Blame view
block/cfq-iosched.c
53.1 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> |
1cc9be68e
|
10 11 |
#include <linux/blkdev.h> #include <linux/elevator.h> |
1da177e4c
|
12 13 |
#include <linux/hash.h> #include <linux/rbtree.h> |
22e2c507c
|
14 |
#include <linux/ioprio.h> |
1da177e4c
|
15 16 17 18 |
/* * tunables */ |
64100099e
|
19 |
static const int cfq_quantum = 4; /* max queue in one round of service */ |
64100099e
|
20 21 22 |
static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 }; static const int cfq_back_max = 16 * 1024; /* maximum backwards seek, in KiB */ static const int cfq_back_penalty = 2; /* penalty of a backwards seek */ |
1da177e4c
|
23 |
|
64100099e
|
24 |
static const int cfq_slice_sync = HZ / 10; |
3b18152c3
|
25 |
static int cfq_slice_async = HZ / 25; |
64100099e
|
26 |
static const int cfq_slice_async_rq = 2; |
caaa5f9f0
|
27 |
static int cfq_slice_idle = HZ / 125; |
22e2c507c
|
28 29 30 31 32 |
#define CFQ_IDLE_GRACE (HZ / 10) #define CFQ_SLICE_SCALE (5) #define CFQ_KEY_ASYNC (0) |
22e2c507c
|
33 |
|
1da177e4c
|
34 35 36 37 38 39 |
/* * for the hash of cfqq inside the cfqd */ #define CFQ_QHASH_SHIFT 6 #define CFQ_QHASH_ENTRIES (1 << CFQ_QHASH_SHIFT) #define list_entry_qhash(entry) hlist_entry((entry), struct cfq_queue, cfq_hash) |
1da177e4c
|
40 |
#define list_entry_cfqq(ptr) list_entry((ptr), struct cfq_queue, cfq_list) |
5e7053747
|
41 42 |
#define RQ_CIC(rq) ((struct cfq_io_context*)(rq)->elevator_private) #define RQ_CFQQ(rq) ((rq)->elevator_private2) |
1da177e4c
|
43 |
|
e18b890bb
|
44 45 |
static struct kmem_cache *cfq_pool; static struct kmem_cache *cfq_ioc_pool; |
1da177e4c
|
46 |
|
4050cf167
|
47 |
static DEFINE_PER_CPU(unsigned long, ioc_count); |
334e94de9
|
48 |
static struct completion *ioc_gone; |
22e2c507c
|
49 50 |
#define CFQ_PRIO_LISTS IOPRIO_BE_NR #define cfq_class_idle(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE) |
22e2c507c
|
51 |
#define cfq_class_rt(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_RT) |
3b18152c3
|
52 53 54 55 56 57 58 59 60 61 |
#define ASYNC (0) #define SYNC (1) #define cfq_cfqq_dispatched(cfqq) \ ((cfqq)->on_dispatch[ASYNC] + (cfqq)->on_dispatch[SYNC]) #define cfq_cfqq_class_sync(cfqq) ((cfqq)->key != CFQ_KEY_ASYNC) #define cfq_cfqq_sync(cfqq) \ (cfq_cfqq_class_sync(cfqq) || (cfqq)->on_dispatch[SYNC]) |
22e2c507c
|
62 |
|
206dc69b3
|
63 |
#define sample_valid(samples) ((samples) > 80) |
22e2c507c
|
64 65 66 |
/* * Per block device queue structure */ |
1da177e4c
|
67 |
struct cfq_data { |
22e2c507c
|
68 69 70 71 72 73 74 75 76 77 78 79 |
request_queue_t *queue; /* * rr list of queues with requests and the count of them */ struct list_head rr_list[CFQ_PRIO_LISTS]; struct list_head busy_rr; struct list_head cur_rr; struct list_head idle_rr; unsigned int busy_queues; /* |
22e2c507c
|
80 81 |
* cfqq lookup hash */ |
1da177e4c
|
82 |
struct hlist_head *cfq_hash; |
1da177e4c
|
83 |
|
22e2c507c
|
84 |
int rq_in_driver; |
25776e359
|
85 |
int hw_tag; |
1da177e4c
|
86 |
|
22e2c507c
|
87 |
/* |
22e2c507c
|
88 89 90 91 |
* idle window management */ struct timer_list idle_slice_timer; struct work_struct unplug_work; |
1da177e4c
|
92 |
|
22e2c507c
|
93 94 95 96 97 98 |
struct cfq_queue *active_queue; struct cfq_io_context *active_cic; int cur_prio, cur_end_prio; unsigned int dispatch_slice; struct timer_list idle_class_timer; |
1da177e4c
|
99 100 |
sector_t last_sector; |
22e2c507c
|
101 |
unsigned long last_end_request; |
1da177e4c
|
102 |
|
1da177e4c
|
103 104 105 106 |
/* * tunables, see top of file */ unsigned int cfq_quantum; |
22e2c507c
|
107 |
unsigned int cfq_fifo_expire[2]; |
1da177e4c
|
108 109 |
unsigned int cfq_back_penalty; unsigned int cfq_back_max; |
22e2c507c
|
110 111 112 |
unsigned int cfq_slice[2]; unsigned int cfq_slice_async_rq; unsigned int cfq_slice_idle; |
d9ff41879
|
113 114 |
struct list_head cic_list; |
1da177e4c
|
115 |
}; |
22e2c507c
|
116 117 118 |
/* * Per process-grouping structure */ |
1da177e4c
|
119 120 121 122 123 |
struct cfq_queue { /* reference count */ atomic_t ref; /* parent cfq_data */ struct cfq_data *cfqd; |
22e2c507c
|
124 |
/* cfqq lookup hash */ |
1da177e4c
|
125 126 |
struct hlist_node cfq_hash; /* hash key */ |
22e2c507c
|
127 |
unsigned int key; |
981a79730
|
128 |
/* member of the rr/busy/cur/idle cfqd list */ |
1da177e4c
|
129 130 131 132 |
struct list_head cfq_list; /* sorted list of pending requests */ struct rb_root sort_list; /* if fifo isn't expired, next request to serve */ |
5e7053747
|
133 |
struct request *next_rq; |
1da177e4c
|
134 135 136 137 |
/* requests queued in sort_list */ int queued[2]; /* currently allocated requests */ int allocated[2]; |
374f84ac3
|
138 139 |
/* pending metadata requests */ int meta_pending; |
1da177e4c
|
140 |
/* fifo list of requests in sort_list */ |
22e2c507c
|
141 |
struct list_head fifo; |
1da177e4c
|
142 |
|
22e2c507c
|
143 |
unsigned long slice_end; |
99f9628ab
|
144 |
unsigned long service_last; |
c5b680f3b
|
145 |
long slice_resid; |
1da177e4c
|
146 |
|
3b18152c3
|
147 148 |
/* number of requests that are on the dispatch list */ int on_dispatch[2]; |
22e2c507c
|
149 150 151 152 |
/* io prio of this group */ unsigned short ioprio, org_ioprio; unsigned short ioprio_class, org_ioprio_class; |
3b18152c3
|
153 154 |
/* various state flags, see below */ unsigned int flags; |
1da177e4c
|
155 |
}; |
3b18152c3
|
156 |
enum cfqq_state_flags { |
b0b8d7494
|
157 158 159 160 161 162 163 164 165 |
CFQ_CFQQ_FLAG_on_rr = 0, /* on round-robin busy list */ CFQ_CFQQ_FLAG_wait_request, /* waiting for a request */ CFQ_CFQQ_FLAG_must_alloc, /* must be allowed rq alloc */ CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */ CFQ_CFQQ_FLAG_must_dispatch, /* must dispatch, even if expired */ 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 */ CFQ_CFQQ_FLAG_queue_new, /* queue never been serviced */ |
44f7c1606
|
166 |
CFQ_CFQQ_FLAG_slice_new, /* no requests dispatched in slice */ |
3b18152c3
|
167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 |
}; #define CFQ_CFQQ_FNS(name) \ static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq) \ { \ cfqq->flags |= (1 << CFQ_CFQQ_FLAG_##name); \ } \ static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq) \ { \ cfqq->flags &= ~(1 << CFQ_CFQQ_FLAG_##name); \ } \ static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq) \ { \ return (cfqq->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0; \ } CFQ_CFQQ_FNS(on_rr); CFQ_CFQQ_FNS(wait_request); CFQ_CFQQ_FNS(must_alloc); CFQ_CFQQ_FNS(must_alloc_slice); CFQ_CFQQ_FNS(must_dispatch); CFQ_CFQQ_FNS(fifo_expire); CFQ_CFQQ_FNS(idle_window); CFQ_CFQQ_FNS(prio_changed); |
53b03744e
|
191 |
CFQ_CFQQ_FNS(queue_new); |
44f7c1606
|
192 |
CFQ_CFQQ_FNS(slice_new); |
3b18152c3
|
193 |
#undef CFQ_CFQQ_FNS |
3b18152c3
|
194 |
static struct cfq_queue *cfq_find_cfq_hash(struct cfq_data *, unsigned int, unsigned short); |
5e7053747
|
195 |
static void cfq_dispatch_insert(request_queue_t *, struct request *); |
6f325a134
|
196 |
static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, unsigned int key, struct task_struct *tsk, gfp_t gfp_mask); |
1da177e4c
|
197 |
|
1da177e4c
|
198 |
/* |
99f95e528
|
199 200 201 202 203 |
* scheduler run of queue, if there are requests pending and no one in the * driver that will restart queueing */ static inline void cfq_schedule_dispatch(struct cfq_data *cfqd) { |
7b14e3b52
|
204 |
if (cfqd->busy_queues) |
99f95e528
|
205 206 207 208 209 210 |
kblockd_schedule_work(&cfqd->unplug_work); } static int cfq_queue_empty(request_queue_t *q) { struct cfq_data *cfqd = q->elevator->elevator_data; |
b4878f245
|
211 |
return !cfqd->busy_queues; |
99f95e528
|
212 |
} |
7749a8d42
|
213 |
static inline pid_t cfq_queue_pid(struct task_struct *task, int rw, int is_sync) |
206dc69b3
|
214 |
{ |
7749a8d42
|
215 216 217 218 |
/* * Use the per-process queue, for read requests and syncronous writes */ if (!(rw & REQ_RW) || is_sync) |
206dc69b3
|
219 220 221 222 |
return task->pid; return CFQ_KEY_ASYNC; } |
1da177e4c
|
223 |
/* |
44f7c1606
|
224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 |
* 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. */ static inline int cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq) { const int base_slice = cfqd->cfq_slice[cfq_cfqq_sync(cfqq)]; WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR); return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - cfqq->ioprio)); } static inline void cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq) { cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies; |
c5b680f3b
|
242 243 244 245 246 247 248 249 |
cfqq->slice_end += cfqq->slice_resid; /* * Don't carry over residual for more than one slice, we only want * to slightly correct the fairness. Carrying over forever would * easily introduce oscillations. */ cfqq->slice_resid = 0; |
44f7c1606
|
250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 |
} /* * 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. */ static inline int cfq_slice_used(struct cfq_queue *cfqq) { if (cfq_cfqq_slice_new(cfqq)) return 0; if (time_before(jiffies, cfqq->slice_end)) return 0; return 1; } /* |
5e7053747
|
268 |
* Lifted from AS - choose which of rq1 and rq2 that is best served now. |
1da177e4c
|
269 |
* We choose the request that is closest to the head right now. Distance |
e8a99053e
|
270 |
* behind the head is penalized and only allowed to a certain extent. |
1da177e4c
|
271 |
*/ |
5e7053747
|
272 273 |
static struct request * cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2) |
1da177e4c
|
274 275 |
{ sector_t last, s1, s2, d1 = 0, d2 = 0; |
1da177e4c
|
276 |
unsigned long back_max; |
e8a99053e
|
277 278 279 |
#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
|
280 |
|
5e7053747
|
281 282 283 284 |
if (rq1 == NULL || rq1 == rq2) return rq2; if (rq2 == NULL) return rq1; |
9c2c38a12
|
285 |
|
5e7053747
|
286 287 288 289 |
if (rq_is_sync(rq1) && !rq_is_sync(rq2)) return rq1; else if (rq_is_sync(rq2) && !rq_is_sync(rq1)) return rq2; |
374f84ac3
|
290 291 292 293 |
if (rq_is_meta(rq1) && !rq_is_meta(rq2)) return rq1; else if (rq_is_meta(rq2) && !rq_is_meta(rq1)) return rq2; |
1da177e4c
|
294 |
|
5e7053747
|
295 296 |
s1 = rq1->sector; s2 = rq2->sector; |
1da177e4c
|
297 298 |
last = cfqd->last_sector; |
1da177e4c
|
299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 |
/* * 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
|
314 |
wrap |= CFQ_RQ1_WRAP; |
1da177e4c
|
315 316 317 318 319 320 |
if (s2 >= last) d2 = s2 - last; else if (s2 + back_max >= last) d2 = (last - s2) * cfqd->cfq_back_penalty; else |
e8a99053e
|
321 |
wrap |= CFQ_RQ2_WRAP; |
1da177e4c
|
322 323 |
/* Found required data */ |
e8a99053e
|
324 325 326 327 328 329 |
/* * By doing switch() on the bit mask "wrap" we avoid having to * check two variables for all permutations: --> faster! */ switch (wrap) { |
5e7053747
|
330 |
case 0: /* common case for CFQ: rq1 and rq2 not wrapped */ |
e8a99053e
|
331 |
if (d1 < d2) |
5e7053747
|
332 |
return rq1; |
e8a99053e
|
333 |
else if (d2 < d1) |
5e7053747
|
334 |
return rq2; |
e8a99053e
|
335 336 |
else { if (s1 >= s2) |
5e7053747
|
337 |
return rq1; |
e8a99053e
|
338 |
else |
5e7053747
|
339 |
return rq2; |
e8a99053e
|
340 |
} |
1da177e4c
|
341 |
|
e8a99053e
|
342 |
case CFQ_RQ2_WRAP: |
5e7053747
|
343 |
return rq1; |
e8a99053e
|
344 |
case CFQ_RQ1_WRAP: |
5e7053747
|
345 346 |
return rq2; case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */ |
e8a99053e
|
347 348 349 350 351 352 353 354 |
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
|
355 |
return rq1; |
1da177e4c
|
356 |
else |
5e7053747
|
357 |
return rq2; |
1da177e4c
|
358 359 360 361 362 363 |
} } /* * would be nice to take fifo expire time into account as well */ |
5e7053747
|
364 365 366 |
static struct request * cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq, struct request *last) |
1da177e4c
|
367 |
{ |
21183b07e
|
368 369 |
struct rb_node *rbnext = rb_next(&last->rb_node); struct rb_node *rbprev = rb_prev(&last->rb_node); |
5e7053747
|
370 |
struct request *next = NULL, *prev = NULL; |
1da177e4c
|
371 |
|
21183b07e
|
372 |
BUG_ON(RB_EMPTY_NODE(&last->rb_node)); |
1da177e4c
|
373 374 |
if (rbprev) |
5e7053747
|
375 |
prev = rb_entry_rq(rbprev); |
1da177e4c
|
376 |
|
21183b07e
|
377 |
if (rbnext) |
5e7053747
|
378 |
next = rb_entry_rq(rbnext); |
21183b07e
|
379 380 381 |
else { rbnext = rb_first(&cfqq->sort_list); if (rbnext && rbnext != &last->rb_node) |
5e7053747
|
382 |
next = rb_entry_rq(rbnext); |
21183b07e
|
383 |
} |
1da177e4c
|
384 |
|
21183b07e
|
385 |
return cfq_choose_req(cfqd, next, prev); |
1da177e4c
|
386 |
} |
22e2c507c
|
387 |
static void cfq_resort_rr_list(struct cfq_queue *cfqq, int preempted) |
1da177e4c
|
388 |
{ |
22e2c507c
|
389 |
struct cfq_data *cfqd = cfqq->cfqd; |
99f9628ab
|
390 391 |
struct list_head *list, *n; struct cfq_queue *__cfqq; |
1da177e4c
|
392 |
|
98e41c7df
|
393 394 395 396 397 |
/* * Resorting requires the cfqq to be on the RR list already. */ if (!cfq_cfqq_on_rr(cfqq)) return; |
1da177e4c
|
398 |
|
22e2c507c
|
399 |
list_del(&cfqq->cfq_list); |
1da177e4c
|
400 |
|
22e2c507c
|
401 402 403 404 405 406 407 408 409 410 411 412 |
if (cfq_class_rt(cfqq)) list = &cfqd->cur_rr; else if (cfq_class_idle(cfqq)) list = &cfqd->idle_rr; else { /* * if cfqq has requests in flight, don't allow it to be * found in cfq_set_active_queue before it has finished them. * this is done to increase fairness between a process that * has lots of io pending vs one that only generates one * sporadically or synchronously */ |
3b18152c3
|
413 |
if (cfq_cfqq_dispatched(cfqq)) |
22e2c507c
|
414 415 416 |
list = &cfqd->busy_rr; else list = &cfqd->rr_list[cfqq->ioprio]; |
1da177e4c
|
417 |
} |
53b03744e
|
418 |
if (preempted || cfq_cfqq_queue_new(cfqq)) { |
99f9628ab
|
419 420 421 422 423 424 |
/* * If this queue was preempted or is new (never been serviced), * let it be added first for fairness but beind other new * queues. */ n = list; |
53b03744e
|
425 426 427 428 |
while (n->next != list) { __cfqq = list_entry_cfqq(n->next); if (!cfq_cfqq_queue_new(__cfqq)) break; |
1da177e4c
|
429 |
|
53b03744e
|
430 431 |
n = n->next; } |
99f9628ab
|
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 |
list_add_tail(&cfqq->cfq_list, n); } else if (!cfq_cfqq_class_sync(cfqq)) { /* * async queue always goes to the end. this wont be overly * unfair to writes, as the sort of the sync queue wont be * allowed to pass the async queue again. */ list_add_tail(&cfqq->cfq_list, list); } else { /* * sort by last service, but don't cross a new or async * queue. we don't cross a new queue because it hasn't been * service before, and we don't cross an async queue because * it gets added to the end on expire. */ n = list; while ((n = n->prev) != list) { struct cfq_queue *__cfqq = list_entry_cfqq(n); |
1da177e4c
|
450 |
|
99f9628ab
|
451 452 453 454 455 456 |
if (!cfq_cfqq_class_sync(cfqq) || !__cfqq->service_last) break; if (time_before(__cfqq->service_last, cfqq->service_last)) break; } list_add(&cfqq->cfq_list, n); |
1da177e4c
|
457 |
} |
1da177e4c
|
458 459 460 461 |
} /* * add to busy list of queues for service, trying to be fair in ordering |
22e2c507c
|
462 |
* the pending list according to last request service |
1da177e4c
|
463 464 |
*/ static inline void |
b4878f245
|
465 |
cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
1da177e4c
|
466 |
{ |
3b18152c3
|
467 468 |
BUG_ON(cfq_cfqq_on_rr(cfqq)); cfq_mark_cfqq_on_rr(cfqq); |
1da177e4c
|
469 |
cfqd->busy_queues++; |
b4878f245
|
470 |
cfq_resort_rr_list(cfqq, 0); |
1da177e4c
|
471 472 473 474 475 |
} static inline void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) { |
3b18152c3
|
476 477 |
BUG_ON(!cfq_cfqq_on_rr(cfqq)); cfq_clear_cfqq_on_rr(cfqq); |
981a79730
|
478 |
list_del_init(&cfqq->cfq_list); |
1da177e4c
|
479 480 481 482 483 484 485 486 |
BUG_ON(!cfqd->busy_queues); cfqd->busy_queues--; } /* * rb tree support functions */ |
5e7053747
|
487 |
static inline void cfq_del_rq_rb(struct request *rq) |
1da177e4c
|
488 |
{ |
5e7053747
|
489 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
b4878f245
|
490 |
struct cfq_data *cfqd = cfqq->cfqd; |
5e7053747
|
491 |
const int sync = rq_is_sync(rq); |
1da177e4c
|
492 |
|
b4878f245
|
493 494 |
BUG_ON(!cfqq->queued[sync]); cfqq->queued[sync]--; |
1da177e4c
|
495 |
|
5e7053747
|
496 |
elv_rb_del(&cfqq->sort_list, rq); |
1da177e4c
|
497 |
|
dd67d0515
|
498 |
if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list)) |
b4878f245
|
499 |
cfq_del_cfqq_rr(cfqd, cfqq); |
1da177e4c
|
500 |
} |
5e7053747
|
501 |
static void cfq_add_rq_rb(struct request *rq) |
1da177e4c
|
502 |
{ |
5e7053747
|
503 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
1da177e4c
|
504 |
struct cfq_data *cfqd = cfqq->cfqd; |
21183b07e
|
505 |
struct request *__alias; |
1da177e4c
|
506 |
|
5380a101d
|
507 |
cfqq->queued[rq_is_sync(rq)]++; |
1da177e4c
|
508 509 510 511 512 |
/* * looks a little odd, but the first insert might return an alias. * if that happens, put the alias on the dispatch list */ |
21183b07e
|
513 |
while ((__alias = elv_rb_add(&cfqq->sort_list, rq)) != NULL) |
5e7053747
|
514 |
cfq_dispatch_insert(cfqd->queue, __alias); |
5fccbf61b
|
515 516 517 |
if (!cfq_cfqq_on_rr(cfqq)) cfq_add_cfqq_rr(cfqd, cfqq); |
5044eed48
|
518 519 520 521 522 523 |
/* * check if this request is a better next-serve candidate */ cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq); BUG_ON(!cfqq->next_rq); |
1da177e4c
|
524 525 526 |
} static inline void |
5e7053747
|
527 |
cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq) |
1da177e4c
|
528 |
{ |
5380a101d
|
529 530 |
elv_rb_del(&cfqq->sort_list, rq); cfqq->queued[rq_is_sync(rq)]--; |
5e7053747
|
531 |
cfq_add_rq_rb(rq); |
1da177e4c
|
532 |
} |
206dc69b3
|
533 534 |
static struct request * cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio) |
1da177e4c
|
535 |
{ |
206dc69b3
|
536 |
struct task_struct *tsk = current; |
7749a8d42
|
537 |
pid_t key = cfq_queue_pid(tsk, bio_data_dir(bio), bio_sync(bio)); |
206dc69b3
|
538 |
struct cfq_queue *cfqq; |
1da177e4c
|
539 |
|
206dc69b3
|
540 |
cfqq = cfq_find_cfq_hash(cfqd, key, tsk->ioprio); |
89850f7ee
|
541 542 |
if (cfqq) { sector_t sector = bio->bi_sector + bio_sectors(bio); |
21183b07e
|
543 |
return elv_rb_find(&cfqq->sort_list, sector); |
89850f7ee
|
544 |
} |
1da177e4c
|
545 |
|
1da177e4c
|
546 547 |
return NULL; } |
b4878f245
|
548 |
static void cfq_activate_request(request_queue_t *q, struct request *rq) |
1da177e4c
|
549 |
{ |
22e2c507c
|
550 |
struct cfq_data *cfqd = q->elevator->elevator_data; |
3b18152c3
|
551 |
|
b4878f245
|
552 |
cfqd->rq_in_driver++; |
25776e359
|
553 554 555 556 557 558 559 560 561 |
/* * If the depth is larger 1, it really could be queueing. But lets * make the mark a little higher - idling could still be good for * low queueing, and a low queueing number could also just indicate * a SCSI mid layer like behaviour where limit+1 is often seen. */ if (!cfqd->hw_tag && cfqd->rq_in_driver > 4) cfqd->hw_tag = 1; |
1da177e4c
|
562 |
} |
b4878f245
|
563 |
static void cfq_deactivate_request(request_queue_t *q, struct request *rq) |
1da177e4c
|
564 |
{ |
b4878f245
|
565 566 567 568 |
struct cfq_data *cfqd = q->elevator->elevator_data; WARN_ON(!cfqd->rq_in_driver); cfqd->rq_in_driver--; |
1da177e4c
|
569 |
} |
b4878f245
|
570 |
static void cfq_remove_request(struct request *rq) |
1da177e4c
|
571 |
{ |
5e7053747
|
572 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
21183b07e
|
573 |
|
5e7053747
|
574 575 |
if (cfqq->next_rq == rq) cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq); |
1da177e4c
|
576 |
|
b4878f245
|
577 |
list_del_init(&rq->queuelist); |
5e7053747
|
578 |
cfq_del_rq_rb(rq); |
374f84ac3
|
579 580 581 582 583 |
if (rq_is_meta(rq)) { WARN_ON(!cfqq->meta_pending); cfqq->meta_pending--; } |
1da177e4c
|
584 585 586 587 588 589 590 |
} static int cfq_merge(request_queue_t *q, struct request **req, struct bio *bio) { struct cfq_data *cfqd = q->elevator->elevator_data; struct request *__rq; |
1da177e4c
|
591 |
|
206dc69b3
|
592 |
__rq = cfq_find_rq_fmerge(cfqd, bio); |
22e2c507c
|
593 |
if (__rq && elv_rq_merge_ok(__rq, bio)) { |
9817064b6
|
594 595 |
*req = __rq; return ELEVATOR_FRONT_MERGE; |
1da177e4c
|
596 597 598 |
} return ELEVATOR_NO_MERGE; |
1da177e4c
|
599 |
} |
21183b07e
|
600 601 |
static void cfq_merged_request(request_queue_t *q, struct request *req, int type) |
1da177e4c
|
602 |
{ |
21183b07e
|
603 |
if (type == ELEVATOR_FRONT_MERGE) { |
5e7053747
|
604 |
struct cfq_queue *cfqq = RQ_CFQQ(req); |
1da177e4c
|
605 |
|
5e7053747
|
606 |
cfq_reposition_rq_rb(cfqq, req); |
1da177e4c
|
607 |
} |
1da177e4c
|
608 609 610 611 612 613 |
} static void cfq_merged_requests(request_queue_t *q, struct request *rq, struct request *next) { |
22e2c507c
|
614 615 616 617 618 619 |
/* * reposition in fifo if next is older than rq */ if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) && time_before(next->start_time, rq->start_time)) list_move(&rq->queuelist, &next->queuelist); |
b4878f245
|
620 |
cfq_remove_request(next); |
22e2c507c
|
621 |
} |
da7752650
|
622 623 624 625 626 627 628 629 630 |
static int cfq_allow_merge(request_queue_t *q, struct request *rq, struct bio *bio) { struct cfq_data *cfqd = q->elevator->elevator_data; const int rw = bio_data_dir(bio); struct cfq_queue *cfqq; pid_t key; /* |
ec8acb690
|
631 |
* Disallow merge of a sync bio into an async request. |
da7752650
|
632 |
*/ |
ec8acb690
|
633 |
if ((bio_data_dir(bio) == READ || bio_sync(bio)) && !rq_is_sync(rq)) |
da7752650
|
634 635 636 |
return 0; /* |
719d34027
|
637 638 |
* Lookup the cfqq that this bio will be queued with. Allow * merge only if rq is queued there. |
da7752650
|
639 |
*/ |
719d34027
|
640 |
key = cfq_queue_pid(current, rw, bio_sync(bio)); |
da7752650
|
641 |
cfqq = cfq_find_cfq_hash(cfqd, key, current->ioprio); |
719d34027
|
642 643 644 |
if (cfqq == RQ_CFQQ(rq)) return 1; |
da7752650
|
645 |
|
ec8acb690
|
646 |
return 0; |
da7752650
|
647 |
} |
22e2c507c
|
648 649 650 651 652 653 654 655 |
static inline void __cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq) { if (cfqq) { /* * stop potential idle class queues waiting service */ del_timer(&cfqd->idle_class_timer); |
22e2c507c
|
656 |
cfqq->slice_end = 0; |
3b18152c3
|
657 658 |
cfq_clear_cfqq_must_alloc_slice(cfqq); cfq_clear_cfqq_fifo_expire(cfqq); |
44f7c1606
|
659 |
cfq_mark_cfqq_slice_new(cfqq); |
22e2c507c
|
660 661 662 663 664 665 |
} cfqd->active_queue = cfqq; } /* |
7b14e3b52
|
666 667 668 669 |
* 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, |
3c6bd2f87
|
670 |
int preempted, int timed_out) |
7b14e3b52
|
671 |
{ |
7b14e3b52
|
672 673 |
if (cfq_cfqq_wait_request(cfqq)) del_timer(&cfqd->idle_slice_timer); |
7b14e3b52
|
674 675 |
cfq_clear_cfqq_must_dispatch(cfqq); cfq_clear_cfqq_wait_request(cfqq); |
53b03744e
|
676 |
cfq_clear_cfqq_queue_new(cfqq); |
7b14e3b52
|
677 678 679 680 681 |
/* * store what was left of this slice, if the queue idled out * or was preempted */ |
3c6bd2f87
|
682 |
if (timed_out && !cfq_cfqq_slice_new(cfqq)) |
c5b680f3b
|
683 |
cfqq->slice_resid = cfqq->slice_end - jiffies; |
7b14e3b52
|
684 |
|
98e41c7df
|
685 |
cfq_resort_rr_list(cfqq, preempted); |
7b14e3b52
|
686 687 688 689 690 691 692 693 694 695 696 |
if (cfqq == cfqd->active_queue) cfqd->active_queue = NULL; if (cfqd->active_cic) { put_io_context(cfqd->active_cic->ioc); cfqd->active_cic = NULL; } cfqd->dispatch_slice = 0; } |
3c6bd2f87
|
697 698 |
static inline void cfq_slice_expired(struct cfq_data *cfqd, int preempted, int timed_out) |
7b14e3b52
|
699 700 701 702 |
{ struct cfq_queue *cfqq = cfqd->active_queue; if (cfqq) |
3c6bd2f87
|
703 |
__cfq_slice_expired(cfqd, cfqq, preempted, timed_out); |
7b14e3b52
|
704 705 706 |
} /* |
22e2c507c
|
707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 |
* 0 * 0,1 * 0,1,2 * 0,1,2,3 * 0,1,2,3,4 * 0,1,2,3,4,5 * 0,1,2,3,4,5,6 * 0,1,2,3,4,5,6,7 */ static int cfq_get_next_prio_level(struct cfq_data *cfqd) { int prio, wrap; prio = -1; wrap = 0; do { int p; for (p = cfqd->cur_prio; p <= cfqd->cur_end_prio; p++) { if (!list_empty(&cfqd->rr_list[p])) { prio = p; break; } } if (prio != -1) break; cfqd->cur_prio = 0; if (++cfqd->cur_end_prio == CFQ_PRIO_LISTS) { cfqd->cur_end_prio = 0; if (wrap) break; wrap = 1; |
1da177e4c
|
740 |
} |
22e2c507c
|
741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 |
} while (1); if (unlikely(prio == -1)) return -1; BUG_ON(prio >= CFQ_PRIO_LISTS); list_splice_init(&cfqd->rr_list[prio], &cfqd->cur_rr); cfqd->cur_prio = prio + 1; if (cfqd->cur_prio > cfqd->cur_end_prio) { cfqd->cur_end_prio = cfqd->cur_prio; cfqd->cur_prio = 0; } if (cfqd->cur_end_prio == CFQ_PRIO_LISTS) { cfqd->cur_prio = 0; cfqd->cur_end_prio = 0; |
1da177e4c
|
758 |
} |
22e2c507c
|
759 760 |
return prio; } |
3b18152c3
|
761 |
static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd) |
22e2c507c
|
762 |
{ |
7b14e3b52
|
763 |
struct cfq_queue *cfqq = NULL; |
22e2c507c
|
764 |
|
89850f7ee
|
765 766 767 768 769 770 |
if (!list_empty(&cfqd->cur_rr) || cfq_get_next_prio_level(cfqd) != -1) { /* * if current list is non-empty, grab first entry. if it is * empty, get next prio level and grab first entry then if any * are spliced */ |
22e2c507c
|
771 |
cfqq = list_entry_cfqq(cfqd->cur_rr.next); |
89850f7ee
|
772 773 774 775 776 |
} else if (!list_empty(&cfqd->busy_rr)) { /* * If no new queues are available, check if the busy list has * some before falling back to idle io. */ |
e0de0206a
|
777 |
cfqq = list_entry_cfqq(cfqd->busy_rr.next); |
89850f7ee
|
778 779 780 781 782 783 |
} else if (!list_empty(&cfqd->idle_rr)) { /* * if we have idle queues and no rt or be queues had pending * requests, either allow immediate service if the grace period * has passed or arm the idle grace timer */ |
22e2c507c
|
784 785 786 787 788 789 790 791 792 |
unsigned long end = cfqd->last_end_request + CFQ_IDLE_GRACE; if (time_after_eq(jiffies, end)) cfqq = list_entry_cfqq(cfqd->idle_rr.next); else mod_timer(&cfqd->idle_class_timer, end); } __cfq_set_active_queue(cfqd, cfqq); |
3b18152c3
|
793 |
return cfqq; |
22e2c507c
|
794 |
} |
caaa5f9f0
|
795 |
#define CIC_SEEKY(cic) ((cic)->seek_mean > (128 * 1024)) |
1792669cc
|
796 |
static int cfq_arm_slice_timer(struct cfq_data *cfqd) |
22e2c507c
|
797 |
{ |
1792669cc
|
798 |
struct cfq_queue *cfqq = cfqd->active_queue; |
206dc69b3
|
799 |
struct cfq_io_context *cic; |
7b14e3b52
|
800 |
unsigned long sl; |
dd67d0515
|
801 |
WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list)); |
22e2c507c
|
802 803 804 805 806 807 |
/* * idle is disabled, either manually or by past process history */ if (!cfqd->cfq_slice_idle) return 0; |
3b18152c3
|
808 |
if (!cfq_cfqq_idle_window(cfqq)) |
22e2c507c
|
809 810 811 812 |
return 0; /* * task has exited, don't wait */ |
206dc69b3
|
813 814 |
cic = cfqd->active_cic; if (!cic || !cic->ioc->task) |
22e2c507c
|
815 |
return 0; |
3b18152c3
|
816 817 |
cfq_mark_cfqq_must_dispatch(cfqq); cfq_mark_cfqq_wait_request(cfqq); |
22e2c507c
|
818 |
|
7b14e3b52
|
819 |
sl = min(cfqq->slice_end - 1, (unsigned long) cfqd->cfq_slice_idle); |
206dc69b3
|
820 821 822 823 824 825 |
/* * we don't want to idle for seeks, but we do want to allow * fair distribution of slice time for a process doing back-to-back * seeks. so allow a little bit of time for him to submit a new rq */ |
caaa5f9f0
|
826 |
if (sample_valid(cic->seek_samples) && CIC_SEEKY(cic)) |
44eb12312
|
827 |
sl = min(sl, msecs_to_jiffies(2)); |
206dc69b3
|
828 |
|
7b14e3b52
|
829 |
mod_timer(&cfqd->idle_slice_timer, jiffies + sl); |
22e2c507c
|
830 |
return 1; |
1da177e4c
|
831 |
} |
5e7053747
|
832 |
static void cfq_dispatch_insert(request_queue_t *q, struct request *rq) |
1da177e4c
|
833 834 |
{ struct cfq_data *cfqd = q->elevator->elevator_data; |
5e7053747
|
835 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
22e2c507c
|
836 |
|
5380a101d
|
837 838 839 |
cfq_remove_request(rq); cfqq->on_dispatch[rq_is_sync(rq)]++; elv_dispatch_sort(q, rq); |
fd61af038
|
840 841 842 |
rq = list_entry(q->queue_head.prev, struct request, queuelist); cfqd->last_sector = rq->sector + rq->nr_sectors; |
1da177e4c
|
843 844 845 846 847 |
} /* * return expired entry, or NULL to just start from scratch in rbtree */ |
5e7053747
|
848 |
static inline struct request *cfq_check_fifo(struct cfq_queue *cfqq) |
1da177e4c
|
849 850 |
{ struct cfq_data *cfqd = cfqq->cfqd; |
22e2c507c
|
851 |
struct request *rq; |
89850f7ee
|
852 |
int fifo; |
1da177e4c
|
853 |
|
3b18152c3
|
854 |
if (cfq_cfqq_fifo_expire(cfqq)) |
1da177e4c
|
855 |
return NULL; |
cb8874119
|
856 857 |
cfq_mark_cfqq_fifo_expire(cfqq); |
89850f7ee
|
858 859 |
if (list_empty(&cfqq->fifo)) return NULL; |
1da177e4c
|
860 |
|
89850f7ee
|
861 862 |
fifo = cfq_cfqq_class_sync(cfqq); rq = rq_entry_fifo(cfqq->fifo.next); |
1da177e4c
|
863 |
|
cb8874119
|
864 |
if (time_after(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo])) |
89850f7ee
|
865 |
return rq; |
1da177e4c
|
866 867 868 |
return NULL; } |
22e2c507c
|
869 870 871 872 |
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
|
873 |
|
22e2c507c
|
874 |
WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR); |
1da177e4c
|
875 |
|
22e2c507c
|
876 |
return 2 * (base_rq + base_rq * (CFQ_PRIO_LISTS - 1 - cfqq->ioprio)); |
1da177e4c
|
877 |
} |
22e2c507c
|
878 879 880 |
/* * get next queue for service */ |
1b5ed5e1f
|
881 |
static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd) |
1da177e4c
|
882 |
{ |
1da177e4c
|
883 |
struct cfq_queue *cfqq; |
1da177e4c
|
884 |
|
22e2c507c
|
885 886 887 |
cfqq = cfqd->active_queue; if (!cfqq) goto new_queue; |
1da177e4c
|
888 |
|
22e2c507c
|
889 890 891 |
/* * slice has expired */ |
44f7c1606
|
892 |
if (!cfq_cfqq_must_dispatch(cfqq) && cfq_slice_used(cfqq)) |
3b18152c3
|
893 |
goto expire; |
1da177e4c
|
894 |
|
22e2c507c
|
895 896 897 898 |
/* * if queue has requests, dispatch one. if not, check if * enough slice is left to wait for one */ |
dd67d0515
|
899 |
if (!RB_EMPTY_ROOT(&cfqq->sort_list)) |
22e2c507c
|
900 |
goto keep_queue; |
44f7c1606
|
901 |
else if (cfq_cfqq_slice_new(cfqq) || cfq_cfqq_dispatched(cfqq)) { |
caaa5f9f0
|
902 903 904 |
cfqq = NULL; goto keep_queue; } else if (cfq_cfqq_class_sync(cfqq)) { |
1792669cc
|
905 |
if (cfq_arm_slice_timer(cfqd)) |
22e2c507c
|
906 907 |
return NULL; } |
3b18152c3
|
908 |
expire: |
3c6bd2f87
|
909 |
cfq_slice_expired(cfqd, 0, 0); |
3b18152c3
|
910 911 |
new_queue: cfqq = cfq_set_active_queue(cfqd); |
22e2c507c
|
912 |
keep_queue: |
3b18152c3
|
913 |
return cfqq; |
22e2c507c
|
914 915 916 917 918 919 920 |
} static int __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq, int max_dispatch) { int dispatched = 0; |
dd67d0515
|
921 |
BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list)); |
22e2c507c
|
922 923 |
do { |
5e7053747
|
924 |
struct request *rq; |
1da177e4c
|
925 926 |
/* |
22e2c507c
|
927 |
* follow expired path, else get first next available |
1da177e4c
|
928 |
*/ |
5e7053747
|
929 930 |
if ((rq = cfq_check_fifo(cfqq)) == NULL) rq = cfqq->next_rq; |
22e2c507c
|
931 932 933 934 |
/* * finally, insert request into driver dispatch list */ |
5e7053747
|
935 |
cfq_dispatch_insert(cfqd->queue, rq); |
1da177e4c
|
936 |
|
22e2c507c
|
937 938 |
cfqd->dispatch_slice++; dispatched++; |
1da177e4c
|
939 |
|
22e2c507c
|
940 |
if (!cfqd->active_cic) { |
5e7053747
|
941 942 |
atomic_inc(&RQ_CIC(rq)->ioc->refcount); cfqd->active_cic = RQ_CIC(rq); |
22e2c507c
|
943 |
} |
1da177e4c
|
944 |
|
dd67d0515
|
945 |
if (RB_EMPTY_ROOT(&cfqq->sort_list)) |
22e2c507c
|
946 947 948 949 950 |
break; } while (dispatched < max_dispatch); /* |
22e2c507c
|
951 952 953 |
* expire an async queue immediately if it has used up its slice. idle * queue always expire after 1 dispatch round. */ |
a99380065
|
954 |
if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) && |
22e2c507c
|
955 |
cfqd->dispatch_slice >= cfq_prio_to_maxrq(cfqd, cfqq)) || |
a99380065
|
956 |
cfq_class_idle(cfqq))) { |
44f7c1606
|
957 |
cfqq->slice_end = jiffies + 1; |
3c6bd2f87
|
958 |
cfq_slice_expired(cfqd, 0, 0); |
44f7c1606
|
959 |
} |
22e2c507c
|
960 961 962 963 964 |
return dispatched; } static int |
1b5ed5e1f
|
965 966 |
cfq_forced_dispatch_cfqqs(struct list_head *list) { |
1b5ed5e1f
|
967 |
struct cfq_queue *cfqq, *next; |
caaa5f9f0
|
968 |
int dispatched; |
1b5ed5e1f
|
969 |
|
caaa5f9f0
|
970 |
dispatched = 0; |
1b5ed5e1f
|
971 |
list_for_each_entry_safe(cfqq, next, list, cfq_list) { |
5e7053747
|
972 973 |
while (cfqq->next_rq) { cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq); |
1b5ed5e1f
|
974 975 976 977 |
dispatched++; } BUG_ON(!list_empty(&cfqq->fifo)); } |
caaa5f9f0
|
978 |
|
1b5ed5e1f
|
979 980 981 982 983 984 985 986 987 988 989 990 991 992 |
return dispatched; } static int cfq_forced_dispatch(struct cfq_data *cfqd) { int i, dispatched = 0; for (i = 0; i < CFQ_PRIO_LISTS; i++) dispatched += cfq_forced_dispatch_cfqqs(&cfqd->rr_list[i]); dispatched += cfq_forced_dispatch_cfqqs(&cfqd->busy_rr); dispatched += cfq_forced_dispatch_cfqqs(&cfqd->cur_rr); dispatched += cfq_forced_dispatch_cfqqs(&cfqd->idle_rr); |
3c6bd2f87
|
993 |
cfq_slice_expired(cfqd, 0, 0); |
1b5ed5e1f
|
994 995 996 997 998 999 1000 |
BUG_ON(cfqd->busy_queues); return dispatched; } static int |
b4878f245
|
1001 |
cfq_dispatch_requests(request_queue_t *q, int force) |
22e2c507c
|
1002 1003 |
{ struct cfq_data *cfqd = q->elevator->elevator_data; |
caaa5f9f0
|
1004 1005 |
struct cfq_queue *cfqq, *prev_cfqq; int dispatched; |
22e2c507c
|
1006 1007 1008 |
if (!cfqd->busy_queues) return 0; |
1b5ed5e1f
|
1009 1010 |
if (unlikely(force)) return cfq_forced_dispatch(cfqd); |
caaa5f9f0
|
1011 1012 1013 |
dispatched = 0; prev_cfqq = NULL; while ((cfqq = cfq_select_queue(cfqd)) != NULL) { |
b4878f245
|
1014 |
int max_dispatch; |
a99380065
|
1015 1016 1017 1018 1019 1020 |
if (cfqd->busy_queues > 1) { /* * Don't repeat dispatch from the previous queue. */ if (prev_cfqq == cfqq) break; |
caaa5f9f0
|
1021 |
|
a99380065
|
1022 1023 1024 1025 1026 1027 1028 1029 |
/* * So we have dispatched before in this round, if the * next queue has idling enabled (must be sync), don't * allow it service until the previous have continued. */ if (cfqd->rq_in_driver && cfq_cfqq_idle_window(cfqq)) break; } |
9ede209e8
|
1030 |
|
3b18152c3
|
1031 1032 |
cfq_clear_cfqq_must_dispatch(cfqq); cfq_clear_cfqq_wait_request(cfqq); |
22e2c507c
|
1033 |
del_timer(&cfqd->idle_slice_timer); |
1b5ed5e1f
|
1034 1035 1036 |
max_dispatch = cfqd->cfq_quantum; if (cfq_class_idle(cfqq)) max_dispatch = 1; |
1da177e4c
|
1037 |
|
caaa5f9f0
|
1038 |
dispatched += __cfq_dispatch_requests(cfqd, cfqq, max_dispatch); |
caaa5f9f0
|
1039 |
prev_cfqq = cfqq; |
1da177e4c
|
1040 |
} |
caaa5f9f0
|
1041 |
return dispatched; |
1da177e4c
|
1042 |
} |
1da177e4c
|
1043 |
/* |
5e7053747
|
1044 1045 |
* 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
|
1046 1047 1048 1049 1050 |
* * queue lock must be held here. */ static void cfq_put_queue(struct cfq_queue *cfqq) { |
22e2c507c
|
1051 1052 1053 |
struct cfq_data *cfqd = cfqq->cfqd; BUG_ON(atomic_read(&cfqq->ref) <= 0); |
1da177e4c
|
1054 1055 1056 1057 1058 |
if (!atomic_dec_and_test(&cfqq->ref)) return; BUG_ON(rb_first(&cfqq->sort_list)); |
22e2c507c
|
1059 |
BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]); |
3b18152c3
|
1060 |
BUG_ON(cfq_cfqq_on_rr(cfqq)); |
1da177e4c
|
1061 |
|
28f95cbc3
|
1062 |
if (unlikely(cfqd->active_queue == cfqq)) { |
3c6bd2f87
|
1063 |
__cfq_slice_expired(cfqd, cfqq, 0, 0); |
28f95cbc3
|
1064 1065 |
cfq_schedule_dispatch(cfqd); } |
22e2c507c
|
1066 |
|
1da177e4c
|
1067 1068 1069 1070 1071 1072 1073 |
/* * it's on the empty list and still hashed */ list_del(&cfqq->cfq_list); hlist_del(&cfqq->cfq_hash); kmem_cache_free(cfq_pool, cfqq); } |
1ea25ecb7
|
1074 |
static struct cfq_queue * |
3b18152c3
|
1075 1076 |
__cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned int prio, const int hashval) |
1da177e4c
|
1077 1078 |
{ struct hlist_head *hash_list = &cfqd->cfq_hash[hashval]; |
206dc69b3
|
1079 1080 |
struct hlist_node *entry; struct cfq_queue *__cfqq; |
1da177e4c
|
1081 |
|
206dc69b3
|
1082 |
hlist_for_each_entry(__cfqq, entry, hash_list, cfq_hash) { |
b0a6916bc
|
1083 |
const unsigned short __p = IOPRIO_PRIO_VALUE(__cfqq->org_ioprio_class, __cfqq->org_ioprio); |
1da177e4c
|
1084 |
|
206dc69b3
|
1085 |
if (__cfqq->key == key && (__p == prio || !prio)) |
1da177e4c
|
1086 1087 1088 1089 1090 1091 1092 |
return __cfqq; } return NULL; } static struct cfq_queue * |
3b18152c3
|
1093 |
cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned short prio) |
1da177e4c
|
1094 |
{ |
3b18152c3
|
1095 |
return __cfq_find_cfq_hash(cfqd, key, prio, hash_long(key, CFQ_QHASH_SHIFT)); |
1da177e4c
|
1096 |
} |
e2d74ac06
|
1097 |
static void cfq_free_io_context(struct io_context *ioc) |
1da177e4c
|
1098 |
{ |
22e2c507c
|
1099 |
struct cfq_io_context *__cic; |
e2d74ac06
|
1100 1101 |
struct rb_node *n; int freed = 0; |
1da177e4c
|
1102 |
|
e2d74ac06
|
1103 1104 1105 |
while ((n = rb_first(&ioc->cic_root)) != NULL) { __cic = rb_entry(n, struct cfq_io_context, rb_node); rb_erase(&__cic->rb_node, &ioc->cic_root); |
22e2c507c
|
1106 |
kmem_cache_free(cfq_ioc_pool, __cic); |
334e94de9
|
1107 |
freed++; |
1da177e4c
|
1108 |
} |
4050cf167
|
1109 1110 1111 |
elv_ioc_count_mod(ioc_count, -freed); if (ioc_gone && !elv_ioc_count_read(ioc_count)) |
334e94de9
|
1112 |
complete(ioc_gone); |
1da177e4c
|
1113 |
} |
89850f7ee
|
1114 |
static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq) |
1da177e4c
|
1115 |
{ |
28f95cbc3
|
1116 |
if (unlikely(cfqq == cfqd->active_queue)) { |
3c6bd2f87
|
1117 |
__cfq_slice_expired(cfqd, cfqq, 0, 0); |
28f95cbc3
|
1118 1119 |
cfq_schedule_dispatch(cfqd); } |
22e2c507c
|
1120 |
|
89850f7ee
|
1121 1122 |
cfq_put_queue(cfqq); } |
22e2c507c
|
1123 |
|
89850f7ee
|
1124 1125 1126 |
static void __cfq_exit_single_io_context(struct cfq_data *cfqd, struct cfq_io_context *cic) { |
fc46379da
|
1127 1128 1129 |
list_del_init(&cic->queue_list); smp_wmb(); cic->key = NULL; |
12a057321
|
1130 |
if (cic->cfqq[ASYNC]) { |
89850f7ee
|
1131 |
cfq_exit_cfqq(cfqd, cic->cfqq[ASYNC]); |
12a057321
|
1132 1133 1134 1135 |
cic->cfqq[ASYNC] = NULL; } if (cic->cfqq[SYNC]) { |
89850f7ee
|
1136 |
cfq_exit_cfqq(cfqd, cic->cfqq[SYNC]); |
12a057321
|
1137 1138 |
cic->cfqq[SYNC] = NULL; } |
89850f7ee
|
1139 1140 1141 1142 1143 1144 1145 1146 1147 |
} /* * Called with interrupts disabled */ static void cfq_exit_single_io_context(struct cfq_io_context *cic) { struct cfq_data *cfqd = cic->key; |
89850f7ee
|
1148 1149 |
if (cfqd) { request_queue_t *q = cfqd->queue; |
fc46379da
|
1150 |
spin_lock_irq(q->queue_lock); |
89850f7ee
|
1151 |
__cfq_exit_single_io_context(cfqd, cic); |
fc46379da
|
1152 |
spin_unlock_irq(q->queue_lock); |
89850f7ee
|
1153 |
} |
1da177e4c
|
1154 |
} |
e2d74ac06
|
1155 |
static void cfq_exit_io_context(struct io_context *ioc) |
1da177e4c
|
1156 |
{ |
22e2c507c
|
1157 |
struct cfq_io_context *__cic; |
e2d74ac06
|
1158 |
struct rb_node *n; |
22e2c507c
|
1159 |
|
1da177e4c
|
1160 1161 1162 |
/* * put the reference this task is holding to the various queues */ |
e2d74ac06
|
1163 1164 1165 1166 |
n = rb_first(&ioc->cic_root); while (n != NULL) { __cic = rb_entry(n, struct cfq_io_context, rb_node); |
22e2c507c
|
1167 |
cfq_exit_single_io_context(__cic); |
e2d74ac06
|
1168 |
n = rb_next(n); |
1da177e4c
|
1169 |
} |
1da177e4c
|
1170 |
} |
22e2c507c
|
1171 |
static struct cfq_io_context * |
8267e268e
|
1172 |
cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask) |
1da177e4c
|
1173 |
{ |
b5deef901
|
1174 |
struct cfq_io_context *cic; |
1da177e4c
|
1175 |
|
b5deef901
|
1176 |
cic = kmem_cache_alloc_node(cfq_ioc_pool, gfp_mask, cfqd->queue->node); |
1da177e4c
|
1177 |
if (cic) { |
553698f94
|
1178 |
memset(cic, 0, sizeof(*cic)); |
22e2c507c
|
1179 |
cic->last_end_request = jiffies; |
553698f94
|
1180 |
INIT_LIST_HEAD(&cic->queue_list); |
22e2c507c
|
1181 1182 |
cic->dtor = cfq_free_io_context; cic->exit = cfq_exit_io_context; |
4050cf167
|
1183 |
elv_ioc_count_inc(ioc_count); |
1da177e4c
|
1184 1185 1186 1187 |
} return cic; } |
22e2c507c
|
1188 1189 1190 1191 |
static void cfq_init_prio_data(struct cfq_queue *cfqq) { struct task_struct *tsk = current; int ioprio_class; |
3b18152c3
|
1192 |
if (!cfq_cfqq_prio_changed(cfqq)) |
22e2c507c
|
1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217 |
return; ioprio_class = IOPRIO_PRIO_CLASS(tsk->ioprio); switch (ioprio_class) { default: printk(KERN_ERR "cfq: bad prio %x ", ioprio_class); case IOPRIO_CLASS_NONE: /* * no prio set, place us in the middle of the BE classes */ cfqq->ioprio = task_nice_ioprio(tsk); cfqq->ioprio_class = IOPRIO_CLASS_BE; break; case IOPRIO_CLASS_RT: cfqq->ioprio = task_ioprio(tsk); cfqq->ioprio_class = IOPRIO_CLASS_RT; break; case IOPRIO_CLASS_BE: cfqq->ioprio = task_ioprio(tsk); cfqq->ioprio_class = IOPRIO_CLASS_BE; break; case IOPRIO_CLASS_IDLE: cfqq->ioprio_class = IOPRIO_CLASS_IDLE; cfqq->ioprio = 7; |
3b18152c3
|
1218 |
cfq_clear_cfqq_idle_window(cfqq); |
22e2c507c
|
1219 1220 1221 1222 1223 1224 1225 1226 1227 |
break; } /* * keep track of original prio settings in case we have to temporarily * elevate the priority of this queue */ cfqq->org_ioprio = cfqq->ioprio; cfqq->org_ioprio_class = cfqq->ioprio_class; |
98e41c7df
|
1228 |
cfq_resort_rr_list(cfqq, 0); |
3b18152c3
|
1229 |
cfq_clear_cfqq_prio_changed(cfqq); |
22e2c507c
|
1230 |
} |
478a82b0e
|
1231 |
static inline void changed_ioprio(struct cfq_io_context *cic) |
22e2c507c
|
1232 |
{ |
478a82b0e
|
1233 1234 |
struct cfq_data *cfqd = cic->key; struct cfq_queue *cfqq; |
c1b707d25
|
1235 |
unsigned long flags; |
35e6077cb
|
1236 |
|
caaa5f9f0
|
1237 1238 |
if (unlikely(!cfqd)) return; |
c1b707d25
|
1239 |
spin_lock_irqsave(cfqd->queue->queue_lock, flags); |
caaa5f9f0
|
1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 |
cfqq = cic->cfqq[ASYNC]; if (cfqq) { struct cfq_queue *new_cfqq; new_cfqq = cfq_get_queue(cfqd, CFQ_KEY_ASYNC, cic->ioc->task, GFP_ATOMIC); if (new_cfqq) { cic->cfqq[ASYNC] = new_cfqq; cfq_put_queue(cfqq); } |
22e2c507c
|
1250 |
} |
caaa5f9f0
|
1251 1252 1253 1254 |
cfqq = cic->cfqq[SYNC]; if (cfqq) cfq_mark_cfqq_prio_changed(cfqq); |
c1b707d25
|
1255 |
spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); |
22e2c507c
|
1256 |
} |
fc46379da
|
1257 |
static void cfq_ioc_set_ioprio(struct io_context *ioc) |
22e2c507c
|
1258 |
{ |
a6a0763a6
|
1259 |
struct cfq_io_context *cic; |
e2d74ac06
|
1260 |
struct rb_node *n; |
a6a0763a6
|
1261 |
|
fc46379da
|
1262 |
ioc->ioprio_changed = 0; |
a6a0763a6
|
1263 |
|
e2d74ac06
|
1264 1265 1266 |
n = rb_first(&ioc->cic_root); while (n != NULL) { cic = rb_entry(n, struct cfq_io_context, rb_node); |
3793c65c1
|
1267 |
|
478a82b0e
|
1268 |
changed_ioprio(cic); |
e2d74ac06
|
1269 1270 |
n = rb_next(n); } |
22e2c507c
|
1271 1272 1273 |
} static struct cfq_queue * |
6f325a134
|
1274 |
cfq_get_queue(struct cfq_data *cfqd, unsigned int key, struct task_struct *tsk, |
8267e268e
|
1275 |
gfp_t gfp_mask) |
22e2c507c
|
1276 1277 1278 |
{ const int hashval = hash_long(key, CFQ_QHASH_SHIFT); struct cfq_queue *cfqq, *new_cfqq = NULL; |
6f325a134
|
1279 |
unsigned short ioprio; |
22e2c507c
|
1280 1281 |
retry: |
6f325a134
|
1282 |
ioprio = tsk->ioprio; |
3b18152c3
|
1283 |
cfqq = __cfq_find_cfq_hash(cfqd, key, ioprio, hashval); |
22e2c507c
|
1284 1285 1286 1287 1288 1289 |
if (!cfqq) { if (new_cfqq) { cfqq = new_cfqq; new_cfqq = NULL; } else if (gfp_mask & __GFP_WAIT) { |
89850f7ee
|
1290 1291 1292 1293 1294 1295 |
/* * Inform the allocator of the fact that we will * just repeat this allocation if it fails, to allow * the allocator to do whatever it needs to attempt to * free memory. */ |
22e2c507c
|
1296 |
spin_unlock_irq(cfqd->queue->queue_lock); |
b5deef901
|
1297 |
new_cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask|__GFP_NOFAIL, cfqd->queue->node); |
22e2c507c
|
1298 1299 1300 |
spin_lock_irq(cfqd->queue->queue_lock); goto retry; } else { |
b5deef901
|
1301 |
cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask, cfqd->queue->node); |
22e2c507c
|
1302 1303 1304 1305 1306 1307 1308 1309 |
if (!cfqq) goto out; } memset(cfqq, 0, sizeof(*cfqq)); INIT_HLIST_NODE(&cfqq->cfq_hash); INIT_LIST_HEAD(&cfqq->cfq_list); |
22e2c507c
|
1310 1311 1312 1313 1314 1315 |
INIT_LIST_HEAD(&cfqq->fifo); cfqq->key = key; hlist_add_head(&cfqq->cfq_hash, &cfqd->cfq_hash[hashval]); atomic_set(&cfqq->ref, 0); cfqq->cfqd = cfqd; |
c5b680f3b
|
1316 |
|
a99380065
|
1317 1318 |
if (key != CFQ_KEY_ASYNC) cfq_mark_cfqq_idle_window(cfqq); |
3b18152c3
|
1319 |
cfq_mark_cfqq_prio_changed(cfqq); |
53b03744e
|
1320 |
cfq_mark_cfqq_queue_new(cfqq); |
3b18152c3
|
1321 |
cfq_init_prio_data(cfqq); |
22e2c507c
|
1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 |
} if (new_cfqq) kmem_cache_free(cfq_pool, new_cfqq); atomic_inc(&cfqq->ref); out: WARN_ON((gfp_mask & __GFP_WAIT) && !cfqq); return cfqq; } |
dbecf3ab4
|
1332 1333 1334 |
static void cfq_drop_dead_cic(struct io_context *ioc, struct cfq_io_context *cic) { |
fc46379da
|
1335 |
WARN_ON(!list_empty(&cic->queue_list)); |
dbecf3ab4
|
1336 |
rb_erase(&cic->rb_node, &ioc->cic_root); |
dbecf3ab4
|
1337 |
kmem_cache_free(cfq_ioc_pool, cic); |
4050cf167
|
1338 |
elv_ioc_count_dec(ioc_count); |
dbecf3ab4
|
1339 |
} |
e2d74ac06
|
1340 1341 1342 |
static struct cfq_io_context * cfq_cic_rb_lookup(struct cfq_data *cfqd, struct io_context *ioc) { |
dbecf3ab4
|
1343 |
struct rb_node *n; |
e2d74ac06
|
1344 |
struct cfq_io_context *cic; |
be3b07535
|
1345 |
void *k, *key = cfqd; |
e2d74ac06
|
1346 |
|
dbecf3ab4
|
1347 1348 |
restart: n = ioc->cic_root.rb_node; |
e2d74ac06
|
1349 1350 |
while (n) { cic = rb_entry(n, struct cfq_io_context, rb_node); |
be3b07535
|
1351 1352 1353 |
/* ->key must be copied to avoid race with cfq_exit_queue() */ k = cic->key; if (unlikely(!k)) { |
dbecf3ab4
|
1354 1355 1356 |
cfq_drop_dead_cic(ioc, cic); goto restart; } |
e2d74ac06
|
1357 |
|
be3b07535
|
1358 |
if (key < k) |
e2d74ac06
|
1359 |
n = n->rb_left; |
be3b07535
|
1360 |
else if (key > k) |
e2d74ac06
|
1361 1362 1363 1364 1365 1366 1367 1368 1369 1370 1371 1372 |
n = n->rb_right; else return cic; } return NULL; } static inline void cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc, struct cfq_io_context *cic) { |
dbecf3ab4
|
1373 1374 |
struct rb_node **p; struct rb_node *parent; |
e2d74ac06
|
1375 |
struct cfq_io_context *__cic; |
0261d6886
|
1376 |
unsigned long flags; |
be3b07535
|
1377 |
void *k; |
e2d74ac06
|
1378 |
|
e2d74ac06
|
1379 1380 |
cic->ioc = ioc; cic->key = cfqd; |
dbecf3ab4
|
1381 1382 1383 |
restart: parent = NULL; p = &ioc->cic_root.rb_node; |
e2d74ac06
|
1384 1385 1386 |
while (*p) { parent = *p; __cic = rb_entry(parent, struct cfq_io_context, rb_node); |
be3b07535
|
1387 1388 1389 |
/* ->key must be copied to avoid race with cfq_exit_queue() */ k = __cic->key; if (unlikely(!k)) { |
be33c3a67
|
1390 |
cfq_drop_dead_cic(ioc, __cic); |
dbecf3ab4
|
1391 1392 |
goto restart; } |
e2d74ac06
|
1393 |
|
be3b07535
|
1394 |
if (cic->key < k) |
e2d74ac06
|
1395 |
p = &(*p)->rb_left; |
be3b07535
|
1396 |
else if (cic->key > k) |
e2d74ac06
|
1397 1398 1399 1400 1401 1402 1403 |
p = &(*p)->rb_right; else BUG(); } rb_link_node(&cic->rb_node, parent, p); rb_insert_color(&cic->rb_node, &ioc->cic_root); |
fc46379da
|
1404 |
|
0261d6886
|
1405 |
spin_lock_irqsave(cfqd->queue->queue_lock, flags); |
e2d74ac06
|
1406 |
list_add(&cic->queue_list, &cfqd->cic_list); |
0261d6886
|
1407 |
spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); |
e2d74ac06
|
1408 |
} |
1da177e4c
|
1409 1410 1411 |
/* * 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
|
1412 |
* than one device managed by cfq. |
1da177e4c
|
1413 1414 |
*/ static struct cfq_io_context * |
e2d74ac06
|
1415 |
cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask) |
1da177e4c
|
1416 |
{ |
22e2c507c
|
1417 |
struct io_context *ioc = NULL; |
1da177e4c
|
1418 |
struct cfq_io_context *cic; |
1da177e4c
|
1419 |
|
22e2c507c
|
1420 |
might_sleep_if(gfp_mask & __GFP_WAIT); |
1da177e4c
|
1421 |
|
b5deef901
|
1422 |
ioc = get_io_context(gfp_mask, cfqd->queue->node); |
1da177e4c
|
1423 1424 |
if (!ioc) return NULL; |
e2d74ac06
|
1425 1426 1427 |
cic = cfq_cic_rb_lookup(cfqd, ioc); if (cic) goto out; |
1da177e4c
|
1428 |
|
e2d74ac06
|
1429 1430 1431 |
cic = cfq_alloc_io_context(cfqd, gfp_mask); if (cic == NULL) goto err; |
1da177e4c
|
1432 |
|
e2d74ac06
|
1433 |
cfq_cic_link(cfqd, ioc, cic); |
1da177e4c
|
1434 |
out: |
fc46379da
|
1435 1436 1437 |
smp_read_barrier_depends(); if (unlikely(ioc->ioprio_changed)) cfq_ioc_set_ioprio(ioc); |
1da177e4c
|
1438 1439 1440 1441 1442 |
return cic; err: put_io_context(ioc); return NULL; } |
22e2c507c
|
1443 1444 |
static void cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_io_context *cic) |
1da177e4c
|
1445 |
{ |
aaf1228dd
|
1446 1447 |
unsigned long elapsed = jiffies - cic->last_end_request; unsigned long ttime = min(elapsed, 2UL * cfqd->cfq_slice_idle); |
db3b5848e
|
1448 |
|
22e2c507c
|
1449 1450 1451 1452 |
cic->ttime_samples = (7*cic->ttime_samples + 256) / 8; cic->ttime_total = (7*cic->ttime_total + 256*ttime) / 8; cic->ttime_mean = (cic->ttime_total + 128) / cic->ttime_samples; } |
1da177e4c
|
1453 |
|
206dc69b3
|
1454 |
static void |
bb37b94c6
|
1455 |
cfq_update_io_seektime(struct cfq_io_context *cic, struct request *rq) |
206dc69b3
|
1456 1457 1458 |
{ sector_t sdist; u64 total; |
5e7053747
|
1459 1460 |
if (cic->last_request_pos < rq->sector) sdist = rq->sector - cic->last_request_pos; |
206dc69b3
|
1461 |
else |
5e7053747
|
1462 |
sdist = cic->last_request_pos - rq->sector; |
206dc69b3
|
1463 1464 1465 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477 1478 |
/* * Don't allow the seek distance to get too large from the * odd fragment, pagein, etc */ if (cic->seek_samples <= 60) /* second&third seek */ sdist = min(sdist, (cic->seek_mean * 4) + 2*1024*1024); else sdist = min(sdist, (cic->seek_mean * 4) + 2*1024*64); cic->seek_samples = (7*cic->seek_samples + 256) / 8; cic->seek_total = (7*cic->seek_total + (u64)256*sdist) / 8; total = cic->seek_total + (cic->seek_samples/2); do_div(total, cic->seek_samples); cic->seek_mean = (sector_t)total; } |
1da177e4c
|
1479 |
|
22e2c507c
|
1480 1481 1482 1483 1484 1485 1486 1487 |
/* * 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) { |
3b18152c3
|
1488 |
int enable_idle = cfq_cfqq_idle_window(cfqq); |
1da177e4c
|
1489 |
|
caaa5f9f0
|
1490 1491 |
if (!cic->ioc->task || !cfqd->cfq_slice_idle || (cfqd->hw_tag && CIC_SEEKY(cic))) |
22e2c507c
|
1492 1493 1494 1495 1496 1497 |
enable_idle = 0; else if (sample_valid(cic->ttime_samples)) { if (cic->ttime_mean > cfqd->cfq_slice_idle) enable_idle = 0; else enable_idle = 1; |
1da177e4c
|
1498 |
} |
3b18152c3
|
1499 1500 1501 1502 |
if (enable_idle) cfq_mark_cfqq_idle_window(cfqq); else cfq_clear_cfqq_idle_window(cfqq); |
22e2c507c
|
1503 |
} |
1da177e4c
|
1504 |
|
22e2c507c
|
1505 1506 1507 1508 1509 1510 |
/* * 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. */ static int cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq, |
5e7053747
|
1511 |
struct request *rq) |
22e2c507c
|
1512 1513 1514 1515 1516 1517 1518 |
{ struct cfq_queue *cfqq = cfqd->active_queue; if (cfq_class_idle(new_cfqq)) return 0; if (!cfqq) |
caaa5f9f0
|
1519 |
return 0; |
22e2c507c
|
1520 1521 1522 |
if (cfq_class_idle(cfqq)) return 1; |
3b18152c3
|
1523 |
if (!cfq_cfqq_wait_request(new_cfqq)) |
22e2c507c
|
1524 1525 |
return 0; /* |
374f84ac3
|
1526 1527 1528 |
* if the new request is sync, but the currently running queue is * not, let the sync request have priority. */ |
5e7053747
|
1529 |
if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq)) |
22e2c507c
|
1530 |
return 1; |
374f84ac3
|
1531 1532 1533 1534 1535 1536 |
/* * 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. */ if (rq_is_meta(rq) && !cfqq->meta_pending) return 1; |
22e2c507c
|
1537 1538 1539 1540 1541 1542 1543 1544 1545 1546 |
return 0; } /* * 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) { |
3c6bd2f87
|
1547 |
cfq_slice_expired(cfqd, 1, 1); |
22e2c507c
|
1548 |
|
bf5722567
|
1549 1550 1551 1552 1553 1554 |
/* * 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)); list_move(&cfqq->cfq_list, &cfqd->cur_rr); |
44f7c1606
|
1555 1556 |
cfqq->slice_end = 0; cfq_mark_cfqq_slice_new(cfqq); |
22e2c507c
|
1557 1558 1559 |
} /* |
5e7053747
|
1560 |
* Called when a new fs request (rq) is added (to cfqq). Check if there's |
22e2c507c
|
1561 1562 1563 |
* something we should do about it */ static void |
5e7053747
|
1564 1565 |
cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq, struct request *rq) |
22e2c507c
|
1566 |
{ |
5e7053747
|
1567 |
struct cfq_io_context *cic = RQ_CIC(rq); |
12e9fddd6
|
1568 |
|
374f84ac3
|
1569 1570 |
if (rq_is_meta(rq)) cfqq->meta_pending++; |
9c2c38a12
|
1571 1572 1573 1574 |
/* * we never wait for an async request and we don't allow preemption * of an async request. so just return early */ |
5e7053747
|
1575 |
if (!rq_is_sync(rq)) { |
12e9fddd6
|
1576 1577 1578 1579 1580 1581 |
/* * sync process issued an async request, if it's waiting * then expire it and kick rq handling. */ if (cic == cfqd->active_cic && del_timer(&cfqd->idle_slice_timer)) { |
3c6bd2f87
|
1582 |
cfq_slice_expired(cfqd, 0, 0); |
dc72ef4ae
|
1583 |
blk_start_queueing(cfqd->queue); |
12e9fddd6
|
1584 |
} |
9c2c38a12
|
1585 |
return; |
12e9fddd6
|
1586 |
} |
22e2c507c
|
1587 |
|
9c2c38a12
|
1588 |
cfq_update_io_thinktime(cfqd, cic); |
bb37b94c6
|
1589 |
cfq_update_io_seektime(cic, rq); |
9c2c38a12
|
1590 |
cfq_update_idle_window(cfqd, cfqq, cic); |
5e7053747
|
1591 |
cic->last_request_pos = rq->sector + rq->nr_sectors; |
22e2c507c
|
1592 1593 1594 1595 1596 1597 1598 |
if (cfqq == cfqd->active_queue) { /* * if we are waiting for a request for this queue, let it rip * immediately and flag that we must not expire this queue * just now */ |
3b18152c3
|
1599 1600 |
if (cfq_cfqq_wait_request(cfqq)) { cfq_mark_cfqq_must_dispatch(cfqq); |
22e2c507c
|
1601 |
del_timer(&cfqd->idle_slice_timer); |
dc72ef4ae
|
1602 |
blk_start_queueing(cfqd->queue); |
22e2c507c
|
1603 |
} |
5e7053747
|
1604 |
} else if (cfq_should_preempt(cfqd, cfqq, rq)) { |
22e2c507c
|
1605 1606 1607 1608 1609 1610 |
/* * not the active queue - expire current slice if it is * idle and has expired it's mean thinktime or this new queue * has some old slice time left and is of higher priority */ cfq_preempt_queue(cfqd, cfqq); |
3b18152c3
|
1611 |
cfq_mark_cfqq_must_dispatch(cfqq); |
dc72ef4ae
|
1612 |
blk_start_queueing(cfqd->queue); |
22e2c507c
|
1613 |
} |
1da177e4c
|
1614 |
} |
b4878f245
|
1615 |
static void cfq_insert_request(request_queue_t *q, struct request *rq) |
1da177e4c
|
1616 |
{ |
b4878f245
|
1617 |
struct cfq_data *cfqd = q->elevator->elevator_data; |
5e7053747
|
1618 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
22e2c507c
|
1619 1620 |
cfq_init_prio_data(cfqq); |
1da177e4c
|
1621 |
|
5e7053747
|
1622 |
cfq_add_rq_rb(rq); |
1da177e4c
|
1623 |
|
22e2c507c
|
1624 |
list_add_tail(&rq->queuelist, &cfqq->fifo); |
5e7053747
|
1625 |
cfq_rq_enqueued(cfqd, cfqq, rq); |
1da177e4c
|
1626 |
} |
1da177e4c
|
1627 1628 |
static void cfq_completed_request(request_queue_t *q, struct request *rq) { |
5e7053747
|
1629 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
b4878f245
|
1630 |
struct cfq_data *cfqd = cfqq->cfqd; |
5380a101d
|
1631 |
const int sync = rq_is_sync(rq); |
b4878f245
|
1632 |
unsigned long now; |
1da177e4c
|
1633 |
|
b4878f245
|
1634 |
now = jiffies; |
1da177e4c
|
1635 |
|
b4878f245
|
1636 1637 1638 1639 |
WARN_ON(!cfqd->rq_in_driver); WARN_ON(!cfqq->on_dispatch[sync]); cfqd->rq_in_driver--; cfqq->on_dispatch[sync]--; |
99f9628ab
|
1640 |
cfqq->service_last = now; |
1da177e4c
|
1641 |
|
b4878f245
|
1642 1643 |
if (!cfq_class_idle(cfqq)) cfqd->last_end_request = now; |
3b18152c3
|
1644 |
|
98e41c7df
|
1645 |
cfq_resort_rr_list(cfqq, 0); |
1da177e4c
|
1646 |
|
caaa5f9f0
|
1647 |
if (sync) |
5e7053747
|
1648 |
RQ_CIC(rq)->last_end_request = now; |
caaa5f9f0
|
1649 1650 1651 1652 1653 1654 |
/* * 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) { |
44f7c1606
|
1655 1656 1657 1658 1659 |
if (cfq_cfqq_slice_new(cfqq)) { cfq_set_prio_slice(cfqd, cfqq); cfq_clear_cfqq_slice_new(cfqq); } if (cfq_slice_used(cfqq)) |
3c6bd2f87
|
1660 |
cfq_slice_expired(cfqd, 0, 1); |
dd67d0515
|
1661 |
else if (sync && RB_EMPTY_ROOT(&cfqq->sort_list)) { |
1792669cc
|
1662 |
if (!cfq_arm_slice_timer(cfqd)) |
caaa5f9f0
|
1663 1664 1665 |
cfq_schedule_dispatch(cfqd); } } |
1da177e4c
|
1666 |
} |
22e2c507c
|
1667 1668 1669 1670 1671 |
/* * we temporarily boost lower priority queues if they are holding fs exclusive * resources. they are boosted to normal prio (CLASS_BE/4) */ static void cfq_prio_boost(struct cfq_queue *cfqq) |
1da177e4c
|
1672 |
{ |
22e2c507c
|
1673 1674 |
const int ioprio_class = cfqq->ioprio_class; const int ioprio = cfqq->ioprio; |
1da177e4c
|
1675 |
|
22e2c507c
|
1676 1677 1678 1679 1680 1681 1682 1683 1684 1685 1686 1687 1688 1689 1690 1691 1692 1693 |
if (has_fs_excl()) { /* * boost idle prio on transactions that would lock out other * users of the filesystem */ if (cfq_class_idle(cfqq)) cfqq->ioprio_class = IOPRIO_CLASS_BE; if (cfqq->ioprio > IOPRIO_NORM) cfqq->ioprio = IOPRIO_NORM; } else { /* * check if we need to unboost the queue */ if (cfqq->ioprio_class != cfqq->org_ioprio_class) cfqq->ioprio_class = cfqq->org_ioprio_class; if (cfqq->ioprio != cfqq->org_ioprio) cfqq->ioprio = cfqq->org_ioprio; } |
1da177e4c
|
1694 |
|
22e2c507c
|
1695 1696 1697 |
/* * refile between round-robin lists if we moved the priority class */ |
98e41c7df
|
1698 |
if ((ioprio_class != cfqq->ioprio_class || ioprio != cfqq->ioprio)) |
22e2c507c
|
1699 1700 |
cfq_resort_rr_list(cfqq, 0); } |
1da177e4c
|
1701 |
|
89850f7ee
|
1702 |
static inline int __cfq_may_queue(struct cfq_queue *cfqq) |
22e2c507c
|
1703 |
{ |
3b18152c3
|
1704 |
if ((cfq_cfqq_wait_request(cfqq) || cfq_cfqq_must_alloc(cfqq)) && |
99f95e528
|
1705 |
!cfq_cfqq_must_alloc_slice(cfqq)) { |
3b18152c3
|
1706 |
cfq_mark_cfqq_must_alloc_slice(cfqq); |
22e2c507c
|
1707 |
return ELV_MQUEUE_MUST; |
3b18152c3
|
1708 |
} |
1da177e4c
|
1709 |
|
22e2c507c
|
1710 |
return ELV_MQUEUE_MAY; |
22e2c507c
|
1711 |
} |
cb78b285c
|
1712 |
static int cfq_may_queue(request_queue_t *q, int rw) |
22e2c507c
|
1713 1714 1715 1716 |
{ struct cfq_data *cfqd = q->elevator->elevator_data; struct task_struct *tsk = current; struct cfq_queue *cfqq; |
7749a8d42
|
1717 1718 1719 |
unsigned int key; key = cfq_queue_pid(tsk, rw, rw & REQ_RW_SYNC); |
22e2c507c
|
1720 1721 1722 1723 1724 1725 1726 |
/* * 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 */ |
7749a8d42
|
1727 |
cfqq = cfq_find_cfq_hash(cfqd, key, tsk->ioprio); |
22e2c507c
|
1728 1729 1730 |
if (cfqq) { cfq_init_prio_data(cfqq); cfq_prio_boost(cfqq); |
89850f7ee
|
1731 |
return __cfq_may_queue(cfqq); |
22e2c507c
|
1732 1733 1734 |
} return ELV_MQUEUE_MAY; |
1da177e4c
|
1735 |
} |
1da177e4c
|
1736 1737 1738 |
/* * queue lock held here */ |
bb37b94c6
|
1739 |
static void cfq_put_request(struct request *rq) |
1da177e4c
|
1740 |
{ |
5e7053747
|
1741 |
struct cfq_queue *cfqq = RQ_CFQQ(rq); |
1da177e4c
|
1742 |
|
5e7053747
|
1743 |
if (cfqq) { |
22e2c507c
|
1744 |
const int rw = rq_data_dir(rq); |
1da177e4c
|
1745 |
|
22e2c507c
|
1746 1747 |
BUG_ON(!cfqq->allocated[rw]); cfqq->allocated[rw]--; |
1da177e4c
|
1748 |
|
5e7053747
|
1749 |
put_io_context(RQ_CIC(rq)->ioc); |
1da177e4c
|
1750 |
|
1da177e4c
|
1751 |
rq->elevator_private = NULL; |
5e7053747
|
1752 |
rq->elevator_private2 = NULL; |
1da177e4c
|
1753 |
|
1da177e4c
|
1754 1755 1756 1757 1758 |
cfq_put_queue(cfqq); } } /* |
22e2c507c
|
1759 |
* Allocate cfq data structures associated with this request. |
1da177e4c
|
1760 |
*/ |
22e2c507c
|
1761 |
static int |
cb78b285c
|
1762 |
cfq_set_request(request_queue_t *q, struct request *rq, gfp_t gfp_mask) |
1da177e4c
|
1763 1764 |
{ struct cfq_data *cfqd = q->elevator->elevator_data; |
3b18152c3
|
1765 |
struct task_struct *tsk = current; |
1da177e4c
|
1766 1767 |
struct cfq_io_context *cic; const int rw = rq_data_dir(rq); |
7749a8d42
|
1768 1769 |
const int is_sync = rq_is_sync(rq); pid_t key = cfq_queue_pid(tsk, rw, is_sync); |
22e2c507c
|
1770 |
struct cfq_queue *cfqq; |
1da177e4c
|
1771 1772 1773 |
unsigned long flags; might_sleep_if(gfp_mask & __GFP_WAIT); |
e2d74ac06
|
1774 |
cic = cfq_get_io_context(cfqd, gfp_mask); |
22e2c507c
|
1775 |
|
1da177e4c
|
1776 |
spin_lock_irqsave(q->queue_lock, flags); |
22e2c507c
|
1777 1778 |
if (!cic) goto queue_fail; |
12a057321
|
1779 |
if (!cic->cfqq[is_sync]) { |
6f325a134
|
1780 |
cfqq = cfq_get_queue(cfqd, key, tsk, gfp_mask); |
22e2c507c
|
1781 1782 |
if (!cfqq) goto queue_fail; |
1da177e4c
|
1783 |
|
12a057321
|
1784 |
cic->cfqq[is_sync] = cfqq; |
22e2c507c
|
1785 |
} else |
12a057321
|
1786 |
cfqq = cic->cfqq[is_sync]; |
1da177e4c
|
1787 1788 |
cfqq->allocated[rw]++; |
3b18152c3
|
1789 |
cfq_clear_cfqq_must_alloc(cfqq); |
22e2c507c
|
1790 |
atomic_inc(&cfqq->ref); |
1da177e4c
|
1791 |
|
5e7053747
|
1792 |
spin_unlock_irqrestore(q->queue_lock, flags); |
3b18152c3
|
1793 |
|
5e7053747
|
1794 1795 1796 |
rq->elevator_private = cic; rq->elevator_private2 = cfqq; return 0; |
1da177e4c
|
1797 |
|
22e2c507c
|
1798 1799 1800 |
queue_fail: if (cic) put_io_context(cic->ioc); |
89850f7ee
|
1801 |
|
3b18152c3
|
1802 |
cfq_schedule_dispatch(cfqd); |
1da177e4c
|
1803 1804 1805 |
spin_unlock_irqrestore(q->queue_lock, flags); return 1; } |
65f27f384
|
1806 |
static void cfq_kick_queue(struct work_struct *work) |
22e2c507c
|
1807 |
{ |
65f27f384
|
1808 1809 1810 |
struct cfq_data *cfqd = container_of(work, struct cfq_data, unplug_work); request_queue_t *q = cfqd->queue; |
22e2c507c
|
1811 1812 1813 |
unsigned long flags; spin_lock_irqsave(q->queue_lock, flags); |
dc72ef4ae
|
1814 |
blk_start_queueing(q); |
22e2c507c
|
1815 1816 1817 1818 1819 1820 1821 1822 1823 1824 1825 |
spin_unlock_irqrestore(q->queue_lock, flags); } /* * 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
|
1826 |
int timed_out = 1; |
22e2c507c
|
1827 1828 1829 1830 |
spin_lock_irqsave(cfqd->queue->queue_lock, flags); if ((cfqq = cfqd->active_queue) != NULL) { |
3c6bd2f87
|
1831 |
timed_out = 0; |
22e2c507c
|
1832 1833 1834 |
/* * expired */ |
44f7c1606
|
1835 |
if (cfq_slice_used(cfqq)) |
22e2c507c
|
1836 1837 1838 1839 1840 1841 |
goto expire; /* * only expire and reinvoke request handler, if there are * other queues with pending requests */ |
caaa5f9f0
|
1842 |
if (!cfqd->busy_queues) |
22e2c507c
|
1843 |
goto out_cont; |
22e2c507c
|
1844 1845 1846 1847 |
/* * not expired and it has a request pending, let it dispatch */ |
dd67d0515
|
1848 |
if (!RB_EMPTY_ROOT(&cfqq->sort_list)) { |
3b18152c3
|
1849 |
cfq_mark_cfqq_must_dispatch(cfqq); |
22e2c507c
|
1850 1851 1852 1853 |
goto out_kick; } } expire: |
3c6bd2f87
|
1854 |
cfq_slice_expired(cfqd, 0, timed_out); |
22e2c507c
|
1855 |
out_kick: |
3b18152c3
|
1856 |
cfq_schedule_dispatch(cfqd); |
22e2c507c
|
1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 |
out_cont: spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); } /* * Timer running if an idle class queue is waiting for service */ static void cfq_idle_class_timer(unsigned long data) { struct cfq_data *cfqd = (struct cfq_data *) data; unsigned long flags, end; spin_lock_irqsave(cfqd->queue->queue_lock, flags); /* * race with a non-idle queue, reset timer */ end = cfqd->last_end_request + CFQ_IDLE_GRACE; |
ae818a38d
|
1875 1876 1877 |
if (!time_after_eq(jiffies, end)) mod_timer(&cfqd->idle_class_timer, end); else |
3b18152c3
|
1878 |
cfq_schedule_dispatch(cfqd); |
22e2c507c
|
1879 1880 1881 |
spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); } |
3b18152c3
|
1882 1883 1884 1885 1886 1887 |
static void cfq_shutdown_timer_wq(struct cfq_data *cfqd) { del_timer_sync(&cfqd->idle_slice_timer); del_timer_sync(&cfqd->idle_class_timer); blk_sync_queue(cfqd->queue); } |
22e2c507c
|
1888 |
|
1da177e4c
|
1889 1890 |
static void cfq_exit_queue(elevator_t *e) { |
22e2c507c
|
1891 |
struct cfq_data *cfqd = e->elevator_data; |
d9ff41879
|
1892 |
request_queue_t *q = cfqd->queue; |
22e2c507c
|
1893 |
|
3b18152c3
|
1894 |
cfq_shutdown_timer_wq(cfqd); |
e2d74ac06
|
1895 |
|
d9ff41879
|
1896 |
spin_lock_irq(q->queue_lock); |
e2d74ac06
|
1897 |
|
d9ff41879
|
1898 |
if (cfqd->active_queue) |
3c6bd2f87
|
1899 |
__cfq_slice_expired(cfqd, cfqd->active_queue, 0, 0); |
e2d74ac06
|
1900 1901 |
while (!list_empty(&cfqd->cic_list)) { |
d9ff41879
|
1902 1903 1904 |
struct cfq_io_context *cic = list_entry(cfqd->cic_list.next, struct cfq_io_context, queue_list); |
89850f7ee
|
1905 1906 |
__cfq_exit_single_io_context(cfqd, cic); |
d9ff41879
|
1907 |
} |
e2d74ac06
|
1908 |
|
d9ff41879
|
1909 |
spin_unlock_irq(q->queue_lock); |
a90d742e4
|
1910 1911 |
cfq_shutdown_timer_wq(cfqd); |
a90d742e4
|
1912 1913 |
kfree(cfqd->cfq_hash); kfree(cfqd); |
1da177e4c
|
1914 |
} |
bb37b94c6
|
1915 |
static void *cfq_init_queue(request_queue_t *q) |
1da177e4c
|
1916 1917 1918 |
{ struct cfq_data *cfqd; int i; |
b5deef901
|
1919 |
cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node); |
1da177e4c
|
1920 |
if (!cfqd) |
bc1c11697
|
1921 |
return NULL; |
1da177e4c
|
1922 1923 |
memset(cfqd, 0, sizeof(*cfqd)); |
22e2c507c
|
1924 1925 1926 1927 1928 1929 1930 |
for (i = 0; i < CFQ_PRIO_LISTS; i++) INIT_LIST_HEAD(&cfqd->rr_list[i]); INIT_LIST_HEAD(&cfqd->busy_rr); INIT_LIST_HEAD(&cfqd->cur_rr); INIT_LIST_HEAD(&cfqd->idle_rr); |
d9ff41879
|
1931 |
INIT_LIST_HEAD(&cfqd->cic_list); |
1da177e4c
|
1932 |
|
b5deef901
|
1933 |
cfqd->cfq_hash = kmalloc_node(sizeof(struct hlist_head) * CFQ_QHASH_ENTRIES, GFP_KERNEL, q->node); |
1da177e4c
|
1934 |
if (!cfqd->cfq_hash) |
5e7053747
|
1935 |
goto out_free; |
1da177e4c
|
1936 |
|
1da177e4c
|
1937 1938 |
for (i = 0; i < CFQ_QHASH_ENTRIES; i++) INIT_HLIST_HEAD(&cfqd->cfq_hash[i]); |
1da177e4c
|
1939 |
cfqd->queue = q; |
1da177e4c
|
1940 |
|
22e2c507c
|
1941 1942 1943 1944 1945 1946 1947 |
init_timer(&cfqd->idle_slice_timer); cfqd->idle_slice_timer.function = cfq_idle_slice_timer; cfqd->idle_slice_timer.data = (unsigned long) cfqd; init_timer(&cfqd->idle_class_timer); cfqd->idle_class_timer.function = cfq_idle_class_timer; cfqd->idle_class_timer.data = (unsigned long) cfqd; |
65f27f384
|
1948 |
INIT_WORK(&cfqd->unplug_work, cfq_kick_queue); |
22e2c507c
|
1949 |
|
1da177e4c
|
1950 |
cfqd->cfq_quantum = cfq_quantum; |
22e2c507c
|
1951 1952 |
cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0]; cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1]; |
1da177e4c
|
1953 1954 |
cfqd->cfq_back_max = cfq_back_max; cfqd->cfq_back_penalty = cfq_back_penalty; |
22e2c507c
|
1955 1956 1957 1958 |
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; |
3b18152c3
|
1959 |
|
bc1c11697
|
1960 |
return cfqd; |
5e7053747
|
1961 |
out_free: |
1da177e4c
|
1962 |
kfree(cfqd); |
bc1c11697
|
1963 |
return NULL; |
1da177e4c
|
1964 1965 1966 1967 |
} static void cfq_slab_kill(void) { |
1da177e4c
|
1968 1969 1970 1971 1972 1973 1974 1975 |
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) { |
1da177e4c
|
1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 |
cfq_pool = kmem_cache_create("cfq_pool", sizeof(struct cfq_queue), 0, 0, NULL, NULL); if (!cfq_pool) goto fail; cfq_ioc_pool = kmem_cache_create("cfq_ioc_pool", sizeof(struct cfq_io_context), 0, 0, NULL, NULL); if (!cfq_ioc_pool) goto fail; return 0; fail: cfq_slab_kill(); return -ENOMEM; } |
1da177e4c
|
1991 1992 1993 |
/* * sysfs parts below --> */ |
1da177e4c
|
1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 |
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
|
2010 |
#define SHOW_FUNCTION(__FUNC, __VAR, __CONV) \ |
3d1ab40f4
|
2011 |
static ssize_t __FUNC(elevator_t *e, char *page) \ |
1da177e4c
|
2012 |
{ \ |
3d1ab40f4
|
2013 |
struct cfq_data *cfqd = e->elevator_data; \ |
1da177e4c
|
2014 2015 2016 2017 2018 2019 |
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
|
2020 2021 |
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
|
2022 2023 |
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
|
2024 2025 2026 2027 |
SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1); 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); |
1da177e4c
|
2028 2029 2030 |
#undef SHOW_FUNCTION #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV) \ |
3d1ab40f4
|
2031 |
static ssize_t __FUNC(elevator_t *e, const char *page, size_t count) \ |
1da177e4c
|
2032 |
{ \ |
3d1ab40f4
|
2033 |
struct cfq_data *cfqd = e->elevator_data; \ |
1da177e4c
|
2034 2035 2036 2037 2038 2039 2040 2041 2042 2043 2044 2045 2046 |
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); |
22e2c507c
|
2047 2048 |
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
|
2049 2050 |
STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0); STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1, UINT_MAX, 0); |
22e2c507c
|
2051 2052 2053 2054 |
STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1); 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); STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1, UINT_MAX, 0); |
1da177e4c
|
2055 |
#undef STORE_FUNCTION |
e572ec7e4
|
2056 2057 2058 2059 2060 |
#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
|
2061 2062 2063 2064 2065 2066 2067 2068 |
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), |
e572ec7e4
|
2069 |
__ATTR_NULL |
1da177e4c
|
2070 |
}; |
1da177e4c
|
2071 2072 2073 2074 2075 |
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
|
2076 |
.elevator_allow_merge_fn = cfq_allow_merge, |
b4878f245
|
2077 |
.elevator_dispatch_fn = cfq_dispatch_requests, |
1da177e4c
|
2078 |
.elevator_add_req_fn = cfq_insert_request, |
b4878f245
|
2079 |
.elevator_activate_req_fn = cfq_activate_request, |
1da177e4c
|
2080 2081 2082 |
.elevator_deactivate_req_fn = cfq_deactivate_request, .elevator_queue_empty_fn = cfq_queue_empty, .elevator_completed_req_fn = cfq_completed_request, |
21183b07e
|
2083 2084 |
.elevator_former_req_fn = elv_rb_former_request, .elevator_latter_req_fn = elv_rb_latter_request, |
1da177e4c
|
2085 2086 2087 2088 2089 |
.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
|
2090 |
.trim = cfq_free_io_context, |
1da177e4c
|
2091 |
}, |
3d1ab40f4
|
2092 |
.elevator_attrs = cfq_attrs, |
1da177e4c
|
2093 2094 2095 2096 2097 2098 2099 |
.elevator_name = "cfq", .elevator_owner = THIS_MODULE, }; static int __init cfq_init(void) { int ret; |
22e2c507c
|
2100 2101 2102 2103 2104 2105 2106 |
/* * could be 0 on HZ < 1000 setups */ if (!cfq_slice_async) cfq_slice_async = 1; if (!cfq_slice_idle) cfq_slice_idle = 1; |
1da177e4c
|
2107 2108 2109 2110 |
if (cfq_slab_setup()) return -ENOMEM; ret = elv_register(&iosched_cfq); |
22e2c507c
|
2111 2112 |
if (ret) cfq_slab_kill(); |
1da177e4c
|
2113 |
|
1da177e4c
|
2114 2115 2116 2117 2118 |
return ret; } static void __exit cfq_exit(void) { |
6e9a4738c
|
2119 |
DECLARE_COMPLETION_ONSTACK(all_gone); |
1da177e4c
|
2120 |
elv_unregister(&iosched_cfq); |
334e94de9
|
2121 |
ioc_gone = &all_gone; |
fba822722
|
2122 2123 |
/* ioc_gone's update must be visible before reading ioc_count */ smp_wmb(); |
4050cf167
|
2124 |
if (elv_ioc_count_read(ioc_count)) |
fba822722
|
2125 |
wait_for_completion(ioc_gone); |
334e94de9
|
2126 |
synchronize_rcu(); |
83521d3eb
|
2127 |
cfq_slab_kill(); |
1da177e4c
|
2128 2129 2130 2131 2132 2133 2134 2135 |
} module_init(cfq_init); module_exit(cfq_exit); MODULE_AUTHOR("Jens Axboe"); MODULE_LICENSE("GPL"); MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler"); |