4 * Common Block IO controller cgroup interface
6 * Based on ideas and code from CFQ, CFS and BFQ:
7 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
9 * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it>
10 * Paolo Valente <paolo.valente@unimore.it>
12 * Copyright (C) 2009 Vivek Goyal <vgoyal@redhat.com>
13 * Nauman Rafique <nauman@google.com>
16 #include <linux/cgroup.h>
17 #include <linux/u64_stats_sync.h>
18 #include <linux/seq_file.h>
19 #include <linux/radix-tree.h>
20 #include <linux/blkdev.h>
22 /* Max limits for throttle policy */
23 #define THROTL_IOPS_MAX UINT_MAX
25 /* CFQ specific, out here for blkcg->cfq_weight */
26 #define CFQ_WEIGHT_MIN 10
27 #define CFQ_WEIGHT_MAX 1000
28 #define CFQ_WEIGHT_DEFAULT 500
30 #ifdef CONFIG_BLK_CGROUP
32 enum blkg_rwstat_type {
39 BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR,
45 struct cgroup_subsys_state css;
48 struct radix_tree_root blkg_tree;
49 struct blkcg_gq *blkg_hint;
50 struct hlist_head blkg_list;
52 /* for policies to test whether associated blkcg has changed */
55 /* TODO: per-policy storage in blkcg */
56 unsigned int cfq_weight; /* belongs to cfq */
60 struct u64_stats_sync syncp;
65 struct u64_stats_sync syncp;
66 uint64_t cnt[BLKG_RWSTAT_NR];
70 * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a
71 * request_queue (q). This is used by blkcg policies which need to track
72 * information per blkcg - q pair.
74 * There can be multiple active blkcg policies and each has its private
75 * data on each blkg, the size of which is determined by
76 * blkcg_policy->pd_size. blkcg core allocates and frees such areas
77 * together with blkg and invokes pd_init/exit_fn() methods.
79 * Such private data must embed struct blkg_policy_data (pd) at the
80 * beginning and pd_size can't be smaller than pd.
82 struct blkg_policy_data {
83 /* the blkg this per-policy data belongs to */
84 struct blkcg_gq *blkg;
86 /* used during policy activation */
87 struct list_head alloc_node;
90 /* association between a blk cgroup and a request queue */
92 /* Pointer to the associated request_queue */
93 struct request_queue *q;
94 struct list_head q_node;
95 struct hlist_node blkcg_node;
98 /* all non-root blkcg_gq's are guaranteed to have access to parent */
99 struct blkcg_gq *parent;
101 /* request allocation list for this blkcg-q pair */
102 struct request_list rl;
104 /* reference count */
107 struct blkg_policy_data *pd[BLKCG_MAX_POLS];
109 struct rcu_head rcu_head;
112 typedef void (blkcg_pol_init_pd_fn)(struct blkcg_gq *blkg);
113 typedef void (blkcg_pol_exit_pd_fn)(struct blkcg_gq *blkg);
114 typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkcg_gq *blkg);
116 struct blkcg_policy {
118 /* policy specific private data size */
120 /* cgroup files for the policy */
121 struct cftype *cftypes;
124 blkcg_pol_init_pd_fn *pd_init_fn;
125 blkcg_pol_exit_pd_fn *pd_exit_fn;
126 blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn;
129 extern struct blkcg blkcg_root;
131 struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, struct request_queue *q);
132 struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
133 struct request_queue *q);
134 int blkcg_init_queue(struct request_queue *q);
135 void blkcg_drain_queue(struct request_queue *q);
136 void blkcg_exit_queue(struct request_queue *q);
138 /* Blkio controller policy registration */
139 int blkcg_policy_register(struct blkcg_policy *pol);
140 void blkcg_policy_unregister(struct blkcg_policy *pol);
141 int blkcg_activate_policy(struct request_queue *q,
142 const struct blkcg_policy *pol);
143 void blkcg_deactivate_policy(struct request_queue *q,
144 const struct blkcg_policy *pol);
146 void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
147 u64 (*prfill)(struct seq_file *,
148 struct blkg_policy_data *, int),
149 const struct blkcg_policy *pol, int data,
151 u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v);
152 u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
153 const struct blkg_rwstat *rwstat);
154 u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off);
155 u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
158 struct blkg_conf_ctx {
159 struct gendisk *disk;
160 struct blkcg_gq *blkg;
164 int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
165 const char *input, struct blkg_conf_ctx *ctx);
166 void blkg_conf_finish(struct blkg_conf_ctx *ctx);
169 static inline struct blkcg *cgroup_to_blkcg(struct cgroup *cgroup)
171 return container_of(cgroup_subsys_state(cgroup, blkio_subsys_id),
175 static inline struct blkcg *task_blkcg(struct task_struct *tsk)
177 return container_of(task_subsys_state(tsk, blkio_subsys_id),
181 static inline struct blkcg *bio_blkcg(struct bio *bio)
183 if (bio && bio->bi_css)
184 return container_of(bio->bi_css, struct blkcg, css);
185 return task_blkcg(current);
189 * blkcg_parent - get the parent of a blkcg
190 * @blkcg: blkcg of interest
192 * Return the parent blkcg of @blkcg. Can be called anytime.
194 static inline struct blkcg *blkcg_parent(struct blkcg *blkcg)
196 struct cgroup *pcg = blkcg->css.cgroup->parent;
198 return pcg ? cgroup_to_blkcg(pcg) : NULL;
202 * blkg_to_pdata - get policy private data
203 * @blkg: blkg of interest
204 * @pol: policy of interest
206 * Return pointer to private data associated with the @blkg-@pol pair.
208 static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
209 struct blkcg_policy *pol)
211 return blkg ? blkg->pd[pol->plid] : NULL;
215 * pdata_to_blkg - get blkg associated with policy private data
216 * @pd: policy private data of interest
218 * @pd is policy private data. Determine the blkg it's associated with.
220 static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd)
222 return pd ? pd->blkg : NULL;
226 * blkg_path - format cgroup path of blkg
227 * @blkg: blkg of interest
228 * @buf: target buffer
229 * @buflen: target buffer length
231 * Format the path of the cgroup of @blkg into @buf.
233 static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen)
238 ret = cgroup_path(blkg->blkcg->css.cgroup, buf, buflen);
241 strncpy(buf, "<unavailable>", buflen);
246 * blkg_get - get a blkg reference
249 * The caller should be holding queue_lock and an existing reference.
251 static inline void blkg_get(struct blkcg_gq *blkg)
253 lockdep_assert_held(blkg->q->queue_lock);
254 WARN_ON_ONCE(!blkg->refcnt);
258 void __blkg_release(struct blkcg_gq *blkg);
261 * blkg_put - put a blkg reference
264 * The caller should be holding queue_lock.
266 static inline void blkg_put(struct blkcg_gq *blkg)
268 lockdep_assert_held(blkg->q->queue_lock);
269 WARN_ON_ONCE(blkg->refcnt <= 0);
271 __blkg_release(blkg);
275 * blk_get_rl - get request_list to use
276 * @q: request_queue of interest
277 * @bio: bio which will be attached to the allocated request (may be %NULL)
279 * The caller wants to allocate a request from @q to use for @bio. Find
280 * the request_list to use and obtain a reference on it. Should be called
281 * under queue_lock. This function is guaranteed to return non-%NULL
284 static inline struct request_list *blk_get_rl(struct request_queue *q,
288 struct blkcg_gq *blkg;
292 blkcg = bio_blkcg(bio);
294 /* bypass blkg lookup and use @q->root_rl directly for root */
295 if (blkcg == &blkcg_root)
299 * Try to use blkg->rl. blkg lookup may fail under memory pressure
300 * or if either the blkcg or queue is going away. Fall back to
301 * root_rl in such cases.
303 blkg = blkg_lookup_create(blkcg, q);
304 if (unlikely(IS_ERR(blkg)))
316 * blk_put_rl - put request_list
317 * @rl: request_list to put
319 * Put the reference acquired by blk_get_rl(). Should be called under
322 static inline void blk_put_rl(struct request_list *rl)
324 /* root_rl may not have blkg set */
325 if (rl->blkg && rl->blkg->blkcg != &blkcg_root)
330 * blk_rq_set_rl - associate a request with a request_list
331 * @rq: request of interest
332 * @rl: target request_list
334 * Associate @rq with @rl so that accounting and freeing can know the
335 * request_list @rq came from.
337 static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl)
343 * blk_rq_rl - return the request_list a request came from
344 * @rq: request of interest
346 * Return the request_list @rq is allocated from.
348 static inline struct request_list *blk_rq_rl(struct request *rq)
353 struct request_list *__blk_queue_next_rl(struct request_list *rl,
354 struct request_queue *q);
356 * blk_queue_for_each_rl - iterate through all request_lists of a request_queue
358 * Should be used under queue_lock.
360 #define blk_queue_for_each_rl(rl, q) \
361 for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q)))
364 * blkg_stat_add - add a value to a blkg_stat
365 * @stat: target blkg_stat
368 * Add @val to @stat. The caller is responsible for synchronizing calls to
371 static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val)
373 u64_stats_update_begin(&stat->syncp);
375 u64_stats_update_end(&stat->syncp);
379 * blkg_stat_read - read the current value of a blkg_stat
380 * @stat: blkg_stat to read
382 * Read the current value of @stat. This function can be called without
383 * synchroniztion and takes care of u64 atomicity.
385 static inline uint64_t blkg_stat_read(struct blkg_stat *stat)
391 start = u64_stats_fetch_begin(&stat->syncp);
393 } while (u64_stats_fetch_retry(&stat->syncp, start));
399 * blkg_stat_reset - reset a blkg_stat
400 * @stat: blkg_stat to reset
402 static inline void blkg_stat_reset(struct blkg_stat *stat)
408 * blkg_rwstat_add - add a value to a blkg_rwstat
409 * @rwstat: target blkg_rwstat
410 * @rw: mask of REQ_{WRITE|SYNC}
413 * Add @val to @rwstat. The counters are chosen according to @rw. The
414 * caller is responsible for synchronizing calls to this function.
416 static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat,
417 int rw, uint64_t val)
419 u64_stats_update_begin(&rwstat->syncp);
422 rwstat->cnt[BLKG_RWSTAT_WRITE] += val;
424 rwstat->cnt[BLKG_RWSTAT_READ] += val;
426 rwstat->cnt[BLKG_RWSTAT_SYNC] += val;
428 rwstat->cnt[BLKG_RWSTAT_ASYNC] += val;
430 u64_stats_update_end(&rwstat->syncp);
434 * blkg_rwstat_read - read the current values of a blkg_rwstat
435 * @rwstat: blkg_rwstat to read
437 * Read the current snapshot of @rwstat and return it as the return value.
438 * This function can be called without synchronization and takes care of
441 static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat)
444 struct blkg_rwstat tmp;
447 start = u64_stats_fetch_begin(&rwstat->syncp);
449 } while (u64_stats_fetch_retry(&rwstat->syncp, start));
455 * blkg_rwstat_sum - read the total count of a blkg_rwstat
456 * @rwstat: blkg_rwstat to read
458 * Return the total count of @rwstat regardless of the IO direction. This
459 * function can be called without synchronization and takes care of u64
462 static inline uint64_t blkg_rwstat_sum(struct blkg_rwstat *rwstat)
464 struct blkg_rwstat tmp = blkg_rwstat_read(rwstat);
466 return tmp.cnt[BLKG_RWSTAT_READ] + tmp.cnt[BLKG_RWSTAT_WRITE];
470 * blkg_rwstat_reset - reset a blkg_rwstat
471 * @rwstat: blkg_rwstat to reset
473 static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat)
475 memset(rwstat->cnt, 0, sizeof(rwstat->cnt));
478 #else /* CONFIG_BLK_CGROUP */
483 struct blkg_policy_data {
489 struct blkcg_policy {
492 static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; }
493 static inline int blkcg_init_queue(struct request_queue *q) { return 0; }
494 static inline void blkcg_drain_queue(struct request_queue *q) { }
495 static inline void blkcg_exit_queue(struct request_queue *q) { }
496 static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; }
497 static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { }
498 static inline int blkcg_activate_policy(struct request_queue *q,
499 const struct blkcg_policy *pol) { return 0; }
500 static inline void blkcg_deactivate_policy(struct request_queue *q,
501 const struct blkcg_policy *pol) { }
503 static inline struct blkcg *cgroup_to_blkcg(struct cgroup *cgroup) { return NULL; }
504 static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; }
506 static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
507 struct blkcg_policy *pol) { return NULL; }
508 static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; }
509 static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; }
510 static inline void blkg_get(struct blkcg_gq *blkg) { }
511 static inline void blkg_put(struct blkcg_gq *blkg) { }
513 static inline struct request_list *blk_get_rl(struct request_queue *q,
514 struct bio *bio) { return &q->root_rl; }
515 static inline void blk_put_rl(struct request_list *rl) { }
516 static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { }
517 static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; }
519 #define blk_queue_for_each_rl(rl, q) \
520 for ((rl) = &(q)->root_rl; (rl); (rl) = NULL)
522 #endif /* CONFIG_BLK_CGROUP */
523 #endif /* _BLK_CGROUP_H */