]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - include/linux/blk-cgroup.h
ceph: quota: add initial infrastructure to support cephfs quotas
[mirror_ubuntu-bionic-kernel.git] / include / linux / blk-cgroup.h
CommitLineData
b2441318 1/* SPDX-License-Identifier: GPL-2.0 */
31e4c28d
VG
2#ifndef _BLK_CGROUP_H
3#define _BLK_CGROUP_H
4/*
5 * Common Block IO controller cgroup interface
6 *
7 * Based on ideas and code from CFQ, CFS and BFQ:
8 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
9 *
10 * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it>
11 * Paolo Valente <paolo.valente@unimore.it>
12 *
13 * Copyright (C) 2009 Vivek Goyal <vgoyal@redhat.com>
14 * Nauman Rafique <nauman@google.com>
15 */
16
17#include <linux/cgroup.h>
24bdb8ef 18#include <linux/percpu_counter.h>
829fdb50 19#include <linux/seq_file.h>
a637120e 20#include <linux/radix-tree.h>
a051661c 21#include <linux/blkdev.h>
a5049a8a 22#include <linux/atomic.h>
902ec5b6 23#include <linux/kthread.h>
31e4c28d 24
24bdb8ef
TH
25/* percpu_counter batch for blkg_[rw]stats, per-cpu drift doesn't matter */
26#define BLKG_STAT_CPU_BATCH (INT_MAX / 2)
27
9355aede
VG
28/* Max limits for throttle policy */
29#define THROTL_IOPS_MAX UINT_MAX
30
f48ec1d7
TH
31#ifdef CONFIG_BLK_CGROUP
32
edcb0722
TH
33enum blkg_rwstat_type {
34 BLKG_RWSTAT_READ,
35 BLKG_RWSTAT_WRITE,
36 BLKG_RWSTAT_SYNC,
37 BLKG_RWSTAT_ASYNC,
38
39 BLKG_RWSTAT_NR,
40 BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR,
303a3acb
DS
41};
42
a637120e
TH
43struct blkcg_gq;
44
3c798398 45struct blkcg {
36558c8a
TH
46 struct cgroup_subsys_state css;
47 spinlock_t lock;
a637120e
TH
48
49 struct radix_tree_root blkg_tree;
55679c8d 50 struct blkcg_gq __rcu *blkg_hint;
36558c8a 51 struct hlist_head blkg_list;
9a9e8a26 52
81437648 53 struct blkcg_policy_data *cpd[BLKCG_MAX_POLS];
52ebea74 54
7876f930 55 struct list_head all_blkcgs_node;
52ebea74
TH
56#ifdef CONFIG_CGROUP_WRITEBACK
57 struct list_head cgwb_list;
58#endif
31e4c28d
VG
59};
60
e6269c44
TH
61/*
62 * blkg_[rw]stat->aux_cnt is excluded for local stats but included for
24bdb8ef
TH
63 * recursive. Used to carry stats of dead children, and, for blkg_rwstat,
64 * to carry result values from read and sum operations.
e6269c44 65 */
edcb0722 66struct blkg_stat {
24bdb8ef 67 struct percpu_counter cpu_cnt;
e6269c44 68 atomic64_t aux_cnt;
edcb0722
TH
69};
70
71struct blkg_rwstat {
24bdb8ef 72 struct percpu_counter cpu_cnt[BLKG_RWSTAT_NR];
e6269c44 73 atomic64_t aux_cnt[BLKG_RWSTAT_NR];
edcb0722
TH
74};
75
f95a04af
TH
76/*
77 * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a
78 * request_queue (q). This is used by blkcg policies which need to track
79 * information per blkcg - q pair.
80 *
001bea73
TH
81 * There can be multiple active blkcg policies and each blkg:policy pair is
82 * represented by a blkg_policy_data which is allocated and freed by each
83 * policy's pd_alloc/free_fn() methods. A policy can allocate private data
84 * area by allocating larger data structure which embeds blkg_policy_data
85 * at the beginning.
f95a04af 86 */
0381411e 87struct blkg_policy_data {
b276a876 88 /* the blkg and policy id this per-policy data belongs to */
3c798398 89 struct blkcg_gq *blkg;
b276a876 90 int plid;
0381411e
TH
91};
92
e48453c3 93/*
e4a9bde9
TH
94 * Policies that need to keep per-blkcg data which is independent from any
95 * request_queue associated to it should implement cpd_alloc/free_fn()
96 * methods. A policy can allocate private data area by allocating larger
97 * data structure which embeds blkcg_policy_data at the beginning.
98 * cpd_init() is invoked to let each policy handle per-blkcg data.
e48453c3
AA
99 */
100struct blkcg_policy_data {
81437648
TH
101 /* the blkcg and policy id this per-policy data belongs to */
102 struct blkcg *blkcg;
e48453c3 103 int plid;
e48453c3
AA
104};
105
3c798398
TH
106/* association between a blk cgroup and a request queue */
107struct blkcg_gq {
c875f4d0 108 /* Pointer to the associated request_queue */
36558c8a
TH
109 struct request_queue *q;
110 struct list_head q_node;
111 struct hlist_node blkcg_node;
3c798398 112 struct blkcg *blkcg;
3c547865 113
ce7acfea
TH
114 /*
115 * Each blkg gets congested separately and the congestion state is
116 * propagated to the matching bdi_writeback_congested.
117 */
118 struct bdi_writeback_congested *wb_congested;
119
3c547865
TH
120 /* all non-root blkcg_gq's are guaranteed to have access to parent */
121 struct blkcg_gq *parent;
122
a051661c
TH
123 /* request allocation list for this blkcg-q pair */
124 struct request_list rl;
3c547865 125
1adaf3dd 126 /* reference count */
a5049a8a 127 atomic_t refcnt;
22084190 128
f427d909
TH
129 /* is this blkg online? protected by both blkcg and q locks */
130 bool online;
131
77ea7338
TH
132 struct blkg_rwstat stat_bytes;
133 struct blkg_rwstat stat_ios;
134
36558c8a 135 struct blkg_policy_data *pd[BLKCG_MAX_POLS];
1adaf3dd 136
36558c8a 137 struct rcu_head rcu_head;
31e4c28d
VG
138};
139
e4a9bde9 140typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp);
81437648 141typedef void (blkcg_pol_init_cpd_fn)(struct blkcg_policy_data *cpd);
e4a9bde9 142typedef void (blkcg_pol_free_cpd_fn)(struct blkcg_policy_data *cpd);
69d7fde5 143typedef void (blkcg_pol_bind_cpd_fn)(struct blkcg_policy_data *cpd);
001bea73 144typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node);
a9520cd6
TH
145typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd);
146typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd);
147typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd);
001bea73 148typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd);
a9520cd6 149typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd);
3e252066 150
3c798398 151struct blkcg_policy {
36558c8a 152 int plid;
36558c8a 153 /* cgroup files for the policy */
2ee867dc 154 struct cftype *dfl_cftypes;
880f50e2 155 struct cftype *legacy_cftypes;
f9fcc2d3
TH
156
157 /* operations */
e4a9bde9 158 blkcg_pol_alloc_cpd_fn *cpd_alloc_fn;
e48453c3 159 blkcg_pol_init_cpd_fn *cpd_init_fn;
e4a9bde9 160 blkcg_pol_free_cpd_fn *cpd_free_fn;
69d7fde5 161 blkcg_pol_bind_cpd_fn *cpd_bind_fn;
e4a9bde9 162
001bea73 163 blkcg_pol_alloc_pd_fn *pd_alloc_fn;
f9fcc2d3 164 blkcg_pol_init_pd_fn *pd_init_fn;
f427d909
TH
165 blkcg_pol_online_pd_fn *pd_online_fn;
166 blkcg_pol_offline_pd_fn *pd_offline_fn;
001bea73 167 blkcg_pol_free_pd_fn *pd_free_fn;
f9fcc2d3 168 blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn;
3e252066
VG
169};
170
3c798398 171extern struct blkcg blkcg_root;
496d5e75 172extern struct cgroup_subsys_state * const blkcg_root_css;
36558c8a 173
24f29046
TH
174struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg,
175 struct request_queue *q, bool update_hint);
3c798398
TH
176struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
177 struct request_queue *q);
36558c8a
TH
178int blkcg_init_queue(struct request_queue *q);
179void blkcg_drain_queue(struct request_queue *q);
180void blkcg_exit_queue(struct request_queue *q);
5efd6113 181
3e252066 182/* Blkio controller policy registration */
d5bf0291 183int blkcg_policy_register(struct blkcg_policy *pol);
3c798398 184void blkcg_policy_unregister(struct blkcg_policy *pol);
36558c8a 185int blkcg_activate_policy(struct request_queue *q,
3c798398 186 const struct blkcg_policy *pol);
36558c8a 187void blkcg_deactivate_policy(struct request_queue *q,
3c798398 188 const struct blkcg_policy *pol);
3e252066 189
dd165eb3 190const char *blkg_dev_name(struct blkcg_gq *blkg);
3c798398 191void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
f95a04af
TH
192 u64 (*prfill)(struct seq_file *,
193 struct blkg_policy_data *, int),
3c798398 194 const struct blkcg_policy *pol, int data,
ec399347 195 bool show_total);
f95a04af
TH
196u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v);
197u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
829fdb50 198 const struct blkg_rwstat *rwstat);
f95a04af
TH
199u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off);
200u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
201 int off);
77ea7338
TH
202int blkg_print_stat_bytes(struct seq_file *sf, void *v);
203int blkg_print_stat_ios(struct seq_file *sf, void *v);
204int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v);
205int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v);
829fdb50 206
f12c74ca
TH
207u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg,
208 struct blkcg_policy *pol, int off);
209struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg,
210 struct blkcg_policy *pol, int off);
16b3de66 211
829fdb50 212struct blkg_conf_ctx {
36558c8a 213 struct gendisk *disk;
3c798398 214 struct blkcg_gq *blkg;
36aa9e5f 215 char *body;
829fdb50
TH
216};
217
3c798398 218int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
36aa9e5f 219 char *input, struct blkg_conf_ctx *ctx);
829fdb50
TH
220void blkg_conf_finish(struct blkg_conf_ctx *ctx);
221
222
a7c6d554
TH
223static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css)
224{
225 return css ? container_of(css, struct blkcg, css) : NULL;
226}
227
b1208b56
TH
228static inline struct blkcg *bio_blkcg(struct bio *bio)
229{
902ec5b6
SL
230 struct cgroup_subsys_state *css;
231
b1208b56 232 if (bio && bio->bi_css)
a7c6d554 233 return css_to_blkcg(bio->bi_css);
902ec5b6
SL
234 css = kthread_blkcg();
235 if (css)
236 return css_to_blkcg(css);
237 return css_to_blkcg(task_css(current, io_cgrp_id));
fd383c2d
TH
238}
239
3c547865
TH
240/**
241 * blkcg_parent - get the parent of a blkcg
242 * @blkcg: blkcg of interest
243 *
244 * Return the parent blkcg of @blkcg. Can be called anytime.
245 */
246static inline struct blkcg *blkcg_parent(struct blkcg *blkcg)
247{
5c9d535b 248 return css_to_blkcg(blkcg->css.parent);
3c547865
TH
249}
250
24f29046
TH
251/**
252 * __blkg_lookup - internal version of blkg_lookup()
253 * @blkcg: blkcg of interest
254 * @q: request_queue of interest
255 * @update_hint: whether to update lookup hint with the result or not
256 *
257 * This is internal version and shouldn't be used by policy
258 * implementations. Looks up blkgs for the @blkcg - @q pair regardless of
259 * @q's bypass state. If @update_hint is %true, the caller should be
260 * holding @q->queue_lock and lookup hint is updated on success.
261 */
262static inline struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg,
263 struct request_queue *q,
264 bool update_hint)
265{
266 struct blkcg_gq *blkg;
267
85b6bc9d
TH
268 if (blkcg == &blkcg_root)
269 return q->root_blkg;
270
24f29046
TH
271 blkg = rcu_dereference(blkcg->blkg_hint);
272 if (blkg && blkg->q == q)
273 return blkg;
274
275 return blkg_lookup_slowpath(blkcg, q, update_hint);
276}
277
278/**
279 * blkg_lookup - lookup blkg for the specified blkcg - q pair
280 * @blkcg: blkcg of interest
281 * @q: request_queue of interest
282 *
283 * Lookup blkg for the @blkcg - @q pair. This function should be called
284 * under RCU read lock and is guaranteed to return %NULL if @q is bypassing
285 * - see blk_queue_bypass_start() for details.
286 */
287static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg,
288 struct request_queue *q)
289{
290 WARN_ON_ONCE(!rcu_read_lock_held());
291
292 if (unlikely(blk_queue_bypass(q)))
293 return NULL;
294 return __blkg_lookup(blkcg, q, false);
295}
296
0381411e
TH
297/**
298 * blkg_to_pdata - get policy private data
299 * @blkg: blkg of interest
300 * @pol: policy of interest
301 *
302 * Return pointer to private data associated with the @blkg-@pol pair.
303 */
f95a04af
TH
304static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
305 struct blkcg_policy *pol)
0381411e 306{
f95a04af 307 return blkg ? blkg->pd[pol->plid] : NULL;
0381411e
TH
308}
309
e48453c3
AA
310static inline struct blkcg_policy_data *blkcg_to_cpd(struct blkcg *blkcg,
311 struct blkcg_policy *pol)
312{
81437648 313 return blkcg ? blkcg->cpd[pol->plid] : NULL;
e48453c3
AA
314}
315
0381411e
TH
316/**
317 * pdata_to_blkg - get blkg associated with policy private data
f95a04af 318 * @pd: policy private data of interest
0381411e 319 *
f95a04af 320 * @pd is policy private data. Determine the blkg it's associated with.
0381411e 321 */
f95a04af 322static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd)
0381411e 323{
f95a04af 324 return pd ? pd->blkg : NULL;
0381411e
TH
325}
326
81437648
TH
327static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd)
328{
329 return cpd ? cpd->blkcg : NULL;
330}
331
54e7ed12
TH
332/**
333 * blkg_path - format cgroup path of blkg
334 * @blkg: blkg of interest
335 * @buf: target buffer
336 * @buflen: target buffer length
337 *
338 * Format the path of the cgroup of @blkg into @buf.
339 */
3c798398 340static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen)
afc24d49 341{
4c737b41 342 return cgroup_path(blkg->blkcg->css.cgroup, buf, buflen);
afc24d49
VG
343}
344
1adaf3dd
TH
345/**
346 * blkg_get - get a blkg reference
347 * @blkg: blkg to get
348 *
a5049a8a 349 * The caller should be holding an existing reference.
1adaf3dd 350 */
3c798398 351static inline void blkg_get(struct blkcg_gq *blkg)
1adaf3dd 352{
a5049a8a
TH
353 WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0);
354 atomic_inc(&blkg->refcnt);
1adaf3dd
TH
355}
356
2a4fd070 357void __blkg_release_rcu(struct rcu_head *rcu);
1adaf3dd
TH
358
359/**
360 * blkg_put - put a blkg reference
361 * @blkg: blkg to put
1adaf3dd 362 */
3c798398 363static inline void blkg_put(struct blkcg_gq *blkg)
1adaf3dd 364{
a5049a8a
TH
365 WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0);
366 if (atomic_dec_and_test(&blkg->refcnt))
2a4fd070 367 call_rcu(&blkg->rcu_head, __blkg_release_rcu);
1adaf3dd
TH
368}
369
dd4a4ffc
TH
370/**
371 * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants
372 * @d_blkg: loop cursor pointing to the current descendant
492eb21b 373 * @pos_css: used for iteration
dd4a4ffc
TH
374 * @p_blkg: target blkg to walk descendants of
375 *
376 * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU
377 * read locked. If called under either blkcg or queue lock, the iteration
378 * is guaranteed to include all and only online blkgs. The caller may
492eb21b 379 * update @pos_css by calling css_rightmost_descendant() to skip subtree.
bd8815a6 380 * @p_blkg is included in the iteration and the first node to be visited.
dd4a4ffc 381 */
492eb21b
TH
382#define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \
383 css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \
384 if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \
dd4a4ffc
TH
385 (p_blkg)->q, false)))
386
aa539cb3
TH
387/**
388 * blkg_for_each_descendant_post - post-order walk of a blkg's descendants
389 * @d_blkg: loop cursor pointing to the current descendant
492eb21b 390 * @pos_css: used for iteration
aa539cb3
TH
391 * @p_blkg: target blkg to walk descendants of
392 *
393 * Similar to blkg_for_each_descendant_pre() but performs post-order
bd8815a6
TH
394 * traversal instead. Synchronization rules are the same. @p_blkg is
395 * included in the iteration and the last node to be visited.
aa539cb3 396 */
492eb21b
TH
397#define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \
398 css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \
399 if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \
aa539cb3
TH
400 (p_blkg)->q, false)))
401
a051661c
TH
402/**
403 * blk_get_rl - get request_list to use
404 * @q: request_queue of interest
405 * @bio: bio which will be attached to the allocated request (may be %NULL)
406 *
407 * The caller wants to allocate a request from @q to use for @bio. Find
408 * the request_list to use and obtain a reference on it. Should be called
409 * under queue_lock. This function is guaranteed to return non-%NULL
410 * request_list.
411 */
412static inline struct request_list *blk_get_rl(struct request_queue *q,
413 struct bio *bio)
414{
415 struct blkcg *blkcg;
416 struct blkcg_gq *blkg;
417
418 rcu_read_lock();
419
420 blkcg = bio_blkcg(bio);
421
422 /* bypass blkg lookup and use @q->root_rl directly for root */
423 if (blkcg == &blkcg_root)
424 goto root_rl;
425
426 /*
427 * Try to use blkg->rl. blkg lookup may fail under memory pressure
428 * or if either the blkcg or queue is going away. Fall back to
429 * root_rl in such cases.
430 */
ae118896
TH
431 blkg = blkg_lookup(blkcg, q);
432 if (unlikely(!blkg))
a051661c
TH
433 goto root_rl;
434
435 blkg_get(blkg);
436 rcu_read_unlock();
437 return &blkg->rl;
438root_rl:
439 rcu_read_unlock();
440 return &q->root_rl;
441}
442
443/**
444 * blk_put_rl - put request_list
445 * @rl: request_list to put
446 *
447 * Put the reference acquired by blk_get_rl(). Should be called under
448 * queue_lock.
449 */
450static inline void blk_put_rl(struct request_list *rl)
451{
401efbf8 452 if (rl->blkg->blkcg != &blkcg_root)
a051661c
TH
453 blkg_put(rl->blkg);
454}
455
456/**
457 * blk_rq_set_rl - associate a request with a request_list
458 * @rq: request of interest
459 * @rl: target request_list
460 *
461 * Associate @rq with @rl so that accounting and freeing can know the
462 * request_list @rq came from.
463 */
464static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl)
465{
466 rq->rl = rl;
467}
468
469/**
470 * blk_rq_rl - return the request_list a request came from
471 * @rq: request of interest
472 *
473 * Return the request_list @rq is allocated from.
474 */
475static inline struct request_list *blk_rq_rl(struct request *rq)
476{
477 return rq->rl;
478}
479
480struct request_list *__blk_queue_next_rl(struct request_list *rl,
481 struct request_queue *q);
482/**
483 * blk_queue_for_each_rl - iterate through all request_lists of a request_queue
484 *
485 * Should be used under queue_lock.
486 */
487#define blk_queue_for_each_rl(rl, q) \
488 for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q)))
489
24bdb8ef 490static inline int blkg_stat_init(struct blkg_stat *stat, gfp_t gfp)
90d3839b 491{
24bdb8ef
TH
492 int ret;
493
494 ret = percpu_counter_init(&stat->cpu_cnt, 0, gfp);
495 if (ret)
496 return ret;
497
e6269c44 498 atomic64_set(&stat->aux_cnt, 0);
24bdb8ef
TH
499 return 0;
500}
501
502static inline void blkg_stat_exit(struct blkg_stat *stat)
503{
504 percpu_counter_destroy(&stat->cpu_cnt);
90d3839b
PZ
505}
506
edcb0722
TH
507/**
508 * blkg_stat_add - add a value to a blkg_stat
509 * @stat: target blkg_stat
510 * @val: value to add
511 *
24bdb8ef
TH
512 * Add @val to @stat. The caller must ensure that IRQ on the same CPU
513 * don't re-enter this function for the same counter.
edcb0722
TH
514 */
515static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val)
516{
104b4e51 517 percpu_counter_add_batch(&stat->cpu_cnt, val, BLKG_STAT_CPU_BATCH);
edcb0722
TH
518}
519
520/**
521 * blkg_stat_read - read the current value of a blkg_stat
522 * @stat: blkg_stat to read
edcb0722
TH
523 */
524static inline uint64_t blkg_stat_read(struct blkg_stat *stat)
525{
24bdb8ef 526 return percpu_counter_sum_positive(&stat->cpu_cnt);
edcb0722
TH
527}
528
529/**
530 * blkg_stat_reset - reset a blkg_stat
531 * @stat: blkg_stat to reset
532 */
533static inline void blkg_stat_reset(struct blkg_stat *stat)
534{
24bdb8ef 535 percpu_counter_set(&stat->cpu_cnt, 0);
e6269c44 536 atomic64_set(&stat->aux_cnt, 0);
edcb0722
TH
537}
538
16b3de66 539/**
e6269c44 540 * blkg_stat_add_aux - add a blkg_stat into another's aux count
16b3de66
TH
541 * @to: the destination blkg_stat
542 * @from: the source
543 *
e6269c44 544 * Add @from's count including the aux one to @to's aux count.
16b3de66 545 */
e6269c44
TH
546static inline void blkg_stat_add_aux(struct blkg_stat *to,
547 struct blkg_stat *from)
16b3de66 548{
e6269c44
TH
549 atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt),
550 &to->aux_cnt);
16b3de66
TH
551}
552
24bdb8ef 553static inline int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp)
90d3839b 554{
24bdb8ef
TH
555 int i, ret;
556
557 for (i = 0; i < BLKG_RWSTAT_NR; i++) {
558 ret = percpu_counter_init(&rwstat->cpu_cnt[i], 0, gfp);
559 if (ret) {
560 while (--i >= 0)
561 percpu_counter_destroy(&rwstat->cpu_cnt[i]);
562 return ret;
563 }
564 atomic64_set(&rwstat->aux_cnt[i], 0);
565 }
566 return 0;
567}
e6269c44 568
24bdb8ef
TH
569static inline void blkg_rwstat_exit(struct blkg_rwstat *rwstat)
570{
571 int i;
e6269c44
TH
572
573 for (i = 0; i < BLKG_RWSTAT_NR; i++)
24bdb8ef 574 percpu_counter_destroy(&rwstat->cpu_cnt[i]);
90d3839b
PZ
575}
576
edcb0722
TH
577/**
578 * blkg_rwstat_add - add a value to a blkg_rwstat
579 * @rwstat: target blkg_rwstat
ef295ecf 580 * @op: REQ_OP and flags
edcb0722
TH
581 * @val: value to add
582 *
583 * Add @val to @rwstat. The counters are chosen according to @rw. The
584 * caller is responsible for synchronizing calls to this function.
585 */
586static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat,
ef295ecf 587 unsigned int op, uint64_t val)
edcb0722 588{
24bdb8ef 589 struct percpu_counter *cnt;
edcb0722 590
63a4cc24 591 if (op_is_write(op))
24bdb8ef 592 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_WRITE];
edcb0722 593 else
24bdb8ef
TH
594 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_READ];
595
104b4e51 596 percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH);
24bdb8ef 597
d71d9ae1 598 if (op_is_sync(op))
24bdb8ef 599 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_SYNC];
edcb0722 600 else
24bdb8ef 601 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_ASYNC];
edcb0722 602
104b4e51 603 percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH);
edcb0722
TH
604}
605
606/**
607 * blkg_rwstat_read - read the current values of a blkg_rwstat
608 * @rwstat: blkg_rwstat to read
609 *
24bdb8ef 610 * Read the current snapshot of @rwstat and return it in the aux counts.
edcb0722 611 */
c94bed89 612static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat)
edcb0722 613{
24bdb8ef
TH
614 struct blkg_rwstat result;
615 int i;
edcb0722 616
24bdb8ef
TH
617 for (i = 0; i < BLKG_RWSTAT_NR; i++)
618 atomic64_set(&result.aux_cnt[i],
619 percpu_counter_sum_positive(&rwstat->cpu_cnt[i]));
620 return result;
edcb0722
TH
621}
622
623/**
4d5e80a7 624 * blkg_rwstat_total - read the total count of a blkg_rwstat
edcb0722
TH
625 * @rwstat: blkg_rwstat to read
626 *
627 * Return the total count of @rwstat regardless of the IO direction. This
628 * function can be called without synchronization and takes care of u64
629 * atomicity.
630 */
4d5e80a7 631static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat)
edcb0722
TH
632{
633 struct blkg_rwstat tmp = blkg_rwstat_read(rwstat);
634
24bdb8ef
TH
635 return atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) +
636 atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]);
edcb0722
TH
637}
638
639/**
640 * blkg_rwstat_reset - reset a blkg_rwstat
641 * @rwstat: blkg_rwstat to reset
642 */
643static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat)
644{
e6269c44
TH
645 int i;
646
24bdb8ef
TH
647 for (i = 0; i < BLKG_RWSTAT_NR; i++) {
648 percpu_counter_set(&rwstat->cpu_cnt[i], 0);
e6269c44 649 atomic64_set(&rwstat->aux_cnt[i], 0);
24bdb8ef 650 }
edcb0722
TH
651}
652
16b3de66 653/**
e6269c44 654 * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count
16b3de66
TH
655 * @to: the destination blkg_rwstat
656 * @from: the source
657 *
e6269c44 658 * Add @from's count including the aux one to @to's aux count.
16b3de66 659 */
e6269c44
TH
660static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to,
661 struct blkg_rwstat *from)
16b3de66
TH
662{
663 struct blkg_rwstat v = blkg_rwstat_read(from);
664 int i;
665
16b3de66 666 for (i = 0; i < BLKG_RWSTAT_NR; i++)
24bdb8ef
TH
667 atomic64_add(atomic64_read(&v.aux_cnt[i]) +
668 atomic64_read(&from->aux_cnt[i]),
e6269c44 669 &to->aux_cnt[i]);
16b3de66
TH
670}
671
ae118896
TH
672#ifdef CONFIG_BLK_DEV_THROTTLING
673extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
674 struct bio *bio);
675#else
676static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
677 struct bio *bio) { return false; }
678#endif
679
680static inline bool blkcg_bio_issue_check(struct request_queue *q,
681 struct bio *bio)
682{
683 struct blkcg *blkcg;
684 struct blkcg_gq *blkg;
685 bool throtl = false;
686
687 rcu_read_lock();
688 blkcg = bio_blkcg(bio);
689
007cc56b
SL
690 /* associate blkcg if bio hasn't attached one */
691 bio_associate_blkcg(bio, &blkcg->css);
692
ae118896
TH
693 blkg = blkg_lookup(blkcg, q);
694 if (unlikely(!blkg)) {
695 spin_lock_irq(q->queue_lock);
696 blkg = blkg_lookup_create(blkcg, q);
697 if (IS_ERR(blkg))
698 blkg = NULL;
699 spin_unlock_irq(q->queue_lock);
700 }
701
702 throtl = blk_throtl_bio(q, blkg, bio);
703
77ea7338
TH
704 if (!throtl) {
705 blkg = blkg ?: q->root_blkg;
ef295ecf 706 blkg_rwstat_add(&blkg->stat_bytes, bio->bi_opf,
77ea7338 707 bio->bi_iter.bi_size);
ef295ecf 708 blkg_rwstat_add(&blkg->stat_ios, bio->bi_opf, 1);
77ea7338
TH
709 }
710
ae118896
TH
711 rcu_read_unlock();
712 return !throtl;
713}
714
36558c8a
TH
715#else /* CONFIG_BLK_CGROUP */
716
efa7d1c7
TH
717struct blkcg {
718};
2f5ea477 719
f95a04af
TH
720struct blkg_policy_data {
721};
722
e48453c3
AA
723struct blkcg_policy_data {
724};
725
3c798398 726struct blkcg_gq {
2f5ea477
JA
727};
728
3c798398 729struct blkcg_policy {
3e252066
VG
730};
731
496d5e75
TH
732#define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL))
733
efa7d1c7
TH
734#ifdef CONFIG_BLOCK
735
3c798398 736static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; }
5efd6113
TH
737static inline int blkcg_init_queue(struct request_queue *q) { return 0; }
738static inline void blkcg_drain_queue(struct request_queue *q) { }
739static inline void blkcg_exit_queue(struct request_queue *q) { }
d5bf0291 740static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; }
3c798398 741static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { }
a2b1693b 742static inline int blkcg_activate_policy(struct request_queue *q,
3c798398 743 const struct blkcg_policy *pol) { return 0; }
a2b1693b 744static inline void blkcg_deactivate_policy(struct request_queue *q,
3c798398
TH
745 const struct blkcg_policy *pol) { }
746
b1208b56 747static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; }
a051661c 748
f95a04af
TH
749static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
750 struct blkcg_policy *pol) { return NULL; }
751static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; }
3c798398
TH
752static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; }
753static inline void blkg_get(struct blkcg_gq *blkg) { }
754static inline void blkg_put(struct blkcg_gq *blkg) { }
afc24d49 755
a051661c
TH
756static inline struct request_list *blk_get_rl(struct request_queue *q,
757 struct bio *bio) { return &q->root_rl; }
758static inline void blk_put_rl(struct request_list *rl) { }
759static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { }
760static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; }
761
ae118896
TH
762static inline bool blkcg_bio_issue_check(struct request_queue *q,
763 struct bio *bio) { return true; }
764
a051661c
TH
765#define blk_queue_for_each_rl(rl, q) \
766 for ((rl) = &(q)->root_rl; (rl); (rl) = NULL)
767
efa7d1c7 768#endif /* CONFIG_BLOCK */
36558c8a
TH
769#endif /* CONFIG_BLK_CGROUP */
770#endif /* _BLK_CGROUP_H */