]> git.ipfire.org Git - thirdparty/linux.git/blob - include/linux/backing-dev.h
io_uring: reset -EBUSY error when io sq thread is waken up
[thirdparty/linux.git] / include / linux / backing-dev.h
1 /* SPDX-License-Identifier: GPL-2.0 */
2 /*
3 * include/linux/backing-dev.h
4 *
5 * low-level device information and state which is propagated up through
6 * to high-level code.
7 */
8
9 #ifndef _LINUX_BACKING_DEV_H
10 #define _LINUX_BACKING_DEV_H
11
12 #include <linux/kernel.h>
13 #include <linux/fs.h>
14 #include <linux/sched.h>
15 #include <linux/blkdev.h>
16 #include <linux/device.h>
17 #include <linux/writeback.h>
18 #include <linux/blk-cgroup.h>
19 #include <linux/backing-dev-defs.h>
20 #include <linux/slab.h>
21
22 static inline struct backing_dev_info *bdi_get(struct backing_dev_info *bdi)
23 {
24 kref_get(&bdi->refcnt);
25 return bdi;
26 }
27
28 struct backing_dev_info *bdi_get_by_id(u64 id);
29 void bdi_put(struct backing_dev_info *bdi);
30
31 __printf(2, 3)
32 int bdi_register(struct backing_dev_info *bdi, const char *fmt, ...);
33 __printf(2, 0)
34 int bdi_register_va(struct backing_dev_info *bdi, const char *fmt,
35 va_list args);
36 int bdi_register_owner(struct backing_dev_info *bdi, struct device *owner);
37 void bdi_unregister(struct backing_dev_info *bdi);
38
39 struct backing_dev_info *bdi_alloc_node(gfp_t gfp_mask, int node_id);
40 static inline struct backing_dev_info *bdi_alloc(gfp_t gfp_mask)
41 {
42 return bdi_alloc_node(gfp_mask, NUMA_NO_NODE);
43 }
44
45 void wb_start_background_writeback(struct bdi_writeback *wb);
46 void wb_workfn(struct work_struct *work);
47 void wb_wakeup_delayed(struct bdi_writeback *wb);
48
49 void wb_wait_for_completion(struct wb_completion *done);
50
51 extern spinlock_t bdi_lock;
52 extern struct list_head bdi_list;
53
54 extern struct workqueue_struct *bdi_wq;
55 extern struct workqueue_struct *bdi_async_bio_wq;
56
57 static inline bool wb_has_dirty_io(struct bdi_writeback *wb)
58 {
59 return test_bit(WB_has_dirty_io, &wb->state);
60 }
61
62 static inline bool bdi_has_dirty_io(struct backing_dev_info *bdi)
63 {
64 /*
65 * @bdi->tot_write_bandwidth is guaranteed to be > 0 if there are
66 * any dirty wbs. See wb_update_write_bandwidth().
67 */
68 return atomic_long_read(&bdi->tot_write_bandwidth);
69 }
70
71 static inline void __add_wb_stat(struct bdi_writeback *wb,
72 enum wb_stat_item item, s64 amount)
73 {
74 percpu_counter_add_batch(&wb->stat[item], amount, WB_STAT_BATCH);
75 }
76
77 static inline void inc_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
78 {
79 __add_wb_stat(wb, item, 1);
80 }
81
82 static inline void dec_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
83 {
84 __add_wb_stat(wb, item, -1);
85 }
86
87 static inline s64 wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
88 {
89 return percpu_counter_read_positive(&wb->stat[item]);
90 }
91
92 static inline s64 wb_stat_sum(struct bdi_writeback *wb, enum wb_stat_item item)
93 {
94 return percpu_counter_sum_positive(&wb->stat[item]);
95 }
96
97 extern void wb_writeout_inc(struct bdi_writeback *wb);
98
99 /*
100 * maximal error of a stat counter.
101 */
102 static inline unsigned long wb_stat_error(void)
103 {
104 #ifdef CONFIG_SMP
105 return nr_cpu_ids * WB_STAT_BATCH;
106 #else
107 return 1;
108 #endif
109 }
110
111 int bdi_set_min_ratio(struct backing_dev_info *bdi, unsigned int min_ratio);
112 int bdi_set_max_ratio(struct backing_dev_info *bdi, unsigned int max_ratio);
113
114 /*
115 * Flags in backing_dev_info::capability
116 *
117 * The first three flags control whether dirty pages will contribute to the
118 * VM's accounting and whether writepages() should be called for dirty pages
119 * (something that would not, for example, be appropriate for ramfs)
120 *
121 * WARNING: these flags are closely related and should not normally be
122 * used separately. The BDI_CAP_NO_ACCT_AND_WRITEBACK combines these
123 * three flags into a single convenience macro.
124 *
125 * BDI_CAP_NO_ACCT_DIRTY: Dirty pages shouldn't contribute to accounting
126 * BDI_CAP_NO_WRITEBACK: Don't write pages back
127 * BDI_CAP_NO_ACCT_WB: Don't automatically account writeback pages
128 * BDI_CAP_STRICTLIMIT: Keep number of dirty pages below bdi threshold.
129 *
130 * BDI_CAP_CGROUP_WRITEBACK: Supports cgroup-aware writeback.
131 * BDI_CAP_SYNCHRONOUS_IO: Device is so fast that asynchronous IO would be
132 * inefficient.
133 */
134 #define BDI_CAP_NO_ACCT_DIRTY 0x00000001
135 #define BDI_CAP_NO_WRITEBACK 0x00000002
136 #define BDI_CAP_NO_ACCT_WB 0x00000004
137 #define BDI_CAP_STABLE_WRITES 0x00000008
138 #define BDI_CAP_STRICTLIMIT 0x00000010
139 #define BDI_CAP_CGROUP_WRITEBACK 0x00000020
140 #define BDI_CAP_SYNCHRONOUS_IO 0x00000040
141
142 #define BDI_CAP_NO_ACCT_AND_WRITEBACK \
143 (BDI_CAP_NO_WRITEBACK | BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_ACCT_WB)
144
145 extern struct backing_dev_info noop_backing_dev_info;
146
147 /**
148 * writeback_in_progress - determine whether there is writeback in progress
149 * @wb: bdi_writeback of interest
150 *
151 * Determine whether there is writeback waiting to be handled against a
152 * bdi_writeback.
153 */
154 static inline bool writeback_in_progress(struct bdi_writeback *wb)
155 {
156 return test_bit(WB_writeback_running, &wb->state);
157 }
158
159 static inline struct backing_dev_info *inode_to_bdi(struct inode *inode)
160 {
161 struct super_block *sb;
162
163 if (!inode)
164 return &noop_backing_dev_info;
165
166 sb = inode->i_sb;
167 #ifdef CONFIG_BLOCK
168 if (sb_is_blkdev_sb(sb))
169 return I_BDEV(inode)->bd_bdi;
170 #endif
171 return sb->s_bdi;
172 }
173
174 static inline int wb_congested(struct bdi_writeback *wb, int cong_bits)
175 {
176 struct backing_dev_info *bdi = wb->bdi;
177
178 if (bdi->congested_fn)
179 return bdi->congested_fn(bdi->congested_data, cong_bits);
180 return wb->congested->state & cong_bits;
181 }
182
183 long congestion_wait(int sync, long timeout);
184 long wait_iff_congested(int sync, long timeout);
185
186 static inline bool bdi_cap_synchronous_io(struct backing_dev_info *bdi)
187 {
188 return bdi->capabilities & BDI_CAP_SYNCHRONOUS_IO;
189 }
190
191 static inline bool bdi_cap_stable_pages_required(struct backing_dev_info *bdi)
192 {
193 return bdi->capabilities & BDI_CAP_STABLE_WRITES;
194 }
195
196 static inline bool bdi_cap_writeback_dirty(struct backing_dev_info *bdi)
197 {
198 return !(bdi->capabilities & BDI_CAP_NO_WRITEBACK);
199 }
200
201 static inline bool bdi_cap_account_dirty(struct backing_dev_info *bdi)
202 {
203 return !(bdi->capabilities & BDI_CAP_NO_ACCT_DIRTY);
204 }
205
206 static inline bool bdi_cap_account_writeback(struct backing_dev_info *bdi)
207 {
208 /* Paranoia: BDI_CAP_NO_WRITEBACK implies BDI_CAP_NO_ACCT_WB */
209 return !(bdi->capabilities & (BDI_CAP_NO_ACCT_WB |
210 BDI_CAP_NO_WRITEBACK));
211 }
212
213 static inline bool mapping_cap_writeback_dirty(struct address_space *mapping)
214 {
215 return bdi_cap_writeback_dirty(inode_to_bdi(mapping->host));
216 }
217
218 static inline bool mapping_cap_account_dirty(struct address_space *mapping)
219 {
220 return bdi_cap_account_dirty(inode_to_bdi(mapping->host));
221 }
222
223 static inline int bdi_sched_wait(void *word)
224 {
225 schedule();
226 return 0;
227 }
228
229 #ifdef CONFIG_CGROUP_WRITEBACK
230
231 struct bdi_writeback_congested *
232 wb_congested_get_create(struct backing_dev_info *bdi, int blkcg_id, gfp_t gfp);
233 void wb_congested_put(struct bdi_writeback_congested *congested);
234 struct bdi_writeback *wb_get_lookup(struct backing_dev_info *bdi,
235 struct cgroup_subsys_state *memcg_css);
236 struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi,
237 struct cgroup_subsys_state *memcg_css,
238 gfp_t gfp);
239 void wb_memcg_offline(struct mem_cgroup *memcg);
240 void wb_blkcg_offline(struct blkcg *blkcg);
241 int inode_congested(struct inode *inode, int cong_bits);
242
243 /**
244 * inode_cgwb_enabled - test whether cgroup writeback is enabled on an inode
245 * @inode: inode of interest
246 *
247 * cgroup writeback requires support from both the bdi and filesystem.
248 * Also, both memcg and iocg have to be on the default hierarchy. Test
249 * whether all conditions are met.
250 *
251 * Note that the test result may change dynamically on the same inode
252 * depending on how memcg and iocg are configured.
253 */
254 static inline bool inode_cgwb_enabled(struct inode *inode)
255 {
256 struct backing_dev_info *bdi = inode_to_bdi(inode);
257
258 return cgroup_subsys_on_dfl(memory_cgrp_subsys) &&
259 cgroup_subsys_on_dfl(io_cgrp_subsys) &&
260 bdi_cap_account_dirty(bdi) &&
261 (bdi->capabilities & BDI_CAP_CGROUP_WRITEBACK) &&
262 (inode->i_sb->s_iflags & SB_I_CGROUPWB);
263 }
264
265 /**
266 * wb_find_current - find wb for %current on a bdi
267 * @bdi: bdi of interest
268 *
269 * Find the wb of @bdi which matches both the memcg and blkcg of %current.
270 * Must be called under rcu_read_lock() which protects the returend wb.
271 * NULL if not found.
272 */
273 static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
274 {
275 struct cgroup_subsys_state *memcg_css;
276 struct bdi_writeback *wb;
277
278 memcg_css = task_css(current, memory_cgrp_id);
279 if (!memcg_css->parent)
280 return &bdi->wb;
281
282 wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
283
284 /*
285 * %current's blkcg equals the effective blkcg of its memcg. No
286 * need to use the relatively expensive cgroup_get_e_css().
287 */
288 if (likely(wb && wb->blkcg_css == task_css(current, io_cgrp_id)))
289 return wb;
290 return NULL;
291 }
292
293 /**
294 * wb_get_create_current - get or create wb for %current on a bdi
295 * @bdi: bdi of interest
296 * @gfp: allocation mask
297 *
298 * Equivalent to wb_get_create() on %current's memcg. This function is
299 * called from a relatively hot path and optimizes the common cases using
300 * wb_find_current().
301 */
302 static inline struct bdi_writeback *
303 wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
304 {
305 struct bdi_writeback *wb;
306
307 rcu_read_lock();
308 wb = wb_find_current(bdi);
309 if (wb && unlikely(!wb_tryget(wb)))
310 wb = NULL;
311 rcu_read_unlock();
312
313 if (unlikely(!wb)) {
314 struct cgroup_subsys_state *memcg_css;
315
316 memcg_css = task_get_css(current, memory_cgrp_id);
317 wb = wb_get_create(bdi, memcg_css, gfp);
318 css_put(memcg_css);
319 }
320 return wb;
321 }
322
323 /**
324 * inode_to_wb_is_valid - test whether an inode has a wb associated
325 * @inode: inode of interest
326 *
327 * Returns %true if @inode has a wb associated. May be called without any
328 * locking.
329 */
330 static inline bool inode_to_wb_is_valid(struct inode *inode)
331 {
332 return inode->i_wb;
333 }
334
335 /**
336 * inode_to_wb - determine the wb of an inode
337 * @inode: inode of interest
338 *
339 * Returns the wb @inode is currently associated with. The caller must be
340 * holding either @inode->i_lock, the i_pages lock, or the
341 * associated wb's list_lock.
342 */
343 static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
344 {
345 #ifdef CONFIG_LOCKDEP
346 WARN_ON_ONCE(debug_locks &&
347 (!lockdep_is_held(&inode->i_lock) &&
348 !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
349 !lockdep_is_held(&inode->i_wb->list_lock)));
350 #endif
351 return inode->i_wb;
352 }
353
354 /**
355 * unlocked_inode_to_wb_begin - begin unlocked inode wb access transaction
356 * @inode: target inode
357 * @cookie: output param, to be passed to the end function
358 *
359 * The caller wants to access the wb associated with @inode but isn't
360 * holding inode->i_lock, the i_pages lock or wb->list_lock. This
361 * function determines the wb associated with @inode and ensures that the
362 * association doesn't change until the transaction is finished with
363 * unlocked_inode_to_wb_end().
364 *
365 * The caller must call unlocked_inode_to_wb_end() with *@cookie afterwards and
366 * can't sleep during the transaction. IRQs may or may not be disabled on
367 * return.
368 */
369 static inline struct bdi_writeback *
370 unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
371 {
372 rcu_read_lock();
373
374 /*
375 * Paired with store_release in inode_switch_wbs_work_fn() and
376 * ensures that we see the new wb if we see cleared I_WB_SWITCH.
377 */
378 cookie->locked = smp_load_acquire(&inode->i_state) & I_WB_SWITCH;
379
380 if (unlikely(cookie->locked))
381 xa_lock_irqsave(&inode->i_mapping->i_pages, cookie->flags);
382
383 /*
384 * Protected by either !I_WB_SWITCH + rcu_read_lock() or the i_pages
385 * lock. inode_to_wb() will bark. Deref directly.
386 */
387 return inode->i_wb;
388 }
389
390 /**
391 * unlocked_inode_to_wb_end - end inode wb access transaction
392 * @inode: target inode
393 * @cookie: @cookie from unlocked_inode_to_wb_begin()
394 */
395 static inline void unlocked_inode_to_wb_end(struct inode *inode,
396 struct wb_lock_cookie *cookie)
397 {
398 if (unlikely(cookie->locked))
399 xa_unlock_irqrestore(&inode->i_mapping->i_pages, cookie->flags);
400
401 rcu_read_unlock();
402 }
403
404 #else /* CONFIG_CGROUP_WRITEBACK */
405
406 static inline bool inode_cgwb_enabled(struct inode *inode)
407 {
408 return false;
409 }
410
411 static inline struct bdi_writeback_congested *
412 wb_congested_get_create(struct backing_dev_info *bdi, int blkcg_id, gfp_t gfp)
413 {
414 refcount_inc(&bdi->wb_congested->refcnt);
415 return bdi->wb_congested;
416 }
417
418 static inline void wb_congested_put(struct bdi_writeback_congested *congested)
419 {
420 if (refcount_dec_and_test(&congested->refcnt))
421 kfree(congested);
422 }
423
424 static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
425 {
426 return &bdi->wb;
427 }
428
429 static inline struct bdi_writeback *
430 wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
431 {
432 return &bdi->wb;
433 }
434
435 static inline bool inode_to_wb_is_valid(struct inode *inode)
436 {
437 return true;
438 }
439
440 static inline struct bdi_writeback *inode_to_wb(struct inode *inode)
441 {
442 return &inode_to_bdi(inode)->wb;
443 }
444
445 static inline struct bdi_writeback *
446 unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
447 {
448 return inode_to_wb(inode);
449 }
450
451 static inline void unlocked_inode_to_wb_end(struct inode *inode,
452 struct wb_lock_cookie *cookie)
453 {
454 }
455
456 static inline void wb_memcg_offline(struct mem_cgroup *memcg)
457 {
458 }
459
460 static inline void wb_blkcg_offline(struct blkcg *blkcg)
461 {
462 }
463
464 static inline int inode_congested(struct inode *inode, int cong_bits)
465 {
466 return wb_congested(&inode_to_bdi(inode)->wb, cong_bits);
467 }
468
469 #endif /* CONFIG_CGROUP_WRITEBACK */
470
471 static inline int inode_read_congested(struct inode *inode)
472 {
473 return inode_congested(inode, 1 << WB_sync_congested);
474 }
475
476 static inline int inode_write_congested(struct inode *inode)
477 {
478 return inode_congested(inode, 1 << WB_async_congested);
479 }
480
481 static inline int inode_rw_congested(struct inode *inode)
482 {
483 return inode_congested(inode, (1 << WB_sync_congested) |
484 (1 << WB_async_congested));
485 }
486
487 static inline int bdi_congested(struct backing_dev_info *bdi, int cong_bits)
488 {
489 return wb_congested(&bdi->wb, cong_bits);
490 }
491
492 static inline int bdi_read_congested(struct backing_dev_info *bdi)
493 {
494 return bdi_congested(bdi, 1 << WB_sync_congested);
495 }
496
497 static inline int bdi_write_congested(struct backing_dev_info *bdi)
498 {
499 return bdi_congested(bdi, 1 << WB_async_congested);
500 }
501
502 static inline int bdi_rw_congested(struct backing_dev_info *bdi)
503 {
504 return bdi_congested(bdi, (1 << WB_sync_congested) |
505 (1 << WB_async_congested));
506 }
507
508 extern const char *bdi_unknown_name;
509
510 static inline const char *bdi_dev_name(struct backing_dev_info *bdi)
511 {
512 if (!bdi || !bdi->dev)
513 return bdi_unknown_name;
514 return dev_name(bdi->dev);
515 }
516
517 #endif /* _LINUX_BACKING_DEV_H */