]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - block/blk-cgroup.h
block: prepare for multiple request_lists
[mirror_ubuntu-zesty-kernel.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>
a637120e 19#include <linux/radix-tree.h>
31e4c28d 20
9355aede
VG
21/* Max limits for throttle policy */
22#define THROTL_IOPS_MAX UINT_MAX
23
3381cb8d
TH
24/* CFQ specific, out here for blkcg->cfq_weight */
25#define CFQ_WEIGHT_MIN 10
26#define CFQ_WEIGHT_MAX 1000
27#define CFQ_WEIGHT_DEFAULT 500
28
f48ec1d7
TH
29#ifdef CONFIG_BLK_CGROUP
30
edcb0722
TH
31enum blkg_rwstat_type {
32 BLKG_RWSTAT_READ,
33 BLKG_RWSTAT_WRITE,
34 BLKG_RWSTAT_SYNC,
35 BLKG_RWSTAT_ASYNC,
36
37 BLKG_RWSTAT_NR,
38 BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR,
303a3acb
DS
39};
40
a637120e
TH
41struct blkcg_gq;
42
3c798398 43struct blkcg {
36558c8a
TH
44 struct cgroup_subsys_state css;
45 spinlock_t lock;
a637120e
TH
46
47 struct radix_tree_root blkg_tree;
48 struct blkcg_gq *blkg_hint;
36558c8a 49 struct hlist_head blkg_list;
9a9e8a26
TH
50
51 /* for policies to test whether associated blkcg has changed */
36558c8a 52 uint64_t id;
3381cb8d 53
3c798398 54 /* TODO: per-policy storage in blkcg */
36558c8a 55 unsigned int cfq_weight; /* belongs to cfq */
31e4c28d
VG
56};
57
edcb0722
TH
58struct blkg_stat {
59 struct u64_stats_sync syncp;
60 uint64_t cnt;
61};
62
63struct blkg_rwstat {
64 struct u64_stats_sync syncp;
65 uint64_t cnt[BLKG_RWSTAT_NR];
66};
67
f95a04af
TH
68/*
69 * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a
70 * request_queue (q). This is used by blkcg policies which need to track
71 * information per blkcg - q pair.
72 *
73 * There can be multiple active blkcg policies and each has its private
74 * data on each blkg, the size of which is determined by
75 * blkcg_policy->pd_size. blkcg core allocates and frees such areas
76 * together with blkg and invokes pd_init/exit_fn() methods.
77 *
78 * Such private data must embed struct blkg_policy_data (pd) at the
79 * beginning and pd_size can't be smaller than pd.
80 */
0381411e
TH
81struct blkg_policy_data {
82 /* the blkg this per-policy data belongs to */
3c798398 83 struct blkcg_gq *blkg;
0381411e 84
a2b1693b 85 /* used during policy activation */
36558c8a 86 struct list_head alloc_node;
0381411e
TH
87};
88
3c798398
TH
89/* association between a blk cgroup and a request queue */
90struct blkcg_gq {
c875f4d0 91 /* Pointer to the associated request_queue */
36558c8a
TH
92 struct request_queue *q;
93 struct list_head q_node;
94 struct hlist_node blkcg_node;
3c798398 95 struct blkcg *blkcg;
1adaf3dd 96 /* reference count */
36558c8a 97 int refcnt;
22084190 98
36558c8a 99 struct blkg_policy_data *pd[BLKCG_MAX_POLS];
1adaf3dd 100
36558c8a 101 struct rcu_head rcu_head;
31e4c28d
VG
102};
103
3c798398
TH
104typedef void (blkcg_pol_init_pd_fn)(struct blkcg_gq *blkg);
105typedef void (blkcg_pol_exit_pd_fn)(struct blkcg_gq *blkg);
106typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkcg_gq *blkg);
3e252066 107
3c798398 108struct blkcg_policy {
36558c8a
TH
109 int plid;
110 /* policy specific private data size */
f95a04af 111 size_t pd_size;
36558c8a
TH
112 /* cgroup files for the policy */
113 struct cftype *cftypes;
f9fcc2d3
TH
114
115 /* operations */
116 blkcg_pol_init_pd_fn *pd_init_fn;
117 blkcg_pol_exit_pd_fn *pd_exit_fn;
118 blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn;
3e252066
VG
119};
120
3c798398 121extern struct blkcg blkcg_root;
36558c8a 122
3c798398
TH
123struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, struct request_queue *q);
124struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
125 struct request_queue *q);
36558c8a
TH
126int blkcg_init_queue(struct request_queue *q);
127void blkcg_drain_queue(struct request_queue *q);
128void blkcg_exit_queue(struct request_queue *q);
5efd6113 129
3e252066 130/* Blkio controller policy registration */
3c798398
TH
131int blkcg_policy_register(struct blkcg_policy *pol);
132void blkcg_policy_unregister(struct blkcg_policy *pol);
36558c8a 133int blkcg_activate_policy(struct request_queue *q,
3c798398 134 const struct blkcg_policy *pol);
36558c8a 135void blkcg_deactivate_policy(struct request_queue *q,
3c798398 136 const struct blkcg_policy *pol);
3e252066 137
3c798398 138void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
f95a04af
TH
139 u64 (*prfill)(struct seq_file *,
140 struct blkg_policy_data *, int),
3c798398 141 const struct blkcg_policy *pol, int data,
ec399347 142 bool show_total);
f95a04af
TH
143u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v);
144u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
829fdb50 145 const struct blkg_rwstat *rwstat);
f95a04af
TH
146u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off);
147u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
148 int off);
829fdb50
TH
149
150struct blkg_conf_ctx {
36558c8a 151 struct gendisk *disk;
3c798398 152 struct blkcg_gq *blkg;
36558c8a 153 u64 v;
829fdb50
TH
154};
155
3c798398
TH
156int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
157 const char *input, struct blkg_conf_ctx *ctx);
829fdb50
TH
158void blkg_conf_finish(struct blkg_conf_ctx *ctx);
159
160
b1208b56
TH
161static inline struct blkcg *cgroup_to_blkcg(struct cgroup *cgroup)
162{
163 return container_of(cgroup_subsys_state(cgroup, blkio_subsys_id),
164 struct blkcg, css);
165}
166
167static inline struct blkcg *task_blkcg(struct task_struct *tsk)
168{
169 return container_of(task_subsys_state(tsk, blkio_subsys_id),
170 struct blkcg, css);
171}
172
173static inline struct blkcg *bio_blkcg(struct bio *bio)
174{
175 if (bio && bio->bi_css)
176 return container_of(bio->bi_css, struct blkcg, css);
177 return task_blkcg(current);
178}
179
0381411e
TH
180/**
181 * blkg_to_pdata - get policy private data
182 * @blkg: blkg of interest
183 * @pol: policy of interest
184 *
185 * Return pointer to private data associated with the @blkg-@pol pair.
186 */
f95a04af
TH
187static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
188 struct blkcg_policy *pol)
0381411e 189{
f95a04af 190 return blkg ? blkg->pd[pol->plid] : NULL;
0381411e
TH
191}
192
193/**
194 * pdata_to_blkg - get blkg associated with policy private data
f95a04af 195 * @pd: policy private data of interest
0381411e 196 *
f95a04af 197 * @pd is policy private data. Determine the blkg it's associated with.
0381411e 198 */
f95a04af 199static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd)
0381411e 200{
f95a04af 201 return pd ? pd->blkg : NULL;
0381411e
TH
202}
203
54e7ed12
TH
204/**
205 * blkg_path - format cgroup path of blkg
206 * @blkg: blkg of interest
207 * @buf: target buffer
208 * @buflen: target buffer length
209 *
210 * Format the path of the cgroup of @blkg into @buf.
211 */
3c798398 212static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen)
afc24d49 213{
54e7ed12
TH
214 int ret;
215
216 rcu_read_lock();
217 ret = cgroup_path(blkg->blkcg->css.cgroup, buf, buflen);
218 rcu_read_unlock();
219 if (ret)
220 strncpy(buf, "<unavailable>", buflen);
221 return ret;
afc24d49
VG
222}
223
1adaf3dd
TH
224/**
225 * blkg_get - get a blkg reference
226 * @blkg: blkg to get
227 *
228 * The caller should be holding queue_lock and an existing reference.
229 */
3c798398 230static inline void blkg_get(struct blkcg_gq *blkg)
1adaf3dd
TH
231{
232 lockdep_assert_held(blkg->q->queue_lock);
233 WARN_ON_ONCE(!blkg->refcnt);
234 blkg->refcnt++;
235}
236
3c798398 237void __blkg_release(struct blkcg_gq *blkg);
1adaf3dd
TH
238
239/**
240 * blkg_put - put a blkg reference
241 * @blkg: blkg to put
242 *
243 * The caller should be holding queue_lock.
244 */
3c798398 245static inline void blkg_put(struct blkcg_gq *blkg)
1adaf3dd
TH
246{
247 lockdep_assert_held(blkg->q->queue_lock);
248 WARN_ON_ONCE(blkg->refcnt <= 0);
249 if (!--blkg->refcnt)
250 __blkg_release(blkg);
251}
252
edcb0722
TH
253/**
254 * blkg_stat_add - add a value to a blkg_stat
255 * @stat: target blkg_stat
256 * @val: value to add
257 *
258 * Add @val to @stat. The caller is responsible for synchronizing calls to
259 * this function.
260 */
261static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val)
262{
263 u64_stats_update_begin(&stat->syncp);
264 stat->cnt += val;
265 u64_stats_update_end(&stat->syncp);
266}
267
268/**
269 * blkg_stat_read - read the current value of a blkg_stat
270 * @stat: blkg_stat to read
271 *
272 * Read the current value of @stat. This function can be called without
273 * synchroniztion and takes care of u64 atomicity.
274 */
275static inline uint64_t blkg_stat_read(struct blkg_stat *stat)
276{
277 unsigned int start;
278 uint64_t v;
279
280 do {
281 start = u64_stats_fetch_begin(&stat->syncp);
282 v = stat->cnt;
283 } while (u64_stats_fetch_retry(&stat->syncp, start));
284
285 return v;
286}
287
288/**
289 * blkg_stat_reset - reset a blkg_stat
290 * @stat: blkg_stat to reset
291 */
292static inline void blkg_stat_reset(struct blkg_stat *stat)
293{
294 stat->cnt = 0;
295}
296
297/**
298 * blkg_rwstat_add - add a value to a blkg_rwstat
299 * @rwstat: target blkg_rwstat
300 * @rw: mask of REQ_{WRITE|SYNC}
301 * @val: value to add
302 *
303 * Add @val to @rwstat. The counters are chosen according to @rw. The
304 * caller is responsible for synchronizing calls to this function.
305 */
306static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat,
307 int rw, uint64_t val)
308{
309 u64_stats_update_begin(&rwstat->syncp);
310
311 if (rw & REQ_WRITE)
312 rwstat->cnt[BLKG_RWSTAT_WRITE] += val;
313 else
314 rwstat->cnt[BLKG_RWSTAT_READ] += val;
315 if (rw & REQ_SYNC)
316 rwstat->cnt[BLKG_RWSTAT_SYNC] += val;
317 else
318 rwstat->cnt[BLKG_RWSTAT_ASYNC] += val;
319
320 u64_stats_update_end(&rwstat->syncp);
321}
322
323/**
324 * blkg_rwstat_read - read the current values of a blkg_rwstat
325 * @rwstat: blkg_rwstat to read
326 *
327 * Read the current snapshot of @rwstat and return it as the return value.
328 * This function can be called without synchronization and takes care of
329 * u64 atomicity.
330 */
c94bed89 331static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat)
edcb0722
TH
332{
333 unsigned int start;
334 struct blkg_rwstat tmp;
335
336 do {
337 start = u64_stats_fetch_begin(&rwstat->syncp);
338 tmp = *rwstat;
339 } while (u64_stats_fetch_retry(&rwstat->syncp, start));
340
341 return tmp;
342}
343
344/**
345 * blkg_rwstat_sum - read the total count of a blkg_rwstat
346 * @rwstat: blkg_rwstat to read
347 *
348 * Return the total count of @rwstat regardless of the IO direction. This
349 * function can be called without synchronization and takes care of u64
350 * atomicity.
351 */
352static inline uint64_t blkg_rwstat_sum(struct blkg_rwstat *rwstat)
353{
354 struct blkg_rwstat tmp = blkg_rwstat_read(rwstat);
355
356 return tmp.cnt[BLKG_RWSTAT_READ] + tmp.cnt[BLKG_RWSTAT_WRITE];
357}
358
359/**
360 * blkg_rwstat_reset - reset a blkg_rwstat
361 * @rwstat: blkg_rwstat to reset
362 */
363static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat)
364{
365 memset(rwstat->cnt, 0, sizeof(rwstat->cnt));
366}
367
36558c8a
TH
368#else /* CONFIG_BLK_CGROUP */
369
370struct cgroup;
b1208b56 371struct blkcg;
2f5ea477 372
f95a04af
TH
373struct blkg_policy_data {
374};
375
3c798398 376struct blkcg_gq {
2f5ea477
JA
377};
378
3c798398 379struct blkcg_policy {
3e252066
VG
380};
381
3c798398 382static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; }
5efd6113
TH
383static inline int blkcg_init_queue(struct request_queue *q) { return 0; }
384static inline void blkcg_drain_queue(struct request_queue *q) { }
385static inline void blkcg_exit_queue(struct request_queue *q) { }
3c798398
TH
386static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; }
387static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { }
a2b1693b 388static inline int blkcg_activate_policy(struct request_queue *q,
3c798398 389 const struct blkcg_policy *pol) { return 0; }
a2b1693b 390static inline void blkcg_deactivate_policy(struct request_queue *q,
3c798398
TH
391 const struct blkcg_policy *pol) { }
392
b1208b56
TH
393static inline struct blkcg *cgroup_to_blkcg(struct cgroup *cgroup) { return NULL; }
394static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; }
f95a04af
TH
395static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
396 struct blkcg_policy *pol) { return NULL; }
397static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; }
3c798398
TH
398static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; }
399static inline void blkg_get(struct blkcg_gq *blkg) { }
400static inline void blkg_put(struct blkcg_gq *blkg) { }
afc24d49 401
36558c8a
TH
402#endif /* CONFIG_BLK_CGROUP */
403#endif /* _BLK_CGROUP_H */