Merge tag 'omap-fixes-audio-clock-and-modem-signed' of git://git.kernel.org/pub/scm...
[linux-2.6-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
d03f6cdc
JK
20static inline struct backing_dev_info *bdi_get(struct backing_dev_info *bdi)
21{
22 kref_get(&bdi->refcnt);
23 return bdi;
24}
25
34f8fe50 26struct backing_dev_info *bdi_get_by_id(u64 id);
d03f6cdc 27void bdi_put(struct backing_dev_info *bdi);
b2e8fb6e 28
7c4cc300
JK
29__printf(2, 3)
30int bdi_register(struct backing_dev_info *bdi, const char *fmt, ...);
a93f00b3 31__printf(2, 0)
7c4cc300
JK
32int bdi_register_va(struct backing_dev_info *bdi, const char *fmt,
33 va_list args);
3c5d202b 34void bdi_set_owner(struct backing_dev_info *bdi, struct device *owner);
b02176f3
TH
35void bdi_unregister(struct backing_dev_info *bdi);
36
aef33c2f 37struct backing_dev_info *bdi_alloc(int node_id);
b02176f3 38
9ecf4866 39void wb_start_background_writeback(struct bdi_writeback *wb);
f0054bb1 40void wb_workfn(struct work_struct *work);
f0054bb1 41void wb_wakeup_delayed(struct bdi_writeback *wb);
cf0ca9fe 42
5b9cce4c
TH
43void wb_wait_for_completion(struct wb_completion *done);
44
03ba3782 45extern spinlock_t bdi_lock;
66f3b8e2
JA
46extern struct list_head bdi_list;
47
839a8e86
TH
48extern struct workqueue_struct *bdi_wq;
49
d6c10f1f 50static inline bool wb_has_dirty_io(struct bdi_writeback *wb)
03ba3782 51{
d6c10f1f 52 return test_bit(WB_has_dirty_io, &wb->state);
03ba3782
JA
53}
54
95a46c65
TH
55static inline bool bdi_has_dirty_io(struct backing_dev_info *bdi)
56{
57 /*
58 * @bdi->tot_write_bandwidth is guaranteed to be > 0 if there are
59 * any dirty wbs. See wb_update_write_bandwidth().
60 */
61 return atomic_long_read(&bdi->tot_write_bandwidth);
03ba3782
JA
62}
63
bd3488e7 64static inline void wb_stat_mod(struct bdi_writeback *wb,
93f78d88 65 enum wb_stat_item item, s64 amount)
b2e8fb6e 66{
104b4e51 67 percpu_counter_add_batch(&wb->stat[item], amount, WB_STAT_BATCH);
b2e8fb6e
PZ
68}
69
93f78d88 70static inline void inc_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
b2e8fb6e 71{
bd3488e7 72 wb_stat_mod(wb, item, 1);
b2e8fb6e
PZ
73}
74
93f78d88 75static inline void dec_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
b2e8fb6e 76{
bd3488e7 77 wb_stat_mod(wb, item, -1);
b2e8fb6e
PZ
78}
79
93f78d88 80static inline s64 wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
b2e8fb6e 81{
93f78d88 82 return percpu_counter_read_positive(&wb->stat[item]);
b2e8fb6e
PZ
83}
84
93f78d88 85static inline s64 wb_stat_sum(struct bdi_writeback *wb, enum wb_stat_item item)
e0bf68dd 86{
e3d3910a 87 return percpu_counter_sum_positive(&wb->stat[item]);
e0bf68dd
PZ
88}
89
93f78d88 90extern void wb_writeout_inc(struct bdi_writeback *wb);
dd5656e5 91
b2e8fb6e
PZ
92/*
93 * maximal error of a stat counter.
94 */
2bce774e 95static inline unsigned long wb_stat_error(void)
e0bf68dd 96{
b2e8fb6e 97#ifdef CONFIG_SMP
93f78d88 98 return nr_cpu_ids * WB_STAT_BATCH;
b2e8fb6e
PZ
99#else
100 return 1;
101#endif
e0bf68dd 102}
1da177e4 103
ae82291e
SR
104/* BDI ratio is expressed as part per 1000000 for finer granularity. */
105#define BDI_RATIO_SCALE 10000
106
712c00d6 107u64 bdi_get_min_bytes(struct backing_dev_info *bdi);
00df7d51 108u64 bdi_get_max_bytes(struct backing_dev_info *bdi);
189d3c4a 109int bdi_set_min_ratio(struct backing_dev_info *bdi, unsigned int min_ratio);
a42dde04 110int bdi_set_max_ratio(struct backing_dev_info *bdi, unsigned int max_ratio);
2c44af4f 111int bdi_set_min_ratio_no_scale(struct backing_dev_info *bdi, unsigned int min_ratio);
4e230b40 112int bdi_set_max_ratio_no_scale(struct backing_dev_info *bdi, unsigned int max_ratio);
803c9805 113int bdi_set_min_bytes(struct backing_dev_info *bdi, u64 min_bytes);
1bf27e98 114int bdi_set_max_bytes(struct backing_dev_info *bdi, u64 max_bytes);
8e9d5ead 115int bdi_set_strict_limit(struct backing_dev_info *bdi, unsigned int strict_limit);
189d3c4a 116
1da177e4
LT
117/*
118 * Flags in backing_dev_info::capability
e4ad08fe 119 *
f56753ac
CH
120 * BDI_CAP_WRITEBACK: Supports dirty page writeback, and dirty pages
121 * should contribute to accounting
122 * BDI_CAP_WRITEBACK_ACCT: Automatically account writeback pages
123 * BDI_CAP_STRICTLIMIT: Keep number of dirty pages below bdi threshold
1da177e4 124 */
f56753ac
CH
125#define BDI_CAP_WRITEBACK (1 << 0)
126#define BDI_CAP_WRITEBACK_ACCT (1 << 1)
127#define BDI_CAP_STRICTLIMIT (1 << 2)
e4ad08fe 128
5129a469 129extern struct backing_dev_info noop_backing_dev_info;
1da177e4 130
4bca7e80
JK
131int bdi_init(struct backing_dev_info *bdi);
132
bc05873d
TH
133/**
134 * writeback_in_progress - determine whether there is writeback in progress
135 * @wb: bdi_writeback of interest
136 *
137 * Determine whether there is writeback waiting to be handled against a
138 * bdi_writeback.
139 */
140static inline bool writeback_in_progress(struct bdi_writeback *wb)
1da177e4 141{
bc05873d 142 return test_bit(WB_writeback_running, &wb->state);
1da177e4
LT
143}
144
ccdf7741 145struct backing_dev_info *inode_to_bdi(struct inode *inode);
1da177e4 146
f56753ac 147static inline bool mapping_can_writeback(struct address_space *mapping)
e4ad08fe 148{
f56753ac 149 return inode_to_bdi(mapping->host)->capabilities & BDI_CAP_WRITEBACK;
e4ad08fe 150}
1da177e4 151
89e9b9e0
TH
152#ifdef CONFIG_CGROUP_WRITEBACK
153
ed288dc0
TH
154struct bdi_writeback *wb_get_lookup(struct backing_dev_info *bdi,
155 struct cgroup_subsys_state *memcg_css);
52ebea74
TH
156struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi,
157 struct cgroup_subsys_state *memcg_css,
158 gfp_t gfp);
52ebea74 159void wb_memcg_offline(struct mem_cgroup *memcg);
dec223c9 160void wb_blkcg_offline(struct cgroup_subsys_state *css);
52ebea74 161
89e9b9e0
TH
162/**
163 * inode_cgwb_enabled - test whether cgroup writeback is enabled on an inode
164 * @inode: inode of interest
165 *
ed7b6b4f
CH
166 * Cgroup writeback requires support from the filesystem. Also, both memcg and
167 * iocg have to be on the default hierarchy. Test whether all conditions are
168 * met.
9badce00
TH
169 *
170 * Note that the test result may change dynamically on the same inode
171 * depending on how memcg and iocg are configured.
89e9b9e0
TH
172 */
173static inline bool inode_cgwb_enabled(struct inode *inode)
174{
175 struct backing_dev_info *bdi = inode_to_bdi(inode);
176
c0522908
TH
177 return cgroup_subsys_on_dfl(memory_cgrp_subsys) &&
178 cgroup_subsys_on_dfl(io_cgrp_subsys) &&
f56753ac 179 (bdi->capabilities & BDI_CAP_WRITEBACK) &&
46b15caa 180 (inode->i_sb->s_iflags & SB_I_CGROUPWB);
89e9b9e0
TH
181}
182
52ebea74
TH
183/**
184 * wb_find_current - find wb for %current on a bdi
185 * @bdi: bdi of interest
186 *
187 * Find the wb of @bdi which matches both the memcg and blkcg of %current.
188 * Must be called under rcu_read_lock() which protects the returend wb.
189 * NULL if not found.
190 */
191static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
192{
193 struct cgroup_subsys_state *memcg_css;
194 struct bdi_writeback *wb;
195
196 memcg_css = task_css(current, memory_cgrp_id);
197 if (!memcg_css->parent)
198 return &bdi->wb;
199
200 wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
201
202 /*
203 * %current's blkcg equals the effective blkcg of its memcg. No
204 * need to use the relatively expensive cgroup_get_e_css().
205 */
c165b3e3 206 if (likely(wb && wb->blkcg_css == task_css(current, io_cgrp_id)))
52ebea74
TH
207 return wb;
208 return NULL;
209}
210
211/**
212 * wb_get_create_current - get or create wb for %current on a bdi
213 * @bdi: bdi of interest
214 * @gfp: allocation mask
215 *
216 * Equivalent to wb_get_create() on %current's memcg. This function is
217 * called from a relatively hot path and optimizes the common cases using
218 * wb_find_current().
219 */
220static inline struct bdi_writeback *
221wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
222{
223 struct bdi_writeback *wb;
224
225 rcu_read_lock();
226 wb = wb_find_current(bdi);
227 if (wb && unlikely(!wb_tryget(wb)))
228 wb = NULL;
229 rcu_read_unlock();
230
231 if (unlikely(!wb)) {
232 struct cgroup_subsys_state *memcg_css;
233
234 memcg_css = task_get_css(current, memory_cgrp_id);
235 wb = wb_get_create(bdi, memcg_css, gfp);
236 css_put(memcg_css);
237 }
238 return wb;
239}
240
52ebea74
TH
241/**
242 * inode_to_wb - determine the wb of an inode
243 * @inode: inode of interest
244 *
aaa2cacf 245 * Returns the wb @inode is currently associated with. The caller must be
b93b0163 246 * holding either @inode->i_lock, the i_pages lock, or the
aaa2cacf 247 * associated wb's list_lock.
52ebea74 248 */
05b93801 249static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
52ebea74 250{
aaa2cacf
TH
251#ifdef CONFIG_LOCKDEP
252 WARN_ON_ONCE(debug_locks &&
253 (!lockdep_is_held(&inode->i_lock) &&
b93b0163 254 !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
aaa2cacf
TH
255 !lockdep_is_held(&inode->i_wb->list_lock)));
256#endif
52ebea74
TH
257 return inode->i_wb;
258}
259
fee468fd
JK
260static inline struct bdi_writeback *inode_to_wb_wbc(
261 struct inode *inode,
262 struct writeback_control *wbc)
263{
264 /*
265 * If wbc does not have inode attached, it means cgroup writeback was
266 * disabled when wbc started. Just use the default wb in that case.
267 */
268 return wbc->wb ? wbc->wb : &inode_to_bdi(inode)->wb;
269}
270
682aa8e1
TH
271/**
272 * unlocked_inode_to_wb_begin - begin unlocked inode wb access transaction
273 * @inode: target inode
2e898e4c 274 * @cookie: output param, to be passed to the end function
682aa8e1
TH
275 *
276 * The caller wants to access the wb associated with @inode but isn't
b93b0163 277 * holding inode->i_lock, the i_pages lock or wb->list_lock. This
682aa8e1
TH
278 * function determines the wb associated with @inode and ensures that the
279 * association doesn't change until the transaction is finished with
280 * unlocked_inode_to_wb_end().
281 *
2e898e4c
GT
282 * The caller must call unlocked_inode_to_wb_end() with *@cookie afterwards and
283 * can't sleep during the transaction. IRQs may or may not be disabled on
284 * return.
682aa8e1
TH
285 */
286static inline struct bdi_writeback *
2e898e4c 287unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
682aa8e1
TH
288{
289 rcu_read_lock();
290
291 /*
a9519def 292 * Paired with store_release in inode_switch_wbs_work_fn() and
682aa8e1
TH
293 * ensures that we see the new wb if we see cleared I_WB_SWITCH.
294 */
2e898e4c 295 cookie->locked = smp_load_acquire(&inode->i_state) & I_WB_SWITCH;
682aa8e1 296
2e898e4c
GT
297 if (unlikely(cookie->locked))
298 xa_lock_irqsave(&inode->i_mapping->i_pages, cookie->flags);
aaa2cacf
TH
299
300 /*
b93b0163
MW
301 * Protected by either !I_WB_SWITCH + rcu_read_lock() or the i_pages
302 * lock. inode_to_wb() will bark. Deref directly.
aaa2cacf
TH
303 */
304 return inode->i_wb;
682aa8e1
TH
305}
306
307/**
308 * unlocked_inode_to_wb_end - end inode wb access transaction
309 * @inode: target inode
2e898e4c 310 * @cookie: @cookie from unlocked_inode_to_wb_begin()
682aa8e1 311 */
2e898e4c
GT
312static inline void unlocked_inode_to_wb_end(struct inode *inode,
313 struct wb_lock_cookie *cookie)
682aa8e1 314{
2e898e4c
GT
315 if (unlikely(cookie->locked))
316 xa_unlock_irqrestore(&inode->i_mapping->i_pages, cookie->flags);
682aa8e1
TH
317
318 rcu_read_unlock();
319}
320
89e9b9e0
TH
321#else /* CONFIG_CGROUP_WRITEBACK */
322
323static inline bool inode_cgwb_enabled(struct inode *inode)
324{
325 return false;
326}
327
52ebea74
TH
328static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
329{
330 return &bdi->wb;
331}
332
333static inline struct bdi_writeback *
334wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
335{
336 return &bdi->wb;
337}
338
52ebea74
TH
339static inline struct bdi_writeback *inode_to_wb(struct inode *inode)
340{
341 return &inode_to_bdi(inode)->wb;
342}
343
fee468fd
JK
344static inline struct bdi_writeback *inode_to_wb_wbc(
345 struct inode *inode,
346 struct writeback_control *wbc)
347{
348 return inode_to_wb(inode);
349}
350
351
682aa8e1 352static inline struct bdi_writeback *
2e898e4c 353unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
682aa8e1
TH
354{
355 return inode_to_wb(inode);
356}
357
2e898e4c
GT
358static inline void unlocked_inode_to_wb_end(struct inode *inode,
359 struct wb_lock_cookie *cookie)
682aa8e1
TH
360{
361}
362
52ebea74
TH
363static inline void wb_memcg_offline(struct mem_cgroup *memcg)
364{
365}
366
dec223c9 367static inline void wb_blkcg_offline(struct cgroup_subsys_state *css)
52ebea74
TH
368{
369}
370
89e9b9e0
TH
371#endif /* CONFIG_CGROUP_WRITEBACK */
372
eb7ae5e0 373const char *bdi_dev_name(struct backing_dev_info *bdi);
68f23b89 374
89e9b9e0 375#endif /* _LINUX_BACKING_DEV_H */