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> |
2e58f173 | 17 | |
c92ff1bd | 18 | #include <asm/mmu.h> |
5b99cd0e | 19 | |
4f9a58d7 OH |
20 | #ifndef AT_VECTOR_SIZE_ARCH |
21 | #define AT_VECTOR_SIZE_ARCH 0 | |
22 | #endif | |
23 | #define AT_VECTOR_SIZE (2*(AT_VECTOR_SIZE_ARCH + AT_VECTOR_SIZE_BASE + 1)) | |
24 | ||
1c8f4220 SJ |
25 | typedef int vm_fault_t; |
26 | ||
5b99cd0e | 27 | struct address_space; |
1306a85a | 28 | struct mem_cgroup; |
133ff0ea | 29 | struct hmm; |
5b99cd0e HC |
30 | |
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 | |
be50015d MW |
36 | * who is mapping it. If you allocate the page using alloc_pages(), you |
37 | * can use some of the space in struct page for your own purposes. | |
38 | * | |
39 | * Pages that were once in the page cache may be found under the RCU lock | |
40 | * even after they have been recycled to a different purpose. The page | |
41 | * cache reads and writes some of the fields in struct page to pin the | |
42 | * page before checking that it's still in the page cache. It is vital | |
43 | * that all users of struct page: | |
44 | * 1. Use the first word as PageFlags. | |
45 | * 2. Clear or preserve bit 0 of page->compound_head. It is used as | |
46 | * PageTail for compound pages, and the page cache must not see false | |
47 | * positives. Some users put a pointer here (guaranteed to be at least | |
48 | * 4-byte aligned), other users avoid using the field altogether. | |
49 | * 3. page->_refcount must either not be used, or must be used in such a | |
50 | * way that other CPUs temporarily incrementing and then decrementing the | |
51 | * refcount does not cause problems. On receiving the page from | |
52 | * alloc_pages(), the refcount will be positive. | |
53 | * 4. Either preserve page->_mapcount or restore it to -1 before freeing it. | |
54 | * | |
55 | * If you allocate pages of order > 0, you can use the fields in the struct | |
56 | * page associated with each page, but bear in mind that the pages may have | |
57 | * been inserted individually into the page cache, so you must use the above | |
58 | * four fields in a compatible way for each struct page. | |
fc9bb8c7 | 59 | * |
4cf7c8bf MW |
60 | * SLUB uses cmpxchg_double() to atomically update its freelist and |
61 | * counters. That requires that freelist & counters be adjacent and | |
62 | * double-word aligned. We align all struct pages to double-word | |
63 | * boundaries, and ensure that 'freelist' is aligned within the | |
64 | * struct. | |
5b99cd0e | 65 | */ |
e20df2c6 MW |
66 | #ifdef CONFIG_HAVE_ALIGNED_STRUCT_PAGE |
67 | #define _struct_page_alignment __aligned(2 * sizeof(unsigned long)) | |
0dd4da5b MW |
68 | #if defined(CONFIG_HAVE_CMPXCHG_DOUBLE) |
69 | #define _slub_counter_t unsigned long | |
e20df2c6 | 70 | #else |
0dd4da5b | 71 | #define _slub_counter_t unsigned int |
e20df2c6 | 72 | #endif |
0dd4da5b MW |
73 | #else /* !CONFIG_HAVE_ALIGNED_STRUCT_PAGE */ |
74 | #define _struct_page_alignment | |
75 | #define _slub_counter_t unsigned int | |
76 | #endif /* !CONFIG_HAVE_ALIGNED_STRUCT_PAGE */ | |
e20df2c6 | 77 | |
5b99cd0e | 78 | struct page { |
fc9bb8c7 | 79 | /* First double word block */ |
5b99cd0e HC |
80 | unsigned long flags; /* Atomic flags, some possibly |
81 | * updated asynchronously */ | |
8456a648 | 82 | union { |
b26435a0 MW |
83 | /* See page-flags.h for the definition of PAGE_MAPPING_FLAGS */ |
84 | struct address_space *mapping; | |
85 | ||
8456a648 | 86 | void *s_mem; /* slab first object */ |
53f9263b | 87 | atomic_t compound_mapcount; /* first tail page */ |
9a982250 | 88 | /* page_deferred_list().next -- second tail page */ |
8456a648 JK |
89 | }; |
90 | ||
fc9bb8c7 | 91 | /* Second double word */ |
99691add VD |
92 | union { |
93 | pgoff_t index; /* Our offset within mapping. */ | |
94 | void *freelist; /* sl[aou]b first free object */ | |
95 | /* page_deferred_list().prev -- second tail page */ | |
96 | }; | |
013e8963 | 97 | |
99691add | 98 | union { |
0dd4da5b | 99 | _slub_counter_t counters; |
ca9c88c7 MW |
100 | unsigned int active; /* SLAB */ |
101 | struct { /* SLUB */ | |
102 | unsigned inuse:16; | |
103 | unsigned objects:15; | |
104 | unsigned frozen:1; | |
105 | }; | |
106 | int units; /* SLOB */ | |
013e8963 | 107 | |
ca9c88c7 MW |
108 | struct { /* Page cache */ |
109 | /* | |
110 | * Count of ptes mapped in mms, to show when | |
111 | * page is mapped & limit reverse map searches. | |
112 | * | |
113 | * Extra information about page type may be | |
114 | * stored here for pages that are never mapped, | |
115 | * in which case the value MUST BE <= -2. | |
116 | * See page-flags.h for more details. | |
117 | */ | |
118 | atomic_t _mapcount; | |
99691add | 119 | |
99691add VD |
120 | /* |
121 | * Usage count, *USE WRAPPER FUNCTION* when manual | |
122 | * accounting. See page_ref.h | |
123 | */ | |
124 | atomic_t _refcount; | |
39b26464 | 125 | }; |
81819f0f | 126 | }; |
fc9bb8c7 | 127 | |
1d798ca3 | 128 | /* |
1d798ca3 KS |
129 | * WARNING: bit 0 of the first word encode PageTail(). That means |
130 | * the rest users of the storage space MUST NOT use the bit to | |
131 | * avoid collision and false-positive PageTail(). | |
132 | */ | |
49e22585 CL |
133 | union { |
134 | struct list_head lru; /* Pageout list, eg. active_list | |
a52633d8 | 135 | * protected by zone_lru_lock ! |
34bf6ef9 DH |
136 | * Can be used as a generic list |
137 | * by the page owner. | |
fc9bb8c7 | 138 | */ |
5c2c2587 DW |
139 | struct dev_pagemap *pgmap; /* ZONE_DEVICE pages are never on an |
140 | * lru or handled by a slab | |
141 | * allocator, this points to the | |
142 | * hosting device page map. | |
143 | */ | |
49e22585 CL |
144 | struct { /* slub per cpu partial pages */ |
145 | struct page *next; /* Next partial slab */ | |
146 | #ifdef CONFIG_64BIT | |
147 | int pages; /* Nr of partial slabs left */ | |
148 | int pobjects; /* Approximate # of objects */ | |
149 | #else | |
150 | short int pages; | |
151 | short int pobjects; | |
152 | #endif | |
153 | }; | |
b8c24c4a | 154 | |
68126702 JK |
155 | struct rcu_head rcu_head; /* Used by SLAB |
156 | * when destroying via RCU | |
157 | */ | |
1d798ca3 | 158 | /* Tail pages of compound page */ |
e4b294c2 | 159 | struct { |
1d798ca3 KS |
160 | unsigned long compound_head; /* If bit zero is set */ |
161 | ||
162 | /* First tail page only */ | |
036e7aa4 MW |
163 | unsigned char compound_dtor; |
164 | unsigned char compound_order; | |
165 | /* two/six bytes available here */ | |
e4b294c2 KS |
166 | }; |
167 | ||
7aa555bf | 168 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && USE_SPLIT_PMD_PTLOCKS |
1d798ca3 KS |
169 | struct { |
170 | unsigned long __pad; /* do not overlay pmd_huge_pte | |
171 | * with compound_head to avoid | |
172 | * possible bit 0 collision. | |
173 | */ | |
174 | pgtable_t pmd_huge_pte; /* protected by page->ptl */ | |
175 | }; | |
7aa555bf | 176 | #endif |
49e22585 | 177 | }; |
fc9bb8c7 | 178 | |
5b99cd0e | 179 | union { |
ab8928b7 MW |
180 | /* |
181 | * Mapping-private opaque data: | |
182 | * Usually used for buffer_heads if PagePrivate | |
183 | * Used for swp_entry_t if PageSwapCache | |
184 | * Indicates order in the buddy system if PageBuddy | |
185 | */ | |
186 | unsigned long private; | |
57c1ffce | 187 | #if USE_SPLIT_PTE_PTLOCKS |
597d795a | 188 | #if ALLOC_SPLIT_PTLOCKS |
539edb58 PZ |
189 | spinlock_t *ptl; |
190 | #else | |
191 | spinlock_t ptl; | |
192 | #endif | |
5b99cd0e | 193 | #endif |
1b4f59e3 | 194 | struct kmem_cache *slab_cache; /* SL[AU]B: Pointer to slab */ |
81819f0f | 195 | }; |
fc9bb8c7 | 196 | |
1306a85a JW |
197 | #ifdef CONFIG_MEMCG |
198 | struct mem_cgroup *mem_cgroup; | |
199 | #endif | |
200 | ||
5b99cd0e HC |
201 | /* |
202 | * On machines where all RAM is mapped into kernel address space, | |
203 | * we can simply calculate the virtual address. On machines with | |
204 | * highmem some memory is mapped into kernel virtual memory | |
205 | * dynamically, so we need a place to store that address. | |
206 | * Note that this field could be 16 bits on x86 ... ;) | |
207 | * | |
208 | * Architectures with slow multiplication can define | |
209 | * WANT_PAGE_VIRTUAL in asm/page.h | |
210 | */ | |
211 | #if defined(WANT_PAGE_VIRTUAL) | |
212 | void *virtual; /* Kernel virtual address (NULL if | |
213 | not kmapped, ie. highmem) */ | |
214 | #endif /* WANT_PAGE_VIRTUAL */ | |
dfec072e | 215 | |
90572890 PZ |
216 | #ifdef LAST_CPUPID_NOT_IN_PAGE_FLAGS |
217 | int _last_cpupid; | |
57e0a030 | 218 | #endif |
e20df2c6 | 219 | } _struct_page_alignment; |
5b99cd0e | 220 | |
b63ae8ca AD |
221 | #define PAGE_FRAG_CACHE_MAX_SIZE __ALIGN_MASK(32768, ~PAGE_MASK) |
222 | #define PAGE_FRAG_CACHE_MAX_ORDER get_order(PAGE_FRAG_CACHE_MAX_SIZE) | |
223 | ||
224 | struct page_frag_cache { | |
225 | void * va; | |
226 | #if (PAGE_SIZE < PAGE_FRAG_CACHE_MAX_SIZE) | |
227 | __u16 offset; | |
228 | __u16 size; | |
229 | #else | |
230 | __u32 offset; | |
231 | #endif | |
232 | /* we maintain a pagecount bias, so that we dont dirty cache line | |
0139aa7b | 233 | * containing page->_refcount every time we allocate a fragment. |
b63ae8ca AD |
234 | */ |
235 | unsigned int pagecnt_bias; | |
236 | bool pfmemalloc; | |
237 | }; | |
238 | ||
64b990d2 | 239 | typedef unsigned long vm_flags_t; |
ca16d140 | 240 | |
8feae131 DH |
241 | /* |
242 | * A region containing a mapping of a non-memory backed file under NOMMU | |
243 | * conditions. These are held in a global tree and are pinned by the VMAs that | |
244 | * map parts of them. | |
245 | */ | |
246 | struct vm_region { | |
247 | struct rb_node vm_rb; /* link in global region tree */ | |
ca16d140 | 248 | vm_flags_t vm_flags; /* VMA vm_flags */ |
8feae131 DH |
249 | unsigned long vm_start; /* start address of region */ |
250 | unsigned long vm_end; /* region initialised to here */ | |
dd8632a1 | 251 | unsigned long vm_top; /* region allocated to here */ |
8feae131 DH |
252 | unsigned long vm_pgoff; /* the offset in vm_file corresponding to vm_start */ |
253 | struct file *vm_file; /* the backing file or NULL */ | |
254 | ||
1e2ae599 | 255 | int vm_usage; /* region usage count (access under nommu_region_sem) */ |
cfe79c00 MF |
256 | bool vm_icache_flushed : 1; /* true if the icache has been flushed for |
257 | * this region */ | |
8feae131 DH |
258 | }; |
259 | ||
745f234b AA |
260 | #ifdef CONFIG_USERFAULTFD |
261 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) { NULL, }) | |
262 | struct vm_userfaultfd_ctx { | |
263 | struct userfaultfd_ctx *ctx; | |
264 | }; | |
265 | #else /* CONFIG_USERFAULTFD */ | |
266 | #define NULL_VM_UFFD_CTX ((struct vm_userfaultfd_ctx) {}) | |
267 | struct vm_userfaultfd_ctx {}; | |
268 | #endif /* CONFIG_USERFAULTFD */ | |
269 | ||
c92ff1bd MS |
270 | /* |
271 | * This struct defines a memory VMM memory area. There is one of these | |
272 | * per VM-area/task. A VM area is any part of the process virtual memory | |
273 | * space that has a special rule for the page-fault handlers (ie a shared | |
274 | * library, the executable area etc). | |
275 | */ | |
276 | struct vm_area_struct { | |
e4c6bfd2 RR |
277 | /* The first cache line has the info for VMA tree walking. */ |
278 | ||
c92ff1bd MS |
279 | unsigned long vm_start; /* Our start address within vm_mm. */ |
280 | unsigned long vm_end; /* The first byte after our end address | |
281 | within vm_mm. */ | |
282 | ||
283 | /* linked list of VM areas per task, sorted by address */ | |
297c5eee | 284 | struct vm_area_struct *vm_next, *vm_prev; |
c92ff1bd | 285 | |
c92ff1bd MS |
286 | struct rb_node vm_rb; |
287 | ||
d3737187 ML |
288 | /* |
289 | * Largest free memory gap in bytes to the left of this VMA. | |
290 | * Either between this VMA and vma->vm_prev, or between one of the | |
291 | * VMAs below us in the VMA rbtree and its ->vm_prev. This helps | |
292 | * get_unmapped_area find a free area of the right size. | |
293 | */ | |
294 | unsigned long rb_subtree_gap; | |
295 | ||
e4c6bfd2 RR |
296 | /* Second cache line starts here. */ |
297 | ||
298 | struct mm_struct *vm_mm; /* The address space we belong to. */ | |
299 | pgprot_t vm_page_prot; /* Access permissions of this VMA. */ | |
300 | unsigned long vm_flags; /* Flags, see mm.h. */ | |
301 | ||
c92ff1bd MS |
302 | /* |
303 | * For areas with an address space and backing store, | |
27ba0644 | 304 | * linkage into the address_space->i_mmap interval tree. |
c92ff1bd | 305 | */ |
ac51b934 KS |
306 | struct { |
307 | struct rb_node rb; | |
308 | unsigned long rb_subtree_last; | |
c92ff1bd MS |
309 | } shared; |
310 | ||
311 | /* | |
312 | * A file's MAP_PRIVATE vma can be in both i_mmap tree and anon_vma | |
313 | * list, after a COW of one of the file pages. A MAP_SHARED vma | |
314 | * can only be in the i_mmap tree. An anonymous MAP_PRIVATE, stack | |
315 | * or brk vma (with NULL file) can only be in an anon_vma list. | |
316 | */ | |
5beb4930 RR |
317 | struct list_head anon_vma_chain; /* Serialized by mmap_sem & |
318 | * page_table_lock */ | |
c92ff1bd MS |
319 | struct anon_vma *anon_vma; /* Serialized by page_table_lock */ |
320 | ||
321 | /* Function pointers to deal with this struct. */ | |
f0f37e2f | 322 | const struct vm_operations_struct *vm_ops; |
c92ff1bd MS |
323 | |
324 | /* Information about our backing store: */ | |
325 | unsigned long vm_pgoff; /* Offset (within vm_file) in PAGE_SIZE | |
ea1754a0 | 326 | units */ |
c92ff1bd MS |
327 | struct file * vm_file; /* File we map to (can be NULL). */ |
328 | void * vm_private_data; /* was vm_pte (shared mem) */ | |
c92ff1bd | 329 | |
ec560175 | 330 | atomic_long_t swap_readahead_info; |
c92ff1bd | 331 | #ifndef CONFIG_MMU |
8feae131 | 332 | struct vm_region *vm_region; /* NOMMU mapping region */ |
c92ff1bd MS |
333 | #endif |
334 | #ifdef CONFIG_NUMA | |
335 | struct mempolicy *vm_policy; /* NUMA policy for the VMA */ | |
336 | #endif | |
745f234b | 337 | struct vm_userfaultfd_ctx vm_userfaultfd_ctx; |
3859a271 | 338 | } __randomize_layout; |
c92ff1bd | 339 | |
b564daf8 ON |
340 | struct core_thread { |
341 | struct task_struct *task; | |
342 | struct core_thread *next; | |
343 | }; | |
344 | ||
32ecb1f2 | 345 | struct core_state { |
c5f1cc8c | 346 | atomic_t nr_threads; |
b564daf8 | 347 | struct core_thread dumper; |
32ecb1f2 ON |
348 | struct completion startup; |
349 | }; | |
350 | ||
db446a08 | 351 | struct kioctx_table; |
c92ff1bd | 352 | struct mm_struct { |
615d6e87 | 353 | struct vm_area_struct *mmap; /* list of VMAs */ |
c92ff1bd | 354 | struct rb_root mm_rb; |
615d6e87 | 355 | u32 vmacache_seqnum; /* per-thread vmacache */ |
efc1a3b1 | 356 | #ifdef CONFIG_MMU |
c92ff1bd MS |
357 | unsigned long (*get_unmapped_area) (struct file *filp, |
358 | unsigned long addr, unsigned long len, | |
359 | unsigned long pgoff, unsigned long flags); | |
efc1a3b1 | 360 | #endif |
c92ff1bd | 361 | unsigned long mmap_base; /* base of mmap area */ |
41aacc1e | 362 | unsigned long mmap_legacy_base; /* base of mmap area in bottom-up allocations */ |
1b028f78 DS |
363 | #ifdef CONFIG_HAVE_ARCH_COMPAT_MMAP_BASES |
364 | /* Base adresses for compatible mmap() */ | |
365 | unsigned long mmap_compat_base; | |
366 | unsigned long mmap_compat_legacy_base; | |
367 | #endif | |
c92ff1bd | 368 | unsigned long task_size; /* size of task vm space */ |
d3737187 | 369 | unsigned long highest_vm_end; /* highest vma end address */ |
c92ff1bd | 370 | pgd_t * pgd; |
b279ddc3 VN |
371 | |
372 | /** | |
373 | * @mm_users: The number of users including userspace. | |
374 | * | |
375 | * Use mmget()/mmget_not_zero()/mmput() to modify. When this drops | |
376 | * to 0 (i.e. when the task exits and there are no other temporary | |
377 | * reference holders), we also release a reference on @mm_count | |
378 | * (which may then free the &struct mm_struct if @mm_count also | |
379 | * drops to 0). | |
380 | */ | |
381 | atomic_t mm_users; | |
382 | ||
383 | /** | |
384 | * @mm_count: The number of references to &struct mm_struct | |
385 | * (@mm_users count as 1). | |
386 | * | |
387 | * Use mmgrab()/mmdrop() to modify. When this drops to 0, the | |
388 | * &struct mm_struct is freed. | |
389 | */ | |
390 | atomic_t mm_count; | |
391 | ||
c4812909 | 392 | #ifdef CONFIG_MMU |
af5b0f6a | 393 | atomic_long_t pgtables_bytes; /* PTE page table pages */ |
5a3fbef3 | 394 | #endif |
c92ff1bd | 395 | int map_count; /* number of VMAs */ |
481b4bb5 | 396 | |
c92ff1bd | 397 | spinlock_t page_table_lock; /* Protects page tables and some counters */ |
481b4bb5 | 398 | struct rw_semaphore mmap_sem; |
c92ff1bd MS |
399 | |
400 | struct list_head mmlist; /* List of maybe swapped mm's. These are globally strung | |
401 | * together off init_mm.mmlist, and are protected | |
402 | * by mmlist_lock | |
403 | */ | |
404 | ||
c92ff1bd MS |
405 | |
406 | unsigned long hiwater_rss; /* High-watermark of RSS usage */ | |
407 | unsigned long hiwater_vm; /* High-water virtual memory usage */ | |
408 | ||
e10d59f2 CL |
409 | unsigned long total_vm; /* Total pages mapped */ |
410 | unsigned long locked_vm; /* Pages that have PG_mlocked set */ | |
411 | unsigned long pinned_vm; /* Refcount permanently increased */ | |
30bdbb78 KK |
412 | unsigned long data_vm; /* VM_WRITE & ~VM_SHARED & ~VM_STACK */ |
413 | unsigned long exec_vm; /* VM_EXEC & ~VM_WRITE & ~VM_STACK */ | |
414 | unsigned long stack_vm; /* VM_STACK */ | |
e10d59f2 | 415 | unsigned long def_flags; |
88aa7cc6 YS |
416 | |
417 | spinlock_t arg_lock; /* protect the below fields */ | |
c92ff1bd MS |
418 | unsigned long start_code, end_code, start_data, end_data; |
419 | unsigned long start_brk, brk, start_stack; | |
420 | unsigned long arg_start, arg_end, env_start, env_end; | |
421 | ||
422 | unsigned long saved_auxv[AT_VECTOR_SIZE]; /* for /proc/PID/auxv */ | |
423 | ||
d559db08 KH |
424 | /* |
425 | * Special counters, in some configurations protected by the | |
426 | * page_table_lock, in other configurations by being atomic. | |
427 | */ | |
428 | struct mm_rss_stat rss_stat; | |
429 | ||
801460d0 HS |
430 | struct linux_binfmt *binfmt; |
431 | ||
6345d24d LT |
432 | cpumask_var_t cpu_vm_mask_var; |
433 | ||
c92ff1bd MS |
434 | /* Architecture-specific MM context */ |
435 | mm_context_t context; | |
436 | ||
c92ff1bd MS |
437 | unsigned long flags; /* Must use atomic bitops to access the bits */ |
438 | ||
a94e2d40 | 439 | struct core_state *core_state; /* coredumping support */ |
a961e409 MD |
440 | #ifdef CONFIG_MEMBARRIER |
441 | atomic_t membarrier_state; | |
442 | #endif | |
858f0993 | 443 | #ifdef CONFIG_AIO |
db446a08 BL |
444 | spinlock_t ioctx_lock; |
445 | struct kioctx_table __rcu *ioctx_table; | |
858f0993 | 446 | #endif |
f98bafa0 | 447 | #ifdef CONFIG_MEMCG |
4cd1a8fc KM |
448 | /* |
449 | * "owner" points to a task that is regarded as the canonical | |
450 | * user/owner of this mm. All of the following must be true in | |
451 | * order for it to be changed: | |
452 | * | |
453 | * current == mm->owner | |
454 | * current->mm != mm | |
455 | * new_owner->mm == mm | |
456 | * new_owner->alloc_lock is held | |
457 | */ | |
4d2deb40 | 458 | struct task_struct __rcu *owner; |
78fb7466 | 459 | #endif |
bfedb589 | 460 | struct user_namespace *user_ns; |
925d1c40 | 461 | |
925d1c40 | 462 | /* store ref to file /proc/<pid>/exe symlink points to */ |
90f31d0e | 463 | struct file __rcu *exe_file; |
cddb8a5c AA |
464 | #ifdef CONFIG_MMU_NOTIFIER |
465 | struct mmu_notifier_mm *mmu_notifier_mm; | |
e7a00c45 | 466 | #endif |
e009bb30 | 467 | #if defined(CONFIG_TRANSPARENT_HUGEPAGE) && !USE_SPLIT_PMD_PTLOCKS |
e7a00c45 | 468 | pgtable_t pmd_huge_pte; /* protected by page_table_lock */ |
cddb8a5c | 469 | #endif |
6345d24d LT |
470 | #ifdef CONFIG_CPUMASK_OFFSTACK |
471 | struct cpumask cpumask_allocation; | |
cbee9f88 PZ |
472 | #endif |
473 | #ifdef CONFIG_NUMA_BALANCING | |
474 | /* | |
34f0315a MG |
475 | * numa_next_scan is the next time that the PTEs will be marked |
476 | * pte_numa. NUMA hinting faults will gather statistics and migrate | |
477 | * pages to new nodes if necessary. | |
cbee9f88 PZ |
478 | */ |
479 | unsigned long numa_next_scan; | |
480 | ||
6e5fb223 PZ |
481 | /* Restart point for scanning and setting pte_numa */ |
482 | unsigned long numa_scan_offset; | |
483 | ||
cbee9f88 PZ |
484 | /* numa_scan_seq prevents two threads setting pte_numa */ |
485 | int numa_scan_seq; | |
20841405 | 486 | #endif |
20841405 RR |
487 | /* |
488 | * An operation with batched TLB flushing is going on. Anything that | |
489 | * can move process memory needs to flush the TLB when moving a | |
490 | * PROT_NONE or PROT_NUMA mapped page. | |
491 | */ | |
16af97dc | 492 | atomic_t tlb_flush_pending; |
3ea27719 MG |
493 | #ifdef CONFIG_ARCH_WANT_BATCHED_UNMAP_TLB_FLUSH |
494 | /* See flush_tlb_batched_pending() */ | |
495 | bool tlb_flush_batched; | |
6345d24d | 496 | #endif |
d4b3b638 | 497 | struct uprobes_state uprobes_state; |
5d317b2b NH |
498 | #ifdef CONFIG_HUGETLB_PAGE |
499 | atomic_long_t hugetlb_usage; | |
500 | #endif | |
ec8d7c14 | 501 | struct work_struct async_put_work; |
133ff0ea JG |
502 | |
503 | #if IS_ENABLED(CONFIG_HMM) | |
504 | /* HMM needs to track a few things per mm */ | |
505 | struct hmm *hmm; | |
506 | #endif | |
3859a271 | 507 | } __randomize_layout; |
c92ff1bd | 508 | |
abe722a1 IM |
509 | extern struct mm_struct init_mm; |
510 | ||
6345d24d LT |
511 | static inline void mm_init_cpumask(struct mm_struct *mm) |
512 | { | |
513 | #ifdef CONFIG_CPUMASK_OFFSTACK | |
514 | mm->cpu_vm_mask_var = &mm->cpumask_allocation; | |
515 | #endif | |
41f727fd | 516 | cpumask_clear(mm->cpu_vm_mask_var); |
6345d24d LT |
517 | } |
518 | ||
45e575ab | 519 | /* Future-safe accessor for struct mm_struct's cpu_vm_mask. */ |
de03c72c KM |
520 | static inline cpumask_t *mm_cpumask(struct mm_struct *mm) |
521 | { | |
522 | return mm->cpu_vm_mask_var; | |
523 | } | |
45e575ab | 524 | |
56236a59 MK |
525 | struct mmu_gather; |
526 | extern void tlb_gather_mmu(struct mmu_gather *tlb, struct mm_struct *mm, | |
527 | unsigned long start, unsigned long end); | |
528 | extern void tlb_finish_mmu(struct mmu_gather *tlb, | |
529 | unsigned long start, unsigned long end); | |
530 | ||
16af97dc | 531 | static inline void init_tlb_flush_pending(struct mm_struct *mm) |
20841405 | 532 | { |
16af97dc | 533 | atomic_set(&mm->tlb_flush_pending, 0); |
20841405 | 534 | } |
16af97dc NA |
535 | |
536 | static inline void inc_tlb_flush_pending(struct mm_struct *mm) | |
20841405 | 537 | { |
16af97dc | 538 | atomic_inc(&mm->tlb_flush_pending); |
af2c1401 | 539 | /* |
8b1b436d PZ |
540 | * The only time this value is relevant is when there are indeed pages |
541 | * to flush. And we'll only flush pages after changing them, which | |
542 | * requires the PTL. | |
543 | * | |
544 | * So the ordering here is: | |
545 | * | |
040cca3a | 546 | * atomic_inc(&mm->tlb_flush_pending); |
8b1b436d PZ |
547 | * spin_lock(&ptl); |
548 | * ... | |
549 | * set_pte_at(); | |
550 | * spin_unlock(&ptl); | |
551 | * | |
552 | * spin_lock(&ptl) | |
553 | * mm_tlb_flush_pending(); | |
554 | * .... | |
555 | * spin_unlock(&ptl); | |
556 | * | |
557 | * flush_tlb_range(); | |
040cca3a | 558 | * atomic_dec(&mm->tlb_flush_pending); |
8b1b436d | 559 | * |
0e709703 PZ |
560 | * Where the increment if constrained by the PTL unlock, it thus |
561 | * ensures that the increment is visible if the PTE modification is | |
562 | * visible. After all, if there is no PTE modification, nobody cares | |
563 | * about TLB flushes either. | |
564 | * | |
565 | * This very much relies on users (mm_tlb_flush_pending() and | |
566 | * mm_tlb_flush_nested()) only caring about _specific_ PTEs (and | |
567 | * therefore specific PTLs), because with SPLIT_PTE_PTLOCKS and RCpc | |
568 | * locks (PPC) the unlock of one doesn't order against the lock of | |
569 | * another PTL. | |
570 | * | |
571 | * The decrement is ordered by the flush_tlb_range(), such that | |
572 | * mm_tlb_flush_pending() will not return false unless all flushes have | |
573 | * completed. | |
af2c1401 | 574 | */ |
20841405 | 575 | } |
16af97dc | 576 | |
16af97dc | 577 | static inline void dec_tlb_flush_pending(struct mm_struct *mm) |
20841405 | 578 | { |
0a2c4048 | 579 | /* |
0e709703 PZ |
580 | * See inc_tlb_flush_pending(). |
581 | * | |
582 | * This cannot be smp_mb__before_atomic() because smp_mb() simply does | |
583 | * not order against TLB invalidate completion, which is what we need. | |
584 | * | |
585 | * Therefore we must rely on tlb_flush_*() to guarantee order. | |
0a2c4048 | 586 | */ |
16af97dc | 587 | atomic_dec(&mm->tlb_flush_pending); |
20841405 | 588 | } |
20841405 | 589 | |
0e709703 PZ |
590 | static inline bool mm_tlb_flush_pending(struct mm_struct *mm) |
591 | { | |
592 | /* | |
593 | * Must be called after having acquired the PTL; orders against that | |
594 | * PTLs release and therefore ensures that if we observe the modified | |
595 | * PTE we must also observe the increment from inc_tlb_flush_pending(). | |
596 | * | |
597 | * That is, it only guarantees to return true if there is a flush | |
598 | * pending for _this_ PTL. | |
599 | */ | |
600 | return atomic_read(&mm->tlb_flush_pending); | |
601 | } | |
602 | ||
603 | static inline bool mm_tlb_flush_nested(struct mm_struct *mm) | |
604 | { | |
605 | /* | |
606 | * Similar to mm_tlb_flush_pending(), we must have acquired the PTL | |
607 | * for which there is a TLB flush pending in order to guarantee | |
608 | * we've seen both that PTE modification and the increment. | |
609 | * | |
610 | * (no requirement on actually still holding the PTL, that is irrelevant) | |
611 | */ | |
612 | return atomic_read(&mm->tlb_flush_pending) > 1; | |
613 | } | |
614 | ||
f872f540 AL |
615 | struct vm_fault; |
616 | ||
617 | struct vm_special_mapping { | |
618 | const char *name; /* The name, e.g. "[vdso]". */ | |
619 | ||
620 | /* | |
621 | * If .fault is not provided, this points to a | |
622 | * NULL-terminated array of pages that back the special mapping. | |
623 | * | |
624 | * This must not be NULL unless .fault is provided. | |
625 | */ | |
a62c34bd | 626 | struct page **pages; |
f872f540 AL |
627 | |
628 | /* | |
629 | * If non-NULL, then this is called to resolve page faults | |
630 | * on the special mapping. If used, .pages is not checked. | |
631 | */ | |
b3ec9f33 SJ |
632 | vm_fault_t (*fault)(const struct vm_special_mapping *sm, |
633 | struct vm_area_struct *vma, | |
634 | struct vm_fault *vmf); | |
b059a453 DS |
635 | |
636 | int (*mremap)(const struct vm_special_mapping *sm, | |
637 | struct vm_area_struct *new_vma); | |
a62c34bd AL |
638 | }; |
639 | ||
d17d8f9d DH |
640 | enum tlb_flush_reason { |
641 | TLB_FLUSH_ON_TASK_SWITCH, | |
642 | TLB_REMOTE_SHOOTDOWN, | |
643 | TLB_LOCAL_SHOOTDOWN, | |
644 | TLB_LOCAL_MM_SHOOTDOWN, | |
5b74283a | 645 | TLB_REMOTE_SEND_IPI, |
d17d8f9d DH |
646 | NR_TLB_FLUSH_REASONS, |
647 | }; | |
648 | ||
bd6dace7 TH |
649 | /* |
650 | * A swap entry has to fit into a "unsigned long", as the entry is hidden | |
651 | * in the "index" field of the swapper address space. | |
652 | */ | |
653 | typedef struct { | |
654 | unsigned long val; | |
655 | } swp_entry_t; | |
656 | ||
5b99cd0e | 657 | #endif /* _LINUX_MM_TYPES_H */ |