]>
Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/mm/madvise.c | |
3 | * | |
4 | * Copyright (C) 1999 Linus Torvalds | |
5 | * Copyright (C) 2002 Christoph Hellwig | |
6 | */ | |
7 | ||
8 | #include <linux/mman.h> | |
9 | #include <linux/pagemap.h> | |
10 | #include <linux/syscalls.h> | |
05b74384 | 11 | #include <linux/mempolicy.h> |
afcf938e | 12 | #include <linux/page-isolation.h> |
1da177e4 | 13 | #include <linux/hugetlb.h> |
3f31d075 | 14 | #include <linux/falloc.h> |
e8edc6e0 | 15 | #include <linux/sched.h> |
f8af4da3 | 16 | #include <linux/ksm.h> |
3f31d075 | 17 | #include <linux/fs.h> |
9ab4233d | 18 | #include <linux/file.h> |
1998cc04 | 19 | #include <linux/blkdev.h> |
66114cad | 20 | #include <linux/backing-dev.h> |
1998cc04 SL |
21 | #include <linux/swap.h> |
22 | #include <linux/swapops.h> | |
854e9ed0 MK |
23 | #include <linux/mmu_notifier.h> |
24 | ||
25 | #include <asm/tlb.h> | |
1da177e4 | 26 | |
0a27a14a NP |
27 | /* |
28 | * Any behaviour which results in changes to the vma->vm_flags needs to | |
29 | * take mmap_sem for writing. Others, which simply traverse vmas, need | |
30 | * to only take it for reading. | |
31 | */ | |
32 | static int madvise_need_mmap_write(int behavior) | |
33 | { | |
34 | switch (behavior) { | |
35 | case MADV_REMOVE: | |
36 | case MADV_WILLNEED: | |
37 | case MADV_DONTNEED: | |
854e9ed0 | 38 | case MADV_FREE: |
0a27a14a NP |
39 | return 0; |
40 | default: | |
41 | /* be safe, default to 1. list exceptions explicitly */ | |
42 | return 1; | |
43 | } | |
44 | } | |
45 | ||
1da177e4 LT |
46 | /* |
47 | * We can potentially split a vm area into separate | |
48 | * areas, each area with its own behavior. | |
49 | */ | |
ec9bed9d | 50 | static long madvise_behavior(struct vm_area_struct *vma, |
05b74384 PM |
51 | struct vm_area_struct **prev, |
52 | unsigned long start, unsigned long end, int behavior) | |
1da177e4 | 53 | { |
ec9bed9d | 54 | struct mm_struct *mm = vma->vm_mm; |
1da177e4 | 55 | int error = 0; |
05b74384 | 56 | pgoff_t pgoff; |
3866ea90 | 57 | unsigned long new_flags = vma->vm_flags; |
e798c6e8 PM |
58 | |
59 | switch (behavior) { | |
f8225661 MT |
60 | case MADV_NORMAL: |
61 | new_flags = new_flags & ~VM_RAND_READ & ~VM_SEQ_READ; | |
62 | break; | |
e798c6e8 | 63 | case MADV_SEQUENTIAL: |
f8225661 | 64 | new_flags = (new_flags & ~VM_RAND_READ) | VM_SEQ_READ; |
e798c6e8 PM |
65 | break; |
66 | case MADV_RANDOM: | |
f8225661 | 67 | new_flags = (new_flags & ~VM_SEQ_READ) | VM_RAND_READ; |
e798c6e8 | 68 | break; |
f8225661 MT |
69 | case MADV_DONTFORK: |
70 | new_flags |= VM_DONTCOPY; | |
71 | break; | |
72 | case MADV_DOFORK: | |
3866ea90 HD |
73 | if (vma->vm_flags & VM_IO) { |
74 | error = -EINVAL; | |
75 | goto out; | |
76 | } | |
f8225661 | 77 | new_flags &= ~VM_DONTCOPY; |
e798c6e8 | 78 | break; |
accb61fe | 79 | case MADV_DONTDUMP: |
0103bd16 | 80 | new_flags |= VM_DONTDUMP; |
accb61fe JB |
81 | break; |
82 | case MADV_DODUMP: | |
0103bd16 KK |
83 | if (new_flags & VM_SPECIAL) { |
84 | error = -EINVAL; | |
85 | goto out; | |
86 | } | |
87 | new_flags &= ~VM_DONTDUMP; | |
accb61fe | 88 | break; |
f8af4da3 HD |
89 | case MADV_MERGEABLE: |
90 | case MADV_UNMERGEABLE: | |
91 | error = ksm_madvise(vma, start, end, behavior, &new_flags); | |
92 | if (error) | |
93 | goto out; | |
94 | break; | |
0af4e98b | 95 | case MADV_HUGEPAGE: |
a664b2d8 | 96 | case MADV_NOHUGEPAGE: |
60ab3244 | 97 | error = hugepage_madvise(vma, &new_flags, behavior); |
0af4e98b AA |
98 | if (error) |
99 | goto out; | |
100 | break; | |
e798c6e8 PM |
101 | } |
102 | ||
05b74384 PM |
103 | if (new_flags == vma->vm_flags) { |
104 | *prev = vma; | |
836d5ffd | 105 | goto out; |
05b74384 PM |
106 | } |
107 | ||
108 | pgoff = vma->vm_pgoff + ((start - vma->vm_start) >> PAGE_SHIFT); | |
109 | *prev = vma_merge(mm, *prev, start, end, new_flags, vma->anon_vma, | |
19a809af AA |
110 | vma->vm_file, pgoff, vma_policy(vma), |
111 | vma->vm_userfaultfd_ctx); | |
05b74384 PM |
112 | if (*prev) { |
113 | vma = *prev; | |
114 | goto success; | |
115 | } | |
116 | ||
117 | *prev = vma; | |
1da177e4 LT |
118 | |
119 | if (start != vma->vm_start) { | |
120 | error = split_vma(mm, vma, start, 1); | |
121 | if (error) | |
122 | goto out; | |
123 | } | |
124 | ||
125 | if (end != vma->vm_end) { | |
126 | error = split_vma(mm, vma, end, 0); | |
127 | if (error) | |
128 | goto out; | |
129 | } | |
130 | ||
836d5ffd | 131 | success: |
1da177e4 LT |
132 | /* |
133 | * vm_flags is protected by the mmap_sem held in write mode. | |
134 | */ | |
e798c6e8 | 135 | vma->vm_flags = new_flags; |
1da177e4 LT |
136 | |
137 | out: | |
138 | if (error == -ENOMEM) | |
139 | error = -EAGAIN; | |
140 | return error; | |
141 | } | |
142 | ||
1998cc04 SL |
143 | #ifdef CONFIG_SWAP |
144 | static int swapin_walk_pmd_entry(pmd_t *pmd, unsigned long start, | |
145 | unsigned long end, struct mm_walk *walk) | |
146 | { | |
147 | pte_t *orig_pte; | |
148 | struct vm_area_struct *vma = walk->private; | |
149 | unsigned long index; | |
150 | ||
151 | if (pmd_none_or_trans_huge_or_clear_bad(pmd)) | |
152 | return 0; | |
153 | ||
154 | for (index = start; index != end; index += PAGE_SIZE) { | |
155 | pte_t pte; | |
156 | swp_entry_t entry; | |
157 | struct page *page; | |
158 | spinlock_t *ptl; | |
159 | ||
160 | orig_pte = pte_offset_map_lock(vma->vm_mm, pmd, start, &ptl); | |
161 | pte = *(orig_pte + ((index - start) / PAGE_SIZE)); | |
162 | pte_unmap_unlock(orig_pte, ptl); | |
163 | ||
0661a336 | 164 | if (pte_present(pte) || pte_none(pte)) |
1998cc04 SL |
165 | continue; |
166 | entry = pte_to_swp_entry(pte); | |
167 | if (unlikely(non_swap_entry(entry))) | |
168 | continue; | |
169 | ||
170 | page = read_swap_cache_async(entry, GFP_HIGHUSER_MOVABLE, | |
171 | vma, index); | |
172 | if (page) | |
173 | page_cache_release(page); | |
174 | } | |
175 | ||
176 | return 0; | |
177 | } | |
178 | ||
179 | static void force_swapin_readahead(struct vm_area_struct *vma, | |
180 | unsigned long start, unsigned long end) | |
181 | { | |
182 | struct mm_walk walk = { | |
183 | .mm = vma->vm_mm, | |
184 | .pmd_entry = swapin_walk_pmd_entry, | |
185 | .private = vma, | |
186 | }; | |
187 | ||
188 | walk_page_range(start, end, &walk); | |
189 | ||
190 | lru_add_drain(); /* Push any new pages onto the LRU now */ | |
191 | } | |
192 | ||
193 | static void force_shm_swapin_readahead(struct vm_area_struct *vma, | |
194 | unsigned long start, unsigned long end, | |
195 | struct address_space *mapping) | |
196 | { | |
197 | pgoff_t index; | |
198 | struct page *page; | |
199 | swp_entry_t swap; | |
200 | ||
201 | for (; start < end; start += PAGE_SIZE) { | |
202 | index = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
203 | ||
55231e5c | 204 | page = find_get_entry(mapping, index); |
1998cc04 SL |
205 | if (!radix_tree_exceptional_entry(page)) { |
206 | if (page) | |
207 | page_cache_release(page); | |
208 | continue; | |
209 | } | |
210 | swap = radix_to_swp_entry(page); | |
211 | page = read_swap_cache_async(swap, GFP_HIGHUSER_MOVABLE, | |
212 | NULL, 0); | |
213 | if (page) | |
214 | page_cache_release(page); | |
215 | } | |
216 | ||
217 | lru_add_drain(); /* Push any new pages onto the LRU now */ | |
218 | } | |
219 | #endif /* CONFIG_SWAP */ | |
220 | ||
1da177e4 LT |
221 | /* |
222 | * Schedule all required I/O operations. Do not wait for completion. | |
223 | */ | |
ec9bed9d VC |
224 | static long madvise_willneed(struct vm_area_struct *vma, |
225 | struct vm_area_struct **prev, | |
1da177e4 LT |
226 | unsigned long start, unsigned long end) |
227 | { | |
228 | struct file *file = vma->vm_file; | |
229 | ||
1998cc04 | 230 | #ifdef CONFIG_SWAP |
97b713ba | 231 | if (!file) { |
1998cc04 | 232 | *prev = vma; |
97b713ba | 233 | force_swapin_readahead(vma, start, end); |
1998cc04 SL |
234 | return 0; |
235 | } | |
1998cc04 | 236 | |
97b713ba CH |
237 | if (shmem_mapping(file->f_mapping)) { |
238 | *prev = vma; | |
239 | force_shm_swapin_readahead(vma, start, end, | |
240 | file->f_mapping); | |
241 | return 0; | |
242 | } | |
243 | #else | |
1bef4003 S |
244 | if (!file) |
245 | return -EBADF; | |
97b713ba | 246 | #endif |
1bef4003 | 247 | |
e748dcd0 | 248 | if (IS_DAX(file_inode(file))) { |
fe77ba6f CO |
249 | /* no bad return value, but ignore advice */ |
250 | return 0; | |
251 | } | |
252 | ||
05b74384 | 253 | *prev = vma; |
1da177e4 LT |
254 | start = ((start - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; |
255 | if (end > vma->vm_end) | |
256 | end = vma->vm_end; | |
257 | end = ((end - vma->vm_start) >> PAGE_SHIFT) + vma->vm_pgoff; | |
258 | ||
f7e839dd | 259 | force_page_cache_readahead(file->f_mapping, file, start, end - start); |
1da177e4 LT |
260 | return 0; |
261 | } | |
262 | ||
854e9ed0 MK |
263 | static int madvise_free_pte_range(pmd_t *pmd, unsigned long addr, |
264 | unsigned long end, struct mm_walk *walk) | |
265 | ||
266 | { | |
267 | struct mmu_gather *tlb = walk->private; | |
268 | struct mm_struct *mm = tlb->mm; | |
269 | struct vm_area_struct *vma = walk->vma; | |
270 | spinlock_t *ptl; | |
271 | pte_t *orig_pte, *pte, ptent; | |
272 | struct page *page; | |
273 | ||
274 | split_huge_pmd(vma, pmd, addr); | |
275 | if (pmd_trans_unstable(pmd)) | |
276 | return 0; | |
277 | ||
278 | orig_pte = pte = pte_offset_map_lock(mm, pmd, addr, &ptl); | |
279 | arch_enter_lazy_mmu_mode(); | |
280 | for (; addr != end; pte++, addr += PAGE_SIZE) { | |
281 | ptent = *pte; | |
282 | ||
283 | if (!pte_present(ptent)) | |
284 | continue; | |
285 | ||
286 | page = vm_normal_page(vma, addr, ptent); | |
287 | if (!page) | |
288 | continue; | |
289 | ||
290 | /* | |
291 | * If pmd isn't transhuge but the page is THP and | |
292 | * is owned by only this process, split it and | |
293 | * deactivate all pages. | |
294 | */ | |
295 | if (PageTransCompound(page)) { | |
296 | if (page_mapcount(page) != 1) | |
297 | goto out; | |
298 | get_page(page); | |
299 | if (!trylock_page(page)) { | |
300 | put_page(page); | |
301 | goto out; | |
302 | } | |
303 | pte_unmap_unlock(orig_pte, ptl); | |
304 | if (split_huge_page(page)) { | |
305 | unlock_page(page); | |
306 | put_page(page); | |
307 | pte_offset_map_lock(mm, pmd, addr, &ptl); | |
308 | goto out; | |
309 | } | |
310 | put_page(page); | |
311 | unlock_page(page); | |
312 | pte = pte_offset_map_lock(mm, pmd, addr, &ptl); | |
313 | pte--; | |
314 | addr -= PAGE_SIZE; | |
315 | continue; | |
316 | } | |
317 | ||
318 | VM_BUG_ON_PAGE(PageTransCompound(page), page); | |
319 | ||
320 | if (PageSwapCache(page) || PageDirty(page)) { | |
321 | if (!trylock_page(page)) | |
322 | continue; | |
323 | /* | |
324 | * If page is shared with others, we couldn't clear | |
325 | * PG_dirty of the page. | |
326 | */ | |
327 | if (page_mapcount(page) != 1) { | |
328 | unlock_page(page); | |
329 | continue; | |
330 | } | |
331 | ||
332 | if (PageSwapCache(page) && !try_to_free_swap(page)) { | |
333 | unlock_page(page); | |
334 | continue; | |
335 | } | |
336 | ||
337 | ClearPageDirty(page); | |
338 | unlock_page(page); | |
339 | } | |
340 | ||
341 | if (pte_young(ptent) || pte_dirty(ptent)) { | |
342 | /* | |
343 | * Some of architecture(ex, PPC) don't update TLB | |
344 | * with set_pte_at and tlb_remove_tlb_entry so for | |
345 | * the portability, remap the pte with old|clean | |
346 | * after pte clearing. | |
347 | */ | |
348 | ptent = ptep_get_and_clear_full(mm, addr, pte, | |
349 | tlb->fullmm); | |
350 | ||
351 | ptent = pte_mkold(ptent); | |
352 | ptent = pte_mkclean(ptent); | |
353 | set_pte_at(mm, addr, pte, ptent); | |
354 | tlb_remove_tlb_entry(tlb, pte, addr); | |
355 | } | |
356 | } | |
357 | out: | |
358 | arch_leave_lazy_mmu_mode(); | |
359 | pte_unmap_unlock(orig_pte, ptl); | |
360 | cond_resched(); | |
361 | return 0; | |
362 | } | |
363 | ||
364 | static void madvise_free_page_range(struct mmu_gather *tlb, | |
365 | struct vm_area_struct *vma, | |
366 | unsigned long addr, unsigned long end) | |
367 | { | |
368 | struct mm_walk free_walk = { | |
369 | .pmd_entry = madvise_free_pte_range, | |
370 | .mm = vma->vm_mm, | |
371 | .private = tlb, | |
372 | }; | |
373 | ||
374 | tlb_start_vma(tlb, vma); | |
375 | walk_page_range(addr, end, &free_walk); | |
376 | tlb_end_vma(tlb, vma); | |
377 | } | |
378 | ||
379 | static int madvise_free_single_vma(struct vm_area_struct *vma, | |
380 | unsigned long start_addr, unsigned long end_addr) | |
381 | { | |
382 | unsigned long start, end; | |
383 | struct mm_struct *mm = vma->vm_mm; | |
384 | struct mmu_gather tlb; | |
385 | ||
386 | if (vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP)) | |
387 | return -EINVAL; | |
388 | ||
389 | /* MADV_FREE works for only anon vma at the moment */ | |
390 | if (!vma_is_anonymous(vma)) | |
391 | return -EINVAL; | |
392 | ||
393 | start = max(vma->vm_start, start_addr); | |
394 | if (start >= vma->vm_end) | |
395 | return -EINVAL; | |
396 | end = min(vma->vm_end, end_addr); | |
397 | if (end <= vma->vm_start) | |
398 | return -EINVAL; | |
399 | ||
400 | lru_add_drain(); | |
401 | tlb_gather_mmu(&tlb, mm, start, end); | |
402 | update_hiwater_rss(mm); | |
403 | ||
404 | mmu_notifier_invalidate_range_start(mm, start, end); | |
405 | madvise_free_page_range(&tlb, vma, start, end); | |
406 | mmu_notifier_invalidate_range_end(mm, start, end); | |
407 | tlb_finish_mmu(&tlb, start, end); | |
408 | ||
409 | return 0; | |
410 | } | |
411 | ||
412 | static long madvise_free(struct vm_area_struct *vma, | |
413 | struct vm_area_struct **prev, | |
414 | unsigned long start, unsigned long end) | |
415 | { | |
416 | *prev = vma; | |
417 | return madvise_free_single_vma(vma, start, end); | |
418 | } | |
419 | ||
1da177e4 LT |
420 | /* |
421 | * Application no longer needs these pages. If the pages are dirty, | |
422 | * it's OK to just throw them away. The app will be more careful about | |
423 | * data it wants to keep. Be sure to free swap resources too. The | |
7e6cbea3 | 424 | * zap_page_range call sets things up for shrink_active_list to actually free |
1da177e4 LT |
425 | * these pages later if no one else has touched them in the meantime, |
426 | * although we could add these pages to a global reuse list for | |
7e6cbea3 | 427 | * shrink_active_list to pick up before reclaiming other pages. |
1da177e4 LT |
428 | * |
429 | * NB: This interface discards data rather than pushes it out to swap, | |
430 | * as some implementations do. This has performance implications for | |
431 | * applications like large transactional databases which want to discard | |
432 | * pages in anonymous maps after committing to backing store the data | |
433 | * that was kept in them. There is no reason to write this data out to | |
434 | * the swap area if the application is discarding it. | |
435 | * | |
436 | * An interface that causes the system to free clean pages and flush | |
437 | * dirty pages is already available as msync(MS_INVALIDATE). | |
438 | */ | |
ec9bed9d VC |
439 | static long madvise_dontneed(struct vm_area_struct *vma, |
440 | struct vm_area_struct **prev, | |
1da177e4 LT |
441 | unsigned long start, unsigned long end) |
442 | { | |
05b74384 | 443 | *prev = vma; |
6aab341e | 444 | if (vma->vm_flags & (VM_LOCKED|VM_HUGETLB|VM_PFNMAP)) |
1da177e4 LT |
445 | return -EINVAL; |
446 | ||
8a5f14a2 | 447 | zap_page_range(vma, start, end - start, NULL); |
1da177e4 LT |
448 | return 0; |
449 | } | |
450 | ||
f6b3ec23 BP |
451 | /* |
452 | * Application wants to free up the pages and associated backing store. | |
453 | * This is effectively punching a hole into the middle of a file. | |
f6b3ec23 BP |
454 | */ |
455 | static long madvise_remove(struct vm_area_struct *vma, | |
00e9fa2d | 456 | struct vm_area_struct **prev, |
f6b3ec23 BP |
457 | unsigned long start, unsigned long end) |
458 | { | |
3f31d075 | 459 | loff_t offset; |
90ed52eb | 460 | int error; |
9ab4233d | 461 | struct file *f; |
f6b3ec23 | 462 | |
90ed52eb | 463 | *prev = NULL; /* tell sys_madvise we drop mmap_sem */ |
00e9fa2d | 464 | |
72079ba0 | 465 | if (vma->vm_flags & VM_LOCKED) |
f6b3ec23 BP |
466 | return -EINVAL; |
467 | ||
9ab4233d AL |
468 | f = vma->vm_file; |
469 | ||
470 | if (!f || !f->f_mapping || !f->f_mapping->host) { | |
f6b3ec23 BP |
471 | return -EINVAL; |
472 | } | |
473 | ||
69cf0fac HD |
474 | if ((vma->vm_flags & (VM_SHARED|VM_WRITE)) != (VM_SHARED|VM_WRITE)) |
475 | return -EACCES; | |
476 | ||
f6b3ec23 BP |
477 | offset = (loff_t)(start - vma->vm_start) |
478 | + ((loff_t)vma->vm_pgoff << PAGE_SHIFT); | |
90ed52eb | 479 | |
9ab4233d AL |
480 | /* |
481 | * Filesystem's fallocate may need to take i_mutex. We need to | |
482 | * explicitly grab a reference because the vma (and hence the | |
483 | * vma's reference to the file) can go away as soon as we drop | |
484 | * mmap_sem. | |
485 | */ | |
486 | get_file(f); | |
0a27a14a | 487 | up_read(¤t->mm->mmap_sem); |
72c72bdf | 488 | error = vfs_fallocate(f, |
3f31d075 HD |
489 | FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE, |
490 | offset, end - start); | |
9ab4233d | 491 | fput(f); |
0a27a14a | 492 | down_read(¤t->mm->mmap_sem); |
90ed52eb | 493 | return error; |
f6b3ec23 BP |
494 | } |
495 | ||
9893e49d AK |
496 | #ifdef CONFIG_MEMORY_FAILURE |
497 | /* | |
498 | * Error injection support for memory error handling. | |
499 | */ | |
afcf938e | 500 | static int madvise_hwpoison(int bhv, unsigned long start, unsigned long end) |
9893e49d | 501 | { |
20cb6cab | 502 | struct page *p; |
9893e49d AK |
503 | if (!capable(CAP_SYS_ADMIN)) |
504 | return -EPERM; | |
20cb6cab WL |
505 | for (; start < end; start += PAGE_SIZE << |
506 | compound_order(compound_head(p))) { | |
325c4ef5 AM |
507 | int ret; |
508 | ||
509 | ret = get_user_pages_fast(start, 1, 0, &p); | |
9893e49d AK |
510 | if (ret != 1) |
511 | return ret; | |
325c4ef5 | 512 | |
29b4eede WL |
513 | if (PageHWPoison(p)) { |
514 | put_page(p); | |
515 | continue; | |
516 | } | |
afcf938e | 517 | if (bhv == MADV_SOFT_OFFLINE) { |
b194b8cd | 518 | pr_info("Soft offlining page %#lx at %#lx\n", |
afcf938e AK |
519 | page_to_pfn(p), start); |
520 | ret = soft_offline_page(p, MF_COUNT_INCREASED); | |
521 | if (ret) | |
8302423b | 522 | return ret; |
afcf938e AK |
523 | continue; |
524 | } | |
b194b8cd | 525 | pr_info("Injecting memory failure for page %#lx at %#lx\n", |
9893e49d AK |
526 | page_to_pfn(p), start); |
527 | /* Ignore return value for now */ | |
cd42f4a3 | 528 | memory_failure(page_to_pfn(p), 0, MF_COUNT_INCREASED); |
9893e49d | 529 | } |
325c4ef5 | 530 | return 0; |
9893e49d AK |
531 | } |
532 | #endif | |
533 | ||
165cd402 | 534 | static long |
535 | madvise_vma(struct vm_area_struct *vma, struct vm_area_struct **prev, | |
536 | unsigned long start, unsigned long end, int behavior) | |
1da177e4 | 537 | { |
1da177e4 | 538 | switch (behavior) { |
f6b3ec23 | 539 | case MADV_REMOVE: |
3866ea90 | 540 | return madvise_remove(vma, prev, start, end); |
1da177e4 | 541 | case MADV_WILLNEED: |
3866ea90 | 542 | return madvise_willneed(vma, prev, start, end); |
854e9ed0 MK |
543 | case MADV_FREE: |
544 | /* | |
545 | * XXX: In this implementation, MADV_FREE works like | |
546 | * MADV_DONTNEED on swapless system or full swap. | |
547 | */ | |
548 | if (get_nr_swap_pages() > 0) | |
549 | return madvise_free(vma, prev, start, end); | |
550 | /* passthrough */ | |
1da177e4 | 551 | case MADV_DONTNEED: |
3866ea90 | 552 | return madvise_dontneed(vma, prev, start, end); |
1da177e4 | 553 | default: |
3866ea90 | 554 | return madvise_behavior(vma, prev, start, end, behavior); |
1da177e4 | 555 | } |
1da177e4 LT |
556 | } |
557 | ||
1ecef9ed | 558 | static bool |
75927af8 NP |
559 | madvise_behavior_valid(int behavior) |
560 | { | |
561 | switch (behavior) { | |
562 | case MADV_DOFORK: | |
563 | case MADV_DONTFORK: | |
564 | case MADV_NORMAL: | |
565 | case MADV_SEQUENTIAL: | |
566 | case MADV_RANDOM: | |
567 | case MADV_REMOVE: | |
568 | case MADV_WILLNEED: | |
569 | case MADV_DONTNEED: | |
854e9ed0 | 570 | case MADV_FREE: |
f8af4da3 HD |
571 | #ifdef CONFIG_KSM |
572 | case MADV_MERGEABLE: | |
573 | case MADV_UNMERGEABLE: | |
0af4e98b AA |
574 | #endif |
575 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
576 | case MADV_HUGEPAGE: | |
a664b2d8 | 577 | case MADV_NOHUGEPAGE: |
f8af4da3 | 578 | #endif |
accb61fe JB |
579 | case MADV_DONTDUMP: |
580 | case MADV_DODUMP: | |
1ecef9ed | 581 | return true; |
75927af8 NP |
582 | |
583 | default: | |
1ecef9ed | 584 | return false; |
75927af8 NP |
585 | } |
586 | } | |
3866ea90 | 587 | |
1da177e4 LT |
588 | /* |
589 | * The madvise(2) system call. | |
590 | * | |
591 | * Applications can use madvise() to advise the kernel how it should | |
592 | * handle paging I/O in this VM area. The idea is to help the kernel | |
593 | * use appropriate read-ahead and caching techniques. The information | |
594 | * provided is advisory only, and can be safely disregarded by the | |
595 | * kernel without affecting the correct operation of the application. | |
596 | * | |
597 | * behavior values: | |
598 | * MADV_NORMAL - the default behavior is to read clusters. This | |
599 | * results in some read-ahead and read-behind. | |
600 | * MADV_RANDOM - the system should read the minimum amount of data | |
601 | * on any access, since it is unlikely that the appli- | |
602 | * cation will need more than what it asks for. | |
603 | * MADV_SEQUENTIAL - pages in the given range will probably be accessed | |
604 | * once, so they can be aggressively read ahead, and | |
605 | * can be freed soon after they are accessed. | |
606 | * MADV_WILLNEED - the application is notifying the system to read | |
607 | * some pages ahead. | |
608 | * MADV_DONTNEED - the application is finished with the given range, | |
609 | * so the kernel can free resources associated with it. | |
f6b3ec23 BP |
610 | * MADV_REMOVE - the application wants to free up the given range of |
611 | * pages and associated backing store. | |
3866ea90 HD |
612 | * MADV_DONTFORK - omit this area from child's address space when forking: |
613 | * typically, to avoid COWing pages pinned by get_user_pages(). | |
614 | * MADV_DOFORK - cancel MADV_DONTFORK: no longer omit this area when forking. | |
f8af4da3 HD |
615 | * MADV_MERGEABLE - the application recommends that KSM try to merge pages in |
616 | * this area with pages of identical content from other such areas. | |
617 | * MADV_UNMERGEABLE- cancel MADV_MERGEABLE: no longer merge pages with others. | |
1da177e4 LT |
618 | * |
619 | * return values: | |
620 | * zero - success | |
621 | * -EINVAL - start + len < 0, start is not page-aligned, | |
622 | * "behavior" is not a valid value, or application | |
623 | * is attempting to release locked or shared pages. | |
624 | * -ENOMEM - addresses in the specified range are not currently | |
625 | * mapped, or are outside the AS of the process. | |
626 | * -EIO - an I/O error occurred while paging in data. | |
627 | * -EBADF - map exists, but area maps something that isn't a file. | |
628 | * -EAGAIN - a kernel resource was temporarily unavailable. | |
629 | */ | |
3480b257 | 630 | SYSCALL_DEFINE3(madvise, unsigned long, start, size_t, len_in, int, behavior) |
1da177e4 | 631 | { |
05b74384 | 632 | unsigned long end, tmp; |
ec9bed9d | 633 | struct vm_area_struct *vma, *prev; |
1da177e4 LT |
634 | int unmapped_error = 0; |
635 | int error = -EINVAL; | |
f7977793 | 636 | int write; |
1da177e4 | 637 | size_t len; |
1998cc04 | 638 | struct blk_plug plug; |
1da177e4 | 639 | |
9893e49d | 640 | #ifdef CONFIG_MEMORY_FAILURE |
afcf938e AK |
641 | if (behavior == MADV_HWPOISON || behavior == MADV_SOFT_OFFLINE) |
642 | return madvise_hwpoison(behavior, start, start+len_in); | |
9893e49d | 643 | #endif |
75927af8 NP |
644 | if (!madvise_behavior_valid(behavior)) |
645 | return error; | |
646 | ||
1da177e4 | 647 | if (start & ~PAGE_MASK) |
84d96d89 | 648 | return error; |
1da177e4 LT |
649 | len = (len_in + ~PAGE_MASK) & PAGE_MASK; |
650 | ||
651 | /* Check to see whether len was rounded up from small -ve to zero */ | |
652 | if (len_in && !len) | |
84d96d89 | 653 | return error; |
1da177e4 LT |
654 | |
655 | end = start + len; | |
656 | if (end < start) | |
84d96d89 | 657 | return error; |
1da177e4 LT |
658 | |
659 | error = 0; | |
660 | if (end == start) | |
84d96d89 RV |
661 | return error; |
662 | ||
663 | write = madvise_need_mmap_write(behavior); | |
664 | if (write) | |
665 | down_write(¤t->mm->mmap_sem); | |
666 | else | |
667 | down_read(¤t->mm->mmap_sem); | |
1da177e4 LT |
668 | |
669 | /* | |
670 | * If the interval [start,end) covers some unmapped address | |
671 | * ranges, just ignore them, but return -ENOMEM at the end. | |
05b74384 | 672 | * - different from the way of handling in mlock etc. |
1da177e4 | 673 | */ |
05b74384 | 674 | vma = find_vma_prev(current->mm, start, &prev); |
836d5ffd HD |
675 | if (vma && start > vma->vm_start) |
676 | prev = vma; | |
677 | ||
1998cc04 | 678 | blk_start_plug(&plug); |
1da177e4 LT |
679 | for (;;) { |
680 | /* Still start < end. */ | |
681 | error = -ENOMEM; | |
682 | if (!vma) | |
84d96d89 | 683 | goto out; |
1da177e4 | 684 | |
05b74384 | 685 | /* Here start < (end|vma->vm_end). */ |
1da177e4 LT |
686 | if (start < vma->vm_start) { |
687 | unmapped_error = -ENOMEM; | |
688 | start = vma->vm_start; | |
05b74384 | 689 | if (start >= end) |
84d96d89 | 690 | goto out; |
1da177e4 LT |
691 | } |
692 | ||
05b74384 PM |
693 | /* Here vma->vm_start <= start < (end|vma->vm_end) */ |
694 | tmp = vma->vm_end; | |
695 | if (end < tmp) | |
696 | tmp = end; | |
1da177e4 | 697 | |
05b74384 PM |
698 | /* Here vma->vm_start <= start < tmp <= (end|vma->vm_end). */ |
699 | error = madvise_vma(vma, &prev, start, tmp, behavior); | |
1da177e4 | 700 | if (error) |
84d96d89 | 701 | goto out; |
05b74384 | 702 | start = tmp; |
90ed52eb | 703 | if (prev && start < prev->vm_end) |
05b74384 PM |
704 | start = prev->vm_end; |
705 | error = unmapped_error; | |
706 | if (start >= end) | |
84d96d89 | 707 | goto out; |
90ed52eb HD |
708 | if (prev) |
709 | vma = prev->vm_next; | |
710 | else /* madvise_remove dropped mmap_sem */ | |
711 | vma = find_vma(current->mm, start); | |
1da177e4 | 712 | } |
1da177e4 | 713 | out: |
84d96d89 | 714 | blk_finish_plug(&plug); |
f7977793 | 715 | if (write) |
0a27a14a NP |
716 | up_write(¤t->mm->mmap_sem); |
717 | else | |
718 | up_read(¤t->mm->mmap_sem); | |
719 | ||
1da177e4 LT |
720 | return error; |
721 | } |