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> |
ec8d7c14 | 15 | #include <linux/workqueue.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 | 24 | struct address_space; |
1306a85a | 25 | struct mem_cgroup; |
5b99cd0e | 26 | |
57c1ffce | 27 | #define USE_SPLIT_PTE_PTLOCKS (NR_CPUS >= CONFIG_SPLIT_PTLOCK_CPUS) |
e009bb30 KS |
28 | #define USE_SPLIT_PMD_PTLOCKS (USE_SPLIT_PTE_PTLOCKS && \ |
29 | IS_ENABLED(CONFIG_ARCH_ENABLE_SPLIT_PMD_PTLOCK)) | |
597d795a | 30 | #define ALLOC_SPLIT_PTLOCKS (SPINLOCK_SIZE > BITS_PER_LONG/8) |
f7d0b926 | 31 | |
5b99cd0e HC |
32 | /* |
33 | * Each physical page in the system has a struct page associated with | |
34 | * it to keep track of whatever it is we are using the page for at the | |
35 | * moment. Note that we have no way to track which tasks are using | |
36 | * a page, though if it is a pagecache page, rmap structures can tell us | |
37 | * who is mapping it. | |
fc9bb8c7 CL |
38 | * |
39 | * The objects in struct page are organized in double word blocks in | |
40 | * order to allows us to use atomic double word operations on portions | |
41 | * of struct page. That is currently only used by slub but the arrangement | |
42 | * allows the use of atomic double word operations on the flags/mapping | |
43 | * and lru list pointers also. | |
5b99cd0e HC |
44 | */ |
45 | struct page { | |
fc9bb8c7 | 46 | /* First double word block */ |
5b99cd0e HC |
47 | unsigned long flags; /* Atomic flags, some possibly |
48 | * updated asynchronously */ | |
8456a648 JK |
49 | union { |
50 | struct address_space *mapping; /* If low bit clear, points to | |
51 | * inode address_space, or NULL. | |
52 | * If page mapped as anonymous | |
53 | * memory, low bit is set, and | |
54 | * it points to anon_vma object: | |
55 | * see PAGE_MAPPING_ANON below. | |
56 | */ | |
57 | void *s_mem; /* slab first object */ | |
53f9263b | 58 | atomic_t compound_mapcount; /* first tail page */ |
9a982250 | 59 | /* page_deferred_list().next -- second tail page */ |
8456a648 JK |
60 | }; |
61 | ||
fc9bb8c7 | 62 | /* Second double word */ |
013e8963 CL |
63 | struct { |
64 | union { | |
fc9bb8c7 | 65 | pgoff_t index; /* Our offset within mapping. */ |
8456a648 | 66 | void *freelist; /* sl[aou]b first free object */ |
9a982250 | 67 | /* page_deferred_list().prev -- second tail page */ |
013e8963 CL |
68 | }; |
69 | ||
70 | union { | |
abca7c49 PS |
71 | #if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) && \ |
72 | defined(CONFIG_HAVE_ALIGNED_STRUCT_PAGE) | |
013e8963 CL |
73 | /* Used for cmpxchg_double in slub */ |
74 | unsigned long counters; | |
abca7c49 PS |
75 | #else |
76 | /* | |
0139aa7b JK |
77 | * Keep _refcount separate from slub cmpxchg_double |
78 | * data. As the rest of the double word is protected by | |
79 | * slab_lock but _refcount is not. | |
abca7c49 PS |
80 | */ |
81 | unsigned counters; | |
82 | #endif | |
013e8963 CL |
83 | |
84 | struct { | |
85 | ||
86 | union { | |
70b50f94 | 87 | /* |
ddc58f27 KS |
88 | * Count of ptes mapped in mms, to show |
89 | * when page is mapped & limit reverse | |
90 | * map searches. | |
70b50f94 AA |
91 | */ |
92 | atomic_t _mapcount; | |
fc9bb8c7 | 93 | |
b8c24c4a | 94 | struct { /* SLUB */ |
013e8963 CL |
95 | unsigned inuse:16; |
96 | unsigned objects:15; | |
97 | unsigned frozen:1; | |
98 | }; | |
b8c24c4a | 99 | int units; /* SLOB */ |
3adf004d | 100 | }; |
0139aa7b JK |
101 | /* |
102 | * Usage count, *USE WRAPPER FUNCTION* | |
103 | * when manual accounting. See page_ref.h | |
104 | */ | |
105 | atomic_t _refcount; | |
fc9bb8c7 | 106 | }; |
8456a648 | 107 | unsigned int active; /* SLAB */ |
39b26464 | 108 | }; |
81819f0f | 109 | }; |
fc9bb8c7 | 110 | |
1d798ca3 KS |
111 | /* |
112 | * Third double word block | |
113 | * | |
114 | * WARNING: bit 0 of the first word encode PageTail(). That means | |
115 | * the rest users of the storage space MUST NOT use the bit to | |
116 | * avoid collision and false-positive PageTail(). | |
117 | */ | |
49e22585 CL |
118 | union { |
119 | struct list_head lru; /* Pageout list, eg. active_list | |
fc9bb8c7 | 120 | * protected by zone->lru_lock ! |
34bf6ef9 DH |
121 | * Can be used as a generic list |
122 | * by the page owner. | |
fc9bb8c7 | 123 | */ |
5c2c2587 DW |
124 | struct dev_pagemap *pgmap; /* ZONE_DEVICE pages are never on an |
125 | * lru or handled by a slab | |
126 | * allocator, this points to the | |
127 | * hosting device page map. | |
128 | */ | |
49e22585 CL |
129 | struct { /* slub per cpu partial pages */ |
130 | struct page *next; /* Next partial slab */ | |
131 | #ifdef CONFIG_64BIT | |
132 | int pages; /* Nr of partial slabs left */ | |
133 | int pobjects; /* Approximate # of objects */ | |
134 | #else | |
135 | short int pages; | |
136 | short int pobjects; | |
137 | #endif | |
138 | }; | |
b8c24c4a | 139 | |
68126702 JK |
140 | struct rcu_head rcu_head; /* Used by SLAB |
141 | * when destroying via RCU | |
142 | */ | |
1d798ca3 | 143 | /* Tail pages of compound page */ |
e4b294c2 | 144 | struct { |
1d798ca3 KS |
145 | unsigned long compound_head; /* If bit zero is set */ |
146 | ||
147 | /* First tail page only */ | |
1965c8b7 KS |
148 | #ifdef CONFIG_64BIT |
149 | /* | |
150 | * On 64 bit system we have enough space in struct page | |
151 | * to encode compound_dtor and compound_order with | |
152 | * unsigned int. It can help compiler generate better or | |
153 | * smaller code on some archtectures. | |
154 | */ | |
155 | unsigned int compound_dtor; | |
156 | unsigned int compound_order; | |
157 | #else | |
f1e61557 KS |
158 | unsigned short int compound_dtor; |
159 | unsigned short int compound_order; | |
1965c8b7 | 160 | #endif |
e4b294c2 KS |
161 | }; |
162 | ||
7aa555bf | 163 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && USE_SPLIT_PMD_PTLOCKS |
1d798ca3 KS |
164 | struct { |
165 | unsigned long __pad; /* do not overlay pmd_huge_pte | |
166 | * with compound_head to avoid | |
167 | * possible bit 0 collision. | |
168 | */ | |
169 | pgtable_t pmd_huge_pte; /* protected by page->ptl */ | |
170 | }; | |
7aa555bf | 171 | #endif |
49e22585 | 172 | }; |
fc9bb8c7 CL |
173 | |
174 | /* Remainder is not double word aligned */ | |
5b99cd0e | 175 | union { |
5b99cd0e HC |
176 | unsigned long private; /* Mapping-private opaque data: |
177 | * usually used for buffer_heads | |
178 | * if PagePrivate set; used for | |
179 | * swp_entry_t if PageSwapCache; | |
180 | * indicates order in the buddy | |
181 | * system if PG_buddy is set. | |
182 | */ | |
57c1ffce | 183 | #if USE_SPLIT_PTE_PTLOCKS |
597d795a | 184 | #if ALLOC_SPLIT_PTLOCKS |
539edb58 PZ |
185 | spinlock_t *ptl; |
186 | #else | |
187 | spinlock_t ptl; | |
188 | #endif | |
5b99cd0e | 189 | #endif |
1b4f59e3 | 190 | struct kmem_cache *slab_cache; /* SL[AU]B: Pointer to slab */ |
81819f0f | 191 | }; |
fc9bb8c7 | 192 | |
1306a85a JW |
193 | #ifdef CONFIG_MEMCG |
194 | struct mem_cgroup *mem_cgroup; | |
195 | #endif | |
196 | ||
5b99cd0e HC |
197 | /* |
198 | * On machines where all RAM is mapped into kernel address space, | |
199 | * we can simply calculate the virtual address. On machines with | |
200 | * highmem some memory is mapped into kernel virtual memory | |
201 | * dynamically, so we need a place to store that address. | |
202 | * Note that this field could be 16 bits on x86 ... ;) | |
203 | * | |
204 | * Architectures with slow multiplication can define | |
205 | * WANT_PAGE_VIRTUAL in asm/page.h | |
206 | */ | |
207 | #if defined(WANT_PAGE_VIRTUAL) | |
208 | void *virtual; /* Kernel virtual address (NULL if | |
209 | not kmapped, ie. highmem) */ | |
210 | #endif /* WANT_PAGE_VIRTUAL */ | |
dfec072e VN |
211 | |
212 | #ifdef CONFIG_KMEMCHECK | |
213 | /* | |
214 | * kmemcheck wants to track the status of each byte in a page; this | |
215 | * is a pointer to such a status block. NULL if not tracked. | |
216 | */ | |
217 | void *shadow; | |
218 | #endif | |
57e0a030 | 219 | |
90572890 PZ |
220 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
221 | int _last_cpupid; | |
57e0a030 | 222 | #endif |
fc9bb8c7 CL |
223 | } |
224 | /* | |
43570fd2 HC |
225 | * The struct page can be forced to be double word aligned so that atomic ops |
226 | * on double words work. The SLUB allocator can make use of such a feature. | |
fc9bb8c7 | 227 | */ |
43570fd2 HC |
228 | #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE |
229 | __aligned(2 * sizeof(unsigned long)) | |
fc9bb8c7 CL |
230 | #endif |
231 | ; | |
5b99cd0e | 232 | |
30d3c128 IC |
233 | struct page_frag { |
234 | struct page *page; | |
235 | #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536) | |
236 | __u32 offset; | |
237 | __u32 size; | |
238 | #else | |
239 | __u16 offset; | |
240 | __u16 size; | |
241 | #endif | |
242 | }; | |
243 | ||
b63ae8ca AD |
244 | #define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK) |
245 | #define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE) | |
246 | ||
247 | struct page_frag_cache { | |
248 | void * va; | |
249 | #if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE) | |
250 | __u16 offset; | |
251 | __u16 size; | |
252 | #else | |
253 | __u32 offset; | |
254 | #endif | |
255 | /* we maintain a pagecount bias, so that we dont dirty cache line | |
0139aa7b | 256 | * containing page->_refcount every time we allocate a fragment. |
b63ae8ca AD |
257 | */ |
258 | unsigned int pagecnt_bias; | |
259 | bool pfmemalloc; | |
260 | }; | |
261 | ||
64b990d2 | 262 | typedef unsigned long vm_flags_t; |
ca16d140 | 263 | |
8feae131 DH |
264 | /* |
265 | * A region containing a mapping of a non-memory backed file under NOMMU | |
266 | * conditions. These are held in a global tree and are pinned by the VMAs that | |
267 | * map parts of them. | |
268 | */ | |
269 | struct vm_region { | |
270 | struct rb_node vm_rb; /* link in global region tree */ | |
ca16d140 | 271 | vm_flags_t vm_flags; /* VMA vm_flags */ |
8feae131 DH |
272 | unsigned long vm_start; /* start address of region */ |
273 | unsigned long vm_end; /* region initialised to here */ | |
dd8632a1 | 274 | unsigned long vm_top; /* region allocated to here */ |
8feae131 DH |
275 | unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ |
276 | struct file *vm_file; /* the backing file or NULL */ | |
277 | ||
1e2ae599 | 278 | int vm_usage; /* region usage count (access under nommu_region_sem) */ |
cfe79c00 MF |
279 | bool vm_icache_flushed : 1; /* true if the icache has been flushed for |
280 | * this region */ | |
8feae131 DH |
281 | }; |
282 | ||
745f234b AA |
283 | #ifdef CONFIG_USERFAULTFD |
284 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, }) | |
285 | struct vm_userfaultfd_ctx { | |
286 | struct userfaultfd_ctx *ctx; | |
287 | }; | |
288 | #else /* CONFIG_USERFAULTFD */ | |
289 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {}) | |
290 | struct vm_userfaultfd_ctx {}; | |
291 | #endif /* CONFIG_USERFAULTFD */ | |
292 | ||
c92ff1bd MS |
293 | /* |
294 | * This struct defines a memory VMM memory area. There is one of these | |
295 | * per VM-area/task. A VM area is any part of the process virtual memory | |
296 | * space that has a special rule for the page-fault handlers (ie a shared | |
297 | * library, the executable area etc). | |
298 | */ | |
299 | struct vm_area_struct { | |
e4c6bfd2 RR |
300 | /* The first cache line has the info for VMA tree walking. */ |
301 | ||
c92ff1bd MS |
302 | unsigned long vm_start; /* Our start address within vm_mm. */ |
303 | unsigned long vm_end; /* The first byte after our end address | |
304 | within vm_mm. */ | |
305 | ||
306 | /* linked list of VM areas per task, sorted by address */ | |
297c5eee | 307 | struct vm_area_struct *vm_next, *vm_prev; |
c92ff1bd | 308 | |
c92ff1bd MS |
309 | struct rb_node vm_rb; |
310 | ||
d3737187 ML |
311 | /* |
312 | * Largest free memory gap in bytes to the left of this VMA. | |
313 | * Either between this VMA and vma->vm_prev, or between one of the | |
314 | * VMAs below us in the VMA rbtree and its ->vm_prev. This helps | |
315 | * get_unmapped_area find a free area of the right size. | |
316 | */ | |
317 | unsigned long rb_subtree_gap; | |
318 | ||
e4c6bfd2 RR |
319 | /* Second cache line starts here. */ |
320 | ||
321 | struct mm_struct *vm_mm; /* The address space we belong to. */ | |
322 | pgprot_t vm_page_prot; /* Access permissions of this VMA. */ | |
323 | unsigned long vm_flags; /* Flags, see mm.h. */ | |
324 | ||
c92ff1bd MS |
325 | /* |
326 | * For areas with an address space and backing store, | |
27ba0644 | 327 | * linkage into the address_space->i_mmap interval tree. |
c92ff1bd | 328 | */ |
ac51b934 KS |
329 | struct { |
330 | struct rb_node rb; | |
331 | unsigned long rb_subtree_last; | |
c92ff1bd MS |
332 | } shared; |
333 | ||
334 | /* | |
335 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma | |
336 | * list, after a COW of one of the file pages. A MAP_SHARED vma | |
337 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack | |
338 | * or brk vma (with NULL file) can only be in an anon_vma list. | |
339 | */ | |
5beb4930 RR |
340 | struct list_head anon_vma_chain; /* Serialized by mmap_sem & |
341 | * page_table_lock */ | |
c92ff1bd MS |
342 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
343 | ||
344 | /* Function pointers to deal with this struct. */ | |
f0f37e2f | 345 | const struct vm_operations_struct *vm_ops; |
c92ff1bd MS |
346 | |
347 | /* Information about our backing store: */ | |
348 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE | |
ea1754a0 | 349 | units */ |
c92ff1bd MS |
350 | struct file * vm_file; /* File we map to (can be NULL). */ |
351 | void * vm_private_data; /* was vm_pte (shared mem) */ | |
c92ff1bd MS |
352 | |
353 | #ifndef CONFIG_MMU | |
8feae131 | 354 | struct vm_region *vm_region; /* NOMMU mapping region */ |
c92ff1bd MS |
355 | #endif |
356 | #ifdef CONFIG_NUMA | |
357 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ | |
358 | #endif | |
745f234b | 359 | struct vm_userfaultfd_ctx vm_userfaultfd_ctx; |
c92ff1bd MS |
360 | }; |
361 | ||
b564daf8 ON |
362 | struct core_thread { |
363 | struct task_struct *task; | |
364 | struct core_thread *next; | |
365 | }; | |
366 | ||
32ecb1f2 | 367 | struct core_state { |
c5f1cc8c | 368 | atomic_t nr_threads; |
b564daf8 | 369 | struct core_thread dumper; |
32ecb1f2 ON |
370 | struct completion startup; |
371 | }; | |
372 | ||
d559db08 | 373 | enum { |
eca56ff9 JM |
374 | MM_FILEPAGES, /* Resident file mapping pages */ |
375 | MM_ANONPAGES, /* Resident anonymous pages */ | |
376 | MM_SWAPENTS, /* Anonymous swap entries */ | |
377 | MM_SHMEMPAGES, /* Resident shared memory pages */ | |
d559db08 KH |
378 | NR_MM_COUNTERS |
379 | }; | |
380 | ||
57c1ffce | 381 | #if USE_SPLIT_PTE_PTLOCKS && defined(CONFIG_MMU) |
34e55232 | 382 | #define SPLIT_RSS_COUNTING |
34e55232 KH |
383 | /* per-thread cached information, */ |
384 | struct task_rss_stat { | |
385 | int events; /* for synchronization threshold */ | |
386 | int count[NR_MM_COUNTERS]; | |
387 | }; | |
57c1ffce | 388 | #endif /* USE_SPLIT_PTE_PTLOCKS */ |
172703b0 | 389 | |
d559db08 | 390 | struct mm_rss_stat { |
172703b0 | 391 | atomic_long_t count[NR_MM_COUNTERS]; |
d559db08 | 392 | }; |
d559db08 | 393 | |
db446a08 | 394 | struct kioctx_table; |
c92ff1bd | 395 | struct mm_struct { |
615d6e87 | 396 | struct vm_area_struct *mmap; /* list of VMAs */ |
c92ff1bd | 397 | struct rb_root mm_rb; |
615d6e87 | 398 | u32 vmacache_seqnum; /* per-thread vmacache */ |
efc1a3b1 | 399 | #ifdef CONFIG_MMU |
c92ff1bd MS |
400 | unsigned long (*get_unmapped_area) (struct file *filp, |
401 | unsigned long addr, unsigned long len, | |
402 | unsigned long pgoff, unsigned long flags); | |
efc1a3b1 | 403 | #endif |
c92ff1bd | 404 | unsigned long mmap_base; /* base of mmap area */ |
41aacc1e | 405 | unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */ |
c92ff1bd | 406 | unsigned long task_size; /* size of task vm space */ |
d3737187 | 407 | unsigned long highest_vm_end; /* highest vma end address */ |
c92ff1bd MS |
408 | pgd_t * pgd; |
409 | atomic_t mm_users; /* How many users with user space? */ | |
410 | atomic_t mm_count; /* How many references to "struct mm_struct" (users count as 1) */ | |
dc6c9a35 | 411 | atomic_long_t nr_ptes; /* PTE page table pages */ |
5a3fbef3 | 412 | #if CONFIG_PGTABLE_LEVELS > 2 |
dc6c9a35 | 413 | atomic_long_t nr_pmds; /* PMD page table pages */ |
5a3fbef3 | 414 | #endif |
c92ff1bd | 415 | int map_count; /* number of VMAs */ |
481b4bb5 | 416 | |
c92ff1bd | 417 | spinlock_t page_table_lock; /* Protects page tables and some counters */ |
481b4bb5 | 418 | struct rw_semaphore mmap_sem; |
c92ff1bd MS |
419 | |
420 | struct list_head mmlist; /* List of maybe swapped mm's. These are globally strung | |
421 | * together off init_mm.mmlist, and are protected | |
422 | * by mmlist_lock | |
423 | */ | |
424 | ||
c92ff1bd MS |
425 | |
426 | unsigned long hiwater_rss; /* High-watermark of RSS usage */ | |
427 | unsigned long hiwater_vm; /* High-water virtual memory usage */ | |
428 | ||
e10d59f2 CL |
429 | unsigned long total_vm; /* Total pages mapped */ |
430 | unsigned long locked_vm; /* Pages that have PG_mlocked set */ | |
431 | unsigned long pinned_vm; /* Refcount permanently increased */ | |
30bdbb78 KK |
432 | unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */ |
433 | unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */ | |
434 | unsigned long stack_vm; /* VM_STACK */ | |
e10d59f2 | 435 | unsigned long def_flags; |
c92ff1bd MS |
436 | unsigned long start_code, end_code, start_data, end_data; |
437 | unsigned long start_brk, brk, start_stack; | |
438 | unsigned long arg_start, arg_end, env_start, env_end; | |
439 | ||
440 | unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ | |
441 | ||
d559db08 KH |
442 | /* |
443 | * Special counters, in some configurations protected by the | |
444 | * page_table_lock, in other configurations by being atomic. | |
445 | */ | |
446 | struct mm_rss_stat rss_stat; | |
447 | ||
801460d0 HS |
448 | struct linux_binfmt *binfmt; |
449 | ||
6345d24d LT |
450 | cpumask_var_t cpu_vm_mask_var; |
451 | ||
c92ff1bd MS |
452 | /* Architecture-specific MM context */ |
453 | mm_context_t context; | |
454 | ||
c92ff1bd MS |
455 | unsigned long flags; /* Must use atomic bitops to access the bits */ |
456 | ||
a94e2d40 | 457 | struct core_state *core_state; /* coredumping support */ |
858f0993 | 458 | #ifdef CONFIG_AIO |
db446a08 BL |
459 | spinlock_t ioctx_lock; |
460 | struct kioctx_table __rcu *ioctx_table; | |
858f0993 | 461 | #endif |
f98bafa0 | 462 | #ifdef CONFIG_MEMCG |
4cd1a8fc KM |
463 | /* |
464 | * "owner" points to a task that is regarded as the canonical | |
465 | * user/owner of this mm. All of the following must be true in | |
466 | * order for it to be changed: | |
467 | * | |
468 | * current == mm->owner | |
469 | * current->mm != mm | |
470 | * new_owner->mm == mm | |
471 | * new_owner->alloc_lock is held | |
472 | */ | |
4d2deb40 | 473 | struct task_struct __rcu *owner; |
78fb7466 | 474 | #endif |
925d1c40 | 475 | |
925d1c40 | 476 | /* store ref to file /proc/<pid>/exe symlink points to */ |
90f31d0e | 477 | struct file __rcu *exe_file; |
cddb8a5c AA |
478 | #ifdef CONFIG_MMU_NOTIFIER |
479 | struct mmu_notifier_mm *mmu_notifier_mm; | |
e7a00c45 | 480 | #endif |
e009bb30 | 481 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
e7a00c45 | 482 | pgtable_t pmd_huge_pte; /* protected by page_table_lock */ |
cddb8a5c | 483 | #endif |
6345d24d LT |
484 | #ifdef CONFIG_CPUMASK_OFFSTACK |
485 | struct cpumask cpumask_allocation; | |
cbee9f88 PZ |
486 | #endif |
487 | #ifdef CONFIG_NUMA_BALANCING | |
488 | /* | |
34f0315a MG |
489 | * numa_next_scan is the next time that the PTEs will be marked |
490 | * pte_numa. NUMA hinting faults will gather statistics and migrate | |
491 | * pages to new nodes if necessary. | |
cbee9f88 PZ |
492 | */ |
493 | unsigned long numa_next_scan; | |
494 | ||
6e5fb223 PZ |
495 | /* Restart point for scanning and setting pte_numa */ |
496 | unsigned long numa_scan_offset; | |
497 | ||
cbee9f88 PZ |
498 | /* numa_scan_seq prevents two threads setting pte_numa */ |
499 | int numa_scan_seq; | |
20841405 RR |
500 | #endif |
501 | #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) | |
502 | /* | |
503 | * An operation with batched TLB flushing is going on. Anything that | |
504 | * can move process memory needs to flush the TLB when moving a | |
505 | * PROT_NONE or PROT_NUMA mapped page. | |
506 | */ | |
507 | bool tlb_flush_pending; | |
6345d24d | 508 | #endif |
d4b3b638 | 509 | struct uprobes_state uprobes_state; |
fe3d197f DH |
510 | #ifdef CONFIG_X86_INTEL_MPX |
511 | /* address of the bounds directory */ | |
512 | void __user *bd_addr; | |
513 | #endif | |
5d317b2b NH |
514 | #ifdef CONFIG_HUGETLB_PAGE |
515 | atomic_long_t hugetlb_usage; | |
516 | #endif | |
7ef949d7 | 517 | #ifdef CONFIG_MMU |
ec8d7c14 | 518 | struct work_struct async_put_work; |
7ef949d7 | 519 | #endif |
c92ff1bd MS |
520 | }; |
521 | ||
6345d24d LT |
522 | static inline void mm_init_cpumask(struct mm_struct *mm) |
523 | { | |
524 | #ifdef CONFIG_CPUMASK_OFFSTACK | |
525 | mm->cpu_vm_mask_var = &mm->cpumask_allocation; | |
526 | #endif | |
41f727fd | 527 | cpumask_clear(mm->cpu_vm_mask_var); |
6345d24d LT |
528 | } |
529 | ||
45e575ab | 530 | /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ |
de03c72c KM |
531 | static inline cpumask_t *mm_cpumask(struct mm_struct *mm) |
532 | { | |
533 | return mm->cpu_vm_mask_var; | |
534 | } | |
45e575ab | 535 | |
20841405 RR |
536 | #if defined(CONFIG_NUMA_BALANCING) || defined(CONFIG_COMPACTION) |
537 | /* | |
538 | * Memory barriers to keep this state in sync are graciously provided by | |
539 | * the page table locks, outside of which no page table modifications happen. | |
540 | * The barriers below prevent the compiler from re-ordering the instructions | |
541 | * around the memory barriers that are already present in the code. | |
542 | */ | |
543 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) | |
544 | { | |
545 | barrier(); | |
546 | return mm->tlb_flush_pending; | |
547 | } | |
548 | static inline void set_tlb_flush_pending(struct mm_struct *mm) | |
549 | { | |
550 | mm->tlb_flush_pending = true; | |
af2c1401 MG |
551 | |
552 | /* | |
553 | * Guarantee that the tlb_flush_pending store does not leak into the | |
554 | * critical section updating the page tables | |
555 | */ | |
556 | smp_mb__before_spinlock(); | |
20841405 RR |
557 | } |
558 | /* Clearing is done after a TLB flush, which also provides a barrier. */ | |
559 | static inline void clear_tlb_flush_pending(struct mm_struct *mm) | |
560 | { | |
561 | barrier(); | |
562 | mm->tlb_flush_pending = false; | |
563 | } | |
564 | #else | |
565 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) | |
566 | { | |
567 | return false; | |
568 | } | |
569 | static inline void set_tlb_flush_pending(struct mm_struct *mm) | |
570 | { | |
571 | } | |
572 | static inline void clear_tlb_flush_pending(struct mm_struct *mm) | |
573 | { | |
574 | } | |
575 | #endif | |
576 | ||
f872f540 AL |
577 | struct vm_fault; |
578 | ||
579 | struct vm_special_mapping { | |
580 | const char *name; /* The name, e.g. "[vdso]". */ | |
581 | ||
582 | /* | |
583 | * If .fault is not provided, this points to a | |
584 | * NULL-terminated array of pages that back the special mapping. | |
585 | * | |
586 | * This must not be NULL unless .fault is provided. | |
587 | */ | |
a62c34bd | 588 | struct page **pages; |
f872f540 AL |
589 | |
590 | /* | |
591 | * If non-NULL, then this is called to resolve page faults | |
592 | * on the special mapping. If used, .pages is not checked. | |
593 | */ | |
594 | int (*fault)(const struct vm_special_mapping *sm, | |
595 | struct vm_area_struct *vma, | |
596 | struct vm_fault *vmf); | |
a62c34bd AL |
597 | }; |
598 | ||
d17d8f9d DH |
599 | enum tlb_flush_reason { |
600 | TLB_FLUSH_ON_TASK_SWITCH, | |
601 | TLB_REMOTE_SHOOTDOWN, | |
602 | TLB_LOCAL_SHOOTDOWN, | |
603 | TLB_LOCAL_MM_SHOOTDOWN, | |
5b74283a | 604 | TLB_REMOTE_SEND_IPI, |
d17d8f9d DH |
605 | NR_TLB_FLUSH_REASONS, |
606 | }; | |
607 | ||
bd6dace7 TH |
608 | /* |
609 | * A swap entry has to fit into a "unsigned long", as the entry is hidden | |
610 | * in the "index" field of the swapper address space. | |
611 | */ | |
612 | typedef struct { | |
613 | unsigned long val; | |
614 | } swp_entry_t; | |
615 | ||
5b99cd0e | 616 | #endif /* _LINUX_MM_TYPES_H */ |