ | |
204 | #define VM_HIGH_ARCH_BIT_3 35 /* bit only usable on 64-bit architectures */ | |
205 | #define VM_HIGH_ARCH_0 BIT(VM_HIGH_ARCH_BIT_0) | |
206 | #define VM_HIGH_ARCH_1 BIT(VM_HIGH_ARCH_BIT_1) | |
207 | #define VM_HIGH_ARCH_2 BIT(VM_HIGH_ARCH_BIT_2) | |
208 | #define VM_HIGH_ARCH_3 BIT(VM_HIGH_ARCH_BIT_3) | |
209 | #endif /* CONFIG_ARCH_USES_HIGH_VMA_FLAGS */ | |
210 | ||
211 | #if defined(CONFIG_X86) | |
212 | # define VM_PAT VM_ARCH_1 /* PAT reserves whole VMA at once (x86) */ | |
213 | #if defined (CONFIG_X86_INTEL_MEMORY_PROTECTION_KEYS) | |
214 | # define VM_PKEY_SHIFT VM_HIGH_ARCH_BIT_0 | |
215 | # define VM_PKEY_BIT0 VM_HIGH_ARCH_0 /* A protection key is a 4-bit value */ | |
216 | # define VM_PKEY_BIT1 VM_HIGH_ARCH_1 | |
217 | # define VM_PKEY_BIT2 VM_HIGH_ARCH_2 | |
218 | # define VM_PKEY_BIT3 VM_HIGH_ARCH_3 | |
219 | #endif | |
220 | #elif defined(CONFIG_PPC) | |
221 | # define VM_SAO VM_ARCH_1 /* Strong Access Ordering (powerpc) */ | |
222 | #elif defined(CONFIG_PARISC) | |
223 | # define VM_GROWSUP VM_ARCH_1 | |
224 | #elif defined(CONFIG_METAG) | |
225 | # define VM_GROWSUP VM_ARCH_1 | |
226 | #elif defined(CONFIG_IA64) | |
227 | # define VM_GROWSUP VM_ARCH_1 | |
228 | #elif !defined(CONFIG_MMU) | |
229 | # define VM_MAPPED_COPY VM_ARCH_1 /* T if mapped copy of data (nommu mmap) */ | |
230 | #endif | |
231 | ||
232 | #if defined(CONFIG_X86) | |
233 | /* MPX specific bounds table or bounds directory */ | |
234 | # define VM_MPX VM_ARCH_2 | |
235 | #endif | |
236 | ||
237 | #ifndef VM_GROWSUP | |
238 | # define VM_GROWSUP VM_NONE | |
239 | #endif | |
240 | ||
241 | /* Bits set in the VMA until the stack is in its final location */ | |
242 | #define VM_STACK_INCOMPLETE_SETUP (VM_RAND_READ | VM_SEQ_READ) | |
243 | ||
244 | #ifndef VM_STACK_DEFAULT_FLAGS /* arch can override this */ | |
245 | #define VM_STACK_DEFAULT_FLAGS VM_DATA_DEFAULT_FLAGS | |
246 | #endif | |
247 | ||
248 | #ifdef CONFIG_STACK_GROWSUP | |
249 | #define VM_STACK VM_GROWSUP | |
250 | #else | |
251 | #define VM_STACK VM_GROWSDOWN | |
252 | #endif | |
253 | ||
254 | #define VM_STACK_FLAGS (VM_STACK | VM_STACK_DEFAULT_FLAGS | VM_ACCOUNT) | |
255 | ||
256 | /* | |
257 | * Special vmas that are non-mergable, non-mlock()able. | |
258 | * Note: mm/huge_memory.c VM_NO_THP depends on this definition. | |
259 | */ | |
260 | #define VM_SPECIAL (VM_IO | VM_DONTEXPAND | VM_PFNMAP | VM_MIXEDMAP) | |
261 | ||
262 | /* This mask defines which mm->def_flags a process can inherit its parent */ | |
263 | #define VM_INIT_DEF_MASK VM_NOHUGEPAGE | |
264 | ||
265 | /* This mask is used to clear all the VMA flags used by mlock */ | |
266 | #define VM_LOCKED_CLEAR_MASK (~(VM_LOCKED | VM_LOCKONFAULT)) | |
267 | ||
268 | /* | |
269 | * mapping from the currently active vm_flags protection bits (the | |
270 | * low four bits) to a page protection mask.. | |
271 | */ | |
272 | extern pgprot_t protection_map[16]; | |
273 | ||
274 | #define FAULT_FLAG_WRITE 0x01 /* Fault was a write access */ | |
275 | #define FAULT_FLAG_MKWRITE 0x02 /* Fault was mkwrite of existing pte */ | |
276 | #define FAULT_FLAG_ALLOW_RETRY 0x04 /* Retry fault if blocking */ | |
277 | #define FAULT_FLAG_RETRY_NOWAIT 0x08 /* Don't drop mmap_sem and wait when retrying */ | |
278 | #define FAULT_FLAG_KILLABLE 0x10 /* The fault task is in SIGKILL killable region */ | |
279 | #define FAULT_FLAG_TRIED 0x20 /* Second try */ | |
280 | #define FAULT_FLAG_USER 0x40 /* The fault originated in userspace */ | |
281 | #define FAULT_FLAG_REMOTE 0x80 /* faulting for non current tsk/mm */ | |
282 | #define FAULT_FLAG_INSTRUCTION 0x100 /* The fault was during an instruction fetch */ | |
283 | ||
284 | /* | |
285 | * vm_fault is filled by the the pagefault handler and passed to the vma's | |
286 | * ->fault function. The vma's ->fault is responsible for returning a bitmask | |
287 | * of VM_FAULT_xxx flags that give details about how the fault was handled. | |
288 | * | |
289 | * MM layer fills up gfp_mask for page allocations but fault handler might | |
290 | * alter it if its implementation requires a different allocation context. | |
291 | * | |
292 | * pgoff should be used in favour of virtual_address, if possible. | |
293 | */ | |
294 | struct vm_fault { | |
295 | struct vm_area_struct *vma; /* Target VMA */ | |
296 | unsigned int flags; /* FAULT_FLAG_xxx flags */ | |
297 | gfp_t gfp_mask; /* gfp mask to be used for allocations */ | |
298 | pgoff_t pgoff; /* Logical page offset based on vma */ | |
299 | unsigned long address; /* Faulting virtual address */ | |
300 | pmd_t *pmd; /* Pointer to pmd entry matching | |
301 | * the 'address' | |
302 | */ | |
303 | ||
304 | struct page *cow_page; /* Handler may choose to COW */ | |
305 | struct page *page; /* ->fault handlers should return a | |
306 | * page here, unless VM_FAULT_NOPAGE | |
307 | * is set (which is also implied by | |
308 | * VM_FAULT_ERROR). | |
309 | */ | |
310 | void *entry; /* ->fault handler can alternatively | |
311 | * return locked DAX entry. In that | |
312 | * case handler should return | |
313 | * VM_FAULT_DAX_LOCKED and fill in | |
314 | * entry here. | |
315 | */ | |
316 | /* These three entries are valid only while holding ptl lock */ | |
317 | pte_t *pte; /* Pointer to pte entry matching | |
318 | * the 'address'. NULL if the page | |
319 | * table hasn't been allocated. | |
320 | */ | |
321 | spinlock_t *ptl; /* Page table lock. | |
322 | * Protects pte page table if 'pte' | |
323 | * is not NULL, otherwise pmd. | |
324 | */ | |
325 | pgtable_t prealloc_pte; /* Pre-allocated pte page table. | |
326 | * vm_ops->map_pages() calls | |
327 | * alloc_set_pte() from atomic context. | |
328 | * do_fault_around() pre-allocates | |
329 | * page table to avoid allocation from | |
330 | * atomic context. | |
331 | */ | |
332 | }; | |
333 | ||
334 | /* | |
335 | * These are the virtual MM functions - opening of an area, closing and | |
336 | * unmapping it (needed to keep files on disk up-to-date etc), pointer | |
337 | * to the functions called when a no-page or a wp-page exception occurs. | |
338 | */ | |
339 | struct vm_operations_struct { | |
340 | void (*open)(struct vm_area_struct * area); | |
341 | void (*close)(struct vm_area_struct * area); | |
342 | int (*mremap)(struct vm_area_struct * area); | |
343 | int (*fault)(struct vm_area_struct *vma, struct vm_fault *vmf); | |
344 | int (*pmd_fault)(struct vm_area_struct *, unsigned long address, | |
345 | pmd_t *, unsigned int flags); | |
346 | void (*map_pages)(struct vm_fault *vmf, | |
347 | pgoff_t start_pgoff, pgoff_t end_pgoff); | |
348 | ||
349 | /* notification that a previously read-only page is about to become | |
350 | * writable, if an error is returned it will cause a SIGBUS */ | |
351 | int (*page_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); | |
352 | ||
353 | /* same as page_mkwrite when using VM_PFNMAP|VM_MIXEDMAP */ | |
354 | int (*pfn_mkwrite)(struct vm_area_struct *vma, struct vm_fault *vmf); | |
355 | ||
356 | /* called by access_process_vm when get_user_pages() fails, typically | |
357 | * for use by special VMAs that can switch between memory and hardware | |
358 | */ | |
359 | int (*access)(struct vm_area_struct *vma, unsigned long addr, | |
360 | void *buf, int len, int write); | |
361 | ||
362 | /* Called by the /proc/PID/maps code to ask the vma whether it | |
363 | * has a special name. Returning non-NULL will also cause this | |
364 | * vma to be dumped unconditionally. */ | |
365 | const char *(*name)(struct vm_area_struct *vma); | |
366 | ||
367 | #ifdef CONFIG_NUMA | |
368 | /* | |
369 | * set_policy() op must add a reference to any non-NULL @new mempolicy | |
370 | * to hold the policy upon return. Caller should pass NULL @new to | |
371 | * remove a policy and fall back to surrounding context--i.e. do not | |
372 | * install a MPOL_DEFAULT policy, nor the task or system default | |
373 | * mempolicy. | |
374 | */ | |
375 | int (*set_policy)(struct vm_area_struct *vma, struct mempolicy *new); | |
376 | ||
377 | /* | |
378 | * get_policy() op must add reference [mpol_get()] to any policy at | |
379 | * (vma,addr) marked as MPOL_SHARED. The shared policy infrastructure | |
380 | * in mm/mempolicy.c will do this automatically. | |
381 | * get_policy() must NOT add a ref if the policy at (vma,addr) is not | |
382 | * marked as MPOL_SHARED. vma policies are protected by the mmap_sem. | |
383 | * If no [shared/vma] mempolicy exists at the addr, get_policy() op | |
384 | * must return NULL--i.e., do not "fallback" to task or system default | |
385 | * policy. | |
386 | */ | |
387 | struct mempolicy *(*get_policy)(struct vm_area_struct *vma, | |
388 | unsigned long addr); | |
389 | #endif | |
390 | /* | |
391 | * Called by vm_normal_page() for special PTEs to find the | |
392 | * page for @addr. This is useful if the default behavior | |
393 | * (using pte_page()) would not find the correct page. | |
394 | */ | |
395 | struct page *(*find_special_page)(struct vm_area_struct *vma, | |
396 | unsigned long addr); | |
397 | }; | |
398 | ||
399 | struct mmu_gather; | |
400 | struct inode; | |
401 | ||
402 | #define page_private(page) ((page)->private) | |
403 | #define set_page_private(page, v) ((page)->private = (v)) | |
404 | ||
405 | #if !defined(__HAVE_ARCH_PTE_DEVMAP) || !defined(CONFIG_TRANSPARENT_HUGEPAGE) | |
406 | static inline int pmd_devmap(pmd_t pmd) | |
407 | { | |
408 | return 0; | |
409 | } | |
410 | #endif | |
411 | ||
412 | /* | |
413 | * FIXME: take this include out, include page-flags.h in | |
414 | * files which need it (119 of them) | |
415 | */ | |
416 | #include <linux/page-flags.h> | |
417 | #include <linux/huge_mm.h> | |
418 | ||
419 | /* | |
420 | * Methods to modify the page usage count. | |
421 | * | |
422 | * What counts for a page usage: | |
423 | * - cache mapping (page->mapping) | |
424 | * - private data (page->private) | |
425 | * - page mapped in a task's page tables, each mapping | |
426 | * is counted separately | |
427 | * | |
428 | * Also, many kernel routines increase the page count before a critical | |
429 | * routine so they can be sure the page doesn't go away from under them. | |
430 | */ | |
431 | ||
432 | /* | |
433 | * Drop a ref, return true if the refcount fell to zero (the page has no users) | |
434 | */ | |
435 | static inline int put_page_testzero(struct page *page) | |
436 | { | |
437 | VM_BUG_ON_PAGE(page_ref_count(page) == 0, page); | |
438 | return page_ref_dec_and_test(page); | |
439 | } | |
440 | ||
441 | /* | |
442 | * Try to grab a ref unless the page has a refcount of zero, return false if | |
443 | * that is the case. | |
444 | * This can be called when MMU is off so it must not access | |
445 | * any of the virtual mappings. | |
446 | */ | |
447 | static inline int get_page_unless_zero(struct page *page) | |
448 | { | |
449 | return page_ref_add_unless(page, 1, 0); | |
450 | } | |
451 | ||
452 | extern int page_is_ram(unsigned long pfn); | |
453 | ||
454 | enum { | |
455 | REGION_INTERSECTS, | |
456 | REGION_DISJOINT, | |
457 | REGION_MIXED, | |
458 | }; | |
459 | ||
460 | int region_intersects(resource_size_t offset, size_t size, unsigned long flags, | |
461 | unsigned long desc); | |
462 | ||
463 | /* Support for virtually mapped pages */ | |
464 | struct page *vmalloc_to_page(const void *addr); | |
465 | unsigned long vmalloc_to_pfn(const void *addr); | |
466 | ||
467 | /* | |
468 | * Determine if an address is within the vmalloc range | |
469 | * | |
470 | * On nommu, vmalloc/vfree wrap through kmalloc/kfree directly, so there | |
471 | * is no special casing required. | |
472 | */ | |
473 | static inline bool is_vmalloc_addr(const void *x) | |
474 | { | |
475 | #ifdef CONFIG_MMU | |
476 | unsigned long addr = (unsigned long)x; | |
477 | ||
478 | return addr >= VMALLOC_START && addr < VMALLOC_END; | |
479 | #else | |
480 | return false; | |
481 | #endif | |
482 | } | |
483 | #ifdef CONFIG_MMU | |
484 | extern int is_vmalloc_or_module_addr(const void *x); | |
485 | #else | |
486 | static inline int is_vmalloc_or_module_addr(const void *x) | |
487 | { | |
488 | return 0; | |
489 | } | |
490 | #endif | |
491 | ||
492 | extern void kvfree(const void *addr); | |
493 | ||
494 | static inline atomic_t *compound_mapcount_ptr(struct page *page) | |
495 | { | |
496 | return &page[1].compound_mapcount; | |
497 | } | |
498 | ||
499 | static inline int compound_mapcount(struct page *page) | |
500 | { | |
501 | VM_BUG_ON_PAGE(!PageCompound(page), page); | |
502 | page = compound_head(page); | |
503 | return atomic_read(compound_mapcount_ptr(page)) + 1; | |
504 | } | |
505 | ||
506 | /* | |
507 | * The atomic page->_mapcount, starts from -1: so that transitions | |
508 | * both from it and to it can be tracked, using atomic_inc_and_test | |
509 | * and atomic_add_negative(-1). | |
510 | */ | |
511 | static inline void page_mapcount_reset(struct page *page) | |
512 | { | |
513 | atomic_set(&(page)->_mapcount, -1); | |
514 | } | |
515 | ||
516 | int __page_mapcount(struct page *page); | |
517 | ||
518 | static inline int page_mapcount(struct page *page) | |
519 | { | |
520 | VM_BUG_ON_PAGE(PageSlab(page), page); | |
521 | ||
522 | if (unlikely(PageCompound(page))) | |
523 | return __page_mapcount(page); | |
524 | return atomic_read(&page->_mapcount) + 1; | |
525 | } | |
526 | ||
527 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
528 | int total_mapcount(struct page *page); | |
529 | int page_trans_huge_mapcount(struct page *page, int *total_mapcount); | |
530 | #else | |
531 | static inline int total_mapcount(struct page *page) | |
532 | { | |
533 | return page_mapcount(page); | |
534 | } | |
535 | static inline int page_trans_huge_mapcount(struct page *page, | |
536 | int *total_mapcount) | |
537 | { | |
538 | int mapcount = page_mapcount(page); | |
539 | if (total_mapcount) | |
540 | *total_mapcount = mapcount; | |
541 | return mapcount; | |
542 | } | |
543 | #endif | |
544 | ||
545 | static inline struct page *virt_to_head_page(const void *x) | |
546 | { | |
547 | struct page *page = virt_to_page(x); | |
548 | ||
549 | return compound_head(page); | |
550 | } | |
551 | ||
552 | void __put_page(struct page *page); | |
553 | ||
554 | void put_pages_list(struct list_head *pages); | |
555 | ||
556 | void split_page(struct page *page, unsigned int order); | |
557 | ||
558 | /* | |
559 | * Compound pages have a destructor function. Provide a | |
560 | * prototype for that function and accessor functions. | |
561 | * These are _only_ valid on the head of a compound page. | |
562 | */ | |
563 | typedef void compound_page_dtor(struct page *); | |
564 | ||
565 | /* Keep the enum in sync with compound_page_dtors array in mm/page_alloc.c */ | |
566 | enum compound_dtor_id { | |
567 | NULL_COMPOUND_DTOR, | |
568 | COMPOUND_PAGE_DTOR, | |
569 | #ifdef CONFIG_HUGETLB_PAGE | |
570 | HUGETLB_PAGE_DTOR, | |
571 | #endif | |
572 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
573 | TRANSHUGE_PAGE_DTOR, | |
574 | #endif | |
575 | NR_COMPOUND_DTORS, | |
576 | }; | |
577 | extern compound_page_dtor * const compound_page_dtors[]; | |
578 | ||
579 | static inline void set_compound_page_dtor(struct page *page, | |
580 | enum compound_dtor_id compound_dtor) | |
581 | { | |
582 | VM_BUG_ON_PAGE(compound_dtor >= NR_COMPOUND_DTORS, page); | |
583 | page[1].compound_dtor = compound_dtor; | |
584 | } | |
585 | ||
586 | static inline compound_page_dtor *get_compound_page_dtor(struct page *page) | |
587 | { | |
588 | VM_BUG_ON_PAGE(page[1].compound_dtor >= NR_COMPOUND_DTORS, page); | |
589 | return compound_page_dtors[page[1].compound_dtor]; | |
590 | } | |
591 | ||
592 | static inline unsigned int compound_order(struct page *page) | |
593 | { | |
594 | if (!PageHead(page)) | |
595 | return 0; | |
596 | return page[1].compound_order; | |
597 | } | |
598 | ||
599 | static inline void set_compound_order(struct page *page, unsigned int order) | |
600 | { | |
601 | page[1].compound_order = order; | |
602 | } | |
603 | ||
604 | void free_compound_page(struct page *page); | |
605 | ||
606 | #ifdef CONFIG_MMU | |
607 | /* | |
608 | * Do pte_mkwrite, but only if the vma says VM_WRITE. We do this when | |
609 | * servicing faults for write access. In the normal case, do always want | |
610 | * pte_mkwrite. But get_user_pages can cause write faults for mappings | |
611 | * that do not have writing enabled, when used by access_process_vm. | |
612 | */ | |
613 | static inline pte_t maybe_mkwrite(pte_t pte, struct vm_area_struct *vma) | |
614 | { | |
615 | if (likely(vma->vm_flags & VM_WRITE)) | |
616 | pte = pte_mkwrite(pte); | |
617 | return pte; | |
618 | } | |
619 | ||
620 | int alloc_set_pte(struct vm_fault *vmf, struct mem_cgroup *memcg, | |
621 | struct page *page); | |
622 | #endif | |
623 | ||
624 | /* | |
625 | * Multiple processes may "see" the same page. E.g. for untouched | |
626 | * mappings of /dev/null, all processes see the same page full of | |
627 | * zeroes, and text pages of executables and shared libraries have | |
628 | * only one copy in memory, at most, normally. | |
629 | * | |
630 | * For the non-reserved pages, page_count(page) denotes a reference count. | |
631 | * page_count() == 0 means the page is free. page->lru is then used for | |
632 | * freelist management in the buddy allocator. | |
633 | * page_count() > 0 means the page has been allocated. | |
634 | * | |
635 | * Pages are allocated by the slab allocator in order to provide memory | |
636 | * to kmalloc and kmem_cache_alloc. In this case, the management of the | |
637 | * page, and the fields in 'struct page' are the responsibility of mm/slab.c | |
638 | * unless a particular usage is carefully commented. (the responsibility of | |
639 | * freeing the kmalloc memory is the caller's, of course). | |
640 | * | |
641 | * A page may be used by anyone else who does a __get_free_page(). | |
642 | * In this case, page_count still tracks the references, and should only | |
643 | * be used through the normal accessor functions. The top bits of page->flags | |
644 | * and page->virtual store page management information, but all other fields | |
645 | * are unused and could be used privately, carefully. The management of this | |
646 | * page is the responsibility of the one who allocated it, and those who have | |
647 | * subsequently been given references to it. | |
648 | * | |
649 | * The other pages (we may call them "pagecache pages") are completely | |
650 | * managed by the Linux memory manager: I/O, buffers, swapping etc. | |
651 | * The following discussion applies only to them. | |
652 | * | |
653 | * A pagecache page contains an opaque `private' member, which belongs to the | |
654 | * page's address_space. Usually, this is the address of a circular list of | |
655 | * the page's disk buffers. PG_private must be set to tell the VM to call | |
656 | * into the filesystem to release these pages. | |
657 | * | |
658 | * A page may belong to an inode's memory mapping. In this case, page->mapping | |
659 | * is the pointer to the inode, and page->index is the file offset of the page, | |
660 | * in units of PAGE_SIZE. | |
661 | * | |
662 | * If pagecache pages are not associated with an inode, they are said to be | |
663 | * anonymous pages. These may become associated with the swapcache, and in that | |
664 | * case PG_swapcache is set, and page->private is an offset into the swapcache. | |
665 | * | |
666 | * In either case (swapcache or inode backed), the pagecache itself holds one | |
667 | * reference to the page. Setting PG_private should also increment the | |
668 | * refcount. The each user mapping also has a reference to the page. | |
669 | * | |
670 | * The pagecache pages are stored in a per-mapping radix tree, which is | |
671 | * rooted at mapping->page_tree, and indexed by offset. | |
672 | * Where 2.4 and early 2.6 kernels kept dirty/clean pages in per-address_space | |
673 | * lists, we instead now tag pages as dirty/writeback in the radix tree. | |
674 | * | |
675 | * All pagecache pages may be subject to I/O: | |
676 | * - inode pages may need to be read from disk, | |
677 | * - inode pages which have been modified and are MAP_SHARED may need | |
678 | * to be written back to the inode on disk, | |
679 | * - anonymous pages (including MAP_PRIVATE file mappings) which have been | |
680 | * modified may need to be swapped out to swap space and (later) to be read | |
681 | * back into memory. | |
682 | */ | |
683 | ||
684 | /* | |
685 | * The zone field is never updated after free_area_init_core() | |
686 | * sets it, so none of the operations on it need to be atomic. | |
687 | */ | |
688 | ||
689 | /* Page flags: | [SECTION] | [NODE] | ZONE | [LAST_CPUPID] | ... | FLAGS | */ | |
690 | #define SECTIONS_PGOFF ((sizeof(unsigned long)*8) - SECTIONS_WIDTH) | |
691 | #define NODES_PGOFF (SECTIONS_PGOFF - NODES_WIDTH) | |
692 | #define ZONES_PGOFF (NODES_PGOFF - ZONES_WIDTH) | |
693 | #define LAST_CPUPID_PGOFF (ZONES_PGOFF - LAST_CPUPID_WIDTH) | |
694 | ||
695 | /* | |
696 | * Define the bit shifts to access each section. For non-existent | |
697 | * sections we define the shift as 0; that plus a 0 mask ensures | |
698 | * the compiler will optimise away reference to them. | |
699 | */ | |
700 | #define SECTIONS_PGSHIFT (SECTIONS_PGOFF * (SECTIONS_WIDTH != 0)) | |
701 | #define NODES_PGSHIFT (NODES_PGOFF * (NODES_WIDTH != 0)) | |
702 | #define ZONES_PGSHIFT (ZONES_PGOFF * (ZONES_WIDTH != 0)) | |
703 | #define LAST_CPUPID_PGSHIFT (LAST_CPUPID_PGOFF * (LAST_CPUPID_WIDTH != 0)) | |
704 | ||
705 | /* NODE:ZONE or SECTION:ZONE is used to ID a zone for the buddy allocator */ | |
706 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
707 | #define ZONEID_SHIFT (SECTIONS_SHIFT + ZONES_SHIFT) | |
708 | #define ZONEID_PGOFF ((SECTIONS_PGOFF < ZONES_PGOFF)? \ | |
709 | SECTIONS_PGOFF : ZONES_PGOFF) | |
710 | #else | |
711 | #define ZONEID_SHIFT (NODES_SHIFT + ZONES_SHIFT) | |
712 | #define ZONEID_PGOFF ((NODES_PGOFF < ZONES_PGOFF)? \ | |
713 | NODES_PGOFF : ZONES_PGOFF) | |
714 | #endif | |
715 | ||
716 | #define ZONEID_PGSHIFT (ZONEID_PGOFF * (ZONEID_SHIFT != 0)) | |
717 | ||
718 | #if SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | |
719 | #error SECTIONS_WIDTH+NODES_WIDTH+ZONES_WIDTH > BITS_PER_LONG - NR_PAGEFLAGS | |
720 | #endif | |
721 | ||
722 | #define ZONES_MASK ((1UL << ZONES_WIDTH) - 1) | |
723 | #define NODES_MASK ((1UL << NODES_WIDTH) - 1) | |
724 | #define SECTIONS_MASK ((1UL << SECTIONS_WIDTH) - 1) | |
725 | #define LAST_CPUPID_MASK ((1UL << LAST_CPUPID_SHIFT) - 1) | |
726 | #define ZONEID_MASK ((1UL << ZONEID_SHIFT) - 1) | |
727 | ||
728 | static inline enum zone_type page_zonenum(const struct page *page) | |
729 | { | |
730 | return (page->flags >> ZONES_PGSHIFT) & ZONES_MASK; | |
731 | } | |
732 | ||
733 | #ifdef CONFIG_ZONE_DEVICE | |
734 | void get_zone_device_page(struct page *page); | |
735 | void put_zone_device_page(struct page *page); | |
736 | static inline bool is_zone_device_page(const struct page *page) | |
737 | { | |
738 | return page_zonenum(page) == ZONE_DEVICE; | |
739 | } | |
740 | #else | |
741 | static inline void get_zone_device_page(struct page *page) | |
742 | { | |
743 | } | |
744 | static inline void put_zone_device_page(struct page *page) | |
745 | { | |
746 | } | |
747 | static inline bool is_zone_device_page(const struct page *page) | |
748 | { | |
749 | return false; | |
750 | } | |
751 | #endif | |
752 | ||
753 | static inline void get_page(struct page *page) | |
754 | { | |
755 | page = compound_head(page); | |
756 | /* | |
757 | * Getting a normal page or the head of a compound page | |
758 | * requires to already have an elevated page->_refcount. | |
759 | */ | |
760 | VM_BUG_ON_PAGE(page_ref_count(page) <= 0, page); | |
761 | page_ref_inc(page); | |
762 | ||
763 | if (unlikely(is_zone_device_page(page))) | |
764 | get_zone_device_page(page); | |
765 | } | |
766 | ||
767 | static inline void put_page(struct page *page) | |
768 | { | |
769 | page = compound_head(page); | |
770 | ||
771 | if (put_page_testzero(page)) | |
772 | __put_page(page); | |
773 | ||
774 | if (unlikely(is_zone_device_page(page))) | |
775 | put_zone_device_page(page); | |
776 | } | |
777 | ||
778 | #if defined(CONFIG_SPARSEMEM) && !defined(CONFIG_SPARSEMEM_VMEMMAP) | |
779 | #define SECTION_IN_PAGE_FLAGS | |
780 | #endif | |
781 | ||
782 | /* | |
783 | * The identification function is mainly used by the buddy allocator for | |
784 | * determining if two pages could be buddies. We are not really identifying | |
785 | * the zone since we could be using the section number id if we do not have | |
786 | * node id available in page flags. | |
787 | * We only guarantee that it will return the same value for two combinable | |
788 | * pages in a zone. | |
789 | */ | |
790 | static inline int page_zone_id(struct page *page) | |
791 | { | |
792 | return (page->flags >> ZONEID_PGSHIFT) & ZONEID_MASK; | |
793 | } | |
794 | ||
795 | static inline int zone_to_nid(struct zone *zone) | |
796 | { | |
797 | #ifdef CONFIG_NUMA | |
798 | return zone->node; | |
799 | #else | |
800 | return 0; | |
801 | #endif | |
802 | } | |
803 | ||
804 | #ifdef NODE_NOT_IN_PAGE_FLAGS | |
805 | extern int page_to_nid(const struct page *page); | |
806 | #else | |
807 | static inline int page_to_nid(const struct page *page) | |
808 | { | |
809 | return (page->flags >> NODES_PGSHIFT) & NODES_MASK; | |
810 | } | |
811 | #endif | |
812 | ||
813 | #ifdef CONFIG_NUMA_BALANCING | |
814 | static inline int cpu_pid_to_cpupid(int cpu, int pid) | |
815 | { | |
816 | return ((cpu & LAST__CPU_MASK) << LAST__PID_SHIFT) | (pid & LAST__PID_MASK); | |
817 | } | |
818 | ||
819 | static inline int cpupid_to_pid(int cpupid) | |
820 | { | |
821 | return cpupid & LAST__PID_MASK; | |
822 | } | |
823 | ||
824 | static inline int cpupid_to_cpu(int cpupid) | |
825 | { | |
826 | return (cpupid >> LAST__PID_SHIFT) & LAST__CPU_MASK; | |
827 | } | |
828 | ||
829 | static inline int cpupid_to_nid(int cpupid) | |
830 | { | |
831 | return cpu_to_node(cpupid_to_cpu(cpupid)); | |
832 | } | |
833 | ||
834 | static inline bool cpupid_pid_unset(int cpupid) | |
835 | { | |
836 | return cpupid_to_pid(cpupid) == (-1 & LAST__PID_MASK); | |
837 | } | |
838 | ||
839 | static inline bool cpupid_cpu_unset(int cpupid) | |
840 | { | |
841 | return cpupid_to_cpu(cpupid) == (-1 & LAST__CPU_MASK); | |
842 | } | |
843 | ||
844 | static inline bool __cpupid_match_pid(pid_t task_pid, int cpupid) | |
845 | { | |
846 | return (task_pid & LAST__PID_MASK) == cpupid_to_pid(cpupid); | |
847 | } | |
848 | ||
849 | #define cpupid_match_pid(task, cpupid) __cpupid_match_pid(task->pid, cpupid) | |
850 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS | |
851 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
852 | { | |
853 | return xchg(&page->_last_cpupid, cpupid & LAST_CPUPID_MASK); | |
854 | } | |
855 | ||
856 | static inline int page_cpupid_last(struct page *page) | |
857 | { | |
858 | return page->_last_cpupid; | |
859 | } | |
860 | static inline void page_cpupid_reset_last(struct page *page) | |
861 | { | |
862 | page->_last_cpupid = -1 & LAST_CPUPID_MASK; | |
863 | } | |
864 | #else | |
865 | static inline int page_cpupid_last(struct page *page) | |
866 | { | |
867 | return (page->flags >> LAST_CPUPID_PGSHIFT) & LAST_CPUPID_MASK; | |
868 | } | |
869 | ||
870 | extern int page_cpupid_xchg_last(struct page *page, int cpupid); | |
871 | ||
872 | static inline void page_cpupid_reset_last(struct page *page) | |
873 | { | |
874 | page->flags |= LAST_CPUPID_MASK << LAST_CPUPID_PGSHIFT; | |
875 | } | |
876 | #endif /* LAST_CPUPID_NOT_IN_PAGE_FLAGS */ | |
877 | #else /* !CONFIG_NUMA_BALANCING */ | |
878 | static inline int page_cpupid_xchg_last(struct page *page, int cpupid) | |
879 | { | |
880 | return page_to_nid(page); /* XXX */ | |
881 | } | |
882 | ||
883 | static inline int page_cpupid_last(struct page *page) | |
884 | { | |
885 | return page_to_nid(page); /* XXX */ | |
886 | } | |
887 | ||
888 | static inline int cpupid_to_nid(int cpupid) | |
889 | { | |
890 | return -1; | |
891 | } | |
892 | ||
893 | static inline int cpupid_to_pid(int cpupid) | |
894 | { | |
895 | return -1; | |
896 | } | |
897 | ||
898 | static inline int cpupid_to_cpu(int cpupid) | |
899 | { | |
900 | return -1; | |
901 | } | |
902 | ||
903 | static inline int cpu_pid_to_cpupid(int nid, int pid) | |
904 | { | |
905 | return -1; | |
906 | } | |
907 | ||
908 | static inline bool cpupid_pid_unset(int cpupid) | |
909 | { | |
910 | return 1; | |
911 | } | |
912 | ||
913 | static inline void page_cpupid_reset_last(struct page *page) | |
914 | { | |
915 | } | |
916 | ||
917 | static inline bool cpupid_match_pid(struct task_struct *task, int cpupid) | |
918 | { | |
919 | return false; | |
920 | } | |
921 | #endif /* CONFIG_NUMA_BALANCING */ | |
922 | ||
923 | static inline struct zone *page_zone(const struct page *page) | |
924 | { | |
925 | return &NODE_DATA(page_to_nid(page))->node_zones[page_zonenum(page)]; | |
926 | } | |
927 | ||
928 | static inline pg_data_t *page_pgdat(const struct page *page) | |
929 | { | |
930 | return NODE_DATA(page_to_nid(page)); | |
931 | } | |
932 | ||
933 | #ifdef SECTION_IN_PAGE_FLAGS | |
934 | static inline void set_page_section(struct page *page, unsigned long section) | |
935 | { | |
936 | page->flags &= ~(SECTIONS_MASK << SECTIONS_PGSHIFT); | |
937 | page->flags |= (section & SECTIONS_MASK) << SECTIONS_PGSHIFT; | |
938 | } | |
939 | ||
940 | static inline unsigned long page_to_section(const struct page *page) | |
941 | { | |
942 | return (page->flags >> SECTIONS_PGSHIFT) & SECTIONS_MASK; | |
943 | } | |
944 | #endif | |
945 | ||
946 | static inline void set_page_zone(struct page *page, enum zone_type zone) | |
947 | { | |
948 | page->flags &= ~(ZONES_MASK << ZONES_PGSHIFT); | |
949 | page->flags |= (zone & ZONES_MASK) << ZONES_PGSHIFT; | |
950 | } | |
951 | ||
952 | static inline void set_page_node(struct page *page, unsigned long node) | |
953 | { | |
954 | page->flags &= ~(NODES_MASK << NODES_PGSHIFT); | |
955 | page->flags |= (node & NODES_MASK) << NODES_PGSHIFT; | |
956 | } | |
957 | ||
958 | static inline void set_page_links(struct page *page, enum zone_type zone, | |
959 | unsigned long node, unsigned long pfn) | |
960 | { | |
961 | set_page_zone(page, zone); | |
962 | set_page_node(page, node); | |
963 | #ifdef SECTION_IN_PAGE_FLAGS | |
964 | set_page_section(page, pfn_to_section_nr(pfn)); | |
965 | #endif | |
966 | } | |
967 | ||
968 | #ifdef CONFIG_MEMCG | |
969 | static inline struct mem_cgroup *page_memcg(struct page *page) | |
970 | { | |
971 | return page->mem_cgroup; | |
972 | } | |
973 | static inline struct mem_cgroup *page_memcg_rcu(struct page *page) | |
974 | { | |
975 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
976 | return READ_ONCE(page->mem_cgroup); | |
977 | } | |
978 | #else | |
979 | static inline struct mem_cgroup *page_memcg(struct page *page) | |
980 | { | |
981 | return NULL; | |
982 | } | |
983 | static inline struct mem_cgroup *page_memcg_rcu(struct page *page) | |
984 | { | |
985 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
986 | return NULL; | |
987 | } | |
988 | #endif | |
989 | ||
990 | /* | |
991 | * Some inline functions in vmstat.h depend on page_zone() | |
992 | */ | |
993 | #include <linux/vmstat.h> | |
994 | ||
995 | static __always_inline void *lowmem_page_address(const struct page *page) | |
996 | { | |
997 | return page_to_virt(page); | |
998 | } | |
999 | ||
1000 | #if defined(CONFIG_HIGHMEM) && !defined(WANT_PAGE_VIRTUAL) | |
1001 | #define HASHED_PAGE_VIRTUAL | |
1002 | #endif | |
1003 | ||
1004 | #if defined(WANT_PAGE_VIRTUAL) | |
1005 | static inline void *page_address(const struct page *page) | |
1006 | { | |
1007 | return page->virtual; | |
1008 | } | |
1009 | static inline void set_page_address(struct page *page, void *address) | |
1010 | { | |
1011 | page->virtual = address; | |
1012 | } | |
1013 | #define page_address_init() do { } while(0) | |
1014 | #endif | |
1015 | ||
1016 | #if defined(HASHED_PAGE_VIRTUAL) | |
1017 | void *page_address(const struct page *page); | |
1018 | void set_page_address(struct page *page, void *virtual); | |
1019 | void page_address_init(void); | |
1020 | #endif | |
1021 | ||
1022 | #if !defined(HASHED_PAGE_VIRTUAL) && !defined(WANT_PAGE_VIRTUAL) | |
1023 | #define page_address(page) lowmem_page_address(page) | |
1024 | #define set_page_address(page, address) do { } while(0) | |
1025 | #define page_address_init() do { } while(0) | |
1026 | #endif | |
1027 | ||
1028 | extern void *page_rmapping(struct page *page); | |
1029 | extern struct anon_vma *page_anon_vma(struct page *page); | |
1030 | extern struct address_space *page_mapping(struct page *page); | |
1031 | ||
1032 | extern struct address_space *__page_file_mapping(struct page *); | |
1033 | ||
1034 | static inline | |
1035 | struct address_space *page_file_mapping(struct page *page) | |
1036 | { | |
1037 | if (unlikely(PageSwapCache(page))) | |
1038 | return __page_file_mapping(page); | |
1039 | ||
1040 | return page->mapping; | |
1041 | } | |
1042 | ||
1043 | extern pgoff_t __page_file_index(struct page *page); | |
1044 | ||
1045 | /* | |
1046 | * Return the pagecache index of the passed page. Regular pagecache pages | |
1047 | * use ->index whereas swapcache pages use swp_offset(->private) | |
1048 | */ | |
1049 | static inline pgoff_t page_index(struct page *page) | |
1050 | { | |
1051 | if (unlikely(PageSwapCache(page))) | |
1052 | return __page_file_index(page); | |
1053 | return page->index; | |
1054 | } | |
1055 | ||
1056 | bool page_mapped(struct page *page); | |
1057 | struct address_space *page_mapping(struct page *page); | |
1058 | ||
1059 | /* | |
1060 | * Return true only if the page has been allocated with | |
1061 | * ALLOC_NO_WATERMARKS and the low watermark was not | |
1062 | * met implying that the system is under some pressure. | |
1063 | */ | |
1064 | static inline bool page_is_pfmemalloc(struct page *page) | |
1065 | { | |
1066 | /* | |
1067 | * Page index cannot be this large so this must be | |
1068 | * a pfmemalloc page. | |
1069 | */ | |
1070 | return page->index == -1UL; | |
1071 | } | |
1072 | ||
1073 | /* | |
1074 | * Only to be called by the page allocator on a freshly allocated | |
1075 | * page. | |
1076 | */ | |
1077 | static inline void set_page_pfmemalloc(struct page *page) | |
1078 | { | |
1079 | page->index = -1UL; | |
1080 | } | |
1081 | ||
1082 | static inline void clear_page_pfmemalloc(struct page *page) | |
1083 | { | |
1084 | page->index = 0; | |
1085 | } | |
1086 | ||
1087 | /* | |
1088 | * Different kinds of faults, as returned by handle_mm_fault(). | |
1089 | * Used to decide whether a process gets delivered SIGBUS or | |
1090 | * just gets major/minor fault counters bumped up. | |
1091 | */ | |
1092 | ||
1093 | #define VM_FAULT_OOM 0x0001 | |
1094 | #define VM_FAULT_SIGBUS 0x0002 | |
1095 | #define VM_FAULT_MAJOR 0x0004 | |
1096 | #define VM_FAULT_WRITE 0x0008 /* Special case for get_user_pages */ | |
1097 | #define VM_FAULT_HWPOISON 0x0010 /* Hit poisoned small page */ | |
1098 | #define VM_FAULT_HWPOISON_LARGE 0x0020 /* Hit poisoned large page. Index encoded in upper bits */ | |
1099 | #define VM_FAULT_SIGSEGV 0x0040 | |
1100 | ||
1101 | #define VM_FAULT_NOPAGE 0x0100 /* ->fault installed the pte, not return page */ | |
1102 | #define VM_FAULT_LOCKED 0x0200 /* ->fault locked the returned page */ | |
1103 | #define VM_FAULT_RETRY 0x0400 /* ->fault blocked, must retry */ | |
1104 | #define VM_FAULT_FALLBACK 0x0800 /* huge page fault failed, fall back to small */ | |
1105 | #define VM_FAULT_DAX_LOCKED 0x1000 /* ->fault has locked DAX entry */ | |
1106 | ||
1107 | #define VM_FAULT_HWPOISON_LARGE_MASK 0xf000 /* encodes hpage index for large hwpoison */ | |
1108 | ||
1109 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | VM_FAULT_SIGSEGV | \ | |
1110 | VM_FAULT_HWPOISON | VM_FAULT_HWPOISON_LARGE | \ | |
1111 | VM_FAULT_FALLBACK) | |
1112 | ||
1113 | /* Encode hstate index for a hwpoisoned large page */ | |
1114 | #define VM_FAULT_SET_HINDEX(x) ((x) << 12) | |
1115 | #define VM_FAULT_GET_HINDEX(x) (((x) >> 12) & 0xf) | |
1116 | ||
1117 | /* | |
1118 | * Can be called by the pagefault handler when it gets a VM_FAULT_OOM. | |
1119 | */ | |
1120 | extern void pagefault_out_of_memory(void); | |
1121 | ||
1122 | #define offset_in_page(p) ((unsigned long)(p) & ~PAGE_MASK) | |
1123 | ||
1124 | /* | |
1125 | * Flags passed to show_mem() and show_free_areas() to suppress output in | |
1126 | * various contexts. | |
1127 | */ | |
1128 | #define SHOW_MEM_FILTER_NODES (0x0001u) /* disallowed nodes */ | |
1129 | ||
1130 | extern void show_free_areas(unsigned int flags); | |
1131 | extern bool skip_free_areas_node(unsigned int flags, int nid); | |
1132 | ||
1133 | int shmem_zero_setup(struct vm_area_struct *); | |
1134 | #ifdef CONFIG_SHMEM | |
1135 | bool shmem_mapping(struct address_space *mapping); | |
1136 | #else | |
1137 | static inline bool shmem_mapping(struct address_space *mapping) | |
1138 | { | |
1139 | return false; | |
1140 | } | |
1141 | #endif | |
1142 | ||
1143 | extern bool can_do_mlock(void); | |
1144 | extern int user_shm_lock(size_t, struct user_struct *); | |
1145 | extern void user_shm_unlock(size_t, struct user_struct *); | |
1146 | ||
1147 | /* | |
1148 | * Parameter block passed down to zap_pte_range in exceptional cases. | |
1149 | */ | |
1150 | struct zap_details { | |
1151 | struct address_space *check_mapping; /* Check page->mapping if set */ | |
1152 | pgoff_t first_index; /* Lowest page->index to unmap */ | |
1153 | pgoff_t last_index; /* Highest page->index to unmap */ | |
1154 | bool ignore_dirty; /* Ignore dirty pages */ | |
1155 | bool check_swap_entries; /* Check also swap entries */ | |
1156 | }; | |
1157 | ||
1158 | struct page *vm_normal_page(struct vm_area_struct *vma, unsigned long addr, | |
1159 | pte_t pte); | |
1160 | struct page *vm_normal_page_pmd(struct vm_area_struct *vma, unsigned long addr, | |
1161 | pmd_t pmd); | |
1162 | ||
1163 | int zap_vma_ptes(struct vm_area_struct *vma, unsigned long address, | |
1164 | unsigned long size); | |
1165 | void zap_page_range(struct vm_area_struct *vma, unsigned long address, | |
1166 | unsigned long size, struct zap_details *); | |
1167 | void unmap_vmas(struct mmu_gather *tlb, struct vm_area_struct *start_vma, | |
1168 | unsigned long start, unsigned long end); | |
1169 | ||
1170 | /** | |
1171 | * mm_walk - callbacks for walk_page_range | |
1172 | * @pmd_entry: if set, called for each non-empty PMD (3rd-level) entry | |
1173 | * this handler is required to be able to handle | |
1174 | * pmd_trans_huge() pmds. They may simply choose to | |
1175 | * split_huge_page() instead of handling it explicitly. | |
1176 | * @pte_entry: if set, called for each non-empty PTE (4th-level) entry | |
1177 | * @pte_hole: if set, called for each hole at all levels | |
1178 | * @hugetlb_entry: if set, called for each hugetlb entry | |
1179 | * @test_walk: caller specific callback function to determine whether | |
1180 | * we walk over the current vma or not. Returning 0 | |
1181 | * value means "do page table walk over the current vma," | |
1182 | * and a negative one means "abort current page table walk | |
1183 | * right now." 1 means "skip the current vma." | |
1184 | * @mm: mm_struct representing the target process of page table walk | |
1185 | * @vma: vma currently walked (NULL if walking outside vmas) | |
1186 | * @private: private data for callbacks' usage | |
1187 | * | |
1188 | * (see the comment on walk_page_range() for more details) | |
1189 | */ | |
1190 | struct mm_walk { | |
1191 | int (*pmd_entry)(pmd_t *pmd, unsigned long addr, | |
1192 | unsigned long next, struct mm_walk *walk); | |
1193 | int (*pte_entry)(pte_t *pte, unsigned long addr, | |
1194 | unsigned long next, struct mm_walk *walk); | |
1195 | int (*pte_hole)(unsigned long addr, unsigned long next, | |
1196 | struct mm_walk *walk); | |
1197 | int (*hugetlb_entry)(pte_t *pte, unsigned long hmask, | |
1198 | unsigned long addr, unsigned long next, | |
1199 | struct mm_walk *walk); | |
1200 | int (*test_walk)(unsigned long addr, unsigned long next, | |
1201 | struct mm_walk *walk); | |
1202 | struct mm_struct *mm; | |
1203 | struct vm_area_struct *vma; | |
1204 | void *private; | |
1205 | }; | |
1206 | ||
1207 | int walk_page_range(unsigned long addr, unsigned long end, | |
1208 | struct mm_walk *walk); | |
1209 | int walk_page_vma(struct vm_area_struct *vma, struct mm_walk *walk); | |
1210 | void free_pgd_range(struct mmu_gather *tlb, unsigned long addr, | |
1211 | unsigned long end, unsigned long floor, unsigned long ceiling); | |
1212 | int copy_page_range(struct mm_struct *dst, struct mm_struct *src, | |
1213 | struct vm_area_struct *vma); | |
1214 | void unmap_mapping_range(struct address_space *mapping, | |
1215 | loff_t const holebegin, loff_t const holelen, int even_cows); | |
1216 | int follow_pfn(struct vm_area_struct *vma, unsigned long address, | |
1217 | unsigned long *pfn); | |
1218 | int follow_phys(struct vm_area_struct *vma, unsigned long address, | |
1219 | unsigned int flags, unsigned long *prot, resource_size_t *phys); | |
1220 | int generic_access_phys(struct vm_area_struct *vma, unsigned long addr, | |
1221 | void *buf, int len, int write); | |
1222 | ||
1223 | static inline void unmap_shared_mapping_range(struct address_space *mapping, | |
1224 | loff_t const holebegin, loff_t const holelen) | |
1225 | { | |
1226 | unmap_mapping_range(mapping, holebegin, holelen, 0); | |
1227 | } | |
1228 | ||
1229 | extern void truncate_pagecache(struct inode *inode, loff_t new); | |
1230 | extern void truncate_setsize(struct inode *inode, loff_t newsize); | |
1231 | void pagecache_isize_extended(struct inode *inode, loff_t from, loff_t to); | |
1232 | void truncate_pagecache_range(struct inode *inode, loff_t offset, loff_t end); | |
1233 | int truncate_inode_page(struct address_space *mapping, struct page *page); | |
1234 | int generic_error_remove_page(struct address_space *mapping, struct page *page); | |
1235 | int invalidate_inode_page(struct page *page); | |
1236 | ||
1237 | #ifdef CONFIG_MMU | |
1238 | extern int handle_mm_fault(struct vm_area_struct *vma, unsigned long address, | |
1239 | unsigned int flags); | |
1240 | extern int fixup_user_fault(struct task_struct *tsk, struct mm_struct *mm, | |
1241 | unsigned long address, unsigned int fault_flags, | |
1242 | bool *unlocked); | |
1243 | #else | |
1244 | static inline int handle_mm_fault(struct vm_area_struct *vma, | |
1245 | unsigned long address, unsigned int flags) | |
1246 | { | |
1247 | /* should never happen if there's no MMU */ | |
1248 | BUG(); | |
1249 | return VM_FAULT_SIGBUS; | |
1250 | } | |
1251 | static inline int fixup_user_fault(struct task_struct *tsk, | |
1252 | struct mm_struct *mm, unsigned long address, | |
1253 | unsigned int fault_flags, bool *unlocked) | |
1254 | { | |
1255 | /* should never happen if there's no MMU */ | |
1256 | BUG(); | |
1257 | return -EFAULT; | |
1258 | } | |
1259 | #endif | |
1260 | ||
1261 | extern int access_process_vm(struct task_struct *tsk, unsigned long addr, void *buf, int len, | |
1262 | unsigned int gup_flags); | |
1263 | extern int access_remote_vm(struct mm_struct *mm, unsigned long addr, | |
1264 | void *buf, int len, unsigned int gup_flags); | |
1265 | ||
1266 | long get_user_pages_remote(struct task_struct *tsk, struct mm_struct *mm, | |
1267 | unsigned long start, unsigned long nr_pages, | |
1268 | unsigned int gup_flags, struct page **pages, | |
1269 | struct vm_area_struct **vmas, int *locked); | |
1270 | long get_user_pages(unsigned long start, unsigned long nr_pages, | |
1271 | unsigned int gup_flags, struct page **pages, | |
1272 | struct vm_area_struct **vmas); | |
1273 | long get_user_pages_locked(unsigned long start, unsigned long nr_pages, | |
1274 | unsigned int gup_flags, struct page **pages, int *locked); | |
1275 | long get_user_pages_unlocked(unsigned long start, unsigned long nr_pages, | |
1276 | struct page **pages, unsigned int gup_flags); | |
1277 | int get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
1278 | struct page **pages); | |
1279 | ||
1280 | /* Container for pinned pfns / pages */ | |
1281 | struct frame_vector { | |
1282 | unsigned int nr_allocated; /* Number of frames we have space for */ | |
1283 | unsigned int nr_frames; /* Number of frames stored in ptrs array */ | |
1284 | bool got_ref; /* Did we pin pages by getting page ref? */ | |
1285 | bool is_pfns; /* Does array contain pages or pfns? */ | |
1286 | void *ptrs[0]; /* Array of pinned pfns / pages. Use | |
1287 | * pfns_vector_pages() or pfns_vector_pfns() | |
1288 | * for access */ | |
1289 | }; | |
1290 | ||
1291 | struct frame_vector *frame_vector_create(unsigned int nr_frames); | |
1292 | void frame_vector_destroy(struct frame_vector *vec); | |
1293 | int get_vaddr_frames(unsigned long start, unsigned int nr_pfns, | |
1294 | unsigned int gup_flags, struct frame_vector *vec); | |
1295 | void put_vaddr_frames(struct frame_vector *vec); | |
1296 | int frame_vector_to_pages(struct frame_vector *vec); | |
1297 | void frame_vector_to_pfns(struct frame_vector *vec); | |
1298 | ||
1299 | static inline unsigned int frame_vector_count(struct frame_vector *vec) | |
1300 | { | |
1301 | return vec->nr_frames; | |
1302 | } | |
1303 | ||
1304 | static inline struct page **frame_vector_pages(struct frame_vector *vec) | |
1305 | { | |
1306 | if (vec->is_pfns) { | |
1307 | int err = frame_vector_to_pages(vec); | |
1308 | ||
1309 | if (err) | |
1310 | return ERR_PTR(err); | |
1311 | } | |
1312 | return (struct page **)(vec->ptrs); | |
1313 | } | |
1314 | ||
1315 | static inline unsigned long *frame_vector_pfns(struct frame_vector *vec) | |
1316 | { | |
1317 | if (!vec->is_pfns) | |
1318 | frame_vector_to_pfns(vec); | |
1319 | return (unsigned long *)(vec->ptrs); | |
1320 | } | |
1321 | ||
1322 | struct kvec; | |
1323 | int get_kernel_pages(const struct kvec *iov, int nr_pages, int write, | |
1324 | struct page **pages); | |
1325 | int get_kernel_page(unsigned long start, int write, struct page **pages); | |
1326 | struct page *get_dump_page(unsigned long addr); | |
1327 | ||
1328 | extern int try_to_release_page(struct page * page, gfp_t gfp_mask); | |
1329 | extern void do_invalidatepage(struct page *page, unsigned int offset, | |
1330 | unsigned int length); | |
1331 | ||
1332 | int __set_page_dirty_nobuffers(struct page *page); | |
1333 | int __set_page_dirty_no_writeback(struct page *page); | |
1334 | int redirty_page_for_writepage(struct writeback_control *wbc, | |
1335 | struct page *page); | |
1336 | void account_page_dirtied(struct page *page, struct address_space *mapping); | |
1337 | void account_page_cleaned(struct page *page, struct address_space *mapping, | |
1338 | struct bdi_writeback *wb); | |
1339 | int set_page_dirty(struct page *page); | |
1340 | int set_page_dirty_lock(struct page *page); | |
1341 | void cancel_dirty_page(struct page *page); | |
1342 | int clear_page_dirty_for_io(struct page *page); | |
1343 | ||
1344 | int get_cmdline(struct task_struct *task, char *buffer, int buflen); | |
1345 | ||
1346 | /* Is the vma a continuation of the stack vma above it? */ | |
1347 | static inline int vma_growsdown(struct vm_area_struct *vma, unsigned long addr) | |
1348 | { | |
1349 | return vma && (vma->vm_end == addr) && (vma->vm_flags & VM_GROWSDOWN); | |
1350 | } | |
1351 | ||
1352 | static inline bool vma_is_anonymous(struct vm_area_struct *vma) | |
1353 | { | |
1354 | return !vma->vm_ops; | |
1355 | } | |
1356 | ||
1357 | static inline int stack_guard_page_start(struct vm_area_struct *vma, | |
1358 | unsigned long addr) | |
1359 | { | |
1360 | return (vma->vm_flags & VM_GROWSDOWN) && | |
1361 | (vma->vm_start == addr) && | |
1362 | !vma_growsdown(vma->vm_prev, addr); | |
1363 | } | |
1364 | ||
1365 | /* Is the vma a continuation of the stack vma below it? */ | |
1366 | static inline int vma_growsup(struct vm_area_struct *vma, unsigned long addr) | |
1367 | { | |
1368 | return vma && (vma->vm_start == addr) && (vma->vm_flags & VM_GROWSUP); | |
1369 | } | |
1370 | ||
1371 | static inline int stack_guard_page_end(struct vm_area_struct *vma, | |
1372 | unsigned long addr) | |
1373 | { | |
1374 | return (vma->vm_flags & VM_GROWSUP) && | |
1375 | (vma->vm_end == addr) && | |
1376 | !vma_growsup(vma->vm_next, addr); | |
1377 | } | |
1378 | ||
1379 | int vma_is_stack_for_current(struct vm_area_struct *vma); | |
1380 | ||
1381 | extern unsigned long move_page_tables(struct vm_area_struct *vma, | |
1382 | unsigned long old_addr, struct vm_area_struct *new_vma, | |
1383 | unsigned long new_addr, unsigned long len, | |
1384 | bool need_rmap_locks); | |
1385 | extern unsigned long change_protection(struct vm_area_struct *vma, unsigned long start, | |
1386 | unsigned long end, pgprot_t newprot, | |
1387 | int dirty_accountable, int prot_numa); | |
1388 | extern int mprotect_fixup(struct vm_area_struct *vma, | |
1389 | struct vm_area_struct **pprev, unsigned long start, | |
1390 | unsigned long end, unsigned long newflags); | |
1391 | ||
1392 | /* | |
1393 | * doesn't attempt to fault and will return short. | |
1394 | */ | |
1395 | int __get_user_pages_fast(unsigned long start, int nr_pages, int write, | |
1396 | struct page **pages); | |
1397 | /* | |
1398 | * per-process(per-mm_struct) statistics. | |
1399 | */ | |
1400 | static inline unsigned long get_mm_counter(struct mm_struct *mm, int member) | |
1401 | { | |
1402 | long val = atomic_long_read(&mm->rss_stat.count[member]); | |
1403 | ||
1404 | #ifdef SPLIT_RSS_COUNTING | |
1405 | /* | |
1406 | * counter is updated in asynchronous manner and may go to minus. | |
1407 | * But it's never be expected number for users. | |
1408 | */ | |
1409 | if (val < 0) | |
1410 | val = 0; | |
1411 | #endif | |
1412 | return (unsigned long)val; | |
1413 | } | |
1414 | ||
1415 | static inline void add_mm_counter(struct mm_struct *mm, int member, long value) | |
1416 | { | |
1417 | atomic_long_add(value, &mm->rss_stat.count[member]); | |
1418 | } | |
1419 | ||
1420 | static inline void inc_mm_counter(struct mm_struct *mm, int member) | |
1421 | { | |
1422 | atomic_long_inc(&mm->rss_stat.count[member]); | |
1423 | } | |
1424 | ||
1425 | static inline void dec_mm_counter(struct mm_struct *mm, int member) | |
1426 | { | |
1427 | atomic_long_dec(&mm->rss_stat.count[member]); | |
1428 | } | |
1429 | ||
1430 | /* Optimized variant when page is already known not to be PageAnon */ | |
1431 | static inline int mm_counter_file(struct page *page) | |
1432 | { | |
1433 | if (PageSwapBacked(page)) | |
1434 | return MM_SHMEMPAGES; | |
1435 | return MM_FILEPAGES; | |
1436 | } | |
1437 | ||
1438 | static inline int mm_counter(struct page *page) | |
1439 | { | |
1440 | if (PageAnon(page)) | |
1441 | return MM_ANONPAGES; | |
1442 | return mm_counter_file(page); | |
1443 | } | |
1444 | ||
1445 | static inline unsigned long get_mm_rss(struct mm_struct *mm) | |
1446 | { | |
1447 | return get_mm_counter(mm, MM_FILEPAGES) + | |
1448 | get_mm_counter(mm, MM_ANONPAGES) + | |
1449 | get_mm_counter(mm, MM_SHMEMPAGES); | |
1450 | } | |
1451 | ||
1452 | static inline unsigned long get_mm_hiwater_rss(struct mm_struct *mm) | |
1453 | { | |
1454 | return max(mm->hiwater_rss, get_mm_rss(mm)); | |
1455 | } | |
1456 | ||
1457 | static inline unsigned long get_mm_hiwater_vm(struct mm_struct *mm) | |
1458 | { | |
1459 | return max(mm->hiwater_vm, mm->total_vm); | |
1460 | } | |
1461 | ||
1462 | static inline void update_hiwater_rss(struct mm_struct *mm) | |
1463 | { | |
1464 | unsigned long _rss = get_mm_rss(mm); | |
1465 | ||
1466 | if ((mm)->hiwater_rss < _rss) | |
1467 | (mm)->hiwater_rss = _rss; | |
1468 | } | |
1469 | ||
1470 | static inline void update_hiwater_vm(struct mm_struct *mm) | |
1471 | { | |
1472 | if (mm->hiwater_vm < mm->total_vm) | |
1473 | mm->hiwater_vm = mm->total_vm; | |
1474 | } | |
1475 | ||
1476 | static inline void reset_mm_hiwater_rss(struct mm_struct *mm) | |
1477 | { | |
1478 | mm->hiwater_rss = get_mm_rss(mm); | |
1479 | } | |
1480 | ||
1481 | static inline void setmax_mm_hiwater_rss(unsigned long *maxrss, | |
1482 | struct mm_struct *mm) | |
1483 | { | |
1484 | unsigned long hiwater_rss = get_mm_hiwater_rss(mm); | |
1485 | ||
1486 | if (*maxrss < hiwater_rss) | |
1487 | *maxrss = hiwater_rss; | |
1488 | } | |
1489 | ||
1490 | #if defined(SPLIT_RSS_COUNTING) | |
1491 | void sync_mm_rss(struct mm_struct *mm); | |
1492 | #else | |
1493 | static inline void sync_mm_rss(struct mm_struct *mm) | |
1494 | { | |
1495 | } | |
1496 | #endif | |
1497 | ||
1498 | #ifndef __HAVE_ARCH_PTE_DEVMAP | |
1499 | static inline int pte_devmap(pte_t pte) | |
1500 | { | |
1501 | return 0; | |
1502 | } | |
1503 | #endif | |
1504 | ||
1505 | int vma_wants_writenotify(struct vm_area_struct *vma, pgprot_t vm_page_prot); | |
1506 | ||
1507 | extern pte_t *__get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
1508 | spinlock_t **ptl); | |
1509 | static inline pte_t *get_locked_pte(struct mm_struct *mm, unsigned long addr, | |
1510 | spinlock_t **ptl) | |
1511 | { | |
1512 | pte_t *ptep; | |
1513 | __cond_lock(*ptl, ptep = __get_locked_pte(mm, addr, ptl)); | |
1514 | return ptep; | |
1515 | } | |
1516 | ||
1517 | #ifdef __PAGETABLE_PUD_FOLDED | |
1518 | static inline int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, | |
1519 | unsigned long address) | |
1520 | { | |
1521 | return 0; | |
1522 | } | |
1523 | #else | |
1524 | int __pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address); | |
1525 | #endif | |
1526 | ||
1527 | #if defined(__PAGETABLE_PMD_FOLDED) || !defined(CONFIG_MMU) | |
1528 | static inline int __pmd_alloc(struct mm_struct *mm, pud_t *pud, | |
1529 | unsigned long address) | |
1530 | { | |
1531 | return 0; | |
1532 | } | |
1533 | ||
1534 | static inline void mm_nr_pmds_init(struct mm_struct *mm) {} | |
1535 | ||
1536 | static inline unsigned long mm_nr_pmds(struct mm_struct *mm) | |
1537 | { | |
1538 | return 0; | |
1539 | } | |
1540 | ||
1541 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) {} | |
1542 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) {} | |
1543 | ||
1544 | #else | |
1545 | int __pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address); | |
1546 | ||
1547 | static inline void mm_nr_pmds_init(struct mm_struct *mm) | |
1548 | { | |
1549 | atomic_long_set(&mm->nr_pmds, 0); | |
1550 | } | |
1551 | ||
1552 | static inline unsigned long mm_nr_pmds(struct mm_struct *mm) | |
1553 | { | |
1554 | return atomic_long_read(&mm->nr_pmds); | |
1555 | } | |
1556 | ||
1557 | static inline void mm_inc_nr_pmds(struct mm_struct *mm) | |
1558 | { | |
1559 | atomic_long_inc(&mm->nr_pmds); | |
1560 | } | |
1561 | ||
1562 | static inline void mm_dec_nr_pmds(struct mm_struct *mm) | |
1563 | { | |
1564 | atomic_long_dec(&mm->nr_pmds); | |
1565 | } | |
1566 | #endif | |
1567 | ||
1568 | int __pte_alloc(struct mm_struct *mm, pmd_t *pmd, unsigned long address); | |
1569 | int __pte_alloc_kernel(pmd_t *pmd, unsigned long address); | |
1570 | ||
1571 | /* | |
1572 | * The following ifdef needed to get the 4level-fixup.h header to work. | |
1573 | * Remove it when 4level-fixup.h has been removed. | |
1574 | */ | |
1575 | #if defined(CONFIG_MMU) && !defined(__ARCH_HAS_4LEVEL_HACK) | |
1576 | static inline pud_t *pud_alloc(struct mm_struct *mm, pgd_t *pgd, unsigned long address) | |
1577 | { | |
1578 | return (unlikely(pgd_none(*pgd)) && __pud_alloc(mm, pgd, address))? | |
1579 | NULL: pud_offset(pgd, address); | |
1580 | } | |
1581 | ||
1582 | static inline pmd_t *pmd_alloc(struct mm_struct *mm, pud_t *pud, unsigned long address) | |
1583 | { | |
1584 | return (unlikely(pud_none(*pud)) && __pmd_alloc(mm, pud, address))? | |
1585 | NULL: pmd_offset(pud, address); | |
1586 | } | |
1587 | #endif /* CONFIG_MMU && !__ARCH_HAS_4LEVEL_HACK */ | |
1588 | ||
1589 | #if USE_SPLIT_PTE_PTLOCKS | |
1590 | #if ALLOC_SPLIT_PTLOCKS | |
1591 | void __init ptlock_cache_init(void); | |
1592 | extern bool ptlock_alloc(struct page *page); | |
1593 | extern void ptlock_free(struct page *page); | |
1594 | ||
1595 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
1596 | { | |
1597 | return page->ptl; | |
1598 | } | |
1599 | #else /* ALLOC_SPLIT_PTLOCKS */ | |
1600 | static inline void ptlock_cache_init(void) | |
1601 | { | |
1602 | } | |
1603 | ||
1604 | static inline bool ptlock_alloc(struct page *page) | |
1605 | { | |
1606 | return true; | |
1607 | } | |
1608 | ||
1609 | static inline void ptlock_free(struct page *page) | |
1610 | { | |
1611 | } | |
1612 | ||
1613 | static inline spinlock_t *ptlock_ptr(struct page *page) | |
1614 | { | |
1615 | return &page->ptl; | |
1616 | } | |
1617 | #endif /* ALLOC_SPLIT_PTLOCKS */ | |
1618 | ||
1619 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1620 | { | |
1621 | return ptlock_ptr(pmd_page(*pmd)); | |
1622 | } | |
1623 | ||
1624 | static inline bool ptlock_init(struct page *page) | |
1625 | { | |
1626 | /* | |
1627 | * prep_new_page() initialize page->private (and therefore page->ptl) | |
1628 | * with 0. Make sure nobody took it in use in between. | |
1629 | * | |
1630 | * It can happen if arch try to use slab for page table allocation: | |
1631 | * slab code uses page->slab_cache, which share storage with page->ptl. | |
1632 | */ | |
1633 | VM_BUG_ON_PAGE(*(unsigned long *)&page->ptl, page); | |
1634 | if (!ptlock_alloc(page)) | |
1635 | return false; | |
1636 | spin_lock_init(ptlock_ptr(page)); | |
1637 | return true; | |
1638 | } | |
1639 | ||
1640 | /* Reset page->mapping so free_pages_check won't complain. */ | |
1641 | static inline void pte_lock_deinit(struct page *page) | |
1642 | { | |
1643 | page->mapping = NULL; | |
1644 | ptlock_free(page); | |
1645 | } | |
1646 | ||
1647 | #else /* !USE_SPLIT_PTE_PTLOCKS */ | |
1648 | /* | |
1649 | * We use mm->page_table_lock to guard all pagetable pages of the mm. | |
1650 | */ | |
1651 | static inline spinlock_t *pte_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1652 | { | |
1653 | return &mm->page_table_lock; | |
1654 | } | |
1655 | static inline void ptlock_cache_init(void) {} | |
1656 | static inline bool ptlock_init(struct page *page) { return true; } | |
1657 | static inline void pte_lock_deinit(struct page *page) {} | |
1658 | #endif /* USE_SPLIT_PTE_PTLOCKS */ | |
1659 | ||
1660 | static inline void pgtable_init(void) | |
1661 | { | |
1662 | ptlock_cache_init(); | |
1663 | pgtable_cache_init(); | |
1664 | } | |
1665 | ||
1666 | static inline bool pgtable_page_ctor(struct page *page) | |
1667 | { | |
1668 | if (!ptlock_init(page)) | |
1669 | return false; | |
1670 | inc_zone_page_state(page, NR_PAGETABLE); | |
1671 | return true; | |
1672 | } | |
1673 | ||
1674 | static inline void pgtable_page_dtor(struct page *page) | |
1675 | { | |
1676 | pte_lock_deinit(page); | |
1677 | dec_zone_page_state(page, NR_PAGETABLE); | |
1678 | } | |
1679 | ||
1680 | #define pte_offset_map_lock(mm, pmd, address, ptlp) \ | |
1681 | ({ \ | |
1682 | spinlock_t *__ptl = pte_lockptr(mm, pmd); \ | |
1683 | pte_t *__pte = pte_offset_map(pmd, address); \ | |
1684 | *(ptlp) = __ptl; \ | |
1685 | spin_lock(__ptl); \ | |
1686 | __pte; \ | |
1687 | }) | |
1688 | ||
1689 | #define pte_unmap_unlock(pte, ptl) do { \ | |
1690 | spin_unlock(ptl); \ | |
1691 | pte_unmap(pte); \ | |
1692 | } while (0) | |
1693 | ||
1694 | #define pte_alloc(mm, pmd, address) \ | |
1695 | (unlikely(pmd_none(*(pmd))) && __pte_alloc(mm, pmd, address)) | |
1696 | ||
1697 | #define pte_alloc_map(mm, pmd, address) \ | |
1698 | (pte_alloc(mm, pmd, address) ? NULL : pte_offset_map(pmd, address)) | |
1699 | ||
1700 | #define pte_alloc_map_lock(mm, pmd, address, ptlp) \ | |
1701 | (pte_alloc(mm, pmd, address) ? \ | |
1702 | NULL : pte_offset_map_lock(mm, pmd, address, ptlp)) | |
1703 | ||
1704 | #define pte_alloc_kernel(pmd, address) \ | |
1705 | ((unlikely(pmd_none(*(pmd))) && __pte_alloc_kernel(pmd, address))? \ | |
1706 | NULL: pte_offset_kernel(pmd, address)) | |
1707 | ||
1708 | #if USE_SPLIT_PMD_PTLOCKS | |
1709 | ||
1710 | static struct page *pmd_to_page(pmd_t *pmd) | |
1711 | { | |
1712 | unsigned long mask = ~(PTRS_PER_PMD * sizeof(pmd_t) - 1); | |
1713 | return virt_to_page((void *)((unsigned long) pmd & mask)); | |
1714 | } | |
1715 | ||
1716 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1717 | { | |
1718 | return ptlock_ptr(pmd_to_page(pmd)); | |
1719 | } | |
1720 | ||
1721 | static inline bool pgtable_pmd_page_ctor(struct page *page) | |
1722 | { | |
1723 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
1724 | page->pmd_huge_pte = NULL; | |
1725 | #endif | |
1726 | return ptlock_init(page); | |
1727 | } | |
1728 | ||
1729 | static inline void pgtable_pmd_page_dtor(struct page *page) | |
1730 | { | |
1731 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
1732 | VM_BUG_ON_PAGE(page->pmd_huge_pte, page); | |
1733 | #endif | |
1734 | ptlock_free(page); | |
1735 | } | |
1736 | ||
1737 | #define pmd_huge_pte(mm, pmd) (pmd_to_page(pmd)->pmd_huge_pte) | |
1738 | ||
1739 | #else | |
1740 | ||
1741 | static inline spinlock_t *pmd_lockptr(struct mm_struct *mm, pmd_t *pmd) | |
1742 | { | |
1743 | return &mm->page_table_lock; | |
1744 | } | |
1745 | ||
1746 | static inline bool pgtable_pmd_page_ctor(struct page *page) { return true; } | |
1747 | static inline void pgtable_pmd_page_dtor(struct page *page) {} | |
1748 | ||
1749 | #define pmd_huge_pte(mm, pmd) ((mm)->pmd_huge_pte) | |
1750 | ||
1751 | #endif | |
1752 | ||
1753 | static inline spinlock_t *pmd_lock(struct mm_struct *mm, pmd_t *pmd) | |
1754 | { | |
1755 | spinlock_t *ptl = pmd_lockptr(mm, pmd); | |
1756 | spin_lock(ptl); | |
1757 | return ptl; | |
1758 | } | |
1759 | ||
1760 | extern void free_area_init(unsigned long * zones_size); | |
1761 | extern void free_area_init_node(int nid, unsigned long * zones_size, | |
1762 | unsigned long zone_start_pfn, unsigned long *zholes_size); | |
1763 | extern void free_initmem(void); | |
1764 | ||
1765 | /* | |
1766 | * Free reserved pages within range [PAGE_ALIGN(start), end & PAGE_MASK) | |
1767 | * into the buddy system. The freed pages will be poisoned with pattern | |
1768 | * "poison" if it's within range [0, UCHAR_MAX]. | |
1769 | * Return pages freed into the buddy system. | |
1770 | */ | |
1771 | extern unsigned long free_reserved_area(void *start, void *end, | |
1772 | int poison, char *s); | |
1773 | ||
1774 | #ifdef CONFIG_HIGHMEM | |
1775 | /* | |
1776 | * Free a highmem page into the buddy system, adjusting totalhigh_pages | |
1777 | * and totalram_pages. | |
1778 | */ | |
1779 | extern void free_highmem_page(struct page *page); | |
1780 | #endif | |
1781 | ||
1782 | extern void adjust_managed_page_count(struct page *page, long count); | |
1783 | extern void mem_init_print_info(const char *str); | |
1784 | ||
1785 | extern void reserve_bootmem_region(phys_addr_t start, phys_addr_t end); | |
1786 | ||
1787 | /* Free the reserved page into the buddy system, so it gets managed. */ | |
1788 | static inline void __free_reserved_page(struct page *page) | |
1789 | { | |
1790 | ClearPageReserved(page); | |
1791 | init_page_count(page); | |
1792 | __free_page(page); | |
1793 | } | |
1794 | ||
1795 | static inline void free_reserved_page(struct page *page) | |
1796 | { | |
1797 | __free_reserved_page(page); | |
1798 | adjust_managed_page_count(page, 1); | |
1799 | } | |
1800 | ||
1801 | static inline void mark_page_reserved(struct page *page) | |
1802 | { | |
1803 | SetPageReserved(page); | |
1804 | adjust_managed_page_count(page, -1); | |
1805 | } | |
1806 | ||
1807 | /* | |
1808 | * Default method to free all the __init memory into the buddy system. | |
1809 | * The freed pages will be poisoned with pattern "poison" if it's within | |
1810 | * range [0, UCHAR_MAX]. | |
1811 | * Return pages freed into the buddy system. | |
1812 | */ | |
1813 | static inline unsigned long free_initmem_default(int poison) | |
1814 | { | |
1815 | extern char __init_begin[], __init_end[]; | |
1816 | ||
1817 | return free_reserved_area(&__init_begin, &__init_end, | |
1818 | poison, "unused kernel"); | |
1819 | } | |
1820 | ||
1821 | static inline unsigned long get_num_physpages(void) | |
1822 | { | |
1823 | int nid; | |
1824 | unsigned long phys_pages = 0; | |
1825 | ||
1826 | for_each_online_node(nid) | |
1827 | phys_pages += node_present_pages(nid); | |
1828 | ||
1829 | return phys_pages; | |
1830 | } | |
1831 | ||
1832 | #ifdef CONFIG_HAVE_MEMBLOCK_NODE_MAP | |
1833 | /* | |
1834 | * With CONFIG_HAVE_MEMBLOCK_NODE_MAP set, an architecture may initialise its | |
1835 | * zones, allocate the backing mem_map and account for memory holes in a more | |
1836 | * architecture independent manner. This is a substitute for creating the | |
1837 | * zone_sizes[] and zholes_size[] arrays and passing them to | |
1838 | * free_area_init_node() | |
1839 | * | |
1840 | * An architecture is expected to register range of page frames backed by | |
1841 | * physical memory with memblock_add[_node]() before calling | |
1842 | * free_area_init_nodes() passing in the PFN each zone ends at. At a basic | |
1843 | * usage, an architecture is expected to do something like | |
1844 | * | |
1845 | * unsigned long max_zone_pfns[MAX_NR_ZONES] = {max_dma, max_normal_pfn, | |
1846 | * max_highmem_pfn}; | |
1847 | * for_each_valid_physical_page_range() | |
1848 | * memblock_add_node(base, size, nid) | |
1849 | * free_area_init_nodes(max_zone_pfns); | |
1850 | * | |
1851 | * free_bootmem_with_active_regions() calls free_bootmem_node() for each | |
1852 | * registered physical page range. Similarly | |
1853 | * sparse_memory_present_with_active_regions() calls memory_present() for | |
1854 | * each range when SPARSEMEM is enabled. | |
1855 | * | |
1856 | * See mm/page_alloc.c for more information on each function exposed by | |
1857 | * CONFIG_HAVE_MEMBLOCK_NODE_MAP. | |
1858 | */ | |
1859 | extern void free_area_init_nodes(unsigned long *max_zone_pfn); | |
1860 | unsigned long node_map_pfn_alignment(void); | |
1861 | unsigned long __absent_pages_in_range(int nid, unsigned long start_pfn, | |
1862 | unsigned long end_pfn); | |
1863 | extern unsigned long absent_pages_in_range(unsigned long start_pfn, | |
1864 | unsigned long end_pfn); | |
1865 | extern void get_pfn_range_for_nid(unsigned int nid, | |
1866 | unsigned long *start_pfn, unsigned long *end_pfn); | |
1867 | extern unsigned long find_min_pfn_with_active_regions(void); | |
1868 | extern void free_bootmem_with_active_regions(int nid, | |
1869 | unsigned long max_low_pfn); | |
1870 | extern void sparse_memory_present_with_active_regions(int nid); | |
1871 | ||
1872 | #endif /* CONFIG_HAVE_MEMBLOCK_NODE_MAP */ | |
1873 | ||
1874 | #if !defined(CONFIG_HAVE_MEMBLOCK_NODE_MAP) && \ | |
1875 | !defined(CONFIG_HAVE_ARCH_EARLY_PFN_TO_NID) | |
1876 | static inline int __early_pfn_to_nid(unsigned long pfn, | |
1877 | struct mminit_pfnnid_cache *state) | |
1878 | { | |
1879 | return 0; | |
1880 | } | |
1881 | #else | |
1882 | /* please see mm/page_alloc.c */ | |
1883 | extern int __meminit early_pfn_to_nid(unsigned long pfn); | |
1884 | /* there is a per-arch backend function. */ | |
1885 | extern int __meminit __early_pfn_to_nid(unsigned long pfn, | |
1886 | struct mminit_pfnnid_cache *state); | |
1887 | #endif | |
1888 | ||
1889 | extern void set_dma_reserve(unsigned long new_dma_reserve); | |
1890 | extern void memmap_init_zone(unsigned long, int, unsigned long, | |
1891 | unsigned long, enum memmap_context); | |
1892 | extern void setup_per_zone_wmarks(void); | |
1893 | extern int __meminit init_per_zone_wmark_min(void); | |
1894 | extern void mem_init(void); | |
1895 | extern void __init mmap_init(void); | |
1896 | extern void show_mem(unsigned int flags); | |
1897 | extern long si_mem_available(void); | |
1898 | extern void si_meminfo(struct sysinfo * val); | |
1899 | extern void si_meminfo_node(struct sysinfo *val, int nid); | |
1900 | #ifdef __HAVE_ARCH_RESERVED_KERNEL_PAGES | |
1901 | extern unsigned long arch_reserved_kernel_pages(void); | |
1902 | #endif | |
1903 | ||
1904 | extern __printf(2, 3) | |
1905 | void warn_alloc(gfp_t gfp_mask, const char *fmt, ...); | |
1906 | ||
1907 | extern void setup_per_cpu_pageset(void); | |
1908 | ||
1909 | extern void zone_pcp_update(struct zone *zone); | |
1910 | extern void zone_pcp_reset(struct zone *zone); | |
1911 | ||
1912 | /* page_alloc.c */ | |
1913 | extern int min_free_kbytes; | |
1914 | extern int watermark_scale_factor; | |
1915 | ||
1916 | /* nommu.c */ | |
1917 | extern atomic_long_t mmap_pages_allocated; | |
1918 | extern int nommu_shrink_inode_mappings(struct inode *, size_t, size_t); | |
1919 | ||
1920 | /* interval_tree.c */ | |
1921 | void vma_interval_tree_insert(struct vm_area_struct *node, | |
1922 | struct rb_root *root); | |
1923 | void vma_interval_tree_insert_after(struct vm_area_struct *node, | |
1924 | struct vm_area_struct *prev, | |
1925 | struct rb_root *root); | |
1926 | void vma_interval_tree_remove(struct vm_area_struct *node, | |
1927 | struct rb_root *root); | |
1928 | struct vm_area_struct *vma_interval_tree_iter_first(struct rb_root *root, | |
1929 | unsigned long start, unsigned long last); | |
1930 | struct vm_area_struct *vma_interval_tree_iter_next(struct vm_area_struct *node, | |
1931 | unsigned long start, unsigned long last); | |
1932 | ||
1933 | #define vma_interval_tree_foreach(vma, root, start, last) \ | |
1934 | for (vma = vma_interval_tree_iter_first(root, start, last); \ | |
1935 | vma; vma = vma_interval_tree_iter_next(vma, start, last)) | |
1936 | ||
1937 | void anon_vma_interval_tree_insert(struct anon_vma_chain *node, | |
1938 | struct rb_root *root); | |
1939 | void anon_vma_interval_tree_remove(struct anon_vma_chain *node, | |
1940 | struct rb_root *root); | |
1941 | struct anon_vma_chain *anon_vma_interval_tree_iter_first( | |
1942 | struct rb_root *root, unsigned long start, unsigned long last); | |
1943 | struct anon_vma_chain *anon_vma_interval_tree_iter_next( | |
1944 | struct anon_vma_chain *node, unsigned long start, unsigned long last); | |
1945 | #ifdef CONFIG_DEBUG_VM_RB | |
1946 | void anon_vma_interval_tree_verify(struct anon_vma_chain *node); | |
1947 | #endif | |
1948 | ||
1949 | #define anon_vma_interval_tree_foreach(avc, root, start, last) \ | |
1950 | for (avc = anon_vma_interval_tree_iter_first(root, start, last); \ | |
1951 | avc; avc = anon_vma_interval_tree_iter_next(avc, start, last)) | |
1952 | ||
1953 | /* mmap.c */ | |
1954 | extern int __vm_enough_memory(struct mm_struct *mm, long pages, int cap_sys_admin); | |
1955 | extern int __vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
1956 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert, | |
1957 | struct vm_area_struct *expand); | |
1958 | static inline int vma_adjust(struct vm_area_struct *vma, unsigned long start, | |
1959 | unsigned long end, pgoff_t pgoff, struct vm_area_struct *insert) | |
1960 | { | |
1961 | return __vma_adjust(vma, start, end, pgoff, insert, NULL); | |
1962 | } | |
1963 | extern struct vm_area_struct *vma_merge(struct mm_struct *, | |
1964 | struct vm_area_struct *prev, unsigned long addr, unsigned long end, | |
1965 | unsigned long vm_flags, struct anon_vma *, struct file *, pgoff_t, | |
1966 | struct mempolicy *, struct vm_userfaultfd_ctx); | |
1967 | extern struct anon_vma *find_mergeable_anon_vma(struct vm_area_struct *); | |
1968 | extern int split_vma(struct mm_struct *, | |
1969 | struct vm_area_struct *, unsigned long addr, int new_below); | |
1970 | extern int insert_vm_struct(struct mm_struct *, struct vm_area_struct *); | |
1971 | extern void __vma_link_rb(struct mm_struct *, struct vm_area_struct *, | |
1972 | struct rb_node **, struct rb_node *); | |
1973 | extern void unlink_file_vma(struct vm_area_struct *); | |
1974 | extern struct vm_area_struct *copy_vma(struct vm_area_struct **, | |
1975 | unsigned long addr, unsigned long len, pgoff_t pgoff, | |
1976 | bool *need_rmap_locks); | |
1977 | extern void exit_mmap(struct mm_struct *); | |
1978 | ||
1979 | static inline int check_data_rlimit(unsigned long rlim, | |
1980 | unsigned long new, | |
1981 | unsigned long start, | |
1982 | unsigned long end_data, | |
1983 | unsigned long start_data) | |
1984 | { | |
1985 | if (rlim < RLIM_INFINITY) { | |
1986 | if (((new - start) + (end_data - start_data)) > rlim) | |
1987 | return -ENOSPC; | |
1988 | } | |
1989 | ||
1990 | return 0; | |
1991 | } | |
1992 | ||
1993 | extern int mm_take_all_locks(struct mm_struct *mm); | |
1994 | extern void mm_drop_all_locks(struct mm_struct *mm); | |
1995 | ||
1996 | extern void set_mm_exe_file(struct mm_struct *mm, struct file *new_exe_file); | |
1997 | extern struct file *get_mm_exe_file(struct mm_struct *mm); | |
1998 | extern struct file *get_task_exe_file(struct task_struct *task); | |
1999 | ||
2000 | extern bool may_expand_vm(struct mm_struct *, vm_flags_t, unsigned long npages); | |
2001 | extern void vm_stat_account(struct mm_struct *, vm_flags_t, long npages); | |
2002 | ||
2003 | extern bool vma_is_special_mapping(const struct vm_area_struct *vma, | |
2004 | const struct vm_special_mapping *sm); | |
2005 | extern struct vm_area_struct *_install_special_mapping(struct mm_struct *mm, | |
2006 | unsigned long addr, unsigned long len, | |
2007 | unsigned long flags, | |
2008 | const struct vm_special_mapping *spec); | |
2009 | /* This is an obsolete alternative to _install_special_mapping. */ | |
2010 | extern int install_special_mapping(struct mm_struct *mm, | |
2011 | unsigned long addr, unsigned long len, | |
2012 | unsigned long flags, struct page **pages); | |
2013 | ||
2014 | extern unsigned long get_unmapped_area(struct file *, unsigned long, unsigned long, unsigned long, unsigned long); | |
2015 | ||
2016 | extern unsigned long mmap_region(struct file *file, unsigned long addr, | |
2017 | unsigned long len, vm_flags_t vm_flags, unsigned long pgoff); | |
2018 | extern unsigned long do_mmap(struct file *file, unsigned long addr, | |
2019 | unsigned long len, unsigned long prot, unsigned long flags, | |
2020 | vm_flags_t vm_flags, unsigned long pgoff, unsigned long *populate); | |
2021 | extern int do_munmap(struct mm_struct *, unsigned long, size_t); | |
2022 | ||
2023 | static inline unsigned long | |
2024 | do_mmap_pgoff(struct file *file, unsigned long addr, | |
2025 | unsigned long len, unsigned long prot, unsigned long flags, | |
2026 | unsigned long pgoff, unsigned long *populate) | |
2027 | { | |
2028 | return do_mmap(file, addr, len, prot, flags, 0, pgoff, populate); | |
2029 | } | |
2030 | ||
2031 | #ifdef CONFIG_MMU | |
2032 | extern int __mm_populate(unsigned long addr, unsigned long len, | |
2033 | int ignore_errors); | |
2034 | static inline void mm_populate(unsigned long addr, unsigned long len) | |
2035 | { | |
2036 | /* Ignore errors */ | |
2037 | (void) __mm_populate(addr, len, 1); | |
2038 | } | |
2039 | #else | |
2040 | static inline void mm_populate(unsigned long addr, unsigned long len) {} | |
2041 | #endif | |
2042 | ||
2043 | /* These take the mm semaphore themselves */ | |
2044 | extern int __must_check vm_brk(unsigned long, unsigned long); | |
2045 | extern int vm_munmap(unsigned long, size_t); | |
2046 | extern unsigned long __must_check vm_mmap(struct file *, unsigned long, | |
2047 | unsigned long, unsigned long, | |
2048 | unsigned long, unsigned long); | |
2049 | ||
2050 | struct vm_unmapped_area_info { | |
2051 | #define VM_UNMAPPED_AREA_TOPDOWN 1 | |
2052 | unsigned long flags; | |
2053 | unsigned long length; | |
2054 | unsigned long low_limit; | |
2055 | unsigned long high_limit; | |
2056 | unsigned long align_mask; | |
2057 | unsigned long align_offset; | |
2058 | }; | |
2059 | ||
2060 | extern unsigned long unmapped_area(struct vm_unmapped_area_info *info); | |
2061 | extern unsigned long unmapped_area_topdown(struct vm_unmapped_area_info *info); | |
2062 | ||
2063 | /* | |
2064 | * Search for an unmapped address range. | |
2065 | * | |
2066 | * We are looking for a range that: | |
2067 | * - does not intersect with any VMA; | |
2068 | * - is contained within the [low_limit, high_limit) interval; | |
2069 | * - is at least the desired size. | |
2070 | * - satisfies (begin_addr & align_mask) == (align_offset & align_mask) | |
2071 | */ | |
2072 | static inline unsigned long | |
2073 | vm_unmapped_area(struct vm_unmapped_area_info *info) | |
2074 | { | |
2075 | if (info->flags & VM_UNMAPPED_AREA_TOPDOWN) | |
2076 | return unmapped_area_topdown(info); | |
2077 | else | |
2078 | return unmapped_area(info); | |
2079 | } | |
2080 | ||
2081 | /* truncate.c */ | |
2082 | extern void truncate_inode_pages(struct address_space *, loff_t); | |
2083 | extern void truncate_inode_pages_range(struct address_space *, | |
2084 | loff_t lstart, loff_t lend); | |
2085 | extern void truncate_inode_pages_final(struct address_space *); | |
2086 | ||
2087 | /* generic vm_area_ops exported for stackable file systems */ | |
2088 | extern int filemap_fault(struct vm_area_struct *, struct vm_fault *); | |
2089 | extern void filemap_map_pages(struct vm_fault *vmf, | |
2090 | pgoff_t start_pgoff, pgoff_t end_pgoff); | |
2091 | extern int filemap_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); | |
2092 | ||
2093 | /* mm/page-writeback.c */ | |
2094 | int write_one_page(struct page *page, int wait); | |
2095 | void task_dirty_inc(struct task_struct *tsk); | |
2096 | ||
2097 | /* readahead.c */ | |
2098 | #define VM_MAX_READAHEAD 128 /* kbytes */ | |
2099 | #define VM_MIN_READAHEAD 16 /* kbytes (includes current page) */ | |
2100 | ||
2101 | int force_page_cache_readahead(struct address_space *mapping, struct file *filp, | |
2102 | pgoff_t offset, unsigned long nr_to_read); | |
2103 | ||
2104 | void page_cache_sync_readahead(struct address_space *mapping, | |
2105 | struct file_ra_state *ra, | |
2106 | struct file *filp, | |
2107 | pgoff_t offset, | |
2108 | unsigned long size); | |
2109 | ||
2110 | void page_cache_async_readahead(struct address_space *mapping, | |
2111 | struct file_ra_state *ra, | |
2112 | struct file *filp, | |
2113 | struct page *pg, | |
2114 | pgoff_t offset, | |
2115 | unsigned long size); | |
2116 | ||
2117 | /* Generic expand stack which grows the stack according to GROWS{UP,DOWN} */ | |
2118 | extern int expand_stack(struct vm_area_struct *vma, unsigned long address); | |
2119 | ||
2120 | /* CONFIG_STACK_GROWSUP still needs to to grow downwards at some places */ | |
2121 | extern int expand_downwards(struct vm_area_struct *vma, | |
2122 | unsigned long address); | |
2123 | #if VM_GROWSUP | |
2124 | extern int expand_upwards(struct vm_area_struct *vma, unsigned long address); | |
2125 | #else | |
2126 | #define expand_upwards(vma, address) (0) | |
2127 | #endif | |
2128 | ||
2129 | /* Look up the first VMA which satisfies addr < vm_end, NULL if none. */ | |
2130 | extern struct vm_area_struct * find_vma(struct mm_struct * mm, unsigned long addr); | |
2131 | extern struct vm_area_struct * find_vma_prev(struct mm_struct * mm, unsigned long addr, | |
2132 | struct vm_area_struct **pprev); | |
2133 | ||
2134 | /* Look up the first VMA which intersects the interval start_addr..end_addr-1, | |
2135 | NULL if none. Assume start_addr < end_addr. */ | |
2136 | static inline struct vm_area_struct * find_vma_intersection(struct mm_struct * mm, unsigned long start_addr, unsigned long end_addr) | |
2137 | { | |
2138 | struct vm_area_struct * vma = find_vma(mm,start_addr); | |
2139 | ||
2140 | if (vma && end_addr <= vma->vm_start) | |
2141 | vma = NULL; | |
2142 | return vma; | |
2143 | } | |
2144 | ||
2145 | static inline unsigned long vma_pages(struct vm_area_struct *vma) | |
2146 | { | |
2147 | return (vma->vm_end - vma->vm_start) >> PAGE_SHIFT; | |
2148 | } | |
2149 | ||
2150 | /* Look up the first VMA which exactly match the interval vm_start ... vm_end */ | |
2151 | static inline struct vm_area_struct *find_exact_vma(struct mm_struct *mm, | |
2152 | unsigned long vm_start, unsigned long vm_end) | |
2153 | { | |
2154 | struct vm_area_struct *vma = find_vma(mm, vm_start); | |
2155 | ||
2156 | if (vma && (vma->vm_start != vm_start || vma->vm_end != vm_end)) | |
2157 | vma = NULL; | |
2158 | ||
2159 | return vma; | |
2160 | } | |
2161 | ||
2162 | #ifdef CONFIG_MMU | |
2163 | pgprot_t vm_get_page_prot(unsigned long vm_flags); | |
2164 | void vma_set_page_prot(struct vm_area_struct *vma); | |
2165 | #else | |
2166 | static inline pgprot_t vm_get_page_prot(unsigned long vm_flags) | |
2167 | { | |
2168 | return __pgprot(0); | |
2169 | } | |
2170 | static inline void vma_set_page_prot(struct vm_area_struct *vma) | |
2171 | { | |
2172 | vma->vm_page_prot = vm_get_page_prot(vma->vm_flags); | |
2173 | } | |
2174 | #endif | |
2175 | ||
2176 | #ifdef CONFIG_NUMA_BALANCING | |
2177 | unsigned long change_prot_numa(struct vm_area_struct *vma, | |
2178 | unsigned long start, unsigned long end); | |
2179 | #endif | |
2180 | ||
2181 | struct vm_area_struct *find_extend_vma(struct mm_struct *, unsigned long addr); | |
2182 | int remap_pfn_range(struct vm_area_struct *, unsigned long addr, | |
2183 | unsigned long pfn, unsigned long size, pgprot_t); | |
2184 | int vm_insert_page(struct vm_area_struct *, unsigned long addr, struct page *); | |
2185 | int vm_insert_pfn(struct vm_area_struct *vma, unsigned long addr, | |
2186 | unsigned long pfn); | |
2187 | int vm_insert_pfn_prot(struct vm_area_struct *vma, unsigned long addr, | |
2188 | unsigned long pfn, pgprot_t pgprot); | |
2189 | int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, | |
2190 | pfn_t pfn); | |
2191 | int vm_iomap_memory(struct vm_area_struct *vma, phys_addr_t start, unsigned long len); | |
2192 | ||
2193 | ||
2194 | struct page *follow_page_mask(struct vm_area_struct *vma, | |
2195 | unsigned long address, unsigned int foll_flags, | |
2196 | unsigned int *page_mask); | |
2197 | ||
2198 | static inline struct page *follow_page(struct vm_area_struct *vma, | |
2199 | unsigned long address, unsigned int foll_flags) | |
2200 | { | |
2201 | unsigned int unused_page_mask; | |
2202 | return follow_page_mask(vma, address, foll_flags, &unused_page_mask); | |
2203 | } | |
2204 | ||
2205 | #define FOLL_WRITE 0x01 /* check pte is writable */ | |
2206 | #define FOLL_TOUCH 0x02 /* mark page accessed */ | |
2207 | #define FOLL_GET 0x04 /* do get_page on page */ | |
2208 | #define FOLL_DUMP 0x08 /* give error on hole if it would be zero */ | |
2209 | #define FOLL_FORCE 0x10 /* get_user_pages read/write w/o permission */ | |
2210 | #define FOLL_NOWAIT 0x20 /* if a disk transfer is needed, start the IO | |
2211 | * and return without waiting upon it */ | |
2212 | #define FOLL_POPULATE 0x40 /* fault in page */ | |
2213 | #define FOLL_SPLIT 0x80 /* don't return transhuge pages, split them */ | |
2214 | #define FOLL_HWPOISON 0x100 /* check page is hwpoisoned */ | |
2215 | #define FOLL_NUMA 0x200 /* force NUMA hinting page fault */ | |
2216 | #define FOLL_MIGRATION 0x400 /* wait for page to replace migration entry */ | |
2217 | #define FOLL_TRIED 0x800 /* a retry, previous pass started an IO */ | |
2218 | #define FOLL_MLOCK 0x1000 /* lock present pages */ | |
2219 | #define FOLL_REMOTE 0x2000 /* we are working on non-current tsk/mm */ | |
2220 | #define FOLL_COW 0x4000 /* internal GUP flag */ | |
2221 | ||
2222 | typedef int (*pte_fn_t)(pte_t *pte, pgtable_t token, unsigned long addr, | |
2223 | void *data); | |
2224 | extern int apply_to_page_range(struct mm_struct *mm, unsigned long address, | |
2225 | unsigned long size, pte_fn_t fn, void *data); | |
2226 | ||
2227 | ||
2228 | #ifdef CONFIG_PAGE_POISONING | |
2229 | extern bool page_poisoning_enabled(void); | |
2230 | extern void kernel_poison_pages(struct page *page, int numpages, int enable); | |
2231 | extern bool page_is_poisoned(struct page *page); | |
2232 | #else | |
2233 | static inline bool page_poisoning_enabled(void) { return false; } | |
2234 | static inline void kernel_poison_pages(struct page *page, int numpages, | |
2235 | int enable) { } | |
2236 | static inline bool page_is_poisoned(struct page *page) { return false; } | |
2237 | #endif | |
2238 | ||
2239 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
2240 | extern bool _debug_pagealloc_enabled; | |
2241 | extern void __kernel_map_pages(struct page *page, int numpages, int enable); | |
2242 | ||
2243 | static inline bool debug_pagealloc_enabled(void) | |
2244 | { | |
2245 | return _debug_pagealloc_enabled; | |
2246 | } | |
2247 | ||
2248 | static inline void | |
2249 | kernel_map_pages(struct page *page, int numpages, int enable) | |
2250 | { | |
2251 | if (!debug_pagealloc_enabled()) | |
2252 | return; | |
2253 | ||
2254 | __kernel_map_pages(page, numpages, enable); | |
2255 | } | |
2256 | #ifdef CONFIG_HIBERNATION | |
2257 | extern bool kernel_page_present(struct page *page); | |
2258 | #endif /* CONFIG_HIBERNATION */ | |
2259 | #else /* CONFIG_DEBUG_PAGEALLOC */ | |
2260 | static inline void | |
2261 | kernel_map_pages(struct page *page, int numpages, int enable) {} | |
2262 | #ifdef CONFIG_HIBERNATION | |
2263 | static inline bool kernel_page_present(struct page *page) { return true; } | |
2264 | #endif /* CONFIG_HIBERNATION */ | |
2265 | static inline bool debug_pagealloc_enabled(void) | |
2266 | { | |
2267 | return false; | |
2268 | } | |
2269 | #endif /* CONFIG_DEBUG_PAGEALLOC */ | |
2270 | ||
2271 | #ifdef __HAVE_ARCH_GATE_AREA | |
2272 | extern struct vm_area_struct *get_gate_vma(struct mm_struct *mm); | |
2273 | extern int in_gate_area_no_mm(unsigned long addr); | |
2274 | extern int in_gate_area(struct mm_struct *mm, unsigned long addr); | |
2275 | #else | |
2276 | static inline struct vm_area_struct *get_gate_vma(struct mm_struct *mm) | |
2277 | { | |
2278 | return NULL; | |
2279 | } | |
2280 | static inline int in_gate_area_no_mm(unsigned long addr) { return 0; } | |
2281 | static inline int in_gate_area(struct mm_struct *mm, unsigned long addr) | |
2282 | { | |
2283 | return 0; | |
2284 | } | |
2285 | #endif /* __HAVE_ARCH_GATE_AREA */ | |
2286 | ||
2287 | extern bool process_shares_mm(struct task_struct *p, struct mm_struct *mm); | |
2288 | ||
2289 | #ifdef CONFIG_SYSCTL | |
2290 | extern int sysctl_drop_caches; | |
2291 | int drop_caches_sysctl_handler(struct ctl_table *, int, | |
2292 | void __user *, size_t *, loff_t *); | |
2293 | #endif | |
2294 | ||
2295 | void drop_slab(void); | |
2296 | void drop_slab_node(int nid); | |
2297 | ||
2298 | #ifndef CONFIG_MMU | |
2299 | #define randomize_va_space 0 | |
2300 | #else | |
2301 | extern int randomize_va_space; | |
2302 | #endif | |
2303 | ||
2304 | const char * arch_vma_name(struct vm_area_struct *vma); | |
2305 | void print_vma_addr(char *prefix, unsigned long rip); | |
2306 | ||
2307 | void sparse_mem_maps_populate_node(struct page **map_map, | |
2308 | unsigned long pnum_begin, | |
2309 | unsigned long pnum_end, | |
2310 | unsigned long map_count, | |
2311 | int nodeid); | |
2312 | ||
2313 | struct page *sparse_mem_map_populate(unsigned long pnum, int nid); | |
2314 | pgd_t *vmemmap_pgd_populate(unsigned long addr, int node); | |
2315 | pud_t *vmemmap_pud_populate(pgd_t *pgd, unsigned long addr, int node); | |
2316 | pmd_t *vmemmap_pmd_populate(pud_t *pud, unsigned long addr, int node); | |
2317 | pte_t *vmemmap_pte_populate(pmd_t *pmd, unsigned long addr, int node); | |
2318 | void *vmemmap_alloc_block(unsigned long size, int node); | |
2319 | struct vmem_altmap; | |
2320 | void *__vmemmap_alloc_block_buf(unsigned long size, int node, | |
2321 | struct vmem_altmap *altmap); | |
2322 | static inline void *vmemmap_alloc_block_buf(unsigned long size, int node) | |
2323 | { | |
2324 | return __vmemmap_alloc_block_buf(size, node, NULL); | |
2325 | } | |
2326 | ||
2327 | void vmemmap_verify(pte_t *, int, unsigned long, unsigned long); | |
2328 | int vmemmap_populate_basepages(unsigned long start, unsigned long end, | |
2329 | int node); | |
2330 | int vmemmap_populate(unsigned long start, unsigned long end, int node); | |
2331 | void vmemmap_populate_print_last(void); | |
2332 | #ifdef CONFIG_MEMORY_HOTPLUG | |
2333 | void vmemmap_free(unsigned long start, unsigned long end); | |
2334 | #endif | |
2335 | void register_page_bootmem_memmap(unsigned long section_nr, struct page *map, | |
2336 | unsigned long size); | |
2337 | ||
2338 | enum mf_flags { | |
2339 | MF_COUNT_INCREASED = 1 << 0, | |
2340 | MF_ACTION_REQUIRED = 1 << 1, | |
2341 | MF_MUST_KILL = 1 << 2, | |
2342 | MF_SOFT_OFFLINE = 1 << 3, | |
2343 | }; | |
2344 | extern int memory_failure(unsigned long pfn, int trapno, int flags); | |
2345 | extern void memory_failure_queue(unsigned long pfn, int trapno, int flags); | |
2346 | extern int unpoison_memory(unsigned long pfn); | |
2347 | extern int get_hwpoison_page(struct page *page); | |
2348 | #define put_hwpoison_page(page) put_page(page) | |
2349 | extern int sysctl_memory_failure_early_kill; | |
2350 | extern int sysctl_memory_failure_recovery; | |
2351 | extern void shake_page(struct page *p, int access); | |
2352 | extern atomic_long_t num_poisoned_pages; | |
2353 | extern int soft_offline_page(struct page *page, int flags); | |
2354 | ||
2355 | ||
2356 | /* | |
2357 | * Error handlers for various types of pages. | |
2358 | */ | |
2359 | enum mf_result { | |
2360 | MF_IGNORED, /* Error: cannot be handled */ | |
2361 | MF_FAILED, /* Error: handling failed */ | |
2362 | MF_DELAYED, /* Will be handled later */ | |
2363 | MF_RECOVERED, /* Successfully recovered */ | |
2364 | }; | |
2365 | ||
2366 | enum mf_action_page_type { | |
2367 | MF_MSG_KERNEL, | |
2368 | MF_MSG_KERNEL_HIGH_ORDER, | |
2369 | MF_MSG_SLAB, | |
2370 | MF_MSG_DIFFERENT_COMPOUND, | |
2371 | MF_MSG_POISONED_HUGE, | |
2372 | MF_MSG_HUGE, | |
2373 | MF_MSG_FREE_HUGE, | |
2374 | MF_MSG_UNMAP_FAILED, | |
2375 | MF_MSG_DIRTY_SWAPCACHE, | |
2376 | MF_MSG_CLEAN_SWAPCACHE, | |
2377 | MF_MSG_DIRTY_MLOCKED_LRU, | |
2378 | MF_MSG_CLEAN_MLOCKED_LRU, | |
2379 | MF_MSG_DIRTY_UNEVICTABLE_LRU, | |
2380 | MF_MSG_CLEAN_UNEVICTABLE_LRU, | |
2381 | MF_MSG_DIRTY_LRU, | |
2382 | MF_MSG_CLEAN_LRU, | |
2383 | MF_MSG_TRUNCATED_LRU, | |
2384 | MF_MSG_BUDDY, | |
2385 | MF_MSG_BUDDY_2ND, | |
2386 | MF_MSG_UNKNOWN, | |
2387 | }; | |
2388 | ||
2389 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) || defined(CONFIG_HUGETLBFS) | |
2390 | extern void clear_huge_page(struct page *page, | |
2391 | unsigned long addr, | |
2392 | unsigned int pages_per_huge_page); | |
2393 | extern void copy_user_huge_page(struct page *dst, struct page *src, | |
2394 | unsigned long addr, struct vm_area_struct *vma, | |
2395 | unsigned int pages_per_huge_page); | |
2396 | #endif /* CONFIG_TRANSPARENT_HUGEPAGE || CONFIG_HUGETLBFS */ | |
2397 | ||
2398 | extern struct page_ext_operations debug_guardpage_ops; | |
2399 | extern struct page_ext_operations page_poisoning_ops; | |
2400 | ||
2401 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
2402 | extern unsigned int _debug_guardpage_minorder; | |
2403 | extern bool _debug_guardpage_enabled; | |
2404 | ||
2405 | static inline unsigned int debug_guardpage_minorder(void) | |
2406 | { | |
2407 | return _debug_guardpage_minorder; | |
2408 | } | |
2409 | ||
2410 | static inline bool debug_guardpage_enabled(void) | |
2411 | { | |
2412 | return _debug_guardpage_enabled; | |
2413 | } | |
2414 | ||
2415 | static inline bool page_is_guard(struct page *page) | |
2416 | { | |
2417 | struct page_ext *page_ext; | |
2418 | ||
2419 | if (!debug_guardpage_enabled()) | |
2420 | return false; | |
2421 | ||
2422 | page_ext = lookup_page_ext(page); | |
2423 | if (unlikely(!page_ext)) | |
2424 | return false; | |
2425 | ||
2426 | return test_bit(PAGE_EXT_DEBUG_GUARD, &page_ext->flags); | |
2427 | } | |
2428 | #else | |
2429 | static inline unsigned int debug_guardpage_minorder(void) { return 0; } | |
2430 | static inline bool debug_guardpage_enabled(void) { return false; } | |
2431 | static inline bool page_is_guard(struct page *page) { return false; } | |
2432 | #endif /* CONFIG_DEBUG_PAGEALLOC */ | |
2433 | ||
2434 | #if MAX_NUMNODES > 1 | |
2435 | void __init setup_nr_node_ids(void); | |
2436 | #else | |
2437 | static inline void setup_nr_node_ids(void) {} | |
2438 | #endif | |
2439 | ||
2440 | #endif /* __KERNEL__ */ | |
2441 | #endif /* _LINUX_MM_H */ |