]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
5b99cd0e HC |
2 | #ifndef _LINUX_MM_TYPES_H |
3 | #define _LINUX_MM_TYPES_H | |
4 | ||
2e58f173 IM |
5 | #include <linux/mm_types_task.h> |
6 | ||
4f9a58d7 | 7 | #include <linux/auxvec.h> |
5b99cd0e HC |
8 | #include <linux/list.h> |
9 | #include <linux/spinlock.h> | |
c92ff1bd MS |
10 | #include <linux/rbtree.h> |
11 | #include <linux/rwsem.h> | |
12 | #include <linux/completion.h> | |
cddb8a5c | 13 | #include <linux/cpumask.h> |
d4b3b638 | 14 | #include <linux/uprobes.h> |
bbeae5b0 | 15 | #include <linux/page-flags-layout.h> |
ec8d7c14 | 16 | #include <linux/workqueue.h> |
57efa1fe | 17 | #include <linux/seqlock.h> |
2e58f173 | 18 | |
c92ff1bd | 19 | #include <asm/mmu.h> |
5b99cd0e | 20 | |
4f9a58d7 OH |
21 | #ifndef AT_VECTOR_SIZE_ARCH |
22 | #define AT_VECTOR_SIZE_ARCH 0 | |
23 | #endif | |
24 | #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1)) | |
25 | ||
1c8f4220 | 26 | |
5b99cd0e | 27 | struct address_space; |
1306a85a | 28 | struct mem_cgroup; |
5b99cd0e HC |
29 | |
30 | /* | |
31 | * Each physical page in the system has a struct page associated with | |
32 | * it to keep track of whatever it is we are using the page for at the | |
33 | * moment. Note that we have no way to track which tasks are using | |
34 | * a page, though if it is a pagecache page, rmap structures can tell us | |
97b4a671 | 35 | * who is mapping it. |
be50015d | 36 | * |
97b4a671 MW |
37 | * If you allocate the page using alloc_pages(), you can use some of the |
38 | * space in struct page for your own purposes. The five words in the main | |
39 | * union are available, except for bit 0 of the first word which must be | |
40 | * kept clear. Many users use this word to store a pointer to an object | |
41 | * which is guaranteed to be aligned. If you use the same storage as | |
42 | * page->mapping, you must restore it to NULL before freeing the page. | |
be50015d | 43 | * |
97b4a671 MW |
44 | * If your page will not be mapped to userspace, you can also use the four |
45 | * bytes in the mapcount union, but you must call page_mapcount_reset() | |
46 | * before freeing it. | |
47 | * | |
48 | * If you want to use the refcount field, it must be used in such a way | |
49 | * that other CPUs temporarily incrementing and then decrementing the | |
50 | * refcount does not cause problems. On receiving the page from | |
51 | * alloc_pages(), the refcount will be positive. | |
52 | * | |
53 | * If you allocate pages of order > 0, you can use some of the fields | |
54 | * in each subpage, but you may need to restore some of their values | |
55 | * afterwards. | |
fc9bb8c7 | 56 | * |
4cf7c8bf MW |
57 | * SLUB uses cmpxchg_double() to atomically update its freelist and |
58 | * counters. That requires that freelist & counters be adjacent and | |
59 | * double-word aligned. We align all struct pages to double-word | |
60 | * boundaries, and ensure that 'freelist' is aligned within the | |
61 | * struct. | |
5b99cd0e | 62 | */ |
e20df2c6 MW |
63 | #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE |
64 | #define _struct_page_alignment __aligned(2 * sizeof(unsigned long)) | |
65 | #else | |
0dd4da5b | 66 | #define _struct_page_alignment |
7d27a04b | 67 | #endif |
e20df2c6 | 68 | |
5b99cd0e HC |
69 | struct page { |
70 | unsigned long flags; /* Atomic flags, some possibly | |
71 | * updated asynchronously */ | |
b7ccc7f8 | 72 | /* |
4da1984e MW |
73 | * Five words (20/40 bytes) are available in this union. |
74 | * WARNING: bit 0 of the first word is used for PageTail(). That | |
75 | * means the other users of this union MUST NOT use the bit to | |
b7ccc7f8 MW |
76 | * avoid collision and false-positive PageTail(). |
77 | */ | |
8456a648 | 78 | union { |
66a6ffd2 | 79 | struct { /* Page cache and anonymous pages */ |
4da1984e MW |
80 | /** |
81 | * @lru: Pageout list, eg. active_list protected by | |
15b44736 | 82 | * lruvec->lru_lock. Sometimes used as a generic list |
4da1984e MW |
83 | * by the page owner. |
84 | */ | |
85 | struct list_head lru; | |
66a6ffd2 MW |
86 | /* See page-flags.h for PAGE_MAPPING_FLAGS */ |
87 | struct address_space *mapping; | |
88 | pgoff_t index; /* Our offset within mapping. */ | |
89 | /** | |
90 | * @private: Mapping-private opaque data. | |
91 | * Usually used for buffer_heads if PagePrivate. | |
92 | * Used for swp_entry_t if PageSwapCache. | |
93 | * Indicates order in the buddy system if PageBuddy. | |
94 | */ | |
95 | unsigned long private; | |
96 | }; | |
c25fff71 JDB |
97 | struct { /* page_pool used by netstack */ |
98 | /** | |
99 | * @dma_addr: might require a 64-bit value even on | |
100 | * 32-bit architectures. | |
101 | */ | |
102 | dma_addr_t dma_addr; | |
103 | }; | |
66a6ffd2 | 104 | struct { /* slab, slob and slub */ |
4da1984e | 105 | union { |
3e05617c | 106 | struct list_head slab_list; |
4da1984e MW |
107 | struct { /* Partial pages */ |
108 | struct page *next; | |
109 | #ifdef CONFIG_64BIT | |
110 | int pages; /* Nr of pages left */ | |
111 | int pobjects; /* Approximate count */ | |
112 | #else | |
113 | short int pages; | |
114 | short int pobjects; | |
115 | #endif | |
116 | }; | |
117 | }; | |
66a6ffd2 MW |
118 | struct kmem_cache *slab_cache; /* not slob */ |
119 | /* Double-word boundary */ | |
120 | void *freelist; /* first free object */ | |
121 | union { | |
122 | void *s_mem; /* slab: first object */ | |
123 | unsigned long counters; /* SLUB */ | |
124 | struct { /* SLUB */ | |
125 | unsigned inuse:16; | |
126 | unsigned objects:15; | |
127 | unsigned frozen:1; | |
128 | }; | |
129 | }; | |
130 | }; | |
4da1984e MW |
131 | struct { /* Tail pages of compound page */ |
132 | unsigned long compound_head; /* Bit zero is set */ | |
133 | ||
134 | /* First tail page only */ | |
135 | unsigned char compound_dtor; | |
136 | unsigned char compound_order; | |
137 | atomic_t compound_mapcount; | |
1378a5ee | 138 | unsigned int compound_nr; /* 1 << compound_order */ |
4da1984e MW |
139 | }; |
140 | struct { /* Second tail page of compound page */ | |
141 | unsigned long _compound_pad_1; /* compound_head */ | |
47e29d32 | 142 | atomic_t hpage_pinned_refcount; |
87eaceb3 | 143 | /* For both global and memcg */ |
4da1984e MW |
144 | struct list_head deferred_list; |
145 | }; | |
66a6ffd2 | 146 | struct { /* Page table pages */ |
4da1984e MW |
147 | unsigned long _pt_pad_1; /* compound_head */ |
148 | pgtable_t pmd_huge_pte; /* protected by page->ptl */ | |
66a6ffd2 | 149 | unsigned long _pt_pad_2; /* mapping */ |
4231aba0 NP |
150 | union { |
151 | struct mm_struct *pt_mm; /* x86 pgds only */ | |
152 | atomic_t pt_frag_refcount; /* powerpc */ | |
153 | }; | |
7d27a04b | 154 | #if ALLOC_SPLIT_PTLOCKS |
66a6ffd2 | 155 | spinlock_t *ptl; |
7d27a04b | 156 | #else |
66a6ffd2 | 157 | spinlock_t ptl; |
7d27a04b | 158 | #endif |
7d27a04b | 159 | }; |
50e7fbc3 MW |
160 | struct { /* ZONE_DEVICE pages */ |
161 | /** @pgmap: Points to the hosting device page map. */ | |
162 | struct dev_pagemap *pgmap; | |
8a164fef | 163 | void *zone_device_data; |
76470ccd RC |
164 | /* |
165 | * ZONE_DEVICE private pages are counted as being | |
166 | * mapped so the next 3 words hold the mapping, index, | |
167 | * and private fields from the source anonymous or | |
168 | * page cache page while the page is migrated to device | |
169 | * private memory. | |
170 | * ZONE_DEVICE MEMORY_DEVICE_FS_DAX pages also | |
171 | * use the mapping, index, and private fields when | |
172 | * pmem backed DAX files are mapped. | |
173 | */ | |
50e7fbc3 | 174 | }; |
4da1984e MW |
175 | |
176 | /** @rcu_head: You can use this to free a page by RCU. */ | |
177 | struct rcu_head rcu_head; | |
7d27a04b MW |
178 | }; |
179 | ||
b21999da MW |
180 | union { /* This union is 4 bytes in size. */ |
181 | /* | |
182 | * If the page can be mapped to userspace, encodes the number | |
183 | * of times this page is referenced by a page table. | |
184 | */ | |
185 | atomic_t _mapcount; | |
186 | ||
6e292b9b MW |
187 | /* |
188 | * If the page is neither PageSlab nor mappable to userspace, | |
189 | * the value stored here may help determine what this page | |
190 | * is used for. See page-flags.h for a list of page types | |
191 | * which are currently stored here. | |
192 | */ | |
193 | unsigned int page_type; | |
194 | ||
ca9c88c7 | 195 | unsigned int active; /* SLAB */ |
ca9c88c7 | 196 | int units; /* SLOB */ |
81819f0f | 197 | }; |
fc9bb8c7 | 198 | |
b21999da MW |
199 | /* Usage count. *DO NOT USE DIRECTLY*. See page_ref.h */ |
200 | atomic_t _refcount; | |
201 | ||
1306a85a | 202 | #ifdef CONFIG_MEMCG |
bcfe06bf | 203 | unsigned long memcg_data; |
1306a85a JW |
204 | #endif |
205 | ||
5b99cd0e HC |
206 | /* |
207 | * On machines where all RAM is mapped into kernel address space, | |
208 | * we can simply calculate the virtual address. On machines with | |
209 | * highmem some memory is mapped into kernel virtual memory | |
210 | * dynamically, so we need a place to store that address. | |
211 | * Note that this field could be 16 bits on x86 ... ;) | |
212 | * | |
213 | * Architectures with slow multiplication can define | |
214 | * WANT_PAGE_VIRTUAL in asm/page.h | |
215 | */ | |
216 | #if defined(WANT_PAGE_VIRTUAL) | |
217 | void *virtual; /* Kernel virtual address (NULL if | |
218 | not kmapped, ie. highmem) */ | |
219 | #endif /* WANT_PAGE_VIRTUAL */ | |
dfec072e | 220 | |
90572890 PZ |
221 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
222 | int _last_cpupid; | |
57e0a030 | 223 | #endif |
e20df2c6 | 224 | } _struct_page_alignment; |
5b99cd0e | 225 | |
169226f7 YS |
226 | static inline atomic_t *compound_mapcount_ptr(struct page *page) |
227 | { | |
228 | return &page[1].compound_mapcount; | |
229 | } | |
230 | ||
47e29d32 JH |
231 | static inline atomic_t *compound_pincount_ptr(struct page *page) |
232 | { | |
233 | return &page[2].hpage_pinned_refcount; | |
234 | } | |
235 | ||
d1402fc7 LG |
236 | /* |
237 | * Used for sizing the vmemmap region on some architectures | |
238 | */ | |
239 | #define STRUCT_PAGE_MAX_SHIFT (order_base_2(sizeof(struct page))) | |
240 | ||
b63ae8ca AD |
241 | #define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK) |
242 | #define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE) | |
243 | ||
b03641af | 244 | #define page_private(page) ((page)->private) |
60e65a6f GJ |
245 | |
246 | static inline void set_page_private(struct page *page, unsigned long private) | |
247 | { | |
248 | page->private = private; | |
249 | } | |
b03641af | 250 | |
b63ae8ca AD |
251 | struct page_frag_cache { |
252 | void * va; | |
253 | #if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE) | |
254 | __u16 offset; | |
255 | __u16 size; | |
256 | #else | |
257 | __u32 offset; | |
258 | #endif | |
259 | /* we maintain a pagecount bias, so that we dont dirty cache line | |
0139aa7b | 260 | * containing page->_refcount every time we allocate a fragment. |
b63ae8ca AD |
261 | */ |
262 | unsigned int pagecnt_bias; | |
263 | bool pfmemalloc; | |
264 | }; | |
265 | ||
64b990d2 | 266 | typedef unsigned long vm_flags_t; |
ca16d140 | 267 | |
8feae131 DH |
268 | /* |
269 | * A region containing a mapping of a non-memory backed file under NOMMU | |
270 | * conditions. These are held in a global tree and are pinned by the VMAs that | |
271 | * map parts of them. | |
272 | */ | |
273 | struct vm_region { | |
274 | struct rb_node vm_rb; /* link in global region tree */ | |
ca16d140 | 275 | vm_flags_t vm_flags; /* VMA vm_flags */ |
8feae131 DH |
276 | unsigned long vm_start; /* start address of region */ |
277 | unsigned long vm_end; /* region initialised to here */ | |
dd8632a1 | 278 | unsigned long vm_top; /* region allocated to here */ |
8feae131 DH |
279 | unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ |
280 | struct file *vm_file; /* the backing file or NULL */ | |
281 | ||
1e2ae599 | 282 | int vm_usage; /* region usage count (access under nommu_region_sem) */ |
cfe79c00 MF |
283 | bool vm_icache_flushed : 1; /* true if the icache has been flushed for |
284 | * this region */ | |
8feae131 DH |
285 | }; |
286 | ||
745f234b AA |
287 | #ifdef CONFIG_USERFAULTFD |
288 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, }) | |
289 | struct vm_userfaultfd_ctx { | |
290 | struct userfaultfd_ctx *ctx; | |
291 | }; | |
292 | #else /* CONFIG_USERFAULTFD */ | |
293 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {}) | |
294 | struct vm_userfaultfd_ctx {}; | |
295 | #endif /* CONFIG_USERFAULTFD */ | |
296 | ||
c92ff1bd | 297 | /* |
552657b7 | 298 | * This struct describes a virtual memory area. There is one of these |
299 | * per VM-area/task. A VM area is any part of the process virtual memory | |
c92ff1bd MS |
300 | * space that has a special rule for the page-fault handlers (ie a shared |
301 | * library, the executable area etc). | |
302 | */ | |
303 | struct vm_area_struct { | |
e4c6bfd2 RR |
304 | /* The first cache line has the info for VMA tree walking. */ |
305 | ||
c92ff1bd MS |
306 | unsigned long vm_start; /* Our start address within vm_mm. */ |
307 | unsigned long vm_end; /* The first byte after our end address | |
308 | within vm_mm. */ | |
309 | ||
310 | /* linked list of VM areas per task, sorted by address */ | |
297c5eee | 311 | struct vm_area_struct *vm_next, *vm_prev; |
c92ff1bd | 312 | |
c92ff1bd MS |
313 | struct rb_node vm_rb; |
314 | ||
d3737187 ML |
315 | /* |
316 | * Largest free memory gap in bytes to the left of this VMA. | |
317 | * Either between this VMA and vma->vm_prev, or between one of the | |
318 | * VMAs below us in the VMA rbtree and its ->vm_prev. This helps | |
319 | * get_unmapped_area find a free area of the right size. | |
320 | */ | |
321 | unsigned long rb_subtree_gap; | |
322 | ||
e4c6bfd2 RR |
323 | /* Second cache line starts here. */ |
324 | ||
325 | struct mm_struct *vm_mm; /* The address space we belong to. */ | |
574c5b3d TH |
326 | |
327 | /* | |
328 | * Access permissions of this VMA. | |
329 | * See vmf_insert_mixed_prot() for discussion. | |
330 | */ | |
331 | pgprot_t vm_page_prot; | |
e4c6bfd2 RR |
332 | unsigned long vm_flags; /* Flags, see mm.h. */ |
333 | ||
c92ff1bd MS |
334 | /* |
335 | * For areas with an address space and backing store, | |
27ba0644 | 336 | * linkage into the address_space->i_mmap interval tree. |
c92ff1bd | 337 | */ |
ac51b934 KS |
338 | struct { |
339 | struct rb_node rb; | |
340 | unsigned long rb_subtree_last; | |
c92ff1bd MS |
341 | } shared; |
342 | ||
343 | /* | |
344 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma | |
345 | * list, after a COW of one of the file pages. A MAP_SHARED vma | |
346 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack | |
347 | * or brk vma (with NULL file) can only be in an anon_vma list. | |
348 | */ | |
c1e8d7c6 | 349 | struct list_head anon_vma_chain; /* Serialized by mmap_lock & |
5beb4930 | 350 | * page_table_lock */ |
c92ff1bd MS |
351 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
352 | ||
353 | /* Function pointers to deal with this struct. */ | |
f0f37e2f | 354 | const struct vm_operations_struct *vm_ops; |
c92ff1bd MS |
355 | |
356 | /* Information about our backing store: */ | |
357 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE | |
ea1754a0 | 358 | units */ |
c92ff1bd MS |
359 | struct file * vm_file; /* File we map to (can be NULL). */ |
360 | void * vm_private_data; /* was vm_pte (shared mem) */ | |
c92ff1bd | 361 | |
219f8a2e | 362 | #ifdef CONFIG_SWAP |
ec560175 | 363 | atomic_long_t swap_readahead_info; |
219f8a2e | 364 | #endif |
c92ff1bd | 365 | #ifndef CONFIG_MMU |
8feae131 | 366 | struct vm_region *vm_region; /* NOMMU mapping region */ |
c92ff1bd MS |
367 | #endif |
368 | #ifdef CONFIG_NUMA | |
369 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ | |
370 | #endif | |
745f234b | 371 | struct vm_userfaultfd_ctx vm_userfaultfd_ctx; |
3859a271 | 372 | } __randomize_layout; |
c92ff1bd | 373 | |
b564daf8 ON |
374 | struct core_thread { |
375 | struct task_struct *task; | |
376 | struct core_thread *next; | |
377 | }; | |
378 | ||
32ecb1f2 | 379 | struct core_state { |
c5f1cc8c | 380 | atomic_t nr_threads; |
b564daf8 | 381 | struct core_thread dumper; |
32ecb1f2 ON |
382 | struct completion startup; |
383 | }; | |
384 | ||
db446a08 | 385 | struct kioctx_table; |
c92ff1bd | 386 | struct mm_struct { |
c1a2f7f0 RR |
387 | struct { |
388 | struct vm_area_struct *mmap; /* list of VMAs */ | |
389 | struct rb_root mm_rb; | |
7a9cdebd | 390 | u64 vmacache_seqnum; /* per-thread vmacache */ |
efc1a3b1 | 391 | #ifdef CONFIG_MMU |
c1a2f7f0 | 392 | unsigned long (*get_unmapped_area) (struct file *filp, |
c92ff1bd MS |
393 | unsigned long addr, unsigned long len, |
394 | unsigned long pgoff, unsigned long flags); | |
efc1a3b1 | 395 | #endif |
c1a2f7f0 RR |
396 | unsigned long mmap_base; /* base of mmap area */ |
397 | unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */ | |
1b028f78 | 398 | #ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES |
c1a2f7f0 RR |
399 | /* Base adresses for compatible mmap() */ |
400 | unsigned long mmap_compat_base; | |
401 | unsigned long mmap_compat_legacy_base; | |
1b028f78 | 402 | #endif |
c1a2f7f0 RR |
403 | unsigned long task_size; /* size of task vm space */ |
404 | unsigned long highest_vm_end; /* highest vma end address */ | |
405 | pgd_t * pgd; | |
406 | ||
227a4aad MD |
407 | #ifdef CONFIG_MEMBARRIER |
408 | /** | |
409 | * @membarrier_state: Flags controlling membarrier behavior. | |
410 | * | |
411 | * This field is close to @pgd to hopefully fit in the same | |
412 | * cache-line, which needs to be touched by switch_mm(). | |
413 | */ | |
414 | atomic_t membarrier_state; | |
415 | #endif | |
416 | ||
c1a2f7f0 RR |
417 | /** |
418 | * @mm_users: The number of users including userspace. | |
419 | * | |
420 | * Use mmget()/mmget_not_zero()/mmput() to modify. When this | |
421 | * drops to 0 (i.e. when the task exits and there are no other | |
422 | * temporary reference holders), we also release a reference on | |
423 | * @mm_count (which may then free the &struct mm_struct if | |
424 | * @mm_count also drops to 0). | |
425 | */ | |
426 | atomic_t mm_users; | |
427 | ||
428 | /** | |
429 | * @mm_count: The number of references to &struct mm_struct | |
430 | * (@mm_users count as 1). | |
431 | * | |
432 | * Use mmgrab()/mmdrop() to modify. When this drops to 0, the | |
433 | * &struct mm_struct is freed. | |
434 | */ | |
435 | atomic_t mm_count; | |
b279ddc3 | 436 | |
008cfe44 PX |
437 | /** |
438 | * @has_pinned: Whether this mm has pinned any pages. This can | |
439 | * be either replaced in the future by @pinned_vm when it | |
440 | * becomes stable, or grow into a counter on its own. We're | |
441 | * aggresive on this bit now - even if the pinned pages were | |
442 | * unpinned later on, we'll still keep this bit set for the | |
443 | * lifecycle of this mm just for simplicity. | |
444 | */ | |
445 | atomic_t has_pinned; | |
446 | ||
57efa1fe JG |
447 | /** |
448 | * @write_protect_seq: Locked when any thread is write | |
449 | * protecting pages mapped by this mm to enforce a later COW, | |
450 | * for instance during page table copying for fork(). | |
451 | */ | |
452 | seqcount_t write_protect_seq; | |
453 | ||
c4812909 | 454 | #ifdef CONFIG_MMU |
c1a2f7f0 | 455 | atomic_long_t pgtables_bytes; /* PTE page table pages */ |
5a3fbef3 | 456 | #endif |
c1a2f7f0 | 457 | int map_count; /* number of VMAs */ |
481b4bb5 | 458 | |
c1a2f7f0 RR |
459 | spinlock_t page_table_lock; /* Protects page tables and some |
460 | * counters | |
461 | */ | |
da1c55f1 | 462 | struct rw_semaphore mmap_lock; |
c92ff1bd | 463 | |
c1a2f7f0 RR |
464 | struct list_head mmlist; /* List of maybe swapped mm's. These |
465 | * are globally strung together off | |
466 | * init_mm.mmlist, and are protected | |
467 | * by mmlist_lock | |
468 | */ | |
c92ff1bd | 469 | |
c92ff1bd | 470 | |
c1a2f7f0 RR |
471 | unsigned long hiwater_rss; /* High-watermark of RSS usage */ |
472 | unsigned long hiwater_vm; /* High-water virtual memory usage */ | |
c92ff1bd | 473 | |
c1a2f7f0 RR |
474 | unsigned long total_vm; /* Total pages mapped */ |
475 | unsigned long locked_vm; /* Pages that have PG_mlocked set */ | |
70f8a3ca | 476 | atomic64_t pinned_vm; /* Refcount permanently increased */ |
c1a2f7f0 RR |
477 | unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */ |
478 | unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */ | |
479 | unsigned long stack_vm; /* VM_STACK */ | |
480 | unsigned long def_flags; | |
88aa7cc6 | 481 | |
c1a2f7f0 RR |
482 | spinlock_t arg_lock; /* protect the below fields */ |
483 | unsigned long start_code, end_code, start_data, end_data; | |
484 | unsigned long start_brk, brk, start_stack; | |
485 | unsigned long arg_start, arg_end, env_start, env_end; | |
c92ff1bd | 486 | |
c1a2f7f0 | 487 | unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ |
c92ff1bd | 488 | |
c1a2f7f0 RR |
489 | /* |
490 | * Special counters, in some configurations protected by the | |
491 | * page_table_lock, in other configurations by being atomic. | |
492 | */ | |
493 | struct mm_rss_stat rss_stat; | |
801460d0 | 494 | |
c1a2f7f0 | 495 | struct linux_binfmt *binfmt; |
6345d24d | 496 | |
c1a2f7f0 RR |
497 | /* Architecture-specific MM context */ |
498 | mm_context_t context; | |
c92ff1bd | 499 | |
c1a2f7f0 | 500 | unsigned long flags; /* Must use atomic bitops to access */ |
c92ff1bd | 501 | |
c1a2f7f0 | 502 | struct core_state *core_state; /* coredumping support */ |
227a4aad | 503 | |
858f0993 | 504 | #ifdef CONFIG_AIO |
c1a2f7f0 RR |
505 | spinlock_t ioctx_lock; |
506 | struct kioctx_table __rcu *ioctx_table; | |
858f0993 | 507 | #endif |
f98bafa0 | 508 | #ifdef CONFIG_MEMCG |
c1a2f7f0 RR |
509 | /* |
510 | * "owner" points to a task that is regarded as the canonical | |
511 | * user/owner of this mm. All of the following must be true in | |
512 | * order for it to be changed: | |
513 | * | |
514 | * current == mm->owner | |
515 | * current->mm != mm | |
516 | * new_owner->mm == mm | |
517 | * new_owner->alloc_lock is held | |
518 | */ | |
519 | struct task_struct __rcu *owner; | |
78fb7466 | 520 | #endif |
c1a2f7f0 | 521 | struct user_namespace *user_ns; |
925d1c40 | 522 | |
c1a2f7f0 RR |
523 | /* store ref to file /proc/<pid>/exe symlink points to */ |
524 | struct file __rcu *exe_file; | |
cddb8a5c | 525 | #ifdef CONFIG_MMU_NOTIFIER |
984cfe4e | 526 | struct mmu_notifier_subscriptions *notifier_subscriptions; |
e7a00c45 | 527 | #endif |
e009bb30 | 528 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
c1a2f7f0 | 529 | pgtable_t pmd_huge_pte; /* protected by page_table_lock */ |
cbee9f88 PZ |
530 | #endif |
531 | #ifdef CONFIG_NUMA_BALANCING | |
c1a2f7f0 RR |
532 | /* |
533 | * numa_next_scan is the next time that the PTEs will be marked | |
534 | * pte_numa. NUMA hinting faults will gather statistics and | |
535 | * migrate pages to new nodes if necessary. | |
536 | */ | |
537 | unsigned long numa_next_scan; | |
cbee9f88 | 538 | |
c1a2f7f0 RR |
539 | /* Restart point for scanning and setting pte_numa */ |
540 | unsigned long numa_scan_offset; | |
6e5fb223 | 541 | |
c1a2f7f0 RR |
542 | /* numa_scan_seq prevents two threads setting pte_numa */ |
543 | int numa_scan_seq; | |
20841405 | 544 | #endif |
c1a2f7f0 RR |
545 | /* |
546 | * An operation with batched TLB flushing is going on. Anything | |
547 | * that can move process memory needs to flush the TLB when | |
548 | * moving a PROT_NONE or PROT_NUMA mapped page. | |
549 | */ | |
550 | atomic_t tlb_flush_pending; | |
3ea27719 | 551 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
c1a2f7f0 RR |
552 | /* See flush_tlb_batched_pending() */ |
553 | bool tlb_flush_batched; | |
6345d24d | 554 | #endif |
c1a2f7f0 | 555 | struct uprobes_state uprobes_state; |
5d317b2b | 556 | #ifdef CONFIG_HUGETLB_PAGE |
c1a2f7f0 | 557 | atomic_long_t hugetlb_usage; |
5d317b2b | 558 | #endif |
c1a2f7f0 | 559 | struct work_struct async_put_work; |
52ad9bc6 FY |
560 | |
561 | #ifdef CONFIG_IOMMU_SUPPORT | |
562 | u32 pasid; | |
563 | #endif | |
c1a2f7f0 RR |
564 | } __randomize_layout; |
565 | ||
566 | /* | |
567 | * The mm_cpumask needs to be at the end of mm_struct, because it | |
568 | * is dynamically sized based on nr_cpu_ids. | |
569 | */ | |
570 | unsigned long cpu_bitmap[]; | |
571 | }; | |
c92ff1bd | 572 | |
abe722a1 IM |
573 | extern struct mm_struct init_mm; |
574 | ||
c1a2f7f0 | 575 | /* Pointer magic because the dynamic array size confuses some compilers. */ |
6345d24d LT |
576 | static inline void mm_init_cpumask(struct mm_struct *mm) |
577 | { | |
c1a2f7f0 RR |
578 | unsigned long cpu_bitmap = (unsigned long)mm; |
579 | ||
580 | cpu_bitmap += offsetof(struct mm_struct, cpu_bitmap); | |
581 | cpumask_clear((struct cpumask *)cpu_bitmap); | |
6345d24d LT |
582 | } |
583 | ||
45e575ab | 584 | /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ |
de03c72c KM |
585 | static inline cpumask_t *mm_cpumask(struct mm_struct *mm) |
586 | { | |
c1a2f7f0 | 587 | return (struct cpumask *)&mm->cpu_bitmap; |
de03c72c | 588 | } |
45e575ab | 589 | |
56236a59 | 590 | struct mmu_gather; |
a72afd87 | 591 | extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm); |
d8b45053 | 592 | extern void tlb_gather_mmu_fullmm(struct mmu_gather *tlb, struct mm_struct *mm); |
ae8eba8b | 593 | extern void tlb_finish_mmu(struct mmu_gather *tlb); |
56236a59 | 594 | |
16af97dc | 595 | static inline void init_tlb_flush_pending(struct mm_struct *mm) |
20841405 | 596 | { |
16af97dc | 597 | atomic_set(&mm->tlb_flush_pending, 0); |
20841405 | 598 | } |
16af97dc NA |
599 | |
600 | static inline void inc_tlb_flush_pending(struct mm_struct *mm) | |
20841405 | 601 | { |
16af97dc | 602 | atomic_inc(&mm->tlb_flush_pending); |
af2c1401 | 603 | /* |
8b1b436d PZ |
604 | * The only time this value is relevant is when there are indeed pages |
605 | * to flush. And we'll only flush pages after changing them, which | |
606 | * requires the PTL. | |
607 | * | |
608 | * So the ordering here is: | |
609 | * | |
040cca3a | 610 | * atomic_inc(&mm->tlb_flush_pending); |
8b1b436d PZ |
611 | * spin_lock(&ptl); |
612 | * ... | |
613 | * set_pte_at(); | |
614 | * spin_unlock(&ptl); | |
615 | * | |
616 | * spin_lock(&ptl) | |
617 | * mm_tlb_flush_pending(); | |
618 | * .... | |
619 | * spin_unlock(&ptl); | |
620 | * | |
621 | * flush_tlb_range(); | |
040cca3a | 622 | * atomic_dec(&mm->tlb_flush_pending); |
8b1b436d | 623 | * |
0e709703 PZ |
624 | * Where the increment if constrained by the PTL unlock, it thus |
625 | * ensures that the increment is visible if the PTE modification is | |
626 | * visible. After all, if there is no PTE modification, nobody cares | |
627 | * about TLB flushes either. | |
628 | * | |
629 | * This very much relies on users (mm_tlb_flush_pending() and | |
630 | * mm_tlb_flush_nested()) only caring about _specific_ PTEs (and | |
631 | * therefore specific PTLs), because with SPLIT_PTE_PTLOCKS and RCpc | |
632 | * locks (PPC) the unlock of one doesn't order against the lock of | |
633 | * another PTL. | |
634 | * | |
635 | * The decrement is ordered by the flush_tlb_range(), such that | |
636 | * mm_tlb_flush_pending() will not return false unless all flushes have | |
637 | * completed. | |
af2c1401 | 638 | */ |
20841405 | 639 | } |
16af97dc | 640 | |
16af97dc | 641 | static inline void dec_tlb_flush_pending(struct mm_struct *mm) |
20841405 | 642 | { |
0a2c4048 | 643 | /* |
0e709703 PZ |
644 | * See inc_tlb_flush_pending(). |
645 | * | |
646 | * This cannot be smp_mb__before_atomic() because smp_mb() simply does | |
647 | * not order against TLB invalidate completion, which is what we need. | |
648 | * | |
649 | * Therefore we must rely on tlb_flush_*() to guarantee order. | |
0a2c4048 | 650 | */ |
16af97dc | 651 | atomic_dec(&mm->tlb_flush_pending); |
20841405 | 652 | } |
20841405 | 653 | |
0e709703 PZ |
654 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) |
655 | { | |
656 | /* | |
657 | * Must be called after having acquired the PTL; orders against that | |
658 | * PTLs release and therefore ensures that if we observe the modified | |
659 | * PTE we must also observe the increment from inc_tlb_flush_pending(). | |
660 | * | |
661 | * That is, it only guarantees to return true if there is a flush | |
662 | * pending for _this_ PTL. | |
663 | */ | |
664 | return atomic_read(&mm->tlb_flush_pending); | |
665 | } | |
666 | ||
667 | static inline bool mm_tlb_flush_nested(struct mm_struct *mm) | |
668 | { | |
669 | /* | |
670 | * Similar to mm_tlb_flush_pending(), we must have acquired the PTL | |
671 | * for which there is a TLB flush pending in order to guarantee | |
672 | * we've seen both that PTE modification and the increment. | |
673 | * | |
674 | * (no requirement on actually still holding the PTL, that is irrelevant) | |
675 | */ | |
676 | return atomic_read(&mm->tlb_flush_pending) > 1; | |
677 | } | |
678 | ||
f872f540 AL |
679 | struct vm_fault; |
680 | ||
3d353901 SJ |
681 | /** |
682 | * typedef vm_fault_t - Return type for page fault handlers. | |
683 | * | |
684 | * Page fault handlers return a bitmask of %VM_FAULT values. | |
685 | */ | |
686 | typedef __bitwise unsigned int vm_fault_t; | |
687 | ||
688 | /** | |
689 | * enum vm_fault_reason - Page fault handlers return a bitmask of | |
690 | * these values to tell the core VM what happened when handling the | |
691 | * fault. Used to decide whether a process gets delivered SIGBUS or | |
692 | * just gets major/minor fault counters bumped up. | |
693 | * | |
694 | * @VM_FAULT_OOM: Out Of Memory | |
695 | * @VM_FAULT_SIGBUS: Bad access | |
696 | * @VM_FAULT_MAJOR: Page read from storage | |
697 | * @VM_FAULT_WRITE: Special case for get_user_pages | |
698 | * @VM_FAULT_HWPOISON: Hit poisoned small page | |
699 | * @VM_FAULT_HWPOISON_LARGE: Hit poisoned large page. Index encoded | |
700 | * in upper bits | |
701 | * @VM_FAULT_SIGSEGV: segmentation fault | |
702 | * @VM_FAULT_NOPAGE: ->fault installed the pte, not return page | |
703 | * @VM_FAULT_LOCKED: ->fault locked the returned page | |
704 | * @VM_FAULT_RETRY: ->fault blocked, must retry | |
705 | * @VM_FAULT_FALLBACK: huge page fault failed, fall back to small | |
706 | * @VM_FAULT_DONE_COW: ->fault has fully handled COW | |
707 | * @VM_FAULT_NEEDDSYNC: ->fault did not modify page tables and needs | |
708 | * fsync() to complete (for synchronous page faults | |
709 | * in DAX) | |
710 | * @VM_FAULT_HINDEX_MASK: mask HINDEX value | |
711 | * | |
712 | */ | |
713 | enum vm_fault_reason { | |
714 | VM_FAULT_OOM = (__force vm_fault_t)0x000001, | |
715 | VM_FAULT_SIGBUS = (__force vm_fault_t)0x000002, | |
716 | VM_FAULT_MAJOR = (__force vm_fault_t)0x000004, | |
717 | VM_FAULT_WRITE = (__force vm_fault_t)0x000008, | |
718 | VM_FAULT_HWPOISON = (__force vm_fault_t)0x000010, | |
719 | VM_FAULT_HWPOISON_LARGE = (__force vm_fault_t)0x000020, | |
720 | VM_FAULT_SIGSEGV = (__force vm_fault_t)0x000040, | |
721 | VM_FAULT_NOPAGE = (__force vm_fault_t)0x000100, | |
722 | VM_FAULT_LOCKED = (__force vm_fault_t)0x000200, | |
723 | VM_FAULT_RETRY = (__force vm_fault_t)0x000400, | |
724 | VM_FAULT_FALLBACK = (__force vm_fault_t)0x000800, | |
725 | VM_FAULT_DONE_COW = (__force vm_fault_t)0x001000, | |
726 | VM_FAULT_NEEDDSYNC = (__force vm_fault_t)0x002000, | |
727 | VM_FAULT_HINDEX_MASK = (__force vm_fault_t)0x0f0000, | |
728 | }; | |
729 | ||
730 | /* Encode hstate index for a hwpoisoned large page */ | |
731 | #define VM_FAULT_SET_HINDEX(x) ((__force vm_fault_t)((x) << 16)) | |
fcae96ff | 732 | #define VM_FAULT_GET_HINDEX(x) (((__force unsigned int)(x) >> 16) & 0xf) |
3d353901 SJ |
733 | |
734 | #define VM_FAULT_ERROR (VM_FAULT_OOM | VM_FAULT_SIGBUS | \ | |
735 | VM_FAULT_SIGSEGV | VM_FAULT_HWPOISON | \ | |
736 | VM_FAULT_HWPOISON_LARGE | VM_FAULT_FALLBACK) | |
737 | ||
738 | #define VM_FAULT_RESULT_TRACE \ | |
739 | { VM_FAULT_OOM, "OOM" }, \ | |
740 | { VM_FAULT_SIGBUS, "SIGBUS" }, \ | |
741 | { VM_FAULT_MAJOR, "MAJOR" }, \ | |
742 | { VM_FAULT_WRITE, "WRITE" }, \ | |
743 | { VM_FAULT_HWPOISON, "HWPOISON" }, \ | |
744 | { VM_FAULT_HWPOISON_LARGE, "HWPOISON_LARGE" }, \ | |
745 | { VM_FAULT_SIGSEGV, "SIGSEGV" }, \ | |
746 | { VM_FAULT_NOPAGE, "NOPAGE" }, \ | |
747 | { VM_FAULT_LOCKED, "LOCKED" }, \ | |
748 | { VM_FAULT_RETRY, "RETRY" }, \ | |
749 | { VM_FAULT_FALLBACK, "FALLBACK" }, \ | |
750 | { VM_FAULT_DONE_COW, "DONE_COW" }, \ | |
751 | { VM_FAULT_NEEDDSYNC, "NEEDDSYNC" } | |
752 | ||
f872f540 AL |
753 | struct vm_special_mapping { |
754 | const char *name; /* The name, e.g. "[vdso]". */ | |
755 | ||
756 | /* | |
757 | * If .fault is not provided, this points to a | |
758 | * NULL-terminated array of pages that back the special mapping. | |
759 | * | |
760 | * This must not be NULL unless .fault is provided. | |
761 | */ | |
a62c34bd | 762 | struct page **pages; |
f872f540 AL |
763 | |
764 | /* | |
765 | * If non-NULL, then this is called to resolve page faults | |
766 | * on the special mapping. If used, .pages is not checked. | |
767 | */ | |
b3ec9f33 SJ |
768 | vm_fault_t (*fault)(const struct vm_special_mapping *sm, |
769 | struct vm_area_struct *vma, | |
770 | struct vm_fault *vmf); | |
b059a453 DS |
771 | |
772 | int (*mremap)(const struct vm_special_mapping *sm, | |
773 | struct vm_area_struct *new_vma); | |
a62c34bd AL |
774 | }; |
775 | ||
d17d8f9d DH |
776 | enum tlb_flush_reason { |
777 | TLB_FLUSH_ON_TASK_SWITCH, | |
778 | TLB_REMOTE_SHOOTDOWN, | |
779 | TLB_LOCAL_SHOOTDOWN, | |
780 | TLB_LOCAL_MM_SHOOTDOWN, | |
5b74283a | 781 | TLB_REMOTE_SEND_IPI, |
d17d8f9d DH |
782 | NR_TLB_FLUSH_REASONS, |
783 | }; | |
784 | ||
bd6dace7 TH |
785 | /* |
786 | * A swap entry has to fit into a "unsigned long", as the entry is hidden | |
787 | * in the "index" field of the swapper address space. | |
788 | */ | |
789 | typedef struct { | |
790 | unsigned long val; | |
791 | } swp_entry_t; | |
792 | ||
5b99cd0e | 793 | #endif /* _LINUX_MM_TYPES_H */ |