Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * include/linux/backing-dev.h | |
3 | * | |
4 | * low-level device information and state which is propagated up through | |
5 | * to high-level code. | |
6 | */ | |
7 | ||
8 | #ifndef _LINUX_BACKING_DEV_H | |
9 | #define _LINUX_BACKING_DEV_H | |
10 | ||
b2e8fb6e PZ |
11 | #include <linux/percpu_counter.h> |
12 | #include <linux/log2.h> | |
eb608e3a | 13 | #include <linux/flex_proportions.h> |
cf0ca9fe | 14 | #include <linux/kernel.h> |
e4ad08fe | 15 | #include <linux/fs.h> |
03ba3782 | 16 | #include <linux/sched.h> |
31373d09 | 17 | #include <linux/timer.h> |
03ba3782 | 18 | #include <linux/writeback.h> |
60063497 | 19 | #include <linux/atomic.h> |
3965c9ae | 20 | #include <linux/sysctl.h> |
839a8e86 | 21 | #include <linux/workqueue.h> |
1da177e4 | 22 | |
3fcfab16 | 23 | struct page; |
cf0ca9fe | 24 | struct device; |
76f1418b | 25 | struct dentry; |
3fcfab16 | 26 | |
1da177e4 | 27 | /* |
4452226e | 28 | * Bits in bdi_writeback.state |
1da177e4 | 29 | */ |
4452226e TH |
30 | enum wb_state { |
31 | WB_async_congested, /* The async (write) queue is getting full */ | |
32 | WB_sync_congested, /* The sync queue is getting full */ | |
33 | WB_registered, /* bdi_register() was done */ | |
34 | WB_writeback_running, /* Writeback is in progress */ | |
1da177e4 LT |
35 | }; |
36 | ||
37 | typedef int (congested_fn)(void *, int); | |
38 | ||
93f78d88 TH |
39 | enum wb_stat_item { |
40 | WB_RECLAIMABLE, | |
41 | WB_WRITEBACK, | |
42 | WB_DIRTIED, | |
43 | WB_WRITTEN, | |
44 | NR_WB_STAT_ITEMS | |
b2e8fb6e PZ |
45 | }; |
46 | ||
93f78d88 | 47 | #define WB_STAT_BATCH (8*(1+ilog2(nr_cpu_ids))) |
b2e8fb6e | 48 | |
03ba3782 | 49 | struct bdi_writeback { |
ecd58403 | 50 | struct backing_dev_info *bdi; /* our parent bdi */ |
03ba3782 | 51 | |
4452226e | 52 | unsigned long state; /* Always use atomic bitops on this */ |
ecd58403 | 53 | unsigned long last_old_flush; /* last old data flush */ |
03ba3782 | 54 | |
839a8e86 | 55 | struct delayed_work dwork; /* work item used for writeback */ |
ecd58403 AB |
56 | struct list_head b_dirty; /* dirty inodes */ |
57 | struct list_head b_io; /* parked for writeback */ | |
58 | struct list_head b_more_io; /* parked for more writeback */ | |
0ae45f63 | 59 | struct list_head b_dirty_time; /* time stamps are dirty */ |
f758eeab | 60 | spinlock_t list_lock; /* protects the b_* lists */ |
93f78d88 TH |
61 | |
62 | struct percpu_counter stat[NR_WB_STAT_ITEMS]; | |
d993831f | 63 | |
e98be2d5 | 64 | unsigned long bw_time_stamp; /* last time write bw is updated */ |
be3ffa27 | 65 | unsigned long dirtied_stamp; |
e98be2d5 WF |
66 | unsigned long written_stamp; /* pages written at bw_time_stamp */ |
67 | unsigned long write_bandwidth; /* the estimated write bandwidth */ | |
68 | unsigned long avg_write_bandwidth; /* further smoothed write bw */ | |
69 | ||
be3ffa27 WF |
70 | /* |
71 | * The base dirty throttle rate, re-calculated on every 200ms. | |
72 | * All the bdi tasks' dirty rate will be curbed under it. | |
7381131c WF |
73 | * @dirty_ratelimit tracks the estimated @balanced_dirty_ratelimit |
74 | * in small steps and is much more smooth/stable than the latter. | |
be3ffa27 WF |
75 | */ |
76 | unsigned long dirty_ratelimit; | |
7381131c | 77 | unsigned long balanced_dirty_ratelimit; |
be3ffa27 | 78 | |
eb608e3a | 79 | struct fprop_local_percpu completions; |
04fbfdc1 | 80 | int dirty_exceeded; |
a88a341a TH |
81 | }; |
82 | ||
83 | struct backing_dev_info { | |
84 | struct list_head bdi_list; | |
85 | unsigned long ra_pages; /* max readahead in PAGE_CACHE_SIZE units */ | |
86 | unsigned int capabilities; /* Device capabilities */ | |
87 | congested_fn *congested_fn; /* Function pointer if device is md/dm */ | |
88 | void *congested_data; /* Pointer to aux data for congested func */ | |
89 | ||
90 | char *name; | |
cf0ca9fe | 91 | |
189d3c4a | 92 | unsigned int min_ratio; |
a42dde04 | 93 | unsigned int max_ratio, max_prop_frac; |
189d3c4a | 94 | |
03ba3782 | 95 | struct bdi_writeback wb; /* default writeback info for this bdi */ |
5acda9d1 | 96 | spinlock_t wb_lock; /* protects work_list & wb.dwork scheduling */ |
76f1418b | 97 | |
03ba3782 JA |
98 | struct list_head work_list; |
99 | ||
100 | struct device *dev; | |
66f3b8e2 | 101 | |
31373d09 MG |
102 | struct timer_list laptop_mode_wb_timer; |
103 | ||
76f1418b MS |
104 | #ifdef CONFIG_DEBUG_FS |
105 | struct dentry *debug_dir; | |
106 | struct dentry *debug_stats; | |
107 | #endif | |
1da177e4 LT |
108 | }; |
109 | ||
de1414a6 CH |
110 | struct backing_dev_info *inode_to_bdi(struct inode *inode); |
111 | ||
8077c0d9 | 112 | int __must_check bdi_init(struct backing_dev_info *bdi); |
b2e8fb6e PZ |
113 | void bdi_destroy(struct backing_dev_info *bdi); |
114 | ||
d2cc4dde | 115 | __printf(3, 4) |
cf0ca9fe PZ |
116 | int bdi_register(struct backing_dev_info *bdi, struct device *parent, |
117 | const char *fmt, ...); | |
118 | int bdi_register_dev(struct backing_dev_info *bdi, dev_t dev); | |
119 | void bdi_unregister(struct backing_dev_info *bdi); | |
b4caecd4 | 120 | int __must_check bdi_setup_and_register(struct backing_dev_info *, char *); |
0e175a18 CW |
121 | void bdi_start_writeback(struct backing_dev_info *bdi, long nr_pages, |
122 | enum wb_reason reason); | |
c5444198 | 123 | void bdi_start_background_writeback(struct backing_dev_info *bdi); |
839a8e86 | 124 | void bdi_writeback_workfn(struct work_struct *work); |
03ba3782 | 125 | int bdi_has_dirty_io(struct backing_dev_info *bdi); |
6467716a | 126 | void bdi_wakeup_thread_delayed(struct backing_dev_info *bdi); |
cf0ca9fe | 127 | |
03ba3782 | 128 | extern spinlock_t bdi_lock; |
66f3b8e2 JA |
129 | extern struct list_head bdi_list; |
130 | ||
839a8e86 TH |
131 | extern struct workqueue_struct *bdi_wq; |
132 | ||
03ba3782 JA |
133 | static inline int wb_has_dirty_io(struct bdi_writeback *wb) |
134 | { | |
135 | return !list_empty(&wb->b_dirty) || | |
136 | !list_empty(&wb->b_io) || | |
137 | !list_empty(&wb->b_more_io); | |
138 | } | |
139 | ||
93f78d88 TH |
140 | static inline void __add_wb_stat(struct bdi_writeback *wb, |
141 | enum wb_stat_item item, s64 amount) | |
b2e8fb6e | 142 | { |
93f78d88 | 143 | __percpu_counter_add(&wb->stat[item], amount, WB_STAT_BATCH); |
b2e8fb6e PZ |
144 | } |
145 | ||
93f78d88 TH |
146 | static inline void __inc_wb_stat(struct bdi_writeback *wb, |
147 | enum wb_stat_item item) | |
b2e8fb6e | 148 | { |
93f78d88 | 149 | __add_wb_stat(wb, item, 1); |
b2e8fb6e PZ |
150 | } |
151 | ||
93f78d88 | 152 | static inline void inc_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item) |
b2e8fb6e PZ |
153 | { |
154 | unsigned long flags; | |
155 | ||
156 | local_irq_save(flags); | |
93f78d88 | 157 | __inc_wb_stat(wb, item); |
b2e8fb6e PZ |
158 | local_irq_restore(flags); |
159 | } | |
160 | ||
93f78d88 TH |
161 | static inline void __dec_wb_stat(struct bdi_writeback *wb, |
162 | enum wb_stat_item item) | |
b2e8fb6e | 163 | { |
93f78d88 | 164 | __add_wb_stat(wb, item, -1); |
b2e8fb6e PZ |
165 | } |
166 | ||
93f78d88 | 167 | static inline void dec_wb_stat(struct bdi_writeback *wb, enum wb_stat_item item) |
b2e8fb6e PZ |
168 | { |
169 | unsigned long flags; | |
170 | ||
171 | local_irq_save(flags); | |
93f78d88 | 172 | __dec_wb_stat(wb, item); |
b2e8fb6e PZ |
173 | local_irq_restore(flags); |
174 | } | |
175 | ||
93f78d88 | 176 | static inline s64 wb_stat(struct bdi_writeback *wb, enum wb_stat_item item) |
b2e8fb6e | 177 | { |
93f78d88 | 178 | return percpu_counter_read_positive(&wb->stat[item]); |
b2e8fb6e PZ |
179 | } |
180 | ||
93f78d88 TH |
181 | static inline s64 __wb_stat_sum(struct bdi_writeback *wb, |
182 | enum wb_stat_item item) | |
b2e8fb6e | 183 | { |
93f78d88 | 184 | return percpu_counter_sum_positive(&wb->stat[item]); |
b2e8fb6e PZ |
185 | } |
186 | ||
93f78d88 | 187 | static inline s64 wb_stat_sum(struct bdi_writeback *wb, enum wb_stat_item item) |
e0bf68dd | 188 | { |
b2e8fb6e PZ |
189 | s64 sum; |
190 | unsigned long flags; | |
191 | ||
192 | local_irq_save(flags); | |
93f78d88 | 193 | sum = __wb_stat_sum(wb, item); |
b2e8fb6e PZ |
194 | local_irq_restore(flags); |
195 | ||
196 | return sum; | |
e0bf68dd PZ |
197 | } |
198 | ||
93f78d88 | 199 | extern void wb_writeout_inc(struct bdi_writeback *wb); |
dd5656e5 | 200 | |
b2e8fb6e PZ |
201 | /* |
202 | * maximal error of a stat counter. | |
203 | */ | |
93f78d88 | 204 | static inline unsigned long wb_stat_error(struct bdi_writeback *wb) |
e0bf68dd | 205 | { |
b2e8fb6e | 206 | #ifdef CONFIG_SMP |
93f78d88 | 207 | return nr_cpu_ids * WB_STAT_BATCH; |
b2e8fb6e PZ |
208 | #else |
209 | return 1; | |
210 | #endif | |
e0bf68dd | 211 | } |
1da177e4 | 212 | |
189d3c4a | 213 | int bdi_set_min_ratio(struct backing_dev_info *bdi, unsigned int min_ratio); |
a42dde04 | 214 | int bdi_set_max_ratio(struct backing_dev_info *bdi, unsigned int max_ratio); |
189d3c4a | 215 | |
1da177e4 LT |
216 | /* |
217 | * Flags in backing_dev_info::capability | |
e4ad08fe MS |
218 | * |
219 | * The first three flags control whether dirty pages will contribute to the | |
220 | * VM's accounting and whether writepages() should be called for dirty pages | |
221 | * (something that would not, for example, be appropriate for ramfs) | |
222 | * | |
223 | * WARNING: these flags are closely related and should not normally be | |
224 | * used separately. The BDI_CAP_NO_ACCT_AND_WRITEBACK combines these | |
225 | * three flags into a single convenience macro. | |
226 | * | |
227 | * BDI_CAP_NO_ACCT_DIRTY: Dirty pages shouldn't contribute to accounting | |
228 | * BDI_CAP_NO_WRITEBACK: Don't write pages back | |
229 | * BDI_CAP_NO_ACCT_WB: Don't automatically account writeback pages | |
5a537485 | 230 | * BDI_CAP_STRICTLIMIT: Keep number of dirty pages below bdi threshold. |
1da177e4 | 231 | */ |
e4ad08fe MS |
232 | #define BDI_CAP_NO_ACCT_DIRTY 0x00000001 |
233 | #define BDI_CAP_NO_WRITEBACK 0x00000002 | |
b4caecd4 CH |
234 | #define BDI_CAP_NO_ACCT_WB 0x00000004 |
235 | #define BDI_CAP_STABLE_WRITES 0x00000008 | |
236 | #define BDI_CAP_STRICTLIMIT 0x00000010 | |
1da177e4 | 237 | |
e4ad08fe MS |
238 | #define BDI_CAP_NO_ACCT_AND_WRITEBACK \ |
239 | (BDI_CAP_NO_WRITEBACK | BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_ACCT_WB) | |
240 | ||
5129a469 | 241 | extern struct backing_dev_info noop_backing_dev_info; |
1da177e4 | 242 | |
1da177e4 | 243 | int writeback_in_progress(struct backing_dev_info *bdi); |
1da177e4 LT |
244 | |
245 | static inline int bdi_congested(struct backing_dev_info *bdi, int bdi_bits) | |
246 | { | |
247 | if (bdi->congested_fn) | |
248 | return bdi->congested_fn(bdi->congested_data, bdi_bits); | |
4452226e | 249 | return (bdi->wb.state & bdi_bits); |
1da177e4 LT |
250 | } |
251 | ||
252 | static inline int bdi_read_congested(struct backing_dev_info *bdi) | |
253 | { | |
4452226e | 254 | return bdi_congested(bdi, 1 << WB_sync_congested); |
1da177e4 LT |
255 | } |
256 | ||
257 | static inline int bdi_write_congested(struct backing_dev_info *bdi) | |
258 | { | |
4452226e | 259 | return bdi_congested(bdi, 1 << WB_async_congested); |
1da177e4 LT |
260 | } |
261 | ||
262 | static inline int bdi_rw_congested(struct backing_dev_info *bdi) | |
263 | { | |
4452226e TH |
264 | return bdi_congested(bdi, (1 << WB_sync_congested) | |
265 | (1 << WB_async_congested)); | |
1da177e4 LT |
266 | } |
267 | ||
373c0a7e TM |
268 | enum { |
269 | BLK_RW_ASYNC = 0, | |
270 | BLK_RW_SYNC = 1, | |
271 | }; | |
272 | ||
8aa7e847 JA |
273 | void clear_bdi_congested(struct backing_dev_info *bdi, int sync); |
274 | void set_bdi_congested(struct backing_dev_info *bdi, int sync); | |
275 | long congestion_wait(int sync, long timeout); | |
0e093d99 | 276 | long wait_iff_congested(struct zone *zone, int sync, long timeout); |
3965c9ae WL |
277 | int pdflush_proc_obsolete(struct ctl_table *table, int write, |
278 | void __user *buffer, size_t *lenp, loff_t *ppos); | |
1da177e4 | 279 | |
7d311cda DW |
280 | static inline bool bdi_cap_stable_pages_required(struct backing_dev_info *bdi) |
281 | { | |
282 | return bdi->capabilities & BDI_CAP_STABLE_WRITES; | |
283 | } | |
284 | ||
e4ad08fe MS |
285 | static inline bool bdi_cap_writeback_dirty(struct backing_dev_info *bdi) |
286 | { | |
287 | return !(bdi->capabilities & BDI_CAP_NO_WRITEBACK); | |
288 | } | |
289 | ||
290 | static inline bool bdi_cap_account_dirty(struct backing_dev_info *bdi) | |
291 | { | |
292 | return !(bdi->capabilities & BDI_CAP_NO_ACCT_DIRTY); | |
293 | } | |
1da177e4 | 294 | |
e4ad08fe MS |
295 | static inline bool bdi_cap_account_writeback(struct backing_dev_info *bdi) |
296 | { | |
297 | /* Paranoia: BDI_CAP_NO_WRITEBACK implies BDI_CAP_NO_ACCT_WB */ | |
298 | return !(bdi->capabilities & (BDI_CAP_NO_ACCT_WB | | |
299 | BDI_CAP_NO_WRITEBACK)); | |
300 | } | |
1da177e4 | 301 | |
e4ad08fe MS |
302 | static inline bool mapping_cap_writeback_dirty(struct address_space *mapping) |
303 | { | |
de1414a6 | 304 | return bdi_cap_writeback_dirty(inode_to_bdi(mapping->host)); |
e4ad08fe | 305 | } |
1da177e4 | 306 | |
e4ad08fe MS |
307 | static inline bool mapping_cap_account_dirty(struct address_space *mapping) |
308 | { | |
de1414a6 | 309 | return bdi_cap_account_dirty(inode_to_bdi(mapping->host)); |
e4ad08fe | 310 | } |
1da177e4 | 311 | |
03ba3782 JA |
312 | static inline int bdi_sched_wait(void *word) |
313 | { | |
314 | schedule(); | |
315 | return 0; | |
316 | } | |
317 | ||
1da177e4 | 318 | #endif /* _LINUX_BACKING_DEV_H */ |