]>
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> |
d4b3b638 | 13 | #include <linux/uprobes.h> |
bbeae5b0 | 14 | #include <linux/page-flags-layout.h> |
c92ff1bd MS |
15 | #include <asm/page.h> |
16 | #include <asm/mmu.h> | |
5b99cd0e | 17 | |
4f9a58d7 OH |
18 | #ifndef AT_VECTOR_SIZE_ARCH |
19 | #define AT_VECTOR_SIZE_ARCH 0 | |
20 | #endif | |
21 | #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1)) | |
22 | ||
5b99cd0e | 23 | struct address_space; |
1306a85a | 24 | struct mem_cgroup; |
5b99cd0e | 25 | |
57c1ffce | 26 | #define USE_SPLIT_PTE_PTLOCKS (NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS) |
e009bb30 KS |
27 | #define USE_SPLIT_PMD_PTLOCKS (USE_SPLIT_PTE_PTLOCKS && \ |
28 | IS_ENABLED(CONFIG_ARCH_ENABLE_SPLIT_PMD_PTLOCK)) | |
597d795a | 29 | #define ALLOC_SPLIT_PTLOCKS (SPINLOCK_SIZE > BITS_PER_LONG/8) |
f7d0b926 | 30 | |
5b99cd0e HC |
31 | /* |
32 | * Each physical page in the system has a struct page associated with | |
33 | * it to keep track of whatever it is we are using the page for at the | |
34 | * moment. Note that we have no way to track which tasks are using | |
35 | * a page, though if it is a pagecache page, rmap structures can tell us | |
36 | * who is mapping it. | |
fc9bb8c7 CL |
37 | * |
38 | * The objects in struct page are organized in double word blocks in | |
39 | * order to allows us to use atomic double word operations on portions | |
40 | * of struct page. That is currently only used by slub but the arrangement | |
41 | * allows the use of atomic double word operations on the flags/mapping | |
42 | * and lru list pointers also. | |
5b99cd0e HC |
43 | */ |
44 | struct page { | |
fc9bb8c7 | 45 | /* First double word block */ |
5b99cd0e HC |
46 | unsigned long flags; /* Atomic flags, some possibly |
47 | * updated asynchronously */ | |
8456a648 JK |
48 | union { |
49 | struct address_space *mapping; /* If low bit clear, points to | |
50 | * inode address_space, or NULL. | |
51 | * If page mapped as anonymous | |
52 | * memory, low bit is set, and | |
53 | * it points to anon_vma object: | |
54 | * see PAGE_MAPPING_ANON below. | |
55 | */ | |
56 | void *s_mem; /* slab first object */ | |
53f9263b | 57 | atomic_t compound_mapcount; /* first tail page */ |
9a982250 | 58 | /* page_deferred_list().next -- second tail page */ |
8456a648 JK |
59 | }; |
60 | ||
fc9bb8c7 | 61 | /* Second double word */ |
013e8963 CL |
62 | struct { |
63 | union { | |
fc9bb8c7 | 64 | pgoff_t index; /* Our offset within mapping. */ |
8456a648 | 65 | void *freelist; /* sl[aou]b first free object */ |
9a982250 | 66 | /* page_deferred_list().prev -- second tail page */ |
013e8963 CL |
67 | }; |
68 | ||
69 | union { | |
abca7c49 PS |
70 | #if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \ |
71 | defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE) | |
013e8963 CL |
72 | /* Used for cmpxchg_double in slub */ |
73 | unsigned long counters; | |
abca7c49 PS |
74 | #else |
75 | /* | |
0139aa7b JK |
76 | * Keep _refcount separate from slub cmpxchg_double |
77 | * data. As the rest of the double word is protected by | |
78 | * slab_lock but _refcount is not. | |
abca7c49 PS |
79 | */ |
80 | unsigned counters; | |
81 | #endif | |
013e8963 CL |
82 | |
83 | struct { | |
84 | ||
85 | union { | |
70b50f94 | 86 | /* |
ddc58f27 KS |
87 | * Count of ptes mapped in mms, to show |
88 | * when page is mapped & limit reverse | |
89 | * map searches. | |
70b50f94 AA |
90 | */ |
91 | atomic_t _mapcount; | |
fc9bb8c7 | 92 | |
b8c24c4a | 93 | struct { /* SLUB */ |
013e8963 CL |
94 | unsigned inuse:16; |
95 | unsigned objects:15; | |
96 | unsigned frozen:1; | |
97 | }; | |
b8c24c4a | 98 | int units; /* SLOB */ |
3adf004d | 99 | }; |
0139aa7b JK |
100 | /* |
101 | * Usage count, *USE WRAPPER FUNCTION* | |
102 | * when manual accounting. See page_ref.h | |
103 | */ | |
104 | atomic_t _refcount; | |
fc9bb8c7 | 105 | }; |
8456a648 | 106 | unsigned int active; /* SLAB */ |
39b26464 | 107 | }; |
81819f0f | 108 | }; |
fc9bb8c7 | 109 | |
1d798ca3 KS |
110 | /* |
111 | * Third double word block | |
112 | * | |
113 | * WARNING: bit 0 of the first word encode PageTail(). That means | |
114 | * the rest users of the storage space MUST NOT use the bit to | |
115 | * avoid collision and false-positive PageTail(). | |
116 | */ | |
49e22585 CL |
117 | union { |
118 | struct list_head lru; /* Pageout list, eg. active_list | |
fc9bb8c7 | 119 | * protected by zone->lru_lock ! |
34bf6ef9 DH |
120 | * Can be used as a generic list |
121 | * by the page owner. | |
fc9bb8c7 | 122 | */ |
5c2c2587 DW |
123 | struct dev_pagemap *pgmap; /* ZONE_DEVICE pages are never on an |
124 | * lru or handled by a slab | |
125 | * allocator, this points to the | |
126 | * hosting device page map. | |
127 | */ | |
49e22585 CL |
128 | struct { /* slub per cpu partial pages */ |
129 | struct page *next; /* Next partial slab */ | |
130 | #ifdef CONFIG_64BIT | |
131 | int pages; /* Nr of partial slabs left */ | |
132 | int pobjects; /* Approximate # of objects */ | |
133 | #else | |
134 | short int pages; | |
135 | short int pobjects; | |
136 | #endif | |
137 | }; | |
b8c24c4a | 138 | |
68126702 JK |
139 | struct rcu_head rcu_head; /* Used by SLAB |
140 | * when destroying via RCU | |
141 | */ | |
1d798ca3 | 142 | /* Tail pages of compound page */ |
e4b294c2 | 143 | struct { |
1d798ca3 KS |
144 | unsigned long compound_head; /* If bit zero is set */ |
145 | ||
146 | /* First tail page only */ | |
1965c8b7 KS |
147 | #ifdef CONFIG_64BIT |
148 | /* | |
149 | * On 64 bit system we have enough space in struct page | |
150 | * to encode compound_dtor and compound_order with | |
151 | * unsigned int. It can help compiler generate better or | |
152 | * smaller code on some archtectures. | |
153 | */ | |
154 | unsigned int compound_dtor; | |
155 | unsigned int compound_order; | |
156 | #else | |
f1e61557 KS |
157 | unsigned short int compound_dtor; |
158 | unsigned short int compound_order; | |
1965c8b7 | 159 | #endif |
e4b294c2 KS |
160 | }; |
161 | ||
7aa555bf | 162 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && USE_SPLIT_PMD_PTLOCKS |
1d798ca3 KS |
163 | struct { |
164 | unsigned long __pad; /* do not overlay pmd_huge_pte | |
165 | * with compound_head to avoid | |
166 | * possible bit 0 collision. | |
167 | */ | |
168 | pgtable_t pmd_huge_pte; /* protected by page->ptl */ | |
169 | }; | |
7aa555bf | 170 | #endif |
49e22585 | 171 | }; |
fc9bb8c7 CL |
172 | |
173 | /* Remainder is not double word aligned */ | |
5b99cd0e | 174 | union { |
5b99cd0e HC |
175 | unsigned long private; /* Mapping-private opaque data: |
176 | * usually used for buffer_heads | |
177 | * if PagePrivate set; used for | |
178 | * swp_entry_t if PageSwapCache; | |
179 | * indicates order in the buddy | |
180 | * system if PG_buddy is set. | |
181 | */ | |
57c1ffce | 182 | #if USE_SPLIT_PTE_PTLOCKS |
597d795a | 183 | #if ALLOC_SPLIT_PTLOCKS |
539edb58 PZ |
184 | spinlock_t *ptl; |
185 | #else | |
186 | spinlock_t ptl; | |
187 | #endif | |
5b99cd0e | 188 | #endif |
1b4f59e3 | 189 | struct kmem_cache *slab_cache; /* SL[AU]B: Pointer to slab */ |
81819f0f | 190 | }; |
fc9bb8c7 | 191 | |
1306a85a JW |
192 | #ifdef CONFIG_MEMCG |
193 | struct mem_cgroup *mem_cgroup; | |
194 | #endif | |
195 | ||
5b99cd0e HC |
196 | /* |
197 | * On machines where all RAM is mapped into kernel address space, | |
198 | * we can simply calculate the virtual address. On machines with | |
199 | * highmem some memory is mapped into kernel virtual memory | |
200 | * dynamically, so we need a place to store that address. | |
201 | * Note that this field could be 16 bits on x86 ... ;) | |
202 | * | |
203 | * Architectures with slow multiplication can define | |
204 | * WANT_PAGE_VIRTUAL in asm/page.h | |
205 | */ | |
206 | #if defined(WANT_PAGE_VIRTUAL) | |
207 | void *virtual; /* Kernel virtual address (NULL if | |
208 | not kmapped, ie. highmem) */ | |
209 | #endif /* WANT_PAGE_VIRTUAL */ | |
dfec072e VN |
210 | |
211 | #ifdef CONFIG_KMEMCHECK | |
212 | /* | |
213 | * kmemcheck wants to track the status of each byte in a page; this | |
214 | * is a pointer to such a status block. NULL if not tracked. | |
215 | */ | |
216 | void *shadow; | |
217 | #endif | |
57e0a030 | 218 | |
90572890 PZ |
219 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
220 | int _last_cpupid; | |
57e0a030 | 221 | #endif |
fc9bb8c7 CL |
222 | } |
223 | /* | |
43570fd2 HC |
224 | * The struct page can be forced to be double word aligned so that atomic ops |
225 | * on double words work. The SLUB allocator can make use of such a feature. | |
fc9bb8c7 | 226 | */ |
43570fd2 HC |
227 | #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE |
228 | __aligned(2 * sizeof(unsigned long)) | |
fc9bb8c7 CL |
229 | #endif |
230 | ; | |
5b99cd0e | 231 | |
30d3c128 IC |
232 | struct page_frag { |
233 | struct page *page; | |
234 | #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) | |
235 | __u32 offset; | |
236 | __u32 size; | |
237 | #else | |
238 | __u16 offset; | |
239 | __u16 size; | |
240 | #endif | |
241 | }; | |
242 | ||
b63ae8ca AD |
243 | #define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK) |
244 | #define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE) | |
245 | ||
246 | struct page_frag_cache { | |
247 | void * va; | |
248 | #if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE) | |
249 | __u16 offset; | |
250 | __u16 size; | |
251 | #else | |
252 | __u32 offset; | |
253 | #endif | |
254 | /* we maintain a pagecount bias, so that we dont dirty cache line | |
0139aa7b | 255 | * containing page->_refcount every time we allocate a fragment. |
b63ae8ca AD |
256 | */ |
257 | unsigned int pagecnt_bias; | |
258 | bool pfmemalloc; | |
259 | }; | |
260 | ||
64b990d2 | 261 | typedef unsigned long vm_flags_t; |
ca16d140 | 262 | |
8feae131 DH |
263 | /* |
264 | * A region containing a mapping of a non-memory backed file under NOMMU | |
265 | * conditions. These are held in a global tree and are pinned by the VMAs that | |
266 | * map parts of them. | |
267 | */ | |
268 | struct vm_region { | |
269 | struct rb_node vm_rb; /* link in global region tree */ | |
ca16d140 | 270 | vm_flags_t vm_flags; /* VMA vm_flags */ |
8feae131 DH |
271 | unsigned long vm_start; /* start address of region */ |
272 | unsigned long vm_end; /* region initialised to here */ | |
dd8632a1 | 273 | unsigned long vm_top; /* region allocated to here */ |
8feae131 DH |
274 | unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ |
275 | struct file *vm_file; /* the backing file or NULL */ | |
276 | ||
1e2ae599 | 277 | int vm_usage; /* region usage count (access under nommu_region_sem) */ |
cfe79c00 MF |
278 | bool vm_icache_flushed : 1; /* true if the icache has been flushed for |
279 | * this region */ | |
8feae131 DH |
280 | }; |
281 | ||
745f234b AA |
282 | #ifdef CONFIG_USERFAULTFD |
283 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, }) | |
284 | struct vm_userfaultfd_ctx { | |
285 | struct userfaultfd_ctx *ctx; | |
286 | }; | |
287 | #else /* CONFIG_USERFAULTFD */ | |
288 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {}) | |
289 | struct vm_userfaultfd_ctx {}; | |
290 | #endif /* CONFIG_USERFAULTFD */ | |
291 | ||
c92ff1bd MS |
292 | /* |
293 | * This struct defines a memory VMM memory area. There is one of these | |
294 | * per VM-area/task. A VM area is any part of the process virtual memory | |
295 | * space that has a special rule for the page-fault handlers (ie a shared | |
296 | * library, the executable area etc). | |
297 | */ | |
298 | struct vm_area_struct { | |
e4c6bfd2 RR |
299 | /* The first cache line has the info for VMA tree walking. */ |
300 | ||
c92ff1bd MS |
301 | unsigned long vm_start; /* Our start address within vm_mm. */ |
302 | unsigned long vm_end; /* The first byte after our end address | |
303 | within vm_mm. */ | |
304 | ||
305 | /* linked list of VM areas per task, sorted by address */ | |
297c5eee | 306 | struct vm_area_struct *vm_next, *vm_prev; |
c92ff1bd | 307 | |
c92ff1bd MS |
308 | struct rb_node vm_rb; |
309 | ||
d3737187 ML |
310 | /* |
311 | * Largest free memory gap in bytes to the left of this VMA. | |
312 | * Either between this VMA and vma->vm_prev, or between one of the | |
313 | * VMAs below us in the VMA rbtree and its ->vm_prev. This helps | |
314 | * get_unmapped_area find a free area of the right size. | |
315 | */ | |
316 | unsigned long rb_subtree_gap; | |
317 | ||
e4c6bfd2 RR |
318 | /* Second cache line starts here. */ |
319 | ||
320 | struct mm_struct *vm_mm; /* The address space we belong to. */ | |
321 | pgprot_t vm_page_prot; /* Access permissions of this VMA. */ | |
322 | unsigned long vm_flags; /* Flags, see mm.h. */ | |
323 | ||
c92ff1bd MS |
324 | /* |
325 | * For areas with an address space and backing store, | |
27ba0644 | 326 | * linkage into the address_space->i_mmap interval tree. |
c92ff1bd | 327 | */ |
ac51b934 KS |
328 | struct { |
329 | struct rb_node rb; | |
330 | unsigned long rb_subtree_last; | |
c92ff1bd MS |
331 | } shared; |
332 | ||
333 | /* | |
334 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma | |
335 | * list, after a COW of one of the file pages. A MAP_SHARED vma | |
336 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack | |
337 | * or brk vma (with NULL file) can only be in an anon_vma list. | |
338 | */ | |
5beb4930 RR |
339 | struct list_head anon_vma_chain; /* Serialized by mmap_sem & |
340 | * page_table_lock */ | |
c92ff1bd MS |
341 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
342 | ||
343 | /* Function pointers to deal with this struct. */ | |
f0f37e2f | 344 | const struct vm_operations_struct *vm_ops; |
c92ff1bd MS |
345 | |
346 | /* Information about our backing store: */ | |
347 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE | |
ea1754a0 | 348 | units */ |
c92ff1bd MS |
349 | struct file * vm_file; /* File we map to (can be NULL). */ |
350 | void * vm_private_data; /* was vm_pte (shared mem) */ | |
c92ff1bd MS |
351 | |
352 | #ifndef CONFIG_MMU | |
8feae131 | 353 | struct vm_region *vm_region; /* NOMMU mapping region */ |
c92ff1bd MS |
354 | #endif |
355 | #ifdef CONFIG_NUMA | |
356 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ | |
357 | #endif | |
745f234b | 358 | struct vm_userfaultfd_ctx vm_userfaultfd_ctx; |
c92ff1bd MS |
359 | }; |
360 | ||
b564daf8 ON |
361 | struct core_thread { |
362 | struct task_struct *task; | |
363 | struct core_thread *next; | |
364 | }; | |
365 | ||
32ecb1f2 | 366 | struct core_state { |
c5f1cc8c | 367 | atomic_t nr_threads; |
b564daf8 | 368 | struct core_thread dumper; |
32ecb1f2 ON |
369 | struct completion startup; |
370 | }; | |
371 | ||
d559db08 | 372 | enum { |
eca56ff9 JM |
373 | MM_FILEPAGES, /* Resident file mapping pages */ |
374 | MM_ANONPAGES, /* Resident anonymous pages */ | |
375 | MM_SWAPENTS, /* Anonymous swap entries */ | |
376 | MM_SHMEMPAGES, /* Resident shared memory pages */ | |
d559db08 KH |
377 | NR_MM_COUNTERS |
378 | }; | |
379 | ||
57c1ffce | 380 | #if USE_SPLIT_PTE_PTLOCKS && defined(CONFIG_MMU) |
34e55232 | 381 | #define SPLIT_RSS_COUNTING |
34e55232 KH |
382 | /* per-thread cached information, */ |
383 | struct task_rss_stat { | |
384 | int events; /* for synchronization threshold */ | |
385 | int count[NR_MM_COUNTERS]; | |
386 | }; | |
57c1ffce | 387 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
172703b0 | 388 | |
d559db08 | 389 | struct mm_rss_stat { |
172703b0 | 390 | atomic_long_t count[NR_MM_COUNTERS]; |
d559db08 | 391 | }; |
d559db08 | 392 | |
db446a08 | 393 | struct kioctx_table; |
c92ff1bd | 394 | struct mm_struct { |
615d6e87 | 395 | struct vm_area_struct *mmap; /* list of VMAs */ |
c92ff1bd | 396 | struct rb_root mm_rb; |
615d6e87 | 397 | u32 vmacache_seqnum; /* per-thread vmacache */ |
efc1a3b1 | 398 | #ifdef CONFIG_MMU |
c92ff1bd MS |
399 | unsigned long (*get_unmapped_area) (struct file *filp, |
400 | unsigned long addr, unsigned long len, | |
401 | unsigned long pgoff, unsigned long flags); | |
efc1a3b1 | 402 | #endif |
c92ff1bd | 403 | unsigned long mmap_base; /* base of mmap area */ |
41aacc1e | 404 | unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */ |
c92ff1bd | 405 | unsigned long task_size; /* size of task vm space */ |
d3737187 | 406 | unsigned long highest_vm_end; /* highest vma end address */ |
c92ff1bd MS |
407 | pgd_t * pgd; |
408 | atomic_t mm_users; /* How many users with user space? */ | |
409 | atomic_t mm_count; /* How many references to "struct mm_struct" (users count as 1) */ | |
dc6c9a35 | 410 | atomic_long_t nr_ptes; /* PTE page table pages */ |
5a3fbef3 | 411 | #if CONFIG_PGTABLE_LEVELS > 2 |
dc6c9a35 | 412 | atomic_long_t nr_pmds; /* PMD page table pages */ |
5a3fbef3 | 413 | #endif |
c92ff1bd | 414 | int map_count; /* number of VMAs */ |
481b4bb5 | 415 | |
c92ff1bd | 416 | spinlock_t page_table_lock; /* Protects page tables and some counters */ |
481b4bb5 | 417 | struct rw_semaphore mmap_sem; |
c92ff1bd MS |
418 | |
419 | struct list_head mmlist; /* List of maybe swapped mm's. These are globally strung | |
420 | * together off init_mm.mmlist, and are protected | |
421 | * by mmlist_lock | |
422 | */ | |
423 | ||
c92ff1bd MS |
424 | |
425 | unsigned long hiwater_rss; /* High-watermark of RSS usage */ | |
426 | unsigned long hiwater_vm; /* High-water virtual memory usage */ | |
427 | ||
e10d59f2 CL |
428 | unsigned long total_vm; /* Total pages mapped */ |
429 | unsigned long locked_vm; /* Pages that have PG_mlocked set */ | |
430 | unsigned long pinned_vm; /* Refcount permanently increased */ | |
30bdbb78 KK |
431 | unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */ |
432 | unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */ | |
433 | unsigned long stack_vm; /* VM_STACK */ | |
e10d59f2 | 434 | unsigned long def_flags; |
c92ff1bd MS |
435 | unsigned long start_code, end_code, start_data, end_data; |
436 | unsigned long start_brk, brk, start_stack; | |
437 | unsigned long arg_start, arg_end, env_start, env_end; | |
438 | ||
439 | unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ | |
440 | ||
d559db08 KH |
441 | /* |
442 | * Special counters, in some configurations protected by the | |
443 | * page_table_lock, in other configurations by being atomic. | |
444 | */ | |
445 | struct mm_rss_stat rss_stat; | |
446 | ||
801460d0 HS |
447 | struct linux_binfmt *binfmt; |
448 | ||
6345d24d LT |
449 | cpumask_var_t cpu_vm_mask_var; |
450 | ||
c92ff1bd MS |
451 | /* Architecture-specific MM context */ |
452 | mm_context_t context; | |
453 | ||
c92ff1bd MS |
454 | unsigned long flags; /* Must use atomic bitops to access the bits */ |
455 | ||
a94e2d40 | 456 | struct core_state *core_state; /* coredumping support */ |
858f0993 | 457 | #ifdef CONFIG_AIO |
db446a08 BL |
458 | spinlock_t ioctx_lock; |
459 | struct kioctx_table __rcu *ioctx_table; | |
858f0993 | 460 | #endif |
f98bafa0 | 461 | #ifdef CONFIG_MEMCG |
4cd1a8fc KM |
462 | /* |
463 | * "owner" points to a task that is regarded as the canonical | |
464 | * user/owner of this mm. All of the following must be true in | |
465 | * order for it to be changed: | |
466 | * | |
467 | * current == mm->owner | |
468 | * current->mm != mm | |
469 | * new_owner->mm == mm | |
470 | * new_owner->alloc_lock is held | |
471 | */ | |
4d2deb40 | 472 | struct task_struct __rcu *owner; |
78fb7466 | 473 | #endif |
925d1c40 | 474 | |
925d1c40 | 475 | /* store ref to file /proc/<pid>/exe symlink points to */ |
90f31d0e | 476 | struct file __rcu *exe_file; |
cddb8a5c AA |
477 | #ifdef CONFIG_MMU_NOTIFIER |
478 | struct mmu_notifier_mm *mmu_notifier_mm; | |
e7a00c45 | 479 | #endif |
e009bb30 | 480 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
e7a00c45 | 481 | pgtable_t pmd_huge_pte; /* protected by page_table_lock */ |
cddb8a5c | 482 | #endif |
6345d24d LT |
483 | #ifdef CONFIG_CPUMASK_OFFSTACK |
484 | struct cpumask cpumask_allocation; | |
cbee9f88 PZ |
485 | #endif |
486 | #ifdef CONFIG_NUMA_BALANCING | |
487 | /* | |
34f0315a MG |
488 | * numa_next_scan is the next time that the PTEs will be marked |
489 | * pte_numa. NUMA hinting faults will gather statistics and migrate | |
490 | * pages to new nodes if necessary. | |
cbee9f88 PZ |
491 | */ |
492 | unsigned long numa_next_scan; | |
493 | ||
6e5fb223 PZ |
494 | /* Restart point for scanning and setting pte_numa */ |
495 | unsigned long numa_scan_offset; | |
496 | ||
cbee9f88 PZ |
497 | /* numa_scan_seq prevents two threads setting pte_numa */ |
498 | int numa_scan_seq; | |
20841405 RR |
499 | #endif |
500 | #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) | |
501 | /* | |
502 | * An operation with batched TLB flushing is going on. Anything that | |
503 | * can move process memory needs to flush the TLB when moving a | |
504 | * PROT_NONE or PROT_NUMA mapped page. | |
505 | */ | |
506 | bool tlb_flush_pending; | |
6345d24d | 507 | #endif |
d4b3b638 | 508 | struct uprobes_state uprobes_state; |
fe3d197f DH |
509 | #ifdef CONFIG_X86_INTEL_MPX |
510 | /* address of the bounds directory */ | |
511 | void __user *bd_addr; | |
512 | #endif | |
5d317b2b NH |
513 | #ifdef CONFIG_HUGETLB_PAGE |
514 | atomic_long_t hugetlb_usage; | |
515 | #endif | |
c92ff1bd MS |
516 | }; |
517 | ||
6345d24d LT |
518 | static inline void mm_init_cpumask(struct mm_struct *mm) |
519 | { | |
520 | #ifdef CONFIG_CPUMASK_OFFSTACK | |
521 | mm->cpu_vm_mask_var = &mm->cpumask_allocation; | |
522 | #endif | |
41f727fd | 523 | cpumask_clear(mm->cpu_vm_mask_var); |
6345d24d LT |
524 | } |
525 | ||
45e575ab | 526 | /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ |
de03c72c KM |
527 | static inline cpumask_t *mm_cpumask(struct mm_struct *mm) |
528 | { | |
529 | return mm->cpu_vm_mask_var; | |
530 | } | |
45e575ab | 531 | |
20841405 RR |
532 | #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) |
533 | /* | |
534 | * Memory barriers to keep this state in sync are graciously provided by | |
535 | * the page table locks, outside of which no page table modifications happen. | |
536 | * The barriers below prevent the compiler from re-ordering the instructions | |
537 | * around the memory barriers that are already present in the code. | |
538 | */ | |
539 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) | |
540 | { | |
541 | barrier(); | |
542 | return mm->tlb_flush_pending; | |
543 | } | |
544 | static inline void set_tlb_flush_pending(struct mm_struct *mm) | |
545 | { | |
546 | mm->tlb_flush_pending = true; | |
af2c1401 MG |
547 | |
548 | /* | |
549 | * Guarantee that the tlb_flush_pending store does not leak into the | |
550 | * critical section updating the page tables | |
551 | */ | |
552 | smp_mb__before_spinlock(); | |
20841405 RR |
553 | } |
554 | /* Clearing is done after a TLB flush, which also provides a barrier. */ | |
555 | static inline void clear_tlb_flush_pending(struct mm_struct *mm) | |
556 | { | |
557 | barrier(); | |
558 | mm->tlb_flush_pending = false; | |
559 | } | |
560 | #else | |
561 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) | |
562 | { | |
563 | return false; | |
564 | } | |
565 | static inline void set_tlb_flush_pending(struct mm_struct *mm) | |
566 | { | |
567 | } | |
568 | static inline void clear_tlb_flush_pending(struct mm_struct *mm) | |
569 | { | |
570 | } | |
571 | #endif | |
572 | ||
f872f540 AL |
573 | struct vm_fault; |
574 | ||
575 | struct vm_special_mapping { | |
576 | const char *name; /* The name, e.g. "[vdso]". */ | |
577 | ||
578 | /* | |
579 | * If .fault is not provided, this points to a | |
580 | * NULL-terminated array of pages that back the special mapping. | |
581 | * | |
582 | * This must not be NULL unless .fault is provided. | |
583 | */ | |
a62c34bd | 584 | struct page **pages; |
f872f540 AL |
585 | |
586 | /* | |
587 | * If non-NULL, then this is called to resolve page faults | |
588 | * on the special mapping. If used, .pages is not checked. | |
589 | */ | |
590 | int (*fault)(const struct vm_special_mapping *sm, | |
591 | struct vm_area_struct *vma, | |
592 | struct vm_fault *vmf); | |
a62c34bd AL |
593 | }; |
594 | ||
d17d8f9d DH |
595 | enum tlb_flush_reason { |
596 | TLB_FLUSH_ON_TASK_SWITCH, | |
597 | TLB_REMOTE_SHOOTDOWN, | |
598 | TLB_LOCAL_SHOOTDOWN, | |
599 | TLB_LOCAL_MM_SHOOTDOWN, | |
5b74283a | 600 | TLB_REMOTE_SEND_IPI, |
d17d8f9d DH |
601 | NR_TLB_FLUSH_REASONS, |
602 | }; | |
603 | ||
bd6dace7 TH |
604 | /* |
605 | * A swap entry has to fit into a "unsigned long", as the entry is hidden | |
606 | * in the "index" field of the swapper address space. | |
607 | */ | |
608 | typedef struct { | |
609 | unsigned long val; | |
610 | } swp_entry_t; | |
611 | ||
5b99cd0e | 612 | #endif /* _LINUX_MM_TYPES_H */ |