blkcg: move blkio_group_conf->iops and ->bps to blk-throttle
[linux-2.6-block.git] / block / blk-cgroup.h
CommitLineData
31e4c28d
VG
1#ifndef _BLK_CGROUP_H
2#define _BLK_CGROUP_H
3/*
4 * Common Block IO controller cgroup interface
5 *
6 * Based on ideas and code from CFQ, CFS and BFQ:
7 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
8 *
9 * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it>
10 * Paolo Valente <paolo.valente@unimore.it>
11 *
12 * Copyright (C) 2009 Vivek Goyal <vgoyal@redhat.com>
13 * Nauman Rafique <nauman@google.com>
14 */
15
16#include <linux/cgroup.h>
575969a0 17#include <linux/u64_stats_sync.h>
829fdb50 18#include <linux/seq_file.h>
31e4c28d 19
062a644d
VG
20enum blkio_policy_id {
21 BLKIO_POLICY_PROP = 0, /* Proportional Bandwidth division */
4c9eefa1 22 BLKIO_POLICY_THROTL, /* Throttling */
035d10b2
TH
23
24 BLKIO_NR_POLICIES,
062a644d
VG
25};
26
9355aede
VG
27/* Max limits for throttle policy */
28#define THROTL_IOPS_MAX UINT_MAX
29
32e380ae 30#ifdef CONFIG_BLK_CGROUP
2f5ea477 31
3381cb8d
TH
32/* CFQ specific, out here for blkcg->cfq_weight */
33#define CFQ_WEIGHT_MIN 10
34#define CFQ_WEIGHT_MAX 1000
35#define CFQ_WEIGHT_DEFAULT 500
36
d3d32e69
TH
37/* cft->private [un]packing for stat printing */
38#define BLKCG_STAT_PRIV(pol, off) (((unsigned)(pol) << 16) | (off))
39#define BLKCG_STAT_POL(prv) ((unsigned)(prv) >> 16)
40#define BLKCG_STAT_OFF(prv) ((unsigned)(prv) & 0xffff)
2aa4a152 41
edcb0722
TH
42enum blkg_rwstat_type {
43 BLKG_RWSTAT_READ,
44 BLKG_RWSTAT_WRITE,
45 BLKG_RWSTAT_SYNC,
46 BLKG_RWSTAT_ASYNC,
47
48 BLKG_RWSTAT_NR,
49 BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR,
303a3acb
DS
50};
51
31e4c28d
VG
52struct blkio_cgroup {
53 struct cgroup_subsys_state css;
31e4c28d
VG
54 spinlock_t lock;
55 struct hlist_head blkg_list;
9a9e8a26
TH
56
57 /* for policies to test whether associated blkcg has changed */
58 uint64_t id;
3381cb8d
TH
59
60 /* TODO: per-policy storage in blkio_cgroup */
61 unsigned int cfq_weight; /* belongs to cfq */
31e4c28d
VG
62};
63
edcb0722
TH
64struct blkg_stat {
65 struct u64_stats_sync syncp;
66 uint64_t cnt;
67};
68
69struct blkg_rwstat {
70 struct u64_stats_sync syncp;
71 uint64_t cnt[BLKG_RWSTAT_NR];
72};
73
0381411e
TH
74/* per-blkg per-policy data */
75struct blkg_policy_data {
76 /* the blkg this per-policy data belongs to */
77 struct blkio_group *blkg;
78
79 /* pol->pdata_size bytes of private data used by policy impl */
80 char pdata[] __aligned(__alignof__(unsigned long long));
81};
82
31e4c28d 83struct blkio_group {
c875f4d0
TH
84 /* Pointer to the associated request_queue */
85 struct request_queue *q;
e8989fae 86 struct list_head q_node;
31e4c28d 87 struct hlist_node blkcg_node;
7ee9c562 88 struct blkio_cgroup *blkcg;
2868ef7b
VG
89 /* Store cgroup path */
90 char path[128];
1adaf3dd
TH
91 /* reference count */
92 int refcnt;
22084190 93
549d3aa8 94 struct blkg_policy_data *pd[BLKIO_NR_POLICIES];
1adaf3dd
TH
95
96 struct rcu_head rcu_head;
31e4c28d
VG
97};
98
0381411e 99typedef void (blkio_init_group_fn)(struct blkio_group *blkg);
9ade5ea4
TH
100typedef void (blkio_exit_group_fn)(struct blkio_group *blkg);
101typedef void (blkio_reset_group_stats_fn)(struct blkio_group *blkg);
3e252066
VG
102
103struct blkio_policy_ops {
0381411e 104 blkio_init_group_fn *blkio_init_group_fn;
9ade5ea4
TH
105 blkio_exit_group_fn *blkio_exit_group_fn;
106 blkio_reset_group_stats_fn *blkio_reset_group_stats_fn;
3e252066
VG
107};
108
109struct blkio_policy_type {
110 struct list_head list;
111 struct blkio_policy_ops ops;
062a644d 112 enum blkio_policy_id plid;
0381411e 113 size_t pdata_size; /* policy specific private data size */
44ea53de 114 struct cftype *cftypes; /* cgroup files for the policy */
3e252066
VG
115};
116
5efd6113
TH
117extern int blkcg_init_queue(struct request_queue *q);
118extern void blkcg_drain_queue(struct request_queue *q);
119extern void blkcg_exit_queue(struct request_queue *q);
120
3e252066
VG
121/* Blkio controller policy registration */
122extern void blkio_policy_register(struct blkio_policy_type *);
123extern void blkio_policy_unregister(struct blkio_policy_type *);
e8989fae
TH
124extern void blkg_destroy_all(struct request_queue *q, bool destroy_root);
125extern void update_root_blkg_pd(struct request_queue *q,
126 enum blkio_policy_id plid);
3e252066 127
829fdb50
TH
128void blkcg_print_blkgs(struct seq_file *sf, struct blkio_cgroup *blkcg,
129 u64 (*prfill)(struct seq_file *, struct blkg_policy_data *, int),
130 int pol, int data, bool show_total);
131u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v);
132u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
133 const struct blkg_rwstat *rwstat);
134int blkcg_print_stat(struct cgroup *cgrp, struct cftype *cft,
135 struct seq_file *sf);
136int blkcg_print_rwstat(struct cgroup *cgrp, struct cftype *cft,
137 struct seq_file *sf);
829fdb50
TH
138
139struct blkg_conf_ctx {
140 struct gendisk *disk;
141 struct blkio_group *blkg;
142 u64 v;
143};
144
145int blkg_conf_prep(struct blkio_cgroup *blkcg, const char *input,
146 struct blkg_conf_ctx *ctx);
147void blkg_conf_finish(struct blkg_conf_ctx *ctx);
148
149
0381411e
TH
150/**
151 * blkg_to_pdata - get policy private data
152 * @blkg: blkg of interest
153 * @pol: policy of interest
154 *
155 * Return pointer to private data associated with the @blkg-@pol pair.
156 */
157static inline void *blkg_to_pdata(struct blkio_group *blkg,
158 struct blkio_policy_type *pol)
159{
549d3aa8 160 return blkg ? blkg->pd[pol->plid]->pdata : NULL;
0381411e
TH
161}
162
163/**
164 * pdata_to_blkg - get blkg associated with policy private data
165 * @pdata: policy private data of interest
0381411e 166 *
aaec55a0 167 * @pdata is policy private data. Determine the blkg it's associated with.
0381411e 168 */
aaec55a0 169static inline struct blkio_group *pdata_to_blkg(void *pdata)
0381411e
TH
170{
171 if (pdata) {
172 struct blkg_policy_data *pd =
173 container_of(pdata, struct blkg_policy_data, pdata);
174 return pd->blkg;
175 }
176 return NULL;
177}
178
afc24d49
VG
179static inline char *blkg_path(struct blkio_group *blkg)
180{
181 return blkg->path;
182}
183
1adaf3dd
TH
184/**
185 * blkg_get - get a blkg reference
186 * @blkg: blkg to get
187 *
188 * The caller should be holding queue_lock and an existing reference.
189 */
190static inline void blkg_get(struct blkio_group *blkg)
191{
192 lockdep_assert_held(blkg->q->queue_lock);
193 WARN_ON_ONCE(!blkg->refcnt);
194 blkg->refcnt++;
195}
196
197void __blkg_release(struct blkio_group *blkg);
198
199/**
200 * blkg_put - put a blkg reference
201 * @blkg: blkg to put
202 *
203 * The caller should be holding queue_lock.
204 */
205static inline void blkg_put(struct blkio_group *blkg)
206{
207 lockdep_assert_held(blkg->q->queue_lock);
208 WARN_ON_ONCE(blkg->refcnt <= 0);
209 if (!--blkg->refcnt)
210 __blkg_release(blkg);
211}
212
edcb0722
TH
213/**
214 * blkg_stat_add - add a value to a blkg_stat
215 * @stat: target blkg_stat
216 * @val: value to add
217 *
218 * Add @val to @stat. The caller is responsible for synchronizing calls to
219 * this function.
220 */
221static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val)
222{
223 u64_stats_update_begin(&stat->syncp);
224 stat->cnt += val;
225 u64_stats_update_end(&stat->syncp);
226}
227
228/**
229 * blkg_stat_read - read the current value of a blkg_stat
230 * @stat: blkg_stat to read
231 *
232 * Read the current value of @stat. This function can be called without
233 * synchroniztion and takes care of u64 atomicity.
234 */
235static inline uint64_t blkg_stat_read(struct blkg_stat *stat)
236{
237 unsigned int start;
238 uint64_t v;
239
240 do {
241 start = u64_stats_fetch_begin(&stat->syncp);
242 v = stat->cnt;
243 } while (u64_stats_fetch_retry(&stat->syncp, start));
244
245 return v;
246}
247
248/**
249 * blkg_stat_reset - reset a blkg_stat
250 * @stat: blkg_stat to reset
251 */
252static inline void blkg_stat_reset(struct blkg_stat *stat)
253{
254 stat->cnt = 0;
255}
256
257/**
258 * blkg_rwstat_add - add a value to a blkg_rwstat
259 * @rwstat: target blkg_rwstat
260 * @rw: mask of REQ_{WRITE|SYNC}
261 * @val: value to add
262 *
263 * Add @val to @rwstat. The counters are chosen according to @rw. The
264 * caller is responsible for synchronizing calls to this function.
265 */
266static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat,
267 int rw, uint64_t val)
268{
269 u64_stats_update_begin(&rwstat->syncp);
270
271 if (rw & REQ_WRITE)
272 rwstat->cnt[BLKG_RWSTAT_WRITE] += val;
273 else
274 rwstat->cnt[BLKG_RWSTAT_READ] += val;
275 if (rw & REQ_SYNC)
276 rwstat->cnt[BLKG_RWSTAT_SYNC] += val;
277 else
278 rwstat->cnt[BLKG_RWSTAT_ASYNC] += val;
279
280 u64_stats_update_end(&rwstat->syncp);
281}
282
283/**
284 * blkg_rwstat_read - read the current values of a blkg_rwstat
285 * @rwstat: blkg_rwstat to read
286 *
287 * Read the current snapshot of @rwstat and return it as the return value.
288 * This function can be called without synchronization and takes care of
289 * u64 atomicity.
290 */
291static struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat)
292{
293 unsigned int start;
294 struct blkg_rwstat tmp;
295
296 do {
297 start = u64_stats_fetch_begin(&rwstat->syncp);
298 tmp = *rwstat;
299 } while (u64_stats_fetch_retry(&rwstat->syncp, start));
300
301 return tmp;
302}
303
304/**
305 * blkg_rwstat_sum - read the total count of a blkg_rwstat
306 * @rwstat: blkg_rwstat to read
307 *
308 * Return the total count of @rwstat regardless of the IO direction. This
309 * function can be called without synchronization and takes care of u64
310 * atomicity.
311 */
312static inline uint64_t blkg_rwstat_sum(struct blkg_rwstat *rwstat)
313{
314 struct blkg_rwstat tmp = blkg_rwstat_read(rwstat);
315
316 return tmp.cnt[BLKG_RWSTAT_READ] + tmp.cnt[BLKG_RWSTAT_WRITE];
317}
318
319/**
320 * blkg_rwstat_reset - reset a blkg_rwstat
321 * @rwstat: blkg_rwstat to reset
322 */
323static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat)
324{
325 memset(rwstat->cnt, 0, sizeof(rwstat->cnt));
326}
327
2f5ea477
JA
328#else
329
330struct blkio_group {
331};
332
3e252066
VG
333struct blkio_policy_type {
334};
335
5efd6113
TH
336static inline int blkcg_init_queue(struct request_queue *q) { return 0; }
337static inline void blkcg_drain_queue(struct request_queue *q) { }
338static inline void blkcg_exit_queue(struct request_queue *q) { }
3e252066
VG
339static inline void blkio_policy_register(struct blkio_policy_type *blkiop) { }
340static inline void blkio_policy_unregister(struct blkio_policy_type *blkiop) { }
03aa264a 341static inline void blkg_destroy_all(struct request_queue *q,
03aa264a 342 bool destory_root) { }
e8989fae
TH
343static inline void update_root_blkg_pd(struct request_queue *q,
344 enum blkio_policy_id plid) { }
3e252066 345
0381411e
TH
346static inline void *blkg_to_pdata(struct blkio_group *blkg,
347 struct blkio_policy_type *pol) { return NULL; }
348static inline struct blkio_group *pdata_to_blkg(void *pdata,
349 struct blkio_policy_type *pol) { return NULL; }
afc24d49 350static inline char *blkg_path(struct blkio_group *blkg) { return NULL; }
1adaf3dd
TH
351static inline void blkg_get(struct blkio_group *blkg) { }
352static inline void blkg_put(struct blkio_group *blkg) { }
afc24d49 353
2f5ea477
JA
354#endif
355
32e380ae 356#ifdef CONFIG_BLK_CGROUP
31e4c28d
VG
357extern struct blkio_cgroup blkio_root_cgroup;
358extern struct blkio_cgroup *cgroup_to_blkio_cgroup(struct cgroup *cgroup);
4f85cb96 359extern struct blkio_cgroup *bio_blkio_cgroup(struct bio *bio);
cd1604fa 360extern struct blkio_group *blkg_lookup(struct blkio_cgroup *blkcg,
e8989fae 361 struct request_queue *q);
cd1604fa
TH
362struct blkio_group *blkg_lookup_create(struct blkio_cgroup *blkcg,
363 struct request_queue *q,
cd1604fa 364 bool for_root);
31e4c28d 365#else
2f5ea477 366struct cgroup;
31e4c28d
VG
367static inline struct blkio_cgroup *
368cgroup_to_blkio_cgroup(struct cgroup *cgroup) { return NULL; }
70087dc3 369static inline struct blkio_cgroup *
4f85cb96 370bio_blkio_cgroup(struct bio *bio) { return NULL; }
31e4c28d 371
cd1604fa
TH
372static inline struct blkio_group *blkg_lookup(struct blkio_cgroup *blkcg,
373 void *key) { return NULL; }
31e4c28d
VG
374#endif
375#endif /* _BLK_CGROUP_H */