1 /* SPDX-License-Identifier: GPL-2.0 */
3 * include/linux/backing-dev.h
5 * low-level device information and state which is propagated up through
9 #ifndef _LINUX_BACKING_DEV_H
10 #define _LINUX_BACKING_DEV_H
12 #include <linux/kernel.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>
22 static inline struct backing_dev_info *bdi_get(struct backing_dev_info *bdi)
24 kref_get(&bdi->refcnt);
28 struct backing_dev_info *bdi_get_by_id(u64 id);
29 void bdi_put(struct backing_dev_info *bdi);
32 int bdi_register(struct backing_dev_info *bdi, const char *fmt, ...);
34 int bdi_register_va(struct backing_dev_info *bdi, const char *fmt,
36 void bdi_set_owner(struct backing_dev_info *bdi, struct device *owner);
37 void bdi_unregister(struct backing_dev_info *bdi);
39 struct backing_dev_info *bdi_alloc(int node_id);
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);
45 void wb_wait_for_completion(struct wb_completion *done);
47 extern spinlock_t bdi_lock;
48 extern struct list_head bdi_list;
50 extern struct workqueue_struct *bdi_wq;
51 extern struct workqueue_struct *bdi_async_bio_wq;
53 static inline bool wb_has_dirty_io(struct bdi_writeback *wb)
55 return test_bit(WB_has_dirty_io, &wb->state);
58 static inline bool bdi_has_dirty_io(struct backing_dev_info *bdi)
61 * @bdi->tot_write_bandwidth is guaranteed to be > 0 if there are
62 * any dirty wbs. See wb_update_write_bandwidth().
64 return atomic_long_read(&bdi->tot_write_bandwidth);
67 static inline void __add_wb_stat(struct bdi_writeback *wb,
68 enum wb_stat_item item, s64 amount)
70 percpu_counter_add_batch(&wb->stat[item], amount, WB_STAT_BATCH);
73 static inline void inc_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
75 __add_wb_stat(wb, item, 1);
78 static inline void dec_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
80 __add_wb_stat(wb, item, -1);
83 static inline s64 wb_stat(struct bdi_writeback *wb, enum wb_stat_item item)
85 return percpu_counter_read_positive(&wb->stat[item]);
88 static inline s64 wb_stat_sum(struct bdi_writeback *wb, enum wb_stat_item item)
90 return percpu_counter_sum_positive(&wb->stat[item]);
93 extern void wb_writeout_inc(struct bdi_writeback *wb);
96 * maximal error of a stat counter.
98 static inline unsigned long wb_stat_error(void)
101 return nr_cpu_ids * WB_STAT_BATCH;
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);
111 * Flags in backing_dev_info::capability
113 * The first three flags control whether dirty pages will contribute to the
114 * VM's accounting and whether writepages() should be called for dirty pages
115 * (something that would not, for example, be appropriate for ramfs)
117 * WARNING: these flags are closely related and should not normally be
118 * used separately. The BDI_CAP_NO_ACCT_AND_WRITEBACK combines these
119 * three flags into a single convenience macro.
121 * BDI_CAP_NO_ACCT_DIRTY: Dirty pages shouldn't contribute to accounting
122 * BDI_CAP_NO_WRITEBACK: Don't write pages back
123 * BDI_CAP_NO_ACCT_WB: Don't automatically account writeback pages
124 * BDI_CAP_STRICTLIMIT: Keep number of dirty pages below bdi threshold.
126 * BDI_CAP_CGROUP_WRITEBACK: Supports cgroup-aware writeback.
127 * BDI_CAP_SYNCHRONOUS_IO: Device is so fast that asynchronous IO would be
130 #define BDI_CAP_NO_ACCT_DIRTY 0x00000001
131 #define BDI_CAP_NO_WRITEBACK 0x00000002
132 #define BDI_CAP_NO_ACCT_WB 0x00000004
133 #define BDI_CAP_STABLE_WRITES 0x00000008
134 #define BDI_CAP_STRICTLIMIT 0x00000010
135 #define BDI_CAP_CGROUP_WRITEBACK 0x00000020
136 #define BDI_CAP_SYNCHRONOUS_IO 0x00000040
138 #define BDI_CAP_NO_ACCT_AND_WRITEBACK \
139 (BDI_CAP_NO_WRITEBACK | BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_ACCT_WB)
141 extern struct backing_dev_info noop_backing_dev_info;
144 * writeback_in_progress - determine whether there is writeback in progress
145 * @wb: bdi_writeback of interest
147 * Determine whether there is writeback waiting to be handled against a
150 static inline bool writeback_in_progress(struct bdi_writeback *wb)
152 return test_bit(WB_writeback_running, &wb->state);
155 static inline struct backing_dev_info *inode_to_bdi(struct inode *inode)
157 struct super_block *sb;
160 return &noop_backing_dev_info;
164 if (sb_is_blkdev_sb(sb))
165 return I_BDEV(inode)->bd_bdi;
170 static inline int wb_congested(struct bdi_writeback *wb, int cong_bits)
172 struct backing_dev_info *bdi = wb->bdi;
174 if (bdi->congested_fn)
175 return bdi->congested_fn(bdi->congested_data, cong_bits);
176 return wb->congested->state & cong_bits;
179 long congestion_wait(int sync, long timeout);
180 long wait_iff_congested(int sync, long timeout);
182 static inline bool bdi_cap_synchronous_io(struct backing_dev_info *bdi)
184 return bdi->capabilities & BDI_CAP_SYNCHRONOUS_IO;
187 static inline bool bdi_cap_stable_pages_required(struct backing_dev_info *bdi)
189 return bdi->capabilities & BDI_CAP_STABLE_WRITES;
192 static inline bool bdi_cap_writeback_dirty(struct backing_dev_info *bdi)
194 return !(bdi->capabilities & BDI_CAP_NO_WRITEBACK);
197 static inline bool bdi_cap_account_dirty(struct backing_dev_info *bdi)
199 return !(bdi->capabilities & BDI_CAP_NO_ACCT_DIRTY);
202 static inline bool bdi_cap_account_writeback(struct backing_dev_info *bdi)
204 /* Paranoia: BDI_CAP_NO_WRITEBACK implies BDI_CAP_NO_ACCT_WB */
205 return !(bdi->capabilities & (BDI_CAP_NO_ACCT_WB |
206 BDI_CAP_NO_WRITEBACK));
209 static inline bool mapping_cap_writeback_dirty(struct address_space *mapping)
211 return bdi_cap_writeback_dirty(inode_to_bdi(mapping->host));
214 static inline bool mapping_cap_account_dirty(struct address_space *mapping)
216 return bdi_cap_account_dirty(inode_to_bdi(mapping->host));
219 static inline int bdi_sched_wait(void *word)
225 #ifdef CONFIG_CGROUP_WRITEBACK
227 struct bdi_writeback_congested *
228 wb_congested_get_create(struct backing_dev_info *bdi, int blkcg_id, gfp_t gfp);
229 void wb_congested_put(struct bdi_writeback_congested *congested);
230 struct bdi_writeback *wb_get_lookup(struct backing_dev_info *bdi,
231 struct cgroup_subsys_state *memcg_css);
232 struct bdi_writeback *wb_get_create(struct backing_dev_info *bdi,
233 struct cgroup_subsys_state *memcg_css,
235 void wb_memcg_offline(struct mem_cgroup *memcg);
236 void wb_blkcg_offline(struct blkcg *blkcg);
237 int inode_congested(struct inode *inode, int cong_bits);
240 * inode_cgwb_enabled - test whether cgroup writeback is enabled on an inode
241 * @inode: inode of interest
243 * cgroup writeback requires support from both the bdi and filesystem.
244 * Also, both memcg and iocg have to be on the default hierarchy. Test
245 * whether all conditions are met.
247 * Note that the test result may change dynamically on the same inode
248 * depending on how memcg and iocg are configured.
250 static inline bool inode_cgwb_enabled(struct inode *inode)
252 struct backing_dev_info *bdi = inode_to_bdi(inode);
254 return cgroup_subsys_on_dfl(memory_cgrp_subsys) &&
255 cgroup_subsys_on_dfl(io_cgrp_subsys) &&
256 bdi_cap_account_dirty(bdi) &&
257 (bdi->capabilities & BDI_CAP_CGROUP_WRITEBACK) &&
258 (inode->i_sb->s_iflags & SB_I_CGROUPWB);
262 * wb_find_current - find wb for %current on a bdi
263 * @bdi: bdi of interest
265 * Find the wb of @bdi which matches both the memcg and blkcg of %current.
266 * Must be called under rcu_read_lock() which protects the returend wb.
269 static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
271 struct cgroup_subsys_state *memcg_css;
272 struct bdi_writeback *wb;
274 memcg_css = task_css(current, memory_cgrp_id);
275 if (!memcg_css->parent)
278 wb = radix_tree_lookup(&bdi->cgwb_tree, memcg_css->id);
281 * %current's blkcg equals the effective blkcg of its memcg. No
282 * need to use the relatively expensive cgroup_get_e_css().
284 if (likely(wb && wb->blkcg_css == task_css(current, io_cgrp_id)))
290 * wb_get_create_current - get or create wb for %current on a bdi
291 * @bdi: bdi of interest
292 * @gfp: allocation mask
294 * Equivalent to wb_get_create() on %current's memcg. This function is
295 * called from a relatively hot path and optimizes the common cases using
298 static inline struct bdi_writeback *
299 wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
301 struct bdi_writeback *wb;
304 wb = wb_find_current(bdi);
305 if (wb && unlikely(!wb_tryget(wb)))
310 struct cgroup_subsys_state *memcg_css;
312 memcg_css = task_get_css(current, memory_cgrp_id);
313 wb = wb_get_create(bdi, memcg_css, gfp);
320 * inode_to_wb_is_valid - test whether an inode has a wb associated
321 * @inode: inode of interest
323 * Returns %true if @inode has a wb associated. May be called without any
326 static inline bool inode_to_wb_is_valid(struct inode *inode)
332 * inode_to_wb - determine the wb of an inode
333 * @inode: inode of interest
335 * Returns the wb @inode is currently associated with. The caller must be
336 * holding either @inode->i_lock, the i_pages lock, or the
337 * associated wb's list_lock.
339 static inline struct bdi_writeback *inode_to_wb(const struct inode *inode)
341 #ifdef CONFIG_LOCKDEP
342 WARN_ON_ONCE(debug_locks &&
343 (!lockdep_is_held(&inode->i_lock) &&
344 !lockdep_is_held(&inode->i_mapping->i_pages.xa_lock) &&
345 !lockdep_is_held(&inode->i_wb->list_lock)));
351 * unlocked_inode_to_wb_begin - begin unlocked inode wb access transaction
352 * @inode: target inode
353 * @cookie: output param, to be passed to the end function
355 * The caller wants to access the wb associated with @inode but isn't
356 * holding inode->i_lock, the i_pages lock or wb->list_lock. This
357 * function determines the wb associated with @inode and ensures that the
358 * association doesn't change until the transaction is finished with
359 * unlocked_inode_to_wb_end().
361 * The caller must call unlocked_inode_to_wb_end() with *@cookie afterwards and
362 * can't sleep during the transaction. IRQs may or may not be disabled on
365 static inline struct bdi_writeback *
366 unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
371 * Paired with store_release in inode_switch_wbs_work_fn() and
372 * ensures that we see the new wb if we see cleared I_WB_SWITCH.
374 cookie->locked = smp_load_acquire(&inode->i_state) & I_WB_SWITCH;
376 if (unlikely(cookie->locked))
377 xa_lock_irqsave(&inode->i_mapping->i_pages, cookie->flags);
380 * Protected by either !I_WB_SWITCH + rcu_read_lock() or the i_pages
381 * lock. inode_to_wb() will bark. Deref directly.
387 * unlocked_inode_to_wb_end - end inode wb access transaction
388 * @inode: target inode
389 * @cookie: @cookie from unlocked_inode_to_wb_begin()
391 static inline void unlocked_inode_to_wb_end(struct inode *inode,
392 struct wb_lock_cookie *cookie)
394 if (unlikely(cookie->locked))
395 xa_unlock_irqrestore(&inode->i_mapping->i_pages, cookie->flags);
400 #else /* CONFIG_CGROUP_WRITEBACK */
402 static inline bool inode_cgwb_enabled(struct inode *inode)
407 static inline struct bdi_writeback_congested *
408 wb_congested_get_create(struct backing_dev_info *bdi, int blkcg_id, gfp_t gfp)
410 refcount_inc(&bdi->wb_congested->refcnt);
411 return bdi->wb_congested;
414 static inline void wb_congested_put(struct bdi_writeback_congested *congested)
416 if (refcount_dec_and_test(&congested->refcnt))
420 static inline struct bdi_writeback *wb_find_current(struct backing_dev_info *bdi)
425 static inline struct bdi_writeback *
426 wb_get_create_current(struct backing_dev_info *bdi, gfp_t gfp)
431 static inline bool inode_to_wb_is_valid(struct inode *inode)
436 static inline struct bdi_writeback *inode_to_wb(struct inode *inode)
438 return &inode_to_bdi(inode)->wb;
441 static inline struct bdi_writeback *
442 unlocked_inode_to_wb_begin(struct inode *inode, struct wb_lock_cookie *cookie)
444 return inode_to_wb(inode);
447 static inline void unlocked_inode_to_wb_end(struct inode *inode,
448 struct wb_lock_cookie *cookie)
452 static inline void wb_memcg_offline(struct mem_cgroup *memcg)
456 static inline void wb_blkcg_offline(struct blkcg *blkcg)
460 static inline int inode_congested(struct inode *inode, int cong_bits)
462 return wb_congested(&inode_to_bdi(inode)->wb, cong_bits);
465 #endif /* CONFIG_CGROUP_WRITEBACK */
467 static inline int inode_read_congested(struct inode *inode)
469 return inode_congested(inode, 1 << WB_sync_congested);
472 static inline int inode_write_congested(struct inode *inode)
474 return inode_congested(inode, 1 << WB_async_congested);
477 static inline int inode_rw_congested(struct inode *inode)
479 return inode_congested(inode, (1 << WB_sync_congested) |
480 (1 << WB_async_congested));
483 static inline int bdi_congested(struct backing_dev_info *bdi, int cong_bits)
485 return wb_congested(&bdi->wb, cong_bits);
488 static inline int bdi_read_congested(struct backing_dev_info *bdi)
490 return bdi_congested(bdi, 1 << WB_sync_congested);
493 static inline int bdi_write_congested(struct backing_dev_info *bdi)
495 return bdi_congested(bdi, 1 << WB_async_congested);
498 static inline int bdi_rw_congested(struct backing_dev_info *bdi)
500 return bdi_congested(bdi, (1 << WB_sync_congested) |
501 (1 << WB_async_congested));
504 const char *bdi_dev_name(struct backing_dev_info *bdi);
506 #endif /* _LINUX_BACKING_DEV_H */