]>
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> | |
575969a0 | 17 | #include <linux/u64_stats_sync.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 | |
9355aede VG |
23 | /* Max limits for throttle policy */ |
24 | #define THROTL_IOPS_MAX UINT_MAX | |
25 | ||
f48ec1d7 TH |
26 | #ifdef CONFIG_BLK_CGROUP |
27 | ||
edcb0722 TH |
28 | enum blkg_rwstat_type { |
29 | BLKG_RWSTAT_READ, | |
30 | BLKG_RWSTAT_WRITE, | |
31 | BLKG_RWSTAT_SYNC, | |
32 | BLKG_RWSTAT_ASYNC, | |
33 | ||
34 | BLKG_RWSTAT_NR, | |
35 | BLKG_RWSTAT_TOTAL = BLKG_RWSTAT_NR, | |
303a3acb DS |
36 | }; |
37 | ||
a637120e TH |
38 | struct blkcg_gq; |
39 | ||
3c798398 | 40 | struct blkcg { |
36558c8a TH |
41 | struct cgroup_subsys_state css; |
42 | spinlock_t lock; | |
a637120e TH |
43 | |
44 | struct radix_tree_root blkg_tree; | |
45 | struct blkcg_gq *blkg_hint; | |
36558c8a | 46 | struct hlist_head blkg_list; |
9a9e8a26 | 47 | |
81437648 | 48 | struct blkcg_policy_data *cpd[BLKCG_MAX_POLS]; |
52ebea74 | 49 | |
7876f930 | 50 | struct list_head all_blkcgs_node; |
52ebea74 TH |
51 | #ifdef CONFIG_CGROUP_WRITEBACK |
52 | struct list_head cgwb_list; | |
53 | #endif | |
31e4c28d VG |
54 | }; |
55 | ||
e6269c44 TH |
56 | /* |
57 | * blkg_[rw]stat->aux_cnt is excluded for local stats but included for | |
58 | * recursive. Used to carry stats of dead children. | |
59 | */ | |
edcb0722 TH |
60 | struct blkg_stat { |
61 | struct u64_stats_sync syncp; | |
62 | uint64_t cnt; | |
e6269c44 | 63 | atomic64_t aux_cnt; |
edcb0722 TH |
64 | }; |
65 | ||
66 | struct blkg_rwstat { | |
67 | struct u64_stats_sync syncp; | |
68 | uint64_t cnt[BLKG_RWSTAT_NR]; | |
e6269c44 | 69 | atomic64_t aux_cnt[BLKG_RWSTAT_NR]; |
edcb0722 TH |
70 | }; |
71 | ||
f95a04af TH |
72 | /* |
73 | * A blkcg_gq (blkg) is association between a block cgroup (blkcg) and a | |
74 | * request_queue (q). This is used by blkcg policies which need to track | |
75 | * information per blkcg - q pair. | |
76 | * | |
001bea73 TH |
77 | * There can be multiple active blkcg policies and each blkg:policy pair is |
78 | * represented by a blkg_policy_data which is allocated and freed by each | |
79 | * policy's pd_alloc/free_fn() methods. A policy can allocate private data | |
80 | * area by allocating larger data structure which embeds blkg_policy_data | |
81 | * at the beginning. | |
f95a04af | 82 | */ |
0381411e | 83 | struct blkg_policy_data { |
b276a876 | 84 | /* the blkg and policy id this per-policy data belongs to */ |
3c798398 | 85 | struct blkcg_gq *blkg; |
b276a876 | 86 | int plid; |
0381411e TH |
87 | }; |
88 | ||
e48453c3 | 89 | /* |
e4a9bde9 TH |
90 | * Policies that need to keep per-blkcg data which is independent from any |
91 | * request_queue associated to it should implement cpd_alloc/free_fn() | |
92 | * methods. A policy can allocate private data area by allocating larger | |
93 | * data structure which embeds blkcg_policy_data at the beginning. | |
94 | * cpd_init() is invoked to let each policy handle per-blkcg data. | |
e48453c3 AA |
95 | */ |
96 | struct blkcg_policy_data { | |
81437648 TH |
97 | /* the blkcg and policy id this per-policy data belongs to */ |
98 | struct blkcg *blkcg; | |
e48453c3 | 99 | int plid; |
e48453c3 AA |
100 | }; |
101 | ||
3c798398 TH |
102 | /* association between a blk cgroup and a request queue */ |
103 | struct blkcg_gq { | |
c875f4d0 | 104 | /* Pointer to the associated request_queue */ |
36558c8a TH |
105 | struct request_queue *q; |
106 | struct list_head q_node; | |
107 | struct hlist_node blkcg_node; | |
3c798398 | 108 | struct blkcg *blkcg; |
3c547865 | 109 | |
ce7acfea TH |
110 | /* |
111 | * Each blkg gets congested separately and the congestion state is | |
112 | * propagated to the matching bdi_writeback_congested. | |
113 | */ | |
114 | struct bdi_writeback_congested *wb_congested; | |
115 | ||
3c547865 TH |
116 | /* all non-root blkcg_gq's are guaranteed to have access to parent */ |
117 | struct blkcg_gq *parent; | |
118 | ||
a051661c TH |
119 | /* request allocation list for this blkcg-q pair */ |
120 | struct request_list rl; | |
3c547865 | 121 | |
1adaf3dd | 122 | /* reference count */ |
a5049a8a | 123 | atomic_t refcnt; |
22084190 | 124 | |
f427d909 TH |
125 | /* is this blkg online? protected by both blkcg and q locks */ |
126 | bool online; | |
127 | ||
36558c8a | 128 | struct blkg_policy_data *pd[BLKCG_MAX_POLS]; |
1adaf3dd | 129 | |
36558c8a | 130 | struct rcu_head rcu_head; |
31e4c28d VG |
131 | }; |
132 | ||
e4a9bde9 | 133 | typedef struct blkcg_policy_data *(blkcg_pol_alloc_cpd_fn)(gfp_t gfp); |
81437648 | 134 | typedef void (blkcg_pol_init_cpd_fn)(struct blkcg_policy_data *cpd); |
e4a9bde9 | 135 | typedef void (blkcg_pol_free_cpd_fn)(struct blkcg_policy_data *cpd); |
001bea73 | 136 | typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, int node); |
a9520cd6 TH |
137 | typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd); |
138 | typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd); | |
139 | typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd); | |
001bea73 | 140 | typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd); |
a9520cd6 | 141 | typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd); |
3e252066 | 142 | |
3c798398 | 143 | struct blkcg_policy { |
36558c8a | 144 | int plid; |
36558c8a TH |
145 | /* cgroup files for the policy */ |
146 | struct cftype *cftypes; | |
f9fcc2d3 TH |
147 | |
148 | /* operations */ | |
e4a9bde9 | 149 | blkcg_pol_alloc_cpd_fn *cpd_alloc_fn; |
e48453c3 | 150 | blkcg_pol_init_cpd_fn *cpd_init_fn; |
e4a9bde9 TH |
151 | blkcg_pol_free_cpd_fn *cpd_free_fn; |
152 | ||
001bea73 | 153 | blkcg_pol_alloc_pd_fn *pd_alloc_fn; |
f9fcc2d3 | 154 | blkcg_pol_init_pd_fn *pd_init_fn; |
f427d909 TH |
155 | blkcg_pol_online_pd_fn *pd_online_fn; |
156 | blkcg_pol_offline_pd_fn *pd_offline_fn; | |
001bea73 | 157 | blkcg_pol_free_pd_fn *pd_free_fn; |
f9fcc2d3 | 158 | blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn; |
3e252066 VG |
159 | }; |
160 | ||
3c798398 | 161 | extern struct blkcg blkcg_root; |
496d5e75 | 162 | extern struct cgroup_subsys_state * const blkcg_root_css; |
36558c8a | 163 | |
24f29046 TH |
164 | struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg, |
165 | struct request_queue *q, bool update_hint); | |
3c798398 TH |
166 | struct blkcg_gq *blkg_lookup_create(struct blkcg *blkcg, |
167 | struct request_queue *q); | |
36558c8a TH |
168 | int blkcg_init_queue(struct request_queue *q); |
169 | void blkcg_drain_queue(struct request_queue *q); | |
170 | void blkcg_exit_queue(struct request_queue *q); | |
5efd6113 | 171 | |
3e252066 | 172 | /* Blkio controller policy registration */ |
d5bf0291 | 173 | int blkcg_policy_register(struct blkcg_policy *pol); |
3c798398 | 174 | void blkcg_policy_unregister(struct blkcg_policy *pol); |
36558c8a | 175 | int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 176 | const struct blkcg_policy *pol); |
36558c8a | 177 | void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 | 178 | const struct blkcg_policy *pol); |
3e252066 | 179 | |
3c798398 | 180 | void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg, |
f95a04af TH |
181 | u64 (*prfill)(struct seq_file *, |
182 | struct blkg_policy_data *, int), | |
3c798398 | 183 | const struct blkcg_policy *pol, int data, |
ec399347 | 184 | bool show_total); |
f95a04af TH |
185 | u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v); |
186 | u64 __blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
829fdb50 | 187 | const struct blkg_rwstat *rwstat); |
f95a04af TH |
188 | u64 blkg_prfill_stat(struct seq_file *sf, struct blkg_policy_data *pd, int off); |
189 | u64 blkg_prfill_rwstat(struct seq_file *sf, struct blkg_policy_data *pd, | |
190 | int off); | |
829fdb50 | 191 | |
16b3de66 TH |
192 | u64 blkg_stat_recursive_sum(struct blkg_policy_data *pd, int off); |
193 | struct blkg_rwstat blkg_rwstat_recursive_sum(struct blkg_policy_data *pd, | |
194 | int off); | |
195 | ||
829fdb50 | 196 | struct blkg_conf_ctx { |
36558c8a | 197 | struct gendisk *disk; |
3c798398 | 198 | struct blkcg_gq *blkg; |
36558c8a | 199 | u64 v; |
829fdb50 TH |
200 | }; |
201 | ||
3c798398 TH |
202 | int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol, |
203 | const char *input, struct blkg_conf_ctx *ctx); | |
829fdb50 TH |
204 | void blkg_conf_finish(struct blkg_conf_ctx *ctx); |
205 | ||
206 | ||
a7c6d554 TH |
207 | static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css) |
208 | { | |
209 | return css ? container_of(css, struct blkcg, css) : NULL; | |
210 | } | |
211 | ||
b1208b56 TH |
212 | static inline struct blkcg *task_blkcg(struct task_struct *tsk) |
213 | { | |
073219e9 | 214 | return css_to_blkcg(task_css(tsk, blkio_cgrp_id)); |
b1208b56 TH |
215 | } |
216 | ||
217 | static inline struct blkcg *bio_blkcg(struct bio *bio) | |
218 | { | |
219 | if (bio && bio->bi_css) | |
a7c6d554 | 220 | return css_to_blkcg(bio->bi_css); |
b1208b56 TH |
221 | return task_blkcg(current); |
222 | } | |
223 | ||
fd383c2d TH |
224 | static inline struct cgroup_subsys_state * |
225 | task_get_blkcg_css(struct task_struct *task) | |
226 | { | |
227 | return task_get_css(task, blkio_cgrp_id); | |
228 | } | |
229 | ||
3c547865 TH |
230 | /** |
231 | * blkcg_parent - get the parent of a blkcg | |
232 | * @blkcg: blkcg of interest | |
233 | * | |
234 | * Return the parent blkcg of @blkcg. Can be called anytime. | |
235 | */ | |
236 | static inline struct blkcg *blkcg_parent(struct blkcg *blkcg) | |
237 | { | |
5c9d535b | 238 | return css_to_blkcg(blkcg->css.parent); |
3c547865 TH |
239 | } |
240 | ||
24f29046 TH |
241 | /** |
242 | * __blkg_lookup - internal version of blkg_lookup() | |
243 | * @blkcg: blkcg of interest | |
244 | * @q: request_queue of interest | |
245 | * @update_hint: whether to update lookup hint with the result or not | |
246 | * | |
247 | * This is internal version and shouldn't be used by policy | |
248 | * implementations. Looks up blkgs for the @blkcg - @q pair regardless of | |
249 | * @q's bypass state. If @update_hint is %true, the caller should be | |
250 | * holding @q->queue_lock and lookup hint is updated on success. | |
251 | */ | |
252 | static inline struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg, | |
253 | struct request_queue *q, | |
254 | bool update_hint) | |
255 | { | |
256 | struct blkcg_gq *blkg; | |
257 | ||
85b6bc9d TH |
258 | if (blkcg == &blkcg_root) |
259 | return q->root_blkg; | |
260 | ||
24f29046 TH |
261 | blkg = rcu_dereference(blkcg->blkg_hint); |
262 | if (blkg && blkg->q == q) | |
263 | return blkg; | |
264 | ||
265 | return blkg_lookup_slowpath(blkcg, q, update_hint); | |
266 | } | |
267 | ||
268 | /** | |
269 | * blkg_lookup - lookup blkg for the specified blkcg - q pair | |
270 | * @blkcg: blkcg of interest | |
271 | * @q: request_queue of interest | |
272 | * | |
273 | * Lookup blkg for the @blkcg - @q pair. This function should be called | |
274 | * under RCU read lock and is guaranteed to return %NULL if @q is bypassing | |
275 | * - see blk_queue_bypass_start() for details. | |
276 | */ | |
277 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, | |
278 | struct request_queue *q) | |
279 | { | |
280 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
281 | ||
282 | if (unlikely(blk_queue_bypass(q))) | |
283 | return NULL; | |
284 | return __blkg_lookup(blkcg, q, false); | |
285 | } | |
286 | ||
0381411e TH |
287 | /** |
288 | * blkg_to_pdata - get policy private data | |
289 | * @blkg: blkg of interest | |
290 | * @pol: policy of interest | |
291 | * | |
292 | * Return pointer to private data associated with the @blkg-@pol pair. | |
293 | */ | |
f95a04af TH |
294 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
295 | struct blkcg_policy *pol) | |
0381411e | 296 | { |
f95a04af | 297 | return blkg ? blkg->pd[pol->plid] : NULL; |
0381411e TH |
298 | } |
299 | ||
e48453c3 AA |
300 | static inline struct blkcg_policy_data *blkcg_to_cpd(struct blkcg *blkcg, |
301 | struct blkcg_policy *pol) | |
302 | { | |
81437648 | 303 | return blkcg ? blkcg->cpd[pol->plid] : NULL; |
e48453c3 AA |
304 | } |
305 | ||
0381411e TH |
306 | /** |
307 | * pdata_to_blkg - get blkg associated with policy private data | |
f95a04af | 308 | * @pd: policy private data of interest |
0381411e | 309 | * |
f95a04af | 310 | * @pd is policy private data. Determine the blkg it's associated with. |
0381411e | 311 | */ |
f95a04af | 312 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) |
0381411e | 313 | { |
f95a04af | 314 | return pd ? pd->blkg : NULL; |
0381411e TH |
315 | } |
316 | ||
81437648 TH |
317 | static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd) |
318 | { | |
319 | return cpd ? cpd->blkcg : NULL; | |
320 | } | |
321 | ||
54e7ed12 TH |
322 | /** |
323 | * blkg_path - format cgroup path of blkg | |
324 | * @blkg: blkg of interest | |
325 | * @buf: target buffer | |
326 | * @buflen: target buffer length | |
327 | * | |
328 | * Format the path of the cgroup of @blkg into @buf. | |
329 | */ | |
3c798398 | 330 | static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen) |
afc24d49 | 331 | { |
e61734c5 | 332 | char *p; |
54e7ed12 | 333 | |
e61734c5 TH |
334 | p = cgroup_path(blkg->blkcg->css.cgroup, buf, buflen); |
335 | if (!p) { | |
54e7ed12 | 336 | strncpy(buf, "<unavailable>", buflen); |
e61734c5 TH |
337 | return -ENAMETOOLONG; |
338 | } | |
339 | ||
340 | memmove(buf, p, buf + buflen - p); | |
341 | return 0; | |
afc24d49 VG |
342 | } |
343 | ||
1adaf3dd TH |
344 | /** |
345 | * blkg_get - get a blkg reference | |
346 | * @blkg: blkg to get | |
347 | * | |
a5049a8a | 348 | * The caller should be holding an existing reference. |
1adaf3dd | 349 | */ |
3c798398 | 350 | static inline void blkg_get(struct blkcg_gq *blkg) |
1adaf3dd | 351 | { |
a5049a8a TH |
352 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
353 | atomic_inc(&blkg->refcnt); | |
1adaf3dd TH |
354 | } |
355 | ||
2a4fd070 | 356 | void __blkg_release_rcu(struct rcu_head *rcu); |
1adaf3dd TH |
357 | |
358 | /** | |
359 | * blkg_put - put a blkg reference | |
360 | * @blkg: blkg to put | |
1adaf3dd | 361 | */ |
3c798398 | 362 | static inline void blkg_put(struct blkcg_gq *blkg) |
1adaf3dd | 363 | { |
a5049a8a TH |
364 | WARN_ON_ONCE(atomic_read(&blkg->refcnt) <= 0); |
365 | if (atomic_dec_and_test(&blkg->refcnt)) | |
2a4fd070 | 366 | call_rcu(&blkg->rcu_head, __blkg_release_rcu); |
1adaf3dd TH |
367 | } |
368 | ||
dd4a4ffc TH |
369 | /** |
370 | * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants | |
371 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 372 | * @pos_css: used for iteration |
dd4a4ffc TH |
373 | * @p_blkg: target blkg to walk descendants of |
374 | * | |
375 | * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU | |
376 | * read locked. If called under either blkcg or queue lock, the iteration | |
377 | * is guaranteed to include all and only online blkgs. The caller may | |
492eb21b | 378 | * update @pos_css by calling css_rightmost_descendant() to skip subtree. |
bd8815a6 | 379 | * @p_blkg is included in the iteration and the first node to be visited. |
dd4a4ffc | 380 | */ |
492eb21b TH |
381 | #define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \ |
382 | css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \ | |
383 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
dd4a4ffc TH |
384 | (p_blkg)->q, false))) |
385 | ||
aa539cb3 TH |
386 | /** |
387 | * blkg_for_each_descendant_post - post-order walk of a blkg's descendants | |
388 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 389 | * @pos_css: used for iteration |
aa539cb3 TH |
390 | * @p_blkg: target blkg to walk descendants of |
391 | * | |
392 | * Similar to blkg_for_each_descendant_pre() but performs post-order | |
bd8815a6 TH |
393 | * traversal instead. Synchronization rules are the same. @p_blkg is |
394 | * included in the iteration and the last node to be visited. | |
aa539cb3 | 395 | */ |
492eb21b TH |
396 | #define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \ |
397 | css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \ | |
398 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
aa539cb3 TH |
399 | (p_blkg)->q, false))) |
400 | ||
a051661c TH |
401 | /** |
402 | * blk_get_rl - get request_list to use | |
403 | * @q: request_queue of interest | |
404 | * @bio: bio which will be attached to the allocated request (may be %NULL) | |
405 | * | |
406 | * The caller wants to allocate a request from @q to use for @bio. Find | |
407 | * the request_list to use and obtain a reference on it. Should be called | |
408 | * under queue_lock. This function is guaranteed to return non-%NULL | |
409 | * request_list. | |
410 | */ | |
411 | static inline struct request_list *blk_get_rl(struct request_queue *q, | |
412 | struct bio *bio) | |
413 | { | |
414 | struct blkcg *blkcg; | |
415 | struct blkcg_gq *blkg; | |
416 | ||
417 | rcu_read_lock(); | |
418 | ||
419 | blkcg = bio_blkcg(bio); | |
420 | ||
421 | /* bypass blkg lookup and use @q->root_rl directly for root */ | |
422 | if (blkcg == &blkcg_root) | |
423 | goto root_rl; | |
424 | ||
425 | /* | |
426 | * Try to use blkg->rl. blkg lookup may fail under memory pressure | |
427 | * or if either the blkcg or queue is going away. Fall back to | |
428 | * root_rl in such cases. | |
429 | */ | |
ae118896 TH |
430 | blkg = blkg_lookup(blkcg, q); |
431 | if (unlikely(!blkg)) | |
a051661c TH |
432 | goto root_rl; |
433 | ||
434 | blkg_get(blkg); | |
435 | rcu_read_unlock(); | |
436 | return &blkg->rl; | |
437 | root_rl: | |
438 | rcu_read_unlock(); | |
439 | return &q->root_rl; | |
440 | } | |
441 | ||
442 | /** | |
443 | * blk_put_rl - put request_list | |
444 | * @rl: request_list to put | |
445 | * | |
446 | * Put the reference acquired by blk_get_rl(). Should be called under | |
447 | * queue_lock. | |
448 | */ | |
449 | static inline void blk_put_rl(struct request_list *rl) | |
450 | { | |
401efbf8 | 451 | if (rl->blkg->blkcg != &blkcg_root) |
a051661c TH |
452 | blkg_put(rl->blkg); |
453 | } | |
454 | ||
455 | /** | |
456 | * blk_rq_set_rl - associate a request with a request_list | |
457 | * @rq: request of interest | |
458 | * @rl: target request_list | |
459 | * | |
460 | * Associate @rq with @rl so that accounting and freeing can know the | |
461 | * request_list @rq came from. | |
462 | */ | |
463 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) | |
464 | { | |
465 | rq->rl = rl; | |
466 | } | |
467 | ||
468 | /** | |
469 | * blk_rq_rl - return the request_list a request came from | |
470 | * @rq: request of interest | |
471 | * | |
472 | * Return the request_list @rq is allocated from. | |
473 | */ | |
474 | static inline struct request_list *blk_rq_rl(struct request *rq) | |
475 | { | |
476 | return rq->rl; | |
477 | } | |
478 | ||
479 | struct request_list *__blk_queue_next_rl(struct request_list *rl, | |
480 | struct request_queue *q); | |
481 | /** | |
482 | * blk_queue_for_each_rl - iterate through all request_lists of a request_queue | |
483 | * | |
484 | * Should be used under queue_lock. | |
485 | */ | |
486 | #define blk_queue_for_each_rl(rl, q) \ | |
487 | for ((rl) = &(q)->root_rl; (rl); (rl) = __blk_queue_next_rl((rl), (q))) | |
488 | ||
90d3839b PZ |
489 | static inline void blkg_stat_init(struct blkg_stat *stat) |
490 | { | |
491 | u64_stats_init(&stat->syncp); | |
e6269c44 | 492 | atomic64_set(&stat->aux_cnt, 0); |
90d3839b PZ |
493 | } |
494 | ||
edcb0722 TH |
495 | /** |
496 | * blkg_stat_add - add a value to a blkg_stat | |
497 | * @stat: target blkg_stat | |
498 | * @val: value to add | |
499 | * | |
500 | * Add @val to @stat. The caller is responsible for synchronizing calls to | |
501 | * this function. | |
502 | */ | |
503 | static inline void blkg_stat_add(struct blkg_stat *stat, uint64_t val) | |
504 | { | |
505 | u64_stats_update_begin(&stat->syncp); | |
506 | stat->cnt += val; | |
507 | u64_stats_update_end(&stat->syncp); | |
508 | } | |
509 | ||
510 | /** | |
511 | * blkg_stat_read - read the current value of a blkg_stat | |
512 | * @stat: blkg_stat to read | |
513 | * | |
e6269c44 TH |
514 | * Read the current value of @stat. The returned value doesn't include the |
515 | * aux count. This function can be called without synchroniztion and takes | |
516 | * care of u64 atomicity. | |
edcb0722 TH |
517 | */ |
518 | static inline uint64_t blkg_stat_read(struct blkg_stat *stat) | |
519 | { | |
520 | unsigned int start; | |
521 | uint64_t v; | |
522 | ||
523 | do { | |
57a7744e | 524 | start = u64_stats_fetch_begin_irq(&stat->syncp); |
edcb0722 | 525 | v = stat->cnt; |
57a7744e | 526 | } while (u64_stats_fetch_retry_irq(&stat->syncp, start)); |
edcb0722 TH |
527 | |
528 | return v; | |
529 | } | |
530 | ||
531 | /** | |
532 | * blkg_stat_reset - reset a blkg_stat | |
533 | * @stat: blkg_stat to reset | |
534 | */ | |
535 | static inline void blkg_stat_reset(struct blkg_stat *stat) | |
536 | { | |
537 | stat->cnt = 0; | |
e6269c44 | 538 | atomic64_set(&stat->aux_cnt, 0); |
edcb0722 TH |
539 | } |
540 | ||
16b3de66 | 541 | /** |
e6269c44 | 542 | * blkg_stat_add_aux - add a blkg_stat into another's aux count |
16b3de66 TH |
543 | * @to: the destination blkg_stat |
544 | * @from: the source | |
545 | * | |
e6269c44 | 546 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 547 | */ |
e6269c44 TH |
548 | static inline void blkg_stat_add_aux(struct blkg_stat *to, |
549 | struct blkg_stat *from) | |
16b3de66 | 550 | { |
e6269c44 TH |
551 | atomic64_add(blkg_stat_read(from) + atomic64_read(&from->aux_cnt), |
552 | &to->aux_cnt); | |
16b3de66 TH |
553 | } |
554 | ||
90d3839b PZ |
555 | static inline void blkg_rwstat_init(struct blkg_rwstat *rwstat) |
556 | { | |
e6269c44 TH |
557 | int i; |
558 | ||
90d3839b | 559 | u64_stats_init(&rwstat->syncp); |
e6269c44 TH |
560 | |
561 | for (i = 0; i < BLKG_RWSTAT_NR; i++) | |
562 | atomic64_set(&rwstat->aux_cnt[i], 0); | |
90d3839b PZ |
563 | } |
564 | ||
edcb0722 TH |
565 | /** |
566 | * blkg_rwstat_add - add a value to a blkg_rwstat | |
567 | * @rwstat: target blkg_rwstat | |
568 | * @rw: mask of REQ_{WRITE|SYNC} | |
569 | * @val: value to add | |
570 | * | |
571 | * Add @val to @rwstat. The counters are chosen according to @rw. The | |
572 | * caller is responsible for synchronizing calls to this function. | |
573 | */ | |
574 | static inline void blkg_rwstat_add(struct blkg_rwstat *rwstat, | |
575 | int rw, uint64_t val) | |
576 | { | |
577 | u64_stats_update_begin(&rwstat->syncp); | |
578 | ||
579 | if (rw & REQ_WRITE) | |
580 | rwstat->cnt[BLKG_RWSTAT_WRITE] += val; | |
581 | else | |
582 | rwstat->cnt[BLKG_RWSTAT_READ] += val; | |
583 | if (rw & REQ_SYNC) | |
584 | rwstat->cnt[BLKG_RWSTAT_SYNC] += val; | |
585 | else | |
586 | rwstat->cnt[BLKG_RWSTAT_ASYNC] += val; | |
587 | ||
588 | u64_stats_update_end(&rwstat->syncp); | |
589 | } | |
590 | ||
591 | /** | |
592 | * blkg_rwstat_read - read the current values of a blkg_rwstat | |
593 | * @rwstat: blkg_rwstat to read | |
594 | * | |
595 | * Read the current snapshot of @rwstat and return it as the return value. | |
596 | * This function can be called without synchronization and takes care of | |
597 | * u64 atomicity. | |
598 | */ | |
c94bed89 | 599 | static inline struct blkg_rwstat blkg_rwstat_read(struct blkg_rwstat *rwstat) |
edcb0722 TH |
600 | { |
601 | unsigned int start; | |
602 | struct blkg_rwstat tmp; | |
603 | ||
604 | do { | |
57a7744e | 605 | start = u64_stats_fetch_begin_irq(&rwstat->syncp); |
edcb0722 | 606 | tmp = *rwstat; |
57a7744e | 607 | } while (u64_stats_fetch_retry_irq(&rwstat->syncp, start)); |
edcb0722 TH |
608 | |
609 | return tmp; | |
610 | } | |
611 | ||
612 | /** | |
4d5e80a7 | 613 | * blkg_rwstat_total - read the total count of a blkg_rwstat |
edcb0722 TH |
614 | * @rwstat: blkg_rwstat to read |
615 | * | |
616 | * Return the total count of @rwstat regardless of the IO direction. This | |
617 | * function can be called without synchronization and takes care of u64 | |
618 | * atomicity. | |
619 | */ | |
4d5e80a7 | 620 | static inline uint64_t blkg_rwstat_total(struct blkg_rwstat *rwstat) |
edcb0722 TH |
621 | { |
622 | struct blkg_rwstat tmp = blkg_rwstat_read(rwstat); | |
623 | ||
624 | return tmp.cnt[BLKG_RWSTAT_READ] + tmp.cnt[BLKG_RWSTAT_WRITE]; | |
625 | } | |
626 | ||
627 | /** | |
628 | * blkg_rwstat_reset - reset a blkg_rwstat | |
629 | * @rwstat: blkg_rwstat to reset | |
630 | */ | |
631 | static inline void blkg_rwstat_reset(struct blkg_rwstat *rwstat) | |
632 | { | |
e6269c44 TH |
633 | int i; |
634 | ||
edcb0722 | 635 | memset(rwstat->cnt, 0, sizeof(rwstat->cnt)); |
e6269c44 TH |
636 | |
637 | for (i = 0; i < BLKG_RWSTAT_NR; i++) | |
638 | atomic64_set(&rwstat->aux_cnt[i], 0); | |
edcb0722 TH |
639 | } |
640 | ||
16b3de66 | 641 | /** |
e6269c44 | 642 | * blkg_rwstat_add_aux - add a blkg_rwstat into another's aux count |
16b3de66 TH |
643 | * @to: the destination blkg_rwstat |
644 | * @from: the source | |
645 | * | |
e6269c44 | 646 | * Add @from's count including the aux one to @to's aux count. |
16b3de66 | 647 | */ |
e6269c44 TH |
648 | static inline void blkg_rwstat_add_aux(struct blkg_rwstat *to, |
649 | struct blkg_rwstat *from) | |
16b3de66 TH |
650 | { |
651 | struct blkg_rwstat v = blkg_rwstat_read(from); | |
652 | int i; | |
653 | ||
16b3de66 | 654 | for (i = 0; i < BLKG_RWSTAT_NR; i++) |
e6269c44 TH |
655 | atomic64_add(v.cnt[i] + atomic64_read(&from->aux_cnt[i]), |
656 | &to->aux_cnt[i]); | |
16b3de66 TH |
657 | } |
658 | ||
ae118896 TH |
659 | #ifdef CONFIG_BLK_DEV_THROTTLING |
660 | extern bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
661 | struct bio *bio); | |
662 | #else | |
663 | static inline bool blk_throtl_bio(struct request_queue *q, struct blkcg_gq *blkg, | |
664 | struct bio *bio) { return false; } | |
665 | #endif | |
666 | ||
667 | static inline bool blkcg_bio_issue_check(struct request_queue *q, | |
668 | struct bio *bio) | |
669 | { | |
670 | struct blkcg *blkcg; | |
671 | struct blkcg_gq *blkg; | |
672 | bool throtl = false; | |
673 | ||
674 | rcu_read_lock(); | |
675 | blkcg = bio_blkcg(bio); | |
676 | ||
677 | blkg = blkg_lookup(blkcg, q); | |
678 | if (unlikely(!blkg)) { | |
679 | spin_lock_irq(q->queue_lock); | |
680 | blkg = blkg_lookup_create(blkcg, q); | |
681 | if (IS_ERR(blkg)) | |
682 | blkg = NULL; | |
683 | spin_unlock_irq(q->queue_lock); | |
684 | } | |
685 | ||
686 | throtl = blk_throtl_bio(q, blkg, bio); | |
687 | ||
688 | rcu_read_unlock(); | |
689 | return !throtl; | |
690 | } | |
691 | ||
36558c8a TH |
692 | #else /* CONFIG_BLK_CGROUP */ |
693 | ||
efa7d1c7 TH |
694 | struct blkcg { |
695 | }; | |
2f5ea477 | 696 | |
f95a04af TH |
697 | struct blkg_policy_data { |
698 | }; | |
699 | ||
e48453c3 AA |
700 | struct blkcg_policy_data { |
701 | }; | |
702 | ||
3c798398 | 703 | struct blkcg_gq { |
2f5ea477 JA |
704 | }; |
705 | ||
3c798398 | 706 | struct blkcg_policy { |
3e252066 VG |
707 | }; |
708 | ||
496d5e75 TH |
709 | #define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL)) |
710 | ||
fd383c2d TH |
711 | static inline struct cgroup_subsys_state * |
712 | task_get_blkcg_css(struct task_struct *task) | |
713 | { | |
714 | return NULL; | |
715 | } | |
716 | ||
efa7d1c7 TH |
717 | #ifdef CONFIG_BLOCK |
718 | ||
3c798398 | 719 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; } |
5efd6113 TH |
720 | static inline int blkcg_init_queue(struct request_queue *q) { return 0; } |
721 | static inline void blkcg_drain_queue(struct request_queue *q) { } | |
722 | static inline void blkcg_exit_queue(struct request_queue *q) { } | |
d5bf0291 | 723 | static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; } |
3c798398 | 724 | static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { } |
a2b1693b | 725 | static inline int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 726 | const struct blkcg_policy *pol) { return 0; } |
a2b1693b | 727 | static inline void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 TH |
728 | const struct blkcg_policy *pol) { } |
729 | ||
b1208b56 | 730 | static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; } |
a051661c | 731 | |
f95a04af TH |
732 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
733 | struct blkcg_policy *pol) { return NULL; } | |
734 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; } | |
3c798398 TH |
735 | static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; } |
736 | static inline void blkg_get(struct blkcg_gq *blkg) { } | |
737 | static inline void blkg_put(struct blkcg_gq *blkg) { } | |
afc24d49 | 738 | |
a051661c TH |
739 | static inline struct request_list *blk_get_rl(struct request_queue *q, |
740 | struct bio *bio) { return &q->root_rl; } | |
741 | static inline void blk_put_rl(struct request_list *rl) { } | |
742 | static inline void blk_rq_set_rl(struct request *rq, struct request_list *rl) { } | |
743 | static inline struct request_list *blk_rq_rl(struct request *rq) { return &rq->q->root_rl; } | |
744 | ||
ae118896 TH |
745 | static inline bool blkcg_bio_issue_check(struct request_queue *q, |
746 | struct bio *bio) { return true; } | |
747 | ||
a051661c TH |
748 | #define blk_queue_for_each_rl(rl, q) \ |
749 | for ((rl) = &(q)->root_rl; (rl); (rl) = NULL) | |
750 | ||
efa7d1c7 | 751 | #endif /* CONFIG_BLOCK */ |
36558c8a TH |
752 | #endif /* CONFIG_BLK_CGROUP */ |
753 | #endif /* _BLK_CGROUP_H */ |