Commit | Line | Data |
---|---|---|
b2441318 | 1 | /* SPDX-License-Identifier: GPL-2.0 */ |
66114cad TH |
2 | #ifndef __LINUX_BACKING_DEV_DEFS_H |
3 | #define __LINUX_BACKING_DEV_DEFS_H | |
4 | ||
5 | #include <linux/list.h> | |
52ebea74 TH |
6 | #include <linux/radix-tree.h> |
7 | #include <linux/rbtree.h> | |
66114cad TH |
8 | #include <linux/spinlock.h> |
9 | #include <linux/percpu_counter.h> | |
52ebea74 | 10 | #include <linux/percpu-refcount.h> |
66114cad TH |
11 | #include <linux/flex_proportions.h> |
12 | #include <linux/timer.h> | |
13 | #include <linux/workqueue.h> | |
d03f6cdc | 14 | #include <linux/kref.h> |
e58dd0de | 15 | #include <linux/refcount.h> |
66114cad TH |
16 | |
17 | struct page; | |
18 | struct device; | |
19 | struct dentry; | |
20 | ||
21 | /* | |
22 | * Bits in bdi_writeback.state | |
23 | */ | |
24 | enum wb_state { | |
66114cad TH |
25 | WB_registered, /* bdi_register() was done */ |
26 | WB_writeback_running, /* Writeback is in progress */ | |
d6c10f1f | 27 | WB_has_dirty_io, /* Dirty inodes on ->b_{dirty|io|more_io} */ |
aac8d41c | 28 | WB_start_all, /* nr_pages == 0 (all) work pending */ |
66114cad TH |
29 | }; |
30 | ||
66114cad TH |
31 | enum wb_stat_item { |
32 | WB_RECLAIMABLE, | |
33 | WB_WRITEBACK, | |
34 | WB_DIRTIED, | |
35 | WB_WRITTEN, | |
36 | NR_WB_STAT_ITEMS | |
37 | }; | |
38 | ||
39 | #define WB_STAT_BATCH (8*(1+ilog2(nr_cpu_ids))) | |
40 | ||
85009b4f JA |
41 | /* |
42 | * why some writeback work was initiated | |
43 | */ | |
44 | enum wb_reason { | |
45 | WB_REASON_BACKGROUND, | |
46 | WB_REASON_VMSCAN, | |
47 | WB_REASON_SYNC, | |
48 | WB_REASON_PERIODIC, | |
49 | WB_REASON_LAPTOP_TIMER, | |
85009b4f JA |
50 | WB_REASON_FS_FREE_SPACE, |
51 | /* | |
52 | * There is no bdi forker thread any more and works are done | |
53 | * by emergency worker, however, this is TPs userland visible | |
54 | * and we'll be exposing exactly the same information, | |
55 | * so it has a mismatch name. | |
56 | */ | |
57 | WB_REASON_FORKER_THREAD, | |
97b27821 | 58 | WB_REASON_FOREIGN_FLUSH, |
85009b4f JA |
59 | |
60 | WB_REASON_MAX, | |
61 | }; | |
62 | ||
5b9cce4c TH |
63 | struct wb_completion { |
64 | atomic_t cnt; | |
65 | wait_queue_head_t *waitq; | |
66 | }; | |
67 | ||
68 | #define __WB_COMPLETION_INIT(_waitq) \ | |
69 | (struct wb_completion){ .cnt = ATOMIC_INIT(1), .waitq = (_waitq) } | |
70 | ||
71 | /* | |
72 | * If one wants to wait for one or more wb_writeback_works, each work's | |
73 | * ->done should be set to a wb_completion defined using the following | |
74 | * macro. Once all work items are issued with wb_queue_work(), the caller | |
75 | * can wait for the completion of all using wb_wait_for_completion(). Work | |
76 | * items which are waited upon aren't freed automatically on completion. | |
77 | */ | |
78 | #define WB_COMPLETION_INIT(bdi) __WB_COMPLETION_INIT(&(bdi)->wb_waitq) | |
79 | ||
80 | #define DEFINE_WB_COMPLETION(cmpl, bdi) \ | |
81 | struct wb_completion cmpl = WB_COMPLETION_INIT(bdi) | |
82 | ||
52ebea74 TH |
83 | /* |
84 | * Each wb (bdi_writeback) can perform writeback operations, is measured | |
85 | * and throttled, independently. Without cgroup writeback, each bdi | |
86 | * (bdi_writeback) is served by its embedded bdi->wb. | |
87 | * | |
88 | * On the default hierarchy, blkcg implicitly enables memcg. This allows | |
89 | * using memcg's page ownership for attributing writeback IOs, and every | |
90 | * memcg - blkcg combination can be served by its own wb by assigning a | |
91 | * dedicated wb to each memcg, which enables isolation across different | |
92 | * cgroups and propagation of IO back pressure down from the IO layer upto | |
93 | * the tasks which are generating the dirty pages to be written back. | |
94 | * | |
95 | * A cgroup wb is indexed on its bdi by the ID of the associated memcg, | |
96 | * refcounted with the number of inodes attached to it, and pins the memcg | |
97 | * and the corresponding blkcg. As the corresponding blkcg for a memcg may | |
98 | * change as blkcg is disabled and enabled higher up in the hierarchy, a wb | |
99 | * is tested for blkcg after lookup and removed from index on mismatch so | |
100 | * that a new wb for the combination can be created. | |
efee1713 CH |
101 | * |
102 | * Each bdi_writeback that is not embedded into the backing_dev_info must hold | |
103 | * a reference to the parent backing_dev_info. See cgwb_create() for details. | |
52ebea74 | 104 | */ |
66114cad TH |
105 | struct bdi_writeback { |
106 | struct backing_dev_info *bdi; /* our parent bdi */ | |
107 | ||
108 | unsigned long state; /* Always use atomic bitops on this */ | |
109 | unsigned long last_old_flush; /* last old data flush */ | |
110 | ||
111 | struct list_head b_dirty; /* dirty inodes */ | |
112 | struct list_head b_io; /* parked for writeback */ | |
113 | struct list_head b_more_io; /* parked for more writeback */ | |
114 | struct list_head b_dirty_time; /* time stamps are dirty */ | |
115 | spinlock_t list_lock; /* protects the b_* lists */ | |
116 | ||
633a2abb | 117 | atomic_t writeback_inodes; /* number of inodes under writeback */ |
66114cad TH |
118 | struct percpu_counter stat[NR_WB_STAT_ITEMS]; |
119 | ||
120 | unsigned long bw_time_stamp; /* last time write bw is updated */ | |
121 | unsigned long dirtied_stamp; | |
122 | unsigned long written_stamp; /* pages written at bw_time_stamp */ | |
123 | unsigned long write_bandwidth; /* the estimated write bandwidth */ | |
95a46c65 | 124 | unsigned long avg_write_bandwidth; /* further smoothed write bw, > 0 */ |
66114cad TH |
125 | |
126 | /* | |
127 | * The base dirty throttle rate, re-calculated on every 200ms. | |
128 | * All the bdi tasks' dirty rate will be curbed under it. | |
129 | * @dirty_ratelimit tracks the estimated @balanced_dirty_ratelimit | |
130 | * in small steps and is much more smooth/stable than the latter. | |
131 | */ | |
132 | unsigned long dirty_ratelimit; | |
133 | unsigned long balanced_dirty_ratelimit; | |
134 | ||
135 | struct fprop_local_percpu completions; | |
136 | int dirty_exceeded; | |
85009b4f | 137 | enum wb_reason start_all_reason; |
66114cad TH |
138 | |
139 | spinlock_t work_lock; /* protects work_list & dwork scheduling */ | |
140 | struct list_head work_list; | |
141 | struct delayed_work dwork; /* work item used for writeback */ | |
45a2966f | 142 | struct delayed_work bw_dwork; /* work item used for bandwidth estimate */ |
52ebea74 | 143 | |
b57d74af JA |
144 | unsigned long dirty_sleep; /* last wait */ |
145 | ||
b817525a TH |
146 | struct list_head bdi_node; /* anchored at bdi->wb_list */ |
147 | ||
52ebea74 TH |
148 | #ifdef CONFIG_CGROUP_WRITEBACK |
149 | struct percpu_ref refcnt; /* used only for !root wb's */ | |
841710aa | 150 | struct fprop_local_percpu memcg_completions; |
52ebea74 TH |
151 | struct cgroup_subsys_state *memcg_css; /* the associated memcg */ |
152 | struct cgroup_subsys_state *blkcg_css; /* and blkcg */ | |
153 | struct list_head memcg_node; /* anchored at memcg->cgwb_list */ | |
154 | struct list_head blkcg_node; /* anchored at blkcg->cgwb_list */ | |
f3b6a6df | 155 | struct list_head b_attached; /* attached inodes, protected by list_lock */ |
c22d70a1 | 156 | struct list_head offline_node; /* anchored at offline_cgwbs */ |
52ebea74 TH |
157 | |
158 | union { | |
159 | struct work_struct release_work; | |
160 | struct rcu_head rcu; | |
161 | }; | |
162 | #endif | |
66114cad TH |
163 | }; |
164 | ||
165 | struct backing_dev_info { | |
34f8fe50 TH |
166 | u64 id; |
167 | struct rb_node rb_node; /* keyed by ->id */ | |
66114cad | 168 | struct list_head bdi_list; |
ea1754a0 | 169 | unsigned long ra_pages; /* max readahead in PAGE_SIZE units */ |
9491ae4a | 170 | unsigned long io_pages; /* max allowed IO size */ |
66114cad | 171 | |
d03f6cdc | 172 | struct kref refcnt; /* Reference counter for the structure */ |
8db378a5 | 173 | unsigned int capabilities; /* Device capabilities */ |
66114cad TH |
174 | unsigned int min_ratio; |
175 | unsigned int max_ratio, max_prop_frac; | |
176 | ||
95a46c65 TH |
177 | /* |
178 | * Sum of avg_write_bw of wbs with dirty inodes. > 0 if there are | |
179 | * any dirty wbs, which is depended upon by bdi_has_dirty(). | |
180 | */ | |
181 | atomic_long_t tot_write_bandwidth; | |
766a9d6e | 182 | |
52ebea74 | 183 | struct bdi_writeback wb; /* the root writeback info for this bdi */ |
b817525a | 184 | struct list_head wb_list; /* list of all wbs */ |
52ebea74 TH |
185 | #ifdef CONFIG_CGROUP_WRITEBACK |
186 | struct radix_tree_root cgwb_tree; /* radix tree of active cgroup wbs */ | |
3ee7e869 | 187 | struct mutex cgwb_release_mutex; /* protect shutdown of wb structs */ |
7fc5854f | 188 | struct rw_semaphore wb_switch_rwsem; /* no cgwb switch while syncing */ |
52ebea74 | 189 | #endif |
cc395d7f TH |
190 | wait_queue_head_t wb_waitq; |
191 | ||
66114cad | 192 | struct device *dev; |
6bd87eec | 193 | char dev_name[64]; |
df08c32c | 194 | struct device *owner; |
66114cad TH |
195 | |
196 | struct timer_list laptop_mode_wb_timer; | |
197 | ||
198 | #ifdef CONFIG_DEBUG_FS | |
199 | struct dentry *debug_dir; | |
66114cad TH |
200 | #endif |
201 | }; | |
202 | ||
2e898e4c GT |
203 | struct wb_lock_cookie { |
204 | bool locked; | |
205 | unsigned long flags; | |
206 | }; | |
207 | ||
21c6321f TH |
208 | #ifdef CONFIG_CGROUP_WRITEBACK |
209 | ||
210 | /** | |
211 | * wb_tryget - try to increment a wb's refcount | |
212 | * @wb: bdi_writeback to get | |
213 | */ | |
214 | static inline bool wb_tryget(struct bdi_writeback *wb) | |
215 | { | |
216 | if (wb != &wb->bdi->wb) | |
217 | return percpu_ref_tryget(&wb->refcnt); | |
218 | return true; | |
219 | } | |
220 | ||
221 | /** | |
222 | * wb_get - increment a wb's refcount | |
223 | * @wb: bdi_writeback to get | |
224 | */ | |
225 | static inline void wb_get(struct bdi_writeback *wb) | |
226 | { | |
227 | if (wb != &wb->bdi->wb) | |
228 | percpu_ref_get(&wb->refcnt); | |
229 | } | |
230 | ||
231 | /** | |
232 | * wb_put - decrement a wb's refcount | |
233 | * @wb: bdi_writeback to put | |
f5fbe6b7 | 234 | * @nr: number of references to put |
21c6321f | 235 | */ |
f5fbe6b7 | 236 | static inline void wb_put_many(struct bdi_writeback *wb, unsigned long nr) |
21c6321f | 237 | { |
368686a9 AR |
238 | if (WARN_ON_ONCE(!wb->bdi)) { |
239 | /* | |
240 | * A driver bug might cause a file to be removed before bdi was | |
241 | * initialized. | |
242 | */ | |
243 | return; | |
244 | } | |
245 | ||
21c6321f | 246 | if (wb != &wb->bdi->wb) |
f5fbe6b7 RG |
247 | percpu_ref_put_many(&wb->refcnt, nr); |
248 | } | |
249 | ||
250 | /** | |
251 | * wb_put - decrement a wb's refcount | |
252 | * @wb: bdi_writeback to put | |
253 | */ | |
254 | static inline void wb_put(struct bdi_writeback *wb) | |
255 | { | |
256 | wb_put_many(wb, 1); | |
21c6321f TH |
257 | } |
258 | ||
e8a7abf5 TH |
259 | /** |
260 | * wb_dying - is a wb dying? | |
261 | * @wb: bdi_writeback of interest | |
262 | * | |
263 | * Returns whether @wb is unlinked and being drained. | |
264 | */ | |
265 | static inline bool wb_dying(struct bdi_writeback *wb) | |
266 | { | |
267 | return percpu_ref_is_dying(&wb->refcnt); | |
268 | } | |
269 | ||
21c6321f TH |
270 | #else /* CONFIG_CGROUP_WRITEBACK */ |
271 | ||
272 | static inline bool wb_tryget(struct bdi_writeback *wb) | |
273 | { | |
274 | return true; | |
275 | } | |
276 | ||
277 | static inline void wb_get(struct bdi_writeback *wb) | |
278 | { | |
279 | } | |
280 | ||
281 | static inline void wb_put(struct bdi_writeback *wb) | |
282 | { | |
283 | } | |
284 | ||
f5fbe6b7 RG |
285 | static inline void wb_put_many(struct bdi_writeback *wb, unsigned long nr) |
286 | { | |
287 | } | |
288 | ||
e8a7abf5 TH |
289 | static inline bool wb_dying(struct bdi_writeback *wb) |
290 | { | |
291 | return false; | |
292 | } | |
293 | ||
21c6321f TH |
294 | #endif /* CONFIG_CGROUP_WRITEBACK */ |
295 | ||
66114cad | 296 | #endif /* __LINUX_BACKING_DEV_DEFS_H */ |