Merge tag 'sound-5.16-rc7' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
[linux-block.git] / include / linux / backing-dev.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
1da177e4
LT
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
cf0ca9fe 12#include <linux/kernel.h>
e4ad08fe 13#include <linux/fs.h>
03ba3782 14#include <linux/sched.h>
68f23b89 15#include <linux/device.h>
03ba3782 16#include <linux/writeback.h>
66114cad 17#include <linux/backing-dev-defs.h>
a13f35e8 18#include <linux/slab.h>
de1414a6 19
e41d12f5
CH
20struct blkcg;
21
d03f6cdc
JK
22static inline struct backing_dev_info *bdi_get(struct backing_dev_info *bdi)
23{
24 kref_get(&bdi->refcnt);
25 return bdi;
26}
27
34f8fe50 28struct backing_dev_info *bdi_get_by_id(u64 id);
d03f6cdc 29void bdi_put(struct backing_dev_info *bdi);
b2e8fb6e 30
7c4cc300
JK
31__printf(2, 3)
32int bdi_register(struct backing_dev_info *bdi, const char *fmt, ...);
a93f00b3 33__printf(2, 0)
7c4cc300
JK
34int bdi_register_va(struct backing_dev_info *bdi, const char *fmt,
35 va_list args);
3c5d202b 36void bdi_set_owner(struct backing_dev_info *bdi, struct device *owner);
b02176f3
TH
37void bdi_unregister(struct backing_dev_info *bdi);
38
aef33c2f 39struct backing_dev_info *bdi_alloc(int node_id);
b02176f3 40
9ecf4866 41void wb_start_background_writeback(struct bdi_writeback *wb);
f0054bb1 42void wb_workfn(struct work_struct *work);
f0054bb1 43void wb_wakeup_delayed(struct bdi_writeback *wb);
cf0ca9fe 44
5b9cce4c
TH
45void wb_wait_for_completion(struct wb_completion *done);
46
03ba3782 47extern spinlock_t bdi_lock;
66f3b8e2
JA
48extern struct list_head bdi_list;
49
839a8e86 50extern struct workqueue_struct *bdi_wq;
d3f77dfd 51extern struct workqueue_struct *bdi_async_bio_wq;
839a8e86 52
d6c10f1f 53static inline bool wb_has_dirty_io(struct bdi_writeback *wb)
03ba3782 54{
d6c10f1f 55 return test_bit(WB_has_dirty_io, &wb->state);
03ba3782
JA
56}
57
95a46c65
TH
58static 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);
03ba3782
JA
65}
66
bd3488e7 67static inline void wb_stat_mod(struct bdi_writeback *wb,
93f78d88 68 enum wb_stat_item item, s64 amount)
b2e8fb6e 69{
104b4e51 70 percpu_counter_add_batch(&wb->stat[item], amount, WB_STAT_BATCH);
b2e8fb6e
PZ
71}
72
93f78d88 73static inline void inc_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
b2e8fb6e 74{
bd3488e7 75 wb_stat_mod(wb, item, 1);
b2e8fb6e
PZ
76}
77
93f78d88 78static inline void dec_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
b2e8fb6e 79{
bd3488e7 80 wb_stat_mod(wb, item, -1);
b2e8fb6e
PZ
81}
82
93f78d88 83static inline s64 wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
b2e8fb6e 84{
93f78d88 85 return percpu_counter_read_positive(&wb->stat[item]);
b2e8fb6e
PZ
86}
87
93f78d88 88static inline s64 wb_stat_sum(struct bdi_writeback *wb, enum wb_stat_item item)
e0bf68dd 89{
e3d3910a 90 return percpu_counter_sum_positive(&wb->stat[item]);
e0bf68dd
PZ
91}
92
93f78d88 93extern void wb_writeout_inc(struct bdi_writeback *wb);
dd5656e5 94
b2e8fb6e
PZ
95/*
96 * maximal error of a stat counter.
97 */
2bce774e 98static inline unsigned long wb_stat_error(void)
e0bf68dd 99{
b2e8fb6e 100#ifdef CONFIG_SMP
93f78d88 101 return nr_cpu_ids * WB_STAT_BATCH;
b2e8fb6e
PZ
102#else
103 return 1;
104#endif
e0bf68dd 105}
1da177e4 106
189d3c4a 107int bdi_set_min_ratio(struct backing_dev_info *bdi, unsigned int min_ratio);
a42dde04 108int bdi_set_max_ratio(struct backing_dev_info *bdi, unsigned int max_ratio);
189d3c4a 109
1da177e4
LT
110/*
111 * Flags in backing_dev_info::capability
e4ad08fe 112 *
f56753ac
CH
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
1da177e4 117 */
f56753ac
CH
118#define BDI_CAP_WRITEBACK (1 << 0)
119#define BDI_CAP_WRITEBACK_ACCT (1 << 1)
120#define BDI_CAP_STRICTLIMIT (1 << 2)
e4ad08fe 121
5129a469 122extern struct backing_dev_info noop_backing_dev_info;
1da177e4 123
bc05873d
TH
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 */
131static inline bool writeback_in_progress(struct bdi_writeback *wb)
1da177e4 132{
bc05873d 133 return test_bit(WB_writeback_running, &wb->state);
1da177e4
LT
134}
135
ccdf7741 136struct backing_dev_info *inode_to_bdi(struct inode *inode);
1da177e4 137
ec8a6f26 138static inline int wb_congested(struct bdi_writeback *wb, int cong_bits)
1da177e4 139{
8c911f3d 140 return wb->congested & cong_bits;
1da177e4 141}
373c0a7e 142
8aa7e847 143long congestion_wait(int sync, long timeout);
1da177e4 144
f56753ac 145static inline bool mapping_can_writeback(struct address_space *mapping)
e4ad08fe 146{
f56753ac 147 return inode_to_bdi(mapping->host)->capabilities & BDI_CAP_WRITEBACK;
e4ad08fe 148}
1da177e4 149
03ba3782
JA
150static inline int bdi_sched_wait(void *word)
151{
152 schedule();
153 return 0;
154}
155
89e9b9e0
TH
156#ifdef CONFIG_CGROUP_WRITEBACK
157
ed288dc0
TH
158struct bdi_writeback *wb_get_lookup(struct backing_dev_info *bdi,
159 struct cgroup_subsys_state *memcg_css);
52ebea74
TH
160struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi,
161 struct cgroup_subsys_state *memcg_css,
162 gfp_t gfp);
52ebea74
TH
163void wb_memcg_offline(struct mem_cgroup *memcg);
164void wb_blkcg_offline(struct blkcg *blkcg);
703c2708 165int inode_congested(struct inode *inode, int cong_bits);
52ebea74 166
89e9b9e0
TH
167/**
168 * inode_cgwb_enabled - test whether cgroup writeback is enabled on an inode
169 * @inode: inode of interest
170 *
ed7b6b4f
CH
171 * Cgroup writeback requires support from the filesystem. Also, both memcg and
172 * iocg have to be on the default hierarchy. Test whether all conditions are
173 * met.
9badce00
TH
174 *
175 * Note that the test result may change dynamically on the same inode
176 * depending on how memcg and iocg are configured.
89e9b9e0
TH
177 */
178static inline bool inode_cgwb_enabled(struct inode *inode)
179{
180 struct backing_dev_info *bdi = inode_to_bdi(inode);
181
c0522908
TH
182 return cgroup_subsys_on_dfl(memory_cgrp_subsys) &&
183 cgroup_subsys_on_dfl(io_cgrp_subsys) &&
f56753ac 184 (bdi->capabilities & BDI_CAP_WRITEBACK) &&
46b15caa 185 (inode->i_sb->s_iflags & SB_I_CGROUPWB);
89e9b9e0
TH
186}
187
52ebea74
TH
188/**
189 * wb_find_current - find wb for %current on a bdi
190 * @bdi: bdi of interest
191 *
192 * Find the wb of @bdi which matches both the memcg and blkcg of %current.
193 * Must be called under rcu_read_lock() which protects the returend wb.
194 * NULL if not found.
195 */
196static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
197{
198 struct cgroup_subsys_state *memcg_css;
199 struct bdi_writeback *wb;
200
201 memcg_css = task_css(current, memory_cgrp_id);
202 if (!memcg_css->parent)
203 return &bdi->wb;
204
205 wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
206
207 /*
208 * %current's blkcg equals the effective blkcg of its memcg. No
209 * need to use the relatively expensive cgroup_get_e_css().
210 */
c165b3e3 211 if (likely(wb && wb->blkcg_css == task_css(current, io_cgrp_id)))
52ebea74
TH
212 return wb;
213 return NULL;
214}
215
216/**
217 * wb_get_create_current - get or create wb for %current on a bdi
218 * @bdi: bdi of interest
219 * @gfp: allocation mask
220 *
221 * Equivalent to wb_get_create() on %current's memcg. This function is
222 * called from a relatively hot path and optimizes the common cases using
223 * wb_find_current().
224 */
225static inline struct bdi_writeback *
226wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
227{
228 struct bdi_writeback *wb;
229
230 rcu_read_lock();
231 wb = wb_find_current(bdi);
232 if (wb && unlikely(!wb_tryget(wb)))
233 wb = NULL;
234 rcu_read_unlock();
235
236 if (unlikely(!wb)) {
237 struct cgroup_subsys_state *memcg_css;
238
239 memcg_css = task_get_css(current, memory_cgrp_id);
240 wb = wb_get_create(bdi, memcg_css, gfp);
241 css_put(memcg_css);
242 }
243 return wb;
244}
245
aaa2cacf
TH
246/**
247 * inode_to_wb_is_valid - test whether an inode has a wb associated
248 * @inode: inode of interest
249 *
250 * Returns %true if @inode has a wb associated. May be called without any
251 * locking.
252 */
253static inline bool inode_to_wb_is_valid(struct inode *inode)
254{
255 return inode->i_wb;
256}
257
52ebea74
TH
258/**
259 * inode_to_wb - determine the wb of an inode
260 * @inode: inode of interest
261 *
aaa2cacf 262 * Returns the wb @inode is currently associated with. The caller must be
b93b0163 263 * holding either @inode->i_lock, the i_pages lock, or the
aaa2cacf 264 * associated wb's list_lock.
52ebea74 265 */
05b93801 266static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
52ebea74 267{
aaa2cacf
TH
268#ifdef CONFIG_LOCKDEP
269 WARN_ON_ONCE(debug_locks &&
270 (!lockdep_is_held(&inode->i_lock) &&
b93b0163 271 !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
aaa2cacf
TH
272 !lockdep_is_held(&inode->i_wb->list_lock)));
273#endif
52ebea74
TH
274 return inode->i_wb;
275}
276
fee468fd
JK
277static inline struct bdi_writeback *inode_to_wb_wbc(
278 struct inode *inode,
279 struct writeback_control *wbc)
280{
281 /*
282 * If wbc does not have inode attached, it means cgroup writeback was
283 * disabled when wbc started. Just use the default wb in that case.
284 */
285 return wbc->wb ? wbc->wb : &inode_to_bdi(inode)->wb;
286}
287
682aa8e1
TH
288/**
289 * unlocked_inode_to_wb_begin - begin unlocked inode wb access transaction
290 * @inode: target inode
2e898e4c 291 * @cookie: output param, to be passed to the end function
682aa8e1
TH
292 *
293 * The caller wants to access the wb associated with @inode but isn't
b93b0163 294 * holding inode->i_lock, the i_pages lock or wb->list_lock. This
682aa8e1
TH
295 * function determines the wb associated with @inode and ensures that the
296 * association doesn't change until the transaction is finished with
297 * unlocked_inode_to_wb_end().
298 *
2e898e4c
GT
299 * The caller must call unlocked_inode_to_wb_end() with *@cookie afterwards and
300 * can't sleep during the transaction. IRQs may or may not be disabled on
301 * return.
682aa8e1
TH
302 */
303static inline struct bdi_writeback *
2e898e4c 304unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
682aa8e1
TH
305{
306 rcu_read_lock();
307
308 /*
a9519def 309 * Paired with store_release in inode_switch_wbs_work_fn() and
682aa8e1
TH
310 * ensures that we see the new wb if we see cleared I_WB_SWITCH.
311 */
2e898e4c 312 cookie->locked = smp_load_acquire(&inode->i_state) & I_WB_SWITCH;
682aa8e1 313
2e898e4c
GT
314 if (unlikely(cookie->locked))
315 xa_lock_irqsave(&inode->i_mapping->i_pages, cookie->flags);
aaa2cacf
TH
316
317 /*
b93b0163
MW
318 * Protected by either !I_WB_SWITCH + rcu_read_lock() or the i_pages
319 * lock. inode_to_wb() will bark. Deref directly.
aaa2cacf
TH
320 */
321 return inode->i_wb;
682aa8e1
TH
322}
323
324/**
325 * unlocked_inode_to_wb_end - end inode wb access transaction
326 * @inode: target inode
2e898e4c 327 * @cookie: @cookie from unlocked_inode_to_wb_begin()
682aa8e1 328 */
2e898e4c
GT
329static inline void unlocked_inode_to_wb_end(struct inode *inode,
330 struct wb_lock_cookie *cookie)
682aa8e1 331{
2e898e4c
GT
332 if (unlikely(cookie->locked))
333 xa_unlock_irqrestore(&inode->i_mapping->i_pages, cookie->flags);
682aa8e1
TH
334
335 rcu_read_unlock();
336}
337
89e9b9e0
TH
338#else /* CONFIG_CGROUP_WRITEBACK */
339
340static inline bool inode_cgwb_enabled(struct inode *inode)
341{
342 return false;
343}
344
52ebea74
TH
345static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
346{
347 return &bdi->wb;
348}
349
350static inline struct bdi_writeback *
351wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
352{
353 return &bdi->wb;
354}
355
aaa2cacf
TH
356static inline bool inode_to_wb_is_valid(struct inode *inode)
357{
358 return true;
359}
360
52ebea74
TH
361static inline struct bdi_writeback *inode_to_wb(struct inode *inode)
362{
363 return &inode_to_bdi(inode)->wb;
364}
365
fee468fd
JK
366static inline struct bdi_writeback *inode_to_wb_wbc(
367 struct inode *inode,
368 struct writeback_control *wbc)
369{
370 return inode_to_wb(inode);
371}
372
373
682aa8e1 374static inline struct bdi_writeback *
2e898e4c 375unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
682aa8e1
TH
376{
377 return inode_to_wb(inode);
378}
379
2e898e4c
GT
380static inline void unlocked_inode_to_wb_end(struct inode *inode,
381 struct wb_lock_cookie *cookie)
682aa8e1
TH
382{
383}
384
52ebea74
TH
385static inline void wb_memcg_offline(struct mem_cgroup *memcg)
386{
387}
388
389static inline void wb_blkcg_offline(struct blkcg *blkcg)
390{
391}
392
703c2708
TH
393static inline int inode_congested(struct inode *inode, int cong_bits)
394{
395 return wb_congested(&inode_to_bdi(inode)->wb, cong_bits);
396}
397
89e9b9e0
TH
398#endif /* CONFIG_CGROUP_WRITEBACK */
399
703c2708
TH
400static inline int inode_read_congested(struct inode *inode)
401{
402 return inode_congested(inode, 1 << WB_sync_congested);
403}
404
405static inline int inode_write_congested(struct inode *inode)
406{
407 return inode_congested(inode, 1 << WB_async_congested);
408}
409
410static inline int inode_rw_congested(struct inode *inode)
411{
412 return inode_congested(inode, (1 << WB_sync_congested) |
413 (1 << WB_async_congested));
414}
415
ec8a6f26
TH
416static inline int bdi_congested(struct backing_dev_info *bdi, int cong_bits)
417{
418 return wb_congested(&bdi->wb, cong_bits);
419}
420
421static inline int bdi_read_congested(struct backing_dev_info *bdi)
422{
423 return bdi_congested(bdi, 1 << WB_sync_congested);
424}
425
426static inline int bdi_write_congested(struct backing_dev_info *bdi)
427{
428 return bdi_congested(bdi, 1 << WB_async_congested);
429}
430
431static inline int bdi_rw_congested(struct backing_dev_info *bdi)
432{
433 return bdi_congested(bdi, (1 << WB_sync_congested) |
434 (1 << WB_async_congested));
435}
436
eb7ae5e0 437const char *bdi_dev_name(struct backing_dev_info *bdi);
68f23b89 438
89e9b9e0 439#endif /* _LINUX_BACKING_DEV_H */