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> |
31e4c28d | 23 | |
24bdb8ef TH |
24 | /* percpu_counter batch for blkg_[rw]stats, per-cpu drift doesn't matter */ |
25 | #define BLKG_STAT_CPU_BATCH (INT_MAX / 2) | |
26 | ||
9355aede VG |
27 | /* Max limits for throttle policy */ |
28 | #define THROTL_IOPS_MAX UINT_MAX | |
29 | ||
f48ec1d7 TH |
30 | #ifdef CONFIG_BLK_CGROUP |
31 | ||
edcb0722 TH |
32 | enum blkg_rwstat_type { |
33 | BLKG_RWSTAT_READ, | |
34 | BLKG_RWSTAT_WRITE, | |
35 | BLKG_RWSTAT_SYNC, | |
36 | BLKG_RWSTAT_ASYNC, | |
37 | ||
38 | BLKG_RWSTAT_NR, | |
39 | BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR, | |
303a3acb DS |
40 | }; |
41 | ||
a637120e TH |
42 | struct blkcg_gq; |
43 | ||
3c798398 | 44 | struct blkcg { |
36558c8a TH |
45 | struct cgroup_subsys_state css; |
46 | spinlock_t lock; | |
a637120e TH |
47 | |
48 | struct radix_tree_root blkg_tree; | |
55679c8d | 49 | struct blkcg_gq __rcu *blkg_hint; |
36558c8a | 50 | struct hlist_head blkg_list; |
9a9e8a26 | 51 | |
81437648 | 52 | struct blkcg_policy_data *cpd[BLKCG_MAX_POLS]; |
52ebea74 | 53 | |
7876f930 | 54 | struct list_head all_blkcgs_node; |
52ebea74 TH |
55 | #ifdef CONFIG_CGROUP_WRITEBACK |
56 | struct list_head cgwb_list; | |
57 | #endif | |
31e4c28d VG |
58 | }; |
59 | ||
e6269c44 TH |
60 | /* |
61 | * blkg_[rw]stat->aux_cnt is excluded for local stats but included for | |
24bdb8ef TH |
62 | * recursive. Used to carry stats of dead children, and, for blkg_rwstat, |
63 | * to carry result values from read and sum operations. | |
e6269c44 | 64 | */ |
edcb0722 | 65 | struct blkg_stat { |
24bdb8ef | 66 | struct percpu_counter cpu_cnt; |
e6269c44 | 67 | atomic64_t aux_cnt; |
edcb0722 TH |
68 | }; |
69 | ||
70 | struct blkg_rwstat { | |
24bdb8ef | 71 | struct percpu_counter cpu_cnt[BLKG_RWSTAT_NR]; |
e6269c44 | 72 | atomic64_t aux_cnt[BLKG_RWSTAT_NR]; |
edcb0722 TH |
73 | }; |
74 | ||
f95a04af TH |
75 | /* |
76 | * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a | |
77 | * request_queue (q). This is used by blkcg policies which need to track | |
78 | * information per blkcg - q pair. | |
79 | * | |
001bea73 TH |
80 | * There can be multiple active blkcg policies and each blkg:policy pair is |
81 | * represented by a blkg_policy_data which is allocated and freed by each | |
82 | * policy's pd_alloc/free_fn() methods. A policy can allocate private data | |
83 | * area by allocating larger data structure which embeds blkg_policy_data | |
84 | * at the beginning. | |
f95a04af | 85 | */ |
0381411e | 86 | struct blkg_policy_data { |
b276a876 | 87 | /* the blkg and policy id this per-policy data belongs to */ |
3c798398 | 88 | struct blkcg_gq *blkg; |
b276a876 | 89 | int plid; |
0381411e TH |
90 | }; |
91 | ||
e48453c3 | 92 | /* |
e4a9bde9 TH |
93 | * Policies that need to keep per-blkcg data which is independent from any |
94 | * request_queue associated to it should implement cpd_alloc/free_fn() | |
95 | * methods. A policy can allocate private data area by allocating larger | |
96 | * data structure which embeds blkcg_policy_data at the beginning. | |
97 | * cpd_init() is invoked to let each policy handle per-blkcg data. | |
e48453c3 AA |
98 | */ |
99 | struct blkcg_policy_data { | |
81437648 TH |
100 | /* the blkcg and policy id this per-policy data belongs to */ |
101 | struct blkcg *blkcg; | |
e48453c3 | 102 | int plid; |
e48453c3 AA |
103 | }; |
104 | ||
3c798398 TH |
105 | /* association between a blk cgroup and a request queue */ |
106 | struct blkcg_gq { | |
c875f4d0 | 107 | /* Pointer to the associated request_queue */ |
36558c8a TH |
108 | struct request_queue *q; |
109 | struct list_head q_node; | |
110 | struct hlist_node blkcg_node; | |
3c798398 | 111 | struct blkcg *blkcg; |
3c547865 | 112 | |
ce7acfea TH |
113 | /* |
114 | * Each blkg gets congested separately and the congestion state is | |
115 | * propagated to the matching bdi_writeback_congested. | |
116 | */ | |
117 | struct bdi_writeback_congested *wb_congested; | |
118 | ||
3c547865 TH |
119 | /* all non-root blkcg_gq's are guaranteed to have access to parent */ |
120 | struct blkcg_gq *parent; | |
121 | ||
a051661c TH |
122 | /* request allocation list for this blkcg-q pair */ |
123 | struct request_list rl; | |
3c547865 | 124 | |
1adaf3dd | 125 | /* reference count */ |
a5049a8a | 126 | atomic_t refcnt; |
22084190 | 127 | |
f427d909 TH |
128 | /* is this blkg online? protected by both blkcg and q locks */ |
129 | bool online; | |
130 | ||
77ea7338 TH |
131 | struct blkg_rwstat stat_bytes; |
132 | struct blkg_rwstat stat_ios; | |
133 | ||
36558c8a | 134 | struct blkg_policy_data *pd[BLKCG_MAX_POLS]; |
1adaf3dd | 135 | |
36558c8a | 136 | struct rcu_head rcu_head; |
31e4c28d VG |
137 | }; |
138 | ||
e4a9bde9 | 139 | typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp); |
81437648 | 140 | typedef void (blkcg_pol_init_cpd_fn)(struct blkcg_policy_data *cpd); |
e4a9bde9 | 141 | typedef void (blkcg_pol_free_cpd_fn)(struct blkcg_policy_data *cpd); |
69d7fde5 | 142 | typedef void (blkcg_pol_bind_cpd_fn)(struct blkcg_policy_data *cpd); |
001bea73 | 143 | typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node); |
a9520cd6 TH |
144 | typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd); |
145 | typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd); | |
146 | typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd); | |
001bea73 | 147 | typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd); |
a9520cd6 | 148 | typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd); |
3e252066 | 149 | |
3c798398 | 150 | struct blkcg_policy { |
36558c8a | 151 | int plid; |
36558c8a | 152 | /* cgroup files for the policy */ |
2ee867dc | 153 | struct cftype *dfl_cftypes; |
880f50e2 | 154 | struct cftype *legacy_cftypes; |
f9fcc2d3 TH |
155 | |
156 | /* operations */ | |
e4a9bde9 | 157 | blkcg_pol_alloc_cpd_fn *cpd_alloc_fn; |
e48453c3 | 158 | blkcg_pol_init_cpd_fn *cpd_init_fn; |
e4a9bde9 | 159 | blkcg_pol_free_cpd_fn *cpd_free_fn; |
69d7fde5 | 160 | blkcg_pol_bind_cpd_fn *cpd_bind_fn; |
e4a9bde9 | 161 | |
001bea73 | 162 | blkcg_pol_alloc_pd_fn *pd_alloc_fn; |
f9fcc2d3 | 163 | blkcg_pol_init_pd_fn *pd_init_fn; |
f427d909 TH |
164 | blkcg_pol_online_pd_fn *pd_online_fn; |
165 | blkcg_pol_offline_pd_fn *pd_offline_fn; | |
001bea73 | 166 | blkcg_pol_free_pd_fn *pd_free_fn; |
f9fcc2d3 | 167 | blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn; |
3e252066 VG |
168 | }; |
169 | ||
3c798398 | 170 | extern struct blkcg blkcg_root; |
496d5e75 | 171 | extern struct cgroup_subsys_state * const blkcg_root_css; |
36558c8a | 172 | |
24f29046 TH |
173 | struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg, |
174 | struct request_queue *q, bool update_hint); | |
3c798398 TH |
175 | struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, |
176 | struct request_queue *q); | |
36558c8a TH |
177 | int blkcg_init_queue(struct request_queue *q); |
178 | void blkcg_drain_queue(struct request_queue *q); | |
179 | void blkcg_exit_queue(struct request_queue *q); | |
5efd6113 | 180 | |
3e252066 | 181 | /* Blkio controller policy registration */ |
d5bf0291 | 182 | int blkcg_policy_register(struct blkcg_policy *pol); |
3c798398 | 183 | void blkcg_policy_unregister(struct blkcg_policy *pol); |
36558c8a | 184 | int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 185 | const struct blkcg_policy *pol); |
36558c8a | 186 | void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 | 187 | const struct blkcg_policy *pol); |
3e252066 | 188 | |
dd165eb3 | 189 | const char *blkg_dev_name(struct blkcg_gq *blkg); |
3c798398 | 190 | void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg, |
f95a04af TH |
191 | u64 (*prfill)(struct seq_file *, |
192 | struct blkg_policy_data *, int), | |
3c798398 | 193 | const struct blkcg_policy *pol, int data, |
ec399347 | 194 | bool show_total); |
f95a04af TH |
195 | u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v); |
196 | u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
829fdb50 | 197 | const struct blkg_rwstat *rwstat); |
f95a04af TH |
198 | u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off); |
199 | u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
200 | int off); | |
77ea7338 TH |
201 | int blkg_print_stat_bytes(struct seq_file *sf, void *v); |
202 | int blkg_print_stat_ios(struct seq_file *sf, void *v); | |
203 | int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v); | |
204 | int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v); | |
829fdb50 | 205 | |
f12c74ca TH |
206 | u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg, |
207 | struct blkcg_policy *pol, int off); | |
208 | struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg, | |
209 | struct blkcg_policy *pol, int off); | |
16b3de66 | 210 | |
829fdb50 | 211 | struct blkg_conf_ctx { |
36558c8a | 212 | struct gendisk *disk; |
3c798398 | 213 | struct blkcg_gq *blkg; |
36aa9e5f | 214 | char *body; |
829fdb50 TH |
215 | }; |
216 | ||
3c798398 | 217 | int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol, |
36aa9e5f | 218 | char *input, struct blkg_conf_ctx *ctx); |
829fdb50 TH |
219 | void blkg_conf_finish(struct blkg_conf_ctx *ctx); |
220 | ||
221 | ||
a7c6d554 TH |
222 | static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css) |
223 | { | |
224 | return css ? container_of(css, struct blkcg, css) : NULL; | |
225 | } | |
226 | ||
b1208b56 TH |
227 | static inline struct blkcg *task_blkcg(struct task_struct *tsk) |
228 | { | |
c165b3e3 | 229 | return css_to_blkcg(task_css(tsk, io_cgrp_id)); |
b1208b56 TH |
230 | } |
231 | ||
232 | static inline struct blkcg *bio_blkcg(struct bio *bio) | |
233 | { | |
234 | if (bio && bio->bi_css) | |
a7c6d554 | 235 | return css_to_blkcg(bio->bi_css); |
b1208b56 TH |
236 | return task_blkcg(current); |
237 | } | |
238 | ||
fd383c2d TH |
239 | static inline struct cgroup_subsys_state * |
240 | task_get_blkcg_css(struct task_struct *task) | |
241 | { | |
c165b3e3 | 242 | return task_get_css(task, io_cgrp_id); |
fd383c2d TH |
243 | } |
244 | ||
3c547865 TH |
245 | /** |
246 | * blkcg_parent - get the parent of a blkcg | |
247 | * @blkcg: blkcg of interest | |
248 | * | |
249 | * Return the parent blkcg of @blkcg. Can be called anytime. | |
250 | */ | |
251 | static inline struct blkcg *blkcg_parent(struct blkcg *blkcg) | |
252 | { | |
5c9d535b | 253 | return css_to_blkcg(blkcg->css.parent); |
3c547865 TH |
254 | } |
255 | ||
24f29046 TH |
256 | /** |
257 | * __blkg_lookup - internal version of blkg_lookup() | |
258 | * @blkcg: blkcg of interest | |
259 | * @q: request_queue of interest | |
260 | * @update_hint: whether to update lookup hint with the result or not | |
261 | * | |
262 | * This is internal version and shouldn't be used by policy | |
263 | * implementations. Looks up blkgs for the @blkcg - @q pair regardless of | |
264 | * @q's bypass state. If @update_hint is %true, the caller should be | |
265 | * holding @q->queue_lock and lookup hint is updated on success. | |
266 | */ | |
267 | static inline struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg, | |
268 | struct request_queue *q, | |
269 | bool update_hint) | |
270 | { | |
271 | struct blkcg_gq *blkg; | |
272 | ||
85b6bc9d TH |
273 | if (blkcg == &blkcg_root) |
274 | return q->root_blkg; | |
275 | ||
24f29046 TH |
276 | blkg = rcu_dereference(blkcg->blkg_hint); |
277 | if (blkg && blkg->q == q) | |
278 | return blkg; | |
279 | ||
280 | return blkg_lookup_slowpath(blkcg, q, update_hint); | |
281 | } | |
282 | ||
283 | /** | |
284 | * blkg_lookup - lookup blkg for the specified blkcg - q pair | |
285 | * @blkcg: blkcg of interest | |
286 | * @q: request_queue of interest | |
287 | * | |
288 | * Lookup blkg for the @blkcg - @q pair. This function should be called | |
289 | * under RCU read lock and is guaranteed to return %NULL if @q is bypassing | |
290 | * - see blk_queue_bypass_start() for details. | |
291 | */ | |
292 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, | |
293 | struct request_queue *q) | |
294 | { | |
295 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
296 | ||
297 | if (unlikely(blk_queue_bypass(q))) | |
298 | return NULL; | |
299 | return __blkg_lookup(blkcg, q, false); | |
300 | } | |
301 | ||
0381411e TH |
302 | /** |
303 | * blkg_to_pdata - get policy private data | |
304 | * @blkg: blkg of interest | |
305 | * @pol: policy of interest | |
306 | * | |
307 | * Return pointer to private data associated with the @blkg-@pol pair. | |
308 | */ | |
f95a04af TH |
309 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
310 | struct blkcg_policy *pol) | |
0381411e | 311 | { |
f95a04af | 312 | return blkg ? blkg->pd[pol->plid] : NULL; |
0381411e TH |
313 | } |
314 | ||
e48453c3 AA |
315 | static inline struct blkcg_policy_data *blkcg_to_cpd(struct blkcg *blkcg, |
316 | struct blkcg_policy *pol) | |
317 | { | |
81437648 | 318 | return blkcg ? blkcg->cpd[pol->plid] : NULL; |
e48453c3 AA |
319 | } |
320 | ||
0381411e TH |
321 | /** |
322 | * pdata_to_blkg - get blkg associated with policy private data | |
f95a04af | 323 | * @pd: policy private data of interest |
0381411e | 324 | * |
f95a04af | 325 | * @pd is policy private data. Determine the blkg it's associated with. |
0381411e | 326 | */ |
f95a04af | 327 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) |
0381411e | 328 | { |
f95a04af | 329 | return pd ? pd->blkg : NULL; |
0381411e TH |
330 | } |
331 | ||
81437648 TH |
332 | static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd) |
333 | { | |
334 | return cpd ? cpd->blkcg : NULL; | |
335 | } | |
336 | ||
54e7ed12 TH |
337 | /** |
338 | * blkg_path - format cgroup path of blkg | |
339 | * @blkg: blkg of interest | |
340 | * @buf: target buffer | |
341 | * @buflen: target buffer length | |
342 | * | |
343 | * Format the path of the cgroup of @blkg into @buf. | |
344 | */ | |
3c798398 | 345 | static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen) |
afc24d49 | 346 | { |
4c737b41 | 347 | return cgroup_path(blkg->blkcg->css.cgroup, buf, buflen); |
afc24d49 VG |
348 | } |
349 | ||
1adaf3dd TH |
350 | /** |
351 | * blkg_get - get a blkg reference | |
352 | * @blkg: blkg to get | |
353 | * | |
a5049a8a | 354 | * The caller should be holding an existing reference. |
1adaf3dd | 355 | */ |
3c798398 | 356 | static inline void blkg_get(struct blkcg_gq *blkg) |
1adaf3dd | 357 | { |
a5049a8a TH |
358 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
359 | atomic_inc(&blkg->refcnt); | |
1adaf3dd TH |
360 | } |
361 | ||
2a4fd070 | 362 | void __blkg_release_rcu(struct rcu_head *rcu); |
1adaf3dd TH |
363 | |
364 | /** | |
365 | * blkg_put - put a blkg reference | |
366 | * @blkg: blkg to put | |
1adaf3dd | 367 | */ |
3c798398 | 368 | static inline void blkg_put(struct blkcg_gq *blkg) |
1adaf3dd | 369 | { |
a5049a8a TH |
370 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
371 | if (atomic_dec_and_test(&blkg->refcnt)) | |
2a4fd070 | 372 | call_rcu(&blkg->rcu_head, __blkg_release_rcu); |
1adaf3dd TH |
373 | } |
374 | ||
dd4a4ffc TH |
375 | /** |
376 | * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants | |
377 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 378 | * @pos_css: used for iteration |
dd4a4ffc TH |
379 | * @p_blkg: target blkg to walk descendants of |
380 | * | |
381 | * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU | |
382 | * read locked. If called under either blkcg or queue lock, the iteration | |
383 | * is guaranteed to include all and only online blkgs. The caller may | |
492eb21b | 384 | * update @pos_css by calling css_rightmost_descendant() to skip subtree. |
bd8815a6 | 385 | * @p_blkg is included in the iteration and the first node to be visited. |
dd4a4ffc | 386 | */ |
492eb21b TH |
387 | #define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \ |
388 | css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \ | |
389 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
dd4a4ffc TH |
390 | (p_blkg)->q, false))) |
391 | ||
aa539cb3 TH |
392 | /** |
393 | * blkg_for_each_descendant_post - post-order walk of a blkg's descendants | |
394 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 395 | * @pos_css: used for iteration |
aa539cb3 TH |
396 | * @p_blkg: target blkg to walk descendants of |
397 | * | |
398 | * Similar to blkg_for_each_descendant_pre() but performs post-order | |
bd8815a6 TH |
399 | * traversal instead. Synchronization rules are the same. @p_blkg is |
400 | * included in the iteration and the last node to be visited. | |
aa539cb3 | 401 | */ |
492eb21b TH |
402 | #define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \ |
403 | css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \ | |
404 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
aa539cb3 TH |
405 | (p_blkg)->q, false))) |
406 | ||
a051661c TH |
407 | /** |
408 | * blk_get_rl - get request_list to use | |
409 | * @q: request_queue of interest | |
410 | * @bio: bio which will be attached to the allocated request (may be %NULL) | |
411 | * | |
412 | * The caller wants to allocate a request from @q to use for @bio. Find | |
413 | * the request_list to use and obtain a reference on it. Should be called | |
414 | * under queue_lock. This function is guaranteed to return non-%NULL | |
415 | * request_list. | |
416 | */ | |
417 | static inline struct request_list *blk_get_rl(struct request_queue *q, | |
418 | struct bio *bio) | |
419 | { | |
420 | struct blkcg *blkcg; | |
421 | struct blkcg_gq *blkg; | |
422 | ||
423 | rcu_read_lock(); | |
424 | ||
425 | blkcg = bio_blkcg(bio); | |
426 | ||
427 | /* bypass blkg lookup and use @q->root_rl directly for root */ | |
428 | if (blkcg == &blkcg_root) | |
429 | goto root_rl; | |
430 | ||
431 | /* | |
432 | * Try to use blkg->rl. blkg lookup may fail under memory pressure | |
433 | * or if either the blkcg or queue is going away. Fall back to | |
434 | * root_rl in such cases. | |
435 | */ | |
ae118896 TH |
436 | blkg = blkg_lookup(blkcg, q); |
437 | if (unlikely(!blkg)) | |
a051661c TH |
438 | goto root_rl; |
439 | ||
440 | blkg_get(blkg); | |
441 | rcu_read_unlock(); | |
442 | return &blkg->rl; | |
443 | root_rl: | |
444 | rcu_read_unlock(); | |
445 | return &q->root_rl; | |
446 | } | |
447 | ||
448 | /** | |
449 | * blk_put_rl - put request_list | |
450 | * @rl: request_list to put | |
451 | * | |
452 | * Put the reference acquired by blk_get_rl(). Should be called under | |
453 | * queue_lock. | |
454 | */ | |
455 | static inline void blk_put_rl(struct request_list *rl) | |
456 | { | |
401efbf8 | 457 | if (rl->blkg->blkcg != &blkcg_root) |
a051661c TH |
458 | blkg_put(rl->blkg); |
459 | } | |
460 | ||
461 | /** | |
462 | * blk_rq_set_rl - associate a request with a request_list | |
463 | * @rq: request of interest | |
464 | * @rl: target request_list | |
465 | * | |
466 | * Associate @rq with @rl so that accounting and freeing can know the | |
467 | * request_list @rq came from. | |
468 | */ | |
469 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) | |
470 | { | |
471 | rq->rl = rl; | |
472 | } | |
473 | ||
474 | /** | |
475 | * blk_rq_rl - return the request_list a request came from | |
476 | * @rq: request of interest | |
477 | * | |
478 | * Return the request_list @rq is allocated from. | |
479 | */ | |
480 | static inline struct request_list *blk_rq_rl(struct request *rq) | |
481 | { | |
482 | return rq->rl; | |
483 | } | |
484 | ||
485 | struct request_list *__blk_queue_next_rl(struct request_list *rl, | |
486 | struct request_queue *q); | |
487 | /** | |
488 | * blk_queue_for_each_rl - iterate through all request_lists of a request_queue | |
489 | * | |
490 | * Should be used under queue_lock. | |
491 | */ | |
492 | #define blk_queue_for_each_rl(rl, q) \ | |
493 | for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q))) | |
494 | ||
24bdb8ef | 495 | static inline int blkg_stat_init(struct blkg_stat *stat, gfp_t gfp) |
90d3839b | 496 | { |
24bdb8ef TH |
497 | int ret; |
498 | ||
499 | ret = percpu_counter_init(&stat->cpu_cnt, 0, gfp); | |
500 | if (ret) | |
501 | return ret; | |
502 | ||
e6269c44 | 503 | atomic64_set(&stat->aux_cnt, 0); |
24bdb8ef TH |
504 | return 0; |
505 | } | |
506 | ||
507 | static inline void blkg_stat_exit(struct blkg_stat *stat) | |
508 | { | |
509 | percpu_counter_destroy(&stat->cpu_cnt); | |
90d3839b PZ |
510 | } |
511 | ||
edcb0722 TH |
512 | /** |
513 | * blkg_stat_add - add a value to a blkg_stat | |
514 | * @stat: target blkg_stat | |
515 | * @val: value to add | |
516 | * | |
24bdb8ef TH |
517 | * Add @val to @stat. The caller must ensure that IRQ on the same CPU |
518 | * don't re-enter this function for the same counter. | |
edcb0722 TH |
519 | */ |
520 | static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val) | |
521 | { | |
104b4e51 | 522 | percpu_counter_add_batch(&stat->cpu_cnt, val, BLKG_STAT_CPU_BATCH); |
edcb0722 TH |
523 | } |
524 | ||
525 | /** | |
526 | * blkg_stat_read - read the current value of a blkg_stat | |
527 | * @stat: blkg_stat to read | |
edcb0722 TH |
528 | */ |
529 | static inline uint64_t blkg_stat_read(struct blkg_stat *stat) | |
530 | { | |
24bdb8ef | 531 | return percpu_counter_sum_positive(&stat->cpu_cnt); |
edcb0722 TH |
532 | } |
533 | ||
534 | /** | |
535 | * blkg_stat_reset - reset a blkg_stat | |
536 | * @stat: blkg_stat to reset | |
537 | */ | |
538 | static inline void blkg_stat_reset(struct blkg_stat *stat) | |
539 | { | |
24bdb8ef | 540 | percpu_counter_set(&stat->cpu_cnt, 0); |
e6269c44 | 541 | atomic64_set(&stat->aux_cnt, 0); |
edcb0722 TH |
542 | } |
543 | ||
16b3de66 | 544 | /** |
e6269c44 | 545 | * blkg_stat_add_aux - add a blkg_stat into another's aux count |
16b3de66 TH |
546 | * @to: the destination blkg_stat |
547 | * @from: the source | |
548 | * | |
e6269c44 | 549 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 550 | */ |
e6269c44 TH |
551 | static inline void blkg_stat_add_aux(struct blkg_stat *to, |
552 | struct blkg_stat *from) | |
16b3de66 | 553 | { |
e6269c44 TH |
554 | atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt), |
555 | &to->aux_cnt); | |
16b3de66 TH |
556 | } |
557 | ||
24bdb8ef | 558 | static inline int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp) |
90d3839b | 559 | { |
24bdb8ef TH |
560 | int i, ret; |
561 | ||
562 | for (i = 0; i < BLKG_RWSTAT_NR; i++) { | |
563 | ret = percpu_counter_init(&rwstat->cpu_cnt[i], 0, gfp); | |
564 | if (ret) { | |
565 | while (--i >= 0) | |
566 | percpu_counter_destroy(&rwstat->cpu_cnt[i]); | |
567 | return ret; | |
568 | } | |
569 | atomic64_set(&rwstat->aux_cnt[i], 0); | |
570 | } | |
571 | return 0; | |
572 | } | |
e6269c44 | 573 | |
24bdb8ef TH |
574 | static inline void blkg_rwstat_exit(struct blkg_rwstat *rwstat) |
575 | { | |
576 | int i; | |
e6269c44 TH |
577 | |
578 | for (i = 0; i < BLKG_RWSTAT_NR; i++) | |
24bdb8ef | 579 | percpu_counter_destroy(&rwstat->cpu_cnt[i]); |
90d3839b PZ |
580 | } |
581 | ||
edcb0722 TH |
582 | /** |
583 | * blkg_rwstat_add - add a value to a blkg_rwstat | |
584 | * @rwstat: target blkg_rwstat | |
ef295ecf | 585 | * @op: REQ_OP and flags |
edcb0722 TH |
586 | * @val: value to add |
587 | * | |
588 | * Add @val to @rwstat. The counters are chosen according to @rw. The | |
589 | * caller is responsible for synchronizing calls to this function. | |
590 | */ | |
591 | static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat, | |
ef295ecf | 592 | unsigned int op, uint64_t val) |
edcb0722 | 593 | { |
24bdb8ef | 594 | struct percpu_counter *cnt; |
edcb0722 | 595 | |
63a4cc24 | 596 | if (op_is_write(op)) |
24bdb8ef | 597 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_WRITE]; |
edcb0722 | 598 | else |
24bdb8ef TH |
599 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_READ]; |
600 | ||
104b4e51 | 601 | percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH); |
24bdb8ef | 602 | |
d71d9ae1 | 603 | if (op_is_sync(op)) |
24bdb8ef | 604 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_SYNC]; |
edcb0722 | 605 | else |
24bdb8ef | 606 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_ASYNC]; |
edcb0722 | 607 | |
104b4e51 | 608 | percpu_counter_add_batch(cnt, val, BLKG_STAT_CPU_BATCH); |
edcb0722 TH |
609 | } |
610 | ||
611 | /** | |
612 | * blkg_rwstat_read - read the current values of a blkg_rwstat | |
613 | * @rwstat: blkg_rwstat to read | |
614 | * | |
24bdb8ef | 615 | * Read the current snapshot of @rwstat and return it in the aux counts. |
edcb0722 | 616 | */ |
c94bed89 | 617 | static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat) |
edcb0722 | 618 | { |
24bdb8ef TH |
619 | struct blkg_rwstat result; |
620 | int i; | |
edcb0722 | 621 | |
24bdb8ef TH |
622 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
623 | atomic64_set(&result.aux_cnt[i], | |
624 | percpu_counter_sum_positive(&rwstat->cpu_cnt[i])); | |
625 | return result; | |
edcb0722 TH |
626 | } |
627 | ||
628 | /** | |
4d5e80a7 | 629 | * blkg_rwstat_total - read the total count of a blkg_rwstat |
edcb0722 TH |
630 | * @rwstat: blkg_rwstat to read |
631 | * | |
632 | * Return the total count of @rwstat regardless of the IO direction. This | |
633 | * function can be called without synchronization and takes care of u64 | |
634 | * atomicity. | |
635 | */ | |
4d5e80a7 | 636 | static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat) |
edcb0722 TH |
637 | { |
638 | struct blkg_rwstat tmp = blkg_rwstat_read(rwstat); | |
639 | ||
24bdb8ef TH |
640 | return atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) + |
641 | atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]); | |
edcb0722 TH |
642 | } |
643 | ||
644 | /** | |
645 | * blkg_rwstat_reset - reset a blkg_rwstat | |
646 | * @rwstat: blkg_rwstat to reset | |
647 | */ | |
648 | static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat) | |
649 | { | |
e6269c44 TH |
650 | int i; |
651 | ||
24bdb8ef TH |
652 | for (i = 0; i < BLKG_RWSTAT_NR; i++) { |
653 | percpu_counter_set(&rwstat->cpu_cnt[i], 0); | |
e6269c44 | 654 | atomic64_set(&rwstat->aux_cnt[i], 0); |
24bdb8ef | 655 | } |
edcb0722 TH |
656 | } |
657 | ||
16b3de66 | 658 | /** |
e6269c44 | 659 | * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count |
16b3de66 TH |
660 | * @to: the destination blkg_rwstat |
661 | * @from: the source | |
662 | * | |
e6269c44 | 663 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 664 | */ |
e6269c44 TH |
665 | static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to, |
666 | struct blkg_rwstat *from) | |
16b3de66 TH |
667 | { |
668 | struct blkg_rwstat v = blkg_rwstat_read(from); | |
669 | int i; | |
670 | ||
16b3de66 | 671 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
24bdb8ef TH |
672 | atomic64_add(atomic64_read(&v.aux_cnt[i]) + |
673 | atomic64_read(&from->aux_cnt[i]), | |
e6269c44 | 674 | &to->aux_cnt[i]); |
16b3de66 TH |
675 | } |
676 | ||
ae118896 TH |
677 | #ifdef CONFIG_BLK_DEV_THROTTLING |
678 | extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
679 | struct bio *bio); | |
680 | #else | |
681 | static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
682 | struct bio *bio) { return false; } | |
683 | #endif | |
684 | ||
685 | static inline bool blkcg_bio_issue_check(struct request_queue *q, | |
686 | struct bio *bio) | |
687 | { | |
688 | struct blkcg *blkcg; | |
689 | struct blkcg_gq *blkg; | |
690 | bool throtl = false; | |
691 | ||
692 | rcu_read_lock(); | |
693 | blkcg = bio_blkcg(bio); | |
694 | ||
007cc56b SL |
695 | /* associate blkcg if bio hasn't attached one */ |
696 | bio_associate_blkcg(bio, &blkcg->css); | |
697 | ||
ae118896 TH |
698 | blkg = blkg_lookup(blkcg, q); |
699 | if (unlikely(!blkg)) { | |
700 | spin_lock_irq(q->queue_lock); | |
701 | blkg = blkg_lookup_create(blkcg, q); | |
702 | if (IS_ERR(blkg)) | |
703 | blkg = NULL; | |
704 | spin_unlock_irq(q->queue_lock); | |
705 | } | |
706 | ||
707 | throtl = blk_throtl_bio(q, blkg, bio); | |
708 | ||
77ea7338 TH |
709 | if (!throtl) { |
710 | blkg = blkg ?: q->root_blkg; | |
ef295ecf | 711 | blkg_rwstat_add(&blkg->stat_bytes, bio->bi_opf, |
77ea7338 | 712 | bio->bi_iter.bi_size); |
ef295ecf | 713 | blkg_rwstat_add(&blkg->stat_ios, bio->bi_opf, 1); |
77ea7338 TH |
714 | } |
715 | ||
ae118896 TH |
716 | rcu_read_unlock(); |
717 | return !throtl; | |
718 | } | |
719 | ||
36558c8a TH |
720 | #else /* CONFIG_BLK_CGROUP */ |
721 | ||
efa7d1c7 TH |
722 | struct blkcg { |
723 | }; | |
2f5ea477 | 724 | |
f95a04af TH |
725 | struct blkg_policy_data { |
726 | }; | |
727 | ||
e48453c3 AA |
728 | struct blkcg_policy_data { |
729 | }; | |
730 | ||
3c798398 | 731 | struct blkcg_gq { |
2f5ea477 JA |
732 | }; |
733 | ||
3c798398 | 734 | struct blkcg_policy { |
3e252066 VG |
735 | }; |
736 | ||
496d5e75 TH |
737 | #define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL)) |
738 | ||
fd383c2d TH |
739 | static inline struct cgroup_subsys_state * |
740 | task_get_blkcg_css(struct task_struct *task) | |
741 | { | |
742 | return NULL; | |
743 | } | |
744 | ||
efa7d1c7 TH |
745 | #ifdef CONFIG_BLOCK |
746 | ||
3c798398 | 747 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; } |
5efd6113 TH |
748 | static inline int blkcg_init_queue(struct request_queue *q) { return 0; } |
749 | static inline void blkcg_drain_queue(struct request_queue *q) { } | |
750 | static inline void blkcg_exit_queue(struct request_queue *q) { } | |
d5bf0291 | 751 | static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; } |
3c798398 | 752 | static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { } |
a2b1693b | 753 | static inline int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 754 | const struct blkcg_policy *pol) { return 0; } |
a2b1693b | 755 | static inline void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 TH |
756 | const struct blkcg_policy *pol) { } |
757 | ||
b1208b56 | 758 | static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; } |
a051661c | 759 | |
f95a04af TH |
760 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
761 | struct blkcg_policy *pol) { return NULL; } | |
762 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; } | |
3c798398 TH |
763 | static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; } |
764 | static inline void blkg_get(struct blkcg_gq *blkg) { } | |
765 | static inline void blkg_put(struct blkcg_gq *blkg) { } | |
afc24d49 | 766 | |
a051661c TH |
767 | static inline struct request_list *blk_get_rl(struct request_queue *q, |
768 | struct bio *bio) { return &q->root_rl; } | |
769 | static inline void blk_put_rl(struct request_list *rl) { } | |
770 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { } | |
771 | static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; } | |
772 | ||
ae118896 TH |
773 | static inline bool blkcg_bio_issue_check(struct request_queue *q, |
774 | struct bio *bio) { return true; } | |
775 | ||
a051661c TH |
776 | #define blk_queue_for_each_rl(rl, q) \ |
777 | for ((rl) = &(q)->root_rl; (rl); (rl) = NULL) | |
778 | ||
efa7d1c7 | 779 | #endif /* CONFIG_BLOCK */ |
36558c8a TH |
780 | #endif /* CONFIG_BLK_CGROUP */ |
781 | #endif /* _BLK_CGROUP_H */ |