Blame view
block/cfq-iosched.c
109 KB
1da177e4c
|
1 |
/* |
1da177e4c
|
2 3 4 5 6 |
* CFQ, or complete fairness queueing, disk scheduler. * * Based on ideas from a previously unfinished io * scheduler (round robin per-process disk scheduling) and Andrea Arcangeli. * |
0fe234795
|
7 |
* Copyright (C) 2003 Jens Axboe <axboe@kernel.dk> |
1da177e4c
|
8 |
*/ |
1da177e4c
|
9 |
#include <linux/module.h> |
5a0e3ad6a
|
10 |
#include <linux/slab.h> |
1cc9be68e
|
11 12 |
#include <linux/blkdev.h> #include <linux/elevator.h> |
ad5ebd2fa
|
13 |
#include <linux/jiffies.h> |
1da177e4c
|
14 |
#include <linux/rbtree.h> |
22e2c507c
|
15 |
#include <linux/ioprio.h> |
7b679138b
|
16 |
#include <linux/blktrace_api.h> |
e98ef89b3
|
17 |
#include "cfq.h" |
1da177e4c
|
18 19 20 21 |
/* * tunables */ |
fe094d98e
|
22 |
/* max queue in one round of service */ |
abc3c744d
|
23 |
static const int cfq_quantum = 8; |
64100099e
|
24 |
static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 }; |
fe094d98e
|
25 26 27 28 |
/* maximum backwards seek, in KiB */ static const int cfq_back_max = 16 * 1024; /* penalty of a backwards seek */ static const int cfq_back_penalty = 2; |
64100099e
|
29 |
static const int cfq_slice_sync = HZ / 10; |
3b18152c3
|
30 |
static int cfq_slice_async = HZ / 25; |
64100099e
|
31 |
static const int cfq_slice_async_rq = 2; |
caaa5f9f0
|
32 |
static int cfq_slice_idle = HZ / 125; |
80bdf0c78
|
33 |
static int cfq_group_idle = HZ / 125; |
5db5d6427
|
34 35 |
static const int cfq_target_latency = HZ * 3/10; /* 300 ms */ static const int cfq_hist_divisor = 4; |
22e2c507c
|
36 |
|
d9e7620e6
|
37 |
/* |
0871714e0
|
38 |
* offset from end of service tree |
d9e7620e6
|
39 |
*/ |
0871714e0
|
40 |
#define CFQ_IDLE_DELAY (HZ / 5) |
d9e7620e6
|
41 42 43 44 45 |
/* * below this threshold, we consider thinktime immediate */ #define CFQ_MIN_TT (2) |
22e2c507c
|
46 |
#define CFQ_SLICE_SCALE (5) |
45333d5a3
|
47 |
#define CFQ_HW_QUEUE_MIN (5) |
25bc6b077
|
48 |
#define CFQ_SERVICE_SHIFT 12 |
22e2c507c
|
49 |
|
3dde36dde
|
50 |
#define CFQQ_SEEK_THR (sector_t)(8 * 100) |
e9ce335df
|
51 |
#define CFQQ_CLOSE_THR (sector_t)(8 * 1024) |
41647e7a9
|
52 |
#define CFQQ_SECT_THR_NONROT (sector_t)(2 * 32) |
3dde36dde
|
53 |
#define CFQQ_SEEKY(cfqq) (hweight32(cfqq->seek_history) > 32/8) |
ae54abed6
|
54 |
|
fe094d98e
|
55 |
#define RQ_CIC(rq) \ |
c186794db
|
56 57 58 |
((struct cfq_io_context *) (rq)->elevator_private[0]) #define RQ_CFQQ(rq) (struct cfq_queue *) ((rq)->elevator_private[1]) #define RQ_CFQG(rq) (struct cfq_group *) ((rq)->elevator_private[2]) |
1da177e4c
|
59 |
|
e18b890bb
|
60 61 |
static struct kmem_cache *cfq_pool; static struct kmem_cache *cfq_ioc_pool; |
1da177e4c
|
62 |
|
245b2e70e
|
63 |
static DEFINE_PER_CPU(unsigned long, cfq_ioc_count); |
334e94de9
|
64 |
static struct completion *ioc_gone; |
9a11b4ed0
|
65 |
static DEFINE_SPINLOCK(ioc_gone_lock); |
334e94de9
|
66 |
|
80b15c738
|
67 68 |
static DEFINE_SPINLOCK(cic_index_lock); static DEFINE_IDA(cic_index_ida); |
22e2c507c
|
69 70 |
#define CFQ_PRIO_LISTS IOPRIO_BE_NR #define cfq_class_idle(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE) |
22e2c507c
|
71 |
#define cfq_class_rt(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_RT) |
206dc69b3
|
72 |
#define sample_valid(samples) ((samples) > 80) |
1fa8f6d68
|
73 |
#define rb_entry_cfqg(node) rb_entry((node), struct cfq_group, rb_node) |
206dc69b3
|
74 |
|
22e2c507c
|
75 |
/* |
cc09e2990
|
76 77 78 79 80 81 82 83 |
* Most of our rbtree usage is for sorting with min extraction, so * if we cache the leftmost node we don't have to walk down the tree * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should * move this into the elevator for the rq sorting as well. */ struct cfq_rb_root { struct rb_root rb; struct rb_node *left; |
aa6f6a3de
|
84 |
unsigned count; |
73e9ffdd0
|
85 |
unsigned total_weight; |
1fa8f6d68
|
86 |
u64 min_vdisktime; |
f5f2b6ceb
|
87 |
struct cfq_ttime ttime; |
cc09e2990
|
88 |
}; |
f5f2b6ceb
|
89 90 |
#define CFQ_RB_ROOT (struct cfq_rb_root) { .rb = RB_ROOT, \ .ttime = {.last_end_request = jiffies,},} |
cc09e2990
|
91 92 |
/* |
6118b70b3
|
93 94 95 96 |
* Per process-grouping structure */ struct cfq_queue { /* reference count */ |
30d7b9448
|
97 |
int ref; |
6118b70b3
|
98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 |
/* various state flags, see below */ unsigned int flags; /* parent cfq_data */ struct cfq_data *cfqd; /* service_tree member */ struct rb_node rb_node; /* service_tree key */ unsigned long rb_key; /* prio tree member */ struct rb_node p_node; /* prio tree root we belong to, if any */ struct rb_root *p_root; /* sorted list of pending requests */ struct rb_root sort_list; /* if fifo isn't expired, next request to serve */ struct request *next_rq; /* requests queued in sort_list */ int queued[2]; /* currently allocated requests */ int allocated[2]; /* fifo list of requests in sort_list */ struct list_head fifo; |
dae739ebc
|
120 121 |
/* time when queue got scheduled in to dispatch first request. */ unsigned long dispatch_start; |
f75edf2dc
|
122 |
unsigned int allocated_slice; |
c4081ba5c
|
123 |
unsigned int slice_dispatch; |
dae739ebc
|
124 125 |
/* time when first request from queue completed and slice started. */ unsigned long slice_start; |
6118b70b3
|
126 127 |
unsigned long slice_end; long slice_resid; |
6118b70b3
|
128 |
|
65299a3b7
|
129 130 |
/* pending priority requests */ int prio_pending; |
6118b70b3
|
131 132 133 134 135 |
/* number of requests that are on the dispatch list or inside driver */ int dispatched; /* io prio of this group */ unsigned short ioprio, org_ioprio; |
4aede84b3
|
136 |
unsigned short ioprio_class; |
6118b70b3
|
137 |
|
c4081ba5c
|
138 |
pid_t pid; |
3dde36dde
|
139 |
u32 seek_history; |
b2c18e1e0
|
140 |
sector_t last_request_pos; |
aa6f6a3de
|
141 |
struct cfq_rb_root *service_tree; |
df5fe3e8e
|
142 |
struct cfq_queue *new_cfqq; |
cdb16e8f7
|
143 |
struct cfq_group *cfqg; |
c4e7893eb
|
144 145 |
/* Number of sectors dispatched from queue in single dispatch round */ unsigned long nr_sectors; |
6118b70b3
|
146 147 148 |
}; /* |
718eee057
|
149 |
* First index in the service_trees. |
c0324a020
|
150 151 152 |
* IDLE is handled separately, so it has negative index */ enum wl_prio_t { |
c0324a020
|
153 |
BE_WORKLOAD = 0, |
615f0259e
|
154 155 |
RT_WORKLOAD = 1, IDLE_WORKLOAD = 2, |
b4627321e
|
156 |
CFQ_PRIO_NR, |
c0324a020
|
157 158 159 |
}; /* |
718eee057
|
160 161 162 163 164 165 166 |
* Second index in the service_trees. */ enum wl_type_t { ASYNC_WORKLOAD = 0, SYNC_NOIDLE_WORKLOAD = 1, SYNC_WORKLOAD = 2 }; |
cdb16e8f7
|
167 168 |
/* This is per cgroup per device grouping structure */ struct cfq_group { |
1fa8f6d68
|
169 170 171 172 173 |
/* group service_tree member */ struct rb_node rb_node; /* group service_tree key */ u64 vdisktime; |
25bc6b077
|
174 |
unsigned int weight; |
8184f93ec
|
175 176 |
unsigned int new_weight; bool needs_update; |
1fa8f6d68
|
177 178 179 |
/* number of cfqq currently on this group */ int nr_cfqq; |
cdb16e8f7
|
180 |
/* |
4495a7d41
|
181 |
* Per group busy queues average. Useful for workload slice calc. We |
b4627321e
|
182 183 184 185 186 187 188 189 190 191 192 |
* create the array for each prio class but at run time it is used * only for RT and BE class and slot for IDLE class remains unused. * This is primarily done to avoid confusion and a gcc warning. */ unsigned int busy_queues_avg[CFQ_PRIO_NR]; /* * rr lists of queues with requests. We maintain service trees for * RT and BE classes. These trees are subdivided in subclasses * of SYNC, SYNC_NOIDLE and ASYNC based on workload type. For IDLE * class there is no subclassification and all the cfq queues go on * a single tree service_tree_idle. |
cdb16e8f7
|
193 194 195 196 |
* Counts are embedded in the cfq_rb_root */ struct cfq_rb_root service_trees[2][3]; struct cfq_rb_root service_tree_idle; |
dae739ebc
|
197 198 199 200 |
unsigned long saved_workload_slice; enum wl_type_t saved_workload; enum wl_prio_t saved_serving_prio; |
25fb5169d
|
201 202 203 |
struct blkio_group blkg; #ifdef CONFIG_CFQ_GROUP_IOSCHED struct hlist_node cfqd_node; |
329a67815
|
204 |
int ref; |
25fb5169d
|
205 |
#endif |
80bdf0c78
|
206 207 |
/* number of requests that are on the dispatch list or inside driver */ int dispatched; |
7700fc4f6
|
208 |
struct cfq_ttime ttime; |
cdb16e8f7
|
209 |
}; |
718eee057
|
210 211 |
/* |
22e2c507c
|
212 213 |
* Per block device queue structure */ |
1da177e4c
|
214 |
struct cfq_data { |
165125e1e
|
215 |
struct request_queue *queue; |
1fa8f6d68
|
216 217 |
/* Root service tree for cfq_groups */ struct cfq_rb_root grp_service_tree; |
cdb16e8f7
|
218 |
struct cfq_group root_group; |
22e2c507c
|
219 220 |
/* |
c0324a020
|
221 |
* The priority currently being served |
22e2c507c
|
222 |
*/ |
c0324a020
|
223 |
enum wl_prio_t serving_prio; |
718eee057
|
224 225 |
enum wl_type_t serving_type; unsigned long workload_expires; |
cdb16e8f7
|
226 |
struct cfq_group *serving_group; |
a36e71f99
|
227 228 229 230 231 232 233 |
/* * Each priority tree is sorted by next_request position. These * trees are used when determining if two or more queues are * interleaving requests (see cfq_close_cooperator). */ struct rb_root prio_trees[CFQ_PRIO_LISTS]; |
22e2c507c
|
234 |
unsigned int busy_queues; |
ef8a41df8
|
235 |
unsigned int busy_sync_queues; |
22e2c507c
|
236 |
|
53c583d22
|
237 238 |
int rq_in_driver; int rq_in_flight[2]; |
45333d5a3
|
239 240 241 242 243 |
/* * queue-depth detection */ int rq_queued; |
25776e359
|
244 |
int hw_tag; |
e459dd08f
|
245 246 247 248 249 250 251 252 |
/* * hw_tag can be * -1 => indeterminate, (cfq will behave as if NCQ is present, to allow better detection) * 1 => NCQ is present (hw_tag_est_depth is the estimated max depth) * 0 => no NCQ */ int hw_tag_est_depth; unsigned int hw_tag_samples; |
1da177e4c
|
253 |
|
22e2c507c
|
254 |
/* |
22e2c507c
|
255 256 257 |
* idle window management */ struct timer_list idle_slice_timer; |
23e018a1b
|
258 |
struct work_struct unplug_work; |
1da177e4c
|
259 |
|
22e2c507c
|
260 261 |
struct cfq_queue *active_queue; struct cfq_io_context *active_cic; |
22e2c507c
|
262 |
|
c2dea2d1f
|
263 264 265 266 267 |
/* * async queue for each priority case */ struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR]; struct cfq_queue *async_idle_cfqq; |
15c31be4d
|
268 |
|
6d048f531
|
269 |
sector_t last_position; |
1da177e4c
|
270 |
|
1da177e4c
|
271 272 273 274 |
/* * tunables, see top of file */ unsigned int cfq_quantum; |
22e2c507c
|
275 |
unsigned int cfq_fifo_expire[2]; |
1da177e4c
|
276 277 |
unsigned int cfq_back_penalty; unsigned int cfq_back_max; |
22e2c507c
|
278 279 280 |
unsigned int cfq_slice[2]; unsigned int cfq_slice_async_rq; unsigned int cfq_slice_idle; |
80bdf0c78
|
281 |
unsigned int cfq_group_idle; |
963b72fc6
|
282 |
unsigned int cfq_latency; |
d9ff41879
|
283 |
|
80b15c738
|
284 |
unsigned int cic_index; |
d9ff41879
|
285 |
struct list_head cic_list; |
1da177e4c
|
286 |
|
6118b70b3
|
287 288 289 290 |
/* * Fallback dummy cfqq for extreme OOM conditions */ struct cfq_queue oom_cfqq; |
365722bb9
|
291 |
|
573412b29
|
292 |
unsigned long last_delayed_sync; |
25fb5169d
|
293 294 295 |
/* List of cfq groups being managed on this device*/ struct hlist_head cfqg_list; |
56edf7d75
|
296 297 298 |
/* Number of groups which are on blkcg->blkg_list */ unsigned int nr_blkcg_linked_grps; |
1da177e4c
|
299 |
}; |
25fb5169d
|
300 |
static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd); |
cdb16e8f7
|
301 302 |
static struct cfq_rb_root *service_tree_for(struct cfq_group *cfqg, enum wl_prio_t prio, |
65b32a573
|
303 |
enum wl_type_t type) |
c0324a020
|
304 |
{ |
1fa8f6d68
|
305 306 |
if (!cfqg) return NULL; |
c0324a020
|
307 |
if (prio == IDLE_WORKLOAD) |
cdb16e8f7
|
308 |
return &cfqg->service_tree_idle; |
c0324a020
|
309 |
|
cdb16e8f7
|
310 |
return &cfqg->service_trees[prio][type]; |
c0324a020
|
311 |
} |
3b18152c3
|
312 |
enum cfqq_state_flags { |
b0b8d7494
|
313 314 |
CFQ_CFQQ_FLAG_on_rr = 0, /* on round-robin busy list */ CFQ_CFQQ_FLAG_wait_request, /* waiting for a request */ |
b029195dd
|
315 |
CFQ_CFQQ_FLAG_must_dispatch, /* must be allowed a dispatch */ |
b0b8d7494
|
316 |
CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */ |
b0b8d7494
|
317 318 319 |
CFQ_CFQQ_FLAG_fifo_expire, /* FIFO checked in this slice */ CFQ_CFQQ_FLAG_idle_window, /* slice idling enabled */ CFQ_CFQQ_FLAG_prio_changed, /* task priority has changed */ |
44f7c1606
|
320 |
CFQ_CFQQ_FLAG_slice_new, /* no requests dispatched in slice */ |
91fac317a
|
321 |
CFQ_CFQQ_FLAG_sync, /* synchronous queue */ |
b3b6d0408
|
322 |
CFQ_CFQQ_FLAG_coop, /* cfqq is shared */ |
ae54abed6
|
323 |
CFQ_CFQQ_FLAG_split_coop, /* shared cfqq will be splitted */ |
76280aff1
|
324 |
CFQ_CFQQ_FLAG_deep, /* sync cfqq experienced large depth */ |
f75edf2dc
|
325 |
CFQ_CFQQ_FLAG_wait_busy, /* Waiting for next request */ |
3b18152c3
|
326 327 328 329 330 |
}; #define CFQ_CFQQ_FNS(name) \ static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq) \ { \ |
fe094d98e
|
331 |
(cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name); \ |
3b18152c3
|
332 333 334 |
} \ static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq) \ { \ |
fe094d98e
|
335 |
(cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name); \ |
3b18152c3
|
336 337 338 |
} \ static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq) \ { \ |
fe094d98e
|
339 |
return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0; \ |
3b18152c3
|
340 341 342 343 |
} CFQ_CFQQ_FNS(on_rr); CFQ_CFQQ_FNS(wait_request); |
b029195dd
|
344 |
CFQ_CFQQ_FNS(must_dispatch); |
3b18152c3
|
345 |
CFQ_CFQQ_FNS(must_alloc_slice); |
3b18152c3
|
346 347 348 |
CFQ_CFQQ_FNS(fifo_expire); CFQ_CFQQ_FNS(idle_window); CFQ_CFQQ_FNS(prio_changed); |
44f7c1606
|
349 |
CFQ_CFQQ_FNS(slice_new); |
91fac317a
|
350 |
CFQ_CFQQ_FNS(sync); |
a36e71f99
|
351 |
CFQ_CFQQ_FNS(coop); |
ae54abed6
|
352 |
CFQ_CFQQ_FNS(split_coop); |
76280aff1
|
353 |
CFQ_CFQQ_FNS(deep); |
f75edf2dc
|
354 |
CFQ_CFQQ_FNS(wait_busy); |
3b18152c3
|
355 |
#undef CFQ_CFQQ_FNS |
afc24d49c
|
356 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED |
2868ef7b3
|
357 358 359 |
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...) \ blk_add_trace_msg((cfqd)->queue, "cfq%d%c %s " fmt, (cfqq)->pid, \ cfq_cfqq_sync((cfqq)) ? 'S' : 'A', \ |
4495a7d41
|
360 |
blkg_path(&(cfqq)->cfqg->blkg), ##args) |
2868ef7b3
|
361 362 363 |
#define cfq_log_cfqg(cfqd, cfqg, fmt, args...) \ blk_add_trace_msg((cfqd)->queue, "%s " fmt, \ |
4495a7d41
|
364 |
blkg_path(&(cfqg)->blkg), ##args) \ |
2868ef7b3
|
365 366 |
#else |
7b679138b
|
367 368 |
#define cfq_log_cfqq(cfqd, cfqq, fmt, args...) \ blk_add_trace_msg((cfqd)->queue, "cfq%d " fmt, (cfqq)->pid, ##args) |
4495a7d41
|
369 |
#define cfq_log_cfqg(cfqd, cfqg, fmt, args...) do {} while (0) |
2868ef7b3
|
370 |
#endif |
7b679138b
|
371 372 |
#define cfq_log(cfqd, fmt, args...) \ blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args) |
615f0259e
|
373 374 375 376 377 378 379 380 381 |
/* Traverses through cfq group service trees */ #define for_each_cfqg_st(cfqg, i, j, st) \ for (i = 0; i <= IDLE_WORKLOAD; i++) \ for (j = 0, st = i < IDLE_WORKLOAD ? &cfqg->service_trees[i][j]\ : &cfqg->service_tree_idle; \ (i < IDLE_WORKLOAD && j <= SYNC_WORKLOAD) || \ (i == IDLE_WORKLOAD && j == 0); \ j++, st = i < IDLE_WORKLOAD ? \ &cfqg->service_trees[i][j]: NULL) \ |
f5f2b6ceb
|
382 383 384 385 386 387 388 389 390 391 392 393 |
static inline bool cfq_io_thinktime_big(struct cfq_data *cfqd, struct cfq_ttime *ttime, bool group_idle) { unsigned long slice; if (!sample_valid(ttime->ttime_samples)) return false; if (group_idle) slice = cfqd->cfq_group_idle; else slice = cfqd->cfq_slice_idle; return ttime->ttime_mean > slice; } |
615f0259e
|
394 |
|
02b35081f
|
395 396 397 398 399 400 401 402 403 404 405 406 407 408 |
static inline bool iops_mode(struct cfq_data *cfqd) { /* * If we are not idling on queues and it is a NCQ drive, parallel * execution of requests is on and measuring time is not possible * in most of the cases until and unless we drive shallower queue * depths and that becomes a performance bottleneck. In such cases * switch to start providing fairness in terms of number of IOs. */ if (!cfqd->cfq_slice_idle && cfqd->hw_tag) return true; else return false; } |
c0324a020
|
409 410 411 412 413 414 415 416 |
static inline enum wl_prio_t cfqq_prio(struct cfq_queue *cfqq) { if (cfq_class_idle(cfqq)) return IDLE_WORKLOAD; if (cfq_class_rt(cfqq)) return RT_WORKLOAD; return BE_WORKLOAD; } |
718eee057
|
417 418 419 420 421 422 423 424 425 |
static enum wl_type_t cfqq_type(struct cfq_queue *cfqq) { if (!cfq_cfqq_sync(cfqq)) return ASYNC_WORKLOAD; if (!cfq_cfqq_idle_window(cfqq)) return SYNC_NOIDLE_WORKLOAD; return SYNC_WORKLOAD; } |
58ff82f34
|
426 427 428 |
static inline int cfq_group_busy_queues_wl(enum wl_prio_t wl, struct cfq_data *cfqd, struct cfq_group *cfqg) |
c0324a020
|
429 430 |
{ if (wl == IDLE_WORKLOAD) |
cdb16e8f7
|
431 |
return cfqg->service_tree_idle.count; |
c0324a020
|
432 |
|
cdb16e8f7
|
433 434 435 |
return cfqg->service_trees[wl][ASYNC_WORKLOAD].count + cfqg->service_trees[wl][SYNC_NOIDLE_WORKLOAD].count + cfqg->service_trees[wl][SYNC_WORKLOAD].count; |
c0324a020
|
436 |
} |
f26bd1f0a
|
437 438 439 440 441 442 |
static inline int cfqg_busy_async_queues(struct cfq_data *cfqd, struct cfq_group *cfqg) { return cfqg->service_trees[RT_WORKLOAD][ASYNC_WORKLOAD].count + cfqg->service_trees[BE_WORKLOAD][ASYNC_WORKLOAD].count; } |
165125e1e
|
443 |
static void cfq_dispatch_insert(struct request_queue *, struct request *); |
a6151c3a5
|
444 |
static struct cfq_queue *cfq_get_queue(struct cfq_data *, bool, |
fd0928df9
|
445 |
struct io_context *, gfp_t); |
4ac845a2e
|
446 |
static struct cfq_io_context *cfq_cic_lookup(struct cfq_data *, |
91fac317a
|
447 448 449 |
struct io_context *); static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_context *cic, |
a6151c3a5
|
450 |
bool is_sync) |
91fac317a
|
451 |
{ |
a6151c3a5
|
452 |
return cic->cfqq[is_sync]; |
91fac317a
|
453 454 455 |
} static inline void cic_set_cfqq(struct cfq_io_context *cic, |
a6151c3a5
|
456 |
struct cfq_queue *cfqq, bool is_sync) |
91fac317a
|
457 |
{ |
a6151c3a5
|
458 |
cic->cfqq[is_sync] = cfqq; |
91fac317a
|
459 |
} |
bca4b914b
|
460 |
#define CIC_DEAD_KEY 1ul |
80b15c738
|
461 |
#define CIC_DEAD_INDEX_SHIFT 1 |
bca4b914b
|
462 463 464 |
static inline void *cfqd_dead_key(struct cfq_data *cfqd) { |
80b15c738
|
465 |
return (void *)(cfqd->cic_index << CIC_DEAD_INDEX_SHIFT | CIC_DEAD_KEY); |
bca4b914b
|
466 467 468 469 470 471 472 473 474 475 476 |
} static inline struct cfq_data *cic_to_cfqd(struct cfq_io_context *cic) { struct cfq_data *cfqd = cic->key; if (unlikely((unsigned long) cfqd & CIC_DEAD_KEY)) return NULL; return cfqd; } |
91fac317a
|
477 478 479 480 |
/* * We regard a request as SYNC, if it's either a read or has the SYNC bit * set (in which case it could also be direct WRITE). */ |
a6151c3a5
|
481 |
static inline bool cfq_bio_sync(struct bio *bio) |
91fac317a
|
482 |
{ |
7b6d91dae
|
483 |
return bio_data_dir(bio) == READ || (bio->bi_rw & REQ_SYNC); |
91fac317a
|
484 |
} |
1da177e4c
|
485 |
|
1da177e4c
|
486 |
/* |
99f95e528
|
487 488 489 |
* scheduler run of queue, if there are requests pending and no one in the * driver that will restart queueing */ |
23e018a1b
|
490 |
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd) |
99f95e528
|
491 |
{ |
7b679138b
|
492 493 |
if (cfqd->busy_queues) { cfq_log(cfqd, "schedule dispatch"); |
23e018a1b
|
494 |
kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work); |
7b679138b
|
495 |
} |
99f95e528
|
496 |
} |
1da177e4c
|
497 |
/* |
44f7c1606
|
498 499 500 501 |
* Scale schedule slice based on io priority. Use the sync time slice only * if a queue is marked sync and has sync io queued. A sync queue with async * io only, should not get full sync slice length. */ |
a6151c3a5
|
502 |
static inline int cfq_prio_slice(struct cfq_data *cfqd, bool sync, |
d9e7620e6
|
503 |
unsigned short prio) |
44f7c1606
|
504 |
{ |
d9e7620e6
|
505 |
const int base_slice = cfqd->cfq_slice[sync]; |
44f7c1606
|
506 |
|
d9e7620e6
|
507 508 509 510 |
WARN_ON(prio >= IOPRIO_BE_NR); return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio)); } |
44f7c1606
|
511 |
|
d9e7620e6
|
512 513 514 515 |
static inline int cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq) { return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio); |
44f7c1606
|
516 |
} |
25bc6b077
|
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 |
static inline u64 cfq_scale_slice(unsigned long delta, struct cfq_group *cfqg) { u64 d = delta << CFQ_SERVICE_SHIFT; d = d * BLKIO_WEIGHT_DEFAULT; do_div(d, cfqg->weight); return d; } static inline u64 max_vdisktime(u64 min_vdisktime, u64 vdisktime) { s64 delta = (s64)(vdisktime - min_vdisktime); if (delta > 0) min_vdisktime = vdisktime; return min_vdisktime; } static inline u64 min_vdisktime(u64 min_vdisktime, u64 vdisktime) { s64 delta = (s64)(vdisktime - min_vdisktime); if (delta < 0) min_vdisktime = vdisktime; return min_vdisktime; } static void update_min_vdisktime(struct cfq_rb_root *st) { |
25bc6b077
|
546 |
struct cfq_group *cfqg; |
25bc6b077
|
547 548 |
if (st->left) { cfqg = rb_entry_cfqg(st->left); |
a60327107
|
549 550 |
st->min_vdisktime = max_vdisktime(st->min_vdisktime, cfqg->vdisktime); |
25bc6b077
|
551 |
} |
25bc6b077
|
552 |
} |
5db5d6427
|
553 554 555 556 557 |
/* * get averaged number of queues of RT/BE priority. * average is updated, with a formula that gives more weight to higher numbers, * to quickly follows sudden increases and decrease slowly */ |
58ff82f34
|
558 559 |
static inline unsigned cfq_group_get_avg_queues(struct cfq_data *cfqd, struct cfq_group *cfqg, bool rt) |
5869619cb
|
560 |
{ |
5db5d6427
|
561 562 563 |
unsigned min_q, max_q; unsigned mult = cfq_hist_divisor - 1; unsigned round = cfq_hist_divisor / 2; |
58ff82f34
|
564 |
unsigned busy = cfq_group_busy_queues_wl(rt, cfqd, cfqg); |
5db5d6427
|
565 |
|
58ff82f34
|
566 567 568 |
min_q = min(cfqg->busy_queues_avg[rt], busy); max_q = max(cfqg->busy_queues_avg[rt], busy); cfqg->busy_queues_avg[rt] = (mult * max_q + min_q + round) / |
5db5d6427
|
569 |
cfq_hist_divisor; |
58ff82f34
|
570 571 572 573 574 575 576 577 578 |
return cfqg->busy_queues_avg[rt]; } static inline unsigned cfq_group_slice(struct cfq_data *cfqd, struct cfq_group *cfqg) { struct cfq_rb_root *st = &cfqd->grp_service_tree; return cfq_target_latency * cfqg->weight / st->total_weight; |
5db5d6427
|
579 |
} |
c553f8e33
|
580 |
static inline unsigned |
ba5bd520f
|
581 |
cfq_scaled_cfqq_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
44f7c1606
|
582 |
{ |
5db5d6427
|
583 584 |
unsigned slice = cfq_prio_to_slice(cfqd, cfqq); if (cfqd->cfq_latency) { |
58ff82f34
|
585 586 587 588 589 590 |
/* * interested queues (we consider only the ones with the same * priority class in the cfq group) */ unsigned iq = cfq_group_get_avg_queues(cfqd, cfqq->cfqg, cfq_class_rt(cfqq)); |
5db5d6427
|
591 592 |
unsigned sync_slice = cfqd->cfq_slice[1]; unsigned expect_latency = sync_slice * iq; |
58ff82f34
|
593 594 595 |
unsigned group_slice = cfq_group_slice(cfqd, cfqq->cfqg); if (expect_latency > group_slice) { |
5db5d6427
|
596 597 598 599 600 601 602 |
unsigned base_low_slice = 2 * cfqd->cfq_slice_idle; /* scale low_slice according to IO priority * and sync vs async */ unsigned low_slice = min(slice, base_low_slice * slice / sync_slice); /* the adapted slice value is scaled to fit all iqs * into the target latency */ |
58ff82f34
|
603 |
slice = max(slice * group_slice / expect_latency, |
5db5d6427
|
604 605 606 |
low_slice); } } |
c553f8e33
|
607 608 609 610 611 612 |
return slice; } static inline void cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq) { |
ba5bd520f
|
613 |
unsigned slice = cfq_scaled_cfqq_slice(cfqd, cfqq); |
c553f8e33
|
614 |
|
dae739ebc
|
615 |
cfqq->slice_start = jiffies; |
5db5d6427
|
616 |
cfqq->slice_end = jiffies + slice; |
f75edf2dc
|
617 |
cfqq->allocated_slice = slice; |
7b679138b
|
618 |
cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies); |
44f7c1606
|
619 620 621 622 623 624 625 |
} /* * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end * isn't valid until the first request from the dispatch is activated * and the slice time set. */ |
a6151c3a5
|
626 |
static inline bool cfq_slice_used(struct cfq_queue *cfqq) |
44f7c1606
|
627 628 |
{ if (cfq_cfqq_slice_new(cfqq)) |
c1e44756f
|
629 |
return false; |
44f7c1606
|
630 |
if (time_before(jiffies, cfqq->slice_end)) |
c1e44756f
|
631 |
return false; |
44f7c1606
|
632 |
|
c1e44756f
|
633 |
return true; |
44f7c1606
|
634 635 636 |
} /* |
5e7053747
|
637 |
* Lifted from AS - choose which of rq1 and rq2 that is best served now. |
1da177e4c
|
638 |
* We choose the request that is closest to the head right now. Distance |
e8a99053e
|
639 |
* behind the head is penalized and only allowed to a certain extent. |
1da177e4c
|
640 |
*/ |
5e7053747
|
641 |
static struct request * |
cf7c25cf9
|
642 |
cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2, sector_t last) |
1da177e4c
|
643 |
{ |
cf7c25cf9
|
644 |
sector_t s1, s2, d1 = 0, d2 = 0; |
1da177e4c
|
645 |
unsigned long back_max; |
e8a99053e
|
646 647 648 |
#define CFQ_RQ1_WRAP 0x01 /* request 1 wraps */ #define CFQ_RQ2_WRAP 0x02 /* request 2 wraps */ unsigned wrap = 0; /* bit mask: requests behind the disk head? */ |
1da177e4c
|
649 |
|
5e7053747
|
650 651 652 653 |
if (rq1 == NULL || rq1 == rq2) return rq2; if (rq2 == NULL) return rq1; |
9c2c38a12
|
654 |
|
229836bd6
|
655 656 |
if (rq_is_sync(rq1) != rq_is_sync(rq2)) return rq_is_sync(rq1) ? rq1 : rq2; |
65299a3b7
|
657 658 |
if ((rq1->cmd_flags ^ rq2->cmd_flags) & REQ_PRIO) return rq1->cmd_flags & REQ_PRIO ? rq1 : rq2; |
b53d1ed73
|
659 |
|
83096ebf1
|
660 661 |
s1 = blk_rq_pos(rq1); s2 = blk_rq_pos(rq2); |
1da177e4c
|
662 |
|
1da177e4c
|
663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 |
/* * by definition, 1KiB is 2 sectors */ back_max = cfqd->cfq_back_max * 2; /* * Strict one way elevator _except_ in the case where we allow * short backward seeks which are biased as twice the cost of a * similar forward seek. */ if (s1 >= last) d1 = s1 - last; else if (s1 + back_max >= last) d1 = (last - s1) * cfqd->cfq_back_penalty; else |
e8a99053e
|
678 |
wrap |= CFQ_RQ1_WRAP; |
1da177e4c
|
679 680 681 682 683 684 |
if (s2 >= last) d2 = s2 - last; else if (s2 + back_max >= last) d2 = (last - s2) * cfqd->cfq_back_penalty; else |
e8a99053e
|
685 |
wrap |= CFQ_RQ2_WRAP; |
1da177e4c
|
686 687 |
/* Found required data */ |
e8a99053e
|
688 689 690 691 692 693 |
/* * By doing switch() on the bit mask "wrap" we avoid having to * check two variables for all permutations: --> faster! */ switch (wrap) { |
5e7053747
|
694 |
case 0: /* common case for CFQ: rq1 and rq2 not wrapped */ |
e8a99053e
|
695 |
if (d1 < d2) |
5e7053747
|
696 |
return rq1; |
e8a99053e
|
697 |
else if (d2 < d1) |
5e7053747
|
698 |
return rq2; |
e8a99053e
|
699 700 |
else { if (s1 >= s2) |
5e7053747
|
701 |
return rq1; |
e8a99053e
|
702 |
else |
5e7053747
|
703 |
return rq2; |
e8a99053e
|
704 |
} |
1da177e4c
|
705 |
|
e8a99053e
|
706 |
case CFQ_RQ2_WRAP: |
5e7053747
|
707 |
return rq1; |
e8a99053e
|
708 |
case CFQ_RQ1_WRAP: |
5e7053747
|
709 710 |
return rq2; case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */ |
e8a99053e
|
711 712 713 714 715 716 717 718 |
default: /* * Since both rqs are wrapped, * start with the one that's further behind head * (--> only *one* back seek required), * since back seek takes more time than forward. */ if (s1 <= s2) |
5e7053747
|
719 |
return rq1; |
1da177e4c
|
720 |
else |
5e7053747
|
721 |
return rq2; |
1da177e4c
|
722 723 |
} } |
498d3aa2b
|
724 725 726 |
/* * The below is leftmost cache rbtree addon */ |
0871714e0
|
727 |
static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root) |
cc09e2990
|
728 |
{ |
615f0259e
|
729 730 731 |
/* Service tree is empty */ if (!root->count) return NULL; |
cc09e2990
|
732 733 |
if (!root->left) root->left = rb_first(&root->rb); |
0871714e0
|
734 735 736 737 |
if (root->left) return rb_entry(root->left, struct cfq_queue, rb_node); return NULL; |
cc09e2990
|
738 |
} |
1fa8f6d68
|
739 740 741 742 743 744 745 746 747 748 |
static struct cfq_group *cfq_rb_first_group(struct cfq_rb_root *root) { if (!root->left) root->left = rb_first(&root->rb); if (root->left) return rb_entry_cfqg(root->left); return NULL; } |
a36e71f99
|
749 750 751 752 753 |
static void rb_erase_init(struct rb_node *n, struct rb_root *root) { rb_erase(n, root); RB_CLEAR_NODE(n); } |
cc09e2990
|
754 755 756 757 |
static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root) { if (root->left == n) root->left = NULL; |
a36e71f99
|
758 |
rb_erase_init(n, &root->rb); |
aa6f6a3de
|
759 |
--root->count; |
cc09e2990
|
760 |
} |
1da177e4c
|
761 762 763 |
/* * would be nice to take fifo expire time into account as well */ |
5e7053747
|
764 765 766 |
static struct request * cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq, struct request *last) |
1da177e4c
|
767 |
{ |
21183b07e
|
768 769 |
struct rb_node *rbnext = rb_next(&last->rb_node); struct rb_node *rbprev = rb_prev(&last->rb_node); |
5e7053747
|
770 |
struct request *next = NULL, *prev = NULL; |
1da177e4c
|
771 |
|
21183b07e
|
772 |
BUG_ON(RB_EMPTY_NODE(&last->rb_node)); |
1da177e4c
|
773 774 |
if (rbprev) |
5e7053747
|
775 |
prev = rb_entry_rq(rbprev); |
1da177e4c
|
776 |
|
21183b07e
|
777 |
if (rbnext) |
5e7053747
|
778 |
next = rb_entry_rq(rbnext); |
21183b07e
|
779 780 781 |
else { rbnext = rb_first(&cfqq->sort_list); if (rbnext && rbnext != &last->rb_node) |
5e7053747
|
782 |
next = rb_entry_rq(rbnext); |
21183b07e
|
783 |
} |
1da177e4c
|
784 |
|
cf7c25cf9
|
785 |
return cfq_choose_req(cfqd, next, prev, blk_rq_pos(last)); |
1da177e4c
|
786 |
} |
d9e7620e6
|
787 788 |
static unsigned long cfq_slice_offset(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
1da177e4c
|
789 |
{ |
d9e7620e6
|
790 791 792 |
/* * just an approximation, should be ok. */ |
cdb16e8f7
|
793 |
return (cfqq->cfqg->nr_cfqq - 1) * (cfq_prio_slice(cfqd, 1, 0) - |
464191c65
|
794 |
cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio)); |
d9e7620e6
|
795 |
} |
1fa8f6d68
|
796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 |
static inline s64 cfqg_key(struct cfq_rb_root *st, struct cfq_group *cfqg) { return cfqg->vdisktime - st->min_vdisktime; } static void __cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg) { struct rb_node **node = &st->rb.rb_node; struct rb_node *parent = NULL; struct cfq_group *__cfqg; s64 key = cfqg_key(st, cfqg); int left = 1; while (*node != NULL) { parent = *node; __cfqg = rb_entry_cfqg(parent); if (key < cfqg_key(st, __cfqg)) node = &parent->rb_left; else { node = &parent->rb_right; left = 0; } } if (left) st->left = &cfqg->rb_node; rb_link_node(&cfqg->rb_node, parent, node); rb_insert_color(&cfqg->rb_node, &st->rb); } static void |
8184f93ec
|
831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 |
cfq_update_group_weight(struct cfq_group *cfqg) { BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node)); if (cfqg->needs_update) { cfqg->weight = cfqg->new_weight; cfqg->needs_update = false; } } static void cfq_group_service_tree_add(struct cfq_rb_root *st, struct cfq_group *cfqg) { BUG_ON(!RB_EMPTY_NODE(&cfqg->rb_node)); cfq_update_group_weight(cfqg); __cfq_group_service_tree_add(st, cfqg); st->total_weight += cfqg->weight; } static void cfq_group_notify_queue_add(struct cfq_data *cfqd, struct cfq_group *cfqg) |
1fa8f6d68
|
852 853 854 855 856 857 |
{ struct cfq_rb_root *st = &cfqd->grp_service_tree; struct cfq_group *__cfqg; struct rb_node *n; cfqg->nr_cfqq++; |
760701bfe
|
858 |
if (!RB_EMPTY_NODE(&cfqg->rb_node)) |
1fa8f6d68
|
859 860 861 862 863 |
return; /* * Currently put the group at the end. Later implement something * so that groups get lesser vtime based on their weights, so that |
25985edce
|
864 |
* if group does not loose all if it was not continuously backlogged. |
1fa8f6d68
|
865 866 867 868 869 870 871 |
*/ n = rb_last(&st->rb); if (n) { __cfqg = rb_entry_cfqg(n); cfqg->vdisktime = __cfqg->vdisktime + CFQ_IDLE_DELAY; } else cfqg->vdisktime = st->min_vdisktime; |
8184f93ec
|
872 873 |
cfq_group_service_tree_add(st, cfqg); } |
1fa8f6d68
|
874 |
|
8184f93ec
|
875 876 877 878 879 880 |
static void cfq_group_service_tree_del(struct cfq_rb_root *st, struct cfq_group *cfqg) { st->total_weight -= cfqg->weight; if (!RB_EMPTY_NODE(&cfqg->rb_node)) cfq_rb_erase(&cfqg->rb_node, st); |
1fa8f6d68
|
881 882 883 |
} static void |
8184f93ec
|
884 |
cfq_group_notify_queue_del(struct cfq_data *cfqd, struct cfq_group *cfqg) |
1fa8f6d68
|
885 886 887 888 889 |
{ struct cfq_rb_root *st = &cfqd->grp_service_tree; BUG_ON(cfqg->nr_cfqq < 1); cfqg->nr_cfqq--; |
25bc6b077
|
890 |
|
1fa8f6d68
|
891 892 893 |
/* If there are other cfq queues under this group, don't delete it */ if (cfqg->nr_cfqq) return; |
2868ef7b3
|
894 |
cfq_log_cfqg(cfqd, cfqg, "del_from_rr group"); |
8184f93ec
|
895 |
cfq_group_service_tree_del(st, cfqg); |
dae739ebc
|
896 |
cfqg->saved_workload_slice = 0; |
e98ef89b3
|
897 |
cfq_blkiocg_update_dequeue_stats(&cfqg->blkg, 1); |
dae739ebc
|
898 |
} |
167400d34
|
899 900 |
static inline unsigned int cfq_cfqq_slice_usage(struct cfq_queue *cfqq, unsigned int *unaccounted_time) |
dae739ebc
|
901 |
{ |
f75edf2dc
|
902 |
unsigned int slice_used; |
dae739ebc
|
903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 |
/* * Queue got expired before even a single request completed or * got expired immediately after first request completion. */ if (!cfqq->slice_start || cfqq->slice_start == jiffies) { /* * Also charge the seek time incurred to the group, otherwise * if there are mutiple queues in the group, each can dispatch * a single request on seeky media and cause lots of seek time * and group will never know it. */ slice_used = max_t(unsigned, (jiffies - cfqq->dispatch_start), 1); } else { slice_used = jiffies - cfqq->slice_start; |
167400d34
|
919 920 |
if (slice_used > cfqq->allocated_slice) { *unaccounted_time = slice_used - cfqq->allocated_slice; |
f75edf2dc
|
921 |
slice_used = cfqq->allocated_slice; |
167400d34
|
922 923 924 925 |
} if (time_after(cfqq->slice_start, cfqq->dispatch_start)) *unaccounted_time += cfqq->slice_start - cfqq->dispatch_start; |
dae739ebc
|
926 |
} |
dae739ebc
|
927 928 929 930 |
return slice_used; } static void cfq_group_served(struct cfq_data *cfqd, struct cfq_group *cfqg, |
e5ff082e8
|
931 |
struct cfq_queue *cfqq) |
dae739ebc
|
932 933 |
{ struct cfq_rb_root *st = &cfqd->grp_service_tree; |
167400d34
|
934 |
unsigned int used_sl, charge, unaccounted_sl = 0; |
f26bd1f0a
|
935 936 937 938 |
int nr_sync = cfqg->nr_cfqq - cfqg_busy_async_queues(cfqd, cfqg) - cfqg->service_tree_idle.count; BUG_ON(nr_sync < 0); |
167400d34
|
939 |
used_sl = charge = cfq_cfqq_slice_usage(cfqq, &unaccounted_sl); |
dae739ebc
|
940 |
|
02b35081f
|
941 942 943 944 |
if (iops_mode(cfqd)) charge = cfqq->slice_dispatch; else if (!cfq_cfqq_sync(cfqq) && !nr_sync) charge = cfqq->allocated_slice; |
dae739ebc
|
945 946 |
/* Can't update vdisktime while group is on service tree */ |
8184f93ec
|
947 |
cfq_group_service_tree_del(st, cfqg); |
02b35081f
|
948 |
cfqg->vdisktime += cfq_scale_slice(charge, cfqg); |
8184f93ec
|
949 950 |
/* If a new weight was requested, update now, off tree */ cfq_group_service_tree_add(st, cfqg); |
dae739ebc
|
951 952 953 954 955 956 957 958 959 |
/* This group is being expired. Save the context */ if (time_after(cfqd->workload_expires, jiffies)) { cfqg->saved_workload_slice = cfqd->workload_expires - jiffies; cfqg->saved_workload = cfqd->serving_type; cfqg->saved_serving_prio = cfqd->serving_prio; } else cfqg->saved_workload_slice = 0; |
2868ef7b3
|
960 961 962 |
cfq_log_cfqg(cfqd, cfqg, "served: vt=%llu min_vt=%llu", cfqg->vdisktime, st->min_vdisktime); |
fd16d2631
|
963 964 965 966 |
cfq_log_cfqq(cfqq->cfqd, cfqq, "sl_used=%u disp=%u charge=%u iops=%u sect=%lu", used_sl, cfqq->slice_dispatch, charge, iops_mode(cfqd), cfqq->nr_sectors); |
167400d34
|
967 968 |
cfq_blkiocg_update_timeslice_used(&cfqg->blkg, used_sl, unaccounted_sl); |
e98ef89b3
|
969 |
cfq_blkiocg_set_start_empty_time(&cfqg->blkg); |
1fa8f6d68
|
970 |
} |
25fb5169d
|
971 972 973 974 975 976 977 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED static inline struct cfq_group *cfqg_of_blkg(struct blkio_group *blkg) { if (blkg) return container_of(blkg, struct cfq_group, blkg); return NULL; } |
8aea45451
|
978 979 |
static void cfq_update_blkio_group_weight(void *key, struct blkio_group *blkg, unsigned int weight) |
f8d461d69
|
980 |
{ |
8184f93ec
|
981 982 983 |
struct cfq_group *cfqg = cfqg_of_blkg(blkg); cfqg->new_weight = weight; cfqg->needs_update = true; |
f8d461d69
|
984 |
} |
f469a7b4d
|
985 986 |
static void cfq_init_add_cfqg_lists(struct cfq_data *cfqd, struct cfq_group *cfqg, struct blkio_cgroup *blkcg) |
25fb5169d
|
987 |
{ |
220841906
|
988 989 |
struct backing_dev_info *bdi = &cfqd->queue->backing_dev_info; unsigned int major, minor; |
25fb5169d
|
990 |
|
f469a7b4d
|
991 992 993 994 995 996 997 |
/* * Add group onto cgroup list. It might happen that bdi->dev is * not initialized yet. Initialize this new group without major * and minor info and this info will be filled in once a new thread * comes for IO. */ if (bdi->dev) { |
a74b2adae
|
998 |
sscanf(dev_name(bdi->dev), "%u:%u", &major, &minor); |
f469a7b4d
|
999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 |
cfq_blkiocg_add_blkio_group(blkcg, &cfqg->blkg, (void *)cfqd, MKDEV(major, minor)); } else cfq_blkiocg_add_blkio_group(blkcg, &cfqg->blkg, (void *)cfqd, 0); cfqd->nr_blkcg_linked_grps++; cfqg->weight = blkcg_get_weight(blkcg, cfqg->blkg.dev); /* Add group on cfqd list */ hlist_add_head(&cfqg->cfqd_node, &cfqd->cfqg_list); } /* * Should be called from sleepable context. No request queue lock as per * cpu stats are allocated dynamically and alloc_percpu needs to be called * from sleepable context. */ static struct cfq_group * cfq_alloc_cfqg(struct cfq_data *cfqd) { struct cfq_group *cfqg = NULL; |
5624a4e44
|
1020 |
int i, j, ret; |
f469a7b4d
|
1021 |
struct cfq_rb_root *st; |
25fb5169d
|
1022 1023 1024 |
cfqg = kzalloc_node(sizeof(*cfqg), GFP_ATOMIC, cfqd->queue->node); if (!cfqg) |
f469a7b4d
|
1025 |
return NULL; |
25fb5169d
|
1026 |
|
25fb5169d
|
1027 1028 1029 |
for_each_cfqg_st(cfqg, i, j, st) *st = CFQ_RB_ROOT; RB_CLEAR_NODE(&cfqg->rb_node); |
7700fc4f6
|
1030 |
cfqg->ttime.last_end_request = jiffies; |
b1c357696
|
1031 1032 1033 1034 1035 1036 |
/* * Take the initial reference that will be released on destroy * This can be thought of a joint reference by cgroup and * elevator which will be dropped by either elevator exit * or cgroup deletion path depending on who is exiting first. */ |
329a67815
|
1037 |
cfqg->ref = 1; |
5624a4e44
|
1038 1039 1040 1041 1042 1043 |
ret = blkio_alloc_blkg_stats(&cfqg->blkg); if (ret) { kfree(cfqg); return NULL; } |
f469a7b4d
|
1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 |
return cfqg; } static struct cfq_group * cfq_find_cfqg(struct cfq_data *cfqd, struct blkio_cgroup *blkcg) { struct cfq_group *cfqg = NULL; void *key = cfqd; struct backing_dev_info *bdi = &cfqd->queue->backing_dev_info; unsigned int major, minor; |
b1c357696
|
1054 |
|
180be2a04
|
1055 |
/* |
f469a7b4d
|
1056 1057 |
* This is the common case when there are no blkio cgroups. * Avoid lookup in this case |
180be2a04
|
1058 |
*/ |
f469a7b4d
|
1059 1060 1061 1062 |
if (blkcg == &blkio_root_cgroup) cfqg = &cfqd->root_group; else cfqg = cfqg_of_blkg(blkiocg_lookup_group(blkcg, key)); |
25fb5169d
|
1063 |
|
f469a7b4d
|
1064 1065 1066 1067 |
if (cfqg && !cfqg->blkg.dev && bdi->dev && dev_name(bdi->dev)) { sscanf(dev_name(bdi->dev), "%u:%u", &major, &minor); cfqg->blkg.dev = MKDEV(major, minor); } |
25fb5169d
|
1068 |
|
25fb5169d
|
1069 1070 1071 1072 |
return cfqg; } /* |
3e59cf9d6
|
1073 1074 |
* Search for the cfq group current task belongs to. request_queue lock must * be held. |
25fb5169d
|
1075 |
*/ |
3e59cf9d6
|
1076 |
static struct cfq_group *cfq_get_cfqg(struct cfq_data *cfqd) |
25fb5169d
|
1077 |
{ |
70087dc38
|
1078 |
struct blkio_cgroup *blkcg; |
f469a7b4d
|
1079 1080 |
struct cfq_group *cfqg = NULL, *__cfqg = NULL; struct request_queue *q = cfqd->queue; |
25fb5169d
|
1081 1082 |
rcu_read_lock(); |
70087dc38
|
1083 |
blkcg = task_blkio_cgroup(current); |
f469a7b4d
|
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 |
cfqg = cfq_find_cfqg(cfqd, blkcg); if (cfqg) { rcu_read_unlock(); return cfqg; } /* * Need to allocate a group. Allocation of group also needs allocation * of per cpu stats which in-turn takes a mutex() and can block. Hence * we need to drop rcu lock and queue_lock before we call alloc. * * Not taking any queue reference here and assuming that queue is * around by the time we return. CFQ queue allocation code does * the same. It might be racy though. */ rcu_read_unlock(); spin_unlock_irq(q->queue_lock); cfqg = cfq_alloc_cfqg(cfqd); spin_lock_irq(q->queue_lock); rcu_read_lock(); blkcg = task_blkio_cgroup(current); /* * If some other thread already allocated the group while we were * not holding queue lock, free up the group */ __cfqg = cfq_find_cfqg(cfqd, blkcg); if (__cfqg) { kfree(cfqg); rcu_read_unlock(); return __cfqg; } |
3e59cf9d6
|
1121 |
if (!cfqg) |
25fb5169d
|
1122 |
cfqg = &cfqd->root_group; |
f469a7b4d
|
1123 1124 |
cfq_init_add_cfqg_lists(cfqd, cfqg, blkcg); |
25fb5169d
|
1125 1126 1127 |
rcu_read_unlock(); return cfqg; } |
7f1dc8a2d
|
1128 1129 |
static inline struct cfq_group *cfq_ref_get_cfqg(struct cfq_group *cfqg) { |
329a67815
|
1130 |
cfqg->ref++; |
7f1dc8a2d
|
1131 1132 |
return cfqg; } |
25fb5169d
|
1133 1134 1135 1136 1137 1138 1139 |
static void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) { /* Currently, all async queues are mapped to root group */ if (!cfq_cfqq_sync(cfqq)) cfqg = &cfqq->cfqd->root_group; cfqq->cfqg = cfqg; |
b1c357696
|
1140 |
/* cfqq reference on cfqg */ |
329a67815
|
1141 |
cfqq->cfqg->ref++; |
b1c357696
|
1142 1143 1144 1145 1146 1147 |
} static void cfq_put_cfqg(struct cfq_group *cfqg) { struct cfq_rb_root *st; int i, j; |
329a67815
|
1148 1149 1150 |
BUG_ON(cfqg->ref <= 0); cfqg->ref--; if (cfqg->ref) |
b1c357696
|
1151 1152 |
return; for_each_cfqg_st(cfqg, i, j, st) |
b54ce60eb
|
1153 |
BUG_ON(!RB_EMPTY_ROOT(&st->rb)); |
5624a4e44
|
1154 |
free_percpu(cfqg->blkg.stats_cpu); |
b1c357696
|
1155 1156 1157 1158 1159 1160 1161 1162 1163 |
kfree(cfqg); } static void cfq_destroy_cfqg(struct cfq_data *cfqd, struct cfq_group *cfqg) { /* Something wrong if we are trying to remove same group twice */ BUG_ON(hlist_unhashed(&cfqg->cfqd_node)); hlist_del_init(&cfqg->cfqd_node); |
a5395b83b
|
1164 1165 |
BUG_ON(cfqd->nr_blkcg_linked_grps <= 0); cfqd->nr_blkcg_linked_grps--; |
b1c357696
|
1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 |
/* * Put the reference taken at the time of creation so that when all * queues are gone, group can be destroyed. */ cfq_put_cfqg(cfqg); } static void cfq_release_cfq_groups(struct cfq_data *cfqd) { struct hlist_node *pos, *n; struct cfq_group *cfqg; hlist_for_each_entry_safe(cfqg, pos, n, &cfqd->cfqg_list, cfqd_node) { /* * If cgroup removal path got to blk_group first and removed * it from cgroup list, then it will take care of destroying * cfqg also. */ |
e98ef89b3
|
1184 |
if (!cfq_blkiocg_del_blkio_group(&cfqg->blkg)) |
b1c357696
|
1185 1186 |
cfq_destroy_cfqg(cfqd, cfqg); } |
25fb5169d
|
1187 |
} |
b1c357696
|
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 |
/* * Blk cgroup controller notification saying that blkio_group object is being * delinked as associated cgroup object is going away. That also means that * no new IO will come in this group. So get rid of this group as soon as * any pending IO in the group is finished. * * This function is called under rcu_read_lock(). key is the rcu protected * pointer. That means "key" is a valid cfq_data pointer as long as we are rcu * read lock. * * "key" was fetched from blkio_group under blkio_cgroup->lock. That means * it should not be NULL as even if elevator was exiting, cgroup deltion * path got to it first. */ |
8aea45451
|
1203 |
static void cfq_unlink_blkio_group(void *key, struct blkio_group *blkg) |
b1c357696
|
1204 1205 1206 1207 1208 1209 1210 1211 |
{ unsigned long flags; struct cfq_data *cfqd = key; spin_lock_irqsave(cfqd->queue->queue_lock, flags); cfq_destroy_cfqg(cfqd, cfqg_of_blkg(blkg)); spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); } |
25fb5169d
|
1212 |
#else /* GROUP_IOSCHED */ |
3e59cf9d6
|
1213 |
static struct cfq_group *cfq_get_cfqg(struct cfq_data *cfqd) |
25fb5169d
|
1214 1215 1216 |
{ return &cfqd->root_group; } |
7f1dc8a2d
|
1217 1218 1219 |
static inline struct cfq_group *cfq_ref_get_cfqg(struct cfq_group *cfqg) { |
50eaeb323
|
1220 |
return cfqg; |
7f1dc8a2d
|
1221 |
} |
25fb5169d
|
1222 1223 1224 1225 |
static inline void cfq_link_cfqq_cfqg(struct cfq_queue *cfqq, struct cfq_group *cfqg) { cfqq->cfqg = cfqg; } |
b1c357696
|
1226 1227 |
static void cfq_release_cfq_groups(struct cfq_data *cfqd) {} static inline void cfq_put_cfqg(struct cfq_group *cfqg) {} |
25fb5169d
|
1228 |
#endif /* GROUP_IOSCHED */ |
498d3aa2b
|
1229 |
/* |
c0324a020
|
1230 |
* The cfqd->service_trees holds all pending cfq_queue's that have |
498d3aa2b
|
1231 1232 1233 |
* requests waiting to be processed. It is sorted in the order that * we will service the queues. */ |
a36e71f99
|
1234 |
static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq, |
a6151c3a5
|
1235 |
bool add_front) |
d9e7620e6
|
1236 |
{ |
0871714e0
|
1237 1238 |
struct rb_node **p, *parent; struct cfq_queue *__cfqq; |
d9e7620e6
|
1239 |
unsigned long rb_key; |
c0324a020
|
1240 |
struct cfq_rb_root *service_tree; |
498d3aa2b
|
1241 |
int left; |
dae739ebc
|
1242 |
int new_cfqq = 1; |
ae30c2865
|
1243 |
|
cdb16e8f7
|
1244 |
service_tree = service_tree_for(cfqq->cfqg, cfqq_prio(cfqq), |
65b32a573
|
1245 |
cfqq_type(cfqq)); |
0871714e0
|
1246 1247 |
if (cfq_class_idle(cfqq)) { rb_key = CFQ_IDLE_DELAY; |
aa6f6a3de
|
1248 |
parent = rb_last(&service_tree->rb); |
0871714e0
|
1249 1250 1251 1252 1253 1254 |
if (parent && parent != &cfqq->rb_node) { __cfqq = rb_entry(parent, struct cfq_queue, rb_node); rb_key += __cfqq->rb_key; } else rb_key += jiffies; } else if (!add_front) { |
b9c8946b1
|
1255 1256 1257 1258 1259 1260 |
/* * Get our rb key offset. Subtract any residual slice * value carried from last service. A negative resid * count indicates slice overrun, and this should position * the next service time further away in the tree. */ |
edd75ffd9
|
1261 |
rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies; |
b9c8946b1
|
1262 |
rb_key -= cfqq->slice_resid; |
edd75ffd9
|
1263 |
cfqq->slice_resid = 0; |
48e025e63
|
1264 1265 |
} else { rb_key = -HZ; |
aa6f6a3de
|
1266 |
__cfqq = cfq_rb_first(service_tree); |
48e025e63
|
1267 1268 |
rb_key += __cfqq ? __cfqq->rb_key : jiffies; } |
1da177e4c
|
1269 |
|
d9e7620e6
|
1270 |
if (!RB_EMPTY_NODE(&cfqq->rb_node)) { |
dae739ebc
|
1271 |
new_cfqq = 0; |
99f9628ab
|
1272 |
/* |
d9e7620e6
|
1273 |
* same position, nothing more to do |
99f9628ab
|
1274 |
*/ |
c0324a020
|
1275 1276 |
if (rb_key == cfqq->rb_key && cfqq->service_tree == service_tree) |
d9e7620e6
|
1277 |
return; |
1da177e4c
|
1278 |
|
aa6f6a3de
|
1279 1280 |
cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree); cfqq->service_tree = NULL; |
1da177e4c
|
1281 |
} |
d9e7620e6
|
1282 |
|
498d3aa2b
|
1283 |
left = 1; |
0871714e0
|
1284 |
parent = NULL; |
aa6f6a3de
|
1285 1286 |
cfqq->service_tree = service_tree; p = &service_tree->rb.rb_node; |
d9e7620e6
|
1287 |
while (*p) { |
67060e379
|
1288 |
struct rb_node **n; |
cc09e2990
|
1289 |
|
d9e7620e6
|
1290 1291 |
parent = *p; __cfqq = rb_entry(parent, struct cfq_queue, rb_node); |
0c534e0a4
|
1292 |
/* |
c0324a020
|
1293 |
* sort by key, that represents service time. |
0c534e0a4
|
1294 |
*/ |
c0324a020
|
1295 |
if (time_before(rb_key, __cfqq->rb_key)) |
67060e379
|
1296 |
n = &(*p)->rb_left; |
c0324a020
|
1297 |
else { |
67060e379
|
1298 |
n = &(*p)->rb_right; |
cc09e2990
|
1299 |
left = 0; |
c0324a020
|
1300 |
} |
67060e379
|
1301 1302 |
p = n; |
d9e7620e6
|
1303 |
} |
cc09e2990
|
1304 |
if (left) |
aa6f6a3de
|
1305 |
service_tree->left = &cfqq->rb_node; |
cc09e2990
|
1306 |
|
d9e7620e6
|
1307 1308 |
cfqq->rb_key = rb_key; rb_link_node(&cfqq->rb_node, parent, p); |
aa6f6a3de
|
1309 1310 |
rb_insert_color(&cfqq->rb_node, &service_tree->rb); service_tree->count++; |
20359f27e
|
1311 |
if (add_front || !new_cfqq) |
dae739ebc
|
1312 |
return; |
8184f93ec
|
1313 |
cfq_group_notify_queue_add(cfqd, cfqq->cfqg); |
1da177e4c
|
1314 |
} |
a36e71f99
|
1315 |
static struct cfq_queue * |
f2d1f0ae7
|
1316 1317 1318 |
cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root, sector_t sector, struct rb_node **ret_parent, struct rb_node ***rb_link) |
a36e71f99
|
1319 |
{ |
a36e71f99
|
1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 |
struct rb_node **p, *parent; struct cfq_queue *cfqq = NULL; parent = NULL; p = &root->rb_node; while (*p) { struct rb_node **n; parent = *p; cfqq = rb_entry(parent, struct cfq_queue, p_node); /* * Sort strictly based on sector. Smallest to the left, * largest to the right. */ |
2e46e8b27
|
1335 |
if (sector > blk_rq_pos(cfqq->next_rq)) |
a36e71f99
|
1336 |
n = &(*p)->rb_right; |
2e46e8b27
|
1337 |
else if (sector < blk_rq_pos(cfqq->next_rq)) |
a36e71f99
|
1338 1339 1340 1341 |
n = &(*p)->rb_left; else break; p = n; |
3ac6c9f8a
|
1342 |
cfqq = NULL; |
a36e71f99
|
1343 1344 1345 1346 1347 |
} *ret_parent = parent; if (rb_link) *rb_link = p; |
3ac6c9f8a
|
1348 |
return cfqq; |
a36e71f99
|
1349 1350 1351 1352 |
} static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq) { |
a36e71f99
|
1353 1354 |
struct rb_node **p, *parent; struct cfq_queue *__cfqq; |
f2d1f0ae7
|
1355 1356 1357 1358 |
if (cfqq->p_root) { rb_erase(&cfqq->p_node, cfqq->p_root); cfqq->p_root = NULL; } |
a36e71f99
|
1359 1360 1361 1362 1363 |
if (cfq_class_idle(cfqq)) return; if (!cfqq->next_rq) return; |
f2d1f0ae7
|
1364 |
cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio]; |
2e46e8b27
|
1365 1366 |
__cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root, blk_rq_pos(cfqq->next_rq), &parent, &p); |
3ac6c9f8a
|
1367 1368 |
if (!__cfqq) { rb_link_node(&cfqq->p_node, parent, p); |
f2d1f0ae7
|
1369 1370 1371 |
rb_insert_color(&cfqq->p_node, cfqq->p_root); } else cfqq->p_root = NULL; |
a36e71f99
|
1372 |
} |
498d3aa2b
|
1373 1374 1375 |
/* * Update cfqq's position in the service tree. */ |
edd75ffd9
|
1376 |
static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
6d048f531
|
1377 |
{ |
6d048f531
|
1378 1379 1380 |
/* * Resorting requires the cfqq to be on the RR list already. */ |
a36e71f99
|
1381 |
if (cfq_cfqq_on_rr(cfqq)) { |
edd75ffd9
|
1382 |
cfq_service_tree_add(cfqd, cfqq, 0); |
a36e71f99
|
1383 1384 |
cfq_prio_tree_add(cfqd, cfqq); } |
6d048f531
|
1385 |
} |
1da177e4c
|
1386 1387 |
/* * add to busy list of queues for service, trying to be fair in ordering |
22e2c507c
|
1388 |
* the pending list according to last request service |
1da177e4c
|
1389 |
*/ |
febffd618
|
1390 |
static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
1da177e4c
|
1391 |
{ |
7b679138b
|
1392 |
cfq_log_cfqq(cfqd, cfqq, "add_to_rr"); |
3b18152c3
|
1393 1394 |
BUG_ON(cfq_cfqq_on_rr(cfqq)); cfq_mark_cfqq_on_rr(cfqq); |
1da177e4c
|
1395 |
cfqd->busy_queues++; |
ef8a41df8
|
1396 1397 |
if (cfq_cfqq_sync(cfqq)) cfqd->busy_sync_queues++; |
1da177e4c
|
1398 |
|
edd75ffd9
|
1399 |
cfq_resort_rr_list(cfqd, cfqq); |
1da177e4c
|
1400 |
} |
498d3aa2b
|
1401 1402 1403 1404 |
/* * Called when the cfqq no longer has requests pending, remove it from * the service tree. */ |
febffd618
|
1405 |
static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
1da177e4c
|
1406 |
{ |
7b679138b
|
1407 |
cfq_log_cfqq(cfqd, cfqq, "del_from_rr"); |
3b18152c3
|
1408 1409 |
BUG_ON(!cfq_cfqq_on_rr(cfqq)); cfq_clear_cfqq_on_rr(cfqq); |
1da177e4c
|
1410 |
|
aa6f6a3de
|
1411 1412 1413 1414 |
if (!RB_EMPTY_NODE(&cfqq->rb_node)) { cfq_rb_erase(&cfqq->rb_node, cfqq->service_tree); cfqq->service_tree = NULL; } |
f2d1f0ae7
|
1415 1416 1417 1418 |
if (cfqq->p_root) { rb_erase(&cfqq->p_node, cfqq->p_root); cfqq->p_root = NULL; } |
d9e7620e6
|
1419 |
|
8184f93ec
|
1420 |
cfq_group_notify_queue_del(cfqd, cfqq->cfqg); |
1da177e4c
|
1421 1422 |
BUG_ON(!cfqd->busy_queues); cfqd->busy_queues--; |
ef8a41df8
|
1423 1424 |
if (cfq_cfqq_sync(cfqq)) cfqd->busy_sync_queues--; |
1da177e4c
|
1425 1426 1427 1428 1429 |
} /* * rb tree support functions */ |
febffd618
|
1430 |
static void cfq_del_rq_rb(struct request *rq) |
1da177e4c
|
1431 |
{ |
5e7053747
|
1432 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
5e7053747
|
1433 |
const int sync = rq_is_sync(rq); |
1da177e4c
|
1434 |
|
b4878f245
|
1435 1436 |
BUG_ON(!cfqq->queued[sync]); cfqq->queued[sync]--; |
1da177e4c
|
1437 |
|
5e7053747
|
1438 |
elv_rb_del(&cfqq->sort_list, rq); |
1da177e4c
|
1439 |
|
f04a64246
|
1440 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 |
if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) { /* * Queue will be deleted from service tree when we actually * expire it later. Right now just remove it from prio tree * as it is empty. */ if (cfqq->p_root) { rb_erase(&cfqq->p_node, cfqq->p_root); cfqq->p_root = NULL; } } |
1da177e4c
|
1451 |
} |
5e7053747
|
1452 |
static void cfq_add_rq_rb(struct request *rq) |
1da177e4c
|
1453 |
{ |
5e7053747
|
1454 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
1da177e4c
|
1455 |
struct cfq_data *cfqd = cfqq->cfqd; |
796d5116c
|
1456 |
struct request *prev; |
1da177e4c
|
1457 |
|
5380a101d
|
1458 |
cfqq->queued[rq_is_sync(rq)]++; |
1da177e4c
|
1459 |
|
796d5116c
|
1460 |
elv_rb_add(&cfqq->sort_list, rq); |
5fccbf61b
|
1461 1462 1463 |
if (!cfq_cfqq_on_rr(cfqq)) cfq_add_cfqq_rr(cfqd, cfqq); |
5044eed48
|
1464 1465 1466 1467 |
/* * check if this request is a better next-serve candidate */ |
a36e71f99
|
1468 |
prev = cfqq->next_rq; |
cf7c25cf9
|
1469 |
cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq, cfqd->last_position); |
a36e71f99
|
1470 1471 1472 1473 1474 1475 |
/* * adjust priority tree position, if ->next_rq changes */ if (prev != cfqq->next_rq) cfq_prio_tree_add(cfqd, cfqq); |
5044eed48
|
1476 |
BUG_ON(!cfqq->next_rq); |
1da177e4c
|
1477 |
} |
febffd618
|
1478 |
static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq) |
1da177e4c
|
1479 |
{ |
5380a101d
|
1480 1481 |
elv_rb_del(&cfqq->sort_list, rq); cfqq->queued[rq_is_sync(rq)]--; |
e98ef89b3
|
1482 1483 |
cfq_blkiocg_update_io_remove_stats(&(RQ_CFQG(rq))->blkg, rq_data_dir(rq), rq_is_sync(rq)); |
5e7053747
|
1484 |
cfq_add_rq_rb(rq); |
e98ef89b3
|
1485 |
cfq_blkiocg_update_io_add_stats(&(RQ_CFQG(rq))->blkg, |
7f1dc8a2d
|
1486 1487 |
&cfqq->cfqd->serving_group->blkg, rq_data_dir(rq), rq_is_sync(rq)); |
1da177e4c
|
1488 |
} |
206dc69b3
|
1489 1490 |
static struct request * cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio) |
1da177e4c
|
1491 |
{ |
206dc69b3
|
1492 |
struct task_struct *tsk = current; |
91fac317a
|
1493 |
struct cfq_io_context *cic; |
206dc69b3
|
1494 |
struct cfq_queue *cfqq; |
1da177e4c
|
1495 |
|
4ac845a2e
|
1496 |
cic = cfq_cic_lookup(cfqd, tsk->io_context); |
91fac317a
|
1497 1498 1499 1500 |
if (!cic) return NULL; cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio)); |
89850f7ee
|
1501 1502 |
if (cfqq) { sector_t sector = bio->bi_sector + bio_sectors(bio); |
21183b07e
|
1503 |
return elv_rb_find(&cfqq->sort_list, sector); |
89850f7ee
|
1504 |
} |
1da177e4c
|
1505 |
|
1da177e4c
|
1506 1507 |
return NULL; } |
165125e1e
|
1508 |
static void cfq_activate_request(struct request_queue *q, struct request *rq) |
1da177e4c
|
1509 |
{ |
22e2c507c
|
1510 |
struct cfq_data *cfqd = q->elevator->elevator_data; |
3b18152c3
|
1511 |
|
53c583d22
|
1512 |
cfqd->rq_in_driver++; |
7b679138b
|
1513 |
cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d", |
53c583d22
|
1514 |
cfqd->rq_in_driver); |
25776e359
|
1515 |
|
5b93629b4
|
1516 |
cfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq); |
1da177e4c
|
1517 |
} |
165125e1e
|
1518 |
static void cfq_deactivate_request(struct request_queue *q, struct request *rq) |
1da177e4c
|
1519 |
{ |
b4878f245
|
1520 |
struct cfq_data *cfqd = q->elevator->elevator_data; |
53c583d22
|
1521 1522 |
WARN_ON(!cfqd->rq_in_driver); cfqd->rq_in_driver--; |
7b679138b
|
1523 |
cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d", |
53c583d22
|
1524 |
cfqd->rq_in_driver); |
1da177e4c
|
1525 |
} |
b4878f245
|
1526 |
static void cfq_remove_request(struct request *rq) |
1da177e4c
|
1527 |
{ |
5e7053747
|
1528 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
21183b07e
|
1529 |
|
5e7053747
|
1530 1531 |
if (cfqq->next_rq == rq) cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq); |
1da177e4c
|
1532 |
|
b4878f245
|
1533 |
list_del_init(&rq->queuelist); |
5e7053747
|
1534 |
cfq_del_rq_rb(rq); |
374f84ac3
|
1535 |
|
45333d5a3
|
1536 |
cfqq->cfqd->rq_queued--; |
e98ef89b3
|
1537 1538 |
cfq_blkiocg_update_io_remove_stats(&(RQ_CFQG(rq))->blkg, rq_data_dir(rq), rq_is_sync(rq)); |
65299a3b7
|
1539 1540 1541 |
if (rq->cmd_flags & REQ_PRIO) { WARN_ON(!cfqq->prio_pending); cfqq->prio_pending--; |
b53d1ed73
|
1542 |
} |
1da177e4c
|
1543 |
} |
165125e1e
|
1544 1545 |
static int cfq_merge(struct request_queue *q, struct request **req, struct bio *bio) |
1da177e4c
|
1546 1547 1548 |
{ struct cfq_data *cfqd = q->elevator->elevator_data; struct request *__rq; |
1da177e4c
|
1549 |
|
206dc69b3
|
1550 |
__rq = cfq_find_rq_fmerge(cfqd, bio); |
22e2c507c
|
1551 |
if (__rq && elv_rq_merge_ok(__rq, bio)) { |
9817064b6
|
1552 1553 |
*req = __rq; return ELEVATOR_FRONT_MERGE; |
1da177e4c
|
1554 1555 1556 |
} return ELEVATOR_NO_MERGE; |
1da177e4c
|
1557 |
} |
165125e1e
|
1558 |
static void cfq_merged_request(struct request_queue *q, struct request *req, |
21183b07e
|
1559 |
int type) |
1da177e4c
|
1560 |
{ |
21183b07e
|
1561 |
if (type == ELEVATOR_FRONT_MERGE) { |
5e7053747
|
1562 |
struct cfq_queue *cfqq = RQ_CFQQ(req); |
1da177e4c
|
1563 |
|
5e7053747
|
1564 |
cfq_reposition_rq_rb(cfqq, req); |
1da177e4c
|
1565 |
} |
1da177e4c
|
1566 |
} |
812d40264
|
1567 1568 1569 |
static void cfq_bio_merged(struct request_queue *q, struct request *req, struct bio *bio) { |
e98ef89b3
|
1570 1571 |
cfq_blkiocg_update_io_merged_stats(&(RQ_CFQG(req))->blkg, bio_data_dir(bio), cfq_bio_sync(bio)); |
812d40264
|
1572 |
} |
1da177e4c
|
1573 |
static void |
165125e1e
|
1574 |
cfq_merged_requests(struct request_queue *q, struct request *rq, |
1da177e4c
|
1575 1576 |
struct request *next) { |
cf7c25cf9
|
1577 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
6ae0516b8
|
1578 |
struct cfq_data *cfqd = q->elevator->elevator_data; |
22e2c507c
|
1579 1580 1581 1582 |
/* * reposition in fifo if next is older than rq */ if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) && |
30996f40b
|
1583 |
time_before(rq_fifo_time(next), rq_fifo_time(rq))) { |
22e2c507c
|
1584 |
list_move(&rq->queuelist, &next->queuelist); |
30996f40b
|
1585 1586 |
rq_set_fifo_time(rq, rq_fifo_time(next)); } |
22e2c507c
|
1587 |
|
cf7c25cf9
|
1588 1589 |
if (cfqq->next_rq == next) cfqq->next_rq = rq; |
b4878f245
|
1590 |
cfq_remove_request(next); |
e98ef89b3
|
1591 1592 |
cfq_blkiocg_update_io_merged_stats(&(RQ_CFQG(rq))->blkg, rq_data_dir(next), rq_is_sync(next)); |
6ae0516b8
|
1593 1594 1595 1596 1597 1598 1599 1600 1601 1602 |
cfqq = RQ_CFQQ(next); /* * all requests of this queue are merged to other queues, delete it * from the service tree. If it's the active_queue, * cfq_dispatch_requests() will choose to expire it or do idle */ if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list) && cfqq != cfqd->active_queue) cfq_del_cfqq_rr(cfqd, cfqq); |
22e2c507c
|
1603 |
} |
165125e1e
|
1604 |
static int cfq_allow_merge(struct request_queue *q, struct request *rq, |
da7752650
|
1605 1606 1607 |
struct bio *bio) { struct cfq_data *cfqd = q->elevator->elevator_data; |
91fac317a
|
1608 |
struct cfq_io_context *cic; |
da7752650
|
1609 |
struct cfq_queue *cfqq; |
da7752650
|
1610 1611 |
/* |
ec8acb690
|
1612 |
* Disallow merge of a sync bio into an async request. |
da7752650
|
1613 |
*/ |
91fac317a
|
1614 |
if (cfq_bio_sync(bio) && !rq_is_sync(rq)) |
a6151c3a5
|
1615 |
return false; |
da7752650
|
1616 1617 |
/* |
719d34027
|
1618 1619 |
* Lookup the cfqq that this bio will be queued with. Allow * merge only if rq is queued there. |
da7752650
|
1620 |
*/ |
4ac845a2e
|
1621 |
cic = cfq_cic_lookup(cfqd, current->io_context); |
91fac317a
|
1622 |
if (!cic) |
a6151c3a5
|
1623 |
return false; |
719d34027
|
1624 |
|
91fac317a
|
1625 |
cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio)); |
a6151c3a5
|
1626 |
return cfqq == RQ_CFQQ(rq); |
da7752650
|
1627 |
} |
812df48d1
|
1628 1629 1630 |
static inline void cfq_del_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq) { del_timer(&cfqd->idle_slice_timer); |
e98ef89b3
|
1631 |
cfq_blkiocg_update_idle_time_stats(&cfqq->cfqg->blkg); |
812df48d1
|
1632 |
} |
febffd618
|
1633 1634 |
static void __cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
22e2c507c
|
1635 1636 |
{ if (cfqq) { |
b1ffe737f
|
1637 1638 |
cfq_log_cfqq(cfqd, cfqq, "set_active wl_prio:%d wl_type:%d", cfqd->serving_prio, cfqd->serving_type); |
62a37f6ba
|
1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651 1652 1653 |
cfq_blkiocg_update_avg_queue_size_stats(&cfqq->cfqg->blkg); cfqq->slice_start = 0; cfqq->dispatch_start = jiffies; cfqq->allocated_slice = 0; cfqq->slice_end = 0; cfqq->slice_dispatch = 0; cfqq->nr_sectors = 0; cfq_clear_cfqq_wait_request(cfqq); cfq_clear_cfqq_must_dispatch(cfqq); cfq_clear_cfqq_must_alloc_slice(cfqq); cfq_clear_cfqq_fifo_expire(cfqq); cfq_mark_cfqq_slice_new(cfqq); cfq_del_timer(cfqd, cfqq); |
22e2c507c
|
1654 1655 1656 1657 1658 1659 |
} cfqd->active_queue = cfqq; } /* |
7b14e3b52
|
1660 1661 1662 1663 |
* current cfqq expired its slice (or was too idle), select new one */ static void __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq, |
e5ff082e8
|
1664 |
bool timed_out) |
7b14e3b52
|
1665 |
{ |
7b679138b
|
1666 |
cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out); |
7b14e3b52
|
1667 |
if (cfq_cfqq_wait_request(cfqq)) |
812df48d1
|
1668 |
cfq_del_timer(cfqd, cfqq); |
7b14e3b52
|
1669 |
|
7b14e3b52
|
1670 |
cfq_clear_cfqq_wait_request(cfqq); |
f75edf2dc
|
1671 |
cfq_clear_cfqq_wait_busy(cfqq); |
7b14e3b52
|
1672 1673 |
/* |
ae54abed6
|
1674 1675 1676 1677 1678 1679 1680 1681 1682 |
* If this cfqq is shared between multiple processes, check to * make sure that those processes are still issuing I/Os within * the mean seek distance. If not, it may be time to break the * queues apart again. */ if (cfq_cfqq_coop(cfqq) && CFQQ_SEEKY(cfqq)) cfq_mark_cfqq_split_coop(cfqq); /* |
6084cdda0
|
1683 |
* store what was left of this slice, if the queue idled/timed out |
7b14e3b52
|
1684 |
*/ |
c553f8e33
|
1685 1686 |
if (timed_out) { if (cfq_cfqq_slice_new(cfqq)) |
ba5bd520f
|
1687 |
cfqq->slice_resid = cfq_scaled_cfqq_slice(cfqd, cfqq); |
c553f8e33
|
1688 1689 |
else cfqq->slice_resid = cfqq->slice_end - jiffies; |
7b679138b
|
1690 1691 |
cfq_log_cfqq(cfqd, cfqq, "resid=%ld", cfqq->slice_resid); } |
7b14e3b52
|
1692 |
|
e5ff082e8
|
1693 |
cfq_group_served(cfqd, cfqq->cfqg, cfqq); |
dae739ebc
|
1694 |
|
f04a64246
|
1695 1696 |
if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) cfq_del_cfqq_rr(cfqd, cfqq); |
edd75ffd9
|
1697 |
cfq_resort_rr_list(cfqd, cfqq); |
7b14e3b52
|
1698 1699 1700 1701 1702 1703 1704 1705 |
if (cfqq == cfqd->active_queue) cfqd->active_queue = NULL; if (cfqd->active_cic) { put_io_context(cfqd->active_cic->ioc); cfqd->active_cic = NULL; } |
7b14e3b52
|
1706 |
} |
e5ff082e8
|
1707 |
static inline void cfq_slice_expired(struct cfq_data *cfqd, bool timed_out) |
7b14e3b52
|
1708 1709 1710 1711 |
{ struct cfq_queue *cfqq = cfqd->active_queue; if (cfqq) |
e5ff082e8
|
1712 |
__cfq_slice_expired(cfqd, cfqq, timed_out); |
7b14e3b52
|
1713 |
} |
498d3aa2b
|
1714 1715 1716 1717 |
/* * Get next queue for service. Unless we have a queue preemption, * we'll simply select the first cfqq in the service tree. */ |
6d048f531
|
1718 |
static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd) |
22e2c507c
|
1719 |
{ |
c0324a020
|
1720 |
struct cfq_rb_root *service_tree = |
cdb16e8f7
|
1721 |
service_tree_for(cfqd->serving_group, cfqd->serving_prio, |
65b32a573
|
1722 |
cfqd->serving_type); |
d9e7620e6
|
1723 |
|
f04a64246
|
1724 1725 |
if (!cfqd->rq_queued) return NULL; |
1fa8f6d68
|
1726 1727 1728 |
/* There is nothing to dispatch */ if (!service_tree) return NULL; |
c0324a020
|
1729 1730 1731 |
if (RB_EMPTY_ROOT(&service_tree->rb)) return NULL; return cfq_rb_first(service_tree); |
6d048f531
|
1732 |
} |
f04a64246
|
1733 1734 |
static struct cfq_queue *cfq_get_next_queue_forced(struct cfq_data *cfqd) { |
25fb5169d
|
1735 |
struct cfq_group *cfqg; |
f04a64246
|
1736 1737 1738 1739 1740 1741 |
struct cfq_queue *cfqq; int i, j; struct cfq_rb_root *st; if (!cfqd->rq_queued) return NULL; |
25fb5169d
|
1742 1743 1744 |
cfqg = cfq_get_next_cfqg(cfqd); if (!cfqg) return NULL; |
f04a64246
|
1745 1746 1747 1748 1749 |
for_each_cfqg_st(cfqg, i, j, st) if ((cfqq = cfq_rb_first(st)) != NULL) return cfqq; return NULL; } |
498d3aa2b
|
1750 1751 1752 |
/* * Get and set a new active queue for service. */ |
a36e71f99
|
1753 1754 |
static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
6d048f531
|
1755 |
{ |
e00ef7997
|
1756 |
if (!cfqq) |
a36e71f99
|
1757 |
cfqq = cfq_get_next_queue(cfqd); |
6d048f531
|
1758 |
|
22e2c507c
|
1759 |
__cfq_set_active_queue(cfqd, cfqq); |
3b18152c3
|
1760 |
return cfqq; |
22e2c507c
|
1761 |
} |
d9e7620e6
|
1762 1763 1764 |
static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd, struct request *rq) { |
83096ebf1
|
1765 1766 |
if (blk_rq_pos(rq) >= cfqd->last_position) return blk_rq_pos(rq) - cfqd->last_position; |
d9e7620e6
|
1767 |
else |
83096ebf1
|
1768 |
return cfqd->last_position - blk_rq_pos(rq); |
d9e7620e6
|
1769 |
} |
b2c18e1e0
|
1770 |
static inline int cfq_rq_close(struct cfq_data *cfqd, struct cfq_queue *cfqq, |
e9ce335df
|
1771 |
struct request *rq) |
6d048f531
|
1772 |
{ |
e9ce335df
|
1773 |
return cfq_dist_from_last(cfqd, rq) <= CFQQ_CLOSE_THR; |
6d048f531
|
1774 |
} |
a36e71f99
|
1775 1776 1777 |
static struct cfq_queue *cfqq_close(struct cfq_data *cfqd, struct cfq_queue *cur_cfqq) { |
f2d1f0ae7
|
1778 |
struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio]; |
a36e71f99
|
1779 1780 1781 1782 1783 1784 1785 1786 1787 1788 1789 |
struct rb_node *parent, *node; struct cfq_queue *__cfqq; sector_t sector = cfqd->last_position; if (RB_EMPTY_ROOT(root)) return NULL; /* * First, if we find a request starting at the end of the last * request, choose it. */ |
f2d1f0ae7
|
1790 |
__cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL); |
a36e71f99
|
1791 1792 1793 1794 1795 1796 1797 1798 |
if (__cfqq) return __cfqq; /* * If the exact sector wasn't found, the parent of the NULL leaf * will contain the closest sector. */ __cfqq = rb_entry(parent, struct cfq_queue, p_node); |
e9ce335df
|
1799 |
if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq)) |
a36e71f99
|
1800 |
return __cfqq; |
2e46e8b27
|
1801 |
if (blk_rq_pos(__cfqq->next_rq) < sector) |
a36e71f99
|
1802 1803 1804 1805 1806 1807 1808 |
node = rb_next(&__cfqq->p_node); else node = rb_prev(&__cfqq->p_node); if (!node) return NULL; __cfqq = rb_entry(node, struct cfq_queue, p_node); |
e9ce335df
|
1809 |
if (cfq_rq_close(cfqd, cur_cfqq, __cfqq->next_rq)) |
a36e71f99
|
1810 1811 1812 1813 1814 1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 |
return __cfqq; return NULL; } /* * cfqd - obvious * cur_cfqq - passed in so that we don't decide that the current queue is * closely cooperating with itself. * * So, basically we're assuming that that cur_cfqq has dispatched at least * one request, and that cfqd->last_position reflects a position on the disk * associated with the I/O issued by cur_cfqq. I'm not sure this is a valid * assumption. */ static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd, |
b3b6d0408
|
1826 |
struct cfq_queue *cur_cfqq) |
6d048f531
|
1827 |
{ |
a36e71f99
|
1828 |
struct cfq_queue *cfqq; |
39c01b219
|
1829 1830 |
if (cfq_class_idle(cur_cfqq)) return NULL; |
e6c5bc737
|
1831 1832 1833 1834 |
if (!cfq_cfqq_sync(cur_cfqq)) return NULL; if (CFQQ_SEEKY(cur_cfqq)) return NULL; |
a36e71f99
|
1835 |
/* |
b9d8f4c73
|
1836 1837 1838 1839 1840 1841 |
* Don't search priority tree if it's the only queue in the group. */ if (cur_cfqq->cfqg->nr_cfqq == 1) return NULL; /* |
d9e7620e6
|
1842 1843 1844 |
* We should notice if some of the queues are cooperating, eg * working closely on the same area of the disk. In that case, * we can group them together and don't waste time idling. |
6d048f531
|
1845 |
*/ |
a36e71f99
|
1846 1847 1848 |
cfqq = cfqq_close(cfqd, cur_cfqq); if (!cfqq) return NULL; |
8682e1f15
|
1849 1850 1851 |
/* If new queue belongs to different cfq_group, don't choose it */ if (cur_cfqq->cfqg != cfqq->cfqg) return NULL; |
df5fe3e8e
|
1852 1853 1854 1855 1856 |
/* * It only makes sense to merge sync queues. */ if (!cfq_cfqq_sync(cfqq)) return NULL; |
e6c5bc737
|
1857 1858 |
if (CFQQ_SEEKY(cfqq)) return NULL; |
df5fe3e8e
|
1859 |
|
c0324a020
|
1860 1861 1862 1863 1864 |
/* * Do not merge queues of different priority classes */ if (cfq_class_rt(cfqq) != cfq_class_rt(cur_cfqq)) return NULL; |
a36e71f99
|
1865 |
return cfqq; |
6d048f531
|
1866 |
} |
a6d44e982
|
1867 1868 1869 1870 1871 1872 1873 |
/* * Determine whether we should enforce idle window for this queue. */ static bool cfq_should_idle(struct cfq_data *cfqd, struct cfq_queue *cfqq) { enum wl_prio_t prio = cfqq_prio(cfqq); |
718eee057
|
1874 |
struct cfq_rb_root *service_tree = cfqq->service_tree; |
a6d44e982
|
1875 |
|
f04a64246
|
1876 1877 |
BUG_ON(!service_tree); BUG_ON(!service_tree->count); |
b6508c161
|
1878 1879 |
if (!cfqd->cfq_slice_idle) return false; |
a6d44e982
|
1880 1881 1882 1883 1884 |
/* We never do for idle class queues. */ if (prio == IDLE_WORKLOAD) return false; /* We do for queues that were marked with idle window flag. */ |
3c764b7a6
|
1885 1886 |
if (cfq_cfqq_idle_window(cfqq) && !(blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag)) |
a6d44e982
|
1887 1888 1889 1890 1891 1892 |
return true; /* * Otherwise, we do only if they are the last ones * in their service tree. */ |
f5f2b6ceb
|
1893 1894 |
if (service_tree->count == 1 && cfq_cfqq_sync(cfqq) && !cfq_io_thinktime_big(cfqd, &service_tree->ttime, false)) |
c1e44756f
|
1895 |
return true; |
b1ffe737f
|
1896 1897 |
cfq_log_cfqq(cfqd, cfqq, "Not idling. st->count:%d", service_tree->count); |
c1e44756f
|
1898 |
return false; |
a6d44e982
|
1899 |
} |
6d048f531
|
1900 |
static void cfq_arm_slice_timer(struct cfq_data *cfqd) |
22e2c507c
|
1901 |
{ |
1792669cc
|
1902 |
struct cfq_queue *cfqq = cfqd->active_queue; |
206dc69b3
|
1903 |
struct cfq_io_context *cic; |
80bdf0c78
|
1904 |
unsigned long sl, group_idle = 0; |
7b14e3b52
|
1905 |
|
a68bbddba
|
1906 |
/* |
f7d7b7a7a
|
1907 1908 1909 |
* SSD device without seek penalty, disable idling. But only do so * for devices that support queuing, otherwise we still have a problem * with sync vs async workloads. |
a68bbddba
|
1910 |
*/ |
f7d7b7a7a
|
1911 |
if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag) |
a68bbddba
|
1912 |
return; |
dd67d0515
|
1913 |
WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list)); |
6d048f531
|
1914 |
WARN_ON(cfq_cfqq_slice_new(cfqq)); |
22e2c507c
|
1915 1916 1917 1918 |
/* * idle is disabled, either manually or by past process history */ |
80bdf0c78
|
1919 1920 1921 1922 1923 1924 1925 |
if (!cfq_should_idle(cfqd, cfqq)) { /* no queue idling. Check for group idling */ if (cfqd->cfq_group_idle) group_idle = cfqd->cfq_group_idle; else return; } |
6d048f531
|
1926 |
|
22e2c507c
|
1927 |
/* |
8e550632c
|
1928 |
* still active requests from this queue, don't idle |
7b679138b
|
1929 |
*/ |
8e550632c
|
1930 |
if (cfqq->dispatched) |
7b679138b
|
1931 1932 1933 |
return; /* |
22e2c507c
|
1934 1935 |
* task has exited, don't wait */ |
206dc69b3
|
1936 |
cic = cfqd->active_cic; |
66dac98ed
|
1937 |
if (!cic || !atomic_read(&cic->ioc->nr_tasks)) |
6d048f531
|
1938 |
return; |
355b659c8
|
1939 1940 1941 1942 1943 |
/* * If our average think time is larger than the remaining time * slice, then don't idle. This avoids overrunning the allotted * time slice. */ |
383cd7213
|
1944 1945 |
if (sample_valid(cic->ttime.ttime_samples) && (cfqq->slice_end - jiffies < cic->ttime.ttime_mean)) { |
fd16d2631
|
1946 |
cfq_log_cfqq(cfqd, cfqq, "Not idling. think_time:%lu", |
383cd7213
|
1947 |
cic->ttime.ttime_mean); |
355b659c8
|
1948 |
return; |
b1ffe737f
|
1949 |
} |
355b659c8
|
1950 |
|
80bdf0c78
|
1951 1952 1953 |
/* There are other queues in the group, don't do group idle */ if (group_idle && cfqq->cfqg->nr_cfqq > 1) return; |
3b18152c3
|
1954 |
cfq_mark_cfqq_wait_request(cfqq); |
22e2c507c
|
1955 |
|
80bdf0c78
|
1956 1957 1958 1959 |
if (group_idle) sl = cfqd->cfq_group_idle; else sl = cfqd->cfq_slice_idle; |
206dc69b3
|
1960 |
|
7b14e3b52
|
1961 |
mod_timer(&cfqd->idle_slice_timer, jiffies + sl); |
e98ef89b3
|
1962 |
cfq_blkiocg_update_set_idle_time_stats(&cfqq->cfqg->blkg); |
80bdf0c78
|
1963 1964 |
cfq_log_cfqq(cfqd, cfqq, "arm_idle: %lu group_idle: %d", sl, group_idle ? 1 : 0); |
1da177e4c
|
1965 |
} |
498d3aa2b
|
1966 1967 1968 |
/* * Move request from internal lists to the request queue dispatch list. */ |
165125e1e
|
1969 |
static void cfq_dispatch_insert(struct request_queue *q, struct request *rq) |
1da177e4c
|
1970 |
{ |
3ed9a2965
|
1971 |
struct cfq_data *cfqd = q->elevator->elevator_data; |
5e7053747
|
1972 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
22e2c507c
|
1973 |
|
7b679138b
|
1974 |
cfq_log_cfqq(cfqd, cfqq, "dispatch_insert"); |
06d218864
|
1975 |
cfqq->next_rq = cfq_find_next_rq(cfqd, cfqq, rq); |
5380a101d
|
1976 |
cfq_remove_request(rq); |
6d048f531
|
1977 |
cfqq->dispatched++; |
80bdf0c78
|
1978 |
(RQ_CFQG(rq))->dispatched++; |
5380a101d
|
1979 |
elv_dispatch_sort(q, rq); |
3ed9a2965
|
1980 |
|
53c583d22
|
1981 |
cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]++; |
c4e7893eb
|
1982 |
cfqq->nr_sectors += blk_rq_sectors(rq); |
e98ef89b3
|
1983 |
cfq_blkiocg_update_dispatch_stats(&cfqq->cfqg->blkg, blk_rq_bytes(rq), |
84c124da9
|
1984 |
rq_data_dir(rq), rq_is_sync(rq)); |
1da177e4c
|
1985 1986 1987 1988 1989 |
} /* * return expired entry, or NULL to just start from scratch in rbtree */ |
febffd618
|
1990 |
static struct request *cfq_check_fifo(struct cfq_queue *cfqq) |
1da177e4c
|
1991 |
{ |
30996f40b
|
1992 |
struct request *rq = NULL; |
1da177e4c
|
1993 |
|
3b18152c3
|
1994 |
if (cfq_cfqq_fifo_expire(cfqq)) |
1da177e4c
|
1995 |
return NULL; |
cb8874119
|
1996 1997 |
cfq_mark_cfqq_fifo_expire(cfqq); |
89850f7ee
|
1998 1999 |
if (list_empty(&cfqq->fifo)) return NULL; |
1da177e4c
|
2000 |
|
89850f7ee
|
2001 |
rq = rq_entry_fifo(cfqq->fifo.next); |
30996f40b
|
2002 |
if (time_before(jiffies, rq_fifo_time(rq))) |
7b679138b
|
2003 |
rq = NULL; |
1da177e4c
|
2004 |
|
30996f40b
|
2005 |
cfq_log_cfqq(cfqq->cfqd, cfqq, "fifo=%p", rq); |
6d048f531
|
2006 |
return rq; |
1da177e4c
|
2007 |
} |
22e2c507c
|
2008 2009 2010 2011 |
static inline int cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq) { const int base_rq = cfqd->cfq_slice_async_rq; |
1da177e4c
|
2012 |
|
22e2c507c
|
2013 |
WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR); |
1da177e4c
|
2014 |
|
b9f8ce059
|
2015 |
return 2 * base_rq * (IOPRIO_BE_NR - cfqq->ioprio); |
1da177e4c
|
2016 |
} |
22e2c507c
|
2017 |
/* |
df5fe3e8e
|
2018 2019 2020 2021 2022 2023 2024 |
* Must be called with the queue_lock held. */ static int cfqq_process_refs(struct cfq_queue *cfqq) { int process_refs, io_refs; io_refs = cfqq->allocated[READ] + cfqq->allocated[WRITE]; |
30d7b9448
|
2025 |
process_refs = cfqq->ref - io_refs; |
df5fe3e8e
|
2026 2027 2028 2029 2030 2031 |
BUG_ON(process_refs < 0); return process_refs; } static void cfq_setup_merge(struct cfq_queue *cfqq, struct cfq_queue *new_cfqq) { |
e6c5bc737
|
2032 |
int process_refs, new_process_refs; |
df5fe3e8e
|
2033 |
struct cfq_queue *__cfqq; |
c10b61f09
|
2034 2035 2036 2037 2038 2039 2040 2041 |
/* * If there are no process references on the new_cfqq, then it is * unsafe to follow the ->new_cfqq chain as other cfqq's in the * chain may have dropped their last reference (not just their * last process reference). */ if (!cfqq_process_refs(new_cfqq)) return; |
df5fe3e8e
|
2042 2043 2044 2045 2046 2047 2048 2049 |
/* Avoid a circular list and skip interim queue merges */ while ((__cfqq = new_cfqq->new_cfqq)) { if (__cfqq == cfqq) return; new_cfqq = __cfqq; } process_refs = cfqq_process_refs(cfqq); |
c10b61f09
|
2050 |
new_process_refs = cfqq_process_refs(new_cfqq); |
df5fe3e8e
|
2051 2052 2053 2054 |
/* * If the process for the cfqq has gone away, there is no * sense in merging the queues. */ |
c10b61f09
|
2055 |
if (process_refs == 0 || new_process_refs == 0) |
df5fe3e8e
|
2056 |
return; |
e6c5bc737
|
2057 2058 2059 |
/* * Merge in the direction of the lesser amount of work. */ |
e6c5bc737
|
2060 2061 |
if (new_process_refs >= process_refs) { cfqq->new_cfqq = new_cfqq; |
30d7b9448
|
2062 |
new_cfqq->ref += process_refs; |
e6c5bc737
|
2063 2064 |
} else { new_cfqq->new_cfqq = cfqq; |
30d7b9448
|
2065 |
cfqq->ref += new_process_refs; |
e6c5bc737
|
2066 |
} |
df5fe3e8e
|
2067 |
} |
cdb16e8f7
|
2068 |
static enum wl_type_t cfq_choose_wl(struct cfq_data *cfqd, |
65b32a573
|
2069 |
struct cfq_group *cfqg, enum wl_prio_t prio) |
718eee057
|
2070 2071 2072 2073 2074 2075 |
{ struct cfq_queue *queue; int i; bool key_valid = false; unsigned long lowest_key = 0; enum wl_type_t cur_best = SYNC_NOIDLE_WORKLOAD; |
65b32a573
|
2076 2077 2078 |
for (i = 0; i <= SYNC_WORKLOAD; ++i) { /* select the one with lowest rb_key */ queue = cfq_rb_first(service_tree_for(cfqg, prio, i)); |
718eee057
|
2079 2080 2081 2082 2083 2084 2085 2086 2087 2088 |
if (queue && (!key_valid || time_before(queue->rb_key, lowest_key))) { lowest_key = queue->rb_key; cur_best = i; key_valid = true; } } return cur_best; } |
cdb16e8f7
|
2089 |
static void choose_service_tree(struct cfq_data *cfqd, struct cfq_group *cfqg) |
718eee057
|
2090 |
{ |
718eee057
|
2091 2092 |
unsigned slice; unsigned count; |
cdb16e8f7
|
2093 |
struct cfq_rb_root *st; |
58ff82f34
|
2094 |
unsigned group_slice; |
e4ea0c16a
|
2095 |
enum wl_prio_t original_prio = cfqd->serving_prio; |
1fa8f6d68
|
2096 |
|
718eee057
|
2097 |
/* Choose next priority. RT > BE > IDLE */ |
58ff82f34
|
2098 |
if (cfq_group_busy_queues_wl(RT_WORKLOAD, cfqd, cfqg)) |
718eee057
|
2099 |
cfqd->serving_prio = RT_WORKLOAD; |
58ff82f34
|
2100 |
else if (cfq_group_busy_queues_wl(BE_WORKLOAD, cfqd, cfqg)) |
718eee057
|
2101 2102 2103 2104 2105 2106 |
cfqd->serving_prio = BE_WORKLOAD; else { cfqd->serving_prio = IDLE_WORKLOAD; cfqd->workload_expires = jiffies + 1; return; } |
e4ea0c16a
|
2107 2108 |
if (original_prio != cfqd->serving_prio) goto new_workload; |
718eee057
|
2109 2110 2111 2112 2113 |
/* * For RT and BE, we have to choose also the type * (SYNC, SYNC_NOIDLE, ASYNC), and to compute a workload * expiration time */ |
65b32a573
|
2114 |
st = service_tree_for(cfqg, cfqd->serving_prio, cfqd->serving_type); |
cdb16e8f7
|
2115 |
count = st->count; |
718eee057
|
2116 2117 |
/* |
65b32a573
|
2118 |
* check workload expiration, and that we still have other queues ready |
718eee057
|
2119 |
*/ |
65b32a573
|
2120 |
if (count && !time_after(jiffies, cfqd->workload_expires)) |
718eee057
|
2121 |
return; |
e4ea0c16a
|
2122 |
new_workload: |
718eee057
|
2123 2124 |
/* otherwise select new workload type */ cfqd->serving_type = |
65b32a573
|
2125 2126 |
cfq_choose_wl(cfqd, cfqg, cfqd->serving_prio); st = service_tree_for(cfqg, cfqd->serving_prio, cfqd->serving_type); |
cdb16e8f7
|
2127 |
count = st->count; |
718eee057
|
2128 2129 2130 2131 2132 2133 |
/* * the workload slice is computed as a fraction of target latency * proportional to the number of queues in that workload, over * all the queues in the same priority class */ |
58ff82f34
|
2134 2135 2136 2137 2138 |
group_slice = cfq_group_slice(cfqd, cfqg); slice = group_slice * count / max_t(unsigned, cfqg->busy_queues_avg[cfqd->serving_prio], cfq_group_busy_queues_wl(cfqd->serving_prio, cfqd, cfqg)); |
718eee057
|
2139 |
|
f26bd1f0a
|
2140 2141 2142 2143 2144 2145 2146 2147 2148 2149 2150 2151 2152 |
if (cfqd->serving_type == ASYNC_WORKLOAD) { unsigned int tmp; /* * Async queues are currently system wide. Just taking * proportion of queues with-in same group will lead to higher * async ratio system wide as generally root group is going * to have higher weight. A more accurate thing would be to * calculate system wide asnc/sync ratio. */ tmp = cfq_target_latency * cfqg_busy_async_queues(cfqd, cfqg); tmp = tmp/cfqd->busy_queues; slice = min_t(unsigned, slice, tmp); |
718eee057
|
2153 2154 2155 |
/* async workload slice is scaled down according to * the sync/async slice ratio. */ slice = slice * cfqd->cfq_slice[0] / cfqd->cfq_slice[1]; |
f26bd1f0a
|
2156 |
} else |
718eee057
|
2157 2158 2159 2160 |
/* sync workload slice is at least 2 * cfq_slice_idle */ slice = max(slice, 2 * cfqd->cfq_slice_idle); slice = max_t(unsigned, slice, CFQ_MIN_TT); |
b1ffe737f
|
2161 |
cfq_log(cfqd, "workload slice:%d", slice); |
718eee057
|
2162 2163 |
cfqd->workload_expires = jiffies + slice; } |
1fa8f6d68
|
2164 2165 2166 |
static struct cfq_group *cfq_get_next_cfqg(struct cfq_data *cfqd) { struct cfq_rb_root *st = &cfqd->grp_service_tree; |
25bc6b077
|
2167 |
struct cfq_group *cfqg; |
1fa8f6d68
|
2168 2169 2170 |
if (RB_EMPTY_ROOT(&st->rb)) return NULL; |
25bc6b077
|
2171 |
cfqg = cfq_rb_first_group(st); |
25bc6b077
|
2172 2173 |
update_min_vdisktime(st); return cfqg; |
1fa8f6d68
|
2174 |
} |
cdb16e8f7
|
2175 2176 |
static void cfq_choose_cfqg(struct cfq_data *cfqd) { |
1fa8f6d68
|
2177 2178 2179 |
struct cfq_group *cfqg = cfq_get_next_cfqg(cfqd); cfqd->serving_group = cfqg; |
dae739ebc
|
2180 2181 2182 2183 2184 2185 |
/* Restore the workload type data */ if (cfqg->saved_workload_slice) { cfqd->workload_expires = jiffies + cfqg->saved_workload_slice; cfqd->serving_type = cfqg->saved_workload; cfqd->serving_prio = cfqg->saved_serving_prio; |
66ae29197
|
2186 2187 |
} else cfqd->workload_expires = jiffies - 1; |
1fa8f6d68
|
2188 |
choose_service_tree(cfqd, cfqg); |
cdb16e8f7
|
2189 |
} |
df5fe3e8e
|
2190 |
/* |
498d3aa2b
|
2191 2192 |
* Select a queue for service. If we have a current active queue, * check whether to continue servicing it, or retrieve and set a new one. |
22e2c507c
|
2193 |
*/ |
1b5ed5e1f
|
2194 |
static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd) |
1da177e4c
|
2195 |
{ |
a36e71f99
|
2196 |
struct cfq_queue *cfqq, *new_cfqq = NULL; |
1da177e4c
|
2197 |
|
22e2c507c
|
2198 2199 2200 |
cfqq = cfqd->active_queue; if (!cfqq) goto new_queue; |
1da177e4c
|
2201 |
|
f04a64246
|
2202 2203 |
if (!cfqd->rq_queued) return NULL; |
c244bb50a
|
2204 2205 2206 2207 2208 2209 |
/* * We were waiting for group to get backlogged. Expire the queue */ if (cfq_cfqq_wait_busy(cfqq) && !RB_EMPTY_ROOT(&cfqq->sort_list)) goto expire; |
22e2c507c
|
2210 |
/* |
6d048f531
|
2211 |
* The active queue has run out of time, expire it and select new. |
22e2c507c
|
2212 |
*/ |
7667aa063
|
2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 |
if (cfq_slice_used(cfqq) && !cfq_cfqq_must_dispatch(cfqq)) { /* * If slice had not expired at the completion of last request * we might not have turned on wait_busy flag. Don't expire * the queue yet. Allow the group to get backlogged. * * The very fact that we have used the slice, that means we * have been idling all along on this queue and it should be * ok to wait for this request to complete. */ |
82bbbf28d
|
2223 2224 2225 |
if (cfqq->cfqg->nr_cfqq == 1 && RB_EMPTY_ROOT(&cfqq->sort_list) && cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) { cfqq = NULL; |
7667aa063
|
2226 |
goto keep_queue; |
82bbbf28d
|
2227 |
} else |
80bdf0c78
|
2228 |
goto check_group_idle; |
7667aa063
|
2229 |
} |
1da177e4c
|
2230 |
|
22e2c507c
|
2231 |
/* |
6d048f531
|
2232 2233 |
* The active queue has requests and isn't expired, allow it to * dispatch. |
22e2c507c
|
2234 |
*/ |
dd67d0515
|
2235 |
if (!RB_EMPTY_ROOT(&cfqq->sort_list)) |
22e2c507c
|
2236 |
goto keep_queue; |
6d048f531
|
2237 2238 |
/* |
a36e71f99
|
2239 2240 2241 |
* If another queue has a request waiting within our mean seek * distance, let it run. The expire code will check for close * cooperators and put the close queue at the front of the service |
df5fe3e8e
|
2242 |
* tree. If possible, merge the expiring queue with the new cfqq. |
a36e71f99
|
2243 |
*/ |
b3b6d0408
|
2244 |
new_cfqq = cfq_close_cooperator(cfqd, cfqq); |
df5fe3e8e
|
2245 2246 2247 |
if (new_cfqq) { if (!cfqq->new_cfqq) cfq_setup_merge(cfqq, new_cfqq); |
a36e71f99
|
2248 |
goto expire; |
df5fe3e8e
|
2249 |
} |
a36e71f99
|
2250 2251 |
/* |
6d048f531
|
2252 2253 2254 2255 |
* No requests pending. If the active queue still has requests in * flight or is idling for a new request, allow either of these * conditions to happen (or time out) before selecting a new queue. */ |
80bdf0c78
|
2256 2257 2258 2259 |
if (timer_pending(&cfqd->idle_slice_timer)) { cfqq = NULL; goto keep_queue; } |
8e1ac6655
|
2260 2261 2262 2263 2264 2265 2266 2267 2268 2269 |
/* * This is a deep seek queue, but the device is much faster than * the queue can deliver, don't idle **/ if (CFQQ_SEEKY(cfqq) && cfq_cfqq_idle_window(cfqq) && (cfq_cfqq_slice_new(cfqq) || (cfqq->slice_end - jiffies > jiffies - cfqq->slice_start))) { cfq_clear_cfqq_deep(cfqq); cfq_clear_cfqq_idle_window(cfqq); } |
80bdf0c78
|
2270 2271 2272 2273 2274 2275 2276 2277 2278 2279 |
if (cfqq->dispatched && cfq_should_idle(cfqd, cfqq)) { cfqq = NULL; goto keep_queue; } /* * If group idle is enabled and there are requests dispatched from * this group, wait for requests to complete. */ check_group_idle: |
7700fc4f6
|
2280 2281 2282 |
if (cfqd->cfq_group_idle && cfqq->cfqg->nr_cfqq == 1 && cfqq->cfqg->dispatched && !cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true)) { |
caaa5f9f0
|
2283 2284 |
cfqq = NULL; goto keep_queue; |
22e2c507c
|
2285 |
} |
3b18152c3
|
2286 |
expire: |
e5ff082e8
|
2287 |
cfq_slice_expired(cfqd, 0); |
3b18152c3
|
2288 |
new_queue: |
718eee057
|
2289 2290 2291 2292 2293 |
/* * Current queue expired. Check if we have to switch to a new * service tree */ if (!new_cfqq) |
cdb16e8f7
|
2294 |
cfq_choose_cfqg(cfqd); |
718eee057
|
2295 |
|
a36e71f99
|
2296 |
cfqq = cfq_set_active_queue(cfqd, new_cfqq); |
22e2c507c
|
2297 |
keep_queue: |
3b18152c3
|
2298 |
return cfqq; |
22e2c507c
|
2299 |
} |
febffd618
|
2300 |
static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq) |
d9e7620e6
|
2301 2302 2303 2304 2305 2306 2307 2308 2309 |
{ int dispatched = 0; while (cfqq->next_rq) { cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq); dispatched++; } BUG_ON(!list_empty(&cfqq->fifo)); |
f04a64246
|
2310 2311 |
/* By default cfqq is not expired if it is empty. Do it explicitly */ |
e5ff082e8
|
2312 |
__cfq_slice_expired(cfqq->cfqd, cfqq, 0); |
d9e7620e6
|
2313 2314 |
return dispatched; } |
498d3aa2b
|
2315 2316 2317 2318 |
/* * Drain our current requests. Used for barriers and when switching * io schedulers on-the-fly. */ |
d9e7620e6
|
2319 |
static int cfq_forced_dispatch(struct cfq_data *cfqd) |
1b5ed5e1f
|
2320 |
{ |
0871714e0
|
2321 |
struct cfq_queue *cfqq; |
d9e7620e6
|
2322 |
int dispatched = 0; |
cdb16e8f7
|
2323 |
|
3440c49f5
|
2324 |
/* Expire the timeslice of the current active queue first */ |
e5ff082e8
|
2325 |
cfq_slice_expired(cfqd, 0); |
3440c49f5
|
2326 2327 |
while ((cfqq = cfq_get_next_queue_forced(cfqd)) != NULL) { __cfq_set_active_queue(cfqd, cfqq); |
f04a64246
|
2328 |
dispatched += __cfq_forced_dispatch_cfqq(cfqq); |
3440c49f5
|
2329 |
} |
1b5ed5e1f
|
2330 |
|
1b5ed5e1f
|
2331 |
BUG_ON(cfqd->busy_queues); |
6923715ae
|
2332 |
cfq_log(cfqd, "forced_dispatch=%d", dispatched); |
1b5ed5e1f
|
2333 2334 |
return dispatched; } |
abc3c744d
|
2335 2336 2337 2338 2339 |
static inline bool cfq_slice_used_soon(struct cfq_data *cfqd, struct cfq_queue *cfqq) { /* the queue hasn't finished any request, can't estimate */ if (cfq_cfqq_slice_new(cfqq)) |
c1e44756f
|
2340 |
return true; |
abc3c744d
|
2341 2342 |
if (time_after(jiffies + cfqd->cfq_slice_idle * cfqq->dispatched, cfqq->slice_end)) |
c1e44756f
|
2343 |
return true; |
abc3c744d
|
2344 |
|
c1e44756f
|
2345 |
return false; |
abc3c744d
|
2346 |
} |
0b182d617
|
2347 |
static bool cfq_may_dispatch(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
2f5cb7381
|
2348 |
{ |
2f5cb7381
|
2349 |
unsigned int max_dispatch; |
22e2c507c
|
2350 |
|
2f5cb7381
|
2351 |
/* |
5ad531db6
|
2352 2353 |
* Drain async requests before we start sync IO */ |
53c583d22
|
2354 |
if (cfq_should_idle(cfqd, cfqq) && cfqd->rq_in_flight[BLK_RW_ASYNC]) |
0b182d617
|
2355 |
return false; |
5ad531db6
|
2356 2357 |
/* |
2f5cb7381
|
2358 2359 |
* If this is an async queue and we have sync IO in flight, let it wait */ |
53c583d22
|
2360 |
if (cfqd->rq_in_flight[BLK_RW_SYNC] && !cfq_cfqq_sync(cfqq)) |
0b182d617
|
2361 |
return false; |
2f5cb7381
|
2362 |
|
abc3c744d
|
2363 |
max_dispatch = max_t(unsigned int, cfqd->cfq_quantum / 2, 1); |
2f5cb7381
|
2364 2365 |
if (cfq_class_idle(cfqq)) max_dispatch = 1; |
b4878f245
|
2366 |
|
2f5cb7381
|
2367 2368 2369 2370 |
/* * Does this cfqq already have too much IO in flight? */ if (cfqq->dispatched >= max_dispatch) { |
ef8a41df8
|
2371 |
bool promote_sync = false; |
2f5cb7381
|
2372 2373 2374 |
/* * idle queue must always only have a single IO in flight */ |
3ed9a2965
|
2375 |
if (cfq_class_idle(cfqq)) |
0b182d617
|
2376 |
return false; |
3ed9a2965
|
2377 |
|
2f5cb7381
|
2378 |
/* |
c4ade94fc
|
2379 2380 |
* If there is only one sync queue * we can ignore async queue here and give the sync |
ef8a41df8
|
2381 2382 2383 2384 |
* queue no dispatch limit. The reason is a sync queue can * preempt async queue, limiting the sync queue doesn't make * sense. This is useful for aiostress test. */ |
c4ade94fc
|
2385 2386 |
if (cfq_cfqq_sync(cfqq) && cfqd->busy_sync_queues == 1) promote_sync = true; |
ef8a41df8
|
2387 2388 |
/* |
2f5cb7381
|
2389 2390 |
* We have other queues, don't allow more IO from this one */ |
ef8a41df8
|
2391 2392 |
if (cfqd->busy_queues > 1 && cfq_slice_used_soon(cfqd, cfqq) && !promote_sync) |
0b182d617
|
2393 |
return false; |
9ede209e8
|
2394 |
|
2f5cb7381
|
2395 |
/* |
474b18ccc
|
2396 |
* Sole queue user, no limit |
365722bb9
|
2397 |
*/ |
ef8a41df8
|
2398 |
if (cfqd->busy_queues == 1 || promote_sync) |
abc3c744d
|
2399 2400 2401 2402 2403 2404 2405 2406 2407 |
max_dispatch = -1; else /* * Normally we start throttling cfqq when cfq_quantum/2 * requests have been dispatched. But we can drive * deeper queue depths at the beginning of slice * subjected to upper limit of cfq_quantum. * */ max_dispatch = cfqd->cfq_quantum; |
8e2967555
|
2408 2409 2410 2411 2412 2413 2414 |
} /* * Async queues must wait a bit before being allowed dispatch. * We also ramp up the dispatch depth gradually for async IO, * based on the last sync IO we serviced */ |
963b72fc6
|
2415 |
if (!cfq_cfqq_sync(cfqq) && cfqd->cfq_latency) { |
573412b29
|
2416 |
unsigned long last_sync = jiffies - cfqd->last_delayed_sync; |
8e2967555
|
2417 |
unsigned int depth; |
365722bb9
|
2418 |
|
61f0c1dca
|
2419 |
depth = last_sync / cfqd->cfq_slice[1]; |
e00c54c36
|
2420 2421 |
if (!depth && !cfqq->dispatched) depth = 1; |
8e2967555
|
2422 2423 |
if (depth < max_dispatch) max_dispatch = depth; |
2f5cb7381
|
2424 |
} |
3ed9a2965
|
2425 |
|
0b182d617
|
2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444 2445 2446 2447 2448 2449 2450 2451 2452 2453 2454 2455 2456 2457 2458 2459 2460 2461 2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475 2476 2477 2478 2479 2480 2481 2482 2483 |
/* * If we're below the current max, allow a dispatch */ return cfqq->dispatched < max_dispatch; } /* * Dispatch a request from cfqq, moving them to the request queue * dispatch list. */ static bool cfq_dispatch_request(struct cfq_data *cfqd, struct cfq_queue *cfqq) { struct request *rq; BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list)); if (!cfq_may_dispatch(cfqd, cfqq)) return false; /* * follow expired path, else get first next available */ rq = cfq_check_fifo(cfqq); if (!rq) rq = cfqq->next_rq; /* * insert request into driver dispatch list */ cfq_dispatch_insert(cfqd->queue, rq); if (!cfqd->active_cic) { struct cfq_io_context *cic = RQ_CIC(rq); atomic_long_inc(&cic->ioc->refcount); cfqd->active_cic = cic; } return true; } /* * Find the cfqq that we need to service and move a request from that to the * dispatch list */ static int cfq_dispatch_requests(struct request_queue *q, int force) { struct cfq_data *cfqd = q->elevator->elevator_data; struct cfq_queue *cfqq; if (!cfqd->busy_queues) return 0; if (unlikely(force)) return cfq_forced_dispatch(cfqd); cfqq = cfq_select_queue(cfqd); if (!cfqq) |
8e2967555
|
2484 |
return 0; |
2f5cb7381
|
2485 |
/* |
0b182d617
|
2486 |
* Dispatch a request from this cfqq, if it is allowed |
2f5cb7381
|
2487 |
*/ |
0b182d617
|
2488 2489 |
if (!cfq_dispatch_request(cfqd, cfqq)) return 0; |
2f5cb7381
|
2490 |
cfqq->slice_dispatch++; |
b029195dd
|
2491 |
cfq_clear_cfqq_must_dispatch(cfqq); |
22e2c507c
|
2492 |
|
2f5cb7381
|
2493 2494 2495 2496 2497 2498 2499 2500 |
/* * expire an async queue immediately if it has used up its slice. idle * queue always expire after 1 dispatch round. */ if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) && cfqq->slice_dispatch >= cfq_prio_to_maxrq(cfqd, cfqq)) || cfq_class_idle(cfqq))) { cfqq->slice_end = jiffies + 1; |
e5ff082e8
|
2501 |
cfq_slice_expired(cfqd, 0); |
1da177e4c
|
2502 |
} |
b217a903a
|
2503 |
cfq_log_cfqq(cfqd, cfqq, "dispatched a request"); |
2f5cb7381
|
2504 |
return 1; |
1da177e4c
|
2505 |
} |
1da177e4c
|
2506 |
/* |
5e7053747
|
2507 2508 |
* task holds one reference to the queue, dropped when task exits. each rq * in-flight on this queue also holds a reference, dropped when rq is freed. |
1da177e4c
|
2509 |
* |
b1c357696
|
2510 |
* Each cfq queue took a reference on the parent group. Drop it now. |
1da177e4c
|
2511 2512 2513 2514 |
* queue lock must be held here. */ static void cfq_put_queue(struct cfq_queue *cfqq) { |
22e2c507c
|
2515 |
struct cfq_data *cfqd = cfqq->cfqd; |
0bbfeb832
|
2516 |
struct cfq_group *cfqg; |
22e2c507c
|
2517 |
|
30d7b9448
|
2518 |
BUG_ON(cfqq->ref <= 0); |
1da177e4c
|
2519 |
|
30d7b9448
|
2520 2521 |
cfqq->ref--; if (cfqq->ref) |
1da177e4c
|
2522 |
return; |
7b679138b
|
2523 |
cfq_log_cfqq(cfqd, cfqq, "put_queue"); |
1da177e4c
|
2524 |
BUG_ON(rb_first(&cfqq->sort_list)); |
22e2c507c
|
2525 |
BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]); |
b1c357696
|
2526 |
cfqg = cfqq->cfqg; |
1da177e4c
|
2527 |
|
28f95cbc3
|
2528 |
if (unlikely(cfqd->active_queue == cfqq)) { |
e5ff082e8
|
2529 |
__cfq_slice_expired(cfqd, cfqq, 0); |
23e018a1b
|
2530 |
cfq_schedule_dispatch(cfqd); |
28f95cbc3
|
2531 |
} |
22e2c507c
|
2532 |
|
f04a64246
|
2533 |
BUG_ON(cfq_cfqq_on_rr(cfqq)); |
1da177e4c
|
2534 |
kmem_cache_free(cfq_pool, cfqq); |
b1c357696
|
2535 |
cfq_put_cfqg(cfqg); |
1da177e4c
|
2536 |
} |
d6de8be71
|
2537 |
/* |
5f45c6958
|
2538 |
* Call func for each cic attached to this ioc. |
d6de8be71
|
2539 |
*/ |
07416d29b
|
2540 |
static void |
5f45c6958
|
2541 2542 |
call_for_each_cic(struct io_context *ioc, void (*func)(struct io_context *, struct cfq_io_context *)) |
07416d29b
|
2543 2544 2545 |
{ struct cfq_io_context *cic; struct hlist_node *n; |
5f45c6958
|
2546 |
rcu_read_lock(); |
07416d29b
|
2547 2548 |
hlist_for_each_entry_rcu(cic, n, &ioc->cic_list, cic_list) func(ioc, cic); |
07416d29b
|
2549 |
|
4ac845a2e
|
2550 |
rcu_read_unlock(); |
34e6bbf23
|
2551 2552 2553 2554 2555 2556 2557 2558 2559 |
} static void cfq_cic_free_rcu(struct rcu_head *head) { struct cfq_io_context *cic; cic = container_of(head, struct cfq_io_context, rcu_head); kmem_cache_free(cfq_ioc_pool, cic); |
245b2e70e
|
2560 |
elv_ioc_count_dec(cfq_ioc_count); |
34e6bbf23
|
2561 |
|
9a11b4ed0
|
2562 2563 2564 2565 2566 2567 2568 |
if (ioc_gone) { /* * CFQ scheduler is exiting, grab exit lock and check * the pending io context count. If it hits zero, * complete ioc_gone and set it back to NULL */ spin_lock(&ioc_gone_lock); |
245b2e70e
|
2569 |
if (ioc_gone && !elv_ioc_count_read(cfq_ioc_count)) { |
9a11b4ed0
|
2570 2571 2572 2573 2574 |
complete(ioc_gone); ioc_gone = NULL; } spin_unlock(&ioc_gone_lock); } |
34e6bbf23
|
2575 |
} |
4ac845a2e
|
2576 |
|
34e6bbf23
|
2577 2578 2579 |
static void cfq_cic_free(struct cfq_io_context *cic) { call_rcu(&cic->rcu_head, cfq_cic_free_rcu); |
4ac845a2e
|
2580 2581 2582 2583 2584 |
} static void cic_free_func(struct io_context *ioc, struct cfq_io_context *cic) { unsigned long flags; |
bca4b914b
|
2585 |
unsigned long dead_key = (unsigned long) cic->key; |
4ac845a2e
|
2586 |
|
bca4b914b
|
2587 |
BUG_ON(!(dead_key & CIC_DEAD_KEY)); |
4ac845a2e
|
2588 2589 |
spin_lock_irqsave(&ioc->lock, flags); |
80b15c738
|
2590 |
radix_tree_delete(&ioc->radix_root, dead_key >> CIC_DEAD_INDEX_SHIFT); |
ffc4e7595
|
2591 |
hlist_del_rcu(&cic->cic_list); |
4ac845a2e
|
2592 |
spin_unlock_irqrestore(&ioc->lock, flags); |
34e6bbf23
|
2593 |
cfq_cic_free(cic); |
4ac845a2e
|
2594 |
} |
d6de8be71
|
2595 2596 2597 2598 2599 |
/* * Must be called with rcu_read_lock() held or preemption otherwise disabled. * Only two callers of this - ->dtor() which is called with the rcu_read_lock(), * and ->trim() which is called with the task lock held */ |
4ac845a2e
|
2600 2601 |
static void cfq_free_io_context(struct io_context *ioc) { |
4ac845a2e
|
2602 |
/* |
34e6bbf23
|
2603 2604 2605 2606 |
* ioc->refcount is zero here, or we are called from elv_unregister(), * so no more cic's are allowed to be linked into this ioc. So it * should be ok to iterate over the known list, we will see all cic's * since no new ones are added. |
4ac845a2e
|
2607 |
*/ |
5f45c6958
|
2608 |
call_for_each_cic(ioc, cic_free_func); |
1da177e4c
|
2609 |
} |
d02a2c077
|
2610 |
static void cfq_put_cooperator(struct cfq_queue *cfqq) |
1da177e4c
|
2611 |
{ |
df5fe3e8e
|
2612 |
struct cfq_queue *__cfqq, *next; |
df5fe3e8e
|
2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 |
/* * If this queue was scheduled to merge with another queue, be * sure to drop the reference taken on that queue (and others in * the merge chain). See cfq_setup_merge and cfq_merge_cfqqs. */ __cfqq = cfqq->new_cfqq; while (__cfqq) { if (__cfqq == cfqq) { WARN(1, "cfqq->new_cfqq loop detected "); break; } next = __cfqq->new_cfqq; cfq_put_queue(__cfqq); __cfqq = next; } |
d02a2c077
|
2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 |
} static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq) { if (unlikely(cfqq == cfqd->active_queue)) { __cfq_slice_expired(cfqd, cfqq, 0); cfq_schedule_dispatch(cfqd); } cfq_put_cooperator(cfqq); |
df5fe3e8e
|
2639 |
|
89850f7ee
|
2640 2641 |
cfq_put_queue(cfqq); } |
22e2c507c
|
2642 |
|
89850f7ee
|
2643 2644 2645 |
static void __cfq_exit_single_io_context(struct cfq_data *cfqd, struct cfq_io_context *cic) { |
4faa3c815
|
2646 |
struct io_context *ioc = cic->ioc; |
fc46379da
|
2647 |
list_del_init(&cic->queue_list); |
4ac845a2e
|
2648 2649 |
/* |
bca4b914b
|
2650 |
* Make sure dead mark is seen for dead queues |
4ac845a2e
|
2651 |
*/ |
fc46379da
|
2652 |
smp_wmb(); |
bca4b914b
|
2653 |
cic->key = cfqd_dead_key(cfqd); |
fc46379da
|
2654 |
|
3181faa85
|
2655 |
rcu_read_lock(); |
9b50902db
|
2656 |
if (rcu_dereference(ioc->ioc_data) == cic) { |
3181faa85
|
2657 |
rcu_read_unlock(); |
9b50902db
|
2658 |
spin_lock(&ioc->lock); |
4faa3c815
|
2659 |
rcu_assign_pointer(ioc->ioc_data, NULL); |
9b50902db
|
2660 |
spin_unlock(&ioc->lock); |
3181faa85
|
2661 2662 |
} else rcu_read_unlock(); |
4faa3c815
|
2663 |
|
ff6657c6c
|
2664 2665 2666 |
if (cic->cfqq[BLK_RW_ASYNC]) { cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_ASYNC]); cic->cfqq[BLK_RW_ASYNC] = NULL; |
12a057321
|
2667 |
} |
ff6657c6c
|
2668 2669 2670 |
if (cic->cfqq[BLK_RW_SYNC]) { cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_SYNC]); cic->cfqq[BLK_RW_SYNC] = NULL; |
12a057321
|
2671 |
} |
89850f7ee
|
2672 |
} |
4ac845a2e
|
2673 2674 |
static void cfq_exit_single_io_context(struct io_context *ioc, struct cfq_io_context *cic) |
89850f7ee
|
2675 |
{ |
bca4b914b
|
2676 |
struct cfq_data *cfqd = cic_to_cfqd(cic); |
89850f7ee
|
2677 |
|
89850f7ee
|
2678 |
if (cfqd) { |
165125e1e
|
2679 |
struct request_queue *q = cfqd->queue; |
4ac845a2e
|
2680 |
unsigned long flags; |
89850f7ee
|
2681 |
|
4ac845a2e
|
2682 |
spin_lock_irqsave(q->queue_lock, flags); |
62c1fe9d9
|
2683 2684 2685 2686 2687 2688 |
/* * Ensure we get a fresh copy of the ->key to prevent * race between exiting task and queue */ smp_read_barrier_depends(); |
bca4b914b
|
2689 |
if (cic->key == cfqd) |
62c1fe9d9
|
2690 |
__cfq_exit_single_io_context(cfqd, cic); |
4ac845a2e
|
2691 |
spin_unlock_irqrestore(q->queue_lock, flags); |
89850f7ee
|
2692 |
} |
1da177e4c
|
2693 |
} |
498d3aa2b
|
2694 2695 2696 2697 |
/* * The process that ioc belongs to has exited, we need to clean up * and put the internal structures we have that belongs to that process. */ |
e2d74ac06
|
2698 |
static void cfq_exit_io_context(struct io_context *ioc) |
1da177e4c
|
2699 |
{ |
4ac845a2e
|
2700 |
call_for_each_cic(ioc, cfq_exit_single_io_context); |
1da177e4c
|
2701 |
} |
22e2c507c
|
2702 |
static struct cfq_io_context * |
8267e268e
|
2703 |
cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask) |
1da177e4c
|
2704 |
{ |
b5deef901
|
2705 |
struct cfq_io_context *cic; |
1da177e4c
|
2706 |
|
94f6030ca
|
2707 2708 |
cic = kmem_cache_alloc_node(cfq_ioc_pool, gfp_mask | __GFP_ZERO, cfqd->queue->node); |
1da177e4c
|
2709 |
if (cic) { |
383cd7213
|
2710 |
cic->ttime.last_end_request = jiffies; |
553698f94
|
2711 |
INIT_LIST_HEAD(&cic->queue_list); |
ffc4e7595
|
2712 |
INIT_HLIST_NODE(&cic->cic_list); |
22e2c507c
|
2713 2714 |
cic->dtor = cfq_free_io_context; cic->exit = cfq_exit_io_context; |
245b2e70e
|
2715 |
elv_ioc_count_inc(cfq_ioc_count); |
1da177e4c
|
2716 2717 2718 2719 |
} return cic; } |
fd0928df9
|
2720 |
static void cfq_init_prio_data(struct cfq_queue *cfqq, struct io_context *ioc) |
22e2c507c
|
2721 2722 2723 |
{ struct task_struct *tsk = current; int ioprio_class; |
3b18152c3
|
2724 |
if (!cfq_cfqq_prio_changed(cfqq)) |
22e2c507c
|
2725 |
return; |
fd0928df9
|
2726 |
ioprio_class = IOPRIO_PRIO_CLASS(ioc->ioprio); |
22e2c507c
|
2727 |
switch (ioprio_class) { |
fe094d98e
|
2728 2729 2730 2731 2732 |
default: printk(KERN_ERR "cfq: bad prio %x ", ioprio_class); case IOPRIO_CLASS_NONE: /* |
6d63c2755
|
2733 |
* no prio set, inherit CPU scheduling settings |
fe094d98e
|
2734 2735 |
*/ cfqq->ioprio = task_nice_ioprio(tsk); |
6d63c2755
|
2736 |
cfqq->ioprio_class = task_nice_ioclass(tsk); |
fe094d98e
|
2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 |
break; case IOPRIO_CLASS_RT: cfqq->ioprio = task_ioprio(ioc); cfqq->ioprio_class = IOPRIO_CLASS_RT; break; case IOPRIO_CLASS_BE: cfqq->ioprio = task_ioprio(ioc); cfqq->ioprio_class = IOPRIO_CLASS_BE; break; case IOPRIO_CLASS_IDLE: cfqq->ioprio_class = IOPRIO_CLASS_IDLE; cfqq->ioprio = 7; cfq_clear_cfqq_idle_window(cfqq); break; |
22e2c507c
|
2751 2752 2753 2754 2755 2756 2757 |
} /* * keep track of original prio settings in case we have to temporarily * elevate the priority of this queue */ cfqq->org_ioprio = cfqq->ioprio; |
3b18152c3
|
2758 |
cfq_clear_cfqq_prio_changed(cfqq); |
22e2c507c
|
2759 |
} |
febffd618
|
2760 |
static void changed_ioprio(struct io_context *ioc, struct cfq_io_context *cic) |
22e2c507c
|
2761 |
{ |
bca4b914b
|
2762 |
struct cfq_data *cfqd = cic_to_cfqd(cic); |
478a82b0e
|
2763 |
struct cfq_queue *cfqq; |
c1b707d25
|
2764 |
unsigned long flags; |
35e6077cb
|
2765 |
|
caaa5f9f0
|
2766 2767 |
if (unlikely(!cfqd)) return; |
c1b707d25
|
2768 |
spin_lock_irqsave(cfqd->queue->queue_lock, flags); |
caaa5f9f0
|
2769 |
|
ff6657c6c
|
2770 |
cfqq = cic->cfqq[BLK_RW_ASYNC]; |
caaa5f9f0
|
2771 2772 |
if (cfqq) { struct cfq_queue *new_cfqq; |
ff6657c6c
|
2773 2774 |
new_cfqq = cfq_get_queue(cfqd, BLK_RW_ASYNC, cic->ioc, GFP_ATOMIC); |
caaa5f9f0
|
2775 |
if (new_cfqq) { |
ff6657c6c
|
2776 |
cic->cfqq[BLK_RW_ASYNC] = new_cfqq; |
caaa5f9f0
|
2777 2778 |
cfq_put_queue(cfqq); } |
22e2c507c
|
2779 |
} |
caaa5f9f0
|
2780 |
|
ff6657c6c
|
2781 |
cfqq = cic->cfqq[BLK_RW_SYNC]; |
caaa5f9f0
|
2782 2783 |
if (cfqq) cfq_mark_cfqq_prio_changed(cfqq); |
c1b707d25
|
2784 |
spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); |
22e2c507c
|
2785 |
} |
fc46379da
|
2786 |
static void cfq_ioc_set_ioprio(struct io_context *ioc) |
22e2c507c
|
2787 |
{ |
4ac845a2e
|
2788 |
call_for_each_cic(ioc, changed_ioprio); |
fc46379da
|
2789 |
ioc->ioprio_changed = 0; |
22e2c507c
|
2790 |
} |
d5036d770
|
2791 |
static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq, |
a6151c3a5
|
2792 |
pid_t pid, bool is_sync) |
d5036d770
|
2793 2794 2795 2796 |
{ RB_CLEAR_NODE(&cfqq->rb_node); RB_CLEAR_NODE(&cfqq->p_node); INIT_LIST_HEAD(&cfqq->fifo); |
30d7b9448
|
2797 |
cfqq->ref = 0; |
d5036d770
|
2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 |
cfqq->cfqd = cfqd; cfq_mark_cfqq_prio_changed(cfqq); if (is_sync) { if (!cfq_class_idle(cfqq)) cfq_mark_cfqq_idle_window(cfqq); cfq_mark_cfqq_sync(cfqq); } cfqq->pid = pid; } |
24610333d
|
2809 2810 2811 2812 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED static void changed_cgroup(struct io_context *ioc, struct cfq_io_context *cic) { struct cfq_queue *sync_cfqq = cic_to_cfqq(cic, 1); |
bca4b914b
|
2813 |
struct cfq_data *cfqd = cic_to_cfqd(cic); |
24610333d
|
2814 2815 2816 2817 2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830 2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 |
unsigned long flags; struct request_queue *q; if (unlikely(!cfqd)) return; q = cfqd->queue; spin_lock_irqsave(q->queue_lock, flags); if (sync_cfqq) { /* * Drop reference to sync queue. A new sync queue will be * assigned in new group upon arrival of a fresh request. */ cfq_log_cfqq(cfqd, sync_cfqq, "changed cgroup"); cic_set_cfqq(cic, NULL, 1); cfq_put_queue(sync_cfqq); } spin_unlock_irqrestore(q->queue_lock, flags); } static void cfq_ioc_set_cgroup(struct io_context *ioc) { call_for_each_cic(ioc, changed_cgroup); ioc->cgroup_changed = 0; } #endif /* CONFIG_CFQ_GROUP_IOSCHED */ |
22e2c507c
|
2843 |
static struct cfq_queue * |
a6151c3a5
|
2844 |
cfq_find_alloc_queue(struct cfq_data *cfqd, bool is_sync, |
fd0928df9
|
2845 |
struct io_context *ioc, gfp_t gfp_mask) |
22e2c507c
|
2846 |
{ |
22e2c507c
|
2847 |
struct cfq_queue *cfqq, *new_cfqq = NULL; |
91fac317a
|
2848 |
struct cfq_io_context *cic; |
cdb16e8f7
|
2849 |
struct cfq_group *cfqg; |
22e2c507c
|
2850 2851 |
retry: |
3e59cf9d6
|
2852 |
cfqg = cfq_get_cfqg(cfqd); |
4ac845a2e
|
2853 |
cic = cfq_cic_lookup(cfqd, ioc); |
91fac317a
|
2854 2855 |
/* cic always exists here */ cfqq = cic_to_cfqq(cic, is_sync); |
22e2c507c
|
2856 |
|
6118b70b3
|
2857 2858 2859 2860 2861 2862 |
/* * Always try a new alloc if we fell back to the OOM cfqq * originally, since it should just be a temporary situation. */ if (!cfqq || cfqq == &cfqd->oom_cfqq) { cfqq = NULL; |
22e2c507c
|
2863 2864 2865 2866 2867 |
if (new_cfqq) { cfqq = new_cfqq; new_cfqq = NULL; } else if (gfp_mask & __GFP_WAIT) { spin_unlock_irq(cfqd->queue->queue_lock); |
94f6030ca
|
2868 |
new_cfqq = kmem_cache_alloc_node(cfq_pool, |
6118b70b3
|
2869 |
gfp_mask | __GFP_ZERO, |
94f6030ca
|
2870 |
cfqd->queue->node); |
22e2c507c
|
2871 |
spin_lock_irq(cfqd->queue->queue_lock); |
6118b70b3
|
2872 2873 |
if (new_cfqq) goto retry; |
22e2c507c
|
2874 |
} else { |
94f6030ca
|
2875 2876 2877 |
cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask | __GFP_ZERO, cfqd->queue->node); |
22e2c507c
|
2878 |
} |
6118b70b3
|
2879 2880 2881 |
if (cfqq) { cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync); cfq_init_prio_data(cfqq, ioc); |
cdb16e8f7
|
2882 |
cfq_link_cfqq_cfqg(cfqq, cfqg); |
6118b70b3
|
2883 2884 2885 |
cfq_log_cfqq(cfqd, cfqq, "alloced"); } else cfqq = &cfqd->oom_cfqq; |
22e2c507c
|
2886 2887 2888 2889 |
} if (new_cfqq) kmem_cache_free(cfq_pool, new_cfqq); |
22e2c507c
|
2890 2891 |
return cfqq; } |
c2dea2d1f
|
2892 2893 2894 |
static struct cfq_queue ** cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio) { |
fe094d98e
|
2895 |
switch (ioprio_class) { |
c2dea2d1f
|
2896 2897 2898 2899 2900 2901 2902 2903 2904 2905 |
case IOPRIO_CLASS_RT: return &cfqd->async_cfqq[0][ioprio]; case IOPRIO_CLASS_BE: return &cfqd->async_cfqq[1][ioprio]; case IOPRIO_CLASS_IDLE: return &cfqd->async_idle_cfqq; default: BUG(); } } |
15c31be4d
|
2906 |
static struct cfq_queue * |
a6151c3a5
|
2907 |
cfq_get_queue(struct cfq_data *cfqd, bool is_sync, struct io_context *ioc, |
15c31be4d
|
2908 2909 |
gfp_t gfp_mask) { |
fd0928df9
|
2910 2911 |
const int ioprio = task_ioprio(ioc); const int ioprio_class = task_ioprio_class(ioc); |
c2dea2d1f
|
2912 |
struct cfq_queue **async_cfqq = NULL; |
15c31be4d
|
2913 |
struct cfq_queue *cfqq = NULL; |
c2dea2d1f
|
2914 2915 2916 2917 |
if (!is_sync) { async_cfqq = cfq_async_queue_prio(cfqd, ioprio_class, ioprio); cfqq = *async_cfqq; } |
6118b70b3
|
2918 |
if (!cfqq) |
fd0928df9
|
2919 |
cfqq = cfq_find_alloc_queue(cfqd, is_sync, ioc, gfp_mask); |
15c31be4d
|
2920 2921 2922 2923 |
/* * pin the queue now that it's allocated, scheduler exit will prune it */ |
c2dea2d1f
|
2924 |
if (!is_sync && !(*async_cfqq)) { |
30d7b9448
|
2925 |
cfqq->ref++; |
c2dea2d1f
|
2926 |
*async_cfqq = cfqq; |
15c31be4d
|
2927 |
} |
30d7b9448
|
2928 |
cfqq->ref++; |
15c31be4d
|
2929 2930 |
return cfqq; } |
498d3aa2b
|
2931 2932 2933 |
/* * We drop cfq io contexts lazily, so we may find a dead one. */ |
dbecf3ab4
|
2934 |
static void |
4ac845a2e
|
2935 2936 |
cfq_drop_dead_cic(struct cfq_data *cfqd, struct io_context *ioc, struct cfq_io_context *cic) |
dbecf3ab4
|
2937 |
{ |
4ac845a2e
|
2938 |
unsigned long flags; |
fc46379da
|
2939 |
WARN_ON(!list_empty(&cic->queue_list)); |
bca4b914b
|
2940 |
BUG_ON(cic->key != cfqd_dead_key(cfqd)); |
597bc485d
|
2941 |
|
4ac845a2e
|
2942 |
spin_lock_irqsave(&ioc->lock, flags); |
726e99ab8
|
2943 2944 |
BUG_ON(rcu_dereference_check(ioc->ioc_data, lockdep_is_held(&ioc->lock)) == cic); |
597bc485d
|
2945 |
|
80b15c738
|
2946 |
radix_tree_delete(&ioc->radix_root, cfqd->cic_index); |
ffc4e7595
|
2947 |
hlist_del_rcu(&cic->cic_list); |
4ac845a2e
|
2948 2949 2950 |
spin_unlock_irqrestore(&ioc->lock, flags); cfq_cic_free(cic); |
dbecf3ab4
|
2951 |
} |
e2d74ac06
|
2952 |
static struct cfq_io_context * |
4ac845a2e
|
2953 |
cfq_cic_lookup(struct cfq_data *cfqd, struct io_context *ioc) |
e2d74ac06
|
2954 |
{ |
e2d74ac06
|
2955 |
struct cfq_io_context *cic; |
d6de8be71
|
2956 |
unsigned long flags; |
e2d74ac06
|
2957 |
|
91fac317a
|
2958 2959 |
if (unlikely(!ioc)) return NULL; |
d6de8be71
|
2960 |
rcu_read_lock(); |
597bc485d
|
2961 2962 2963 |
/* * we maintain a last-hit cache, to avoid browsing over the tree */ |
4ac845a2e
|
2964 |
cic = rcu_dereference(ioc->ioc_data); |
d6de8be71
|
2965 2966 |
if (cic && cic->key == cfqd) { rcu_read_unlock(); |
597bc485d
|
2967 |
return cic; |
d6de8be71
|
2968 |
} |
597bc485d
|
2969 |
|
4ac845a2e
|
2970 |
do { |
80b15c738
|
2971 |
cic = radix_tree_lookup(&ioc->radix_root, cfqd->cic_index); |
4ac845a2e
|
2972 2973 2974 |
rcu_read_unlock(); if (!cic) break; |
bca4b914b
|
2975 |
if (unlikely(cic->key != cfqd)) { |
4ac845a2e
|
2976 |
cfq_drop_dead_cic(cfqd, ioc, cic); |
d6de8be71
|
2977 |
rcu_read_lock(); |
4ac845a2e
|
2978 |
continue; |
dbecf3ab4
|
2979 |
} |
e2d74ac06
|
2980 |
|
d6de8be71
|
2981 |
spin_lock_irqsave(&ioc->lock, flags); |
4ac845a2e
|
2982 |
rcu_assign_pointer(ioc->ioc_data, cic); |
d6de8be71
|
2983 |
spin_unlock_irqrestore(&ioc->lock, flags); |
4ac845a2e
|
2984 2985 |
break; } while (1); |
e2d74ac06
|
2986 |
|
4ac845a2e
|
2987 |
return cic; |
e2d74ac06
|
2988 |
} |
4ac845a2e
|
2989 2990 2991 2992 2993 |
/* * Add cic into ioc, using cfqd as the search key. This enables us to lookup * the process specific cfq io context when entered from the block layer. * Also adds the cic to a per-cfqd list, used when this queue is removed. */ |
febffd618
|
2994 2995 |
static int cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc, struct cfq_io_context *cic, gfp_t gfp_mask) |
e2d74ac06
|
2996 |
{ |
0261d6886
|
2997 |
unsigned long flags; |
4ac845a2e
|
2998 |
int ret; |
e2d74ac06
|
2999 |
|
4ac845a2e
|
3000 3001 3002 3003 |
ret = radix_tree_preload(gfp_mask); if (!ret) { cic->ioc = ioc; cic->key = cfqd; |
e2d74ac06
|
3004 |
|
4ac845a2e
|
3005 3006 |
spin_lock_irqsave(&ioc->lock, flags); ret = radix_tree_insert(&ioc->radix_root, |
80b15c738
|
3007 |
cfqd->cic_index, cic); |
ffc4e7595
|
3008 3009 |
if (!ret) hlist_add_head_rcu(&cic->cic_list, &ioc->cic_list); |
4ac845a2e
|
3010 |
spin_unlock_irqrestore(&ioc->lock, flags); |
e2d74ac06
|
3011 |
|
4ac845a2e
|
3012 3013 3014 3015 3016 3017 3018 |
radix_tree_preload_end(); if (!ret) { spin_lock_irqsave(cfqd->queue->queue_lock, flags); list_add(&cic->queue_list, &cfqd->cic_list); spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); } |
e2d74ac06
|
3019 |
} |
5eb46851d
|
3020 |
if (ret && ret != -EEXIST) |
4ac845a2e
|
3021 3022 |
printk(KERN_ERR "cfq: cic link failed! "); |
fc46379da
|
3023 |
|
4ac845a2e
|
3024 |
return ret; |
e2d74ac06
|
3025 |
} |
1da177e4c
|
3026 3027 3028 |
/* * Setup general io context and cfq io context. There can be several cfq * io contexts per general io context, if this process is doing io to more |
e2d74ac06
|
3029 |
* than one device managed by cfq. |
1da177e4c
|
3030 3031 |
*/ static struct cfq_io_context * |
e2d74ac06
|
3032 |
cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask) |
1da177e4c
|
3033 |
{ |
22e2c507c
|
3034 |
struct io_context *ioc = NULL; |
1da177e4c
|
3035 |
struct cfq_io_context *cic; |
5eb46851d
|
3036 |
int ret; |
1da177e4c
|
3037 |
|
22e2c507c
|
3038 |
might_sleep_if(gfp_mask & __GFP_WAIT); |
1da177e4c
|
3039 |
|
b5deef901
|
3040 |
ioc = get_io_context(gfp_mask, cfqd->queue->node); |
1da177e4c
|
3041 3042 |
if (!ioc) return NULL; |
5eb46851d
|
3043 |
retry: |
4ac845a2e
|
3044 |
cic = cfq_cic_lookup(cfqd, ioc); |
e2d74ac06
|
3045 3046 |
if (cic) goto out; |
1da177e4c
|
3047 |
|
e2d74ac06
|
3048 3049 3050 |
cic = cfq_alloc_io_context(cfqd, gfp_mask); if (cic == NULL) goto err; |
1da177e4c
|
3051 |
|
5eb46851d
|
3052 3053 3054 3055 3056 3057 |
ret = cfq_cic_link(cfqd, ioc, cic, gfp_mask); if (ret == -EEXIST) { /* someone has linked cic to ioc already */ cfq_cic_free(cic); goto retry; } else if (ret) |
4ac845a2e
|
3058 |
goto err_free; |
1da177e4c
|
3059 |
out: |
fc46379da
|
3060 3061 3062 |
smp_read_barrier_depends(); if (unlikely(ioc->ioprio_changed)) cfq_ioc_set_ioprio(ioc); |
24610333d
|
3063 3064 3065 3066 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED if (unlikely(ioc->cgroup_changed)) cfq_ioc_set_cgroup(ioc); #endif |
1da177e4c
|
3067 |
return cic; |
4ac845a2e
|
3068 3069 |
err_free: cfq_cic_free(cic); |
1da177e4c
|
3070 3071 3072 3073 |
err: put_io_context(ioc); return NULL; } |
22e2c507c
|
3074 |
static void |
383cd7213
|
3075 |
__cfq_update_io_thinktime(struct cfq_ttime *ttime, unsigned long slice_idle) |
1da177e4c
|
3076 |
{ |
383cd7213
|
3077 3078 |
unsigned long elapsed = jiffies - ttime->last_end_request; elapsed = min(elapsed, 2UL * slice_idle); |
db3b5848e
|
3079 |
|
383cd7213
|
3080 3081 3082 3083 3084 3085 3086 3087 3088 |
ttime->ttime_samples = (7*ttime->ttime_samples + 256) / 8; ttime->ttime_total = (7*ttime->ttime_total + 256*elapsed) / 8; ttime->ttime_mean = (ttime->ttime_total + 128) / ttime->ttime_samples; } static void cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_queue *cfqq, struct cfq_io_context *cic) { |
f5f2b6ceb
|
3089 |
if (cfq_cfqq_sync(cfqq)) { |
383cd7213
|
3090 |
__cfq_update_io_thinktime(&cic->ttime, cfqd->cfq_slice_idle); |
f5f2b6ceb
|
3091 3092 3093 |
__cfq_update_io_thinktime(&cfqq->service_tree->ttime, cfqd->cfq_slice_idle); } |
7700fc4f6
|
3094 3095 3096 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED __cfq_update_io_thinktime(&cfqq->cfqg->ttime, cfqd->cfq_group_idle); #endif |
22e2c507c
|
3097 |
} |
1da177e4c
|
3098 |
|
206dc69b3
|
3099 |
static void |
b2c18e1e0
|
3100 |
cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_queue *cfqq, |
6d048f531
|
3101 |
struct request *rq) |
206dc69b3
|
3102 |
{ |
3dde36dde
|
3103 |
sector_t sdist = 0; |
41647e7a9
|
3104 |
sector_t n_sec = blk_rq_sectors(rq); |
3dde36dde
|
3105 3106 3107 3108 3109 3110 |
if (cfqq->last_request_pos) { if (cfqq->last_request_pos < blk_rq_pos(rq)) sdist = blk_rq_pos(rq) - cfqq->last_request_pos; else sdist = cfqq->last_request_pos - blk_rq_pos(rq); } |
206dc69b3
|
3111 |
|
3dde36dde
|
3112 |
cfqq->seek_history <<= 1; |
41647e7a9
|
3113 3114 3115 3116 |
if (blk_queue_nonrot(cfqd->queue)) cfqq->seek_history |= (n_sec < CFQQ_SECT_THR_NONROT); else cfqq->seek_history |= (sdist > CFQQ_SEEK_THR); |
206dc69b3
|
3117 |
} |
1da177e4c
|
3118 |
|
22e2c507c
|
3119 3120 3121 3122 3123 3124 3125 3126 |
/* * Disable idle window if the process thinks too long or seeks so much that * it doesn't matter */ static void cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq, struct cfq_io_context *cic) { |
7b679138b
|
3127 |
int old_idle, enable_idle; |
1be92f2fc
|
3128 |
|
0871714e0
|
3129 3130 3131 3132 |
/* * Don't idle for async or idle io prio class */ if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq)) |
1be92f2fc
|
3133 |
return; |
c265a7f41
|
3134 |
enable_idle = old_idle = cfq_cfqq_idle_window(cfqq); |
1da177e4c
|
3135 |
|
76280aff1
|
3136 3137 |
if (cfqq->queued[0] + cfqq->queued[1] >= 4) cfq_mark_cfqq_deep(cfqq); |
749ef9f84
|
3138 3139 3140 |
if (cfqq->next_rq && (cfqq->next_rq->cmd_flags & REQ_NOIDLE)) enable_idle = 0; else if (!atomic_read(&cic->ioc->nr_tasks) || !cfqd->cfq_slice_idle || |
3dde36dde
|
3141 |
(!cfq_cfqq_deep(cfqq) && CFQQ_SEEKY(cfqq))) |
22e2c507c
|
3142 |
enable_idle = 0; |
383cd7213
|
3143 3144 |
else if (sample_valid(cic->ttime.ttime_samples)) { if (cic->ttime.ttime_mean > cfqd->cfq_slice_idle) |
22e2c507c
|
3145 3146 3147 |
enable_idle = 0; else enable_idle = 1; |
1da177e4c
|
3148 |
} |
7b679138b
|
3149 3150 3151 3152 3153 3154 3155 |
if (old_idle != enable_idle) { cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle); if (enable_idle) cfq_mark_cfqq_idle_window(cfqq); else cfq_clear_cfqq_idle_window(cfqq); } |
22e2c507c
|
3156 |
} |
1da177e4c
|
3157 |
|
22e2c507c
|
3158 3159 3160 3161 |
/* * Check if new_cfqq should preempt the currently active queue. Return 0 for * no or if we aren't sure, a 1 will cause a preempt. */ |
a6151c3a5
|
3162 |
static bool |
22e2c507c
|
3163 |
cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq, |
5e7053747
|
3164 |
struct request *rq) |
22e2c507c
|
3165 |
{ |
6d048f531
|
3166 |
struct cfq_queue *cfqq; |
22e2c507c
|
3167 |
|
6d048f531
|
3168 3169 |
cfqq = cfqd->active_queue; if (!cfqq) |
a6151c3a5
|
3170 |
return false; |
22e2c507c
|
3171 |
|
6d048f531
|
3172 |
if (cfq_class_idle(new_cfqq)) |
a6151c3a5
|
3173 |
return false; |
22e2c507c
|
3174 3175 |
if (cfq_class_idle(cfqq)) |
a6151c3a5
|
3176 |
return true; |
1e3335de0
|
3177 |
|
22e2c507c
|
3178 |
/* |
875feb63b
|
3179 3180 3181 3182 3183 3184 |
* Don't allow a non-RT request to preempt an ongoing RT cfqq timeslice. */ if (cfq_class_rt(cfqq) && !cfq_class_rt(new_cfqq)) return false; /* |
374f84ac3
|
3185 3186 3187 |
* if the new request is sync, but the currently running queue is * not, let the sync request have priority. */ |
5e7053747
|
3188 |
if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq)) |
a6151c3a5
|
3189 |
return true; |
1e3335de0
|
3190 |
|
8682e1f15
|
3191 3192 3193 3194 3195 3196 3197 3198 3199 3200 3201 3202 |
if (new_cfqq->cfqg != cfqq->cfqg) return false; if (cfq_slice_used(cfqq)) return true; /* Allow preemption only if we are idling on sync-noidle tree */ if (cfqd->serving_type == SYNC_NOIDLE_WORKLOAD && cfqq_type(new_cfqq) == SYNC_NOIDLE_WORKLOAD && new_cfqq->service_tree->count == 2 && RB_EMPTY_ROOT(&cfqq->sort_list)) return true; |
374f84ac3
|
3203 |
/* |
b53d1ed73
|
3204 3205 3206 |
* So both queues are sync. Let the new request get disk time if * it's a metadata request and the current queue is doing regular IO. */ |
65299a3b7
|
3207 |
if ((rq->cmd_flags & REQ_PRIO) && !cfqq->prio_pending) |
b53d1ed73
|
3208 3209 3210 |
return true; /* |
3a9a3f6cc
|
3211 3212 3213 |
* Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice. */ if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq)) |
a6151c3a5
|
3214 |
return true; |
3a9a3f6cc
|
3215 |
|
d2d59e18a
|
3216 3217 3218 |
/* An idle queue should not be idle now for some reason */ if (RB_EMPTY_ROOT(&cfqq->sort_list) && !cfq_should_idle(cfqd, cfqq)) return true; |
1e3335de0
|
3219 |
if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq)) |
a6151c3a5
|
3220 |
return false; |
1e3335de0
|
3221 3222 3223 3224 3225 |
/* * if this request is as-good as one we would expect from the * current cfqq, let it preempt */ |
e9ce335df
|
3226 |
if (cfq_rq_close(cfqd, cfqq, rq)) |
a6151c3a5
|
3227 |
return true; |
1e3335de0
|
3228 |
|
a6151c3a5
|
3229 |
return false; |
22e2c507c
|
3230 3231 3232 3233 3234 3235 3236 3237 |
} /* * cfqq preempts the active queue. if we allowed preempt with no slice left, * let it have half of its nominal slice. */ static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq) { |
f8ae6e3eb
|
3238 |
struct cfq_queue *old_cfqq = cfqd->active_queue; |
7b679138b
|
3239 |
cfq_log_cfqq(cfqd, cfqq, "preempt"); |
e5ff082e8
|
3240 |
cfq_slice_expired(cfqd, 1); |
22e2c507c
|
3241 |
|
bf5722567
|
3242 |
/* |
f8ae6e3eb
|
3243 3244 3245 3246 3247 3248 3249 |
* workload type is changed, don't save slice, otherwise preempt * doesn't happen */ if (cfqq_type(old_cfqq) != cfqq_type(cfqq)) cfqq->cfqg->saved_workload_slice = 0; /* |
bf5722567
|
3250 3251 3252 3253 |
* Put the new queue at the front of the of the current list, * so we know that it will be selected next. */ BUG_ON(!cfq_cfqq_on_rr(cfqq)); |
edd75ffd9
|
3254 3255 |
cfq_service_tree_add(cfqd, cfqq, 1); |
eda5e0c91
|
3256 |
|
62a37f6ba
|
3257 3258 |
cfqq->slice_end = 0; cfq_mark_cfqq_slice_new(cfqq); |
22e2c507c
|
3259 3260 3261 |
} /* |
5e7053747
|
3262 |
* Called when a new fs request (rq) is added (to cfqq). Check if there's |
22e2c507c
|
3263 3264 3265 |
* something we should do about it */ static void |
5e7053747
|
3266 3267 |
cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq, struct request *rq) |
22e2c507c
|
3268 |
{ |
5e7053747
|
3269 |
struct cfq_io_context *cic = RQ_CIC(rq); |
12e9fddd6
|
3270 |
|
45333d5a3
|
3271 |
cfqd->rq_queued++; |
65299a3b7
|
3272 3273 |
if (rq->cmd_flags & REQ_PRIO) cfqq->prio_pending++; |
374f84ac3
|
3274 |
|
383cd7213
|
3275 |
cfq_update_io_thinktime(cfqd, cfqq, cic); |
b2c18e1e0
|
3276 |
cfq_update_io_seektime(cfqd, cfqq, rq); |
9c2c38a12
|
3277 |
cfq_update_idle_window(cfqd, cfqq, cic); |
b2c18e1e0
|
3278 |
cfqq->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq); |
22e2c507c
|
3279 3280 3281 |
if (cfqq == cfqd->active_queue) { /* |
b029195dd
|
3282 3283 3284 |
* Remember that we saw a request from this process, but * don't start queuing just yet. Otherwise we risk seeing lots * of tiny requests, because we disrupt the normal plugging |
d6ceb25e8
|
3285 3286 |
* and merging. If the request is already larger than a single * page, let it rip immediately. For that case we assume that |
2d8707229
|
3287 3288 3289 |
* merging is already done. Ditto for a busy system that * has other work pending, don't risk delaying until the * idle timer unplug to continue working. |
22e2c507c
|
3290 |
*/ |
d6ceb25e8
|
3291 |
if (cfq_cfqq_wait_request(cfqq)) { |
2d8707229
|
3292 3293 |
if (blk_rq_bytes(rq) > PAGE_CACHE_SIZE || cfqd->busy_queues > 1) { |
812df48d1
|
3294 |
cfq_del_timer(cfqd, cfqq); |
554554f60
|
3295 |
cfq_clear_cfqq_wait_request(cfqq); |
24ecfbe27
|
3296 |
__blk_run_queue(cfqd->queue); |
a11cdaa7a
|
3297 |
} else { |
e98ef89b3
|
3298 |
cfq_blkiocg_update_idle_time_stats( |
a11cdaa7a
|
3299 |
&cfqq->cfqg->blkg); |
bf7919371
|
3300 |
cfq_mark_cfqq_must_dispatch(cfqq); |
a11cdaa7a
|
3301 |
} |
d6ceb25e8
|
3302 |
} |
5e7053747
|
3303 |
} else if (cfq_should_preempt(cfqd, cfqq, rq)) { |
22e2c507c
|
3304 3305 3306 |
/* * not the active queue - expire current slice if it is * idle and has expired it's mean thinktime or this new queue |
3a9a3f6cc
|
3307 3308 |
* has some old slice time left and is of higher priority or * this new queue is RT and the current one is BE |
22e2c507c
|
3309 3310 |
*/ cfq_preempt_queue(cfqd, cfqq); |
24ecfbe27
|
3311 |
__blk_run_queue(cfqd->queue); |
22e2c507c
|
3312 |
} |
1da177e4c
|
3313 |
} |
165125e1e
|
3314 |
static void cfq_insert_request(struct request_queue *q, struct request *rq) |
1da177e4c
|
3315 |
{ |
b4878f245
|
3316 |
struct cfq_data *cfqd = q->elevator->elevator_data; |
5e7053747
|
3317 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
22e2c507c
|
3318 |
|
7b679138b
|
3319 |
cfq_log_cfqq(cfqd, cfqq, "insert_request"); |
fd0928df9
|
3320 |
cfq_init_prio_data(cfqq, RQ_CIC(rq)->ioc); |
1da177e4c
|
3321 |
|
30996f40b
|
3322 |
rq_set_fifo_time(rq, jiffies + cfqd->cfq_fifo_expire[rq_is_sync(rq)]); |
22e2c507c
|
3323 |
list_add_tail(&rq->queuelist, &cfqq->fifo); |
aa6f6a3de
|
3324 |
cfq_add_rq_rb(rq); |
e98ef89b3
|
3325 |
cfq_blkiocg_update_io_add_stats(&(RQ_CFQG(rq))->blkg, |
cdc1184cf
|
3326 3327 |
&cfqd->serving_group->blkg, rq_data_dir(rq), rq_is_sync(rq)); |
5e7053747
|
3328 |
cfq_rq_enqueued(cfqd, cfqq, rq); |
1da177e4c
|
3329 |
} |
45333d5a3
|
3330 3331 3332 3333 3334 3335 |
/* * Update hw_tag based on peak queue depth over 50 samples under * sufficient load. */ static void cfq_update_hw_tag(struct cfq_data *cfqd) { |
1a1238a7d
|
3336 |
struct cfq_queue *cfqq = cfqd->active_queue; |
53c583d22
|
3337 3338 |
if (cfqd->rq_in_driver > cfqd->hw_tag_est_depth) cfqd->hw_tag_est_depth = cfqd->rq_in_driver; |
e459dd08f
|
3339 3340 3341 |
if (cfqd->hw_tag == 1) return; |
45333d5a3
|
3342 3343 |
if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN && |
53c583d22
|
3344 |
cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN) |
45333d5a3
|
3345 |
return; |
1a1238a7d
|
3346 3347 3348 3349 3350 3351 3352 |
/* * If active queue hasn't enough requests and can idle, cfq might not * dispatch sufficient requests to hardware. Don't zero hw_tag in this * case */ if (cfqq && cfq_cfqq_idle_window(cfqq) && cfqq->dispatched + cfqq->queued[0] + cfqq->queued[1] < |
53c583d22
|
3353 |
CFQ_HW_QUEUE_MIN && cfqd->rq_in_driver < CFQ_HW_QUEUE_MIN) |
1a1238a7d
|
3354 |
return; |
45333d5a3
|
3355 3356 |
if (cfqd->hw_tag_samples++ < 50) return; |
e459dd08f
|
3357 |
if (cfqd->hw_tag_est_depth >= CFQ_HW_QUEUE_MIN) |
45333d5a3
|
3358 3359 3360 |
cfqd->hw_tag = 1; else cfqd->hw_tag = 0; |
45333d5a3
|
3361 |
} |
7667aa063
|
3362 3363 3364 |
static bool cfq_should_wait_busy(struct cfq_data *cfqd, struct cfq_queue *cfqq) { struct cfq_io_context *cic = cfqd->active_cic; |
02a8f01b5
|
3365 3366 3367 |
/* If the queue already has requests, don't wait */ if (!RB_EMPTY_ROOT(&cfqq->sort_list)) return false; |
7667aa063
|
3368 3369 3370 |
/* If there are other queues in the group, don't wait */ if (cfqq->cfqg->nr_cfqq > 1) return false; |
7700fc4f6
|
3371 3372 3373 |
/* the only queue in the group, but think time is big */ if (cfq_io_thinktime_big(cfqd, &cfqq->cfqg->ttime, true)) return false; |
7667aa063
|
3374 3375 3376 3377 |
if (cfq_slice_used(cfqq)) return true; /* if slice left is less than think time, wait busy */ |
383cd7213
|
3378 3379 |
if (cic && sample_valid(cic->ttime.ttime_samples) && (cfqq->slice_end - jiffies < cic->ttime.ttime_mean)) |
7667aa063
|
3380 3381 3382 3383 3384 3385 3386 3387 3388 3389 3390 3391 3392 3393 |
return true; /* * If think times is less than a jiffy than ttime_mean=0 and above * will not be true. It might happen that slice has not expired yet * but will expire soon (4-5 ns) during select_queue(). To cover the * case where think time is less than a jiffy, mark the queue wait * busy if only 1 jiffy is left in the slice. */ if (cfqq->slice_end - jiffies == 1) return true; return false; } |
165125e1e
|
3394 |
static void cfq_completed_request(struct request_queue *q, struct request *rq) |
1da177e4c
|
3395 |
{ |
5e7053747
|
3396 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
b4878f245
|
3397 |
struct cfq_data *cfqd = cfqq->cfqd; |
5380a101d
|
3398 |
const int sync = rq_is_sync(rq); |
b4878f245
|
3399 |
unsigned long now; |
1da177e4c
|
3400 |
|
b4878f245
|
3401 |
now = jiffies; |
33659ebba
|
3402 3403 |
cfq_log_cfqq(cfqd, cfqq, "complete rqnoidle %d", !!(rq->cmd_flags & REQ_NOIDLE)); |
1da177e4c
|
3404 |
|
45333d5a3
|
3405 |
cfq_update_hw_tag(cfqd); |
53c583d22
|
3406 |
WARN_ON(!cfqd->rq_in_driver); |
6d048f531
|
3407 |
WARN_ON(!cfqq->dispatched); |
53c583d22
|
3408 |
cfqd->rq_in_driver--; |
6d048f531
|
3409 |
cfqq->dispatched--; |
80bdf0c78
|
3410 |
(RQ_CFQG(rq))->dispatched--; |
e98ef89b3
|
3411 3412 3413 |
cfq_blkiocg_update_completion_stats(&cfqq->cfqg->blkg, rq_start_time_ns(rq), rq_io_start_time_ns(rq), rq_data_dir(rq), rq_is_sync(rq)); |
1da177e4c
|
3414 |
|
53c583d22
|
3415 |
cfqd->rq_in_flight[cfq_cfqq_sync(cfqq)]--; |
3ed9a2965
|
3416 |
|
365722bb9
|
3417 |
if (sync) { |
f5f2b6ceb
|
3418 |
struct cfq_rb_root *service_tree; |
383cd7213
|
3419 |
RQ_CIC(rq)->ttime.last_end_request = now; |
f5f2b6ceb
|
3420 3421 3422 3423 3424 3425 3426 |
if (cfq_cfqq_on_rr(cfqq)) service_tree = cfqq->service_tree; else service_tree = service_tree_for(cfqq->cfqg, cfqq_prio(cfqq), cfqq_type(cfqq)); service_tree->ttime.last_end_request = now; |
573412b29
|
3427 3428 |
if (!time_after(rq->start_time + cfqd->cfq_fifo_expire[1], now)) cfqd->last_delayed_sync = now; |
365722bb9
|
3429 |
} |
caaa5f9f0
|
3430 |
|
7700fc4f6
|
3431 3432 3433 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED cfqq->cfqg->ttime.last_end_request = now; #endif |
caaa5f9f0
|
3434 3435 3436 3437 3438 |
/* * If this is the active queue, check if it needs to be expired, * or if we want to idle in case it has no pending requests. */ if (cfqd->active_queue == cfqq) { |
a36e71f99
|
3439 |
const bool cfqq_empty = RB_EMPTY_ROOT(&cfqq->sort_list); |
44f7c1606
|
3440 3441 3442 3443 |
if (cfq_cfqq_slice_new(cfqq)) { cfq_set_prio_slice(cfqd, cfqq); cfq_clear_cfqq_slice_new(cfqq); } |
f75edf2dc
|
3444 3445 |
/* |
7667aa063
|
3446 3447 |
* Should we wait for next request to come in before we expire * the queue. |
f75edf2dc
|
3448 |
*/ |
7667aa063
|
3449 |
if (cfq_should_wait_busy(cfqd, cfqq)) { |
80bdf0c78
|
3450 3451 3452 3453 |
unsigned long extend_sl = cfqd->cfq_slice_idle; if (!cfqd->cfq_slice_idle) extend_sl = cfqd->cfq_group_idle; cfqq->slice_end = jiffies + extend_sl; |
f75edf2dc
|
3454 |
cfq_mark_cfqq_wait_busy(cfqq); |
b1ffe737f
|
3455 |
cfq_log_cfqq(cfqd, cfqq, "will busy wait"); |
f75edf2dc
|
3456 |
} |
a36e71f99
|
3457 |
/* |
8e550632c
|
3458 3459 3460 3461 3462 3463 |
* Idling is not enabled on: * - expired queues * - idle-priority queues * - async queues * - queues with still some requests queued * - when there is a close cooperator |
a36e71f99
|
3464 |
*/ |
0871714e0
|
3465 |
if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq)) |
e5ff082e8
|
3466 |
cfq_slice_expired(cfqd, 1); |
8e550632c
|
3467 3468 |
else if (sync && cfqq_empty && !cfq_close_cooperator(cfqd, cfqq)) { |
749ef9f84
|
3469 |
cfq_arm_slice_timer(cfqd); |
8e550632c
|
3470 |
} |
caaa5f9f0
|
3471 |
} |
6d048f531
|
3472 |
|
53c583d22
|
3473 |
if (!cfqd->rq_in_driver) |
23e018a1b
|
3474 |
cfq_schedule_dispatch(cfqd); |
1da177e4c
|
3475 |
} |
89850f7ee
|
3476 |
static inline int __cfq_may_queue(struct cfq_queue *cfqq) |
22e2c507c
|
3477 |
{ |
1b379d8da
|
3478 |
if (cfq_cfqq_wait_request(cfqq) && !cfq_cfqq_must_alloc_slice(cfqq)) { |
3b18152c3
|
3479 |
cfq_mark_cfqq_must_alloc_slice(cfqq); |
22e2c507c
|
3480 |
return ELV_MQUEUE_MUST; |
3b18152c3
|
3481 |
} |
1da177e4c
|
3482 |
|
22e2c507c
|
3483 |
return ELV_MQUEUE_MAY; |
22e2c507c
|
3484 |
} |
165125e1e
|
3485 |
static int cfq_may_queue(struct request_queue *q, int rw) |
22e2c507c
|
3486 3487 3488 |
{ struct cfq_data *cfqd = q->elevator->elevator_data; struct task_struct *tsk = current; |
91fac317a
|
3489 |
struct cfq_io_context *cic; |
22e2c507c
|
3490 3491 3492 3493 3494 3495 3496 3497 |
struct cfq_queue *cfqq; /* * don't force setup of a queue from here, as a call to may_queue * does not necessarily imply that a request actually will be queued. * so just lookup a possibly existing queue, or return 'may queue' * if that fails */ |
4ac845a2e
|
3498 |
cic = cfq_cic_lookup(cfqd, tsk->io_context); |
91fac317a
|
3499 3500 |
if (!cic) return ELV_MQUEUE_MAY; |
b0b78f81a
|
3501 |
cfqq = cic_to_cfqq(cic, rw_is_sync(rw)); |
22e2c507c
|
3502 |
if (cfqq) { |
fd0928df9
|
3503 |
cfq_init_prio_data(cfqq, cic->ioc); |
22e2c507c
|
3504 |
|
89850f7ee
|
3505 |
return __cfq_may_queue(cfqq); |
22e2c507c
|
3506 3507 3508 |
} return ELV_MQUEUE_MAY; |
1da177e4c
|
3509 |
} |
1da177e4c
|
3510 3511 3512 |
/* * queue lock held here */ |
bb37b94c6
|
3513 |
static void cfq_put_request(struct request *rq) |
1da177e4c
|
3514 |
{ |
5e7053747
|
3515 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
1da177e4c
|
3516 |
|
5e7053747
|
3517 |
if (cfqq) { |
22e2c507c
|
3518 |
const int rw = rq_data_dir(rq); |
1da177e4c
|
3519 |
|
22e2c507c
|
3520 3521 |
BUG_ON(!cfqq->allocated[rw]); cfqq->allocated[rw]--; |
1da177e4c
|
3522 |
|
5e7053747
|
3523 |
put_io_context(RQ_CIC(rq)->ioc); |
1da177e4c
|
3524 |
|
c186794db
|
3525 3526 |
rq->elevator_private[0] = NULL; rq->elevator_private[1] = NULL; |
1da177e4c
|
3527 |
|
7f1dc8a2d
|
3528 3529 |
/* Put down rq reference on cfqg */ cfq_put_cfqg(RQ_CFQG(rq)); |
c186794db
|
3530 |
rq->elevator_private[2] = NULL; |
7f1dc8a2d
|
3531 |
|
1da177e4c
|
3532 3533 3534 |
cfq_put_queue(cfqq); } } |
df5fe3e8e
|
3535 3536 3537 3538 3539 3540 |
static struct cfq_queue * cfq_merge_cfqqs(struct cfq_data *cfqd, struct cfq_io_context *cic, struct cfq_queue *cfqq) { cfq_log_cfqq(cfqd, cfqq, "merging with queue %p", cfqq->new_cfqq); cic_set_cfqq(cic, cfqq->new_cfqq, 1); |
b3b6d0408
|
3541 |
cfq_mark_cfqq_coop(cfqq->new_cfqq); |
df5fe3e8e
|
3542 3543 3544 |
cfq_put_queue(cfqq); return cic_to_cfqq(cic, 1); } |
e6c5bc737
|
3545 3546 3547 3548 3549 3550 3551 3552 |
/* * Returns NULL if a new cfqq should be allocated, or the old cfqq if this * was the last process referring to said cfqq. */ static struct cfq_queue * split_cfqq(struct cfq_io_context *cic, struct cfq_queue *cfqq) { if (cfqq_process_refs(cfqq) == 1) { |
e6c5bc737
|
3553 3554 |
cfqq->pid = current->pid; cfq_clear_cfqq_coop(cfqq); |
ae54abed6
|
3555 |
cfq_clear_cfqq_split_coop(cfqq); |
e6c5bc737
|
3556 3557 3558 3559 |
return cfqq; } cic_set_cfqq(cic, NULL, 1); |
d02a2c077
|
3560 3561 |
cfq_put_cooperator(cfqq); |
e6c5bc737
|
3562 3563 3564 |
cfq_put_queue(cfqq); return NULL; } |
1da177e4c
|
3565 |
/* |
22e2c507c
|
3566 |
* Allocate cfq data structures associated with this request. |
1da177e4c
|
3567 |
*/ |
22e2c507c
|
3568 |
static int |
165125e1e
|
3569 |
cfq_set_request(struct request_queue *q, struct request *rq, gfp_t gfp_mask) |
1da177e4c
|
3570 3571 3572 3573 |
{ struct cfq_data *cfqd = q->elevator->elevator_data; struct cfq_io_context *cic; const int rw = rq_data_dir(rq); |
a6151c3a5
|
3574 |
const bool is_sync = rq_is_sync(rq); |
22e2c507c
|
3575 |
struct cfq_queue *cfqq; |
1da177e4c
|
3576 3577 3578 |
unsigned long flags; might_sleep_if(gfp_mask & __GFP_WAIT); |
e2d74ac06
|
3579 |
cic = cfq_get_io_context(cfqd, gfp_mask); |
22e2c507c
|
3580 |
|
1da177e4c
|
3581 |
spin_lock_irqsave(q->queue_lock, flags); |
22e2c507c
|
3582 3583 |
if (!cic) goto queue_fail; |
e6c5bc737
|
3584 |
new_queue: |
91fac317a
|
3585 |
cfqq = cic_to_cfqq(cic, is_sync); |
32f2e807a
|
3586 |
if (!cfqq || cfqq == &cfqd->oom_cfqq) { |
fd0928df9
|
3587 |
cfqq = cfq_get_queue(cfqd, is_sync, cic->ioc, gfp_mask); |
91fac317a
|
3588 |
cic_set_cfqq(cic, cfqq, is_sync); |
df5fe3e8e
|
3589 3590 |
} else { /* |
e6c5bc737
|
3591 3592 |
* If the queue was seeky for too long, break it apart. */ |
ae54abed6
|
3593 |
if (cfq_cfqq_coop(cfqq) && cfq_cfqq_split_coop(cfqq)) { |
e6c5bc737
|
3594 3595 3596 3597 3598 3599 3600 |
cfq_log_cfqq(cfqd, cfqq, "breaking apart cfqq"); cfqq = split_cfqq(cic, cfqq); if (!cfqq) goto new_queue; } /* |
df5fe3e8e
|
3601 3602 3603 3604 3605 3606 3607 |
* Check to see if this queue is scheduled to merge with * another, closely cooperating queue. The merging of * queues happens here as it must be done in process context. * The reference on new_cfqq was taken in merge_cfqqs. */ if (cfqq->new_cfqq) cfqq = cfq_merge_cfqqs(cfqd, cic, cfqq); |
91fac317a
|
3608 |
} |
1da177e4c
|
3609 3610 |
cfqq->allocated[rw]++; |
1da177e4c
|
3611 |
|
6fae9c251
|
3612 |
cfqq->ref++; |
c186794db
|
3613 3614 3615 |
rq->elevator_private[0] = cic; rq->elevator_private[1] = cfqq; rq->elevator_private[2] = cfq_ref_get_cfqg(cfqq->cfqg); |
93803e014
|
3616 |
spin_unlock_irqrestore(q->queue_lock, flags); |
5e7053747
|
3617 |
return 0; |
1da177e4c
|
3618 |
|
22e2c507c
|
3619 |
queue_fail: |
23e018a1b
|
3620 |
cfq_schedule_dispatch(cfqd); |
1da177e4c
|
3621 |
spin_unlock_irqrestore(q->queue_lock, flags); |
7b679138b
|
3622 |
cfq_log(cfqd, "set_request fail"); |
1da177e4c
|
3623 3624 |
return 1; } |
65f27f384
|
3625 |
static void cfq_kick_queue(struct work_struct *work) |
22e2c507c
|
3626 |
{ |
65f27f384
|
3627 |
struct cfq_data *cfqd = |
23e018a1b
|
3628 |
container_of(work, struct cfq_data, unplug_work); |
165125e1e
|
3629 |
struct request_queue *q = cfqd->queue; |
22e2c507c
|
3630 |
|
40bb54d19
|
3631 |
spin_lock_irq(q->queue_lock); |
24ecfbe27
|
3632 |
__blk_run_queue(cfqd->queue); |
40bb54d19
|
3633 |
spin_unlock_irq(q->queue_lock); |
22e2c507c
|
3634 3635 3636 3637 3638 3639 3640 3641 3642 3643 |
} /* * Timer running if the active_queue is currently idling inside its time slice */ static void cfq_idle_slice_timer(unsigned long data) { struct cfq_data *cfqd = (struct cfq_data *) data; struct cfq_queue *cfqq; unsigned long flags; |
3c6bd2f87
|
3644 |
int timed_out = 1; |
22e2c507c
|
3645 |
|
7b679138b
|
3646 |
cfq_log(cfqd, "idle timer fired"); |
22e2c507c
|
3647 |
spin_lock_irqsave(cfqd->queue->queue_lock, flags); |
fe094d98e
|
3648 3649 |
cfqq = cfqd->active_queue; if (cfqq) { |
3c6bd2f87
|
3650 |
timed_out = 0; |
22e2c507c
|
3651 |
/* |
b029195dd
|
3652 3653 3654 3655 3656 3657 |
* We saw a request before the queue expired, let it through */ if (cfq_cfqq_must_dispatch(cfqq)) goto out_kick; /* |
22e2c507c
|
3658 3659 |
* expired */ |
44f7c1606
|
3660 |
if (cfq_slice_used(cfqq)) |
22e2c507c
|
3661 3662 3663 3664 3665 3666 |
goto expire; /* * only expire and reinvoke request handler, if there are * other queues with pending requests */ |
caaa5f9f0
|
3667 |
if (!cfqd->busy_queues) |
22e2c507c
|
3668 |
goto out_cont; |
22e2c507c
|
3669 3670 3671 3672 |
/* * not expired and it has a request pending, let it dispatch */ |
75e50984f
|
3673 |
if (!RB_EMPTY_ROOT(&cfqq->sort_list)) |
22e2c507c
|
3674 |
goto out_kick; |
76280aff1
|
3675 3676 3677 3678 3679 |
/* * Queue depth flag is reset only when the idle didn't succeed */ cfq_clear_cfqq_deep(cfqq); |
22e2c507c
|
3680 3681 |
} expire: |
e5ff082e8
|
3682 |
cfq_slice_expired(cfqd, timed_out); |
22e2c507c
|
3683 |
out_kick: |
23e018a1b
|
3684 |
cfq_schedule_dispatch(cfqd); |
22e2c507c
|
3685 3686 3687 |
out_cont: spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); } |
3b18152c3
|
3688 3689 3690 |
static void cfq_shutdown_timer_wq(struct cfq_data *cfqd) { del_timer_sync(&cfqd->idle_slice_timer); |
23e018a1b
|
3691 |
cancel_work_sync(&cfqd->unplug_work); |
3b18152c3
|
3692 |
} |
22e2c507c
|
3693 |
|
c2dea2d1f
|
3694 3695 3696 3697 3698 3699 3700 3701 3702 |
static void cfq_put_async_queues(struct cfq_data *cfqd) { int i; for (i = 0; i < IOPRIO_BE_NR; i++) { if (cfqd->async_cfqq[0][i]) cfq_put_queue(cfqd->async_cfqq[0][i]); if (cfqd->async_cfqq[1][i]) cfq_put_queue(cfqd->async_cfqq[1][i]); |
c2dea2d1f
|
3703 |
} |
2389d1ef1
|
3704 3705 3706 |
if (cfqd->async_idle_cfqq) cfq_put_queue(cfqd->async_idle_cfqq); |
c2dea2d1f
|
3707 |
} |
b374d18a4
|
3708 |
static void cfq_exit_queue(struct elevator_queue *e) |
1da177e4c
|
3709 |
{ |
22e2c507c
|
3710 |
struct cfq_data *cfqd = e->elevator_data; |
165125e1e
|
3711 |
struct request_queue *q = cfqd->queue; |
56edf7d75
|
3712 |
bool wait = false; |
22e2c507c
|
3713 |
|
3b18152c3
|
3714 |
cfq_shutdown_timer_wq(cfqd); |
e2d74ac06
|
3715 |
|
d9ff41879
|
3716 |
spin_lock_irq(q->queue_lock); |
e2d74ac06
|
3717 |
|
d9ff41879
|
3718 |
if (cfqd->active_queue) |
e5ff082e8
|
3719 |
__cfq_slice_expired(cfqd, cfqd->active_queue, 0); |
e2d74ac06
|
3720 3721 |
while (!list_empty(&cfqd->cic_list)) { |
d9ff41879
|
3722 3723 3724 |
struct cfq_io_context *cic = list_entry(cfqd->cic_list.next, struct cfq_io_context, queue_list); |
89850f7ee
|
3725 3726 |
__cfq_exit_single_io_context(cfqd, cic); |
d9ff41879
|
3727 |
} |
e2d74ac06
|
3728 |
|
c2dea2d1f
|
3729 |
cfq_put_async_queues(cfqd); |
b1c357696
|
3730 |
cfq_release_cfq_groups(cfqd); |
56edf7d75
|
3731 3732 3733 3734 3735 3736 3737 |
/* * If there are groups which we could not unlink from blkcg list, * wait for a rcu period for them to be freed. */ if (cfqd->nr_blkcg_linked_grps) wait = true; |
15c31be4d
|
3738 |
|
d9ff41879
|
3739 |
spin_unlock_irq(q->queue_lock); |
a90d742e4
|
3740 3741 |
cfq_shutdown_timer_wq(cfqd); |
80b15c738
|
3742 3743 3744 |
spin_lock(&cic_index_lock); ida_remove(&cic_index_ida, cfqd->cic_index); spin_unlock(&cic_index_lock); |
56edf7d75
|
3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 |
/* * Wait for cfqg->blkg->key accessors to exit their grace periods. * Do this wait only if there are other unlinked groups out * there. This can happen if cgroup deletion path claimed the * responsibility of cleaning up a group before queue cleanup code * get to the group. * * Do not call synchronize_rcu() unconditionally as there are drivers * which create/delete request queue hundreds of times during scan/boot * and synchronize_rcu() can take significant time and slow down boot. */ if (wait) synchronize_rcu(); |
2abae55f5
|
3758 3759 3760 3761 3762 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED /* Free up per cpu stats for root group */ free_percpu(cfqd->root_group.blkg.stats_cpu); #endif |
56edf7d75
|
3763 |
kfree(cfqd); |
1da177e4c
|
3764 |
} |
80b15c738
|
3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775 3776 3777 3778 3779 3780 3781 |
static int cfq_alloc_cic_index(void) { int index, error; do { if (!ida_pre_get(&cic_index_ida, GFP_KERNEL)) return -ENOMEM; spin_lock(&cic_index_lock); error = ida_get_new(&cic_index_ida, &index); spin_unlock(&cic_index_lock); if (error && error != -EAGAIN) return error; } while (error); return index; } |
165125e1e
|
3782 |
static void *cfq_init_queue(struct request_queue *q) |
1da177e4c
|
3783 3784 |
{ struct cfq_data *cfqd; |
718eee057
|
3785 |
int i, j; |
cdb16e8f7
|
3786 |
struct cfq_group *cfqg; |
615f0259e
|
3787 |
struct cfq_rb_root *st; |
1da177e4c
|
3788 |
|
80b15c738
|
3789 3790 3791 |
i = cfq_alloc_cic_index(); if (i < 0) return NULL; |
94f6030ca
|
3792 |
cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node); |
1547010e6
|
3793 3794 3795 3796 |
if (!cfqd) { spin_lock(&cic_index_lock); ida_remove(&cic_index_ida, i); spin_unlock(&cic_index_lock); |
bc1c11697
|
3797 |
return NULL; |
1547010e6
|
3798 |
} |
1da177e4c
|
3799 |
|
30d7b9448
|
3800 3801 3802 3803 |
/* * Don't need take queue_lock in the routine, since we are * initializing the ioscheduler, and nobody is using cfqd */ |
80b15c738
|
3804 |
cfqd->cic_index = i; |
1fa8f6d68
|
3805 3806 |
/* Init root service tree */ cfqd->grp_service_tree = CFQ_RB_ROOT; |
cdb16e8f7
|
3807 3808 |
/* Init root group */ cfqg = &cfqd->root_group; |
615f0259e
|
3809 3810 |
for_each_cfqg_st(cfqg, i, j, st) *st = CFQ_RB_ROOT; |
1fa8f6d68
|
3811 |
RB_CLEAR_NODE(&cfqg->rb_node); |
26a2ac009
|
3812 |
|
25bc6b077
|
3813 3814 |
/* Give preference to root group over other groups */ cfqg->weight = 2*BLKIO_WEIGHT_DEFAULT; |
25fb5169d
|
3815 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED |
b1c357696
|
3816 |
/* |
56edf7d75
|
3817 3818 3819 3820 3821 |
* Set root group reference to 2. One reference will be dropped when * all groups on cfqd->cfqg_list are being deleted during queue exit. * Other reference will remain there as we don't want to delete this * group as it is statically allocated and gets destroyed when * throtl_data goes away. |
b1c357696
|
3822 |
*/ |
56edf7d75
|
3823 |
cfqg->ref = 2; |
5624a4e44
|
3824 3825 3826 |
if (blkio_alloc_blkg_stats(&cfqg->blkg)) { kfree(cfqg); |
2984ff38c
|
3827 3828 3829 3830 |
spin_lock(&cic_index_lock); ida_remove(&cic_index_ida, cfqd->cic_index); spin_unlock(&cic_index_lock); |
5624a4e44
|
3831 3832 3833 |
kfree(cfqd); return NULL; } |
dcf097b24
|
3834 |
rcu_read_lock(); |
5624a4e44
|
3835 |
|
e98ef89b3
|
3836 3837 |
cfq_blkiocg_add_blkio_group(&blkio_root_cgroup, &cfqg->blkg, (void *)cfqd, 0); |
dcf097b24
|
3838 |
rcu_read_unlock(); |
56edf7d75
|
3839 3840 3841 3842 |
cfqd->nr_blkcg_linked_grps++; /* Add group on cfqd->cfqg_list */ hlist_add_head(&cfqg->cfqd_node, &cfqd->cfqg_list); |
25fb5169d
|
3843 |
#endif |
26a2ac009
|
3844 3845 3846 3847 3848 3849 3850 |
/* * Not strictly needed (since RB_ROOT just clears the node and we * zeroed cfqd on alloc), but better be safe in case someone decides * to add magic to the rb code */ for (i = 0; i < CFQ_PRIO_LISTS; i++) cfqd->prio_trees[i] = RB_ROOT; |
6118b70b3
|
3851 3852 3853 3854 3855 3856 |
/* * Our fallback cfqq if cfq_find_alloc_queue() runs into OOM issues. * Grab a permanent reference to it, so that the normal code flow * will not attempt to free it. */ cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0); |
30d7b9448
|
3857 |
cfqd->oom_cfqq.ref++; |
cdb16e8f7
|
3858 |
cfq_link_cfqq_cfqg(&cfqd->oom_cfqq, &cfqd->root_group); |
6118b70b3
|
3859 |
|
d9ff41879
|
3860 |
INIT_LIST_HEAD(&cfqd->cic_list); |
1da177e4c
|
3861 |
|
1da177e4c
|
3862 |
cfqd->queue = q; |
1da177e4c
|
3863 |
|
22e2c507c
|
3864 3865 3866 |
init_timer(&cfqd->idle_slice_timer); cfqd->idle_slice_timer.function = cfq_idle_slice_timer; cfqd->idle_slice_timer.data = (unsigned long) cfqd; |
23e018a1b
|
3867 |
INIT_WORK(&cfqd->unplug_work, cfq_kick_queue); |
22e2c507c
|
3868 |
|
1da177e4c
|
3869 |
cfqd->cfq_quantum = cfq_quantum; |
22e2c507c
|
3870 3871 |
cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0]; cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1]; |
1da177e4c
|
3872 3873 |
cfqd->cfq_back_max = cfq_back_max; cfqd->cfq_back_penalty = cfq_back_penalty; |
22e2c507c
|
3874 3875 3876 3877 |
cfqd->cfq_slice[0] = cfq_slice_async; cfqd->cfq_slice[1] = cfq_slice_sync; cfqd->cfq_slice_async_rq = cfq_slice_async_rq; cfqd->cfq_slice_idle = cfq_slice_idle; |
80bdf0c78
|
3878 |
cfqd->cfq_group_idle = cfq_group_idle; |
963b72fc6
|
3879 |
cfqd->cfq_latency = 1; |
e459dd08f
|
3880 |
cfqd->hw_tag = -1; |
edc71131c
|
3881 3882 3883 3884 |
/* * we optimistically start assuming sync ops weren't delayed in last * second, in order to have larger depth for async operations. */ |
573412b29
|
3885 |
cfqd->last_delayed_sync = jiffies - HZ; |
bc1c11697
|
3886 |
return cfqd; |
1da177e4c
|
3887 3888 3889 3890 |
} static void cfq_slab_kill(void) { |
d6de8be71
|
3891 3892 3893 3894 |
/* * Caller already ensured that pending RCU callbacks are completed, * so we should have no busy allocations at this point. */ |
1da177e4c
|
3895 3896 3897 3898 3899 3900 3901 3902 |
if (cfq_pool) kmem_cache_destroy(cfq_pool); if (cfq_ioc_pool) kmem_cache_destroy(cfq_ioc_pool); } static int __init cfq_slab_setup(void) { |
0a31bd5f2
|
3903 |
cfq_pool = KMEM_CACHE(cfq_queue, 0); |
1da177e4c
|
3904 3905 |
if (!cfq_pool) goto fail; |
34e6bbf23
|
3906 |
cfq_ioc_pool = KMEM_CACHE(cfq_io_context, 0); |
1da177e4c
|
3907 3908 3909 3910 3911 3912 3913 3914 |
if (!cfq_ioc_pool) goto fail; return 0; fail: cfq_slab_kill(); return -ENOMEM; } |
1da177e4c
|
3915 3916 3917 |
/* * sysfs parts below --> */ |
1da177e4c
|
3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928 3929 3930 3931 3932 |
static ssize_t cfq_var_show(unsigned int var, char *page) { return sprintf(page, "%d ", var); } static ssize_t cfq_var_store(unsigned int *var, const char *page, size_t count) { char *p = (char *) page; *var = simple_strtoul(p, &p, 10); return count; } |
1da177e4c
|
3933 |
#define SHOW_FUNCTION(__FUNC, __VAR, __CONV) \ |
b374d18a4
|
3934 |
static ssize_t __FUNC(struct elevator_queue *e, char *page) \ |
1da177e4c
|
3935 |
{ \ |
3d1ab40f4
|
3936 |
struct cfq_data *cfqd = e->elevator_data; \ |
1da177e4c
|
3937 3938 3939 3940 3941 3942 |
unsigned int __data = __VAR; \ if (__CONV) \ __data = jiffies_to_msecs(__data); \ return cfq_var_show(__data, (page)); \ } SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0); |
22e2c507c
|
3943 3944 |
SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1); SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1); |
e572ec7e4
|
3945 3946 |
SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0); SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0); |
22e2c507c
|
3947 |
SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1); |
80bdf0c78
|
3948 |
SHOW_FUNCTION(cfq_group_idle_show, cfqd->cfq_group_idle, 1); |
22e2c507c
|
3949 3950 3951 |
SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1); SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1); SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0); |
963b72fc6
|
3952 |
SHOW_FUNCTION(cfq_low_latency_show, cfqd->cfq_latency, 0); |
1da177e4c
|
3953 3954 3955 |
#undef SHOW_FUNCTION #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV) \ |
b374d18a4
|
3956 |
static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \ |
1da177e4c
|
3957 |
{ \ |
3d1ab40f4
|
3958 |
struct cfq_data *cfqd = e->elevator_data; \ |
1da177e4c
|
3959 3960 3961 3962 3963 3964 3965 3966 3967 3968 3969 3970 3971 |
unsigned int __data; \ int ret = cfq_var_store(&__data, (page), count); \ if (__data < (MIN)) \ __data = (MIN); \ else if (__data > (MAX)) \ __data = (MAX); \ if (__CONV) \ *(__PTR) = msecs_to_jiffies(__data); \ else \ *(__PTR) = __data; \ return ret; \ } STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0); |
fe094d98e
|
3972 3973 3974 3975 |
STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1, UINT_MAX, 1); STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1, UINT_MAX, 1); |
e572ec7e4
|
3976 |
STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0); |
fe094d98e
|
3977 3978 |
STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1, UINT_MAX, 0); |
22e2c507c
|
3979 |
STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1); |
80bdf0c78
|
3980 |
STORE_FUNCTION(cfq_group_idle_store, &cfqd->cfq_group_idle, 0, UINT_MAX, 1); |
22e2c507c
|
3981 3982 |
STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1); STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1); |
fe094d98e
|
3983 3984 |
STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1, UINT_MAX, 0); |
963b72fc6
|
3985 |
STORE_FUNCTION(cfq_low_latency_store, &cfqd->cfq_latency, 0, 1, 0); |
1da177e4c
|
3986 |
#undef STORE_FUNCTION |
e572ec7e4
|
3987 3988 3989 3990 3991 |
#define CFQ_ATTR(name) \ __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store) static struct elv_fs_entry cfq_attrs[] = { CFQ_ATTR(quantum), |
e572ec7e4
|
3992 3993 3994 3995 3996 3997 3998 3999 |
CFQ_ATTR(fifo_expire_sync), CFQ_ATTR(fifo_expire_async), CFQ_ATTR(back_seek_max), CFQ_ATTR(back_seek_penalty), CFQ_ATTR(slice_sync), CFQ_ATTR(slice_async), CFQ_ATTR(slice_async_rq), CFQ_ATTR(slice_idle), |
80bdf0c78
|
4000 |
CFQ_ATTR(group_idle), |
963b72fc6
|
4001 |
CFQ_ATTR(low_latency), |
e572ec7e4
|
4002 |
__ATTR_NULL |
1da177e4c
|
4003 |
}; |
1da177e4c
|
4004 4005 4006 4007 4008 |
static struct elevator_type iosched_cfq = { .ops = { .elevator_merge_fn = cfq_merge, .elevator_merged_fn = cfq_merged_request, .elevator_merge_req_fn = cfq_merged_requests, |
da7752650
|
4009 |
.elevator_allow_merge_fn = cfq_allow_merge, |
812d40264
|
4010 |
.elevator_bio_merged_fn = cfq_bio_merged, |
b4878f245
|
4011 |
.elevator_dispatch_fn = cfq_dispatch_requests, |
1da177e4c
|
4012 |
.elevator_add_req_fn = cfq_insert_request, |
b4878f245
|
4013 |
.elevator_activate_req_fn = cfq_activate_request, |
1da177e4c
|
4014 |
.elevator_deactivate_req_fn = cfq_deactivate_request, |
1da177e4c
|
4015 |
.elevator_completed_req_fn = cfq_completed_request, |
21183b07e
|
4016 4017 |
.elevator_former_req_fn = elv_rb_former_request, .elevator_latter_req_fn = elv_rb_latter_request, |
1da177e4c
|
4018 4019 4020 4021 4022 |
.elevator_set_req_fn = cfq_set_request, .elevator_put_req_fn = cfq_put_request, .elevator_may_queue_fn = cfq_may_queue, .elevator_init_fn = cfq_init_queue, .elevator_exit_fn = cfq_exit_queue, |
fc46379da
|
4023 |
.trim = cfq_free_io_context, |
1da177e4c
|
4024 |
}, |
3d1ab40f4
|
4025 |
.elevator_attrs = cfq_attrs, |
1da177e4c
|
4026 4027 4028 |
.elevator_name = "cfq", .elevator_owner = THIS_MODULE, }; |
3e2520668
|
4029 4030 4031 4032 4033 4034 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED static struct blkio_policy_type blkio_policy_cfq = { .ops = { .blkio_unlink_group_fn = cfq_unlink_blkio_group, .blkio_update_group_weight_fn = cfq_update_blkio_group_weight, }, |
062a644d6
|
4035 |
.plid = BLKIO_POLICY_PROP, |
3e2520668
|
4036 4037 4038 4039 |
}; #else static struct blkio_policy_type blkio_policy_cfq; #endif |
1da177e4c
|
4040 4041 |
static int __init cfq_init(void) { |
22e2c507c
|
4042 4043 4044 4045 4046 4047 4048 |
/* * could be 0 on HZ < 1000 setups */ if (!cfq_slice_async) cfq_slice_async = 1; if (!cfq_slice_idle) cfq_slice_idle = 1; |
80bdf0c78
|
4049 4050 4051 4052 4053 4054 |
#ifdef CONFIG_CFQ_GROUP_IOSCHED if (!cfq_group_idle) cfq_group_idle = 1; #else cfq_group_idle = 0; #endif |
1da177e4c
|
4055 4056 |
if (cfq_slab_setup()) return -ENOMEM; |
2fdd82bd8
|
4057 |
elv_register(&iosched_cfq); |
3e2520668
|
4058 |
blkio_policy_register(&blkio_policy_cfq); |
1da177e4c
|
4059 |
|
2fdd82bd8
|
4060 |
return 0; |
1da177e4c
|
4061 4062 4063 4064 |
} static void __exit cfq_exit(void) { |
6e9a4738c
|
4065 |
DECLARE_COMPLETION_ONSTACK(all_gone); |
3e2520668
|
4066 |
blkio_policy_unregister(&blkio_policy_cfq); |
1da177e4c
|
4067 |
elv_unregister(&iosched_cfq); |
334e94de9
|
4068 |
ioc_gone = &all_gone; |
fba822722
|
4069 4070 |
/* ioc_gone's update must be visible before reading ioc_count */ smp_wmb(); |
d6de8be71
|
4071 4072 4073 4074 4075 |
/* * this also protects us from entering cfq_slab_kill() with * pending RCU callbacks */ |
245b2e70e
|
4076 |
if (elv_ioc_count_read(cfq_ioc_count)) |
9a11b4ed0
|
4077 |
wait_for_completion(&all_gone); |
80b15c738
|
4078 |
ida_destroy(&cic_index_ida); |
83521d3eb
|
4079 |
cfq_slab_kill(); |
1da177e4c
|
4080 4081 4082 4083 4084 4085 4086 4087 |
} module_init(cfq_init); module_exit(cfq_exit); MODULE_AUTHOR("Jens Axboe"); MODULE_LICENSE("GPL"); MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler"); |