]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blame - include/linux/blk-cgroup.h
blkcg: misc preparations for unified hierarchy interface
[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 150 /* cgroup files for the policy */
880f50e2 151 struct cftype *legacy_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
dd165eb3 185const char *blkg_dev_name(struct blkcg_gq *blkg);
3c798398 186void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg,
f95a04af
TH
187 u64 (*prfill)(struct seq_file *,
188 struct blkg_policy_data *, int),
3c798398 189 const struct blkcg_policy *pol, int data,
ec399347 190 bool show_total);
f95a04af
TH
191u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v);
192u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
829fdb50 193 const struct blkg_rwstat *rwstat);
f95a04af
TH
194u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off);
195u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd,
196 int off);
77ea7338
TH
197int blkg_print_stat_bytes(struct seq_file *sf, void *v);
198int blkg_print_stat_ios(struct seq_file *sf, void *v);
199int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v);
200int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v);
829fdb50 201
f12c74ca
TH
202u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg,
203 struct blkcg_policy *pol, int off);
204struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg,
205 struct blkcg_policy *pol, int off);
16b3de66 206
829fdb50 207struct blkg_conf_ctx {
36558c8a 208 struct gendisk *disk;
3c798398 209 struct blkcg_gq *blkg;
36aa9e5f 210 char *body;
829fdb50
TH
211};
212
3c798398 213int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol,
36aa9e5f 214 char *input, struct blkg_conf_ctx *ctx);
829fdb50
TH
215void blkg_conf_finish(struct blkg_conf_ctx *ctx);
216
217
a7c6d554
TH
218static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css)
219{
220 return css ? container_of(css, struct blkcg, css) : NULL;
221}
222
b1208b56
TH
223static inline struct blkcg *task_blkcg(struct task_struct *tsk)
224{
c165b3e3 225 return css_to_blkcg(task_css(tsk, io_cgrp_id));
b1208b56
TH
226}
227
228static inline struct blkcg *bio_blkcg(struct bio *bio)
229{
230 if (bio && bio->bi_css)
a7c6d554 231 return css_to_blkcg(bio->bi_css);
b1208b56
TH
232 return task_blkcg(current);
233}
234
fd383c2d
TH
235static inline struct cgroup_subsys_state *
236task_get_blkcg_css(struct task_struct *task)
237{
c165b3e3 238 return task_get_css(task, io_cgrp_id);
fd383c2d
TH
239}
240
3c547865
TH
241/**
242 * blkcg_parent - get the parent of a blkcg
243 * @blkcg: blkcg of interest
244 *
245 * Return the parent blkcg of @blkcg. Can be called anytime.
246 */
247static inline struct blkcg *blkcg_parent(struct blkcg *blkcg)
248{
5c9d535b 249 return css_to_blkcg(blkcg->css.parent);
3c547865
TH
250}
251
24f29046
TH
252/**
253 * __blkg_lookup - internal version of blkg_lookup()
254 * @blkcg: blkcg of interest
255 * @q: request_queue of interest
256 * @update_hint: whether to update lookup hint with the result or not
257 *
258 * This is internal version and shouldn't be used by policy
259 * implementations. Looks up blkgs for the @blkcg - @q pair regardless of
260 * @q's bypass state. If @update_hint is %true, the caller should be
261 * holding @q->queue_lock and lookup hint is updated on success.
262 */
263static inline struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg,
264 struct request_queue *q,
265 bool update_hint)
266{
267 struct blkcg_gq *blkg;
268
85b6bc9d
TH
269 if (blkcg == &blkcg_root)
270 return q->root_blkg;
271
24f29046
TH
272 blkg = rcu_dereference(blkcg->blkg_hint);
273 if (blkg && blkg->q == q)
274 return blkg;
275
276 return blkg_lookup_slowpath(blkcg, q, update_hint);
277}
278
279/**
280 * blkg_lookup - lookup blkg for the specified blkcg - q pair
281 * @blkcg: blkcg of interest
282 * @q: request_queue of interest
283 *
284 * Lookup blkg for the @blkcg - @q pair. This function should be called
285 * under RCU read lock and is guaranteed to return %NULL if @q is bypassing
286 * - see blk_queue_bypass_start() for details.
287 */
288static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg,
289 struct request_queue *q)
290{
291 WARN_ON_ONCE(!rcu_read_lock_held());
292
293 if (unlikely(blk_queue_bypass(q)))
294 return NULL;
295 return __blkg_lookup(blkcg, q, false);
296}
297
0381411e
TH
298/**
299 * blkg_to_pdata - get policy private data
300 * @blkg: blkg of interest
301 * @pol: policy of interest
302 *
303 * Return pointer to private data associated with the @blkg-@pol pair.
304 */
f95a04af
TH
305static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
306 struct blkcg_policy *pol)
0381411e 307{
f95a04af 308 return blkg ? blkg->pd[pol->plid] : NULL;
0381411e
TH
309}
310
e48453c3
AA
311static inline struct blkcg_policy_data *blkcg_to_cpd(struct blkcg *blkcg,
312 struct blkcg_policy *pol)
313{
81437648 314 return blkcg ? blkcg->cpd[pol->plid] : NULL;
e48453c3
AA
315}
316
0381411e
TH
317/**
318 * pdata_to_blkg - get blkg associated with policy private data
f95a04af 319 * @pd: policy private data of interest
0381411e 320 *
f95a04af 321 * @pd is policy private data. Determine the blkg it's associated with.
0381411e 322 */
f95a04af 323static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd)
0381411e 324{
f95a04af 325 return pd ? pd->blkg : NULL;
0381411e
TH
326}
327
81437648
TH
328static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd)
329{
330 return cpd ? cpd->blkcg : NULL;
331}
332
54e7ed12
TH
333/**
334 * blkg_path - format cgroup path of blkg
335 * @blkg: blkg of interest
336 * @buf: target buffer
337 * @buflen: target buffer length
338 *
339 * Format the path of the cgroup of @blkg into @buf.
340 */
3c798398 341static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen)
afc24d49 342{
e61734c5 343 char *p;
54e7ed12 344
e61734c5
TH
345 p = cgroup_path(blkg->blkcg->css.cgroup, buf, buflen);
346 if (!p) {
54e7ed12 347 strncpy(buf, "<unavailable>", buflen);
e61734c5
TH
348 return -ENAMETOOLONG;
349 }
350
351 memmove(buf, p, buf + buflen - p);
352 return 0;
afc24d49
VG
353}
354
1adaf3dd
TH
355/**
356 * blkg_get - get a blkg reference
357 * @blkg: blkg to get
358 *
a5049a8a 359 * The caller should be holding an existing reference.
1adaf3dd 360 */
3c798398 361static inline void blkg_get(struct blkcg_gq *blkg)
1adaf3dd 362{
a5049a8a
TH
363 WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0);
364 atomic_inc(&blkg->refcnt);
1adaf3dd
TH
365}
366
2a4fd070 367void __blkg_release_rcu(struct rcu_head *rcu);
1adaf3dd
TH
368
369/**
370 * blkg_put - put a blkg reference
371 * @blkg: blkg to put
1adaf3dd 372 */
3c798398 373static inline void blkg_put(struct blkcg_gq *blkg)
1adaf3dd 374{
a5049a8a
TH
375 WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0);
376 if (atomic_dec_and_test(&blkg->refcnt))
2a4fd070 377 call_rcu(&blkg->rcu_head, __blkg_release_rcu);
1adaf3dd
TH
378}
379
dd4a4ffc
TH
380/**
381 * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants
382 * @d_blkg: loop cursor pointing to the current descendant
492eb21b 383 * @pos_css: used for iteration
dd4a4ffc
TH
384 * @p_blkg: target blkg to walk descendants of
385 *
386 * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU
387 * read locked. If called under either blkcg or queue lock, the iteration
388 * is guaranteed to include all and only online blkgs. The caller may
492eb21b 389 * update @pos_css by calling css_rightmost_descendant() to skip subtree.
bd8815a6 390 * @p_blkg is included in the iteration and the first node to be visited.
dd4a4ffc 391 */
492eb21b
TH
392#define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \
393 css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \
394 if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \
dd4a4ffc
TH
395 (p_blkg)->q, false)))
396
aa539cb3
TH
397/**
398 * blkg_for_each_descendant_post - post-order walk of a blkg's descendants
399 * @d_blkg: loop cursor pointing to the current descendant
492eb21b 400 * @pos_css: used for iteration
aa539cb3
TH
401 * @p_blkg: target blkg to walk descendants of
402 *
403 * Similar to blkg_for_each_descendant_pre() but performs post-order
bd8815a6
TH
404 * traversal instead. Synchronization rules are the same. @p_blkg is
405 * included in the iteration and the last node to be visited.
aa539cb3 406 */
492eb21b
TH
407#define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \
408 css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \
409 if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \
aa539cb3
TH
410 (p_blkg)->q, false)))
411
a051661c
TH
412/**
413 * blk_get_rl - get request_list to use
414 * @q: request_queue of interest
415 * @bio: bio which will be attached to the allocated request (may be %NULL)
416 *
417 * The caller wants to allocate a request from @q to use for @bio. Find
418 * the request_list to use and obtain a reference on it. Should be called
419 * under queue_lock. This function is guaranteed to return non-%NULL
420 * request_list.
421 */
422static inline struct request_list *blk_get_rl(struct request_queue *q,
423 struct bio *bio)
424{
425 struct blkcg *blkcg;
426 struct blkcg_gq *blkg;
427
428 rcu_read_lock();
429
430 blkcg = bio_blkcg(bio);
431
432 /* bypass blkg lookup and use @q->root_rl directly for root */
433 if (blkcg == &blkcg_root)
434 goto root_rl;
435
436 /*
437 * Try to use blkg->rl. blkg lookup may fail under memory pressure
438 * or if either the blkcg or queue is going away. Fall back to
439 * root_rl in such cases.
440 */
ae118896
TH
441 blkg = blkg_lookup(blkcg, q);
442 if (unlikely(!blkg))
a051661c
TH
443 goto root_rl;
444
445 blkg_get(blkg);
446 rcu_read_unlock();
447 return &blkg->rl;
448root_rl:
449 rcu_read_unlock();
450 return &q->root_rl;
451}
452
453/**
454 * blk_put_rl - put request_list
455 * @rl: request_list to put
456 *
457 * Put the reference acquired by blk_get_rl(). Should be called under
458 * queue_lock.
459 */
460static inline void blk_put_rl(struct request_list *rl)
461{
401efbf8 462 if (rl->blkg->blkcg != &blkcg_root)
a051661c
TH
463 blkg_put(rl->blkg);
464}
465
466/**
467 * blk_rq_set_rl - associate a request with a request_list
468 * @rq: request of interest
469 * @rl: target request_list
470 *
471 * Associate @rq with @rl so that accounting and freeing can know the
472 * request_list @rq came from.
473 */
474static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl)
475{
476 rq->rl = rl;
477}
478
479/**
480 * blk_rq_rl - return the request_list a request came from
481 * @rq: request of interest
482 *
483 * Return the request_list @rq is allocated from.
484 */
485static inline struct request_list *blk_rq_rl(struct request *rq)
486{
487 return rq->rl;
488}
489
490struct request_list *__blk_queue_next_rl(struct request_list *rl,
491 struct request_queue *q);
492/**
493 * blk_queue_for_each_rl - iterate through all request_lists of a request_queue
494 *
495 * Should be used under queue_lock.
496 */
497#define blk_queue_for_each_rl(rl, q) \
498 for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q)))
499
24bdb8ef 500static inline int blkg_stat_init(struct blkg_stat *stat, gfp_t gfp)
90d3839b 501{
24bdb8ef
TH
502 int ret;
503
504 ret = percpu_counter_init(&stat->cpu_cnt, 0, gfp);
505 if (ret)
506 return ret;
507
e6269c44 508 atomic64_set(&stat->aux_cnt, 0);
24bdb8ef
TH
509 return 0;
510}
511
512static inline void blkg_stat_exit(struct blkg_stat *stat)
513{
514 percpu_counter_destroy(&stat->cpu_cnt);
90d3839b
PZ
515}
516
edcb0722
TH
517/**
518 * blkg_stat_add - add a value to a blkg_stat
519 * @stat: target blkg_stat
520 * @val: value to add
521 *
24bdb8ef
TH
522 * Add @val to @stat. The caller must ensure that IRQ on the same CPU
523 * don't re-enter this function for the same counter.
edcb0722
TH
524 */
525static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val)
526{
24bdb8ef 527 __percpu_counter_add(&stat->cpu_cnt, val, BLKG_STAT_CPU_BATCH);
edcb0722
TH
528}
529
530/**
531 * blkg_stat_read - read the current value of a blkg_stat
532 * @stat: blkg_stat to read
edcb0722
TH
533 */
534static inline uint64_t blkg_stat_read(struct blkg_stat *stat)
535{
24bdb8ef 536 return percpu_counter_sum_positive(&stat->cpu_cnt);
edcb0722
TH
537}
538
539/**
540 * blkg_stat_reset - reset a blkg_stat
541 * @stat: blkg_stat to reset
542 */
543static inline void blkg_stat_reset(struct blkg_stat *stat)
544{
24bdb8ef 545 percpu_counter_set(&stat->cpu_cnt, 0);
e6269c44 546 atomic64_set(&stat->aux_cnt, 0);
edcb0722
TH
547}
548
16b3de66 549/**
e6269c44 550 * blkg_stat_add_aux - add a blkg_stat into another's aux count
16b3de66
TH
551 * @to: the destination blkg_stat
552 * @from: the source
553 *
e6269c44 554 * Add @from's count including the aux one to @to's aux count.
16b3de66 555 */
e6269c44
TH
556static inline void blkg_stat_add_aux(struct blkg_stat *to,
557 struct blkg_stat *from)
16b3de66 558{
e6269c44
TH
559 atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt),
560 &to->aux_cnt);
16b3de66
TH
561}
562
24bdb8ef 563static inline int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp)
90d3839b 564{
24bdb8ef
TH
565 int i, ret;
566
567 for (i = 0; i < BLKG_RWSTAT_NR; i++) {
568 ret = percpu_counter_init(&rwstat->cpu_cnt[i], 0, gfp);
569 if (ret) {
570 while (--i >= 0)
571 percpu_counter_destroy(&rwstat->cpu_cnt[i]);
572 return ret;
573 }
574 atomic64_set(&rwstat->aux_cnt[i], 0);
575 }
576 return 0;
577}
e6269c44 578
24bdb8ef
TH
579static inline void blkg_rwstat_exit(struct blkg_rwstat *rwstat)
580{
581 int i;
e6269c44
TH
582
583 for (i = 0; i < BLKG_RWSTAT_NR; i++)
24bdb8ef 584 percpu_counter_destroy(&rwstat->cpu_cnt[i]);
90d3839b
PZ
585}
586
edcb0722
TH
587/**
588 * blkg_rwstat_add - add a value to a blkg_rwstat
589 * @rwstat: target blkg_rwstat
590 * @rw: mask of REQ_{WRITE|SYNC}
591 * @val: value to add
592 *
593 * Add @val to @rwstat. The counters are chosen according to @rw. The
594 * caller is responsible for synchronizing calls to this function.
595 */
596static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat,
597 int rw, uint64_t val)
598{
24bdb8ef 599 struct percpu_counter *cnt;
edcb0722
TH
600
601 if (rw & REQ_WRITE)
24bdb8ef 602 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_WRITE];
edcb0722 603 else
24bdb8ef
TH
604 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_READ];
605
606 __percpu_counter_add(cnt, val, BLKG_STAT_CPU_BATCH);
607
edcb0722 608 if (rw & REQ_SYNC)
24bdb8ef 609 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_SYNC];
edcb0722 610 else
24bdb8ef 611 cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_ASYNC];
edcb0722 612
24bdb8ef 613 __percpu_counter_add(cnt, val, BLKG_STAT_CPU_BATCH);
edcb0722
TH
614}
615
616/**
617 * blkg_rwstat_read - read the current values of a blkg_rwstat
618 * @rwstat: blkg_rwstat to read
619 *
24bdb8ef 620 * Read the current snapshot of @rwstat and return it in the aux counts.
edcb0722 621 */
c94bed89 622static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat)
edcb0722 623{
24bdb8ef
TH
624 struct blkg_rwstat result;
625 int i;
edcb0722 626
24bdb8ef
TH
627 for (i = 0; i < BLKG_RWSTAT_NR; i++)
628 atomic64_set(&result.aux_cnt[i],
629 percpu_counter_sum_positive(&rwstat->cpu_cnt[i]));
630 return result;
edcb0722
TH
631}
632
633/**
4d5e80a7 634 * blkg_rwstat_total - read the total count of a blkg_rwstat
edcb0722
TH
635 * @rwstat: blkg_rwstat to read
636 *
637 * Return the total count of @rwstat regardless of the IO direction. This
638 * function can be called without synchronization and takes care of u64
639 * atomicity.
640 */
4d5e80a7 641static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat)
edcb0722
TH
642{
643 struct blkg_rwstat tmp = blkg_rwstat_read(rwstat);
644
24bdb8ef
TH
645 return atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) +
646 atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]);
edcb0722
TH
647}
648
649/**
650 * blkg_rwstat_reset - reset a blkg_rwstat
651 * @rwstat: blkg_rwstat to reset
652 */
653static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat)
654{
e6269c44
TH
655 int i;
656
24bdb8ef
TH
657 for (i = 0; i < BLKG_RWSTAT_NR; i++) {
658 percpu_counter_set(&rwstat->cpu_cnt[i], 0);
e6269c44 659 atomic64_set(&rwstat->aux_cnt[i], 0);
24bdb8ef 660 }
edcb0722
TH
661}
662
16b3de66 663/**
e6269c44 664 * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count
16b3de66
TH
665 * @to: the destination blkg_rwstat
666 * @from: the source
667 *
e6269c44 668 * Add @from's count including the aux one to @to's aux count.
16b3de66 669 */
e6269c44
TH
670static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to,
671 struct blkg_rwstat *from)
16b3de66
TH
672{
673 struct blkg_rwstat v = blkg_rwstat_read(from);
674 int i;
675
16b3de66 676 for (i = 0; i < BLKG_RWSTAT_NR; i++)
24bdb8ef
TH
677 atomic64_add(atomic64_read(&v.aux_cnt[i]) +
678 atomic64_read(&from->aux_cnt[i]),
e6269c44 679 &to->aux_cnt[i]);
16b3de66
TH
680}
681
ae118896
TH
682#ifdef CONFIG_BLK_DEV_THROTTLING
683extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
684 struct bio *bio);
685#else
686static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg,
687 struct bio *bio) { return false; }
688#endif
689
690static inline bool blkcg_bio_issue_check(struct request_queue *q,
691 struct bio *bio)
692{
693 struct blkcg *blkcg;
694 struct blkcg_gq *blkg;
695 bool throtl = false;
696
697 rcu_read_lock();
698 blkcg = bio_blkcg(bio);
699
700 blkg = blkg_lookup(blkcg, q);
701 if (unlikely(!blkg)) {
702 spin_lock_irq(q->queue_lock);
703 blkg = blkg_lookup_create(blkcg, q);
704 if (IS_ERR(blkg))
705 blkg = NULL;
706 spin_unlock_irq(q->queue_lock);
707 }
708
709 throtl = blk_throtl_bio(q, blkg, bio);
710
77ea7338
TH
711 if (!throtl) {
712 blkg = blkg ?: q->root_blkg;
713 blkg_rwstat_add(&blkg->stat_bytes, bio->bi_flags,
714 bio->bi_iter.bi_size);
715 blkg_rwstat_add(&blkg->stat_ios, bio->bi_flags, 1);
716 }
717
ae118896
TH
718 rcu_read_unlock();
719 return !throtl;
720}
721
36558c8a
TH
722#else /* CONFIG_BLK_CGROUP */
723
efa7d1c7
TH
724struct blkcg {
725};
2f5ea477 726
f95a04af
TH
727struct blkg_policy_data {
728};
729
e48453c3
AA
730struct blkcg_policy_data {
731};
732
3c798398 733struct blkcg_gq {
2f5ea477
JA
734};
735
3c798398 736struct blkcg_policy {
3e252066
VG
737};
738
496d5e75
TH
739#define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL))
740
fd383c2d
TH
741static inline struct cgroup_subsys_state *
742task_get_blkcg_css(struct task_struct *task)
743{
744 return NULL;
745}
746
efa7d1c7
TH
747#ifdef CONFIG_BLOCK
748
3c798398 749static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; }
5efd6113
TH
750static inline int blkcg_init_queue(struct request_queue *q) { return 0; }
751static inline void blkcg_drain_queue(struct request_queue *q) { }
752static inline void blkcg_exit_queue(struct request_queue *q) { }
d5bf0291 753static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; }
3c798398 754static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { }
a2b1693b 755static inline int blkcg_activate_policy(struct request_queue *q,
3c798398 756 const struct blkcg_policy *pol) { return 0; }
a2b1693b 757static inline void blkcg_deactivate_policy(struct request_queue *q,
3c798398
TH
758 const struct blkcg_policy *pol) { }
759
b1208b56 760static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; }
a051661c 761
f95a04af
TH
762static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg,
763 struct blkcg_policy *pol) { return NULL; }
764static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; }
3c798398
TH
765static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; }
766static inline void blkg_get(struct blkcg_gq *blkg) { }
767static inline void blkg_put(struct blkcg_gq *blkg) { }
afc24d49 768
a051661c
TH
769static inline struct request_list *blk_get_rl(struct request_queue *q,
770 struct bio *bio) { return &q->root_rl; }
771static inline void blk_put_rl(struct request_list *rl) { }
772static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { }
773static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; }
774
ae118896
TH
775static inline bool blkcg_bio_issue_check(struct request_queue *q,
776 struct bio *bio) { return true; }
777
a051661c
TH
778#define blk_queue_for_each_rl(rl, q) \
779 for ((rl) = &(q)->root_rl; (rl); (rl) = NULL)
780
efa7d1c7 781#endif /* CONFIG_BLOCK */
36558c8a
TH
782#endif /* CONFIG_BLK_CGROUP */
783#endif /* _BLK_CGROUP_H */