]> git.proxmox.com Git - mirror_ubuntu-kernels.git/blob - include/linux/backing-dev.h
Merge tag 'for-5.16/block-2021-10-29' of git://git.kernel.dk/linux-block
[mirror_ubuntu-kernels.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/device.h>
16 #include <linux/writeback.h>
17 #include <linux/backing-dev-defs.h>
18 #include <linux/slab.h>
19
20 struct blkcg;
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 void bdi_set_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(int node_id);
40
41 void wb_start_background_writeback(struct bdi_writeback *wb);
42 void wb_workfn(struct work_struct *work);
43 void wb_wakeup_delayed(struct bdi_writeback *wb);
44
45 void wb_wait_for_completion(struct wb_completion *done);
46
47 extern spinlock_t bdi_lock;
48 extern struct list_head bdi_list;
49
50 extern struct workqueue_struct *bdi_wq;
51 extern struct workqueue_struct *bdi_async_bio_wq;
52
53 static inline bool wb_has_dirty_io(struct bdi_writeback *wb)
54 {
55 return test_bit(WB_has_dirty_io, &wb->state);
56 }
57
58 static inline bool bdi_has_dirty_io(struct backing_dev_info *bdi)
59 {
60 /*
61 * @bdi->tot_write_bandwidth is guaranteed to be > 0 if there are
62 * any dirty wbs. See wb_update_write_bandwidth().
63 */
64 return atomic_long_read(&bdi->tot_write_bandwidth);
65 }
66
67 static inline void wb_stat_mod(struct bdi_writeback *wb,
68 enum wb_stat_item item, s64 amount)
69 {
70 percpu_counter_add_batch(&wb->stat[item], amount, WB_STAT_BATCH);
71 }
72
73 static inline void inc_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
74 {
75 wb_stat_mod(wb, item, 1);
76 }
77
78 static inline void dec_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
79 {
80 wb_stat_mod(wb, item, -1);
81 }
82
83 static inline s64 wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
84 {
85 return percpu_counter_read_positive(&wb->stat[item]);
86 }
87
88 static inline s64 wb_stat_sum(struct bdi_writeback *wb, enum wb_stat_item item)
89 {
90 return percpu_counter_sum_positive(&wb->stat[item]);
91 }
92
93 extern void wb_writeout_inc(struct bdi_writeback *wb);
94
95 /*
96 * maximal error of a stat counter.
97 */
98 static inline unsigned long wb_stat_error(void)
99 {
100 #ifdef CONFIG_SMP
101 return nr_cpu_ids * WB_STAT_BATCH;
102 #else
103 return 1;
104 #endif
105 }
106
107 int bdi_set_min_ratio(struct backing_dev_info *bdi, unsigned int min_ratio);
108 int bdi_set_max_ratio(struct backing_dev_info *bdi, unsigned int max_ratio);
109
110 /*
111 * Flags in backing_dev_info::capability
112 *
113 * BDI_CAP_WRITEBACK: Supports dirty page writeback, and dirty pages
114 * should contribute to accounting
115 * BDI_CAP_WRITEBACK_ACCT: Automatically account writeback pages
116 * BDI_CAP_STRICTLIMIT: Keep number of dirty pages below bdi threshold
117 */
118 #define BDI_CAP_WRITEBACK (1 << 0)
119 #define BDI_CAP_WRITEBACK_ACCT (1 << 1)
120 #define BDI_CAP_STRICTLIMIT (1 << 2)
121
122 extern struct backing_dev_info noop_backing_dev_info;
123
124 /**
125 * writeback_in_progress - determine whether there is writeback in progress
126 * @wb: bdi_writeback of interest
127 *
128 * Determine whether there is writeback waiting to be handled against a
129 * bdi_writeback.
130 */
131 static inline bool writeback_in_progress(struct bdi_writeback *wb)
132 {
133 return test_bit(WB_writeback_running, &wb->state);
134 }
135
136 struct backing_dev_info *inode_to_bdi(struct inode *inode);
137
138 static inline int wb_congested(struct bdi_writeback *wb, int cong_bits)
139 {
140 return wb->congested & cong_bits;
141 }
142
143 long congestion_wait(int sync, long timeout);
144 long wait_iff_congested(int sync, long timeout);
145
146 static inline bool mapping_can_writeback(struct address_space *mapping)
147 {
148 return inode_to_bdi(mapping->host)->capabilities & BDI_CAP_WRITEBACK;
149 }
150
151 static inline int bdi_sched_wait(void *word)
152 {
153 schedule();
154 return 0;
155 }
156
157 #ifdef CONFIG_CGROUP_WRITEBACK
158
159 struct bdi_writeback *wb_get_lookup(struct backing_dev_info *bdi,
160 struct cgroup_subsys_state *memcg_css);
161 struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi,
162 struct cgroup_subsys_state *memcg_css,
163 gfp_t gfp);
164 void wb_memcg_offline(struct mem_cgroup *memcg);
165 void wb_blkcg_offline(struct blkcg *blkcg);
166 int inode_congested(struct inode *inode, int cong_bits);
167
168 /**
169 * inode_cgwb_enabled - test whether cgroup writeback is enabled on an inode
170 * @inode: inode of interest
171 *
172 * Cgroup writeback requires support from the filesystem. Also, both memcg and
173 * iocg have to be on the default hierarchy. Test whether all conditions are
174 * met.
175 *
176 * Note that the test result may change dynamically on the same inode
177 * depending on how memcg and iocg are configured.
178 */
179 static inline bool inode_cgwb_enabled(struct inode *inode)
180 {
181 struct backing_dev_info *bdi = inode_to_bdi(inode);
182
183 return cgroup_subsys_on_dfl(memory_cgrp_subsys) &&
184 cgroup_subsys_on_dfl(io_cgrp_subsys) &&
185 (bdi->capabilities & BDI_CAP_WRITEBACK) &&
186 (inode->i_sb->s_iflags & SB_I_CGROUPWB);
187 }
188
189 /**
190 * wb_find_current - find wb for %current on a bdi
191 * @bdi: bdi of interest
192 *
193 * Find the wb of @bdi which matches both the memcg and blkcg of %current.
194 * Must be called under rcu_read_lock() which protects the returend wb.
195 * NULL if not found.
196 */
197 static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
198 {
199 struct cgroup_subsys_state *memcg_css;
200 struct bdi_writeback *wb;
201
202 memcg_css = task_css(current, memory_cgrp_id);
203 if (!memcg_css->parent)
204 return &bdi->wb;
205
206 wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
207
208 /*
209 * %current's blkcg equals the effective blkcg of its memcg. No
210 * need to use the relatively expensive cgroup_get_e_css().
211 */
212 if (likely(wb && wb->blkcg_css == task_css(current, io_cgrp_id)))
213 return wb;
214 return NULL;
215 }
216
217 /**
218 * wb_get_create_current - get or create wb for %current on a bdi
219 * @bdi: bdi of interest
220 * @gfp: allocation mask
221 *
222 * Equivalent to wb_get_create() on %current's memcg. This function is
223 * called from a relatively hot path and optimizes the common cases using
224 * wb_find_current().
225 */
226 static inline struct bdi_writeback *
227 wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
228 {
229 struct bdi_writeback *wb;
230
231 rcu_read_lock();
232 wb = wb_find_current(bdi);
233 if (wb && unlikely(!wb_tryget(wb)))
234 wb = NULL;
235 rcu_read_unlock();
236
237 if (unlikely(!wb)) {
238 struct cgroup_subsys_state *memcg_css;
239
240 memcg_css = task_get_css(current, memory_cgrp_id);
241 wb = wb_get_create(bdi, memcg_css, gfp);
242 css_put(memcg_css);
243 }
244 return wb;
245 }
246
247 /**
248 * inode_to_wb_is_valid - test whether an inode has a wb associated
249 * @inode: inode of interest
250 *
251 * Returns %true if @inode has a wb associated. May be called without any
252 * locking.
253 */
254 static inline bool inode_to_wb_is_valid(struct inode *inode)
255 {
256 return inode->i_wb;
257 }
258
259 /**
260 * inode_to_wb - determine the wb of an inode
261 * @inode: inode of interest
262 *
263 * Returns the wb @inode is currently associated with. The caller must be
264 * holding either @inode->i_lock, the i_pages lock, or the
265 * associated wb's list_lock.
266 */
267 static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
268 {
269 #ifdef CONFIG_LOCKDEP
270 WARN_ON_ONCE(debug_locks &&
271 (!lockdep_is_held(&inode->i_lock) &&
272 !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
273 !lockdep_is_held(&inode->i_wb->list_lock)));
274 #endif
275 return inode->i_wb;
276 }
277
278 static inline struct bdi_writeback *inode_to_wb_wbc(
279 struct inode *inode,
280 struct writeback_control *wbc)
281 {
282 /*
283 * If wbc does not have inode attached, it means cgroup writeback was
284 * disabled when wbc started. Just use the default wb in that case.
285 */
286 return wbc->wb ? wbc->wb : &inode_to_bdi(inode)->wb;
287 }
288
289 /**
290 * unlocked_inode_to_wb_begin - begin unlocked inode wb access transaction
291 * @inode: target inode
292 * @cookie: output param, to be passed to the end function
293 *
294 * The caller wants to access the wb associated with @inode but isn't
295 * holding inode->i_lock, the i_pages lock or wb->list_lock. This
296 * function determines the wb associated with @inode and ensures that the
297 * association doesn't change until the transaction is finished with
298 * unlocked_inode_to_wb_end().
299 *
300 * The caller must call unlocked_inode_to_wb_end() with *@cookie afterwards and
301 * can't sleep during the transaction. IRQs may or may not be disabled on
302 * return.
303 */
304 static inline struct bdi_writeback *
305 unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
306 {
307 rcu_read_lock();
308
309 /*
310 * Paired with store_release in inode_switch_wbs_work_fn() and
311 * ensures that we see the new wb if we see cleared I_WB_SWITCH.
312 */
313 cookie->locked = smp_load_acquire(&inode->i_state) & I_WB_SWITCH;
314
315 if (unlikely(cookie->locked))
316 xa_lock_irqsave(&inode->i_mapping->i_pages, cookie->flags);
317
318 /*
319 * Protected by either !I_WB_SWITCH + rcu_read_lock() or the i_pages
320 * lock. inode_to_wb() will bark. Deref directly.
321 */
322 return inode->i_wb;
323 }
324
325 /**
326 * unlocked_inode_to_wb_end - end inode wb access transaction
327 * @inode: target inode
328 * @cookie: @cookie from unlocked_inode_to_wb_begin()
329 */
330 static inline void unlocked_inode_to_wb_end(struct inode *inode,
331 struct wb_lock_cookie *cookie)
332 {
333 if (unlikely(cookie->locked))
334 xa_unlock_irqrestore(&inode->i_mapping->i_pages, cookie->flags);
335
336 rcu_read_unlock();
337 }
338
339 #else /* CONFIG_CGROUP_WRITEBACK */
340
341 static inline bool inode_cgwb_enabled(struct inode *inode)
342 {
343 return false;
344 }
345
346 static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
347 {
348 return &bdi->wb;
349 }
350
351 static inline struct bdi_writeback *
352 wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
353 {
354 return &bdi->wb;
355 }
356
357 static inline bool inode_to_wb_is_valid(struct inode *inode)
358 {
359 return true;
360 }
361
362 static inline struct bdi_writeback *inode_to_wb(struct inode *inode)
363 {
364 return &inode_to_bdi(inode)->wb;
365 }
366
367 static inline struct bdi_writeback *inode_to_wb_wbc(
368 struct inode *inode,
369 struct writeback_control *wbc)
370 {
371 return inode_to_wb(inode);
372 }
373
374
375 static inline struct bdi_writeback *
376 unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
377 {
378 return inode_to_wb(inode);
379 }
380
381 static inline void unlocked_inode_to_wb_end(struct inode *inode,
382 struct wb_lock_cookie *cookie)
383 {
384 }
385
386 static inline void wb_memcg_offline(struct mem_cgroup *memcg)
387 {
388 }
389
390 static inline void wb_blkcg_offline(struct blkcg *blkcg)
391 {
392 }
393
394 static inline int inode_congested(struct inode *inode, int cong_bits)
395 {
396 return wb_congested(&inode_to_bdi(inode)->wb, cong_bits);
397 }
398
399 #endif /* CONFIG_CGROUP_WRITEBACK */
400
401 static inline int inode_read_congested(struct inode *inode)
402 {
403 return inode_congested(inode, 1 << WB_sync_congested);
404 }
405
406 static inline int inode_write_congested(struct inode *inode)
407 {
408 return inode_congested(inode, 1 << WB_async_congested);
409 }
410
411 static inline int inode_rw_congested(struct inode *inode)
412 {
413 return inode_congested(inode, (1 << WB_sync_congested) |
414 (1 << WB_async_congested));
415 }
416
417 static inline int bdi_congested(struct backing_dev_info *bdi, int cong_bits)
418 {
419 return wb_congested(&bdi->wb, cong_bits);
420 }
421
422 static inline int bdi_read_congested(struct backing_dev_info *bdi)
423 {
424 return bdi_congested(bdi, 1 << WB_sync_congested);
425 }
426
427 static inline int bdi_write_congested(struct backing_dev_info *bdi)
428 {
429 return bdi_congested(bdi, 1 << WB_async_congested);
430 }
431
432 static inline int bdi_rw_congested(struct backing_dev_info *bdi)
433 {
434 return bdi_congested(bdi, (1 << WB_sync_congested) |
435 (1 << WB_async_congested));
436 }
437
438 const char *bdi_dev_name(struct backing_dev_info *bdi);
439
440 #endif /* _LINUX_BACKING_DEV_H */