Commit | Line | Data |
---|---|---|
31e4c28d VG |
1 | #ifndef _BLK_CGROUP_H |
2 | #define _BLK_CGROUP_H | |
3 | /* | |
4 | * Common Block IO controller cgroup interface | |
5 | * | |
6 | * Based on ideas and code from CFQ, CFS and BFQ: | |
7 | * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk> | |
8 | * | |
9 | * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it> | |
10 | * Paolo Valente <paolo.valente@unimore.it> | |
11 | * | |
12 | * Copyright (C) 2009 Vivek Goyal <vgoyal@redhat.com> | |
13 | * Nauman Rafique <nauman@google.com> | |
14 | */ | |
15 | ||
16 | #include <linux/cgroup.h> | |
24bdb8ef | 17 | #include <linux/percpu_counter.h> |
829fdb50 | 18 | #include <linux/seq_file.h> |
a637120e | 19 | #include <linux/radix-tree.h> |
a051661c | 20 | #include <linux/blkdev.h> |
a5049a8a | 21 | #include <linux/atomic.h> |
31e4c28d | 22 | |
24bdb8ef TH |
23 | /* percpu_counter batch for blkg_[rw]stats, per-cpu drift doesn't matter */ |
24 | #define BLKG_STAT_CPU_BATCH (INT_MAX / 2) | |
25 | ||
9355aede VG |
26 | /* Max limits for throttle policy */ |
27 | #define THROTL_IOPS_MAX UINT_MAX | |
28 | ||
f48ec1d7 TH |
29 | #ifdef CONFIG_BLK_CGROUP |
30 | ||
edcb0722 TH |
31 | enum blkg_rwstat_type { |
32 | BLKG_RWSTAT_READ, | |
33 | BLKG_RWSTAT_WRITE, | |
34 | BLKG_RWSTAT_SYNC, | |
35 | BLKG_RWSTAT_ASYNC, | |
36 | ||
37 | BLKG_RWSTAT_NR, | |
38 | BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR, | |
303a3acb DS |
39 | }; |
40 | ||
a637120e TH |
41 | struct blkcg_gq; |
42 | ||
3c798398 | 43 | struct blkcg { |
36558c8a TH |
44 | struct cgroup_subsys_state css; |
45 | spinlock_t lock; | |
a637120e TH |
46 | |
47 | struct radix_tree_root blkg_tree; | |
48 | struct blkcg_gq *blkg_hint; | |
36558c8a | 49 | struct hlist_head blkg_list; |
9a9e8a26 | 50 | |
81437648 | 51 | struct blkcg_policy_data *cpd[BLKCG_MAX_POLS]; |
52ebea74 | 52 | |
7876f930 | 53 | struct list_head all_blkcgs_node; |
52ebea74 TH |
54 | #ifdef CONFIG_CGROUP_WRITEBACK |
55 | struct list_head cgwb_list; | |
56 | #endif | |
31e4c28d VG |
57 | }; |
58 | ||
e6269c44 TH |
59 | /* |
60 | * blkg_[rw]stat->aux_cnt is excluded for local stats but included for | |
24bdb8ef TH |
61 | * recursive. Used to carry stats of dead children, and, for blkg_rwstat, |
62 | * to carry result values from read and sum operations. | |
e6269c44 | 63 | */ |
edcb0722 | 64 | struct blkg_stat { |
24bdb8ef | 65 | struct percpu_counter cpu_cnt; |
e6269c44 | 66 | atomic64_t aux_cnt; |
edcb0722 TH |
67 | }; |
68 | ||
69 | struct blkg_rwstat { | |
24bdb8ef | 70 | struct percpu_counter cpu_cnt[BLKG_RWSTAT_NR]; |
e6269c44 | 71 | atomic64_t aux_cnt[BLKG_RWSTAT_NR]; |
edcb0722 TH |
72 | }; |
73 | ||
f95a04af TH |
74 | /* |
75 | * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a | |
76 | * request_queue (q). This is used by blkcg policies which need to track | |
77 | * information per blkcg - q pair. | |
78 | * | |
001bea73 TH |
79 | * There can be multiple active blkcg policies and each blkg:policy pair is |
80 | * represented by a blkg_policy_data which is allocated and freed by each | |
81 | * policy's pd_alloc/free_fn() methods. A policy can allocate private data | |
82 | * area by allocating larger data structure which embeds blkg_policy_data | |
83 | * at the beginning. | |
f95a04af | 84 | */ |
0381411e | 85 | struct blkg_policy_data { |
b276a876 | 86 | /* the blkg and policy id this per-policy data belongs to */ |
3c798398 | 87 | struct blkcg_gq *blkg; |
b276a876 | 88 | int plid; |
0381411e TH |
89 | }; |
90 | ||
e48453c3 | 91 | /* |
e4a9bde9 TH |
92 | * Policies that need to keep per-blkcg data which is independent from any |
93 | * request_queue associated to it should implement cpd_alloc/free_fn() | |
94 | * methods. A policy can allocate private data area by allocating larger | |
95 | * data structure which embeds blkcg_policy_data at the beginning. | |
96 | * cpd_init() is invoked to let each policy handle per-blkcg data. | |
e48453c3 AA |
97 | */ |
98 | struct blkcg_policy_data { | |
81437648 TH |
99 | /* the blkcg and policy id this per-policy data belongs to */ |
100 | struct blkcg *blkcg; | |
e48453c3 | 101 | int plid; |
e48453c3 AA |
102 | }; |
103 | ||
3c798398 TH |
104 | /* association between a blk cgroup and a request queue */ |
105 | struct blkcg_gq { | |
c875f4d0 | 106 | /* Pointer to the associated request_queue */ |
36558c8a TH |
107 | struct request_queue *q; |
108 | struct list_head q_node; | |
109 | struct hlist_node blkcg_node; | |
3c798398 | 110 | struct blkcg *blkcg; |
3c547865 | 111 | |
ce7acfea TH |
112 | /* |
113 | * Each blkg gets congested separately and the congestion state is | |
114 | * propagated to the matching bdi_writeback_congested. | |
115 | */ | |
116 | struct bdi_writeback_congested *wb_congested; | |
117 | ||
3c547865 TH |
118 | /* all non-root blkcg_gq's are guaranteed to have access to parent */ |
119 | struct blkcg_gq *parent; | |
120 | ||
a051661c TH |
121 | /* request allocation list for this blkcg-q pair */ |
122 | struct request_list rl; | |
3c547865 | 123 | |
1adaf3dd | 124 | /* reference count */ |
a5049a8a | 125 | atomic_t refcnt; |
22084190 | 126 | |
f427d909 TH |
127 | /* is this blkg online? protected by both blkcg and q locks */ |
128 | bool online; | |
129 | ||
77ea7338 TH |
130 | struct blkg_rwstat stat_bytes; |
131 | struct blkg_rwstat stat_ios; | |
132 | ||
36558c8a | 133 | struct blkg_policy_data *pd[BLKCG_MAX_POLS]; |
1adaf3dd | 134 | |
36558c8a | 135 | struct rcu_head rcu_head; |
31e4c28d VG |
136 | }; |
137 | ||
e4a9bde9 | 138 | typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp); |
81437648 | 139 | typedef void (blkcg_pol_init_cpd_fn)(struct blkcg_policy_data *cpd); |
e4a9bde9 | 140 | typedef void (blkcg_pol_free_cpd_fn)(struct blkcg_policy_data *cpd); |
69d7fde5 | 141 | typedef void (blkcg_pol_bind_cpd_fn)(struct blkcg_policy_data *cpd); |
001bea73 | 142 | typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node); |
a9520cd6 TH |
143 | typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd); |
144 | typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd); | |
145 | typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd); | |
001bea73 | 146 | typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd); |
a9520cd6 | 147 | typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd); |
3e252066 | 148 | |
3c798398 | 149 | struct blkcg_policy { |
36558c8a | 150 | int plid; |
36558c8a | 151 | /* cgroup files for the policy */ |
2ee867dc | 152 | struct cftype *dfl_cftypes; |
880f50e2 | 153 | struct cftype *legacy_cftypes; |
f9fcc2d3 TH |
154 | |
155 | /* operations */ | |
e4a9bde9 | 156 | blkcg_pol_alloc_cpd_fn *cpd_alloc_fn; |
e48453c3 | 157 | blkcg_pol_init_cpd_fn *cpd_init_fn; |
e4a9bde9 | 158 | blkcg_pol_free_cpd_fn *cpd_free_fn; |
69d7fde5 | 159 | blkcg_pol_bind_cpd_fn *cpd_bind_fn; |
e4a9bde9 | 160 | |
001bea73 | 161 | blkcg_pol_alloc_pd_fn *pd_alloc_fn; |
f9fcc2d3 | 162 | blkcg_pol_init_pd_fn *pd_init_fn; |
f427d909 TH |
163 | blkcg_pol_online_pd_fn *pd_online_fn; |
164 | blkcg_pol_offline_pd_fn *pd_offline_fn; | |
001bea73 | 165 | blkcg_pol_free_pd_fn *pd_free_fn; |
f9fcc2d3 | 166 | blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn; |
3e252066 VG |
167 | }; |
168 | ||
3c798398 | 169 | extern struct blkcg blkcg_root; |
496d5e75 | 170 | extern struct cgroup_subsys_state * const blkcg_root_css; |
36558c8a | 171 | |
24f29046 TH |
172 | struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg, |
173 | struct request_queue *q, bool update_hint); | |
3c798398 TH |
174 | struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, |
175 | struct request_queue *q); | |
36558c8a TH |
176 | int blkcg_init_queue(struct request_queue *q); |
177 | void blkcg_drain_queue(struct request_queue *q); | |
178 | void blkcg_exit_queue(struct request_queue *q); | |
5efd6113 | 179 | |
3e252066 | 180 | /* Blkio controller policy registration */ |
d5bf0291 | 181 | int blkcg_policy_register(struct blkcg_policy *pol); |
3c798398 | 182 | void blkcg_policy_unregister(struct blkcg_policy *pol); |
36558c8a | 183 | int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 184 | const struct blkcg_policy *pol); |
36558c8a | 185 | void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 | 186 | const struct blkcg_policy *pol); |
3e252066 | 187 | |
dd165eb3 | 188 | const char *blkg_dev_name(struct blkcg_gq *blkg); |
3c798398 | 189 | void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg, |
f95a04af TH |
190 | u64 (*prfill)(struct seq_file *, |
191 | struct blkg_policy_data *, int), | |
3c798398 | 192 | const struct blkcg_policy *pol, int data, |
ec399347 | 193 | bool show_total); |
f95a04af TH |
194 | u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v); |
195 | u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
829fdb50 | 196 | const struct blkg_rwstat *rwstat); |
f95a04af TH |
197 | u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off); |
198 | u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
199 | int off); | |
77ea7338 TH |
200 | int blkg_print_stat_bytes(struct seq_file *sf, void *v); |
201 | int blkg_print_stat_ios(struct seq_file *sf, void *v); | |
202 | int blkg_print_stat_bytes_recursive(struct seq_file *sf, void *v); | |
203 | int blkg_print_stat_ios_recursive(struct seq_file *sf, void *v); | |
829fdb50 | 204 | |
f12c74ca TH |
205 | u64 blkg_stat_recursive_sum(struct blkcg_gq *blkg, |
206 | struct blkcg_policy *pol, int off); | |
207 | struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkcg_gq *blkg, | |
208 | struct blkcg_policy *pol, int off); | |
16b3de66 | 209 | |
829fdb50 | 210 | struct blkg_conf_ctx { |
36558c8a | 211 | struct gendisk *disk; |
3c798398 | 212 | struct blkcg_gq *blkg; |
36aa9e5f | 213 | char *body; |
829fdb50 TH |
214 | }; |
215 | ||
3c798398 | 216 | int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol, |
36aa9e5f | 217 | char *input, struct blkg_conf_ctx *ctx); |
829fdb50 TH |
218 | void blkg_conf_finish(struct blkg_conf_ctx *ctx); |
219 | ||
220 | ||
a7c6d554 TH |
221 | static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css) |
222 | { | |
223 | return css ? container_of(css, struct blkcg, css) : NULL; | |
224 | } | |
225 | ||
b1208b56 TH |
226 | static inline struct blkcg *task_blkcg(struct task_struct *tsk) |
227 | { | |
c165b3e3 | 228 | return css_to_blkcg(task_css(tsk, io_cgrp_id)); |
b1208b56 TH |
229 | } |
230 | ||
231 | static inline struct blkcg *bio_blkcg(struct bio *bio) | |
232 | { | |
233 | if (bio && bio->bi_css) | |
a7c6d554 | 234 | return css_to_blkcg(bio->bi_css); |
b1208b56 TH |
235 | return task_blkcg(current); |
236 | } | |
237 | ||
fd383c2d TH |
238 | static inline struct cgroup_subsys_state * |
239 | task_get_blkcg_css(struct task_struct *task) | |
240 | { | |
c165b3e3 | 241 | return task_get_css(task, io_cgrp_id); |
fd383c2d TH |
242 | } |
243 | ||
3c547865 TH |
244 | /** |
245 | * blkcg_parent - get the parent of a blkcg | |
246 | * @blkcg: blkcg of interest | |
247 | * | |
248 | * Return the parent blkcg of @blkcg. Can be called anytime. | |
249 | */ | |
250 | static inline struct blkcg *blkcg_parent(struct blkcg *blkcg) | |
251 | { | |
5c9d535b | 252 | return css_to_blkcg(blkcg->css.parent); |
3c547865 TH |
253 | } |
254 | ||
24f29046 TH |
255 | /** |
256 | * __blkg_lookup - internal version of blkg_lookup() | |
257 | * @blkcg: blkcg of interest | |
258 | * @q: request_queue of interest | |
259 | * @update_hint: whether to update lookup hint with the result or not | |
260 | * | |
261 | * This is internal version and shouldn't be used by policy | |
262 | * implementations. Looks up blkgs for the @blkcg - @q pair regardless of | |
263 | * @q's bypass state. If @update_hint is %true, the caller should be | |
264 | * holding @q->queue_lock and lookup hint is updated on success. | |
265 | */ | |
266 | static inline struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg, | |
267 | struct request_queue *q, | |
268 | bool update_hint) | |
269 | { | |
270 | struct blkcg_gq *blkg; | |
271 | ||
85b6bc9d TH |
272 | if (blkcg == &blkcg_root) |
273 | return q->root_blkg; | |
274 | ||
24f29046 TH |
275 | blkg = rcu_dereference(blkcg->blkg_hint); |
276 | if (blkg && blkg->q == q) | |
277 | return blkg; | |
278 | ||
279 | return blkg_lookup_slowpath(blkcg, q, update_hint); | |
280 | } | |
281 | ||
282 | /** | |
283 | * blkg_lookup - lookup blkg for the specified blkcg - q pair | |
284 | * @blkcg: blkcg of interest | |
285 | * @q: request_queue of interest | |
286 | * | |
287 | * Lookup blkg for the @blkcg - @q pair. This function should be called | |
288 | * under RCU read lock and is guaranteed to return %NULL if @q is bypassing | |
289 | * - see blk_queue_bypass_start() for details. | |
290 | */ | |
291 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, | |
292 | struct request_queue *q) | |
293 | { | |
294 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
295 | ||
296 | if (unlikely(blk_queue_bypass(q))) | |
297 | return NULL; | |
298 | return __blkg_lookup(blkcg, q, false); | |
299 | } | |
300 | ||
0381411e TH |
301 | /** |
302 | * blkg_to_pdata - get policy private data | |
303 | * @blkg: blkg of interest | |
304 | * @pol: policy of interest | |
305 | * | |
306 | * Return pointer to private data associated with the @blkg-@pol pair. | |
307 | */ | |
f95a04af TH |
308 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
309 | struct blkcg_policy *pol) | |
0381411e | 310 | { |
f95a04af | 311 | return blkg ? blkg->pd[pol->plid] : NULL; |
0381411e TH |
312 | } |
313 | ||
e48453c3 AA |
314 | static inline struct blkcg_policy_data *blkcg_to_cpd(struct blkcg *blkcg, |
315 | struct blkcg_policy *pol) | |
316 | { | |
81437648 | 317 | return blkcg ? blkcg->cpd[pol->plid] : NULL; |
e48453c3 AA |
318 | } |
319 | ||
0381411e TH |
320 | /** |
321 | * pdata_to_blkg - get blkg associated with policy private data | |
f95a04af | 322 | * @pd: policy private data of interest |
0381411e | 323 | * |
f95a04af | 324 | * @pd is policy private data. Determine the blkg it's associated with. |
0381411e | 325 | */ |
f95a04af | 326 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) |
0381411e | 327 | { |
f95a04af | 328 | return pd ? pd->blkg : NULL; |
0381411e TH |
329 | } |
330 | ||
81437648 TH |
331 | static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd) |
332 | { | |
333 | return cpd ? cpd->blkcg : NULL; | |
334 | } | |
335 | ||
54e7ed12 TH |
336 | /** |
337 | * blkg_path - format cgroup path of blkg | |
338 | * @blkg: blkg of interest | |
339 | * @buf: target buffer | |
340 | * @buflen: target buffer length | |
341 | * | |
342 | * Format the path of the cgroup of @blkg into @buf. | |
343 | */ | |
3c798398 | 344 | static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen) |
afc24d49 | 345 | { |
e61734c5 | 346 | char *p; |
54e7ed12 | 347 | |
e61734c5 TH |
348 | p = cgroup_path(blkg->blkcg->css.cgroup, buf, buflen); |
349 | if (!p) { | |
54e7ed12 | 350 | strncpy(buf, "<unavailable>", buflen); |
e61734c5 TH |
351 | return -ENAMETOOLONG; |
352 | } | |
353 | ||
354 | memmove(buf, p, buf + buflen - p); | |
355 | return 0; | |
afc24d49 VG |
356 | } |
357 | ||
1adaf3dd TH |
358 | /** |
359 | * blkg_get - get a blkg reference | |
360 | * @blkg: blkg to get | |
361 | * | |
a5049a8a | 362 | * The caller should be holding an existing reference. |
1adaf3dd | 363 | */ |
3c798398 | 364 | static inline void blkg_get(struct blkcg_gq *blkg) |
1adaf3dd | 365 | { |
a5049a8a TH |
366 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
367 | atomic_inc(&blkg->refcnt); | |
1adaf3dd TH |
368 | } |
369 | ||
2a4fd070 | 370 | void __blkg_release_rcu(struct rcu_head *rcu); |
1adaf3dd TH |
371 | |
372 | /** | |
373 | * blkg_put - put a blkg reference | |
374 | * @blkg: blkg to put | |
1adaf3dd | 375 | */ |
3c798398 | 376 | static inline void blkg_put(struct blkcg_gq *blkg) |
1adaf3dd | 377 | { |
a5049a8a TH |
378 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
379 | if (atomic_dec_and_test(&blkg->refcnt)) | |
2a4fd070 | 380 | call_rcu(&blkg->rcu_head, __blkg_release_rcu); |
1adaf3dd TH |
381 | } |
382 | ||
dd4a4ffc TH |
383 | /** |
384 | * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants | |
385 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 386 | * @pos_css: used for iteration |
dd4a4ffc TH |
387 | * @p_blkg: target blkg to walk descendants of |
388 | * | |
389 | * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU | |
390 | * read locked. If called under either blkcg or queue lock, the iteration | |
391 | * is guaranteed to include all and only online blkgs. The caller may | |
492eb21b | 392 | * update @pos_css by calling css_rightmost_descendant() to skip subtree. |
bd8815a6 | 393 | * @p_blkg is included in the iteration and the first node to be visited. |
dd4a4ffc | 394 | */ |
492eb21b TH |
395 | #define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \ |
396 | css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \ | |
397 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
dd4a4ffc TH |
398 | (p_blkg)->q, false))) |
399 | ||
aa539cb3 TH |
400 | /** |
401 | * blkg_for_each_descendant_post - post-order walk of a blkg's descendants | |
402 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 403 | * @pos_css: used for iteration |
aa539cb3 TH |
404 | * @p_blkg: target blkg to walk descendants of |
405 | * | |
406 | * Similar to blkg_for_each_descendant_pre() but performs post-order | |
bd8815a6 TH |
407 | * traversal instead. Synchronization rules are the same. @p_blkg is |
408 | * included in the iteration and the last node to be visited. | |
aa539cb3 | 409 | */ |
492eb21b TH |
410 | #define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \ |
411 | css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \ | |
412 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
aa539cb3 TH |
413 | (p_blkg)->q, false))) |
414 | ||
a051661c TH |
415 | /** |
416 | * blk_get_rl - get request_list to use | |
417 | * @q: request_queue of interest | |
418 | * @bio: bio which will be attached to the allocated request (may be %NULL) | |
419 | * | |
420 | * The caller wants to allocate a request from @q to use for @bio. Find | |
421 | * the request_list to use and obtain a reference on it. Should be called | |
422 | * under queue_lock. This function is guaranteed to return non-%NULL | |
423 | * request_list. | |
424 | */ | |
425 | static inline struct request_list *blk_get_rl(struct request_queue *q, | |
426 | struct bio *bio) | |
427 | { | |
428 | struct blkcg *blkcg; | |
429 | struct blkcg_gq *blkg; | |
430 | ||
431 | rcu_read_lock(); | |
432 | ||
433 | blkcg = bio_blkcg(bio); | |
434 | ||
435 | /* bypass blkg lookup and use @q->root_rl directly for root */ | |
436 | if (blkcg == &blkcg_root) | |
437 | goto root_rl; | |
438 | ||
439 | /* | |
440 | * Try to use blkg->rl. blkg lookup may fail under memory pressure | |
441 | * or if either the blkcg or queue is going away. Fall back to | |
442 | * root_rl in such cases. | |
443 | */ | |
ae118896 TH |
444 | blkg = blkg_lookup(blkcg, q); |
445 | if (unlikely(!blkg)) | |
a051661c TH |
446 | goto root_rl; |
447 | ||
448 | blkg_get(blkg); | |
449 | rcu_read_unlock(); | |
450 | return &blkg->rl; | |
451 | root_rl: | |
452 | rcu_read_unlock(); | |
453 | return &q->root_rl; | |
454 | } | |
455 | ||
456 | /** | |
457 | * blk_put_rl - put request_list | |
458 | * @rl: request_list to put | |
459 | * | |
460 | * Put the reference acquired by blk_get_rl(). Should be called under | |
461 | * queue_lock. | |
462 | */ | |
463 | static inline void blk_put_rl(struct request_list *rl) | |
464 | { | |
401efbf8 | 465 | if (rl->blkg->blkcg != &blkcg_root) |
a051661c TH |
466 | blkg_put(rl->blkg); |
467 | } | |
468 | ||
469 | /** | |
470 | * blk_rq_set_rl - associate a request with a request_list | |
471 | * @rq: request of interest | |
472 | * @rl: target request_list | |
473 | * | |
474 | * Associate @rq with @rl so that accounting and freeing can know the | |
475 | * request_list @rq came from. | |
476 | */ | |
477 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) | |
478 | { | |
479 | rq->rl = rl; | |
480 | } | |
481 | ||
482 | /** | |
483 | * blk_rq_rl - return the request_list a request came from | |
484 | * @rq: request of interest | |
485 | * | |
486 | * Return the request_list @rq is allocated from. | |
487 | */ | |
488 | static inline struct request_list *blk_rq_rl(struct request *rq) | |
489 | { | |
490 | return rq->rl; | |
491 | } | |
492 | ||
493 | struct request_list *__blk_queue_next_rl(struct request_list *rl, | |
494 | struct request_queue *q); | |
495 | /** | |
496 | * blk_queue_for_each_rl - iterate through all request_lists of a request_queue | |
497 | * | |
498 | * Should be used under queue_lock. | |
499 | */ | |
500 | #define blk_queue_for_each_rl(rl, q) \ | |
501 | for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q))) | |
502 | ||
24bdb8ef | 503 | static inline int blkg_stat_init(struct blkg_stat *stat, gfp_t gfp) |
90d3839b | 504 | { |
24bdb8ef TH |
505 | int ret; |
506 | ||
507 | ret = percpu_counter_init(&stat->cpu_cnt, 0, gfp); | |
508 | if (ret) | |
509 | return ret; | |
510 | ||
e6269c44 | 511 | atomic64_set(&stat->aux_cnt, 0); |
24bdb8ef TH |
512 | return 0; |
513 | } | |
514 | ||
515 | static inline void blkg_stat_exit(struct blkg_stat *stat) | |
516 | { | |
517 | percpu_counter_destroy(&stat->cpu_cnt); | |
90d3839b PZ |
518 | } |
519 | ||
edcb0722 TH |
520 | /** |
521 | * blkg_stat_add - add a value to a blkg_stat | |
522 | * @stat: target blkg_stat | |
523 | * @val: value to add | |
524 | * | |
24bdb8ef TH |
525 | * Add @val to @stat. The caller must ensure that IRQ on the same CPU |
526 | * don't re-enter this function for the same counter. | |
edcb0722 TH |
527 | */ |
528 | static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val) | |
529 | { | |
24bdb8ef | 530 | __percpu_counter_add(&stat->cpu_cnt, val, BLKG_STAT_CPU_BATCH); |
edcb0722 TH |
531 | } |
532 | ||
533 | /** | |
534 | * blkg_stat_read - read the current value of a blkg_stat | |
535 | * @stat: blkg_stat to read | |
edcb0722 TH |
536 | */ |
537 | static inline uint64_t blkg_stat_read(struct blkg_stat *stat) | |
538 | { | |
24bdb8ef | 539 | return percpu_counter_sum_positive(&stat->cpu_cnt); |
edcb0722 TH |
540 | } |
541 | ||
542 | /** | |
543 | * blkg_stat_reset - reset a blkg_stat | |
544 | * @stat: blkg_stat to reset | |
545 | */ | |
546 | static inline void blkg_stat_reset(struct blkg_stat *stat) | |
547 | { | |
24bdb8ef | 548 | percpu_counter_set(&stat->cpu_cnt, 0); |
e6269c44 | 549 | atomic64_set(&stat->aux_cnt, 0); |
edcb0722 TH |
550 | } |
551 | ||
16b3de66 | 552 | /** |
e6269c44 | 553 | * blkg_stat_add_aux - add a blkg_stat into another's aux count |
16b3de66 TH |
554 | * @to: the destination blkg_stat |
555 | * @from: the source | |
556 | * | |
e6269c44 | 557 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 558 | */ |
e6269c44 TH |
559 | static inline void blkg_stat_add_aux(struct blkg_stat *to, |
560 | struct blkg_stat *from) | |
16b3de66 | 561 | { |
e6269c44 TH |
562 | atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt), |
563 | &to->aux_cnt); | |
16b3de66 TH |
564 | } |
565 | ||
24bdb8ef | 566 | static inline int blkg_rwstat_init(struct blkg_rwstat *rwstat, gfp_t gfp) |
90d3839b | 567 | { |
24bdb8ef TH |
568 | int i, ret; |
569 | ||
570 | for (i = 0; i < BLKG_RWSTAT_NR; i++) { | |
571 | ret = percpu_counter_init(&rwstat->cpu_cnt[i], 0, gfp); | |
572 | if (ret) { | |
573 | while (--i >= 0) | |
574 | percpu_counter_destroy(&rwstat->cpu_cnt[i]); | |
575 | return ret; | |
576 | } | |
577 | atomic64_set(&rwstat->aux_cnt[i], 0); | |
578 | } | |
579 | return 0; | |
580 | } | |
e6269c44 | 581 | |
24bdb8ef TH |
582 | static inline void blkg_rwstat_exit(struct blkg_rwstat *rwstat) |
583 | { | |
584 | int i; | |
e6269c44 TH |
585 | |
586 | for (i = 0; i < BLKG_RWSTAT_NR; i++) | |
24bdb8ef | 587 | percpu_counter_destroy(&rwstat->cpu_cnt[i]); |
90d3839b PZ |
588 | } |
589 | ||
edcb0722 TH |
590 | /** |
591 | * blkg_rwstat_add - add a value to a blkg_rwstat | |
592 | * @rwstat: target blkg_rwstat | |
63a4cc24 MC |
593 | * @op: REQ_OP |
594 | * @op_flags: rq_flag_bits | |
edcb0722 TH |
595 | * @val: value to add |
596 | * | |
597 | * Add @val to @rwstat. The counters are chosen according to @rw. The | |
598 | * caller is responsible for synchronizing calls to this function. | |
599 | */ | |
600 | static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat, | |
63a4cc24 | 601 | int op, int op_flags, uint64_t val) |
edcb0722 | 602 | { |
24bdb8ef | 603 | struct percpu_counter *cnt; |
edcb0722 | 604 | |
63a4cc24 | 605 | if (op_is_write(op)) |
24bdb8ef | 606 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_WRITE]; |
edcb0722 | 607 | else |
24bdb8ef TH |
608 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_READ]; |
609 | ||
610 | __percpu_counter_add(cnt, val, BLKG_STAT_CPU_BATCH); | |
611 | ||
63a4cc24 | 612 | if (op_flags & REQ_SYNC) |
24bdb8ef | 613 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_SYNC]; |
edcb0722 | 614 | else |
24bdb8ef | 615 | cnt = &rwstat->cpu_cnt[BLKG_RWSTAT_ASYNC]; |
edcb0722 | 616 | |
24bdb8ef | 617 | __percpu_counter_add(cnt, val, BLKG_STAT_CPU_BATCH); |
edcb0722 TH |
618 | } |
619 | ||
620 | /** | |
621 | * blkg_rwstat_read - read the current values of a blkg_rwstat | |
622 | * @rwstat: blkg_rwstat to read | |
623 | * | |
24bdb8ef | 624 | * Read the current snapshot of @rwstat and return it in the aux counts. |
edcb0722 | 625 | */ |
c94bed89 | 626 | static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat) |
edcb0722 | 627 | { |
24bdb8ef TH |
628 | struct blkg_rwstat result; |
629 | int i; | |
edcb0722 | 630 | |
24bdb8ef TH |
631 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
632 | atomic64_set(&result.aux_cnt[i], | |
633 | percpu_counter_sum_positive(&rwstat->cpu_cnt[i])); | |
634 | return result; | |
edcb0722 TH |
635 | } |
636 | ||
637 | /** | |
4d5e80a7 | 638 | * blkg_rwstat_total - read the total count of a blkg_rwstat |
edcb0722 TH |
639 | * @rwstat: blkg_rwstat to read |
640 | * | |
641 | * Return the total count of @rwstat regardless of the IO direction. This | |
642 | * function can be called without synchronization and takes care of u64 | |
643 | * atomicity. | |
644 | */ | |
4d5e80a7 | 645 | static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat) |
edcb0722 TH |
646 | { |
647 | struct blkg_rwstat tmp = blkg_rwstat_read(rwstat); | |
648 | ||
24bdb8ef TH |
649 | return atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_READ]) + |
650 | atomic64_read(&tmp.aux_cnt[BLKG_RWSTAT_WRITE]); | |
edcb0722 TH |
651 | } |
652 | ||
653 | /** | |
654 | * blkg_rwstat_reset - reset a blkg_rwstat | |
655 | * @rwstat: blkg_rwstat to reset | |
656 | */ | |
657 | static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat) | |
658 | { | |
e6269c44 TH |
659 | int i; |
660 | ||
24bdb8ef TH |
661 | for (i = 0; i < BLKG_RWSTAT_NR; i++) { |
662 | percpu_counter_set(&rwstat->cpu_cnt[i], 0); | |
e6269c44 | 663 | atomic64_set(&rwstat->aux_cnt[i], 0); |
24bdb8ef | 664 | } |
edcb0722 TH |
665 | } |
666 | ||
16b3de66 | 667 | /** |
e6269c44 | 668 | * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count |
16b3de66 TH |
669 | * @to: the destination blkg_rwstat |
670 | * @from: the source | |
671 | * | |
e6269c44 | 672 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 673 | */ |
e6269c44 TH |
674 | static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to, |
675 | struct blkg_rwstat *from) | |
16b3de66 TH |
676 | { |
677 | struct blkg_rwstat v = blkg_rwstat_read(from); | |
678 | int i; | |
679 | ||
16b3de66 | 680 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
24bdb8ef TH |
681 | atomic64_add(atomic64_read(&v.aux_cnt[i]) + |
682 | atomic64_read(&from->aux_cnt[i]), | |
e6269c44 | 683 | &to->aux_cnt[i]); |
16b3de66 TH |
684 | } |
685 | ||
ae118896 TH |
686 | #ifdef CONFIG_BLK_DEV_THROTTLING |
687 | extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
688 | struct bio *bio); | |
689 | #else | |
690 | static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
691 | struct bio *bio) { return false; } | |
692 | #endif | |
693 | ||
694 | static inline bool blkcg_bio_issue_check(struct request_queue *q, | |
695 | struct bio *bio) | |
696 | { | |
697 | struct blkcg *blkcg; | |
698 | struct blkcg_gq *blkg; | |
699 | bool throtl = false; | |
700 | ||
701 | rcu_read_lock(); | |
702 | blkcg = bio_blkcg(bio); | |
703 | ||
704 | blkg = blkg_lookup(blkcg, q); | |
705 | if (unlikely(!blkg)) { | |
706 | spin_lock_irq(q->queue_lock); | |
707 | blkg = blkg_lookup_create(blkcg, q); | |
708 | if (IS_ERR(blkg)) | |
709 | blkg = NULL; | |
710 | spin_unlock_irq(q->queue_lock); | |
711 | } | |
712 | ||
713 | throtl = blk_throtl_bio(q, blkg, bio); | |
714 | ||
77ea7338 TH |
715 | if (!throtl) { |
716 | blkg = blkg ?: q->root_blkg; | |
1eff9d32 | 717 | blkg_rwstat_add(&blkg->stat_bytes, bio_op(bio), bio->bi_opf, |
77ea7338 | 718 | bio->bi_iter.bi_size); |
1eff9d32 | 719 | blkg_rwstat_add(&blkg->stat_ios, bio_op(bio), bio->bi_opf, 1); |
77ea7338 TH |
720 | } |
721 | ||
ae118896 TH |
722 | rcu_read_unlock(); |
723 | return !throtl; | |
724 | } | |
725 | ||
36558c8a TH |
726 | #else /* CONFIG_BLK_CGROUP */ |
727 | ||
efa7d1c7 TH |
728 | struct blkcg { |
729 | }; | |
2f5ea477 | 730 | |
f95a04af TH |
731 | struct blkg_policy_data { |
732 | }; | |
733 | ||
e48453c3 AA |
734 | struct blkcg_policy_data { |
735 | }; | |
736 | ||
3c798398 | 737 | struct blkcg_gq { |
2f5ea477 JA |
738 | }; |
739 | ||
3c798398 | 740 | struct blkcg_policy { |
3e252066 VG |
741 | }; |
742 | ||
496d5e75 TH |
743 | #define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL)) |
744 | ||
fd383c2d TH |
745 | static inline struct cgroup_subsys_state * |
746 | task_get_blkcg_css(struct task_struct *task) | |
747 | { | |
748 | return NULL; | |
749 | } | |
750 | ||
efa7d1c7 TH |
751 | #ifdef CONFIG_BLOCK |
752 | ||
3c798398 | 753 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; } |
5efd6113 TH |
754 | static inline int blkcg_init_queue(struct request_queue *q) { return 0; } |
755 | static inline void blkcg_drain_queue(struct request_queue *q) { } | |
756 | static inline void blkcg_exit_queue(struct request_queue *q) { } | |
d5bf0291 | 757 | static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; } |
3c798398 | 758 | static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { } |
a2b1693b | 759 | static inline int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 760 | const struct blkcg_policy *pol) { return 0; } |
a2b1693b | 761 | static inline void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 TH |
762 | const struct blkcg_policy *pol) { } |
763 | ||
b1208b56 | 764 | static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; } |
a051661c | 765 | |
f95a04af TH |
766 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
767 | struct blkcg_policy *pol) { return NULL; } | |
768 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; } | |
3c798398 TH |
769 | static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; } |
770 | static inline void blkg_get(struct blkcg_gq *blkg) { } | |
771 | static inline void blkg_put(struct blkcg_gq *blkg) { } | |
afc24d49 | 772 | |
a051661c TH |
773 | static inline struct request_list *blk_get_rl(struct request_queue *q, |
774 | struct bio *bio) { return &q->root_rl; } | |
775 | static inline void blk_put_rl(struct request_list *rl) { } | |
776 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { } | |
777 | static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; } | |
778 | ||
ae118896 TH |
779 | static inline bool blkcg_bio_issue_check(struct request_queue *q, |
780 | struct bio *bio) { return true; } | |
781 | ||
a051661c TH |
782 | #define blk_queue_for_each_rl(rl, q) \ |
783 | for ((rl) = &(q)->root_rl; (rl); (rl) = NULL) | |
784 | ||
efa7d1c7 | 785 | #endif /* CONFIG_BLOCK */ |
36558c8a TH |
786 | #endif /* CONFIG_BLK_CGROUP */ |
787 | #endif /* _BLK_CGROUP_H */ |