Commit | Line | Data |
---|---|---|
5b99cd0e HC |
1 | #ifndef _LINUX_MM_TYPES_H |
2 | #define _LINUX_MM_TYPES_H | |
3 | ||
4f9a58d7 | 4 | #include <linux/auxvec.h> |
5b99cd0e HC |
5 | #include <linux/types.h> |
6 | #include <linux/threads.h> | |
7 | #include <linux/list.h> | |
8 | #include <linux/spinlock.h> | |
c92ff1bd MS |
9 | #include <linux/rbtree.h> |
10 | #include <linux/rwsem.h> | |
11 | #include <linux/completion.h> | |
cddb8a5c | 12 | #include <linux/cpumask.h> |
6a11f75b | 13 | #include <linux/page-debug-flags.h> |
d4b3b638 | 14 | #include <linux/uprobes.h> |
bbeae5b0 | 15 | #include <linux/page-flags-layout.h> |
c92ff1bd MS |
16 | #include <asm/page.h> |
17 | #include <asm/mmu.h> | |
5b99cd0e | 18 | |
4f9a58d7 OH |
19 | #ifndef AT_VECTOR_SIZE_ARCH |
20 | #define AT_VECTOR_SIZE_ARCH 0 | |
21 | #endif | |
22 | #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1)) | |
23 | ||
5b99cd0e HC |
24 | struct address_space; |
25 | ||
f7d0b926 JF |
26 | #define USE_SPLIT_PTLOCKS (NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS) |
27 | ||
5b99cd0e HC |
28 | /* |
29 | * Each physical page in the system has a struct page associated with | |
30 | * it to keep track of whatever it is we are using the page for at the | |
31 | * moment. Note that we have no way to track which tasks are using | |
32 | * a page, though if it is a pagecache page, rmap structures can tell us | |
33 | * who is mapping it. | |
fc9bb8c7 CL |
34 | * |
35 | * The objects in struct page are organized in double word blocks in | |
36 | * order to allows us to use atomic double word operations on portions | |
37 | * of struct page. That is currently only used by slub but the arrangement | |
38 | * allows the use of atomic double word operations on the flags/mapping | |
39 | * and lru list pointers also. | |
5b99cd0e HC |
40 | */ |
41 | struct page { | |
fc9bb8c7 | 42 | /* First double word block */ |
5b99cd0e HC |
43 | unsigned long flags; /* Atomic flags, some possibly |
44 | * updated asynchronously */ | |
fc9bb8c7 CL |
45 | struct address_space *mapping; /* If low bit clear, points to |
46 | * inode address_space, or NULL. | |
47 | * If page mapped as anonymous | |
48 | * memory, low bit is set, and | |
49 | * it points to anon_vma object: | |
50 | * see PAGE_MAPPING_ANON below. | |
51 | */ | |
52 | /* Second double word */ | |
013e8963 CL |
53 | struct { |
54 | union { | |
fc9bb8c7 | 55 | pgoff_t index; /* Our offset within mapping. */ |
b8c24c4a | 56 | void *freelist; /* slub/slob first free object */ |
072bb0aa | 57 | bool pfmemalloc; /* If set by the page allocator, |
b37f1dd0 | 58 | * ALLOC_NO_WATERMARKS was set |
072bb0aa MG |
59 | * and the low watermark was not |
60 | * met implying that the system | |
61 | * is under some pressure. The | |
62 | * caller should try ensure | |
63 | * this page is only used to | |
64 | * free other pages. | |
65 | */ | |
013e8963 CL |
66 | }; |
67 | ||
68 | union { | |
abca7c49 PS |
69 | #if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \ |
70 | defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE) | |
013e8963 CL |
71 | /* Used for cmpxchg_double in slub */ |
72 | unsigned long counters; | |
abca7c49 PS |
73 | #else |
74 | /* | |
75 | * Keep _count separate from slub cmpxchg_double data. | |
76 | * As the rest of the double word is protected by | |
77 | * slab_lock but _count is not. | |
78 | */ | |
79 | unsigned counters; | |
80 | #endif | |
013e8963 CL |
81 | |
82 | struct { | |
83 | ||
84 | union { | |
70b50f94 AA |
85 | /* |
86 | * Count of ptes mapped in | |
87 | * mms, to show when page is | |
88 | * mapped & limit reverse map | |
89 | * searches. | |
90 | * | |
91 | * Used also for tail pages | |
92 | * refcounting instead of | |
93 | * _count. Tail pages cannot | |
94 | * be mapped and keeping the | |
95 | * tail page _count zero at | |
96 | * all times guarantees | |
97 | * get_page_unless_zero() will | |
98 | * never succeed on tail | |
99 | * pages. | |
100 | */ | |
101 | atomic_t _mapcount; | |
fc9bb8c7 | 102 | |
b8c24c4a | 103 | struct { /* SLUB */ |
013e8963 CL |
104 | unsigned inuse:16; |
105 | unsigned objects:15; | |
106 | unsigned frozen:1; | |
107 | }; | |
b8c24c4a | 108 | int units; /* SLOB */ |
3adf004d | 109 | }; |
013e8963 | 110 | atomic_t _count; /* Usage count, see below. */ |
fc9bb8c7 | 111 | }; |
39b26464 | 112 | }; |
81819f0f | 113 | }; |
fc9bb8c7 CL |
114 | |
115 | /* Third double word block */ | |
49e22585 CL |
116 | union { |
117 | struct list_head lru; /* Pageout list, eg. active_list | |
fc9bb8c7 CL |
118 | * protected by zone->lru_lock ! |
119 | */ | |
49e22585 CL |
120 | struct { /* slub per cpu partial pages */ |
121 | struct page *next; /* Next partial slab */ | |
122 | #ifdef CONFIG_64BIT | |
123 | int pages; /* Nr of partial slabs left */ | |
124 | int pobjects; /* Approximate # of objects */ | |
125 | #else | |
126 | short int pages; | |
127 | short int pobjects; | |
128 | #endif | |
129 | }; | |
b8c24c4a CL |
130 | |
131 | struct list_head list; /* slobs list of pages */ | |
1b4f59e3 | 132 | struct slab *slab_page; /* slab fields */ |
49e22585 | 133 | }; |
fc9bb8c7 CL |
134 | |
135 | /* Remainder is not double word aligned */ | |
5b99cd0e | 136 | union { |
5b99cd0e HC |
137 | unsigned long private; /* Mapping-private opaque data: |
138 | * usually used for buffer_heads | |
139 | * if PagePrivate set; used for | |
140 | * swp_entry_t if PageSwapCache; | |
141 | * indicates order in the buddy | |
142 | * system if PG_buddy is set. | |
143 | */ | |
f7d0b926 | 144 | #if USE_SPLIT_PTLOCKS |
fc9bb8c7 | 145 | spinlock_t ptl; |
5b99cd0e | 146 | #endif |
1b4f59e3 | 147 | struct kmem_cache *slab_cache; /* SL[AU]B: Pointer to slab */ |
fc9bb8c7 | 148 | struct page *first_page; /* Compound tail pages */ |
81819f0f | 149 | }; |
fc9bb8c7 | 150 | |
5b99cd0e HC |
151 | /* |
152 | * On machines where all RAM is mapped into kernel address space, | |
153 | * we can simply calculate the virtual address. On machines with | |
154 | * highmem some memory is mapped into kernel virtual memory | |
155 | * dynamically, so we need a place to store that address. | |
156 | * Note that this field could be 16 bits on x86 ... ;) | |
157 | * | |
158 | * Architectures with slow multiplication can define | |
159 | * WANT_PAGE_VIRTUAL in asm/page.h | |
160 | */ | |
161 | #if defined(WANT_PAGE_VIRTUAL) | |
162 | void *virtual; /* Kernel virtual address (NULL if | |
163 | not kmapped, ie. highmem) */ | |
164 | #endif /* WANT_PAGE_VIRTUAL */ | |
ee3b4290 AM |
165 | #ifdef CONFIG_WANT_PAGE_DEBUG_FLAGS |
166 | unsigned long debug_flags; /* Use atomic bitops on this */ | |
167 | #endif | |
dfec072e VN |
168 | |
169 | #ifdef CONFIG_KMEMCHECK | |
170 | /* | |
171 | * kmemcheck wants to track the status of each byte in a page; this | |
172 | * is a pointer to such a status block. NULL if not tracked. | |
173 | */ | |
174 | void *shadow; | |
175 | #endif | |
57e0a030 | 176 | |
75980e97 | 177 | #ifdef LAST_NID_NOT_IN_PAGE_FLAGS |
57e0a030 MG |
178 | int _last_nid; |
179 | #endif | |
fc9bb8c7 CL |
180 | } |
181 | /* | |
43570fd2 HC |
182 | * The struct page can be forced to be double word aligned so that atomic ops |
183 | * on double words work. The SLUB allocator can make use of such a feature. | |
fc9bb8c7 | 184 | */ |
43570fd2 HC |
185 | #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE |
186 | __aligned(2 * sizeof(unsigned long)) | |
fc9bb8c7 CL |
187 | #endif |
188 | ; | |
5b99cd0e | 189 | |
30d3c128 IC |
190 | struct page_frag { |
191 | struct page *page; | |
192 | #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) | |
193 | __u32 offset; | |
194 | __u32 size; | |
195 | #else | |
196 | __u16 offset; | |
197 | __u16 size; | |
198 | #endif | |
199 | }; | |
200 | ||
ca16d140 KM |
201 | typedef unsigned long __nocast vm_flags_t; |
202 | ||
8feae131 DH |
203 | /* |
204 | * A region containing a mapping of a non-memory backed file under NOMMU | |
205 | * conditions. These are held in a global tree and are pinned by the VMAs that | |
206 | * map parts of them. | |
207 | */ | |
208 | struct vm_region { | |
209 | struct rb_node vm_rb; /* link in global region tree */ | |
ca16d140 | 210 | vm_flags_t vm_flags; /* VMA vm_flags */ |
8feae131 DH |
211 | unsigned long vm_start; /* start address of region */ |
212 | unsigned long vm_end; /* region initialised to here */ | |
dd8632a1 | 213 | unsigned long vm_top; /* region allocated to here */ |
8feae131 DH |
214 | unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ |
215 | struct file *vm_file; /* the backing file or NULL */ | |
216 | ||
1e2ae599 | 217 | int vm_usage; /* region usage count (access under nommu_region_sem) */ |
cfe79c00 MF |
218 | bool vm_icache_flushed : 1; /* true if the icache has been flushed for |
219 | * this region */ | |
8feae131 DH |
220 | }; |
221 | ||
c92ff1bd MS |
222 | /* |
223 | * This struct defines a memory VMM memory area. There is one of these | |
224 | * per VM-area/task. A VM area is any part of the process virtual memory | |
225 | * space that has a special rule for the page-fault handlers (ie a shared | |
226 | * library, the executable area etc). | |
227 | */ | |
228 | struct vm_area_struct { | |
e4c6bfd2 RR |
229 | /* The first cache line has the info for VMA tree walking. */ |
230 | ||
c92ff1bd MS |
231 | unsigned long vm_start; /* Our start address within vm_mm. */ |
232 | unsigned long vm_end; /* The first byte after our end address | |
233 | within vm_mm. */ | |
234 | ||
235 | /* linked list of VM areas per task, sorted by address */ | |
297c5eee | 236 | struct vm_area_struct *vm_next, *vm_prev; |
c92ff1bd | 237 | |
c92ff1bd MS |
238 | struct rb_node vm_rb; |
239 | ||
d3737187 ML |
240 | /* |
241 | * Largest free memory gap in bytes to the left of this VMA. | |
242 | * Either between this VMA and vma->vm_prev, or between one of the | |
243 | * VMAs below us in the VMA rbtree and its ->vm_prev. This helps | |
244 | * get_unmapped_area find a free area of the right size. | |
245 | */ | |
246 | unsigned long rb_subtree_gap; | |
247 | ||
e4c6bfd2 RR |
248 | /* Second cache line starts here. */ |
249 | ||
250 | struct mm_struct *vm_mm; /* The address space we belong to. */ | |
251 | pgprot_t vm_page_prot; /* Access permissions of this VMA. */ | |
252 | unsigned long vm_flags; /* Flags, see mm.h. */ | |
253 | ||
c92ff1bd MS |
254 | /* |
255 | * For areas with an address space and backing store, | |
6b2dbba8 | 256 | * linkage into the address_space->i_mmap interval tree, or |
c92ff1bd MS |
257 | * linkage of vma in the address_space->i_mmap_nonlinear list. |
258 | */ | |
259 | union { | |
260 | struct { | |
6b2dbba8 ML |
261 | struct rb_node rb; |
262 | unsigned long rb_subtree_last; | |
263 | } linear; | |
264 | struct list_head nonlinear; | |
c92ff1bd MS |
265 | } shared; |
266 | ||
267 | /* | |
268 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma | |
269 | * list, after a COW of one of the file pages. A MAP_SHARED vma | |
270 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack | |
271 | * or brk vma (with NULL file) can only be in an anon_vma list. | |
272 | */ | |
5beb4930 RR |
273 | struct list_head anon_vma_chain; /* Serialized by mmap_sem & |
274 | * page_table_lock */ | |
c92ff1bd MS |
275 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
276 | ||
277 | /* Function pointers to deal with this struct. */ | |
f0f37e2f | 278 | const struct vm_operations_struct *vm_ops; |
c92ff1bd MS |
279 | |
280 | /* Information about our backing store: */ | |
281 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE | |
282 | units, *not* PAGE_CACHE_SIZE */ | |
283 | struct file * vm_file; /* File we map to (can be NULL). */ | |
284 | void * vm_private_data; /* was vm_pte (shared mem) */ | |
c92ff1bd MS |
285 | |
286 | #ifndef CONFIG_MMU | |
8feae131 | 287 | struct vm_region *vm_region; /* NOMMU mapping region */ |
c92ff1bd MS |
288 | #endif |
289 | #ifdef CONFIG_NUMA | |
290 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ | |
291 | #endif | |
292 | }; | |
293 | ||
b564daf8 ON |
294 | struct core_thread { |
295 | struct task_struct *task; | |
296 | struct core_thread *next; | |
297 | }; | |
298 | ||
32ecb1f2 | 299 | struct core_state { |
c5f1cc8c | 300 | atomic_t nr_threads; |
b564daf8 | 301 | struct core_thread dumper; |
32ecb1f2 ON |
302 | struct completion startup; |
303 | }; | |
304 | ||
d559db08 KH |
305 | enum { |
306 | MM_FILEPAGES, | |
307 | MM_ANONPAGES, | |
b084d435 | 308 | MM_SWAPENTS, |
d559db08 KH |
309 | NR_MM_COUNTERS |
310 | }; | |
311 | ||
53bddb4e | 312 | #if USE_SPLIT_PTLOCKS && defined(CONFIG_MMU) |
34e55232 | 313 | #define SPLIT_RSS_COUNTING |
34e55232 KH |
314 | /* per-thread cached information, */ |
315 | struct task_rss_stat { | |
316 | int events; /* for synchronization threshold */ | |
317 | int count[NR_MM_COUNTERS]; | |
318 | }; | |
172703b0 MF |
319 | #endif /* USE_SPLIT_PTLOCKS */ |
320 | ||
d559db08 | 321 | struct mm_rss_stat { |
172703b0 | 322 | atomic_long_t count[NR_MM_COUNTERS]; |
d559db08 | 323 | }; |
d559db08 | 324 | |
c92ff1bd MS |
325 | struct mm_struct { |
326 | struct vm_area_struct * mmap; /* list of VMAs */ | |
327 | struct rb_root mm_rb; | |
328 | struct vm_area_struct * mmap_cache; /* last find_vma result */ | |
efc1a3b1 | 329 | #ifdef CONFIG_MMU |
c92ff1bd MS |
330 | unsigned long (*get_unmapped_area) (struct file *filp, |
331 | unsigned long addr, unsigned long len, | |
332 | unsigned long pgoff, unsigned long flags); | |
333 | void (*unmap_area) (struct mm_struct *mm, unsigned long addr); | |
efc1a3b1 | 334 | #endif |
c92ff1bd MS |
335 | unsigned long mmap_base; /* base of mmap area */ |
336 | unsigned long task_size; /* size of task vm space */ | |
337 | unsigned long cached_hole_size; /* if non-zero, the largest hole below free_area_cache */ | |
338 | unsigned long free_area_cache; /* first hole of size cached_hole_size or larger */ | |
d3737187 | 339 | unsigned long highest_vm_end; /* highest vma end address */ |
c92ff1bd MS |
340 | pgd_t * pgd; |
341 | atomic_t mm_users; /* How many users with user space? */ | |
342 | atomic_t mm_count; /* How many references to "struct mm_struct" (users count as 1) */ | |
343 | int map_count; /* number of VMAs */ | |
481b4bb5 | 344 | |
c92ff1bd | 345 | spinlock_t page_table_lock; /* Protects page tables and some counters */ |
481b4bb5 | 346 | struct rw_semaphore mmap_sem; |
c92ff1bd MS |
347 | |
348 | struct list_head mmlist; /* List of maybe swapped mm's. These are globally strung | |
349 | * together off init_mm.mmlist, and are protected | |
350 | * by mmlist_lock | |
351 | */ | |
352 | ||
c92ff1bd MS |
353 | |
354 | unsigned long hiwater_rss; /* High-watermark of RSS usage */ | |
355 | unsigned long hiwater_vm; /* High-water virtual memory usage */ | |
356 | ||
e10d59f2 CL |
357 | unsigned long total_vm; /* Total pages mapped */ |
358 | unsigned long locked_vm; /* Pages that have PG_mlocked set */ | |
359 | unsigned long pinned_vm; /* Refcount permanently increased */ | |
360 | unsigned long shared_vm; /* Shared pages (files) */ | |
361 | unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE */ | |
362 | unsigned long stack_vm; /* VM_GROWSUP/DOWN */ | |
e10d59f2 CL |
363 | unsigned long def_flags; |
364 | unsigned long nr_ptes; /* Page table pages */ | |
c92ff1bd MS |
365 | unsigned long start_code, end_code, start_data, end_data; |
366 | unsigned long start_brk, brk, start_stack; | |
367 | unsigned long arg_start, arg_end, env_start, env_end; | |
368 | ||
369 | unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ | |
370 | ||
d559db08 KH |
371 | /* |
372 | * Special counters, in some configurations protected by the | |
373 | * page_table_lock, in other configurations by being atomic. | |
374 | */ | |
375 | struct mm_rss_stat rss_stat; | |
376 | ||
801460d0 HS |
377 | struct linux_binfmt *binfmt; |
378 | ||
6345d24d LT |
379 | cpumask_var_t cpu_vm_mask_var; |
380 | ||
c92ff1bd MS |
381 | /* Architecture-specific MM context */ |
382 | mm_context_t context; | |
383 | ||
c92ff1bd MS |
384 | unsigned long flags; /* Must use atomic bitops to access the bits */ |
385 | ||
a94e2d40 | 386 | struct core_state *core_state; /* coredumping support */ |
858f0993 | 387 | #ifdef CONFIG_AIO |
abf137dd JA |
388 | spinlock_t ioctx_lock; |
389 | struct hlist_head ioctx_list; | |
858f0993 | 390 | #endif |
cf475ad2 | 391 | #ifdef CONFIG_MM_OWNER |
4cd1a8fc KM |
392 | /* |
393 | * "owner" points to a task that is regarded as the canonical | |
394 | * user/owner of this mm. All of the following must be true in | |
395 | * order for it to be changed: | |
396 | * | |
397 | * current == mm->owner | |
398 | * current->mm != mm | |
399 | * new_owner->mm == mm | |
400 | * new_owner->alloc_lock is held | |
401 | */ | |
4d2deb40 | 402 | struct task_struct __rcu *owner; |
78fb7466 | 403 | #endif |
925d1c40 | 404 | |
925d1c40 MH |
405 | /* store ref to file /proc/<pid>/exe symlink points to */ |
406 | struct file *exe_file; | |
cddb8a5c AA |
407 | #ifdef CONFIG_MMU_NOTIFIER |
408 | struct mmu_notifier_mm *mmu_notifier_mm; | |
e7a00c45 AA |
409 | #endif |
410 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
411 | pgtable_t pmd_huge_pte; /* protected by page_table_lock */ | |
cddb8a5c | 412 | #endif |
6345d24d LT |
413 | #ifdef CONFIG_CPUMASK_OFFSTACK |
414 | struct cpumask cpumask_allocation; | |
cbee9f88 PZ |
415 | #endif |
416 | #ifdef CONFIG_NUMA_BALANCING | |
417 | /* | |
34f0315a MG |
418 | * numa_next_scan is the next time that the PTEs will be marked |
419 | * pte_numa. NUMA hinting faults will gather statistics and migrate | |
420 | * pages to new nodes if necessary. | |
cbee9f88 PZ |
421 | */ |
422 | unsigned long numa_next_scan; | |
423 | ||
b8593bfd MG |
424 | /* numa_next_reset is when the PTE scanner period will be reset */ |
425 | unsigned long numa_next_reset; | |
426 | ||
6e5fb223 PZ |
427 | /* Restart point for scanning and setting pte_numa */ |
428 | unsigned long numa_scan_offset; | |
429 | ||
cbee9f88 PZ |
430 | /* numa_scan_seq prevents two threads setting pte_numa */ |
431 | int numa_scan_seq; | |
5bca2303 MG |
432 | |
433 | /* | |
434 | * The first node a task was scheduled on. If a task runs on | |
435 | * a different node than Make PTE Scan Go Now. | |
436 | */ | |
437 | int first_nid; | |
6345d24d | 438 | #endif |
d4b3b638 | 439 | struct uprobes_state uprobes_state; |
c92ff1bd MS |
440 | }; |
441 | ||
5bca2303 MG |
442 | /* first nid will either be a valid NID or one of these values */ |
443 | #define NUMA_PTE_SCAN_INIT -1 | |
444 | #define NUMA_PTE_SCAN_ACTIVE -2 | |
445 | ||
6345d24d LT |
446 | static inline void mm_init_cpumask(struct mm_struct *mm) |
447 | { | |
448 | #ifdef CONFIG_CPUMASK_OFFSTACK | |
449 | mm->cpu_vm_mask_var = &mm->cpumask_allocation; | |
450 | #endif | |
451 | } | |
452 | ||
45e575ab | 453 | /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ |
de03c72c KM |
454 | static inline cpumask_t *mm_cpumask(struct mm_struct *mm) |
455 | { | |
456 | return mm->cpu_vm_mask_var; | |
457 | } | |
45e575ab | 458 | |
5b99cd0e | 459 | #endif /* _LINUX_MM_TYPES_H */ |