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 LT |
27 | /* |
28 | * Bits in backing_dev_info.state | |
29 | */ | |
30 | enum bdi_state { | |
1faa16d2 JA |
31 | BDI_async_congested, /* The async (write) queue is getting full */ |
32 | BDI_sync_congested, /* The sync queue is getting full */ | |
500b067c | 33 | BDI_registered, /* bdi_register() was done */ |
81d73a32 | 34 | BDI_writeback_running, /* Writeback is in progress */ |
1da177e4 LT |
35 | }; |
36 | ||
37 | typedef int (congested_fn)(void *, int); | |
38 | ||
b2e8fb6e | 39 | enum bdi_stat_item { |
c9e51e41 | 40 | BDI_RECLAIMABLE, |
69cb51d1 | 41 | BDI_WRITEBACK, |
c8e28ce0 | 42 | BDI_DIRTIED, |
f7d2b1ec | 43 | BDI_WRITTEN, |
b2e8fb6e PZ |
44 | NR_BDI_STAT_ITEMS |
45 | }; | |
46 | ||
47 | #define BDI_STAT_BATCH (8*(1+ilog2(nr_cpu_ids))) | |
48 | ||
03ba3782 | 49 | struct bdi_writeback { |
ecd58403 | 50 | struct backing_dev_info *bdi; /* our parent bdi */ |
03ba3782 | 51 | |
ecd58403 | 52 | unsigned long last_old_flush; /* last old data flush */ |
03ba3782 | 53 | |
839a8e86 | 54 | struct delayed_work dwork; /* work item used for writeback */ |
ecd58403 AB |
55 | struct list_head b_dirty; /* dirty inodes */ |
56 | struct list_head b_io; /* parked for writeback */ | |
57 | struct list_head b_more_io; /* parked for more writeback */ | |
f758eeab | 58 | spinlock_t list_lock; /* protects the b_* lists */ |
03ba3782 JA |
59 | }; |
60 | ||
1da177e4 | 61 | struct backing_dev_info { |
66f3b8e2 | 62 | struct list_head bdi_list; |
1da177e4 LT |
63 | unsigned long ra_pages; /* max readahead in PAGE_CACHE_SIZE units */ |
64 | unsigned long state; /* Always use atomic bitops on this */ | |
65 | unsigned int capabilities; /* Device capabilities */ | |
66 | congested_fn *congested_fn; /* Function pointer if device is md/dm */ | |
67 | void *congested_data; /* Pointer to aux data for congested func */ | |
b2e8fb6e | 68 | |
d993831f JA |
69 | char *name; |
70 | ||
b2e8fb6e | 71 | struct percpu_counter bdi_stat[NR_BDI_STAT_ITEMS]; |
04fbfdc1 | 72 | |
e98be2d5 | 73 | unsigned long bw_time_stamp; /* last time write bw is updated */ |
be3ffa27 | 74 | unsigned long dirtied_stamp; |
e98be2d5 WF |
75 | unsigned long written_stamp; /* pages written at bw_time_stamp */ |
76 | unsigned long write_bandwidth; /* the estimated write bandwidth */ | |
77 | unsigned long avg_write_bandwidth; /* further smoothed write bw */ | |
78 | ||
be3ffa27 WF |
79 | /* |
80 | * The base dirty throttle rate, re-calculated on every 200ms. | |
81 | * All the bdi tasks' dirty rate will be curbed under it. | |
7381131c WF |
82 | * @dirty_ratelimit tracks the estimated @balanced_dirty_ratelimit |
83 | * in small steps and is much more smooth/stable than the latter. | |
be3ffa27 WF |
84 | */ |
85 | unsigned long dirty_ratelimit; | |
7381131c | 86 | unsigned long balanced_dirty_ratelimit; |
be3ffa27 | 87 | |
eb608e3a | 88 | struct fprop_local_percpu completions; |
04fbfdc1 | 89 | int dirty_exceeded; |
cf0ca9fe | 90 | |
189d3c4a | 91 | unsigned int min_ratio; |
a42dde04 | 92 | unsigned int max_ratio, max_prop_frac; |
189d3c4a | 93 | |
03ba3782 | 94 | struct bdi_writeback wb; /* default writeback info for this bdi */ |
5acda9d1 | 95 | spinlock_t wb_lock; /* protects work_list & wb.dwork scheduling */ |
76f1418b | 96 | |
03ba3782 JA |
97 | struct list_head work_list; |
98 | ||
99 | struct device *dev; | |
66f3b8e2 | 100 | |
31373d09 MG |
101 | struct timer_list laptop_mode_wb_timer; |
102 | ||
76f1418b MS |
103 | #ifdef CONFIG_DEBUG_FS |
104 | struct dentry *debug_dir; | |
105 | struct dentry *debug_stats; | |
106 | #endif | |
1da177e4 LT |
107 | }; |
108 | ||
8077c0d9 | 109 | int __must_check bdi_init(struct backing_dev_info *bdi); |
b2e8fb6e PZ |
110 | void bdi_destroy(struct backing_dev_info *bdi); |
111 | ||
d2cc4dde | 112 | __printf(3, 4) |
cf0ca9fe PZ |
113 | int bdi_register(struct backing_dev_info *bdi, struct device *parent, |
114 | const char *fmt, ...); | |
115 | int bdi_register_dev(struct backing_dev_info *bdi, dev_t dev); | |
116 | void bdi_unregister(struct backing_dev_info *bdi); | |
b4caecd4 | 117 | int __must_check bdi_setup_and_register(struct backing_dev_info *, char *); |
0e175a18 CW |
118 | void bdi_start_writeback(struct backing_dev_info *bdi, long nr_pages, |
119 | enum wb_reason reason); | |
c5444198 | 120 | void bdi_start_background_writeback(struct backing_dev_info *bdi); |
839a8e86 | 121 | void bdi_writeback_workfn(struct work_struct *work); |
03ba3782 | 122 | int bdi_has_dirty_io(struct backing_dev_info *bdi); |
6467716a | 123 | void bdi_wakeup_thread_delayed(struct backing_dev_info *bdi); |
cf0ca9fe | 124 | |
03ba3782 | 125 | extern spinlock_t bdi_lock; |
66f3b8e2 JA |
126 | extern struct list_head bdi_list; |
127 | ||
839a8e86 TH |
128 | extern struct workqueue_struct *bdi_wq; |
129 | ||
03ba3782 JA |
130 | static inline int wb_has_dirty_io(struct bdi_writeback *wb) |
131 | { | |
132 | return !list_empty(&wb->b_dirty) || | |
133 | !list_empty(&wb->b_io) || | |
134 | !list_empty(&wb->b_more_io); | |
135 | } | |
136 | ||
b2e8fb6e PZ |
137 | static inline void __add_bdi_stat(struct backing_dev_info *bdi, |
138 | enum bdi_stat_item item, s64 amount) | |
139 | { | |
140 | __percpu_counter_add(&bdi->bdi_stat[item], amount, BDI_STAT_BATCH); | |
141 | } | |
142 | ||
143 | static inline void __inc_bdi_stat(struct backing_dev_info *bdi, | |
144 | enum bdi_stat_item item) | |
145 | { | |
146 | __add_bdi_stat(bdi, item, 1); | |
147 | } | |
148 | ||
149 | static inline void inc_bdi_stat(struct backing_dev_info *bdi, | |
150 | enum bdi_stat_item item) | |
151 | { | |
152 | unsigned long flags; | |
153 | ||
154 | local_irq_save(flags); | |
155 | __inc_bdi_stat(bdi, item); | |
156 | local_irq_restore(flags); | |
157 | } | |
158 | ||
159 | static inline void __dec_bdi_stat(struct backing_dev_info *bdi, | |
160 | enum bdi_stat_item item) | |
161 | { | |
162 | __add_bdi_stat(bdi, item, -1); | |
163 | } | |
164 | ||
165 | static inline void dec_bdi_stat(struct backing_dev_info *bdi, | |
166 | enum bdi_stat_item item) | |
167 | { | |
168 | unsigned long flags; | |
169 | ||
170 | local_irq_save(flags); | |
171 | __dec_bdi_stat(bdi, item); | |
172 | local_irq_restore(flags); | |
173 | } | |
174 | ||
175 | static inline s64 bdi_stat(struct backing_dev_info *bdi, | |
176 | enum bdi_stat_item item) | |
177 | { | |
178 | return percpu_counter_read_positive(&bdi->bdi_stat[item]); | |
179 | } | |
180 | ||
181 | static inline s64 __bdi_stat_sum(struct backing_dev_info *bdi, | |
182 | enum bdi_stat_item item) | |
183 | { | |
184 | return percpu_counter_sum_positive(&bdi->bdi_stat[item]); | |
185 | } | |
186 | ||
187 | static inline s64 bdi_stat_sum(struct backing_dev_info *bdi, | |
188 | enum bdi_stat_item item) | |
e0bf68dd | 189 | { |
b2e8fb6e PZ |
190 | s64 sum; |
191 | unsigned long flags; | |
192 | ||
193 | local_irq_save(flags); | |
194 | sum = __bdi_stat_sum(bdi, item); | |
195 | local_irq_restore(flags); | |
196 | ||
197 | return sum; | |
e0bf68dd PZ |
198 | } |
199 | ||
dd5656e5 MS |
200 | extern void bdi_writeout_inc(struct backing_dev_info *bdi); |
201 | ||
b2e8fb6e PZ |
202 | /* |
203 | * maximal error of a stat counter. | |
204 | */ | |
205 | static inline unsigned long bdi_stat_error(struct backing_dev_info *bdi) | |
e0bf68dd | 206 | { |
b2e8fb6e PZ |
207 | #ifdef CONFIG_SMP |
208 | return nr_cpu_ids * BDI_STAT_BATCH; | |
209 | #else | |
210 | return 1; | |
211 | #endif | |
e0bf68dd | 212 | } |
1da177e4 | 213 | |
189d3c4a | 214 | int bdi_set_min_ratio(struct backing_dev_info *bdi, unsigned int min_ratio); |
a42dde04 | 215 | int bdi_set_max_ratio(struct backing_dev_info *bdi, unsigned int max_ratio); |
189d3c4a | 216 | |
1da177e4 LT |
217 | /* |
218 | * Flags in backing_dev_info::capability | |
e4ad08fe MS |
219 | * |
220 | * The first three flags control whether dirty pages will contribute to the | |
221 | * VM's accounting and whether writepages() should be called for dirty pages | |
222 | * (something that would not, for example, be appropriate for ramfs) | |
223 | * | |
224 | * WARNING: these flags are closely related and should not normally be | |
225 | * used separately. The BDI_CAP_NO_ACCT_AND_WRITEBACK combines these | |
226 | * three flags into a single convenience macro. | |
227 | * | |
228 | * BDI_CAP_NO_ACCT_DIRTY: Dirty pages shouldn't contribute to accounting | |
229 | * BDI_CAP_NO_WRITEBACK: Don't write pages back | |
230 | * BDI_CAP_NO_ACCT_WB: Don't automatically account writeback pages | |
5a537485 | 231 | * BDI_CAP_STRICTLIMIT: Keep number of dirty pages below bdi threshold. |
1da177e4 | 232 | */ |
e4ad08fe MS |
233 | #define BDI_CAP_NO_ACCT_DIRTY 0x00000001 |
234 | #define BDI_CAP_NO_WRITEBACK 0x00000002 | |
b4caecd4 CH |
235 | #define BDI_CAP_NO_ACCT_WB 0x00000004 |
236 | #define BDI_CAP_STABLE_WRITES 0x00000008 | |
237 | #define BDI_CAP_STRICTLIMIT 0x00000010 | |
1da177e4 | 238 | |
e4ad08fe MS |
239 | #define BDI_CAP_NO_ACCT_AND_WRITEBACK \ |
240 | (BDI_CAP_NO_WRITEBACK | BDI_CAP_NO_ACCT_DIRTY | BDI_CAP_NO_ACCT_WB) | |
241 | ||
1da177e4 | 242 | extern struct backing_dev_info default_backing_dev_info; |
5129a469 | 243 | extern struct backing_dev_info noop_backing_dev_info; |
1da177e4 | 244 | |
1da177e4 | 245 | int writeback_in_progress(struct backing_dev_info *bdi); |
1da177e4 LT |
246 | |
247 | static inline int bdi_congested(struct backing_dev_info *bdi, int bdi_bits) | |
248 | { | |
249 | if (bdi->congested_fn) | |
250 | return bdi->congested_fn(bdi->congested_data, bdi_bits); | |
251 | return (bdi->state & bdi_bits); | |
252 | } | |
253 | ||
254 | static inline int bdi_read_congested(struct backing_dev_info *bdi) | |
255 | { | |
1faa16d2 | 256 | return bdi_congested(bdi, 1 << BDI_sync_congested); |
1da177e4 LT |
257 | } |
258 | ||
259 | static inline int bdi_write_congested(struct backing_dev_info *bdi) | |
260 | { | |
1faa16d2 | 261 | return bdi_congested(bdi, 1 << BDI_async_congested); |
1da177e4 LT |
262 | } |
263 | ||
264 | static inline int bdi_rw_congested(struct backing_dev_info *bdi) | |
265 | { | |
1faa16d2 JA |
266 | return bdi_congested(bdi, (1 << BDI_sync_congested) | |
267 | (1 << BDI_async_congested)); | |
1da177e4 LT |
268 | } |
269 | ||
373c0a7e TM |
270 | enum { |
271 | BLK_RW_ASYNC = 0, | |
272 | BLK_RW_SYNC = 1, | |
273 | }; | |
274 | ||
8aa7e847 JA |
275 | void clear_bdi_congested(struct backing_dev_info *bdi, int sync); |
276 | void set_bdi_congested(struct backing_dev_info *bdi, int sync); | |
277 | long congestion_wait(int sync, long timeout); | |
0e093d99 | 278 | long wait_iff_congested(struct zone *zone, int sync, long timeout); |
3965c9ae WL |
279 | int pdflush_proc_obsolete(struct ctl_table *table, int write, |
280 | void __user *buffer, size_t *lenp, loff_t *ppos); | |
1da177e4 | 281 | |
7d311cda DW |
282 | static inline bool bdi_cap_stable_pages_required(struct backing_dev_info *bdi) |
283 | { | |
284 | return bdi->capabilities & BDI_CAP_STABLE_WRITES; | |
285 | } | |
286 | ||
e4ad08fe MS |
287 | static inline bool bdi_cap_writeback_dirty(struct backing_dev_info *bdi) |
288 | { | |
289 | return !(bdi->capabilities & BDI_CAP_NO_WRITEBACK); | |
290 | } | |
291 | ||
292 | static inline bool bdi_cap_account_dirty(struct backing_dev_info *bdi) | |
293 | { | |
294 | return !(bdi->capabilities & BDI_CAP_NO_ACCT_DIRTY); | |
295 | } | |
1da177e4 | 296 | |
e4ad08fe MS |
297 | static inline bool bdi_cap_account_writeback(struct backing_dev_info *bdi) |
298 | { | |
299 | /* Paranoia: BDI_CAP_NO_WRITEBACK implies BDI_CAP_NO_ACCT_WB */ | |
300 | return !(bdi->capabilities & (BDI_CAP_NO_ACCT_WB | | |
301 | BDI_CAP_NO_WRITEBACK)); | |
302 | } | |
1da177e4 | 303 | |
e4ad08fe MS |
304 | static inline bool mapping_cap_writeback_dirty(struct address_space *mapping) |
305 | { | |
306 | return bdi_cap_writeback_dirty(mapping->backing_dev_info); | |
307 | } | |
1da177e4 | 308 | |
e4ad08fe MS |
309 | static inline bool mapping_cap_account_dirty(struct address_space *mapping) |
310 | { | |
311 | return bdi_cap_account_dirty(mapping->backing_dev_info); | |
312 | } | |
1da177e4 | 313 | |
03ba3782 JA |
314 | static inline int bdi_sched_wait(void *word) |
315 | { | |
316 | schedule(); | |
317 | return 0; | |
318 | } | |
319 | ||
1da177e4 | 320 | #endif /* _LINUX_BACKING_DEV_H */ |