]>
Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
1da177e4 LT |
2 | /* |
3 | * linux/mm/page_io.c | |
4 | * | |
5 | * Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds | |
6 | * | |
7 | * Swap reorganised 29.12.95, | |
8 | * Asynchronous swapping added 30.12.95. Stephen Tweedie | |
9 | * Removed race in async swapping. 14.4.1996. Bruno Haible | |
10 | * Add swap of shared pages through the page cache. 20.2.1998. Stephen Tweedie | |
11 | * Always use brw_page, life becomes simpler. 12 May 1998 Eric Biederman | |
12 | */ | |
13 | ||
14 | #include <linux/mm.h> | |
15 | #include <linux/kernel_stat.h> | |
5a0e3ad6 | 16 | #include <linux/gfp.h> |
1da177e4 LT |
17 | #include <linux/pagemap.h> |
18 | #include <linux/swap.h> | |
19 | #include <linux/bio.h> | |
20 | #include <linux/swapops.h> | |
62c230bc | 21 | #include <linux/buffer_head.h> |
1da177e4 | 22 | #include <linux/writeback.h> |
38b5faf4 | 23 | #include <linux/frontswap.h> |
b430e9d1 | 24 | #include <linux/blkdev.h> |
93779069 | 25 | #include <linux/psi.h> |
e2e40f2c | 26 | #include <linux/uio.h> |
b0ba2d0f | 27 | #include <linux/sched/task.h> |
a3d5dc90 | 28 | #include <linux/delayacct.h> |
1da177e4 | 29 | |
4246a0b6 | 30 | void end_swap_bio_write(struct bio *bio) |
1da177e4 | 31 | { |
263663cd | 32 | struct page *page = bio_first_page_all(bio); |
1da177e4 | 33 | |
4e4cbee9 | 34 | if (bio->bi_status) { |
1da177e4 | 35 | SetPageError(page); |
6ddab3b9 PZ |
36 | /* |
37 | * We failed to write the page out to swap-space. | |
38 | * Re-dirty the page in order to avoid it being reclaimed. | |
39 | * Also print a dire warning that things will go BAD (tm) | |
40 | * very quickly. | |
41 | * | |
575ced1c | 42 | * Also clear PG_reclaim to avoid folio_rotate_reclaimable() |
6ddab3b9 PZ |
43 | */ |
44 | set_page_dirty(page); | |
25eaab43 GD |
45 | pr_alert_ratelimited("Write-error on swap-device (%u:%u:%llu)\n", |
46 | MAJOR(bio_dev(bio)), MINOR(bio_dev(bio)), | |
47 | (unsigned long long)bio->bi_iter.bi_sector); | |
6ddab3b9 PZ |
48 | ClearPageReclaim(page); |
49 | } | |
1da177e4 LT |
50 | end_page_writeback(page); |
51 | bio_put(bio); | |
1da177e4 LT |
52 | } |
53 | ||
4246a0b6 | 54 | static void end_swap_bio_read(struct bio *bio) |
1da177e4 | 55 | { |
263663cd | 56 | struct page *page = bio_first_page_all(bio); |
23955622 | 57 | struct task_struct *waiter = bio->bi_private; |
1da177e4 | 58 | |
4e4cbee9 | 59 | if (bio->bi_status) { |
1da177e4 LT |
60 | SetPageError(page); |
61 | ClearPageUptodate(page); | |
25eaab43 GD |
62 | pr_alert_ratelimited("Read-error on swap-device (%u:%u:%llu)\n", |
63 | MAJOR(bio_dev(bio)), MINOR(bio_dev(bio)), | |
64 | (unsigned long long)bio->bi_iter.bi_sector); | |
b430e9d1 | 65 | goto out; |
1da177e4 | 66 | } |
b430e9d1 MK |
67 | |
68 | SetPageUptodate(page); | |
b430e9d1 | 69 | out: |
1da177e4 | 70 | unlock_page(page); |
23955622 | 71 | WRITE_ONCE(bio->bi_private, NULL); |
1da177e4 | 72 | bio_put(bio); |
87518530 ON |
73 | if (waiter) { |
74 | blk_wake_io_task(waiter); | |
75 | put_task_struct(waiter); | |
76 | } | |
1da177e4 LT |
77 | } |
78 | ||
a509bc1a MG |
79 | int generic_swapfile_activate(struct swap_info_struct *sis, |
80 | struct file *swap_file, | |
81 | sector_t *span) | |
82 | { | |
83 | struct address_space *mapping = swap_file->f_mapping; | |
84 | struct inode *inode = mapping->host; | |
85 | unsigned blocks_per_page; | |
86 | unsigned long page_no; | |
87 | unsigned blkbits; | |
88 | sector_t probe_block; | |
89 | sector_t last_block; | |
90 | sector_t lowest_block = -1; | |
91 | sector_t highest_block = 0; | |
92 | int nr_extents = 0; | |
93 | int ret; | |
94 | ||
95 | blkbits = inode->i_blkbits; | |
96 | blocks_per_page = PAGE_SIZE >> blkbits; | |
97 | ||
98 | /* | |
4efaceb1 | 99 | * Map all the blocks into the extent tree. This code doesn't try |
a509bc1a MG |
100 | * to be very smart. |
101 | */ | |
102 | probe_block = 0; | |
103 | page_no = 0; | |
104 | last_block = i_size_read(inode) >> blkbits; | |
105 | while ((probe_block + blocks_per_page) <= last_block && | |
106 | page_no < sis->max) { | |
107 | unsigned block_in_page; | |
108 | sector_t first_block; | |
109 | ||
7e4411bf MP |
110 | cond_resched(); |
111 | ||
30460e1e CM |
112 | first_block = probe_block; |
113 | ret = bmap(inode, &first_block); | |
114 | if (ret || !first_block) | |
a509bc1a MG |
115 | goto bad_bmap; |
116 | ||
117 | /* | |
118 | * It must be PAGE_SIZE aligned on-disk | |
119 | */ | |
120 | if (first_block & (blocks_per_page - 1)) { | |
121 | probe_block++; | |
122 | goto reprobe; | |
123 | } | |
124 | ||
125 | for (block_in_page = 1; block_in_page < blocks_per_page; | |
126 | block_in_page++) { | |
127 | sector_t block; | |
128 | ||
30460e1e CM |
129 | block = probe_block + block_in_page; |
130 | ret = bmap(inode, &block); | |
131 | if (ret || !block) | |
a509bc1a | 132 | goto bad_bmap; |
30460e1e | 133 | |
a509bc1a MG |
134 | if (block != first_block + block_in_page) { |
135 | /* Discontiguity */ | |
136 | probe_block++; | |
137 | goto reprobe; | |
138 | } | |
139 | } | |
140 | ||
141 | first_block >>= (PAGE_SHIFT - blkbits); | |
142 | if (page_no) { /* exclude the header page */ | |
143 | if (first_block < lowest_block) | |
144 | lowest_block = first_block; | |
145 | if (first_block > highest_block) | |
146 | highest_block = first_block; | |
147 | } | |
148 | ||
149 | /* | |
150 | * We found a PAGE_SIZE-length, PAGE_SIZE-aligned run of blocks | |
151 | */ | |
152 | ret = add_swap_extent(sis, page_no, 1, first_block); | |
153 | if (ret < 0) | |
154 | goto out; | |
155 | nr_extents += ret; | |
156 | page_no++; | |
157 | probe_block += blocks_per_page; | |
158 | reprobe: | |
159 | continue; | |
160 | } | |
161 | ret = nr_extents; | |
162 | *span = 1 + highest_block - lowest_block; | |
163 | if (page_no == 0) | |
164 | page_no = 1; /* force Empty message */ | |
165 | sis->max = page_no; | |
166 | sis->pages = page_no - 1; | |
167 | sis->highest_bit = page_no - 1; | |
168 | out: | |
169 | return ret; | |
170 | bad_bmap: | |
1170532b | 171 | pr_err("swapon: swapfile has holes\n"); |
a509bc1a MG |
172 | ret = -EINVAL; |
173 | goto out; | |
174 | } | |
175 | ||
1da177e4 LT |
176 | /* |
177 | * We may have stale swap cache pages in memory: notice | |
178 | * them here and get rid of the unnecessary final write. | |
179 | */ | |
180 | int swap_writepage(struct page *page, struct writeback_control *wbc) | |
181 | { | |
2f772e6c | 182 | int ret = 0; |
1da177e4 | 183 | |
a2c43eed | 184 | if (try_to_free_swap(page)) { |
1da177e4 LT |
185 | unlock_page(page); |
186 | goto out; | |
187 | } | |
8a84802e SP |
188 | /* |
189 | * Arch code may have to preserve more data than just the page | |
190 | * contents, e.g. memory tags. | |
191 | */ | |
192 | ret = arch_prepare_to_swap(page); | |
193 | if (ret) { | |
194 | set_page_dirty(page); | |
195 | unlock_page(page); | |
196 | goto out; | |
197 | } | |
165c8aed | 198 | if (frontswap_store(page) == 0) { |
38b5faf4 DM |
199 | set_page_writeback(page); |
200 | unlock_page(page); | |
201 | end_page_writeback(page); | |
202 | goto out; | |
203 | } | |
1eec6702 | 204 | ret = __swap_writepage(page, wbc, end_swap_bio_write); |
2f772e6c SJ |
205 | out: |
206 | return ret; | |
207 | } | |
208 | ||
225311a4 HY |
209 | static inline void count_swpout_vm_event(struct page *page) |
210 | { | |
211 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE | |
212 | if (unlikely(PageTransHuge(page))) | |
213 | count_vm_event(THP_SWPOUT); | |
214 | #endif | |
6c357848 | 215 | count_vm_events(PSWPOUT, thp_nr_pages(page)); |
225311a4 HY |
216 | } |
217 | ||
a18b9b15 CH |
218 | #if defined(CONFIG_MEMCG) && defined(CONFIG_BLK_CGROUP) |
219 | static void bio_associate_blkg_from_page(struct bio *bio, struct page *page) | |
220 | { | |
221 | struct cgroup_subsys_state *css; | |
bcfe06bf | 222 | struct mem_cgroup *memcg; |
a18b9b15 | 223 | |
bcfe06bf RG |
224 | memcg = page_memcg(page); |
225 | if (!memcg) | |
a18b9b15 CH |
226 | return; |
227 | ||
228 | rcu_read_lock(); | |
bcfe06bf | 229 | css = cgroup_e_css(memcg->css.cgroup, &io_cgrp_subsys); |
a18b9b15 CH |
230 | bio_associate_blkg_from_css(bio, css); |
231 | rcu_read_unlock(); | |
232 | } | |
233 | #else | |
234 | #define bio_associate_blkg_from_page(bio, page) do { } while (0) | |
235 | #endif /* CONFIG_MEMCG && CONFIG_BLK_CGROUP */ | |
236 | ||
1eec6702 | 237 | int __swap_writepage(struct page *page, struct writeback_control *wbc, |
4246a0b6 | 238 | bio_end_io_t end_write_func) |
2f772e6c SJ |
239 | { |
240 | struct bio *bio; | |
4e49ea4a | 241 | int ret; |
2f772e6c | 242 | struct swap_info_struct *sis = page_swap_info(page); |
62c230bc | 243 | |
cc30c5d6 | 244 | VM_BUG_ON_PAGE(!PageSwapCache(page), page); |
32646315 | 245 | if (data_race(sis->flags & SWP_FS_OPS)) { |
62c230bc MG |
246 | struct kiocb kiocb; |
247 | struct file *swap_file = sis->swap_file; | |
248 | struct address_space *mapping = swap_file->f_mapping; | |
62a8067a AV |
249 | struct bio_vec bv = { |
250 | .bv_page = page, | |
251 | .bv_len = PAGE_SIZE, | |
252 | .bv_offset = 0 | |
253 | }; | |
05afcb77 | 254 | struct iov_iter from; |
62c230bc | 255 | |
aa563d7b | 256 | iov_iter_bvec(&from, WRITE, &bv, 1, PAGE_SIZE); |
62c230bc MG |
257 | init_sync_kiocb(&kiocb, swap_file); |
258 | kiocb.ki_pos = page_file_offset(page); | |
62c230bc | 259 | |
0cdc444a | 260 | set_page_writeback(page); |
62c230bc | 261 | unlock_page(page); |
c8b8e32d | 262 | ret = mapping->a_ops->direct_IO(&kiocb, &from); |
62c230bc MG |
263 | if (ret == PAGE_SIZE) { |
264 | count_vm_event(PSWPOUT); | |
265 | ret = 0; | |
2d30d31e | 266 | } else { |
0cdc444a MG |
267 | /* |
268 | * In the case of swap-over-nfs, this can be a | |
269 | * temporary failure if the system has limited | |
270 | * memory for allocating transmit buffers. | |
271 | * Mark the page dirty and avoid | |
575ced1c | 272 | * folio_rotate_reclaimable but rate-limit the |
0cdc444a MG |
273 | * messages but do not flag PageError like |
274 | * the normal direct-to-bio case as it could | |
275 | * be temporary. | |
276 | */ | |
2d30d31e | 277 | set_page_dirty(page); |
0cdc444a | 278 | ClearPageReclaim(page); |
1170532b JP |
279 | pr_err_ratelimited("Write error on dio swapfile (%llu)\n", |
280 | page_file_offset(page)); | |
62c230bc | 281 | } |
0cdc444a | 282 | end_page_writeback(page); |
62c230bc MG |
283 | return ret; |
284 | } | |
285 | ||
dd6bd0d9 MW |
286 | ret = bdev_write_page(sis->bdev, swap_page_sector(page), page, wbc); |
287 | if (!ret) { | |
225311a4 | 288 | count_swpout_vm_event(page); |
dd6bd0d9 MW |
289 | return 0; |
290 | } | |
291 | ||
07888c66 CH |
292 | bio = bio_alloc(sis->bdev, 1, |
293 | REQ_OP_WRITE | REQ_SWAP | wbc_to_write_flags(wbc), | |
294 | GFP_NOIO); | |
48d15436 | 295 | bio->bi_iter.bi_sector = swap_page_sector(page); |
48d15436 CH |
296 | bio->bi_end_io = end_write_func; |
297 | bio_add_page(bio, page, thp_size(page), 0); | |
298 | ||
6a7f6d86 | 299 | bio_associate_blkg_from_page(bio, page); |
225311a4 | 300 | count_swpout_vm_event(page); |
1da177e4 LT |
301 | set_page_writeback(page); |
302 | unlock_page(page); | |
4e49ea4a | 303 | submit_bio(bio); |
548d9782 ML |
304 | |
305 | return 0; | |
1da177e4 LT |
306 | } |
307 | ||
0bcac06f | 308 | int swap_readpage(struct page *page, bool synchronous) |
1da177e4 LT |
309 | { |
310 | struct bio *bio; | |
311 | int ret = 0; | |
62c230bc | 312 | struct swap_info_struct *sis = page_swap_info(page); |
d8c47cc7 | 313 | bool workingset = PageWorkingset(page); |
93779069 | 314 | unsigned long pflags; |
1da177e4 | 315 | |
0bcac06f | 316 | VM_BUG_ON_PAGE(!PageSwapCache(page) && !synchronous, page); |
309381fe SL |
317 | VM_BUG_ON_PAGE(!PageLocked(page), page); |
318 | VM_BUG_ON_PAGE(PageUptodate(page), page); | |
93779069 MK |
319 | |
320 | /* | |
321 | * Count submission time as memory stall. When the device is congested, | |
322 | * or the submitting cgroup IO-throttled, submission can be a | |
323 | * significant part of overall IO time. | |
324 | */ | |
d8c47cc7 JW |
325 | if (workingset) |
326 | psi_memstall_enter(&pflags); | |
a3d5dc90 | 327 | delayacct_swapin_start(); |
93779069 | 328 | |
165c8aed | 329 | if (frontswap_load(page) == 0) { |
38b5faf4 DM |
330 | SetPageUptodate(page); |
331 | unlock_page(page); | |
332 | goto out; | |
333 | } | |
62c230bc | 334 | |
32646315 | 335 | if (data_race(sis->flags & SWP_FS_OPS)) { |
62c230bc MG |
336 | struct file *swap_file = sis->swap_file; |
337 | struct address_space *mapping = swap_file->f_mapping; | |
338 | ||
339 | ret = mapping->a_ops->readpage(swap_file, page); | |
340 | if (!ret) | |
341 | count_vm_event(PSWPIN); | |
93779069 | 342 | goto out; |
62c230bc MG |
343 | } |
344 | ||
5115db10 CH |
345 | if (sis->flags & SWP_SYNCHRONOUS_IO) { |
346 | ret = bdev_read_page(sis->bdev, swap_page_sector(page), page); | |
347 | if (!ret) { | |
5115db10 CH |
348 | count_vm_event(PSWPIN); |
349 | goto out; | |
350 | } | |
dd6bd0d9 MW |
351 | } |
352 | ||
353 | ret = 0; | |
07888c66 | 354 | bio = bio_alloc(sis->bdev, 1, REQ_OP_READ, GFP_KERNEL); |
48d15436 CH |
355 | bio->bi_iter.bi_sector = swap_page_sector(page); |
356 | bio->bi_end_io = end_swap_bio_read; | |
357 | bio_add_page(bio, page, thp_size(page), 0); | |
b0ba2d0f TH |
358 | /* |
359 | * Keep this task valid during swap readpage because the oom killer may | |
360 | * attempt to access it in the page fault retry time check. | |
361 | */ | |
87518530 | 362 | if (synchronous) { |
6ce913fe | 363 | bio->bi_opf |= REQ_POLLED; |
87518530 ON |
364 | get_task_struct(current); |
365 | bio->bi_private = current; | |
366 | } | |
f8891e5e | 367 | count_vm_event(PSWPIN); |
23955622 | 368 | bio_get(bio); |
3e08773c | 369 | submit_bio(bio); |
0bcac06f | 370 | while (synchronous) { |
1ac5cd49 | 371 | set_current_state(TASK_UNINTERRUPTIBLE); |
23955622 SL |
372 | if (!READ_ONCE(bio->bi_private)) |
373 | break; | |
374 | ||
5a72e899 | 375 | if (!bio_poll(bio, NULL, 0)) |
0f190a7a | 376 | blk_io_schedule(); |
23955622 SL |
377 | } |
378 | __set_current_state(TASK_RUNNING); | |
379 | bio_put(bio); | |
380 | ||
1da177e4 | 381 | out: |
d8c47cc7 JW |
382 | if (workingset) |
383 | psi_memstall_leave(&pflags); | |
a3d5dc90 | 384 | delayacct_swapin_end(); |
1da177e4 LT |
385 | return ret; |
386 | } | |
62c230bc | 387 | |
7e63df00 | 388 | bool swap_dirty_folio(struct address_space *mapping, struct folio *folio) |
62c230bc | 389 | { |
7e63df00 | 390 | struct swap_info_struct *sis = swp_swap_info(folio_swap_entry(folio)); |
62c230bc | 391 | |
32646315 | 392 | if (data_race(sis->flags & SWP_FS_OPS)) { |
7e63df00 MWO |
393 | const struct address_space_operations *aops; |
394 | ||
395 | mapping = sis->swap_file->f_mapping; | |
396 | aops = mapping->a_ops; | |
cc30c5d6 | 397 | |
7e63df00 | 398 | VM_BUG_ON_FOLIO(!folio_test_swapcache(folio), folio); |
3a3bae50 | 399 | return aops->dirty_folio(mapping, folio); |
62c230bc | 400 | } else { |
46de8b97 | 401 | return noop_dirty_folio(mapping, folio); |
62c230bc MG |
402 | } |
403 | } |