1 /* SPDX-License-Identifier: GPL-2.0 */
5 * Common Block IO controller cgroup interface
7 * Based on ideas and code from CFQ, CFS and BFQ:
8 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
10 * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it>
11 * Paolo Valente <paolo.valente@unimore.it>
13 * Copyright (C) 2009 Vivek Goyal <vgoyal@redhat.com>
14 * Nauman Rafique <nauman@google.com>
17 #include <linux/cgroup.h>
18 #include <linux/percpu_counter.h>
19 #include <linux/seq_file.h>
20 #include <linux/radix-tree.h>
21 #include <linux/blkdev.h>
22 #include <linux/atomic.h>
23 #include <linux/kthread.h>
25 /* percpu_counter batch for blkg_[rw]stats, per-cpu drift doesn't matter */
26 #define BLKG_STAT_CPU_BATCH (INT_MAX / 2)
28 /* Max limits for throttle policy */
29 #define THROTL_IOPS_MAX UINT_MAX
31 #ifdef CONFIG_BLK_CGROUP
33 enum blkg_rwstat_type {
41 BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR,
47 struct cgroup_subsys_state css;
50 struct radix_tree_root blkg_tree;
51 struct blkcg_gq __rcu *blkg_hint;
52 struct hlist_head blkg_list;
54 struct blkcg_policy_data *cpd[BLKCG_MAX_POLS];
56 struct list_head all_blkcgs_node;
57 #ifdef CONFIG_CGROUP_WRITEBACK
58 struct list_head cgwb_list;
63 * blkg_[rw]stat->aux_cnt is excluded for local stats but included for
64 * recursive. Used to carry stats of dead children, and, for blkg_rwstat,
65 * to carry result values from read and sum operations.
68 struct percpu_counter cpu_cnt;
73 struct percpu_counter cpu_cnt[BLKG_RWSTAT_NR];
74 atomic64_t aux_cnt[BLKG_RWSTAT_NR];
78 * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a
79 * request_queue (q). This is used by blkcg policies which need to track
80 * information per blkcg - q pair.
82 * There can be multiple active blkcg policies and each blkg:policy pair is
83 * represented by a blkg_policy_data which is allocated and freed by each
84 * policy's pd_alloc/free_fn() methods. A policy can allocate private data
85 * area by allocating larger data structure which embeds blkg_policy_data
88 struct blkg_policy_data {
89 /* the blkg and policy id this per-policy data belongs to */
90 struct blkcg_gq *blkg;
95 * Policies that need to keep per-blkcg data which is independent from any
96 * request_queue associated to it should implement cpd_alloc/free_fn()
97 * methods. A policy can allocate private data area by allocating larger
98 * data structure which embeds blkcg_policy_data at the beginning.
99 * cpd_init() is invoked to let each policy handle per-blkcg data.
101 struct blkcg_policy_data {
102 /* the blkcg and policy id this per-policy data belongs to */
107 /* association between a blk cgroup and a request queue */
109 /* Pointer to the associated request_queue */
110 struct request_queue *q;
111 struct list_head q_node;
112 struct hlist_node blkcg_node;
116 * Each blkg gets congested separately and the congestion state is
117 * propagated to the matching bdi_writeback_congested.
119 struct bdi_writeback_congested *wb_congested;
121 /* all non-root blkcg_gq's are guaranteed to have access to parent */
122 struct blkcg_gq *parent;
124 /* request allocation list for this blkcg-q pair */
125 struct request_list rl;
127 /* reference count */
130 /* is this blkg online? protected by both blkcg and q locks */
133 struct blkg_rwstat stat_bytes;
134 struct blkg_rwstat stat_ios;
136 struct blkg_policy_data *pd[BLKCG_MAX_POLS];
138 struct rcu_head rcu_head;
141 atomic64_t delay_nsec;
142 atomic64_t delay_start;
147 typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp);
148 typedef void (blkcg_pol_init_cpd_fn)(struct blkcg_policy_data *cpd);
149 typedef void (blkcg_pol_free_cpd_fn)(struct blkcg_policy_data *cpd);
150 typedef void (blkcg_pol_bind_cpd_fn)(struct blkcg_policy_data *cpd);
151 typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node);
152 typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd);
153 typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd);
154 typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd);
155 typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd);
156 typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd);
157 typedef size_t (blkcg_pol_stat_pd_fn)(struct blkg_policy_data *pd, char *buf,
160 struct blkcg_policy {
162 /* cgroup files for the policy */
163 struct cftype *dfl_cftypes;
164 struct cftype *legacy_cftypes;
167 blkcg_pol_alloc_cpd_fn *cpd_alloc_fn;
168 blkcg_pol_init_cpd_fn *cpd_init_fn;
169 blkcg_pol_free_cpd_fn *cpd_free_fn;
170 blkcg_pol_bind_cpd_fn *cpd_bind_fn;
172 blkcg_pol_alloc_pd_fn *pd_alloc_fn;
173 blkcg_pol_init_pd_fn *pd_init_fn;
174 blkcg_pol_online_pd_fn *pd_online_fn;
175 blkcg_pol_offline_pd_fn *pd_offline_fn;
176 blkcg_pol_free_pd_fn *pd_free_fn;
177 blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn;
178 blkcg_pol_stat_pd_fn *pd_stat_fn;
181 extern struct blkcg blkcg_root;
182 extern struct cgroup_subsys_state * const blkcg_root_css;
184 struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg,
185 struct request_queue *q, bool update_hint);
186 struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
187 struct request_queue *q);
188 int blkcg_init_queue(struct request_queue *q);
189 void blkcg_drain_queue(struct request_queue *q);
190 void blkcg_exit_queue(struct request_queue *q);
192 /* Blkio controller policy registration */
193 int blkcg_policy_register(struct blkcg_policy *pol);
194 void blkcg_policy_unregister(struct blkcg_policy *pol);
195 int blkcg_activate_policy(struct request_queue *q,
196 const struct blkcg_policy *pol);
197 void blkcg_deactivate_policy(struct request_queue *q,
198 const struct blkcg_policy *pol);
200 const char *blkg_dev_name(struct blkcg_gq *blkg);
201 void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
202 u64 (*prfill)(struct seq_file *,
203 struct blkg_policy_data *, int),
204 const struct blkcg_policy *pol, int data,
206 u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v);
207 u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
208 const struct blkg_rwstat *rwstat);
209 u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off);
210 u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
212 int blkg_print_stat_bytes(struct seq_file *sf, void *v);
213 int blkg_print_stat_ios(struct seq_file *sf, void *v);
214 int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v);
215 int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v);
217 u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg,
218 struct blkcg_policy *pol, int off);
219 struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg,
220 struct blkcg_policy *pol, int off);
222 struct blkg_conf_ctx {
223 struct gendisk *disk;
224 struct blkcg_gq *blkg;
228 int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
229 char *input, struct blkg_conf_ctx *ctx);
230 void blkg_conf_finish(struct blkg_conf_ctx *ctx);
233 static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css)
235 return css ? container_of(css, struct blkcg, css) : NULL;
238 static inline struct blkcg *bio_blkcg(struct bio *bio)
240 struct cgroup_subsys_state *css;
242 if (bio && bio->bi_css)
243 return css_to_blkcg(bio->bi_css);
244 css = kthread_blkcg();
246 return css_to_blkcg(css);
247 return css_to_blkcg(task_css(current, io_cgrp_id));
250 static inline bool blk_cgroup_congested(void)
252 struct cgroup_subsys_state *css;
256 css = kthread_blkcg();
258 css = task_css(current, io_cgrp_id);
260 if (atomic_read(&css->cgroup->congestion_count)) {
271 * bio_issue_as_root_blkg - see if this bio needs to be issued as root blkg
272 * @return: true if this bio needs to be submitted with the root blkg context.
274 * In order to avoid priority inversions we sometimes need to issue a bio as if
275 * it were attached to the root blkg, and then backcharge to the actual owning
276 * blkg. The idea is we do bio_blkcg() to look up the actual context for the
277 * bio and attach the appropriate blkg to the bio. Then we call this helper and
278 * if it is true run with the root blkg for that queue and then do any
279 * backcharging to the originating cgroup once the io is complete.
281 static inline bool bio_issue_as_root_blkg(struct bio *bio)
283 return (bio->bi_opf & (REQ_META | REQ_SWAP)) != 0;
287 * blkcg_parent - get the parent of a blkcg
288 * @blkcg: blkcg of interest
290 * Return the parent blkcg of @blkcg. Can be called anytime.
292 static inline struct blkcg *blkcg_parent(struct blkcg *blkcg)
294 return css_to_blkcg(blkcg->css.parent);
298 * __blkg_lookup - internal version of blkg_lookup()
299 * @blkcg: blkcg of interest
300 * @q: request_queue of interest
301 * @update_hint: whether to update lookup hint with the result or not
303 * This is internal version and shouldn't be used by policy
304 * implementations. Looks up blkgs for the @blkcg - @q pair regardless of
305 * @q's bypass state. If @update_hint is %true, the caller should be
306 * holding @q->queue_lock and lookup hint is updated on success.
308 static inline struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg,
309 struct request_queue *q,
312 struct blkcg_gq *blkg;
314 if (blkcg == &blkcg_root)
317 blkg = rcu_dereference(blkcg->blkg_hint);
318 if (blkg && blkg->q == q)
321 return blkg_lookup_slowpath(blkcg, q, update_hint);
325 * blkg_lookup - lookup blkg for the specified blkcg - q pair
326 * @blkcg: blkcg of interest
327 * @q: request_queue of interest
329 * Lookup blkg for the @blkcg - @q pair. This function should be called
330 * under RCU read lock and is guaranteed to return %NULL if @q is bypassing
331 * - see blk_queue_bypass_start() for details.
333 static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg,
334 struct request_queue *q)
336 WARN_ON_ONCE(!rcu_read_lock_held());
338 if (unlikely(blk_queue_bypass(q)))
340 return __blkg_lookup(blkcg, q, false);
344 * blk_queue_root_blkg - return blkg for the (blkcg_root, @q) pair
345 * @q: request_queue of interest
347 * Lookup blkg for @q at the root level. See also blkg_lookup().
349 static inline struct blkcg_gq *blk_queue_root_blkg(struct request_queue *q)
355 * blkg_to_pdata - get policy private data
356 * @blkg: blkg of interest
357 * @pol: policy of interest
359 * Return pointer to private data associated with the @blkg-@pol pair.
361 static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
362 struct blkcg_policy *pol)
364 return blkg ? blkg->pd[pol->plid] : NULL;
367 static inline struct blkcg_policy_data *blkcg_to_cpd(struct blkcg *blkcg,
368 struct blkcg_policy *pol)
370 return blkcg ? blkcg->cpd[pol->plid] : NULL;
374 * pdata_to_blkg - get blkg associated with policy private data
375 * @pd: policy private data of interest
377 * @pd is policy private data. Determine the blkg it's associated with.
379 static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd)
381 return pd ? pd->blkg : NULL;
384 static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd)
386 return cpd ? cpd->blkcg : NULL;
390 * blkg_path - format cgroup path of blkg
391 * @blkg: blkg of interest
392 * @buf: target buffer
393 * @buflen: target buffer length
395 * Format the path of the cgroup of @blkg into @buf.
397 static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen)
399 return cgroup_path(blkg->blkcg->css.cgroup, buf, buflen);
403 * blkg_get - get a blkg reference
406 * The caller should be holding an existing reference.
408 static inline void blkg_get(struct blkcg_gq *blkg)
410 WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0);
411 atomic_inc(&blkg->refcnt);
415 * blkg_try_get - try and get a blkg reference
418 * This is for use when doing an RCU lookup of the blkg. We may be in the midst
419 * of freeing this blkg, so we can only use it if the refcnt is not zero.
421 static inline struct blkcg_gq *blkg_try_get(struct blkcg_gq *blkg)
423 if (atomic_inc_not_zero(&blkg->refcnt))
429 void __blkg_release_rcu(struct rcu_head *rcu);
432 * blkg_put - put a blkg reference
435 static inline void blkg_put(struct blkcg_gq *blkg)
437 WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0);
438 if (atomic_dec_and_test(&blkg->refcnt))
439 call_rcu(&blkg->rcu_head, __blkg_release_rcu);
443 * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants
444 * @d_blkg: loop cursor pointing to the current descendant
445 * @pos_css: used for iteration
446 * @p_blkg: target blkg to walk descendants of
448 * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU
449 * read locked. If called under either blkcg or queue lock, the iteration
450 * is guaranteed to include all and only online blkgs. The caller may
451 * update @pos_css by calling css_rightmost_descendant() to skip subtree.
452 * @p_blkg is included in the iteration and the first node to be visited.
454 #define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \
455 css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \
456 if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \
457 (p_blkg)->q, false)))
460 * blkg_for_each_descendant_post - post-order walk of a blkg's descendants
461 * @d_blkg: loop cursor pointing to the current descendant
462 * @pos_css: used for iteration
463 * @p_blkg: target blkg to walk descendants of
465 * Similar to blkg_for_each_descendant_pre() but performs post-order
466 * traversal instead. Synchronization rules are the same. @p_blkg is
467 * included in the iteration and the last node to be visited.
469 #define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \
470 css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \
471 if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \
472 (p_blkg)->q, false)))
475 * blk_get_rl - get request_list to use
476 * @q: request_queue of interest
477 * @bio: bio which will be attached to the allocated request (may be %NULL)
479 * The caller wants to allocate a request from @q to use for @bio. Find
480 * the request_list to use and obtain a reference on it. Should be called
481 * under queue_lock. This function is guaranteed to return non-%NULL
484 static inline struct request_list *blk_get_rl(struct request_queue *q,
488 struct blkcg_gq *blkg;
492 blkcg = bio_blkcg(bio);
494 /* bypass blkg lookup and use @q->root_rl directly for root */
495 if (blkcg == &blkcg_root)
499 * Try to use blkg->rl. blkg lookup may fail under memory pressure
500 * or if either the blkcg or queue is going away. Fall back to
501 * root_rl in such cases.
503 blkg = blkg_lookup(blkcg, q);
516 * blk_put_rl - put request_list
517 * @rl: request_list to put
519 * Put the reference acquired by blk_get_rl(). Should be called under
522 static inline void blk_put_rl(struct request_list *rl)
524 if (rl->blkg->blkcg != &blkcg_root)
529 * blk_rq_set_rl - associate a request with a request_list
530 * @rq: request of interest
531 * @rl: target request_list
533 * Associate @rq with @rl so that accounting and freeing can know the
534 * request_list @rq came from.
536 static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl)
542 * blk_rq_rl - return the request_list a request came from
543 * @rq: request of interest
545 * Return the request_list @rq is allocated from.
547 static inline struct request_list *blk_rq_rl(struct request *rq)
552 struct request_list *__blk_queue_next_rl(struct request_list *rl,
553 struct request_queue *q);
555 * blk_queue_for_each_rl - iterate through all request_lists of a request_queue
557 * Should be used under queue_lock.
559 #define blk_queue_for_each_rl(rl, q) \
560 for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q)))
562 static inline int blkg_stat_init(struct blkg_stat *stat, gfp_t gfp)
566 ret = percpu_counter_init(&stat->cpu_cnt, 0, gfp);
570 atomic64_set(&stat->aux_cnt, 0);
574 static inline void blkg_stat_exit(struct blkg_stat *stat)
576 percpu_counter_destroy(&stat->cpu_cnt);
580 * blkg_stat_add - add a value to a blkg_stat
581 * @stat: target blkg_stat
584 * Add @val to @stat. The caller must ensure that IRQ on the same CPU
585 * don't re-enter this function for the same counter.
587 static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val)
589 percpu_counter_add_batch(&stat->cpu_cnt, val, BLKG_STAT_CPU_BATCH);
593 * blkg_stat_read - read the current value of a blkg_stat
594 * @stat: blkg_stat to read
596 static inline uint64_t blkg_stat_read(struct blkg_stat *stat)
598 return percpu_counter_sum_positive(&stat->cpu_cnt);
602 * blkg_stat_reset - reset a blkg_stat
603 * @stat: blkg_stat to reset
605 static inline void blkg_stat_reset(struct blkg_stat *stat)
607 percpu_counter_set(&stat->cpu_cnt, 0);
608 atomic64_set(&stat->aux_cnt, 0);
612 * blkg_stat_add_aux - add a blkg_stat into another's aux count
613 * @to: the destination blkg_stat
616 * Add @from's count including the aux one to @to's aux count.
618 static inline void blkg_stat_add_aux(struct blkg_stat *to,
619 struct blkg_stat *from)
621 atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt),
625 static inline int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp)
629 for (i = 0; i < BLKG_RWSTAT_NR; i++) {
630 ret = percpu_counter_init(&rwstat->cpu_cnt[i], 0, gfp);
633 percpu_counter_destroy(&rwstat->cpu_cnt[i]);
636 atomic64_set(&rwstat->aux_cnt[i], 0);
641 static inline void blkg_rwstat_exit(struct blkg_rwstat *rwstat)
645 for (i = 0; i < BLKG_RWSTAT_NR; i++)
646 percpu_counter_destroy(&rwstat->cpu_cnt[i]);
650 * blkg_rwstat_add - add a value to a blkg_rwstat
651 * @rwstat: target blkg_rwstat
652 * @op: REQ_OP and flags
655 * Add @val to @rwstat. The counters are chosen according to @rw. The
656 * caller is responsible for synchronizing calls to this function.
658 static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat,
659 unsigned int op, uint64_t val)
661 struct percpu_counter *cnt;
663 if (op_is_discard(op))
664 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_DISCARD];
665 else if (op_is_write(op))
666 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_WRITE];
668 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_READ];
670 percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH);
673 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_SYNC];
675 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_ASYNC];
677 percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH);
681 * blkg_rwstat_read - read the current values of a blkg_rwstat
682 * @rwstat: blkg_rwstat to read
684 * Read the current snapshot of @rwstat and return it in the aux counts.
686 static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat)
688 struct blkg_rwstat result;
691 for (i = 0; i < BLKG_RWSTAT_NR; i++)
692 atomic64_set(&result.aux_cnt[i],
693 percpu_counter_sum_positive(&rwstat->cpu_cnt[i]));
698 * blkg_rwstat_total - read the total count of a blkg_rwstat
699 * @rwstat: blkg_rwstat to read
701 * Return the total count of @rwstat regardless of the IO direction. This
702 * function can be called without synchronization and takes care of u64
705 static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat)
707 struct blkg_rwstat tmp = blkg_rwstat_read(rwstat);
709 return atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) +
710 atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]);
714 * blkg_rwstat_reset - reset a blkg_rwstat
715 * @rwstat: blkg_rwstat to reset
717 static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat)
721 for (i = 0; i < BLKG_RWSTAT_NR; i++) {
722 percpu_counter_set(&rwstat->cpu_cnt[i], 0);
723 atomic64_set(&rwstat->aux_cnt[i], 0);
728 * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count
729 * @to: the destination blkg_rwstat
732 * Add @from's count including the aux one to @to's aux count.
734 static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to,
735 struct blkg_rwstat *from)
737 u64 sum[BLKG_RWSTAT_NR];
740 for (i = 0; i < BLKG_RWSTAT_NR; i++)
741 sum[i] = percpu_counter_sum_positive(&from->cpu_cnt[i]);
743 for (i = 0; i < BLKG_RWSTAT_NR; i++)
744 atomic64_add(sum[i] + atomic64_read(&from->aux_cnt[i]),
748 #ifdef CONFIG_BLK_DEV_THROTTLING
749 extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
752 static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
753 struct bio *bio) { return false; }
756 static inline bool blkcg_bio_issue_check(struct request_queue *q,
760 struct blkcg_gq *blkg;
764 blkcg = bio_blkcg(bio);
766 /* associate blkcg if bio hasn't attached one */
767 bio_associate_blkcg(bio, &blkcg->css);
769 blkg = blkg_lookup(blkcg, q);
770 if (unlikely(!blkg)) {
771 spin_lock_irq(q->queue_lock);
772 blkg = blkg_lookup_create(blkcg, q);
775 spin_unlock_irq(q->queue_lock);
778 throtl = blk_throtl_bio(q, blkg, bio);
781 blkg = blkg ?: q->root_blkg;
783 * If the bio is flagged with BIO_QUEUE_ENTERED it means this
784 * is a split bio and we would have already accounted for the
787 if (!bio_flagged(bio, BIO_QUEUE_ENTERED))
788 blkg_rwstat_add(&blkg->stat_bytes, bio->bi_opf,
789 bio->bi_iter.bi_size);
790 blkg_rwstat_add(&blkg->stat_ios, bio->bi_opf, 1);
797 static inline void blkcg_use_delay(struct blkcg_gq *blkg)
799 if (atomic_add_return(1, &blkg->use_delay) == 1)
800 atomic_inc(&blkg->blkcg->css.cgroup->congestion_count);
803 static inline int blkcg_unuse_delay(struct blkcg_gq *blkg)
805 int old = atomic_read(&blkg->use_delay);
811 * We do this song and dance because we can race with somebody else
812 * adding or removing delay. If we just did an atomic_dec we'd end up
813 * negative and we'd already be in trouble. We need to subtract 1 and
814 * then check to see if we were the last delay so we can drop the
815 * congestion count on the cgroup.
818 int cur = atomic_cmpxchg(&blkg->use_delay, old, old - 1);
827 atomic_dec(&blkg->blkcg->css.cgroup->congestion_count);
831 static inline void blkcg_clear_delay(struct blkcg_gq *blkg)
833 int old = atomic_read(&blkg->use_delay);
836 /* We only want 1 person clearing the congestion count for this blkg. */
838 int cur = atomic_cmpxchg(&blkg->use_delay, old, 0);
840 atomic_dec(&blkg->blkcg->css.cgroup->congestion_count);
847 void blkcg_add_delay(struct blkcg_gq *blkg, u64 now, u64 delta);
848 void blkcg_schedule_throttle(struct request_queue *q, bool use_memdelay);
849 void blkcg_maybe_throttle_current(void);
850 #else /* CONFIG_BLK_CGROUP */
855 struct blkg_policy_data {
858 struct blkcg_policy_data {
864 struct blkcg_policy {
867 #define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL))
869 static inline void blkcg_maybe_throttle_current(void) { }
870 static inline bool blk_cgroup_congested(void) { return false; }
874 static inline void blkcg_schedule_throttle(struct request_queue *q, bool use_memdelay) { }
876 static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; }
877 static inline struct blkcg_gq *blk_queue_root_blkg(struct request_queue *q)
879 static inline int blkcg_init_queue(struct request_queue *q) { return 0; }
880 static inline void blkcg_drain_queue(struct request_queue *q) { }
881 static inline void blkcg_exit_queue(struct request_queue *q) { }
882 static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; }
883 static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { }
884 static inline int blkcg_activate_policy(struct request_queue *q,
885 const struct blkcg_policy *pol) { return 0; }
886 static inline void blkcg_deactivate_policy(struct request_queue *q,
887 const struct blkcg_policy *pol) { }
889 static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; }
891 static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
892 struct blkcg_policy *pol) { return NULL; }
893 static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; }
894 static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; }
895 static inline void blkg_get(struct blkcg_gq *blkg) { }
896 static inline void blkg_put(struct blkcg_gq *blkg) { }
898 static inline struct request_list *blk_get_rl(struct request_queue *q,
899 struct bio *bio) { return &q->root_rl; }
900 static inline void blk_put_rl(struct request_list *rl) { }
901 static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { }
902 static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; }
904 static inline bool blkcg_bio_issue_check(struct request_queue *q,
905 struct bio *bio) { return true; }
907 #define blk_queue_for_each_rl(rl, q) \
908 for ((rl) = &(q)->root_rl; (rl); (rl) = NULL)
910 #endif /* CONFIG_BLOCK */
911 #endif /* CONFIG_BLK_CGROUP */
912 #endif /* _BLK_CGROUP_H */