Commit | Line | Data |
---|---|---|
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 |
33 | enum 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 |
43 | struct blkcg_gq; |
44 | ||
3c798398 | 45 | struct 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 | 66 | struct blkg_stat { |
24bdb8ef | 67 | struct percpu_counter cpu_cnt; |
e6269c44 | 68 | atomic64_t aux_cnt; |
edcb0722 TH |
69 | }; |
70 | ||
71 | struct 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 | 87 | struct 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; |
4c699480 | 91 | bool offline; |
0381411e TH |
92 | }; |
93 | ||
e48453c3 | 94 | /* |
e4a9bde9 TH |
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. | |
e48453c3 AA |
100 | */ |
101 | struct blkcg_policy_data { | |
81437648 TH |
102 | /* the blkcg and policy id this per-policy data belongs to */ |
103 | struct blkcg *blkcg; | |
e48453c3 | 104 | int plid; |
e48453c3 AA |
105 | }; |
106 | ||
3c798398 TH |
107 | /* association between a blk cgroup and a request queue */ |
108 | struct blkcg_gq { | |
c875f4d0 | 109 | /* Pointer to the associated request_queue */ |
36558c8a TH |
110 | struct request_queue *q; |
111 | struct list_head q_node; | |
112 | struct hlist_node blkcg_node; | |
3c798398 | 113 | struct blkcg *blkcg; |
3c547865 | 114 | |
ce7acfea TH |
115 | /* |
116 | * Each blkg gets congested separately and the congestion state is | |
117 | * propagated to the matching bdi_writeback_congested. | |
118 | */ | |
119 | struct bdi_writeback_congested *wb_congested; | |
120 | ||
3c547865 TH |
121 | /* all non-root blkcg_gq's are guaranteed to have access to parent */ |
122 | struct blkcg_gq *parent; | |
123 | ||
a051661c TH |
124 | /* request allocation list for this blkcg-q pair */ |
125 | struct request_list rl; | |
3c547865 | 126 | |
1adaf3dd | 127 | /* reference count */ |
a5049a8a | 128 | atomic_t refcnt; |
22084190 | 129 | |
f427d909 TH |
130 | /* is this blkg online? protected by both blkcg and q locks */ |
131 | bool online; | |
132 | ||
77ea7338 TH |
133 | struct blkg_rwstat stat_bytes; |
134 | struct blkg_rwstat stat_ios; | |
135 | ||
36558c8a | 136 | struct blkg_policy_data *pd[BLKCG_MAX_POLS]; |
1adaf3dd | 137 | |
36558c8a | 138 | struct rcu_head rcu_head; |
31e4c28d VG |
139 | }; |
140 | ||
e4a9bde9 | 141 | typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp); |
81437648 | 142 | typedef void (blkcg_pol_init_cpd_fn)(struct blkcg_policy_data *cpd); |
e4a9bde9 | 143 | typedef void (blkcg_pol_free_cpd_fn)(struct blkcg_policy_data *cpd); |
69d7fde5 | 144 | typedef void (blkcg_pol_bind_cpd_fn)(struct blkcg_policy_data *cpd); |
001bea73 | 145 | typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node); |
a9520cd6 TH |
146 | typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd); |
147 | typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd); | |
148 | typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd); | |
001bea73 | 149 | typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd); |
a9520cd6 | 150 | typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd); |
3e252066 | 151 | |
3c798398 | 152 | struct blkcg_policy { |
36558c8a | 153 | int plid; |
36558c8a | 154 | /* cgroup files for the policy */ |
2ee867dc | 155 | struct cftype *dfl_cftypes; |
880f50e2 | 156 | struct cftype *legacy_cftypes; |
f9fcc2d3 TH |
157 | |
158 | /* operations */ | |
e4a9bde9 | 159 | blkcg_pol_alloc_cpd_fn *cpd_alloc_fn; |
e48453c3 | 160 | blkcg_pol_init_cpd_fn *cpd_init_fn; |
e4a9bde9 | 161 | blkcg_pol_free_cpd_fn *cpd_free_fn; |
69d7fde5 | 162 | blkcg_pol_bind_cpd_fn *cpd_bind_fn; |
e4a9bde9 | 163 | |
001bea73 | 164 | blkcg_pol_alloc_pd_fn *pd_alloc_fn; |
f9fcc2d3 | 165 | blkcg_pol_init_pd_fn *pd_init_fn; |
f427d909 TH |
166 | blkcg_pol_online_pd_fn *pd_online_fn; |
167 | blkcg_pol_offline_pd_fn *pd_offline_fn; | |
001bea73 | 168 | blkcg_pol_free_pd_fn *pd_free_fn; |
f9fcc2d3 | 169 | blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn; |
3e252066 VG |
170 | }; |
171 | ||
3c798398 | 172 | extern struct blkcg blkcg_root; |
496d5e75 | 173 | extern struct cgroup_subsys_state * const blkcg_root_css; |
36558c8a | 174 | |
24f29046 TH |
175 | struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg, |
176 | struct request_queue *q, bool update_hint); | |
3c798398 TH |
177 | struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, |
178 | struct request_queue *q); | |
36558c8a TH |
179 | int blkcg_init_queue(struct request_queue *q); |
180 | void blkcg_drain_queue(struct request_queue *q); | |
181 | void blkcg_exit_queue(struct request_queue *q); | |
5efd6113 | 182 | |
3e252066 | 183 | /* Blkio controller policy registration */ |
d5bf0291 | 184 | int blkcg_policy_register(struct blkcg_policy *pol); |
3c798398 | 185 | void blkcg_policy_unregister(struct blkcg_policy *pol); |
36558c8a | 186 | int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 187 | const struct blkcg_policy *pol); |
36558c8a | 188 | void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 | 189 | const struct blkcg_policy *pol); |
3e252066 | 190 | |
dd165eb3 | 191 | const char *blkg_dev_name(struct blkcg_gq *blkg); |
3c798398 | 192 | void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg, |
f95a04af TH |
193 | u64 (*prfill)(struct seq_file *, |
194 | struct blkg_policy_data *, int), | |
3c798398 | 195 | const struct blkcg_policy *pol, int data, |
ec399347 | 196 | bool show_total); |
f95a04af TH |
197 | u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v); |
198 | u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
829fdb50 | 199 | const struct blkg_rwstat *rwstat); |
f95a04af TH |
200 | u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off); |
201 | u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
202 | int off); | |
77ea7338 TH |
203 | int blkg_print_stat_bytes(struct seq_file *sf, void *v); |
204 | int blkg_print_stat_ios(struct seq_file *sf, void *v); | |
205 | int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v); | |
206 | int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v); | |
829fdb50 | 207 | |
f12c74ca TH |
208 | u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg, |
209 | struct blkcg_policy *pol, int off); | |
210 | struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg, | |
211 | struct blkcg_policy *pol, int off); | |
16b3de66 | 212 | |
829fdb50 | 213 | struct blkg_conf_ctx { |
36558c8a | 214 | struct gendisk *disk; |
3c798398 | 215 | struct blkcg_gq *blkg; |
36aa9e5f | 216 | char *body; |
829fdb50 TH |
217 | }; |
218 | ||
3c798398 | 219 | int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol, |
36aa9e5f | 220 | char *input, struct blkg_conf_ctx *ctx); |
829fdb50 TH |
221 | void blkg_conf_finish(struct blkg_conf_ctx *ctx); |
222 | ||
223 | ||
a7c6d554 TH |
224 | static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css) |
225 | { | |
226 | return css ? container_of(css, struct blkcg, css) : NULL; | |
227 | } | |
228 | ||
b1208b56 TH |
229 | static inline struct blkcg *bio_blkcg(struct bio *bio) |
230 | { | |
902ec5b6 SL |
231 | struct cgroup_subsys_state *css; |
232 | ||
b1208b56 | 233 | if (bio && bio->bi_css) |
a7c6d554 | 234 | return css_to_blkcg(bio->bi_css); |
902ec5b6 SL |
235 | css = kthread_blkcg(); |
236 | if (css) | |
237 | return css_to_blkcg(css); | |
238 | return css_to_blkcg(task_css(current, 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 | */ | |
247 | static 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 | */ | |
263 | static 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 | */ | |
288 | static 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 |
305 | static 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 |
311 | static 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 | 323 | static 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 |
328 | static 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 | 341 | static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen) |
afc24d49 | 342 | { |
4c737b41 | 343 | return cgroup_path(blkg->blkcg->css.cgroup, buf, buflen); |
afc24d49 VG |
344 | } |
345 | ||
1adaf3dd TH |
346 | /** |
347 | * blkg_get - get a blkg reference | |
348 | * @blkg: blkg to get | |
349 | * | |
a5049a8a | 350 | * The caller should be holding an existing reference. |
1adaf3dd | 351 | */ |
3c798398 | 352 | static inline void blkg_get(struct blkcg_gq *blkg) |
1adaf3dd | 353 | { |
a5049a8a TH |
354 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
355 | atomic_inc(&blkg->refcnt); | |
1adaf3dd TH |
356 | } |
357 | ||
2a4fd070 | 358 | void __blkg_release_rcu(struct rcu_head *rcu); |
1adaf3dd TH |
359 | |
360 | /** | |
361 | * blkg_put - put a blkg reference | |
362 | * @blkg: blkg to put | |
1adaf3dd | 363 | */ |
3c798398 | 364 | static inline void blkg_put(struct blkcg_gq *blkg) |
1adaf3dd | 365 | { |
a5049a8a TH |
366 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
367 | if (atomic_dec_and_test(&blkg->refcnt)) | |
2a4fd070 | 368 | call_rcu(&blkg->rcu_head, __blkg_release_rcu); |
1adaf3dd TH |
369 | } |
370 | ||
dd4a4ffc TH |
371 | /** |
372 | * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants | |
373 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 374 | * @pos_css: used for iteration |
dd4a4ffc TH |
375 | * @p_blkg: target blkg to walk descendants of |
376 | * | |
377 | * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU | |
378 | * read locked. If called under either blkcg or queue lock, the iteration | |
379 | * is guaranteed to include all and only online blkgs. The caller may | |
492eb21b | 380 | * update @pos_css by calling css_rightmost_descendant() to skip subtree. |
bd8815a6 | 381 | * @p_blkg is included in the iteration and the first node to be visited. |
dd4a4ffc | 382 | */ |
492eb21b TH |
383 | #define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \ |
384 | css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \ | |
385 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
dd4a4ffc TH |
386 | (p_blkg)->q, false))) |
387 | ||
aa539cb3 TH |
388 | /** |
389 | * blkg_for_each_descendant_post - post-order walk of a blkg's descendants | |
390 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 391 | * @pos_css: used for iteration |
aa539cb3 TH |
392 | * @p_blkg: target blkg to walk descendants of |
393 | * | |
394 | * Similar to blkg_for_each_descendant_pre() but performs post-order | |
bd8815a6 TH |
395 | * traversal instead. Synchronization rules are the same. @p_blkg is |
396 | * included in the iteration and the last node to be visited. | |
aa539cb3 | 397 | */ |
492eb21b TH |
398 | #define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \ |
399 | css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \ | |
400 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
aa539cb3 TH |
401 | (p_blkg)->q, false))) |
402 | ||
a051661c TH |
403 | /** |
404 | * blk_get_rl - get request_list to use | |
405 | * @q: request_queue of interest | |
406 | * @bio: bio which will be attached to the allocated request (may be %NULL) | |
407 | * | |
408 | * The caller wants to allocate a request from @q to use for @bio. Find | |
409 | * the request_list to use and obtain a reference on it. Should be called | |
410 | * under queue_lock. This function is guaranteed to return non-%NULL | |
411 | * request_list. | |
412 | */ | |
413 | static inline struct request_list *blk_get_rl(struct request_queue *q, | |
414 | struct bio *bio) | |
415 | { | |
416 | struct blkcg *blkcg; | |
417 | struct blkcg_gq *blkg; | |
418 | ||
419 | rcu_read_lock(); | |
420 | ||
421 | blkcg = bio_blkcg(bio); | |
422 | ||
423 | /* bypass blkg lookup and use @q->root_rl directly for root */ | |
424 | if (blkcg == &blkcg_root) | |
425 | goto root_rl; | |
426 | ||
427 | /* | |
428 | * Try to use blkg->rl. blkg lookup may fail under memory pressure | |
429 | * or if either the blkcg or queue is going away. Fall back to | |
430 | * root_rl in such cases. | |
431 | */ | |
ae118896 TH |
432 | blkg = blkg_lookup(blkcg, q); |
433 | if (unlikely(!blkg)) | |
a051661c TH |
434 | goto root_rl; |
435 | ||
436 | blkg_get(blkg); | |
437 | rcu_read_unlock(); | |
438 | return &blkg->rl; | |
439 | root_rl: | |
440 | rcu_read_unlock(); | |
441 | return &q->root_rl; | |
442 | } | |
443 | ||
444 | /** | |
445 | * blk_put_rl - put request_list | |
446 | * @rl: request_list to put | |
447 | * | |
448 | * Put the reference acquired by blk_get_rl(). Should be called under | |
449 | * queue_lock. | |
450 | */ | |
451 | static inline void blk_put_rl(struct request_list *rl) | |
452 | { | |
401efbf8 | 453 | if (rl->blkg->blkcg != &blkcg_root) |
a051661c TH |
454 | blkg_put(rl->blkg); |
455 | } | |
456 | ||
457 | /** | |
458 | * blk_rq_set_rl - associate a request with a request_list | |
459 | * @rq: request of interest | |
460 | * @rl: target request_list | |
461 | * | |
462 | * Associate @rq with @rl so that accounting and freeing can know the | |
463 | * request_list @rq came from. | |
464 | */ | |
465 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) | |
466 | { | |
467 | rq->rl = rl; | |
468 | } | |
469 | ||
470 | /** | |
471 | * blk_rq_rl - return the request_list a request came from | |
472 | * @rq: request of interest | |
473 | * | |
474 | * Return the request_list @rq is allocated from. | |
475 | */ | |
476 | static inline struct request_list *blk_rq_rl(struct request *rq) | |
477 | { | |
478 | return rq->rl; | |
479 | } | |
480 | ||
481 | struct request_list *__blk_queue_next_rl(struct request_list *rl, | |
482 | struct request_queue *q); | |
483 | /** | |
484 | * blk_queue_for_each_rl - iterate through all request_lists of a request_queue | |
485 | * | |
486 | * Should be used under queue_lock. | |
487 | */ | |
488 | #define blk_queue_for_each_rl(rl, q) \ | |
489 | for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q))) | |
490 | ||
24bdb8ef | 491 | static inline int blkg_stat_init(struct blkg_stat *stat, gfp_t gfp) |
90d3839b | 492 | { |
24bdb8ef TH |
493 | int ret; |
494 | ||
495 | ret = percpu_counter_init(&stat->cpu_cnt, 0, gfp); | |
496 | if (ret) | |
497 | return ret; | |
498 | ||
e6269c44 | 499 | atomic64_set(&stat->aux_cnt, 0); |
24bdb8ef TH |
500 | return 0; |
501 | } | |
502 | ||
503 | static inline void blkg_stat_exit(struct blkg_stat *stat) | |
504 | { | |
505 | percpu_counter_destroy(&stat->cpu_cnt); | |
90d3839b PZ |
506 | } |
507 | ||
edcb0722 TH |
508 | /** |
509 | * blkg_stat_add - add a value to a blkg_stat | |
510 | * @stat: target blkg_stat | |
511 | * @val: value to add | |
512 | * | |
24bdb8ef TH |
513 | * Add @val to @stat. The caller must ensure that IRQ on the same CPU |
514 | * don't re-enter this function for the same counter. | |
edcb0722 TH |
515 | */ |
516 | static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val) | |
517 | { | |
104b4e51 | 518 | percpu_counter_add_batch(&stat->cpu_cnt, val, BLKG_STAT_CPU_BATCH); |
edcb0722 TH |
519 | } |
520 | ||
521 | /** | |
522 | * blkg_stat_read - read the current value of a blkg_stat | |
523 | * @stat: blkg_stat to read | |
edcb0722 TH |
524 | */ |
525 | static inline uint64_t blkg_stat_read(struct blkg_stat *stat) | |
526 | { | |
24bdb8ef | 527 | return percpu_counter_sum_positive(&stat->cpu_cnt); |
edcb0722 TH |
528 | } |
529 | ||
530 | /** | |
531 | * blkg_stat_reset - reset a blkg_stat | |
532 | * @stat: blkg_stat to reset | |
533 | */ | |
534 | static inline void blkg_stat_reset(struct blkg_stat *stat) | |
535 | { | |
24bdb8ef | 536 | percpu_counter_set(&stat->cpu_cnt, 0); |
e6269c44 | 537 | atomic64_set(&stat->aux_cnt, 0); |
edcb0722 TH |
538 | } |
539 | ||
16b3de66 | 540 | /** |
e6269c44 | 541 | * blkg_stat_add_aux - add a blkg_stat into another's aux count |
16b3de66 TH |
542 | * @to: the destination blkg_stat |
543 | * @from: the source | |
544 | * | |
e6269c44 | 545 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 546 | */ |
e6269c44 TH |
547 | static inline void blkg_stat_add_aux(struct blkg_stat *to, |
548 | struct blkg_stat *from) | |
16b3de66 | 549 | { |
e6269c44 TH |
550 | atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt), |
551 | &to->aux_cnt); | |
16b3de66 TH |
552 | } |
553 | ||
24bdb8ef | 554 | static inline int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp) |
90d3839b | 555 | { |
24bdb8ef TH |
556 | int i, ret; |
557 | ||
558 | for (i = 0; i < BLKG_RWSTAT_NR; i++) { | |
559 | ret = percpu_counter_init(&rwstat->cpu_cnt[i], 0, gfp); | |
560 | if (ret) { | |
561 | while (--i >= 0) | |
562 | percpu_counter_destroy(&rwstat->cpu_cnt[i]); | |
563 | return ret; | |
564 | } | |
565 | atomic64_set(&rwstat->aux_cnt[i], 0); | |
566 | } | |
567 | return 0; | |
568 | } | |
e6269c44 | 569 | |
24bdb8ef TH |
570 | static inline void blkg_rwstat_exit(struct blkg_rwstat *rwstat) |
571 | { | |
572 | int i; | |
e6269c44 TH |
573 | |
574 | for (i = 0; i < BLKG_RWSTAT_NR; i++) | |
24bdb8ef | 575 | percpu_counter_destroy(&rwstat->cpu_cnt[i]); |
90d3839b PZ |
576 | } |
577 | ||
edcb0722 TH |
578 | /** |
579 | * blkg_rwstat_add - add a value to a blkg_rwstat | |
580 | * @rwstat: target blkg_rwstat | |
ef295ecf | 581 | * @op: REQ_OP and flags |
edcb0722 TH |
582 | * @val: value to add |
583 | * | |
584 | * Add @val to @rwstat. The counters are chosen according to @rw. The | |
585 | * caller is responsible for synchronizing calls to this function. | |
586 | */ | |
587 | static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat, | |
ef295ecf | 588 | unsigned int op, uint64_t val) |
edcb0722 | 589 | { |
24bdb8ef | 590 | struct percpu_counter *cnt; |
edcb0722 | 591 | |
63a4cc24 | 592 | if (op_is_write(op)) |
24bdb8ef | 593 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_WRITE]; |
edcb0722 | 594 | else |
24bdb8ef TH |
595 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_READ]; |
596 | ||
104b4e51 | 597 | percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH); |
24bdb8ef | 598 | |
d71d9ae1 | 599 | if (op_is_sync(op)) |
24bdb8ef | 600 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_SYNC]; |
edcb0722 | 601 | else |
24bdb8ef | 602 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_ASYNC]; |
edcb0722 | 603 | |
104b4e51 | 604 | percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH); |
edcb0722 TH |
605 | } |
606 | ||
607 | /** | |
608 | * blkg_rwstat_read - read the current values of a blkg_rwstat | |
609 | * @rwstat: blkg_rwstat to read | |
610 | * | |
24bdb8ef | 611 | * Read the current snapshot of @rwstat and return it in the aux counts. |
edcb0722 | 612 | */ |
c94bed89 | 613 | static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat) |
edcb0722 | 614 | { |
24bdb8ef TH |
615 | struct blkg_rwstat result; |
616 | int i; | |
edcb0722 | 617 | |
24bdb8ef TH |
618 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
619 | atomic64_set(&result.aux_cnt[i], | |
620 | percpu_counter_sum_positive(&rwstat->cpu_cnt[i])); | |
621 | return result; | |
edcb0722 TH |
622 | } |
623 | ||
624 | /** | |
4d5e80a7 | 625 | * blkg_rwstat_total - read the total count of a blkg_rwstat |
edcb0722 TH |
626 | * @rwstat: blkg_rwstat to read |
627 | * | |
628 | * Return the total count of @rwstat regardless of the IO direction. This | |
629 | * function can be called without synchronization and takes care of u64 | |
630 | * atomicity. | |
631 | */ | |
4d5e80a7 | 632 | static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat) |
edcb0722 TH |
633 | { |
634 | struct blkg_rwstat tmp = blkg_rwstat_read(rwstat); | |
635 | ||
24bdb8ef TH |
636 | return atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) + |
637 | atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]); | |
edcb0722 TH |
638 | } |
639 | ||
640 | /** | |
641 | * blkg_rwstat_reset - reset a blkg_rwstat | |
642 | * @rwstat: blkg_rwstat to reset | |
643 | */ | |
644 | static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat) | |
645 | { | |
e6269c44 TH |
646 | int i; |
647 | ||
24bdb8ef TH |
648 | for (i = 0; i < BLKG_RWSTAT_NR; i++) { |
649 | percpu_counter_set(&rwstat->cpu_cnt[i], 0); | |
e6269c44 | 650 | atomic64_set(&rwstat->aux_cnt[i], 0); |
24bdb8ef | 651 | } |
edcb0722 TH |
652 | } |
653 | ||
16b3de66 | 654 | /** |
e6269c44 | 655 | * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count |
16b3de66 TH |
656 | * @to: the destination blkg_rwstat |
657 | * @from: the source | |
658 | * | |
e6269c44 | 659 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 660 | */ |
e6269c44 TH |
661 | static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to, |
662 | struct blkg_rwstat *from) | |
16b3de66 | 663 | { |
ddc21231 | 664 | u64 sum[BLKG_RWSTAT_NR]; |
16b3de66 TH |
665 | int i; |
666 | ||
16b3de66 | 667 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
ddc21231 AB |
668 | sum[i] = percpu_counter_sum_positive(&from->cpu_cnt[i]); |
669 | ||
670 | for (i = 0; i < BLKG_RWSTAT_NR; i++) | |
671 | atomic64_add(sum[i] + atomic64_read(&from->aux_cnt[i]), | |
e6269c44 | 672 | &to->aux_cnt[i]); |
16b3de66 TH |
673 | } |
674 | ||
ae118896 TH |
675 | #ifdef CONFIG_BLK_DEV_THROTTLING |
676 | extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
677 | struct bio *bio); | |
678 | #else | |
679 | static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
680 | struct bio *bio) { return false; } | |
681 | #endif | |
682 | ||
683 | static inline bool blkcg_bio_issue_check(struct request_queue *q, | |
684 | struct bio *bio) | |
685 | { | |
686 | struct blkcg *blkcg; | |
687 | struct blkcg_gq *blkg; | |
688 | bool throtl = false; | |
689 | ||
690 | rcu_read_lock(); | |
691 | blkcg = bio_blkcg(bio); | |
692 | ||
007cc56b SL |
693 | /* associate blkcg if bio hasn't attached one */ |
694 | bio_associate_blkcg(bio, &blkcg->css); | |
695 | ||
ae118896 TH |
696 | blkg = blkg_lookup(blkcg, q); |
697 | if (unlikely(!blkg)) { | |
698 | spin_lock_irq(q->queue_lock); | |
699 | blkg = blkg_lookup_create(blkcg, q); | |
700 | if (IS_ERR(blkg)) | |
701 | blkg = NULL; | |
702 | spin_unlock_irq(q->queue_lock); | |
703 | } | |
704 | ||
705 | throtl = blk_throtl_bio(q, blkg, bio); | |
706 | ||
77ea7338 TH |
707 | if (!throtl) { |
708 | blkg = blkg ?: q->root_blkg; | |
ef295ecf | 709 | blkg_rwstat_add(&blkg->stat_bytes, bio->bi_opf, |
77ea7338 | 710 | bio->bi_iter.bi_size); |
ef295ecf | 711 | blkg_rwstat_add(&blkg->stat_ios, bio->bi_opf, 1); |
77ea7338 TH |
712 | } |
713 | ||
ae118896 TH |
714 | rcu_read_unlock(); |
715 | return !throtl; | |
716 | } | |
717 | ||
36558c8a TH |
718 | #else /* CONFIG_BLK_CGROUP */ |
719 | ||
efa7d1c7 TH |
720 | struct blkcg { |
721 | }; | |
2f5ea477 | 722 | |
f95a04af TH |
723 | struct blkg_policy_data { |
724 | }; | |
725 | ||
e48453c3 AA |
726 | struct blkcg_policy_data { |
727 | }; | |
728 | ||
3c798398 | 729 | struct blkcg_gq { |
2f5ea477 JA |
730 | }; |
731 | ||
3c798398 | 732 | struct blkcg_policy { |
3e252066 VG |
733 | }; |
734 | ||
496d5e75 TH |
735 | #define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL)) |
736 | ||
efa7d1c7 TH |
737 | #ifdef CONFIG_BLOCK |
738 | ||
3c798398 | 739 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; } |
5efd6113 TH |
740 | static inline int blkcg_init_queue(struct request_queue *q) { return 0; } |
741 | static inline void blkcg_drain_queue(struct request_queue *q) { } | |
742 | static inline void blkcg_exit_queue(struct request_queue *q) { } | |
d5bf0291 | 743 | static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; } |
3c798398 | 744 | static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { } |
a2b1693b | 745 | static inline int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 746 | const struct blkcg_policy *pol) { return 0; } |
a2b1693b | 747 | static inline void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 TH |
748 | const struct blkcg_policy *pol) { } |
749 | ||
b1208b56 | 750 | static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; } |
a051661c | 751 | |
f95a04af TH |
752 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
753 | struct blkcg_policy *pol) { return NULL; } | |
754 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; } | |
3c798398 TH |
755 | static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; } |
756 | static inline void blkg_get(struct blkcg_gq *blkg) { } | |
757 | static inline void blkg_put(struct blkcg_gq *blkg) { } | |
afc24d49 | 758 | |
a051661c TH |
759 | static inline struct request_list *blk_get_rl(struct request_queue *q, |
760 | struct bio *bio) { return &q->root_rl; } | |
761 | static inline void blk_put_rl(struct request_list *rl) { } | |
762 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { } | |
763 | static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; } | |
764 | ||
ae118896 TH |
765 | static inline bool blkcg_bio_issue_check(struct request_queue *q, |
766 | struct bio *bio) { return true; } | |
767 | ||
a051661c TH |
768 | #define blk_queue_for_each_rl(rl, q) \ |
769 | for ((rl) = &(q)->root_rl; (rl); (rl) = NULL) | |
770 | ||
efa7d1c7 | 771 | #endif /* CONFIG_BLOCK */ |
36558c8a TH |
772 | #endif /* CONFIG_BLK_CGROUP */ |
773 | #endif /* _BLK_CGROUP_H */ |