]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - include/linux/blk-cgroup.h
blkcg: rename subsystem name from blkio to io
[mirror_ubuntu-zesty-kernel.git] / include / linux / 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>
24bdb8ef 17#include <linux/percpu_counter.h>
829fdb50 18#include <linux/seq_file.h>
a637120e 19#include <linux/radix-tree.h>
a051661c 20#include <linux/blkdev.h>
a5049a8a 21#include <linux/atomic.h>
31e4c28d 22
24bdb8ef
TH
23/* percpu_counter batch for blkg_[rw]stats, per-cpu drift doesn't matter */
24#define BLKG_STAT_CPU_BATCH (INT_MAX / 2)
25
9355aede
VG
26/* Max limits for throttle policy */
27#define THROTL_IOPS_MAX UINT_MAX
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 50
81437648 51 struct blkcg_policy_data *cpd[BLKCG_MAX_POLS];
52ebea74 52
7876f930 53 struct list_head all_blkcgs_node;
52ebea74
TH
54#ifdef CONFIG_CGROUP_WRITEBACK
55 struct list_head cgwb_list;
56#endif
31e4c28d
VG
57};
58
e6269c44
TH
59/*
60 * blkg_[rw]stat->aux_cnt is excluded for local stats but included for
24bdb8ef
TH
61 * recursive. Used to carry stats of dead children, and, for blkg_rwstat,
62 * to carry result values from read and sum operations.
e6269c44 63 */
edcb0722 64struct blkg_stat {
24bdb8ef 65 struct percpu_counter cpu_cnt;
e6269c44 66 atomic64_t aux_cnt;
edcb0722
TH
67};
68
69struct blkg_rwstat {
24bdb8ef 70 struct percpu_counter cpu_cnt[BLKG_RWSTAT_NR];
e6269c44 71 atomic64_t aux_cnt[BLKG_RWSTAT_NR];
edcb0722
TH
72};
73
f95a04af
TH
74/*
75 * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a
76 * request_queue (q). This is used by blkcg policies which need to track
77 * information per blkcg - q pair.
78 *
001bea73
TH
79 * There can be multiple active blkcg policies and each blkg:policy pair is
80 * represented by a blkg_policy_data which is allocated and freed by each
81 * policy's pd_alloc/free_fn() methods. A policy can allocate private data
82 * area by allocating larger data structure which embeds blkg_policy_data
83 * at the beginning.
f95a04af 84 */
0381411e 85struct blkg_policy_data {
b276a876 86 /* the blkg and policy id this per-policy data belongs to */
3c798398 87 struct blkcg_gq *blkg;
b276a876 88 int plid;
0381411e
TH
89};
90
e48453c3 91/*
e4a9bde9
TH
92 * Policies that need to keep per-blkcg data which is independent from any
93 * request_queue associated to it should implement cpd_alloc/free_fn()
94 * methods. A policy can allocate private data area by allocating larger
95 * data structure which embeds blkcg_policy_data at the beginning.
96 * cpd_init() is invoked to let each policy handle per-blkcg data.
e48453c3
AA
97 */
98struct blkcg_policy_data {
81437648
TH
99 /* the blkcg and policy id this per-policy data belongs to */
100 struct blkcg *blkcg;
e48453c3 101 int plid;
e48453c3
AA
102};
103
3c798398
TH
104/* association between a blk cgroup and a request queue */
105struct blkcg_gq {
c875f4d0 106 /* Pointer to the associated request_queue */
36558c8a
TH
107 struct request_queue *q;
108 struct list_head q_node;
109 struct hlist_node blkcg_node;
3c798398 110 struct blkcg *blkcg;
3c547865 111
ce7acfea
TH
112 /*
113 * Each blkg gets congested separately and the congestion state is
114 * propagated to the matching bdi_writeback_congested.
115 */
116 struct bdi_writeback_congested *wb_congested;
117
3c547865
TH
118 /* all non-root blkcg_gq's are guaranteed to have access to parent */
119 struct blkcg_gq *parent;
120
a051661c
TH
121 /* request allocation list for this blkcg-q pair */
122 struct request_list rl;
3c547865 123
1adaf3dd 124 /* reference count */
a5049a8a 125 atomic_t refcnt;
22084190 126
f427d909
TH
127 /* is this blkg online? protected by both blkcg and q locks */
128 bool online;
129
77ea7338
TH
130 struct blkg_rwstat stat_bytes;
131 struct blkg_rwstat stat_ios;
132
36558c8a 133 struct blkg_policy_data *pd[BLKCG_MAX_POLS];
1adaf3dd 134
36558c8a 135 struct rcu_head rcu_head;
31e4c28d
VG
136};
137
e4a9bde9 138typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp);
81437648 139typedef void (blkcg_pol_init_cpd_fn)(struct blkcg_policy_data *cpd);
e4a9bde9 140typedef void (blkcg_pol_free_cpd_fn)(struct blkcg_policy_data *cpd);
001bea73 141typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node);
a9520cd6
TH
142typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd);
143typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd);
144typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd);
001bea73 145typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd);
a9520cd6 146typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd);
3e252066 147
3c798398 148struct blkcg_policy {
36558c8a 149 int plid;
36558c8a
TH
150 /* cgroup files for the policy */
151 struct cftype *cftypes;
f9fcc2d3
TH
152
153 /* operations */
e4a9bde9 154 blkcg_pol_alloc_cpd_fn *cpd_alloc_fn;
e48453c3 155 blkcg_pol_init_cpd_fn *cpd_init_fn;
e4a9bde9
TH
156 blkcg_pol_free_cpd_fn *cpd_free_fn;
157
001bea73 158 blkcg_pol_alloc_pd_fn *pd_alloc_fn;
f9fcc2d3 159 blkcg_pol_init_pd_fn *pd_init_fn;
f427d909
TH
160 blkcg_pol_online_pd_fn *pd_online_fn;
161 blkcg_pol_offline_pd_fn *pd_offline_fn;
001bea73 162 blkcg_pol_free_pd_fn *pd_free_fn;
f9fcc2d3 163 blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn;
3e252066
VG
164};
165
3c798398 166extern struct blkcg blkcg_root;
496d5e75 167extern struct cgroup_subsys_state * const blkcg_root_css;
36558c8a 168
24f29046
TH
169struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg,
170 struct request_queue *q, bool update_hint);
3c798398
TH
171struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg,
172 struct request_queue *q);
36558c8a
TH
173int blkcg_init_queue(struct request_queue *q);
174void blkcg_drain_queue(struct request_queue *q);
175void blkcg_exit_queue(struct request_queue *q);
5efd6113 176
3e252066 177/* Blkio controller policy registration */
d5bf0291 178int blkcg_policy_register(struct blkcg_policy *pol);
3c798398 179void blkcg_policy_unregister(struct blkcg_policy *pol);
36558c8a 180int blkcg_activate_policy(struct request_queue *q,
3c798398 181 const struct blkcg_policy *pol);
36558c8a 182void blkcg_deactivate_policy(struct request_queue *q,
3c798398 183 const struct blkcg_policy *pol);
3e252066 184
3c798398 185void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
f95a04af
TH
186 u64 (*prfill)(struct seq_file *,
187 struct blkg_policy_data *, int),
3c798398 188 const struct blkcg_policy *pol, int data,
ec399347 189 bool show_total);
f95a04af
TH
190u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v);
191u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
829fdb50 192 const struct blkg_rwstat *rwstat);
f95a04af
TH
193u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off);
194u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
195 int off);
77ea7338
TH
196int blkg_print_stat_bytes(struct seq_file *sf, void *v);
197int blkg_print_stat_ios(struct seq_file *sf, void *v);
198int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v);
199int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v);
829fdb50 200
f12c74ca
TH
201u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg,
202 struct blkcg_policy *pol, int off);
203struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg,
204 struct blkcg_policy *pol, int off);
16b3de66 205
829fdb50 206struct blkg_conf_ctx {
36558c8a 207 struct gendisk *disk;
3c798398 208 struct blkcg_gq *blkg;
36558c8a 209 u64 v;
829fdb50
TH
210};
211
3c798398
TH
212int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
213 const char *input, struct blkg_conf_ctx *ctx);
829fdb50
TH
214void blkg_conf_finish(struct blkg_conf_ctx *ctx);
215
216
a7c6d554
TH
217static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css)
218{
219 return css ? container_of(css, struct blkcg, css) : NULL;
220}
221
b1208b56
TH
222static inline struct blkcg *task_blkcg(struct task_struct *tsk)
223{
c165b3e3 224 return css_to_blkcg(task_css(tsk, io_cgrp_id));
b1208b56
TH
225}
226
227static inline struct blkcg *bio_blkcg(struct bio *bio)
228{
229 if (bio && bio->bi_css)
a7c6d554 230 return css_to_blkcg(bio->bi_css);
b1208b56
TH
231 return task_blkcg(current);
232}
233
fd383c2d
TH
234static inline struct cgroup_subsys_state *
235task_get_blkcg_css(struct task_struct *task)
236{
c165b3e3 237 return task_get_css(task, 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{
e61734c5 342 char *p;
54e7ed12 343
e61734c5
TH
344 p = cgroup_path(blkg->blkcg->css.cgroup, buf, buflen);
345 if (!p) {
54e7ed12 346 strncpy(buf, "<unavailable>", buflen);
e61734c5
TH
347 return -ENAMETOOLONG;
348 }
349
350 memmove(buf, p, buf + buflen - p);
351 return 0;
afc24d49
VG
352}
353
1adaf3dd
TH
354/**
355 * blkg_get - get a blkg reference
356 * @blkg: blkg to get
357 *
a5049a8a 358 * The caller should be holding an existing reference.
1adaf3dd 359 */
3c798398 360static inline void blkg_get(struct blkcg_gq *blkg)
1adaf3dd 361{
a5049a8a
TH
362 WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0);
363 atomic_inc(&blkg->refcnt);
1adaf3dd
TH
364}
365
2a4fd070 366void __blkg_release_rcu(struct rcu_head *rcu);
1adaf3dd
TH
367
368/**
369 * blkg_put - put a blkg reference
370 * @blkg: blkg to put
1adaf3dd 371 */
3c798398 372static inline void blkg_put(struct blkcg_gq *blkg)
1adaf3dd 373{
a5049a8a
TH
374 WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0);
375 if (atomic_dec_and_test(&blkg->refcnt))
2a4fd070 376 call_rcu(&blkg->rcu_head, __blkg_release_rcu);
1adaf3dd
TH
377}
378
dd4a4ffc
TH
379/**
380 * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants
381 * @d_blkg: loop cursor pointing to the current descendant
492eb21b 382 * @pos_css: used for iteration
dd4a4ffc
TH
383 * @p_blkg: target blkg to walk descendants of
384 *
385 * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU
386 * read locked. If called under either blkcg or queue lock, the iteration
387 * is guaranteed to include all and only online blkgs. The caller may
492eb21b 388 * update @pos_css by calling css_rightmost_descendant() to skip subtree.
bd8815a6 389 * @p_blkg is included in the iteration and the first node to be visited.
dd4a4ffc 390 */
492eb21b
TH
391#define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \
392 css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \
393 if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \
dd4a4ffc
TH
394 (p_blkg)->q, false)))
395
aa539cb3
TH
396/**
397 * blkg_for_each_descendant_post - post-order walk of a blkg's descendants
398 * @d_blkg: loop cursor pointing to the current descendant
492eb21b 399 * @pos_css: used for iteration
aa539cb3
TH
400 * @p_blkg: target blkg to walk descendants of
401 *
402 * Similar to blkg_for_each_descendant_pre() but performs post-order
bd8815a6
TH
403 * traversal instead. Synchronization rules are the same. @p_blkg is
404 * included in the iteration and the last node to be visited.
aa539cb3 405 */
492eb21b
TH
406#define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \
407 css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \
408 if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \
aa539cb3
TH
409 (p_blkg)->q, false)))
410
a051661c
TH
411/**
412 * blk_get_rl - get request_list to use
413 * @q: request_queue of interest
414 * @bio: bio which will be attached to the allocated request (may be %NULL)
415 *
416 * The caller wants to allocate a request from @q to use for @bio. Find
417 * the request_list to use and obtain a reference on it. Should be called
418 * under queue_lock. This function is guaranteed to return non-%NULL
419 * request_list.
420 */
421static inline struct request_list *blk_get_rl(struct request_queue *q,
422 struct bio *bio)
423{
424 struct blkcg *blkcg;
425 struct blkcg_gq *blkg;
426
427 rcu_read_lock();
428
429 blkcg = bio_blkcg(bio);
430
431 /* bypass blkg lookup and use @q->root_rl directly for root */
432 if (blkcg == &blkcg_root)
433 goto root_rl;
434
435 /*
436 * Try to use blkg->rl. blkg lookup may fail under memory pressure
437 * or if either the blkcg or queue is going away. Fall back to
438 * root_rl in such cases.
439 */
ae118896
TH
440 blkg = blkg_lookup(blkcg, q);
441 if (unlikely(!blkg))
a051661c
TH
442 goto root_rl;
443
444 blkg_get(blkg);
445 rcu_read_unlock();
446 return &blkg->rl;
447root_rl:
448 rcu_read_unlock();
449 return &q->root_rl;
450}
451
452/**
453 * blk_put_rl - put request_list
454 * @rl: request_list to put
455 *
456 * Put the reference acquired by blk_get_rl(). Should be called under
457 * queue_lock.
458 */
459static inline void blk_put_rl(struct request_list *rl)
460{
401efbf8 461 if (rl->blkg->blkcg != &blkcg_root)
a051661c
TH
462 blkg_put(rl->blkg);
463}
464
465/**
466 * blk_rq_set_rl - associate a request with a request_list
467 * @rq: request of interest
468 * @rl: target request_list
469 *
470 * Associate @rq with @rl so that accounting and freeing can know the
471 * request_list @rq came from.
472 */
473static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl)
474{
475 rq->rl = rl;
476}
477
478/**
479 * blk_rq_rl - return the request_list a request came from
480 * @rq: request of interest
481 *
482 * Return the request_list @rq is allocated from.
483 */
484static inline struct request_list *blk_rq_rl(struct request *rq)
485{
486 return rq->rl;
487}
488
489struct request_list *__blk_queue_next_rl(struct request_list *rl,
490 struct request_queue *q);
491/**
492 * blk_queue_for_each_rl - iterate through all request_lists of a request_queue
493 *
494 * Should be used under queue_lock.
495 */
496#define blk_queue_for_each_rl(rl, q) \
497 for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q)))
498
24bdb8ef 499static inline int blkg_stat_init(struct blkg_stat *stat, gfp_t gfp)
90d3839b 500{
24bdb8ef
TH
501 int ret;
502
503 ret = percpu_counter_init(&stat->cpu_cnt, 0, gfp);
504 if (ret)
505 return ret;
506
e6269c44 507 atomic64_set(&stat->aux_cnt, 0);
24bdb8ef
TH
508 return 0;
509}
510
511static inline void blkg_stat_exit(struct blkg_stat *stat)
512{
513 percpu_counter_destroy(&stat->cpu_cnt);
90d3839b
PZ
514}
515
edcb0722
TH
516/**
517 * blkg_stat_add - add a value to a blkg_stat
518 * @stat: target blkg_stat
519 * @val: value to add
520 *
24bdb8ef
TH
521 * Add @val to @stat. The caller must ensure that IRQ on the same CPU
522 * don't re-enter this function for the same counter.
edcb0722
TH
523 */
524static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val)
525{
24bdb8ef 526 __percpu_counter_add(&stat->cpu_cnt, val, BLKG_STAT_CPU_BATCH);
edcb0722
TH
527}
528
529/**
530 * blkg_stat_read - read the current value of a blkg_stat
531 * @stat: blkg_stat to read
edcb0722
TH
532 */
533static inline uint64_t blkg_stat_read(struct blkg_stat *stat)
534{
24bdb8ef 535 return percpu_counter_sum_positive(&stat->cpu_cnt);
edcb0722
TH
536}
537
538/**
539 * blkg_stat_reset - reset a blkg_stat
540 * @stat: blkg_stat to reset
541 */
542static inline void blkg_stat_reset(struct blkg_stat *stat)
543{
24bdb8ef 544 percpu_counter_set(&stat->cpu_cnt, 0);
e6269c44 545 atomic64_set(&stat->aux_cnt, 0);
edcb0722
TH
546}
547
16b3de66 548/**
e6269c44 549 * blkg_stat_add_aux - add a blkg_stat into another's aux count
16b3de66
TH
550 * @to: the destination blkg_stat
551 * @from: the source
552 *
e6269c44 553 * Add @from's count including the aux one to @to's aux count.
16b3de66 554 */
e6269c44
TH
555static inline void blkg_stat_add_aux(struct blkg_stat *to,
556 struct blkg_stat *from)
16b3de66 557{
e6269c44
TH
558 atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt),
559 &to->aux_cnt);
16b3de66
TH
560}
561
24bdb8ef 562static inline int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp)
90d3839b 563{
24bdb8ef
TH
564 int i, ret;
565
566 for (i = 0; i < BLKG_RWSTAT_NR; i++) {
567 ret = percpu_counter_init(&rwstat->cpu_cnt[i], 0, gfp);
568 if (ret) {
569 while (--i >= 0)
570 percpu_counter_destroy(&rwstat->cpu_cnt[i]);
571 return ret;
572 }
573 atomic64_set(&rwstat->aux_cnt[i], 0);
574 }
575 return 0;
576}
e6269c44 577
24bdb8ef
TH
578static inline void blkg_rwstat_exit(struct blkg_rwstat *rwstat)
579{
580 int i;
e6269c44
TH
581
582 for (i = 0; i < BLKG_RWSTAT_NR; i++)
24bdb8ef 583 percpu_counter_destroy(&rwstat->cpu_cnt[i]);
90d3839b
PZ
584}
585
edcb0722
TH
586/**
587 * blkg_rwstat_add - add a value to a blkg_rwstat
588 * @rwstat: target blkg_rwstat
589 * @rw: mask of REQ_{WRITE|SYNC}
590 * @val: value to add
591 *
592 * Add @val to @rwstat. The counters are chosen according to @rw. The
593 * caller is responsible for synchronizing calls to this function.
594 */
595static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat,
596 int rw, uint64_t val)
597{
24bdb8ef 598 struct percpu_counter *cnt;
edcb0722
TH
599
600 if (rw & REQ_WRITE)
24bdb8ef 601 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_WRITE];
edcb0722 602 else
24bdb8ef
TH
603 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_READ];
604
605 __percpu_counter_add(cnt, val, BLKG_STAT_CPU_BATCH);
606
edcb0722 607 if (rw & REQ_SYNC)
24bdb8ef 608 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_SYNC];
edcb0722 609 else
24bdb8ef 610 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_ASYNC];
edcb0722 611
24bdb8ef 612 __percpu_counter_add(cnt, val, BLKG_STAT_CPU_BATCH);
edcb0722
TH
613}
614
615/**
616 * blkg_rwstat_read - read the current values of a blkg_rwstat
617 * @rwstat: blkg_rwstat to read
618 *
24bdb8ef 619 * Read the current snapshot of @rwstat and return it in the aux counts.
edcb0722 620 */
c94bed89 621static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat)
edcb0722 622{
24bdb8ef
TH
623 struct blkg_rwstat result;
624 int i;
edcb0722 625
24bdb8ef
TH
626 for (i = 0; i < BLKG_RWSTAT_NR; i++)
627 atomic64_set(&result.aux_cnt[i],
628 percpu_counter_sum_positive(&rwstat->cpu_cnt[i]));
629 return result;
edcb0722
TH
630}
631
632/**
4d5e80a7 633 * blkg_rwstat_total - read the total count of a blkg_rwstat
edcb0722
TH
634 * @rwstat: blkg_rwstat to read
635 *
636 * Return the total count of @rwstat regardless of the IO direction. This
637 * function can be called without synchronization and takes care of u64
638 * atomicity.
639 */
4d5e80a7 640static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat)
edcb0722
TH
641{
642 struct blkg_rwstat tmp = blkg_rwstat_read(rwstat);
643
24bdb8ef
TH
644 return atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) +
645 atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]);
edcb0722
TH
646}
647
648/**
649 * blkg_rwstat_reset - reset a blkg_rwstat
650 * @rwstat: blkg_rwstat to reset
651 */
652static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat)
653{
e6269c44
TH
654 int i;
655
24bdb8ef
TH
656 for (i = 0; i < BLKG_RWSTAT_NR; i++) {
657 percpu_counter_set(&rwstat->cpu_cnt[i], 0);
e6269c44 658 atomic64_set(&rwstat->aux_cnt[i], 0);
24bdb8ef 659 }
edcb0722
TH
660}
661
16b3de66 662/**
e6269c44 663 * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count
16b3de66
TH
664 * @to: the destination blkg_rwstat
665 * @from: the source
666 *
e6269c44 667 * Add @from's count including the aux one to @to's aux count.
16b3de66 668 */
e6269c44
TH
669static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to,
670 struct blkg_rwstat *from)
16b3de66
TH
671{
672 struct blkg_rwstat v = blkg_rwstat_read(from);
673 int i;
674
16b3de66 675 for (i = 0; i < BLKG_RWSTAT_NR; i++)
24bdb8ef
TH
676 atomic64_add(atomic64_read(&v.aux_cnt[i]) +
677 atomic64_read(&from->aux_cnt[i]),
e6269c44 678 &to->aux_cnt[i]);
16b3de66
TH
679}
680
ae118896
TH
681#ifdef CONFIG_BLK_DEV_THROTTLING
682extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
683 struct bio *bio);
684#else
685static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
686 struct bio *bio) { return false; }
687#endif
688
689static inline bool blkcg_bio_issue_check(struct request_queue *q,
690 struct bio *bio)
691{
692 struct blkcg *blkcg;
693 struct blkcg_gq *blkg;
694 bool throtl = false;
695
696 rcu_read_lock();
697 blkcg = bio_blkcg(bio);
698
699 blkg = blkg_lookup(blkcg, q);
700 if (unlikely(!blkg)) {
701 spin_lock_irq(q->queue_lock);
702 blkg = blkg_lookup_create(blkcg, q);
703 if (IS_ERR(blkg))
704 blkg = NULL;
705 spin_unlock_irq(q->queue_lock);
706 }
707
708 throtl = blk_throtl_bio(q, blkg, bio);
709
77ea7338
TH
710 if (!throtl) {
711 blkg = blkg ?: q->root_blkg;
712 blkg_rwstat_add(&blkg->stat_bytes, bio->bi_flags,
713 bio->bi_iter.bi_size);
714 blkg_rwstat_add(&blkg->stat_ios, bio->bi_flags, 1);
715 }
716
ae118896
TH
717 rcu_read_unlock();
718 return !throtl;
719}
720
36558c8a
TH
721#else /* CONFIG_BLK_CGROUP */
722
efa7d1c7
TH
723struct blkcg {
724};
2f5ea477 725
f95a04af
TH
726struct blkg_policy_data {
727};
728
e48453c3
AA
729struct blkcg_policy_data {
730};
731
3c798398 732struct blkcg_gq {
2f5ea477
JA
733};
734
3c798398 735struct blkcg_policy {
3e252066
VG
736};
737
496d5e75
TH
738#define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL))
739
fd383c2d
TH
740static inline struct cgroup_subsys_state *
741task_get_blkcg_css(struct task_struct *task)
742{
743 return NULL;
744}
745
efa7d1c7
TH
746#ifdef CONFIG_BLOCK
747
3c798398 748static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; }
5efd6113
TH
749static inline int blkcg_init_queue(struct request_queue *q) { return 0; }
750static inline void blkcg_drain_queue(struct request_queue *q) { }
751static inline void blkcg_exit_queue(struct request_queue *q) { }
d5bf0291 752static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; }
3c798398 753static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { }
a2b1693b 754static inline int blkcg_activate_policy(struct request_queue *q,
3c798398 755 const struct blkcg_policy *pol) { return 0; }
a2b1693b 756static inline void blkcg_deactivate_policy(struct request_queue *q,
3c798398
TH
757 const struct blkcg_policy *pol) { }
758
b1208b56 759static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; }
a051661c 760
f95a04af
TH
761static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
762 struct blkcg_policy *pol) { return NULL; }
763static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; }
3c798398
TH
764static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; }
765static inline void blkg_get(struct blkcg_gq *blkg) { }
766static inline void blkg_put(struct blkcg_gq *blkg) { }
afc24d49 767
a051661c
TH
768static inline struct request_list *blk_get_rl(struct request_queue *q,
769 struct bio *bio) { return &q->root_rl; }
770static inline void blk_put_rl(struct request_list *rl) { }
771static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { }
772static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; }
773
ae118896
TH
774static inline bool blkcg_bio_issue_check(struct request_queue *q,
775 struct bio *bio) { return true; }
776
a051661c
TH
777#define blk_queue_for_each_rl(rl, q) \
778 for ((rl) = &(q)->root_rl; (rl); (rl) = NULL)
779
efa7d1c7 780#endif /* CONFIG_BLOCK */
36558c8a
TH
781#endif /* CONFIG_BLK_CGROUP */
782#endif /* _BLK_CGROUP_H */