]>
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> | |
f7331648 | 18 | #include <linux/percpu.h> |
24bdb8ef | 19 | #include <linux/percpu_counter.h> |
f7331648 | 20 | #include <linux/u64_stats_sync.h> |
829fdb50 | 21 | #include <linux/seq_file.h> |
a637120e | 22 | #include <linux/radix-tree.h> |
a051661c | 23 | #include <linux/blkdev.h> |
a5049a8a | 24 | #include <linux/atomic.h> |
902ec5b6 | 25 | #include <linux/kthread.h> |
5cdf2e3f | 26 | #include <linux/fs.h> |
31e4c28d | 27 | |
24bdb8ef TH |
28 | /* percpu_counter batch for blkg_[rw]stats, per-cpu drift doesn't matter */ |
29 | #define BLKG_STAT_CPU_BATCH (INT_MAX / 2) | |
30 | ||
9355aede VG |
31 | /* Max limits for throttle policy */ |
32 | #define THROTL_IOPS_MAX UINT_MAX | |
33 | ||
f48ec1d7 TH |
34 | #ifdef CONFIG_BLK_CGROUP |
35 | ||
f7331648 TH |
36 | enum blkg_iostat_type { |
37 | BLKG_IOSTAT_READ, | |
38 | BLKG_IOSTAT_WRITE, | |
39 | BLKG_IOSTAT_DISCARD, | |
40 | ||
41 | BLKG_IOSTAT_NR, | |
42 | }; | |
43 | ||
a637120e TH |
44 | struct blkcg_gq; |
45 | ||
3c798398 | 46 | struct blkcg { |
36558c8a TH |
47 | struct cgroup_subsys_state css; |
48 | spinlock_t lock; | |
d866dbf6 | 49 | refcount_t online_pin; |
a637120e TH |
50 | |
51 | struct radix_tree_root blkg_tree; | |
55679c8d | 52 | struct blkcg_gq __rcu *blkg_hint; |
36558c8a | 53 | struct hlist_head blkg_list; |
9a9e8a26 | 54 | |
81437648 | 55 | struct blkcg_policy_data *cpd[BLKCG_MAX_POLS]; |
52ebea74 | 56 | |
7876f930 | 57 | struct list_head all_blkcgs_node; |
52ebea74 TH |
58 | #ifdef CONFIG_CGROUP_WRITEBACK |
59 | struct list_head cgwb_list; | |
60 | #endif | |
31e4c28d VG |
61 | }; |
62 | ||
f7331648 TH |
63 | struct blkg_iostat { |
64 | u64 bytes[BLKG_IOSTAT_NR]; | |
65 | u64 ios[BLKG_IOSTAT_NR]; | |
66 | }; | |
67 | ||
68 | struct blkg_iostat_set { | |
69 | struct u64_stats_sync sync; | |
70 | struct blkg_iostat cur; | |
71 | struct blkg_iostat last; | |
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 TH |
111 | |
112 | /* all non-root blkcg_gq's are guaranteed to have access to parent */ | |
113 | struct blkcg_gq *parent; | |
114 | ||
1adaf3dd | 115 | /* reference count */ |
7fcf2b03 | 116 | struct percpu_ref refcnt; |
22084190 | 117 | |
f427d909 TH |
118 | /* is this blkg online? protected by both blkcg and q locks */ |
119 | bool online; | |
120 | ||
f7331648 TH |
121 | struct blkg_iostat_set __percpu *iostat_cpu; |
122 | struct blkg_iostat_set iostat; | |
77ea7338 | 123 | |
36558c8a | 124 | struct blkg_policy_data *pd[BLKCG_MAX_POLS]; |
1adaf3dd | 125 | |
d3f77dfd TH |
126 | spinlock_t async_bio_lock; |
127 | struct bio_list async_bios; | |
128 | struct work_struct async_bio_work; | |
d09d8df3 JB |
129 | |
130 | atomic_t use_delay; | |
131 | atomic64_t delay_nsec; | |
132 | atomic64_t delay_start; | |
133 | u64 last_delay; | |
134 | int last_use; | |
d3f77dfd TH |
135 | |
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); |
cf09a8ee TH |
143 | typedef struct blkg_policy_data *(blkcg_pol_alloc_pd_fn)(gfp_t gfp, |
144 | struct request_queue *q, struct blkcg *blkcg); | |
a9520cd6 TH |
145 | typedef void (blkcg_pol_init_pd_fn)(struct blkg_policy_data *pd); |
146 | typedef void (blkcg_pol_online_pd_fn)(struct blkg_policy_data *pd); | |
147 | typedef void (blkcg_pol_offline_pd_fn)(struct blkg_policy_data *pd); | |
001bea73 | 148 | typedef void (blkcg_pol_free_pd_fn)(struct blkg_policy_data *pd); |
a9520cd6 | 149 | typedef void (blkcg_pol_reset_pd_stats_fn)(struct blkg_policy_data *pd); |
903d23f0 JB |
150 | typedef size_t (blkcg_pol_stat_pd_fn)(struct blkg_policy_data *pd, char *buf, |
151 | size_t size); | |
3e252066 | 152 | |
3c798398 | 153 | struct blkcg_policy { |
36558c8a | 154 | int plid; |
36558c8a | 155 | /* cgroup files for the policy */ |
2ee867dc | 156 | struct cftype *dfl_cftypes; |
880f50e2 | 157 | struct cftype *legacy_cftypes; |
f9fcc2d3 TH |
158 | |
159 | /* operations */ | |
e4a9bde9 | 160 | blkcg_pol_alloc_cpd_fn *cpd_alloc_fn; |
e48453c3 | 161 | blkcg_pol_init_cpd_fn *cpd_init_fn; |
e4a9bde9 | 162 | blkcg_pol_free_cpd_fn *cpd_free_fn; |
69d7fde5 | 163 | blkcg_pol_bind_cpd_fn *cpd_bind_fn; |
e4a9bde9 | 164 | |
001bea73 | 165 | blkcg_pol_alloc_pd_fn *pd_alloc_fn; |
f9fcc2d3 | 166 | blkcg_pol_init_pd_fn *pd_init_fn; |
f427d909 TH |
167 | blkcg_pol_online_pd_fn *pd_online_fn; |
168 | blkcg_pol_offline_pd_fn *pd_offline_fn; | |
001bea73 | 169 | blkcg_pol_free_pd_fn *pd_free_fn; |
f9fcc2d3 | 170 | blkcg_pol_reset_pd_stats_fn *pd_reset_stats_fn; |
903d23f0 | 171 | blkcg_pol_stat_pd_fn *pd_stat_fn; |
3e252066 VG |
172 | }; |
173 | ||
3c798398 | 174 | extern struct blkcg blkcg_root; |
496d5e75 | 175 | extern struct cgroup_subsys_state * const blkcg_root_css; |
07b0fdec | 176 | extern bool blkcg_debug_stats; |
36558c8a | 177 | |
24f29046 TH |
178 | struct blkcg_gq *blkg_lookup_slowpath(struct blkcg *blkcg, |
179 | struct request_queue *q, bool update_hint); | |
36558c8a | 180 | int blkcg_init_queue(struct request_queue *q); |
36558c8a | 181 | void blkcg_exit_queue(struct request_queue *q); |
5efd6113 | 182 | |
3e252066 | 183 | /* Blkio controller policy registration */ |
d5bf0291 | 184 | int blkcg_policy_register(struct blkcg_policy *pol); |
3c798398 | 185 | void blkcg_policy_unregister(struct blkcg_policy *pol); |
36558c8a | 186 | int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 187 | const struct blkcg_policy *pol); |
36558c8a | 188 | void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 | 189 | const struct blkcg_policy *pol); |
3e252066 | 190 | |
dd165eb3 | 191 | const char *blkg_dev_name(struct blkcg_gq *blkg); |
3c798398 | 192 | void blkcg_print_blkgs(struct seq_file *sf, struct blkcg *blkcg, |
f95a04af TH |
193 | u64 (*prfill)(struct seq_file *, |
194 | struct blkg_policy_data *, int), | |
3c798398 | 195 | const struct blkcg_policy *pol, int data, |
ec399347 | 196 | bool show_total); |
f95a04af | 197 | u64 __blkg_prfill_u64(struct seq_file *sf, struct blkg_policy_data *pd, u64 v); |
16b3de66 | 198 | |
829fdb50 | 199 | struct blkg_conf_ctx { |
22ae8ce8 | 200 | struct block_device *bdev; |
3c798398 | 201 | struct blkcg_gq *blkg; |
36aa9e5f | 202 | char *body; |
829fdb50 TH |
203 | }; |
204 | ||
22ae8ce8 | 205 | struct block_device *blkcg_conf_open_bdev(char **inputp); |
3c798398 | 206 | int blkg_conf_prep(struct blkcg *blkcg, const struct blkcg_policy *pol, |
36aa9e5f | 207 | char *input, struct blkg_conf_ctx *ctx); |
829fdb50 TH |
208 | void blkg_conf_finish(struct blkg_conf_ctx *ctx); |
209 | ||
0fe061b9 DZ |
210 | /** |
211 | * blkcg_css - find the current css | |
212 | * | |
213 | * Find the css associated with either the kthread or the current task. | |
214 | * This may return a dying css, so it is up to the caller to use tryget logic | |
215 | * to confirm it is alive and well. | |
216 | */ | |
217 | static inline struct cgroup_subsys_state *blkcg_css(void) | |
218 | { | |
219 | struct cgroup_subsys_state *css; | |
220 | ||
221 | css = kthread_blkcg(); | |
222 | if (css) | |
223 | return css; | |
224 | return task_css(current, io_cgrp_id); | |
225 | } | |
226 | ||
a7c6d554 TH |
227 | static inline struct blkcg *css_to_blkcg(struct cgroup_subsys_state *css) |
228 | { | |
229 | return css ? container_of(css, struct blkcg, css) : NULL; | |
230 | } | |
231 | ||
0fe061b9 DZ |
232 | /** |
233 | * __bio_blkcg - internal, inconsistent version to get blkcg | |
234 | * | |
235 | * DO NOT USE. | |
236 | * This function is inconsistent and consequently is dangerous to use. The | |
237 | * first part of the function returns a blkcg where a reference is owned by the | |
238 | * bio. This means it does not need to be rcu protected as it cannot go away | |
239 | * with the bio owning a reference to it. However, the latter potentially gets | |
240 | * it from task_css(). This can race against task migration and the cgroup | |
241 | * dying. It is also semantically different as it must be called rcu protected | |
242 | * and is susceptible to failure when trying to get a reference to it. | |
243 | * Therefore, it is not ok to assume that *_get() will always succeed on the | |
244 | * blkcg returned here. | |
245 | */ | |
246 | static inline struct blkcg *__bio_blkcg(struct bio *bio) | |
27e6fa99 | 247 | { |
db6638d7 DZ |
248 | if (bio && bio->bi_blkg) |
249 | return bio->bi_blkg->blkcg; | |
0fe061b9 DZ |
250 | return css_to_blkcg(blkcg_css()); |
251 | } | |
b5f2954d | 252 | |
0fe061b9 DZ |
253 | /** |
254 | * bio_blkcg - grab the blkcg associated with a bio | |
255 | * @bio: target bio | |
256 | * | |
257 | * This returns the blkcg associated with a bio, %NULL if not associated. | |
258 | * Callers are expected to either handle %NULL or know association has been | |
259 | * done prior to calling this. | |
260 | */ | |
261 | static inline struct blkcg *bio_blkcg(struct bio *bio) | |
262 | { | |
db6638d7 DZ |
263 | if (bio && bio->bi_blkg) |
264 | return bio->bi_blkg->blkcg; | |
0fe061b9 | 265 | return NULL; |
fd383c2d TH |
266 | } |
267 | ||
d09d8df3 JB |
268 | static inline bool blk_cgroup_congested(void) |
269 | { | |
270 | struct cgroup_subsys_state *css; | |
271 | bool ret = false; | |
272 | ||
273 | rcu_read_lock(); | |
274 | css = kthread_blkcg(); | |
275 | if (!css) | |
276 | css = task_css(current, io_cgrp_id); | |
277 | while (css) { | |
278 | if (atomic_read(&css->cgroup->congestion_count)) { | |
279 | ret = true; | |
280 | break; | |
281 | } | |
282 | css = css->parent; | |
283 | } | |
284 | rcu_read_unlock(); | |
285 | return ret; | |
286 | } | |
287 | ||
c7c98fd3 JB |
288 | /** |
289 | * bio_issue_as_root_blkg - see if this bio needs to be issued as root blkg | |
290 | * @return: true if this bio needs to be submitted with the root blkg context. | |
291 | * | |
292 | * In order to avoid priority inversions we sometimes need to issue a bio as if | |
293 | * it were attached to the root blkg, and then backcharge to the actual owning | |
294 | * blkg. The idea is we do bio_blkcg() to look up the actual context for the | |
295 | * bio and attach the appropriate blkg to the bio. Then we call this helper and | |
296 | * if it is true run with the root blkg for that queue and then do any | |
297 | * backcharging to the originating cgroup once the io is complete. | |
298 | */ | |
299 | static inline bool bio_issue_as_root_blkg(struct bio *bio) | |
300 | { | |
0d1e0c7c | 301 | return (bio->bi_opf & (REQ_META | REQ_SWAP)) != 0; |
c7c98fd3 JB |
302 | } |
303 | ||
3c547865 TH |
304 | /** |
305 | * blkcg_parent - get the parent of a blkcg | |
306 | * @blkcg: blkcg of interest | |
307 | * | |
308 | * Return the parent blkcg of @blkcg. Can be called anytime. | |
309 | */ | |
310 | static inline struct blkcg *blkcg_parent(struct blkcg *blkcg) | |
311 | { | |
5c9d535b | 312 | return css_to_blkcg(blkcg->css.parent); |
3c547865 TH |
313 | } |
314 | ||
24f29046 TH |
315 | /** |
316 | * __blkg_lookup - internal version of blkg_lookup() | |
317 | * @blkcg: blkcg of interest | |
318 | * @q: request_queue of interest | |
319 | * @update_hint: whether to update lookup hint with the result or not | |
320 | * | |
321 | * This is internal version and shouldn't be used by policy | |
322 | * implementations. Looks up blkgs for the @blkcg - @q pair regardless of | |
323 | * @q's bypass state. If @update_hint is %true, the caller should be | |
324 | * holding @q->queue_lock and lookup hint is updated on success. | |
325 | */ | |
326 | static inline struct blkcg_gq *__blkg_lookup(struct blkcg *blkcg, | |
327 | struct request_queue *q, | |
328 | bool update_hint) | |
329 | { | |
330 | struct blkcg_gq *blkg; | |
331 | ||
85b6bc9d TH |
332 | if (blkcg == &blkcg_root) |
333 | return q->root_blkg; | |
334 | ||
24f29046 TH |
335 | blkg = rcu_dereference(blkcg->blkg_hint); |
336 | if (blkg && blkg->q == q) | |
337 | return blkg; | |
338 | ||
339 | return blkg_lookup_slowpath(blkcg, q, update_hint); | |
340 | } | |
341 | ||
342 | /** | |
343 | * blkg_lookup - lookup blkg for the specified blkcg - q pair | |
344 | * @blkcg: blkcg of interest | |
345 | * @q: request_queue of interest | |
346 | * | |
347 | * Lookup blkg for the @blkcg - @q pair. This function should be called | |
012d4a65 | 348 | * under RCU read lock. |
24f29046 TH |
349 | */ |
350 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, | |
351 | struct request_queue *q) | |
352 | { | |
353 | WARN_ON_ONCE(!rcu_read_lock_held()); | |
24f29046 TH |
354 | return __blkg_lookup(blkcg, q, false); |
355 | } | |
356 | ||
6bad9b21 | 357 | /** |
b86d865c | 358 | * blk_queue_root_blkg - return blkg for the (blkcg_root, @q) pair |
6bad9b21 BVA |
359 | * @q: request_queue of interest |
360 | * | |
361 | * Lookup blkg for @q at the root level. See also blkg_lookup(). | |
362 | */ | |
b86d865c | 363 | static inline struct blkcg_gq *blk_queue_root_blkg(struct request_queue *q) |
6bad9b21 | 364 | { |
b86d865c | 365 | return q->root_blkg; |
6bad9b21 BVA |
366 | } |
367 | ||
0381411e TH |
368 | /** |
369 | * blkg_to_pdata - get policy private data | |
370 | * @blkg: blkg of interest | |
371 | * @pol: policy of interest | |
372 | * | |
373 | * Return pointer to private data associated with the @blkg-@pol pair. | |
374 | */ | |
f95a04af TH |
375 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
376 | struct blkcg_policy *pol) | |
0381411e | 377 | { |
f95a04af | 378 | return blkg ? blkg->pd[pol->plid] : NULL; |
0381411e TH |
379 | } |
380 | ||
e48453c3 AA |
381 | static inline struct blkcg_policy_data *blkcg_to_cpd(struct blkcg *blkcg, |
382 | struct blkcg_policy *pol) | |
383 | { | |
81437648 | 384 | return blkcg ? blkcg->cpd[pol->plid] : NULL; |
e48453c3 AA |
385 | } |
386 | ||
0381411e TH |
387 | /** |
388 | * pdata_to_blkg - get blkg associated with policy private data | |
f95a04af | 389 | * @pd: policy private data of interest |
0381411e | 390 | * |
f95a04af | 391 | * @pd is policy private data. Determine the blkg it's associated with. |
0381411e | 392 | */ |
f95a04af | 393 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) |
0381411e | 394 | { |
f95a04af | 395 | return pd ? pd->blkg : NULL; |
0381411e TH |
396 | } |
397 | ||
81437648 TH |
398 | static inline struct blkcg *cpd_to_blkcg(struct blkcg_policy_data *cpd) |
399 | { | |
400 | return cpd ? cpd->blkcg : NULL; | |
401 | } | |
402 | ||
59b57717 DZF |
403 | extern void blkcg_destroy_blkgs(struct blkcg *blkcg); |
404 | ||
59b57717 | 405 | /** |
d866dbf6 | 406 | * blkcg_pin_online - pin online state |
59b57717 DZF |
407 | * @blkcg: blkcg of interest |
408 | * | |
d866dbf6 TH |
409 | * While pinned, a blkcg is kept online. This is primarily used to |
410 | * impedance-match blkg and cgwb lifetimes so that blkg doesn't go offline | |
411 | * while an associated cgwb is still active. | |
59b57717 | 412 | */ |
d866dbf6 | 413 | static inline void blkcg_pin_online(struct blkcg *blkcg) |
59b57717 | 414 | { |
d866dbf6 | 415 | refcount_inc(&blkcg->online_pin); |
59b57717 DZF |
416 | } |
417 | ||
418 | /** | |
d866dbf6 | 419 | * blkcg_unpin_online - unpin online state |
59b57717 DZF |
420 | * @blkcg: blkcg of interest |
421 | * | |
d866dbf6 TH |
422 | * This is primarily used to impedance-match blkg and cgwb lifetimes so |
423 | * that blkg doesn't go offline while an associated cgwb is still active. | |
424 | * When this count goes to zero, all active cgwbs have finished so the | |
59b57717 | 425 | * blkcg can continue destruction by calling blkcg_destroy_blkgs(). |
59b57717 | 426 | */ |
d866dbf6 | 427 | static inline void blkcg_unpin_online(struct blkcg *blkcg) |
59b57717 | 428 | { |
4308a434 TH |
429 | do { |
430 | if (!refcount_dec_and_test(&blkcg->online_pin)) | |
431 | break; | |
59b57717 | 432 | blkcg_destroy_blkgs(blkcg); |
4308a434 TH |
433 | blkcg = blkcg_parent(blkcg); |
434 | } while (blkcg); | |
59b57717 DZF |
435 | } |
436 | ||
54e7ed12 TH |
437 | /** |
438 | * blkg_path - format cgroup path of blkg | |
439 | * @blkg: blkg of interest | |
440 | * @buf: target buffer | |
441 | * @buflen: target buffer length | |
442 | * | |
443 | * Format the path of the cgroup of @blkg into @buf. | |
444 | */ | |
3c798398 | 445 | static inline int blkg_path(struct blkcg_gq *blkg, char *buf, int buflen) |
afc24d49 | 446 | { |
4c737b41 | 447 | return cgroup_path(blkg->blkcg->css.cgroup, buf, buflen); |
afc24d49 VG |
448 | } |
449 | ||
1adaf3dd TH |
450 | /** |
451 | * blkg_get - get a blkg reference | |
452 | * @blkg: blkg to get | |
453 | * | |
a5049a8a | 454 | * The caller should be holding an existing reference. |
1adaf3dd | 455 | */ |
3c798398 | 456 | static inline void blkg_get(struct blkcg_gq *blkg) |
1adaf3dd | 457 | { |
7fcf2b03 | 458 | percpu_ref_get(&blkg->refcnt); |
1adaf3dd TH |
459 | } |
460 | ||
d09d8df3 | 461 | /** |
7754f669 | 462 | * blkg_tryget - try and get a blkg reference |
d09d8df3 JB |
463 | * @blkg: blkg to get |
464 | * | |
465 | * This is for use when doing an RCU lookup of the blkg. We may be in the midst | |
466 | * of freeing this blkg, so we can only use it if the refcnt is not zero. | |
467 | */ | |
7754f669 | 468 | static inline bool blkg_tryget(struct blkcg_gq *blkg) |
d09d8df3 | 469 | { |
6ab21879 | 470 | return blkg && percpu_ref_tryget(&blkg->refcnt); |
d09d8df3 JB |
471 | } |
472 | ||
1adaf3dd TH |
473 | /** |
474 | * blkg_put - put a blkg reference | |
475 | * @blkg: blkg to put | |
1adaf3dd | 476 | */ |
3c798398 | 477 | static inline void blkg_put(struct blkcg_gq *blkg) |
1adaf3dd | 478 | { |
7fcf2b03 | 479 | percpu_ref_put(&blkg->refcnt); |
1adaf3dd TH |
480 | } |
481 | ||
dd4a4ffc TH |
482 | /** |
483 | * blkg_for_each_descendant_pre - pre-order walk of a blkg's descendants | |
484 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 485 | * @pos_css: used for iteration |
dd4a4ffc TH |
486 | * @p_blkg: target blkg to walk descendants of |
487 | * | |
488 | * Walk @c_blkg through the descendants of @p_blkg. Must be used with RCU | |
489 | * read locked. If called under either blkcg or queue lock, the iteration | |
490 | * is guaranteed to include all and only online blkgs. The caller may | |
492eb21b | 491 | * update @pos_css by calling css_rightmost_descendant() to skip subtree. |
bd8815a6 | 492 | * @p_blkg is included in the iteration and the first node to be visited. |
dd4a4ffc | 493 | */ |
492eb21b TH |
494 | #define blkg_for_each_descendant_pre(d_blkg, pos_css, p_blkg) \ |
495 | css_for_each_descendant_pre((pos_css), &(p_blkg)->blkcg->css) \ | |
496 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
dd4a4ffc TH |
497 | (p_blkg)->q, false))) |
498 | ||
aa539cb3 TH |
499 | /** |
500 | * blkg_for_each_descendant_post - post-order walk of a blkg's descendants | |
501 | * @d_blkg: loop cursor pointing to the current descendant | |
492eb21b | 502 | * @pos_css: used for iteration |
aa539cb3 TH |
503 | * @p_blkg: target blkg to walk descendants of |
504 | * | |
505 | * Similar to blkg_for_each_descendant_pre() but performs post-order | |
bd8815a6 TH |
506 | * traversal instead. Synchronization rules are the same. @p_blkg is |
507 | * included in the iteration and the last node to be visited. | |
aa539cb3 | 508 | */ |
492eb21b TH |
509 | #define blkg_for_each_descendant_post(d_blkg, pos_css, p_blkg) \ |
510 | css_for_each_descendant_post((pos_css), &(p_blkg)->blkcg->css) \ | |
511 | if (((d_blkg) = __blkg_lookup(css_to_blkcg(pos_css), \ | |
aa539cb3 TH |
512 | (p_blkg)->q, false))) |
513 | ||
d3f77dfd TH |
514 | bool __blkcg_punt_bio_submit(struct bio *bio); |
515 | ||
516 | static inline bool blkcg_punt_bio_submit(struct bio *bio) | |
517 | { | |
518 | if (bio->bi_opf & REQ_CGROUP_PUNT) | |
519 | return __blkcg_punt_bio_submit(bio); | |
520 | else | |
521 | return false; | |
522 | } | |
e439bedf DZ |
523 | |
524 | static inline void blkcg_bio_issue_init(struct bio *bio) | |
525 | { | |
526 | bio_issue_init(&bio->bi_issue, bio_sectors(bio)); | |
527 | } | |
528 | ||
d09d8df3 JB |
529 | static inline void blkcg_use_delay(struct blkcg_gq *blkg) |
530 | { | |
54c52e10 TH |
531 | if (WARN_ON_ONCE(atomic_read(&blkg->use_delay) < 0)) |
532 | return; | |
d09d8df3 JB |
533 | if (atomic_add_return(1, &blkg->use_delay) == 1) |
534 | atomic_inc(&blkg->blkcg->css.cgroup->congestion_count); | |
535 | } | |
536 | ||
537 | static inline int blkcg_unuse_delay(struct blkcg_gq *blkg) | |
538 | { | |
539 | int old = atomic_read(&blkg->use_delay); | |
540 | ||
54c52e10 TH |
541 | if (WARN_ON_ONCE(old < 0)) |
542 | return 0; | |
d09d8df3 JB |
543 | if (old == 0) |
544 | return 0; | |
545 | ||
546 | /* | |
547 | * We do this song and dance because we can race with somebody else | |
548 | * adding or removing delay. If we just did an atomic_dec we'd end up | |
549 | * negative and we'd already be in trouble. We need to subtract 1 and | |
550 | * then check to see if we were the last delay so we can drop the | |
551 | * congestion count on the cgroup. | |
552 | */ | |
553 | while (old) { | |
554 | int cur = atomic_cmpxchg(&blkg->use_delay, old, old - 1); | |
555 | if (cur == old) | |
556 | break; | |
557 | old = cur; | |
558 | } | |
559 | ||
560 | if (old == 0) | |
561 | return 0; | |
562 | if (old == 1) | |
563 | atomic_dec(&blkg->blkcg->css.cgroup->congestion_count); | |
564 | return 1; | |
565 | } | |
566 | ||
54c52e10 TH |
567 | /** |
568 | * blkcg_set_delay - Enable allocator delay mechanism with the specified delay amount | |
569 | * @blkg: target blkg | |
570 | * @delay: delay duration in nsecs | |
571 | * | |
572 | * When enabled with this function, the delay is not decayed and must be | |
573 | * explicitly cleared with blkcg_clear_delay(). Must not be mixed with | |
574 | * blkcg_[un]use_delay() and blkcg_add_delay() usages. | |
575 | */ | |
576 | static inline void blkcg_set_delay(struct blkcg_gq *blkg, u64 delay) | |
577 | { | |
578 | int old = atomic_read(&blkg->use_delay); | |
579 | ||
580 | /* We only want 1 person setting the congestion count for this blkg. */ | |
581 | if (!old && atomic_cmpxchg(&blkg->use_delay, old, -1) == old) | |
582 | atomic_inc(&blkg->blkcg->css.cgroup->congestion_count); | |
583 | ||
584 | atomic64_set(&blkg->delay_nsec, delay); | |
585 | } | |
586 | ||
587 | /** | |
588 | * blkcg_clear_delay - Disable allocator delay mechanism | |
589 | * @blkg: target blkg | |
590 | * | |
591 | * Disable use_delay mechanism. See blkcg_set_delay(). | |
592 | */ | |
d09d8df3 JB |
593 | static inline void blkcg_clear_delay(struct blkcg_gq *blkg) |
594 | { | |
595 | int old = atomic_read(&blkg->use_delay); | |
54c52e10 | 596 | |
d09d8df3 | 597 | /* We only want 1 person clearing the congestion count for this blkg. */ |
54c52e10 TH |
598 | if (old && atomic_cmpxchg(&blkg->use_delay, old, 0) == old) |
599 | atomic_dec(&blkg->blkcg->css.cgroup->congestion_count); | |
d09d8df3 JB |
600 | } |
601 | ||
db18a53e | 602 | void blk_cgroup_bio_start(struct bio *bio); |
d09d8df3 JB |
603 | void blkcg_add_delay(struct blkcg_gq *blkg, u64 now, u64 delta); |
604 | void blkcg_schedule_throttle(struct request_queue *q, bool use_memdelay); | |
605 | void blkcg_maybe_throttle_current(void); | |
36558c8a TH |
606 | #else /* CONFIG_BLK_CGROUP */ |
607 | ||
efa7d1c7 TH |
608 | struct blkcg { |
609 | }; | |
2f5ea477 | 610 | |
f95a04af TH |
611 | struct blkg_policy_data { |
612 | }; | |
613 | ||
e48453c3 AA |
614 | struct blkcg_policy_data { |
615 | }; | |
616 | ||
3c798398 | 617 | struct blkcg_gq { |
2f5ea477 JA |
618 | }; |
619 | ||
3c798398 | 620 | struct blkcg_policy { |
3e252066 VG |
621 | }; |
622 | ||
496d5e75 TH |
623 | #define blkcg_root_css ((struct cgroup_subsys_state *)ERR_PTR(-EINVAL)) |
624 | ||
d09d8df3 JB |
625 | static inline void blkcg_maybe_throttle_current(void) { } |
626 | static inline bool blk_cgroup_congested(void) { return false; } | |
627 | ||
efa7d1c7 TH |
628 | #ifdef CONFIG_BLOCK |
629 | ||
d09d8df3 JB |
630 | static inline void blkcg_schedule_throttle(struct request_queue *q, bool use_memdelay) { } |
631 | ||
3c798398 | 632 | static inline struct blkcg_gq *blkg_lookup(struct blkcg *blkcg, void *key) { return NULL; } |
b86d865c BVA |
633 | static inline struct blkcg_gq *blk_queue_root_blkg(struct request_queue *q) |
634 | { return NULL; } | |
5efd6113 | 635 | static inline int blkcg_init_queue(struct request_queue *q) { return 0; } |
5efd6113 | 636 | static inline void blkcg_exit_queue(struct request_queue *q) { } |
d5bf0291 | 637 | static inline int blkcg_policy_register(struct blkcg_policy *pol) { return 0; } |
3c798398 | 638 | static inline void blkcg_policy_unregister(struct blkcg_policy *pol) { } |
a2b1693b | 639 | static inline int blkcg_activate_policy(struct request_queue *q, |
3c798398 | 640 | const struct blkcg_policy *pol) { return 0; } |
a2b1693b | 641 | static inline void blkcg_deactivate_policy(struct request_queue *q, |
3c798398 TH |
642 | const struct blkcg_policy *pol) { } |
643 | ||
0fe061b9 | 644 | static inline struct blkcg *__bio_blkcg(struct bio *bio) { return NULL; } |
b1208b56 | 645 | static inline struct blkcg *bio_blkcg(struct bio *bio) { return NULL; } |
a051661c | 646 | |
f95a04af TH |
647 | static inline struct blkg_policy_data *blkg_to_pd(struct blkcg_gq *blkg, |
648 | struct blkcg_policy *pol) { return NULL; } | |
649 | static inline struct blkcg_gq *pd_to_blkg(struct blkg_policy_data *pd) { return NULL; } | |
3c798398 TH |
650 | static inline char *blkg_path(struct blkcg_gq *blkg) { return NULL; } |
651 | static inline void blkg_get(struct blkcg_gq *blkg) { } | |
652 | static inline void blkg_put(struct blkcg_gq *blkg) { } | |
afc24d49 | 653 | |
d3f77dfd | 654 | static inline bool blkcg_punt_bio_submit(struct bio *bio) { return false; } |
e439bedf | 655 | static inline void blkcg_bio_issue_init(struct bio *bio) { } |
db18a53e | 656 | static inline void blk_cgroup_bio_start(struct bio *bio) { } |
ae118896 | 657 | |
a051661c TH |
658 | #define blk_queue_for_each_rl(rl, q) \ |
659 | for ((rl) = &(q)->root_rl; (rl); (rl) = NULL) | |
660 | ||
efa7d1c7 | 661 | #endif /* CONFIG_BLOCK */ |
36558c8a TH |
662 | #endif /* CONFIG_BLK_CGROUP */ |
663 | #endif /* _BLK_CGROUP_H */ |