]>
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; |
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 | */ |
100 | struct 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 */ |
107 | struct 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 | 140 | typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp); |
81437648 | 141 | typedef void (blkcg_pol_init_cpd_fn)(struct blkcg_policy_data *cpd); |
e4a9bde9 | 142 | typedef void (blkcg_pol_free_cpd_fn)(struct blkcg_policy_data *cpd); |
69d7fde5 | 143 | typedef void (blkcg_pol_bind_cpd_fn)(struct blkcg_policy_data *cpd); |
001bea73 | 144 | typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node); |
a9520cd6 TH |
145 | typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd); |
146 | typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd); | |
147 | typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd); | |
001bea73 | 148 | typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd); |
a9520cd6 | 149 | typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd); |
3e252066 | 150 | |
3c798398 | 151 | struct 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 | 171 | extern struct blkcg blkcg_root; |
496d5e75 | 172 | extern struct cgroup_subsys_state * const blkcg_root_css; |
36558c8a | 173 | |
24f29046 TH |
174 | struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg, |
175 | struct request_queue *q, bool update_hint); | |
3c798398 TH |
176 | struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, |
177 | struct request_queue *q); | |
36558c8a TH |
178 | int blkcg_init_queue(struct request_queue *q); |
179 | void blkcg_drain_queue(struct request_queue *q); | |
180 | void blkcg_exit_queue(struct request_queue *q); | |
5efd6113 | 181 | |
3e252066 | 182 | /* Blkio controller policy registration */ |
d5bf0291 | 183 | int blkcg_policy_register(struct blkcg_policy *pol); |
3c798398 | 184 | void blkcg_policy_unregister(struct blkcg_policy *pol); |
36558c8a | 185 | int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 186 | const struct blkcg_policy *pol); |
36558c8a | 187 | void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 | 188 | const struct blkcg_policy *pol); |
3e252066 | 189 | |
dd165eb3 | 190 | const char *blkg_dev_name(struct blkcg_gq *blkg); |
3c798398 | 191 | void 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 |
196 | u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v); |
197 | u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
829fdb50 | 198 | const struct blkg_rwstat *rwstat); |
f95a04af TH |
199 | u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off); |
200 | u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
201 | int off); | |
77ea7338 TH |
202 | int blkg_print_stat_bytes(struct seq_file *sf, void *v); |
203 | int blkg_print_stat_ios(struct seq_file *sf, void *v); | |
204 | int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v); | |
205 | int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v); | |
829fdb50 | 206 | |
f12c74ca TH |
207 | u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg, |
208 | struct blkcg_policy *pol, int off); | |
209 | struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg, | |
210 | struct blkcg_policy *pol, int off); | |
16b3de66 | 211 | |
829fdb50 | 212 | struct blkg_conf_ctx { |
36558c8a | 213 | struct gendisk *disk; |
3c798398 | 214 | struct blkcg_gq *blkg; |
36aa9e5f | 215 | char *body; |
829fdb50 TH |
216 | }; |
217 | ||
3c798398 | 218 | int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol, |
36aa9e5f | 219 | char *input, struct blkg_conf_ctx *ctx); |
829fdb50 TH |
220 | void blkg_conf_finish(struct blkg_conf_ctx *ctx); |
221 | ||
222 | ||
a7c6d554 TH |
223 | static 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 |
228 | static 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 | */ | |
246 | static 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 | */ | |
262 | static 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 | */ | |
287 | static 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 | ||
d329af0c BVA |
297 | /** |
298 | * blkg_lookup - look up blkg for the specified request queue | |
299 | * @q: request_queue of interest | |
300 | * | |
301 | * Lookup blkg for @q at the root level. See also blkg_lookup(). | |
302 | */ | |
303 | static inline struct blkcg_gq *blkg_root_lookup(struct request_queue *q) | |
304 | { | |
305 | struct blkcg_gq *blkg; | |
306 | ||
307 | rcu_read_lock(); | |
308 | blkg = blkg_lookup(&blkcg_root, q); | |
309 | rcu_read_unlock(); | |
310 | ||
311 | return blkg; | |
312 | } | |
313 | ||
0381411e TH |
314 | /** |
315 | * blkg_to_pdata - get policy private data | |
316 | * @blkg: blkg of interest | |
317 | * @pol: policy of interest | |
318 | * | |
319 | * Return pointer to private data associated with the @blkg-@pol pair. | |
320 | */ | |
f95a04af TH |
321 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
322 | struct blkcg_policy *pol) | |
0381411e | 323 | { |
f95a04af | 324 | return blkg ? blkg->pd[pol->plid] : NULL; |
0381411e TH |
325 | } |
326 | ||
e48453c3 AA |
327 | static inline struct blkcg_policy_data *blkcg_to_cpd(struct blkcg *blkcg, |
328 | struct blkcg_policy *pol) | |
329 | { | |
81437648 | 330 | return blkcg ? blkcg->cpd[pol->plid] : NULL; |
e48453c3 AA |
331 | } |
332 | ||
0381411e TH |
333 | /** |
334 | * pdata_to_blkg - get blkg associated with policy private data | |
f95a04af | 335 | * @pd: policy private data of interest |
0381411e | 336 | * |
f95a04af | 337 | * @pd is policy private data. Determine the blkg it's associated with. |
0381411e | 338 | */ |
f95a04af | 339 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) |
0381411e | 340 | { |
f95a04af | 341 | return pd ? pd->blkg : NULL; |
0381411e TH |
342 | } |
343 | ||
81437648 TH |
344 | static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd) |
345 | { | |
346 | return cpd ? cpd->blkcg : NULL; | |
347 | } | |
348 | ||
54e7ed12 TH |
349 | /** |
350 | * blkg_path - format cgroup path of blkg | |
351 | * @blkg: blkg of interest | |
352 | * @buf: target buffer | |
353 | * @buflen: target buffer length | |
354 | * | |
355 | * Format the path of the cgroup of @blkg into @buf. | |
356 | */ | |
3c798398 | 357 | static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen) |
afc24d49 | 358 | { |
4c737b41 | 359 | return cgroup_path(blkg->blkcg->css.cgroup, buf, buflen); |
afc24d49 VG |
360 | } |
361 | ||
1adaf3dd TH |
362 | /** |
363 | * blkg_get - get a blkg reference | |
364 | * @blkg: blkg to get | |
365 | * | |
a5049a8a | 366 | * The caller should be holding an existing reference. |
1adaf3dd | 367 | */ |
3c798398 | 368 | static inline void blkg_get(struct blkcg_gq *blkg) |
1adaf3dd | 369 | { |
a5049a8a TH |
370 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
371 | atomic_inc(&blkg->refcnt); | |
1adaf3dd TH |
372 | } |
373 | ||
2a4fd070 | 374 | void __blkg_release_rcu(struct rcu_head *rcu); |
1adaf3dd TH |
375 | |
376 | /** | |
377 | * blkg_put - put a blkg reference | |
378 | * @blkg: blkg to put | |
1adaf3dd | 379 | */ |
3c798398 | 380 | static inline void blkg_put(struct blkcg_gq *blkg) |
1adaf3dd | 381 | { |
a5049a8a TH |
382 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
383 | if (atomic_dec_and_test(&blkg->refcnt)) | |
2a4fd070 | 384 | call_rcu(&blkg->rcu_head, __blkg_release_rcu); |
1adaf3dd TH |
385 | } |
386 | ||
dd4a4ffc TH |
387 | /** |
388 | * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants | |
389 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 390 | * @pos_css: used for iteration |
dd4a4ffc TH |
391 | * @p_blkg: target blkg to walk descendants of |
392 | * | |
393 | * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU | |
394 | * read locked. If called under either blkcg or queue lock, the iteration | |
395 | * is guaranteed to include all and only online blkgs. The caller may | |
492eb21b | 396 | * update @pos_css by calling css_rightmost_descendant() to skip subtree. |
bd8815a6 | 397 | * @p_blkg is included in the iteration and the first node to be visited. |
dd4a4ffc | 398 | */ |
492eb21b TH |
399 | #define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \ |
400 | css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \ | |
401 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
dd4a4ffc TH |
402 | (p_blkg)->q, false))) |
403 | ||
aa539cb3 TH |
404 | /** |
405 | * blkg_for_each_descendant_post - post-order walk of a blkg's descendants | |
406 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 407 | * @pos_css: used for iteration |
aa539cb3 TH |
408 | * @p_blkg: target blkg to walk descendants of |
409 | * | |
410 | * Similar to blkg_for_each_descendant_pre() but performs post-order | |
bd8815a6 TH |
411 | * traversal instead. Synchronization rules are the same. @p_blkg is |
412 | * included in the iteration and the last node to be visited. | |
aa539cb3 | 413 | */ |
492eb21b TH |
414 | #define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \ |
415 | css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \ | |
416 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
aa539cb3 TH |
417 | (p_blkg)->q, false))) |
418 | ||
a051661c TH |
419 | /** |
420 | * blk_get_rl - get request_list to use | |
421 | * @q: request_queue of interest | |
422 | * @bio: bio which will be attached to the allocated request (may be %NULL) | |
423 | * | |
424 | * The caller wants to allocate a request from @q to use for @bio. Find | |
425 | * the request_list to use and obtain a reference on it. Should be called | |
426 | * under queue_lock. This function is guaranteed to return non-%NULL | |
427 | * request_list. | |
428 | */ | |
429 | static inline struct request_list *blk_get_rl(struct request_queue *q, | |
430 | struct bio *bio) | |
431 | { | |
432 | struct blkcg *blkcg; | |
433 | struct blkcg_gq *blkg; | |
434 | ||
435 | rcu_read_lock(); | |
436 | ||
437 | blkcg = bio_blkcg(bio); | |
438 | ||
439 | /* bypass blkg lookup and use @q->root_rl directly for root */ | |
440 | if (blkcg == &blkcg_root) | |
441 | goto root_rl; | |
442 | ||
443 | /* | |
444 | * Try to use blkg->rl. blkg lookup may fail under memory pressure | |
445 | * or if either the blkcg or queue is going away. Fall back to | |
446 | * root_rl in such cases. | |
447 | */ | |
ae118896 TH |
448 | blkg = blkg_lookup(blkcg, q); |
449 | if (unlikely(!blkg)) | |
a051661c TH |
450 | goto root_rl; |
451 | ||
452 | blkg_get(blkg); | |
453 | rcu_read_unlock(); | |
454 | return &blkg->rl; | |
455 | root_rl: | |
456 | rcu_read_unlock(); | |
457 | return &q->root_rl; | |
458 | } | |
459 | ||
460 | /** | |
461 | * blk_put_rl - put request_list | |
462 | * @rl: request_list to put | |
463 | * | |
464 | * Put the reference acquired by blk_get_rl(). Should be called under | |
465 | * queue_lock. | |
466 | */ | |
467 | static inline void blk_put_rl(struct request_list *rl) | |
468 | { | |
401efbf8 | 469 | if (rl->blkg->blkcg != &blkcg_root) |
a051661c TH |
470 | blkg_put(rl->blkg); |
471 | } | |
472 | ||
473 | /** | |
474 | * blk_rq_set_rl - associate a request with a request_list | |
475 | * @rq: request of interest | |
476 | * @rl: target request_list | |
477 | * | |
478 | * Associate @rq with @rl so that accounting and freeing can know the | |
479 | * request_list @rq came from. | |
480 | */ | |
481 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) | |
482 | { | |
483 | rq->rl = rl; | |
484 | } | |
485 | ||
486 | /** | |
487 | * blk_rq_rl - return the request_list a request came from | |
488 | * @rq: request of interest | |
489 | * | |
490 | * Return the request_list @rq is allocated from. | |
491 | */ | |
492 | static inline struct request_list *blk_rq_rl(struct request *rq) | |
493 | { | |
494 | return rq->rl; | |
495 | } | |
496 | ||
497 | struct request_list *__blk_queue_next_rl(struct request_list *rl, | |
498 | struct request_queue *q); | |
499 | /** | |
500 | * blk_queue_for_each_rl - iterate through all request_lists of a request_queue | |
501 | * | |
502 | * Should be used under queue_lock. | |
503 | */ | |
504 | #define blk_queue_for_each_rl(rl, q) \ | |
505 | for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q))) | |
506 | ||
24bdb8ef | 507 | static inline int blkg_stat_init(struct blkg_stat *stat, gfp_t gfp) |
90d3839b | 508 | { |
24bdb8ef TH |
509 | int ret; |
510 | ||
511 | ret = percpu_counter_init(&stat->cpu_cnt, 0, gfp); | |
512 | if (ret) | |
513 | return ret; | |
514 | ||
e6269c44 | 515 | atomic64_set(&stat->aux_cnt, 0); |
24bdb8ef TH |
516 | return 0; |
517 | } | |
518 | ||
519 | static inline void blkg_stat_exit(struct blkg_stat *stat) | |
520 | { | |
521 | percpu_counter_destroy(&stat->cpu_cnt); | |
90d3839b PZ |
522 | } |
523 | ||
edcb0722 TH |
524 | /** |
525 | * blkg_stat_add - add a value to a blkg_stat | |
526 | * @stat: target blkg_stat | |
527 | * @val: value to add | |
528 | * | |
24bdb8ef TH |
529 | * Add @val to @stat. The caller must ensure that IRQ on the same CPU |
530 | * don't re-enter this function for the same counter. | |
edcb0722 TH |
531 | */ |
532 | static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val) | |
533 | { | |
104b4e51 | 534 | percpu_counter_add_batch(&stat->cpu_cnt, val, BLKG_STAT_CPU_BATCH); |
edcb0722 TH |
535 | } |
536 | ||
537 | /** | |
538 | * blkg_stat_read - read the current value of a blkg_stat | |
539 | * @stat: blkg_stat to read | |
edcb0722 TH |
540 | */ |
541 | static inline uint64_t blkg_stat_read(struct blkg_stat *stat) | |
542 | { | |
24bdb8ef | 543 | return percpu_counter_sum_positive(&stat->cpu_cnt); |
edcb0722 TH |
544 | } |
545 | ||
546 | /** | |
547 | * blkg_stat_reset - reset a blkg_stat | |
548 | * @stat: blkg_stat to reset | |
549 | */ | |
550 | static inline void blkg_stat_reset(struct blkg_stat *stat) | |
551 | { | |
24bdb8ef | 552 | percpu_counter_set(&stat->cpu_cnt, 0); |
e6269c44 | 553 | atomic64_set(&stat->aux_cnt, 0); |
edcb0722 TH |
554 | } |
555 | ||
16b3de66 | 556 | /** |
e6269c44 | 557 | * blkg_stat_add_aux - add a blkg_stat into another's aux count |
16b3de66 TH |
558 | * @to: the destination blkg_stat |
559 | * @from: the source | |
560 | * | |
e6269c44 | 561 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 562 | */ |
e6269c44 TH |
563 | static inline void blkg_stat_add_aux(struct blkg_stat *to, |
564 | struct blkg_stat *from) | |
16b3de66 | 565 | { |
e6269c44 TH |
566 | atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt), |
567 | &to->aux_cnt); | |
16b3de66 TH |
568 | } |
569 | ||
24bdb8ef | 570 | static inline int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp) |
90d3839b | 571 | { |
24bdb8ef TH |
572 | int i, ret; |
573 | ||
574 | for (i = 0; i < BLKG_RWSTAT_NR; i++) { | |
575 | ret = percpu_counter_init(&rwstat->cpu_cnt[i], 0, gfp); | |
576 | if (ret) { | |
577 | while (--i >= 0) | |
578 | percpu_counter_destroy(&rwstat->cpu_cnt[i]); | |
579 | return ret; | |
580 | } | |
581 | atomic64_set(&rwstat->aux_cnt[i], 0); | |
582 | } | |
583 | return 0; | |
584 | } | |
e6269c44 | 585 | |
24bdb8ef TH |
586 | static inline void blkg_rwstat_exit(struct blkg_rwstat *rwstat) |
587 | { | |
588 | int i; | |
e6269c44 TH |
589 | |
590 | for (i = 0; i < BLKG_RWSTAT_NR; i++) | |
24bdb8ef | 591 | percpu_counter_destroy(&rwstat->cpu_cnt[i]); |
90d3839b PZ |
592 | } |
593 | ||
edcb0722 TH |
594 | /** |
595 | * blkg_rwstat_add - add a value to a blkg_rwstat | |
596 | * @rwstat: target blkg_rwstat | |
ef295ecf | 597 | * @op: REQ_OP and flags |
edcb0722 TH |
598 | * @val: value to add |
599 | * | |
600 | * Add @val to @rwstat. The counters are chosen according to @rw. The | |
601 | * caller is responsible for synchronizing calls to this function. | |
602 | */ | |
603 | static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat, | |
ef295ecf | 604 | unsigned int op, uint64_t val) |
edcb0722 | 605 | { |
24bdb8ef | 606 | struct percpu_counter *cnt; |
edcb0722 | 607 | |
63a4cc24 | 608 | if (op_is_write(op)) |
24bdb8ef | 609 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_WRITE]; |
edcb0722 | 610 | else |
24bdb8ef TH |
611 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_READ]; |
612 | ||
104b4e51 | 613 | percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH); |
24bdb8ef | 614 | |
d71d9ae1 | 615 | if (op_is_sync(op)) |
24bdb8ef | 616 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_SYNC]; |
edcb0722 | 617 | else |
24bdb8ef | 618 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_ASYNC]; |
edcb0722 | 619 | |
104b4e51 | 620 | percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH); |
edcb0722 TH |
621 | } |
622 | ||
623 | /** | |
624 | * blkg_rwstat_read - read the current values of a blkg_rwstat | |
625 | * @rwstat: blkg_rwstat to read | |
626 | * | |
24bdb8ef | 627 | * Read the current snapshot of @rwstat and return it in the aux counts. |
edcb0722 | 628 | */ |
c94bed89 | 629 | static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat) |
edcb0722 | 630 | { |
24bdb8ef TH |
631 | struct blkg_rwstat result; |
632 | int i; | |
edcb0722 | 633 | |
24bdb8ef TH |
634 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
635 | atomic64_set(&result.aux_cnt[i], | |
636 | percpu_counter_sum_positive(&rwstat->cpu_cnt[i])); | |
637 | return result; | |
edcb0722 TH |
638 | } |
639 | ||
640 | /** | |
4d5e80a7 | 641 | * blkg_rwstat_total - read the total count of a blkg_rwstat |
edcb0722 TH |
642 | * @rwstat: blkg_rwstat to read |
643 | * | |
644 | * Return the total count of @rwstat regardless of the IO direction. This | |
645 | * function can be called without synchronization and takes care of u64 | |
646 | * atomicity. | |
647 | */ | |
4d5e80a7 | 648 | static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat) |
edcb0722 TH |
649 | { |
650 | struct blkg_rwstat tmp = blkg_rwstat_read(rwstat); | |
651 | ||
24bdb8ef TH |
652 | return atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) + |
653 | atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]); | |
edcb0722 TH |
654 | } |
655 | ||
656 | /** | |
657 | * blkg_rwstat_reset - reset a blkg_rwstat | |
658 | * @rwstat: blkg_rwstat to reset | |
659 | */ | |
660 | static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat) | |
661 | { | |
e6269c44 TH |
662 | int i; |
663 | ||
24bdb8ef TH |
664 | for (i = 0; i < BLKG_RWSTAT_NR; i++) { |
665 | percpu_counter_set(&rwstat->cpu_cnt[i], 0); | |
e6269c44 | 666 | atomic64_set(&rwstat->aux_cnt[i], 0); |
24bdb8ef | 667 | } |
edcb0722 TH |
668 | } |
669 | ||
16b3de66 | 670 | /** |
e6269c44 | 671 | * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count |
16b3de66 TH |
672 | * @to: the destination blkg_rwstat |
673 | * @from: the source | |
674 | * | |
e6269c44 | 675 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 676 | */ |
e6269c44 TH |
677 | static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to, |
678 | struct blkg_rwstat *from) | |
16b3de66 TH |
679 | { |
680 | struct blkg_rwstat v = blkg_rwstat_read(from); | |
681 | int i; | |
682 | ||
16b3de66 | 683 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
24bdb8ef TH |
684 | atomic64_add(atomic64_read(&v.aux_cnt[i]) + |
685 | atomic64_read(&from->aux_cnt[i]), | |
e6269c44 | 686 | &to->aux_cnt[i]); |
16b3de66 TH |
687 | } |
688 | ||
ae118896 TH |
689 | #ifdef CONFIG_BLK_DEV_THROTTLING |
690 | extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
691 | struct bio *bio); | |
692 | #else | |
693 | static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
694 | struct bio *bio) { return false; } | |
695 | #endif | |
696 | ||
697 | static inline bool blkcg_bio_issue_check(struct request_queue *q, | |
698 | struct bio *bio) | |
699 | { | |
700 | struct blkcg *blkcg; | |
701 | struct blkcg_gq *blkg; | |
702 | bool throtl = false; | |
703 | ||
704 | rcu_read_lock(); | |
705 | blkcg = bio_blkcg(bio); | |
706 | ||
007cc56b SL |
707 | /* associate blkcg if bio hasn't attached one */ |
708 | bio_associate_blkcg(bio, &blkcg->css); | |
709 | ||
ae118896 TH |
710 | blkg = blkg_lookup(blkcg, q); |
711 | if (unlikely(!blkg)) { | |
712 | spin_lock_irq(q->queue_lock); | |
713 | blkg = blkg_lookup_create(blkcg, q); | |
714 | if (IS_ERR(blkg)) | |
715 | blkg = NULL; | |
716 | spin_unlock_irq(q->queue_lock); | |
717 | } | |
718 | ||
719 | throtl = blk_throtl_bio(q, blkg, bio); | |
720 | ||
77ea7338 TH |
721 | if (!throtl) { |
722 | blkg = blkg ?: q->root_blkg; | |
ef295ecf | 723 | blkg_rwstat_add(&blkg->stat_bytes, bio->bi_opf, |
77ea7338 | 724 | bio->bi_iter.bi_size); |
ef295ecf | 725 | blkg_rwstat_add(&blkg->stat_ios, bio->bi_opf, 1); |
77ea7338 TH |
726 | } |
727 | ||
ae118896 TH |
728 | rcu_read_unlock(); |
729 | return !throtl; | |
730 | } | |
731 | ||
36558c8a TH |
732 | #else /* CONFIG_BLK_CGROUP */ |
733 | ||
efa7d1c7 TH |
734 | struct blkcg { |
735 | }; | |
2f5ea477 | 736 | |
f95a04af TH |
737 | struct blkg_policy_data { |
738 | }; | |
739 | ||
e48453c3 AA |
740 | struct blkcg_policy_data { |
741 | }; | |
742 | ||
3c798398 | 743 | struct blkcg_gq { |
2f5ea477 JA |
744 | }; |
745 | ||
3c798398 | 746 | struct blkcg_policy { |
3e252066 VG |
747 | }; |
748 | ||
496d5e75 TH |
749 | #define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL)) |
750 | ||
efa7d1c7 TH |
751 | #ifdef CONFIG_BLOCK |
752 | ||
3c798398 | 753 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; } |
d329af0c | 754 | static inline struct blkcg_gq *blkg_root_lookup(struct request_queue *q) { return NULL; } |
5efd6113 TH |
755 | static inline int blkcg_init_queue(struct request_queue *q) { return 0; } |
756 | static inline void blkcg_drain_queue(struct request_queue *q) { } | |
757 | static inline void blkcg_exit_queue(struct request_queue *q) { } | |
d5bf0291 | 758 | static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; } |
3c798398 | 759 | static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { } |
a2b1693b | 760 | static inline int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 761 | const struct blkcg_policy *pol) { return 0; } |
a2b1693b | 762 | static inline void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 TH |
763 | const struct blkcg_policy *pol) { } |
764 | ||
b1208b56 | 765 | static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; } |
a051661c | 766 | |
f95a04af TH |
767 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
768 | struct blkcg_policy *pol) { return NULL; } | |
769 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; } | |
3c798398 TH |
770 | static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; } |
771 | static inline void blkg_get(struct blkcg_gq *blkg) { } | |
772 | static inline void blkg_put(struct blkcg_gq *blkg) { } | |
afc24d49 | 773 | |
a051661c TH |
774 | static inline struct request_list *blk_get_rl(struct request_queue *q, |
775 | struct bio *bio) { return &q->root_rl; } | |
776 | static inline void blk_put_rl(struct request_list *rl) { } | |
777 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { } | |
778 | static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; } | |
779 | ||
ae118896 TH |
780 | static inline bool blkcg_bio_issue_check(struct request_queue *q, |
781 | struct bio *bio) { return true; } | |
782 | ||
a051661c TH |
783 | #define blk_queue_for_each_rl(rl, q) \ |
784 | for ((rl) = &(q)->root_rl; (rl); (rl) = NULL) | |
785 | ||
efa7d1c7 | 786 | #endif /* CONFIG_BLOCK */ |
36558c8a TH |
787 | #endif /* CONFIG_BLK_CGROUP */ |
788 | #endif /* _BLK_CGROUP_H */ |