Blame view
block/as-iosched.c
37.7 KB
1da177e4c Linux-2.6.12-rc2 |
1 |
/* |
1da177e4c Linux-2.6.12-rc2 |
2 3 |
* Anticipatory & deadline i/o scheduler. * |
0fe234795 [PATCH] Update ax... |
4 |
* Copyright (C) 2002 Jens Axboe <axboe@kernel.dk> |
f5b3db001 [PATCH] as: coope... |
5 |
* Nick Piggin <nickpiggin@yahoo.com.au> |
1da177e4c Linux-2.6.12-rc2 |
6 7 8 9 10 11 12 |
* */ #include <linux/kernel.h> #include <linux/fs.h> #include <linux/blkdev.h> #include <linux/elevator.h> #include <linux/bio.h> |
1da177e4c Linux-2.6.12-rc2 |
13 14 15 16 |
#include <linux/module.h> #include <linux/slab.h> #include <linux/init.h> #include <linux/compiler.h> |
1da177e4c Linux-2.6.12-rc2 |
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 |
#include <linux/rbtree.h> #include <linux/interrupt.h> #define REQ_SYNC 1 #define REQ_ASYNC 0 /* * See Documentation/block/as-iosched.txt */ /* * max time before a read is submitted. */ #define default_read_expire (HZ / 8) /* * ditto for writes, these limits are not hard, even * if the disk is capable of satisfying them. */ #define default_write_expire (HZ / 4) /* * read_batch_expire describes how long we will allow a stream of reads to * persist before looking to see whether it is time to switch over to writes. */ #define default_read_batch_expire (HZ / 2) /* * write_batch_expire describes how long we want a stream of writes to run for. * This is not a hard limit, but a target we set for the auto-tuning thingy. * See, the problem is: we can send a lot of writes to disk cache / TCQ in * a short amount of time... */ #define default_write_batch_expire (HZ / 8) /* * max time we may wait to anticipate a read (default around 6ms) */ #define default_antic_expire ((HZ / 150) ? HZ / 150 : 1) /* * Keep track of up to 20ms thinktimes. We can go as big as we like here, * however huge values tend to interfere and not decay fast enough. A program * might be in a non-io phase of operation. Waiting on user input for example, * or doing a lengthy computation. A small penalty can be justified there, and * will still catch out those processes that constantly have large thinktimes. */ #define MAX_THINKTIME (HZ/50UL) /* Bits in as_io_context.state */ enum as_io_states { |
f5b3db001 [PATCH] as: coope... |
68 |
AS_TASK_RUNNING=0, /* Process has not exited */ |
1da177e4c Linux-2.6.12-rc2 |
69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 |
AS_TASK_IOSTARTED, /* Process has started some IO */ AS_TASK_IORUNNING, /* Process has completed some IO */ }; enum anticipation_status { ANTIC_OFF=0, /* Not anticipating (normal operation) */ ANTIC_WAIT_REQ, /* The last read has not yet completed */ ANTIC_WAIT_NEXT, /* Currently anticipating a request vs last read (which has completed) */ ANTIC_FINISHED, /* Anticipating but have found a candidate * or timed out */ }; struct as_data { /* * run time data */ struct request_queue *q; /* the "owner" queue */ /* * requests (as_rq s) are present on both sort_list and fifo_list */ struct rb_root sort_list[2]; struct list_head fifo_list[2]; |
8a8e674cb [PATCH] as-iosche... |
94 |
struct request *next_rq[2]; /* next in sort order */ |
1da177e4c Linux-2.6.12-rc2 |
95 |
sector_t last_sector[2]; /* last REQ_SYNC & REQ_ASYNC sectors */ |
1da177e4c Linux-2.6.12-rc2 |
96 97 98 |
unsigned long exit_prob; /* probability a task will exit while being waited on */ |
f5b3db001 [PATCH] as: coope... |
99 100 101 |
unsigned long exit_no_coop; /* probablility an exited task will not be part of a later cooperating request */ |
1da177e4c Linux-2.6.12-rc2 |
102 103 104 105 106 107 108 109 110 111 112 113 114 |
unsigned long new_ttime_total; /* mean thinktime on new proc */ unsigned long new_ttime_mean; u64 new_seek_total; /* mean seek on new proc */ sector_t new_seek_mean; unsigned long current_batch_expires; unsigned long last_check_fifo[2]; int changed_batch; /* 1: waiting for old batch to end */ int new_batch; /* 1: waiting on first read complete */ int batch_data_dir; /* current batch REQ_SYNC / REQ_ASYNC */ int write_batch_count; /* max # of reqs in a write batch */ int current_write_count; /* how many requests left this batch */ int write_batch_idled; /* has the write batch gone idle? */ |
1da177e4c Linux-2.6.12-rc2 |
115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 |
enum anticipation_status antic_status; unsigned long antic_start; /* jiffies: when it started */ struct timer_list antic_timer; /* anticipatory scheduling timer */ struct work_struct antic_work; /* Deferred unplugging */ struct io_context *io_context; /* Identify the expected process */ int ioc_finished; /* IO associated with io_context is finished */ int nr_dispatched; /* * settings that change how the i/o scheduler behaves */ unsigned long fifo_expire[2]; unsigned long batch_expire[2]; unsigned long antic_expire; }; |
1da177e4c Linux-2.6.12-rc2 |
131 132 133 134 135 136 137 138 139 140 141 142 143 144 |
/* * per-request data. */ enum arq_state { AS_RQ_NEW=0, /* New - not referenced and not on any lists */ AS_RQ_QUEUED, /* In the request queue. It belongs to the scheduler */ AS_RQ_DISPATCHED, /* On the dispatch list. It belongs to the driver now */ AS_RQ_PRESCHED, /* Debug poisoning for requests being used */ AS_RQ_REMOVED, AS_RQ_MERGED, AS_RQ_POSTSCHED, /* when they shouldn't be */ }; |
8a8e674cb [PATCH] as-iosche... |
145 146 147 |
#define RQ_IOC(rq) ((struct io_context *) (rq)->elevator_private) #define RQ_STATE(rq) ((enum arq_state)(rq)->elevator_private2) #define RQ_SET_STATE(rq, state) ((rq)->elevator_private2 = (void *) state) |
1da177e4c Linux-2.6.12-rc2 |
148 |
|
e4313dd42 [PATCH] as-iosche... |
149 |
static DEFINE_PER_CPU(unsigned long, ioc_count); |
334e94de9 [PATCH] deal with... |
150 |
static struct completion *ioc_gone; |
8a8e674cb [PATCH] as-iosche... |
151 |
static void as_move_to_dispatch(struct as_data *ad, struct request *rq); |
ef9be1d33 [BLOCK] as-iosche... |
152 |
static void as_antic_stop(struct as_data *ad); |
1da177e4c Linux-2.6.12-rc2 |
153 154 155 156 157 158 159 160 |
/* * IO Context helper functions */ /* Called to deallocate the as_io_context */ static void free_as_io_context(struct as_io_context *aic) { kfree(aic); |
e4313dd42 [PATCH] as-iosche... |
161 162 |
elv_ioc_count_dec(ioc_count); if (ioc_gone && !elv_ioc_count_read(ioc_count)) |
334e94de9 [PATCH] deal with... |
163 |
complete(ioc_gone); |
1da177e4c Linux-2.6.12-rc2 |
164 |
} |
e17a9489b [PATCH] stop elv_... |
165 166 |
static void as_trim(struct io_context *ioc) { |
334e94de9 [PATCH] deal with... |
167 168 |
if (ioc->aic) free_as_io_context(ioc->aic); |
e17a9489b [PATCH] stop elv_... |
169 170 |
ioc->aic = NULL; } |
1da177e4c Linux-2.6.12-rc2 |
171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 |
/* Called when the task exits */ static void exit_as_io_context(struct as_io_context *aic) { WARN_ON(!test_bit(AS_TASK_RUNNING, &aic->state)); clear_bit(AS_TASK_RUNNING, &aic->state); } static struct as_io_context *alloc_as_io_context(void) { struct as_io_context *ret; ret = kmalloc(sizeof(*ret), GFP_ATOMIC); if (ret) { ret->dtor = free_as_io_context; ret->exit = exit_as_io_context; ret->state = 1 << AS_TASK_RUNNING; atomic_set(&ret->nr_queued, 0); atomic_set(&ret->nr_dispatched, 0); spin_lock_init(&ret->lock); ret->ttime_total = 0; ret->ttime_samples = 0; ret->ttime_mean = 0; ret->seek_total = 0; ret->seek_samples = 0; ret->seek_mean = 0; |
e4313dd42 [PATCH] as-iosche... |
196 |
elv_ioc_count_inc(ioc_count); |
1da177e4c Linux-2.6.12-rc2 |
197 198 199 200 201 202 203 204 205 |
} return ret; } /* * If the current task has no AS IO context then create one and initialise it. * Then take a ref on the task's io context and return it. */ |
b5deef901 [PATCH] Make sure... |
206 |
static struct io_context *as_get_io_context(int node) |
1da177e4c Linux-2.6.12-rc2 |
207 |
{ |
b5deef901 [PATCH] Make sure... |
208 |
struct io_context *ioc = get_io_context(GFP_ATOMIC, node); |
1da177e4c Linux-2.6.12-rc2 |
209 210 211 212 213 214 215 216 217 |
if (ioc && !ioc->aic) { ioc->aic = alloc_as_io_context(); if (!ioc->aic) { put_io_context(ioc); ioc = NULL; } } return ioc; } |
8a8e674cb [PATCH] as-iosche... |
218 |
static void as_put_io_context(struct request *rq) |
b4878f245 [PATCH] 02/05: up... |
219 220 |
{ struct as_io_context *aic; |
8a8e674cb [PATCH] as-iosche... |
221 |
if (unlikely(!RQ_IOC(rq))) |
b4878f245 [PATCH] 02/05: up... |
222 |
return; |
8a8e674cb [PATCH] as-iosche... |
223 |
aic = RQ_IOC(rq)->aic; |
b4878f245 [PATCH] 02/05: up... |
224 |
|
8a8e674cb [PATCH] as-iosche... |
225 |
if (rq_is_sync(rq) && aic) { |
b4878f245 [PATCH] 02/05: up... |
226 227 228 229 230 |
spin_lock(&aic->lock); set_bit(AS_TASK_IORUNNING, &aic->state); aic->last_end_request = jiffies; spin_unlock(&aic->lock); } |
8a8e674cb [PATCH] as-iosche... |
231 |
put_io_context(RQ_IOC(rq)); |
b4878f245 [PATCH] 02/05: up... |
232 |
} |
1da177e4c Linux-2.6.12-rc2 |
233 |
/* |
1da177e4c Linux-2.6.12-rc2 |
234 235 |
* rb tree support functions */ |
9e2585a8a [PATCH] as-iosche... |
236 |
#define RQ_RB_ROOT(ad, rq) (&(ad)->sort_list[rq_is_sync((rq))]) |
1da177e4c Linux-2.6.12-rc2 |
237 |
|
8a8e674cb [PATCH] as-iosche... |
238 |
static void as_add_rq_rb(struct as_data *ad, struct request *rq) |
ef9be1d33 [BLOCK] as-iosche... |
239 |
{ |
e37f346e3 [PATCH] as-iosche... |
240 |
struct request *alias; |
ef9be1d33 [BLOCK] as-iosche... |
241 |
|
9e2585a8a [PATCH] as-iosche... |
242 |
while ((unlikely(alias = elv_rb_add(RQ_RB_ROOT(ad, rq), rq)))) { |
8a8e674cb [PATCH] as-iosche... |
243 |
as_move_to_dispatch(ad, alias); |
ef9be1d33 [BLOCK] as-iosche... |
244 245 246 |
as_antic_stop(ad); } } |
8a8e674cb [PATCH] as-iosche... |
247 |
static inline void as_del_rq_rb(struct as_data *ad, struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
248 |
{ |
9e2585a8a [PATCH] as-iosche... |
249 |
elv_rb_del(RQ_RB_ROOT(ad, rq), rq); |
1da177e4c Linux-2.6.12-rc2 |
250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 |
} /* * IO Scheduler proper */ #define MAXBACK (1024 * 1024) /* * Maximum distance the disk will go backward * for a request. */ #define BACK_PENALTY 2 /* * as_choose_req selects the preferred one of two requests of the same data_dir * ignoring time - eg. timeouts, which is the job of as_dispatch_request */ |
8a8e674cb [PATCH] as-iosche... |
267 268 |
static struct request * as_choose_req(struct as_data *ad, struct request *rq1, struct request *rq2) |
1da177e4c Linux-2.6.12-rc2 |
269 270 271 272 273 |
{ int data_dir; sector_t last, s1, s2, d1, d2; int r1_wrap=0, r2_wrap=0; /* requests are behind the disk head */ const sector_t maxback = MAXBACK; |
8a8e674cb [PATCH] as-iosche... |
274 275 276 277 |
if (rq1 == NULL || rq1 == rq2) return rq2; if (rq2 == NULL) return rq1; |
1da177e4c Linux-2.6.12-rc2 |
278 |
|
8a8e674cb [PATCH] as-iosche... |
279 |
data_dir = rq_is_sync(rq1); |
1da177e4c Linux-2.6.12-rc2 |
280 281 |
last = ad->last_sector[data_dir]; |
8a8e674cb [PATCH] as-iosche... |
282 283 |
s1 = rq1->sector; s2 = rq2->sector; |
1da177e4c Linux-2.6.12-rc2 |
284 |
|
8a8e674cb [PATCH] as-iosche... |
285 |
BUG_ON(data_dir != rq_is_sync(rq2)); |
1da177e4c Linux-2.6.12-rc2 |
286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 |
/* * 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+maxback >= last) d1 = (last - s1)*BACK_PENALTY; else { r1_wrap = 1; d1 = 0; /* shut up, gcc */ } if (s2 >= last) d2 = s2 - last; else if (s2+maxback >= last) d2 = (last - s2)*BACK_PENALTY; else { r2_wrap = 1; d2 = 0; } /* Found required data */ if (!r1_wrap && r2_wrap) |
8a8e674cb [PATCH] as-iosche... |
312 |
return rq1; |
1da177e4c Linux-2.6.12-rc2 |
313 |
else if (!r2_wrap && r1_wrap) |
8a8e674cb [PATCH] as-iosche... |
314 |
return rq2; |
1da177e4c Linux-2.6.12-rc2 |
315 316 317 |
else if (r1_wrap && r2_wrap) { /* both behind the head */ if (s1 <= s2) |
8a8e674cb [PATCH] as-iosche... |
318 |
return rq1; |
1da177e4c Linux-2.6.12-rc2 |
319 |
else |
8a8e674cb [PATCH] as-iosche... |
320 |
return rq2; |
1da177e4c Linux-2.6.12-rc2 |
321 322 323 324 |
} /* Both requests in front of the head */ if (d1 < d2) |
8a8e674cb [PATCH] as-iosche... |
325 |
return rq1; |
1da177e4c Linux-2.6.12-rc2 |
326 |
else if (d2 < d1) |
8a8e674cb [PATCH] as-iosche... |
327 |
return rq2; |
1da177e4c Linux-2.6.12-rc2 |
328 329 |
else { if (s1 >= s2) |
8a8e674cb [PATCH] as-iosche... |
330 |
return rq1; |
1da177e4c Linux-2.6.12-rc2 |
331 |
else |
8a8e674cb [PATCH] as-iosche... |
332 |
return rq2; |
1da177e4c Linux-2.6.12-rc2 |
333 334 335 336 |
} } /* |
8a8e674cb [PATCH] as-iosche... |
337 |
* as_find_next_rq finds the next request after @prev in elevator order. |
1da177e4c Linux-2.6.12-rc2 |
338 339 340 |
* this with as_choose_req form the basis for how the scheduler chooses * what request to process next. Anticipation works on top of this. */ |
8a8e674cb [PATCH] as-iosche... |
341 342 |
static struct request * as_find_next_rq(struct as_data *ad, struct request *last) |
1da177e4c Linux-2.6.12-rc2 |
343 |
{ |
1da177e4c Linux-2.6.12-rc2 |
344 345 |
struct rb_node *rbnext = rb_next(&last->rb_node); struct rb_node *rbprev = rb_prev(&last->rb_node); |
8a8e674cb [PATCH] as-iosche... |
346 |
struct request *next = NULL, *prev = NULL; |
1da177e4c Linux-2.6.12-rc2 |
347 |
|
e37f346e3 [PATCH] as-iosche... |
348 |
BUG_ON(RB_EMPTY_NODE(&last->rb_node)); |
1da177e4c Linux-2.6.12-rc2 |
349 350 |
if (rbprev) |
8a8e674cb [PATCH] as-iosche... |
351 |
prev = rb_entry_rq(rbprev); |
1da177e4c Linux-2.6.12-rc2 |
352 353 |
if (rbnext) |
8a8e674cb [PATCH] as-iosche... |
354 |
next = rb_entry_rq(rbnext); |
1da177e4c Linux-2.6.12-rc2 |
355 |
else { |
9e2585a8a [PATCH] as-iosche... |
356 |
const int data_dir = rq_is_sync(last); |
1da177e4c Linux-2.6.12-rc2 |
357 |
|
e37f346e3 [PATCH] as-iosche... |
358 359 |
rbnext = rb_first(&ad->sort_list[data_dir]); if (rbnext && rbnext != &last->rb_node) |
8a8e674cb [PATCH] as-iosche... |
360 |
next = rb_entry_rq(rbnext); |
e37f346e3 [PATCH] as-iosche... |
361 |
} |
1da177e4c Linux-2.6.12-rc2 |
362 |
|
e37f346e3 [PATCH] as-iosche... |
363 |
return as_choose_req(ad, next, prev); |
1da177e4c Linux-2.6.12-rc2 |
364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 |
} /* * anticipatory scheduling functions follow */ /* * as_antic_expired tells us when we have anticipated too long. * The funny "absolute difference" math on the elapsed time is to handle * jiffy wraps, and disks which have been idle for 0x80000000 jiffies. */ static int as_antic_expired(struct as_data *ad) { long delta_jif; delta_jif = jiffies - ad->antic_start; if (unlikely(delta_jif < 0)) delta_jif = -delta_jif; if (delta_jif < ad->antic_expire) return 0; return 1; } /* * as_antic_waitnext starts anticipating that a nice request will soon be * submitted. See also as_antic_waitreq */ static void as_antic_waitnext(struct as_data *ad) { unsigned long timeout; BUG_ON(ad->antic_status != ANTIC_OFF && ad->antic_status != ANTIC_WAIT_REQ); timeout = ad->antic_start + ad->antic_expire; mod_timer(&ad->antic_timer, timeout); ad->antic_status = ANTIC_WAIT_NEXT; } /* * as_antic_waitreq starts anticipating. We don't start timing the anticipation * until the request that we're anticipating on has finished. This means we * are timing from when the candidate process wakes up hopefully. */ static void as_antic_waitreq(struct as_data *ad) { BUG_ON(ad->antic_status == ANTIC_FINISHED); if (ad->antic_status == ANTIC_OFF) { if (!ad->io_context || ad->ioc_finished) as_antic_waitnext(ad); else ad->antic_status = ANTIC_WAIT_REQ; } } /* * This is called directly by the functions in this file to stop anticipation. * We kill the timer and schedule a call to the request_fn asap. */ static void as_antic_stop(struct as_data *ad) { int status = ad->antic_status; if (status == ANTIC_WAIT_REQ || status == ANTIC_WAIT_NEXT) { if (status == ANTIC_WAIT_NEXT) del_timer(&ad->antic_timer); ad->antic_status = ANTIC_FINISHED; /* see as_work_handler */ kblockd_schedule_work(&ad->antic_work); } } /* * as_antic_timeout is the timer function set by as_antic_waitnext. */ static void as_antic_timeout(unsigned long data) { struct request_queue *q = (struct request_queue *)data; struct as_data *ad = q->elevator->elevator_data; unsigned long flags; spin_lock_irqsave(q->queue_lock, flags); if (ad->antic_status == ANTIC_WAIT_REQ || ad->antic_status == ANTIC_WAIT_NEXT) { struct as_io_context *aic = ad->io_context->aic; ad->antic_status = ANTIC_FINISHED; kblockd_schedule_work(&ad->antic_work); if (aic->ttime_samples == 0) { |
f5b3db001 [PATCH] as: coope... |
457 |
/* process anticipated on has exited or timed out*/ |
1da177e4c Linux-2.6.12-rc2 |
458 459 |
ad->exit_prob = (7*ad->exit_prob + 256)/8; } |
f5b3db001 [PATCH] as: coope... |
460 461 462 463 |
if (!test_bit(AS_TASK_RUNNING, &aic->state)) { /* process not "saved" by a cooperating request */ ad->exit_no_coop = (7*ad->exit_no_coop + 256)/8; } |
1da177e4c Linux-2.6.12-rc2 |
464 465 466 |
} spin_unlock_irqrestore(q->queue_lock, flags); } |
f5b3db001 [PATCH] as: coope... |
467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 |
static void as_update_thinktime(struct as_data *ad, struct as_io_context *aic, unsigned long ttime) { /* fixed point: 1.0 == 1<<8 */ if (aic->ttime_samples == 0) { ad->new_ttime_total = (7*ad->new_ttime_total + 256*ttime) / 8; ad->new_ttime_mean = ad->new_ttime_total / 256; ad->exit_prob = (7*ad->exit_prob)/8; } aic->ttime_samples = (7*aic->ttime_samples + 256) / 8; aic->ttime_total = (7*aic->ttime_total + 256*ttime) / 8; aic->ttime_mean = (aic->ttime_total + 128) / aic->ttime_samples; } static void as_update_seekdist(struct as_data *ad, struct as_io_context *aic, sector_t sdist) { u64 total; if (aic->seek_samples == 0) { ad->new_seek_total = (7*ad->new_seek_total + 256*(u64)sdist)/8; ad->new_seek_mean = ad->new_seek_total / 256; } /* * Don't allow the seek distance to get too large from the * odd fragment, pagein, etc */ if (aic->seek_samples <= 60) /* second&third seek */ sdist = min(sdist, (aic->seek_mean * 4) + 2*1024*1024); else sdist = min(sdist, (aic->seek_mean * 4) + 2*1024*64); aic->seek_samples = (7*aic->seek_samples + 256) / 8; aic->seek_total = (7*aic->seek_total + (u64)256*sdist) / 8; total = aic->seek_total + (aic->seek_samples/2); do_div(total, aic->seek_samples); aic->seek_mean = (sector_t)total; } /* * as_update_iohist keeps a decaying histogram of IO thinktimes, and * updates @aic->ttime_mean based on that. It is called when a new * request is queued. */ static void as_update_iohist(struct as_data *ad, struct as_io_context *aic, struct request *rq) { |
9e2585a8a [PATCH] as-iosche... |
516 |
int data_dir = rq_is_sync(rq); |
f5b3db001 [PATCH] as: coope... |
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 |
unsigned long thinktime = 0; sector_t seek_dist; if (aic == NULL) return; if (data_dir == REQ_SYNC) { unsigned long in_flight = atomic_read(&aic->nr_queued) + atomic_read(&aic->nr_dispatched); spin_lock(&aic->lock); if (test_bit(AS_TASK_IORUNNING, &aic->state) || test_bit(AS_TASK_IOSTARTED, &aic->state)) { /* Calculate read -> read thinktime */ if (test_bit(AS_TASK_IORUNNING, &aic->state) && in_flight == 0) { thinktime = jiffies - aic->last_end_request; thinktime = min(thinktime, MAX_THINKTIME-1); } as_update_thinktime(ad, aic, thinktime); /* Calculate read -> read seek distance */ if (aic->last_request_pos < rq->sector) seek_dist = rq->sector - aic->last_request_pos; else seek_dist = aic->last_request_pos - rq->sector; as_update_seekdist(ad, aic, seek_dist); } aic->last_request_pos = rq->sector + rq->nr_sectors; set_bit(AS_TASK_IOSTARTED, &aic->state); spin_unlock(&aic->lock); } } |
1da177e4c Linux-2.6.12-rc2 |
549 550 551 552 |
/* * as_close_req decides if one request is considered "close" to the * previous one issued. */ |
f5b3db001 [PATCH] as: coope... |
553 |
static int as_close_req(struct as_data *ad, struct as_io_context *aic, |
8a8e674cb [PATCH] as-iosche... |
554 |
struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
555 |
{ |
c6a632a2b as: fix antic_exp... |
556 |
unsigned long delay; /* jiffies */ |
1da177e4c Linux-2.6.12-rc2 |
557 |
sector_t last = ad->last_sector[ad->batch_data_dir]; |
8a8e674cb [PATCH] as-iosche... |
558 |
sector_t next = rq->sector; |
1da177e4c Linux-2.6.12-rc2 |
559 |
sector_t delta; /* acceptable close offset (in sectors) */ |
f5b3db001 [PATCH] as: coope... |
560 |
sector_t s; |
1da177e4c Linux-2.6.12-rc2 |
561 562 563 564 |
if (ad->antic_status == ANTIC_OFF || !ad->ioc_finished) delay = 0; else |
c6a632a2b as: fix antic_exp... |
565 |
delay = jiffies - ad->antic_start; |
1da177e4c Linux-2.6.12-rc2 |
566 |
|
f5b3db001 [PATCH] as: coope... |
567 568 |
if (delay == 0) delta = 8192; |
c6a632a2b as: fix antic_exp... |
569 |
else if (delay <= (20 * HZ / 1000) && delay <= ad->antic_expire) |
f5b3db001 [PATCH] as: coope... |
570 |
delta = 8192 << delay; |
1da177e4c Linux-2.6.12-rc2 |
571 572 |
else return 1; |
f5b3db001 [PATCH] as: coope... |
573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 |
if ((last <= next + (delta>>1)) && (next <= last + delta)) return 1; if (last < next) s = next - last; else s = last - next; if (aic->seek_samples == 0) { /* * Process has just started IO. Use past statistics to * gauge success possibility */ if (ad->new_seek_mean > s) { /* this request is better than what we're expecting */ return 1; } } else { if (aic->seek_mean > s) { /* this request is better than what we're expecting */ return 1; } } return 0; |
1da177e4c Linux-2.6.12-rc2 |
599 600 601 602 603 604 605 606 607 608 609 |
} /* * as_can_break_anticipation returns true if we have been anticipating this * request. * * It also returns true if the process against which we are anticipating * submits a write - that's presumably an fsync, O_SYNC write, etc. We want to * dispatch it ASAP, because we know that application will not be submitting * any new reads. * |
f5b3db001 [PATCH] as: coope... |
610 |
* If the task which has submitted the request has exited, break anticipation. |
1da177e4c Linux-2.6.12-rc2 |
611 612 613 |
* * If this task has queued some other IO, do not enter enticipation. */ |
8a8e674cb [PATCH] as-iosche... |
614 |
static int as_can_break_anticipation(struct as_data *ad, struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
615 616 617 |
{ struct io_context *ioc; struct as_io_context *aic; |
1da177e4c Linux-2.6.12-rc2 |
618 619 620 |
ioc = ad->io_context; BUG_ON(!ioc); |
8a8e674cb [PATCH] as-iosche... |
621 |
if (rq && ioc == RQ_IOC(rq)) { |
1da177e4c Linux-2.6.12-rc2 |
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 |
/* request from same process */ return 1; } if (ad->ioc_finished && as_antic_expired(ad)) { /* * In this situation status should really be FINISHED, * however the timer hasn't had the chance to run yet. */ return 1; } aic = ioc->aic; if (!aic) return 0; |
1da177e4c Linux-2.6.12-rc2 |
637 638 639 640 641 642 643 644 645 |
if (atomic_read(&aic->nr_queued) > 0) { /* process has more requests queued */ return 1; } if (atomic_read(&aic->nr_dispatched) > 0) { /* process has more requests dispatched */ return 1; } |
8a8e674cb [PATCH] as-iosche... |
646 |
if (rq && rq_is_sync(rq) && as_close_req(ad, aic, rq)) { |
1da177e4c Linux-2.6.12-rc2 |
647 648 649 |
/* * Found a close request that is not one of ours. * |
f5b3db001 [PATCH] as: coope... |
650 651 |
* This makes close requests from another process update * our IO history. Is generally useful when there are |
1da177e4c Linux-2.6.12-rc2 |
652 653 654 |
* two or more cooperating processes working in the same * area. */ |
f5b3db001 [PATCH] as: coope... |
655 656 657 658 659 660 |
if (!test_bit(AS_TASK_RUNNING, &aic->state)) { if (aic->ttime_samples == 0) ad->exit_prob = (7*ad->exit_prob + 256)/8; ad->exit_no_coop = (7*ad->exit_no_coop)/8; } |
8a8e674cb [PATCH] as-iosche... |
661 |
as_update_iohist(ad, aic, rq); |
1da177e4c Linux-2.6.12-rc2 |
662 663 |
return 1; } |
f5b3db001 [PATCH] as: coope... |
664 665 666 667 668 669 670 671 |
if (!test_bit(AS_TASK_RUNNING, &aic->state)) { /* process anticipated on has exited */ if (aic->ttime_samples == 0) ad->exit_prob = (7*ad->exit_prob + 256)/8; if (ad->exit_no_coop > 128) return 1; } |
1da177e4c Linux-2.6.12-rc2 |
672 673 674 675 |
if (aic->ttime_samples == 0) { if (ad->new_ttime_mean > ad->antic_expire) return 1; |
f5b3db001 [PATCH] as: coope... |
676 |
if (ad->exit_prob * ad->exit_no_coop > 128*256) |
1da177e4c Linux-2.6.12-rc2 |
677 678 679 680 681 |
return 1; } else if (aic->ttime_mean > ad->antic_expire) { /* the process thinks too much between requests */ return 1; } |
1da177e4c Linux-2.6.12-rc2 |
682 683 684 685 |
return 0; } /* |
8a8e674cb [PATCH] as-iosche... |
686 |
* as_can_anticipate indicates whether we should either run rq |
1da177e4c Linux-2.6.12-rc2 |
687 688 |
* or keep anticipating a better request. */ |
8a8e674cb [PATCH] as-iosche... |
689 |
static int as_can_anticipate(struct as_data *ad, struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
690 691 692 693 694 695 696 697 698 699 700 701 |
{ if (!ad->io_context) /* * Last request submitted was a write */ return 0; if (ad->antic_status == ANTIC_FINISHED) /* * Don't restart if we have just finished. Run the next request */ return 0; |
8a8e674cb [PATCH] as-iosche... |
702 |
if (as_can_break_anticipation(ad, rq)) |
1da177e4c Linux-2.6.12-rc2 |
703 704 705 706 707 708 709 710 711 712 713 |
/* * This request is a good candidate. Don't keep anticipating, * run it. */ return 0; /* * OK from here, we haven't finished, and don't have a decent request! * Status is either ANTIC_OFF so start waiting, * ANTIC_WAIT_REQ so continue waiting for request to finish * or ANTIC_WAIT_NEXT so continue waiting for an acceptable request. |
1da177e4c Linux-2.6.12-rc2 |
714 715 716 717 |
*/ return 1; } |
1da177e4c Linux-2.6.12-rc2 |
718 |
/* |
8a8e674cb [PATCH] as-iosche... |
719 |
* as_update_rq must be called whenever a request (rq) is added to |
1da177e4c Linux-2.6.12-rc2 |
720 721 722 |
* the sort_list. This function keeps caches up to date, and checks if the * request might be one we are "anticipating" */ |
8a8e674cb [PATCH] as-iosche... |
723 |
static void as_update_rq(struct as_data *ad, struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
724 |
{ |
8a8e674cb [PATCH] as-iosche... |
725 |
const int data_dir = rq_is_sync(rq); |
1da177e4c Linux-2.6.12-rc2 |
726 |
|
8a8e674cb [PATCH] as-iosche... |
727 728 |
/* keep the next_rq cache up to date */ ad->next_rq[data_dir] = as_choose_req(ad, rq, ad->next_rq[data_dir]); |
1da177e4c Linux-2.6.12-rc2 |
729 730 731 732 733 734 735 736 |
/* * have we been anticipating this request? * or does it come from the same process as the one we are anticipating * for? */ if (ad->antic_status == ANTIC_WAIT_REQ || ad->antic_status == ANTIC_WAIT_NEXT) { |
8a8e674cb [PATCH] as-iosche... |
737 |
if (as_can_break_anticipation(ad, rq)) |
1da177e4c Linux-2.6.12-rc2 |
738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 |
as_antic_stop(ad); } } /* * Gathers timings and resizes the write batch automatically */ static void update_write_batch(struct as_data *ad) { unsigned long batch = ad->batch_expire[REQ_ASYNC]; long write_time; write_time = (jiffies - ad->current_batch_expires) + batch; if (write_time < 0) write_time = 0; if (write_time > batch && !ad->write_batch_idled) { if (write_time > batch * 3) ad->write_batch_count /= 2; else ad->write_batch_count--; } else if (write_time < batch && ad->current_write_count == 0) { if (batch > write_time * 3) ad->write_batch_count *= 2; else ad->write_batch_count++; } if (ad->write_batch_count < 1) ad->write_batch_count = 1; } /* * as_completed_request is to be called when a request has completed and * returned something to the requesting process, be it an error or data. */ static void as_completed_request(request_queue_t *q, struct request *rq) { struct as_data *ad = q->elevator->elevator_data; |
1da177e4c Linux-2.6.12-rc2 |
777 778 |
WARN_ON(!list_empty(&rq->queuelist)); |
8a8e674cb [PATCH] as-iosche... |
779 780 781 |
if (RQ_STATE(rq) != AS_RQ_REMOVED) { printk("rq->state %d ", RQ_STATE(rq)); |
1da177e4c Linux-2.6.12-rc2 |
782 783 784 |
WARN_ON(1); goto out; } |
1da177e4c Linux-2.6.12-rc2 |
785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 |
if (ad->changed_batch && ad->nr_dispatched == 1) { kblockd_schedule_work(&ad->antic_work); ad->changed_batch = 0; if (ad->batch_data_dir == REQ_SYNC) ad->new_batch = 1; } WARN_ON(ad->nr_dispatched == 0); ad->nr_dispatched--; /* * Start counting the batch from when a request of that direction is * actually serviced. This should help devices with big TCQ windows * and writeback caches */ |
9e2585a8a [PATCH] as-iosche... |
800 |
if (ad->new_batch && ad->batch_data_dir == rq_is_sync(rq)) { |
1da177e4c Linux-2.6.12-rc2 |
801 802 803 804 805 |
update_write_batch(ad); ad->current_batch_expires = jiffies + ad->batch_expire[REQ_SYNC]; ad->new_batch = 0; } |
8a8e674cb [PATCH] as-iosche... |
806 |
if (ad->io_context == RQ_IOC(rq) && ad->io_context) { |
1da177e4c Linux-2.6.12-rc2 |
807 808 809 810 811 812 813 814 815 816 |
ad->antic_start = jiffies; ad->ioc_finished = 1; if (ad->antic_status == ANTIC_WAIT_REQ) { /* * We were waiting on this request, now anticipate * the next one */ as_antic_waitnext(ad); } } |
8a8e674cb [PATCH] as-iosche... |
817 |
as_put_io_context(rq); |
1da177e4c Linux-2.6.12-rc2 |
818 |
out: |
8a8e674cb [PATCH] as-iosche... |
819 |
RQ_SET_STATE(rq, AS_RQ_POSTSCHED); |
1da177e4c Linux-2.6.12-rc2 |
820 821 822 823 824 825 826 827 828 829 |
} /* * as_remove_queued_request removes a request from the pre dispatch queue * without updating refcounts. It is expected the caller will drop the * reference unless it replaces the request at somepart of the elevator * (ie. the dispatch queue) */ static void as_remove_queued_request(request_queue_t *q, struct request *rq) { |
9e2585a8a [PATCH] as-iosche... |
830 |
const int data_dir = rq_is_sync(rq); |
1da177e4c Linux-2.6.12-rc2 |
831 |
struct as_data *ad = q->elevator->elevator_data; |
8a8e674cb [PATCH] as-iosche... |
832 |
struct io_context *ioc; |
1da177e4c Linux-2.6.12-rc2 |
833 |
|
8a8e674cb [PATCH] as-iosche... |
834 |
WARN_ON(RQ_STATE(rq) != AS_RQ_QUEUED); |
1da177e4c Linux-2.6.12-rc2 |
835 |
|
8a8e674cb [PATCH] as-iosche... |
836 837 838 839 |
ioc = RQ_IOC(rq); if (ioc && ioc->aic) { BUG_ON(!atomic_read(&ioc->aic->nr_queued)); atomic_dec(&ioc->aic->nr_queued); |
1da177e4c Linux-2.6.12-rc2 |
840 841 842 |
} /* |
8a8e674cb [PATCH] as-iosche... |
843 |
* Update the "next_rq" cache if we are about to remove its |
1da177e4c Linux-2.6.12-rc2 |
844 845 |
* entry */ |
8a8e674cb [PATCH] as-iosche... |
846 847 |
if (ad->next_rq[data_dir] == rq) ad->next_rq[data_dir] = as_find_next_rq(ad, rq); |
1da177e4c Linux-2.6.12-rc2 |
848 |
|
d4f2f4629 [PATCH] as-iosche... |
849 |
rq_fifo_clear(rq); |
8a8e674cb [PATCH] as-iosche... |
850 |
as_del_rq_rb(ad, rq); |
1da177e4c Linux-2.6.12-rc2 |
851 852 853 |
} /* |
1da177e4c Linux-2.6.12-rc2 |
854 855 856 857 858 859 860 861 862 |
* as_fifo_expired returns 0 if there are no expired reads on the fifo, * 1 otherwise. It is ratelimited so that we only perform the check once per * `fifo_expire' interval. Otherwise a large number of expired requests * would create a hopeless seekstorm. * * See as_antic_expired comment. */ static int as_fifo_expired(struct as_data *ad, int adir) { |
d4f2f4629 [PATCH] as-iosche... |
863 |
struct request *rq; |
1da177e4c Linux-2.6.12-rc2 |
864 865 866 867 868 869 870 871 872 873 874 875 |
long delta_jif; delta_jif = jiffies - ad->last_check_fifo[adir]; if (unlikely(delta_jif < 0)) delta_jif = -delta_jif; if (delta_jif < ad->fifo_expire[adir]) return 0; ad->last_check_fifo[adir] = jiffies; if (list_empty(&ad->fifo_list[adir])) return 0; |
d4f2f4629 [PATCH] as-iosche... |
876 |
rq = rq_entry_fifo(ad->fifo_list[adir].next); |
1da177e4c Linux-2.6.12-rc2 |
877 |
|
d4f2f4629 [PATCH] as-iosche... |
878 |
return time_after(jiffies, rq_fifo_time(rq)); |
1da177e4c Linux-2.6.12-rc2 |
879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 |
} /* * as_batch_expired returns true if the current batch has expired. A batch * is a set of reads or a set of writes. */ static inline int as_batch_expired(struct as_data *ad) { if (ad->changed_batch || ad->new_batch) return 0; if (ad->batch_data_dir == REQ_SYNC) /* TODO! add a check so a complete fifo gets written? */ return time_after(jiffies, ad->current_batch_expires); return time_after(jiffies, ad->current_batch_expires) || ad->current_write_count == 0; } /* * move an entry to dispatch queue */ |
8a8e674cb [PATCH] as-iosche... |
901 |
static void as_move_to_dispatch(struct as_data *ad, struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
902 |
{ |
9e2585a8a [PATCH] as-iosche... |
903 |
const int data_dir = rq_is_sync(rq); |
1da177e4c Linux-2.6.12-rc2 |
904 |
|
e37f346e3 [PATCH] as-iosche... |
905 |
BUG_ON(RB_EMPTY_NODE(&rq->rb_node)); |
1da177e4c Linux-2.6.12-rc2 |
906 907 908 909 910 911 |
as_antic_stop(ad); ad->antic_status = ANTIC_OFF; /* * This has to be set in order to be correctly updated by |
8a8e674cb [PATCH] as-iosche... |
912 |
* as_find_next_rq |
1da177e4c Linux-2.6.12-rc2 |
913 914 915 916 |
*/ ad->last_sector[data_dir] = rq->sector + rq->nr_sectors; if (data_dir == REQ_SYNC) { |
8a8e674cb [PATCH] as-iosche... |
917 |
struct io_context *ioc = RQ_IOC(rq); |
1da177e4c Linux-2.6.12-rc2 |
918 |
/* In case we have to anticipate after this */ |
8a8e674cb [PATCH] as-iosche... |
919 |
copy_io_context(&ad->io_context, &ioc); |
1da177e4c Linux-2.6.12-rc2 |
920 921 922 923 924 925 926 927 928 929 |
} else { if (ad->io_context) { put_io_context(ad->io_context); ad->io_context = NULL; } if (ad->current_write_count != 0) ad->current_write_count--; } ad->ioc_finished = 0; |
8a8e674cb [PATCH] as-iosche... |
930 |
ad->next_rq[data_dir] = as_find_next_rq(ad, rq); |
1da177e4c Linux-2.6.12-rc2 |
931 932 933 934 |
/* * take it off the sort and fifo list, add to dispatch queue */ |
1da177e4c Linux-2.6.12-rc2 |
935 |
as_remove_queued_request(ad->q, rq); |
8a8e674cb [PATCH] as-iosche... |
936 |
WARN_ON(RQ_STATE(rq) != AS_RQ_QUEUED); |
1da177e4c Linux-2.6.12-rc2 |
937 |
|
b4878f245 [PATCH] 02/05: up... |
938 |
elv_dispatch_sort(ad->q, rq); |
8a8e674cb [PATCH] as-iosche... |
939 940 941 |
RQ_SET_STATE(rq, AS_RQ_DISPATCHED); if (RQ_IOC(rq) && RQ_IOC(rq)->aic) atomic_inc(&RQ_IOC(rq)->aic->nr_dispatched); |
1da177e4c Linux-2.6.12-rc2 |
942 943 944 945 946 947 948 949 |
ad->nr_dispatched++; } /* * as_dispatch_request selects the best request according to * read/write expire, batch expire, etc, and moves it to the dispatch * queue. Returns 1 if a request was found, 0 otherwise. */ |
b4878f245 [PATCH] 02/05: up... |
950 |
static int as_dispatch_request(request_queue_t *q, int force) |
1da177e4c Linux-2.6.12-rc2 |
951 |
{ |
b4878f245 [PATCH] 02/05: up... |
952 |
struct as_data *ad = q->elevator->elevator_data; |
1da177e4c Linux-2.6.12-rc2 |
953 954 |
const int reads = !list_empty(&ad->fifo_list[REQ_SYNC]); const int writes = !list_empty(&ad->fifo_list[REQ_ASYNC]); |
8a8e674cb [PATCH] as-iosche... |
955 |
struct request *rq; |
1da177e4c Linux-2.6.12-rc2 |
956 |
|
b4878f245 [PATCH] 02/05: up... |
957 958 959 960 961 962 963 964 965 966 967 968 969 |
if (unlikely(force)) { /* * Forced dispatch, accounting is useless. Reset * accounting states and dump fifo_lists. Note that * batch_data_dir is reset to REQ_SYNC to avoid * screwing write batch accounting as write batch * accounting occurs on W->R transition. */ int dispatched = 0; ad->batch_data_dir = REQ_SYNC; ad->changed_batch = 0; ad->new_batch = 0; |
8a8e674cb [PATCH] as-iosche... |
970 971 |
while (ad->next_rq[REQ_SYNC]) { as_move_to_dispatch(ad, ad->next_rq[REQ_SYNC]); |
b4878f245 [PATCH] 02/05: up... |
972 973 974 |
dispatched++; } ad->last_check_fifo[REQ_SYNC] = jiffies; |
8a8e674cb [PATCH] as-iosche... |
975 976 |
while (ad->next_rq[REQ_ASYNC]) { as_move_to_dispatch(ad, ad->next_rq[REQ_ASYNC]); |
b4878f245 [PATCH] 02/05: up... |
977 978 979 980 981 982 |
dispatched++; } ad->last_check_fifo[REQ_ASYNC] = jiffies; return dispatched; } |
1da177e4c Linux-2.6.12-rc2 |
983 984 985 986 987 988 989 990 991 992 993 |
/* Signal that the write batch was uncontended, so we can't time it */ if (ad->batch_data_dir == REQ_ASYNC && !reads) { if (ad->current_write_count == 0 || !writes) ad->write_batch_idled = 1; } if (!(reads || writes) || ad->antic_status == ANTIC_WAIT_REQ || ad->antic_status == ANTIC_WAIT_NEXT || ad->changed_batch) return 0; |
f5b3db001 [PATCH] as: coope... |
994 |
if (!(reads && writes && as_batch_expired(ad))) { |
1da177e4c Linux-2.6.12-rc2 |
995 996 997 |
/* * batch is still running or no reads or no writes */ |
8a8e674cb [PATCH] as-iosche... |
998 |
rq = ad->next_rq[ad->batch_data_dir]; |
1da177e4c Linux-2.6.12-rc2 |
999 1000 1001 1002 |
if (ad->batch_data_dir == REQ_SYNC && ad->antic_expire) { if (as_fifo_expired(ad, REQ_SYNC)) goto fifo_expired; |
8a8e674cb [PATCH] as-iosche... |
1003 |
if (as_can_anticipate(ad, rq)) { |
1da177e4c Linux-2.6.12-rc2 |
1004 1005 1006 1007 |
as_antic_waitreq(ad); return 0; } } |
8a8e674cb [PATCH] as-iosche... |
1008 |
if (rq) { |
1da177e4c Linux-2.6.12-rc2 |
1009 1010 1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 |
/* we have a "next request" */ if (reads && !writes) ad->current_batch_expires = jiffies + ad->batch_expire[REQ_SYNC]; goto dispatch_request; } } /* * at this point we are not running a batch. select the appropriate * data direction (read / write) */ if (reads) { |
dd67d0515 [PATCH] rbtree: s... |
1023 |
BUG_ON(RB_EMPTY_ROOT(&ad->sort_list[REQ_SYNC])); |
1da177e4c Linux-2.6.12-rc2 |
1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 |
if (writes && ad->batch_data_dir == REQ_SYNC) /* * Last batch was a read, switch to writes */ goto dispatch_writes; if (ad->batch_data_dir == REQ_ASYNC) { WARN_ON(ad->new_batch); ad->changed_batch = 1; } ad->batch_data_dir = REQ_SYNC; |
8a8e674cb [PATCH] as-iosche... |
1036 |
rq = rq_entry_fifo(ad->fifo_list[REQ_SYNC].next); |
1da177e4c Linux-2.6.12-rc2 |
1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 |
ad->last_check_fifo[ad->batch_data_dir] = jiffies; goto dispatch_request; } /* * the last batch was a read */ if (writes) { dispatch_writes: |
dd67d0515 [PATCH] rbtree: s... |
1047 |
BUG_ON(RB_EMPTY_ROOT(&ad->sort_list[REQ_ASYNC])); |
1da177e4c Linux-2.6.12-rc2 |
1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 |
if (ad->batch_data_dir == REQ_SYNC) { ad->changed_batch = 1; /* * new_batch might be 1 when the queue runs out of * reads. A subsequent submission of a write might * cause a change of batch before the read is finished. */ ad->new_batch = 0; } ad->batch_data_dir = REQ_ASYNC; ad->current_write_count = ad->write_batch_count; ad->write_batch_idled = 0; |
8a8e674cb [PATCH] as-iosche... |
1062 |
rq = ad->next_rq[ad->batch_data_dir]; |
1da177e4c Linux-2.6.12-rc2 |
1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075 |
goto dispatch_request; } BUG(); return 0; dispatch_request: /* * If a request has expired, service it. */ if (as_fifo_expired(ad, ad->batch_data_dir)) { fifo_expired: |
8a8e674cb [PATCH] as-iosche... |
1076 |
rq = rq_entry_fifo(ad->fifo_list[ad->batch_data_dir].next); |
1da177e4c Linux-2.6.12-rc2 |
1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 |
} if (ad->changed_batch) { WARN_ON(ad->new_batch); if (ad->nr_dispatched) return 0; if (ad->batch_data_dir == REQ_ASYNC) ad->current_batch_expires = jiffies + ad->batch_expire[REQ_ASYNC]; else ad->new_batch = 1; ad->changed_batch = 0; } /* |
8a8e674cb [PATCH] as-iosche... |
1095 |
* rq is the selected appropriate request. |
1da177e4c Linux-2.6.12-rc2 |
1096 |
*/ |
8a8e674cb [PATCH] as-iosche... |
1097 |
as_move_to_dispatch(ad, rq); |
1da177e4c Linux-2.6.12-rc2 |
1098 1099 1100 |
return 1; } |
1da177e4c Linux-2.6.12-rc2 |
1101 |
/* |
8a8e674cb [PATCH] as-iosche... |
1102 |
* add rq to rbtree and fifo |
1da177e4c Linux-2.6.12-rc2 |
1103 |
*/ |
b4878f245 [PATCH] 02/05: up... |
1104 |
static void as_add_request(request_queue_t *q, struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
1105 |
{ |
b4878f245 [PATCH] 02/05: up... |
1106 |
struct as_data *ad = q->elevator->elevator_data; |
1da177e4c Linux-2.6.12-rc2 |
1107 |
int data_dir; |
8a8e674cb [PATCH] as-iosche... |
1108 |
RQ_SET_STATE(rq, AS_RQ_NEW); |
b4878f245 [PATCH] 02/05: up... |
1109 |
|
9e2585a8a [PATCH] as-iosche... |
1110 |
data_dir = rq_is_sync(rq); |
1da177e4c Linux-2.6.12-rc2 |
1111 |
|
b5deef901 [PATCH] Make sure... |
1112 |
rq->elevator_private = as_get_io_context(q->node); |
1da177e4c Linux-2.6.12-rc2 |
1113 |
|
8a8e674cb [PATCH] as-iosche... |
1114 1115 1116 |
if (RQ_IOC(rq)) { as_update_iohist(ad, RQ_IOC(rq)->aic, rq); atomic_inc(&RQ_IOC(rq)->aic->nr_queued); |
1da177e4c Linux-2.6.12-rc2 |
1117 |
} |
8a8e674cb [PATCH] as-iosche... |
1118 |
as_add_rq_rb(ad, rq); |
1da177e4c Linux-2.6.12-rc2 |
1119 |
|
ef9be1d33 [BLOCK] as-iosche... |
1120 1121 1122 |
/* * set expire time (only used for reads) and add to fifo list */ |
d4f2f4629 [PATCH] as-iosche... |
1123 1124 |
rq_set_fifo_time(rq, jiffies + ad->fifo_expire[data_dir]); list_add_tail(&rq->queuelist, &ad->fifo_list[data_dir]); |
1da177e4c Linux-2.6.12-rc2 |
1125 |
|
8a8e674cb [PATCH] as-iosche... |
1126 1127 |
as_update_rq(ad, rq); /* keep state machine up to date */ RQ_SET_STATE(rq, AS_RQ_QUEUED); |
1da177e4c Linux-2.6.12-rc2 |
1128 |
} |
b4878f245 [PATCH] 02/05: up... |
1129 |
static void as_activate_request(request_queue_t *q, struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
1130 |
{ |
8a8e674cb [PATCH] as-iosche... |
1131 1132 1133 1134 |
WARN_ON(RQ_STATE(rq) != AS_RQ_DISPATCHED); RQ_SET_STATE(rq, AS_RQ_REMOVED); if (RQ_IOC(rq) && RQ_IOC(rq)->aic) atomic_dec(&RQ_IOC(rq)->aic->nr_dispatched); |
1da177e4c Linux-2.6.12-rc2 |
1135 |
} |
b4878f245 [PATCH] 02/05: up... |
1136 |
static void as_deactivate_request(request_queue_t *q, struct request *rq) |
1da177e4c Linux-2.6.12-rc2 |
1137 |
{ |
8a8e674cb [PATCH] as-iosche... |
1138 1139 1140 1141 |
WARN_ON(RQ_STATE(rq) != AS_RQ_REMOVED); RQ_SET_STATE(rq, AS_RQ_DISPATCHED); if (RQ_IOC(rq) && RQ_IOC(rq)->aic) atomic_inc(&RQ_IOC(rq)->aic->nr_dispatched); |
1da177e4c Linux-2.6.12-rc2 |
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 |
} /* * as_queue_empty tells us if there are requests left in the device. It may * not be the case that a driver can get the next request even if the queue * is not empty - it is used in the block layer to check for plugging and * merging opportunities */ static int as_queue_empty(request_queue_t *q) { struct as_data *ad = q->elevator->elevator_data; |
b4878f245 [PATCH] 02/05: up... |
1153 1154 |
return list_empty(&ad->fifo_list[REQ_ASYNC]) && list_empty(&ad->fifo_list[REQ_SYNC]); |
1da177e4c Linux-2.6.12-rc2 |
1155 |
} |
1da177e4c Linux-2.6.12-rc2 |
1156 1157 1158 1159 1160 1161 |
static int as_merge(request_queue_t *q, struct request **req, struct bio *bio) { struct as_data *ad = q->elevator->elevator_data; sector_t rb_key = bio->bi_sector + bio_sectors(bio); struct request *__rq; |
1da177e4c Linux-2.6.12-rc2 |
1162 1163 1164 1165 |
/* * check for front merge */ |
e37f346e3 [PATCH] as-iosche... |
1166 |
__rq = elv_rb_find(&ad->sort_list[bio_data_dir(bio)], rb_key); |
9817064b6 [PATCH] elevator:... |
1167 1168 1169 |
if (__rq && elv_rq_merge_ok(__rq, bio)) { *req = __rq; return ELEVATOR_FRONT_MERGE; |
1da177e4c Linux-2.6.12-rc2 |
1170 1171 1172 |
} return ELEVATOR_NO_MERGE; |
1da177e4c Linux-2.6.12-rc2 |
1173 |
} |
e37f346e3 [PATCH] as-iosche... |
1174 |
static void as_merged_request(request_queue_t *q, struct request *req, int type) |
1da177e4c Linux-2.6.12-rc2 |
1175 1176 |
{ struct as_data *ad = q->elevator->elevator_data; |
1da177e4c Linux-2.6.12-rc2 |
1177 1178 |
/* |
1da177e4c Linux-2.6.12-rc2 |
1179 1180 |
* if the merge was a front merge, we need to reposition request */ |
e37f346e3 [PATCH] as-iosche... |
1181 |
if (type == ELEVATOR_FRONT_MERGE) { |
8a8e674cb [PATCH] as-iosche... |
1182 1183 |
as_del_rq_rb(ad, req); as_add_rq_rb(ad, req); |
1da177e4c Linux-2.6.12-rc2 |
1184 1185 1186 1187 1188 1189 |
/* * Note! At this stage of this and the next function, our next * request may not be optimal - eg the request may have "grown" * behind the disk head. We currently don't bother adjusting. */ } |
1da177e4c Linux-2.6.12-rc2 |
1190 |
} |
f5b3db001 [PATCH] as: coope... |
1191 1192 |
static void as_merged_requests(request_queue_t *q, struct request *req, struct request *next) |
1da177e4c Linux-2.6.12-rc2 |
1193 |
{ |
1da177e4c Linux-2.6.12-rc2 |
1194 |
/* |
8a8e674cb [PATCH] as-iosche... |
1195 1196 |
* if next expires before rq, assign its expire time to arq * and move into next position (next will be deleted) in fifo |
1da177e4c Linux-2.6.12-rc2 |
1197 |
*/ |
d4f2f4629 [PATCH] as-iosche... |
1198 1199 |
if (!list_empty(&req->queuelist) && !list_empty(&next->queuelist)) { if (time_before(rq_fifo_time(next), rq_fifo_time(req))) { |
8a8e674cb [PATCH] as-iosche... |
1200 1201 |
struct io_context *rioc = RQ_IOC(req); struct io_context *nioc = RQ_IOC(next); |
d4f2f4629 [PATCH] as-iosche... |
1202 1203 |
list_move(&req->queuelist, &next->queuelist); rq_set_fifo_time(req, rq_fifo_time(next)); |
1da177e4c Linux-2.6.12-rc2 |
1204 1205 1206 1207 |
/* * Don't copy here but swap, because when anext is * removed below, it must contain the unused context */ |
8a8e674cb [PATCH] as-iosche... |
1208 |
swap_io_context(&rioc, &nioc); |
1da177e4c Linux-2.6.12-rc2 |
1209 1210 1211 1212 |
} } /* |
1da177e4c Linux-2.6.12-rc2 |
1213 1214 1215 |
* kill knowledge of next, this one is a goner */ as_remove_queued_request(q, next); |
8a8e674cb [PATCH] as-iosche... |
1216 |
as_put_io_context(next); |
1da177e4c Linux-2.6.12-rc2 |
1217 |
|
8a8e674cb [PATCH] as-iosche... |
1218 |
RQ_SET_STATE(next, AS_RQ_MERGED); |
1da177e4c Linux-2.6.12-rc2 |
1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 |
} /* * This is executed in a "deferred" process context, by kblockd. It calls the * driver's request_fn so the driver can submit that request. * * IMPORTANT! This guy will reenter the elevator, so set up all queue global * state before calling, and don't rely on any state over calls. * * FIXME! dispatch queue is not a queue at all! */ |
65f27f384 WorkStruct: Pass ... |
1230 |
static void as_work_handler(struct work_struct *work) |
1da177e4c Linux-2.6.12-rc2 |
1231 |
{ |
65f27f384 WorkStruct: Pass ... |
1232 1233 |
struct as_data *ad = container_of(work, struct as_data, antic_work); struct request_queue *q = ad->q; |
1da177e4c Linux-2.6.12-rc2 |
1234 1235 1236 |
unsigned long flags; spin_lock_irqsave(q->queue_lock, flags); |
dc72ef4ae [PATCH] Add blk_s... |
1237 |
blk_start_queueing(q); |
1da177e4c Linux-2.6.12-rc2 |
1238 1239 |
spin_unlock_irqrestore(q->queue_lock, flags); } |
cb78b285c [PATCH] Drop usel... |
1240 |
static int as_may_queue(request_queue_t *q, int rw) |
1da177e4c Linux-2.6.12-rc2 |
1241 1242 1243 1244 1245 1246 |
{ int ret = ELV_MQUEUE_MAY; struct as_data *ad = q->elevator->elevator_data; struct io_context *ioc; if (ad->antic_status == ANTIC_WAIT_REQ || ad->antic_status == ANTIC_WAIT_NEXT) { |
b5deef901 [PATCH] Make sure... |
1247 |
ioc = as_get_io_context(q->node); |
1da177e4c Linux-2.6.12-rc2 |
1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 |
if (ad->io_context == ioc) ret = ELV_MQUEUE_MUST; put_io_context(ioc); } return ret; } static void as_exit_queue(elevator_t *e) { struct as_data *ad = e->elevator_data; del_timer_sync(&ad->antic_timer); |
19a75d83f kblockd: use flus... |
1261 |
kblockd_flush_work(&ad->antic_work); |
1da177e4c Linux-2.6.12-rc2 |
1262 1263 1264 |
BUG_ON(!list_empty(&ad->fifo_list[REQ_SYNC])); BUG_ON(!list_empty(&ad->fifo_list[REQ_ASYNC])); |
1da177e4c Linux-2.6.12-rc2 |
1265 |
put_io_context(ad->io_context); |
1da177e4c Linux-2.6.12-rc2 |
1266 1267 1268 1269 |
kfree(ad); } /* |
8a8e674cb [PATCH] as-iosche... |
1270 |
* initialize elevator private data (as_data). |
1da177e4c Linux-2.6.12-rc2 |
1271 |
*/ |
bb37b94c6 [BLOCK] Cleanup u... |
1272 |
static void *as_init_queue(request_queue_t *q) |
1da177e4c Linux-2.6.12-rc2 |
1273 1274 |
{ struct as_data *ad; |
1da177e4c Linux-2.6.12-rc2 |
1275 |
|
94f6030ca Slab allocators: ... |
1276 |
ad = kmalloc_node(sizeof(*ad), GFP_KERNEL | __GFP_ZERO, q->node); |
1da177e4c Linux-2.6.12-rc2 |
1277 |
if (!ad) |
bc1c11697 [PATCH] elevator ... |
1278 |
return NULL; |
1da177e4c Linux-2.6.12-rc2 |
1279 1280 |
ad->q = q; /* Identify what queue the data belongs to */ |
1da177e4c Linux-2.6.12-rc2 |
1281 1282 1283 1284 |
/* anticipatory scheduling helpers */ ad->antic_timer.function = as_antic_timeout; ad->antic_timer.data = (unsigned long)q; init_timer(&ad->antic_timer); |
65f27f384 WorkStruct: Pass ... |
1285 |
INIT_WORK(&ad->antic_work, as_work_handler); |
1da177e4c Linux-2.6.12-rc2 |
1286 |
|
1da177e4c Linux-2.6.12-rc2 |
1287 1288 1289 1290 |
INIT_LIST_HEAD(&ad->fifo_list[REQ_SYNC]); INIT_LIST_HEAD(&ad->fifo_list[REQ_ASYNC]); ad->sort_list[REQ_SYNC] = RB_ROOT; ad->sort_list[REQ_ASYNC] = RB_ROOT; |
1da177e4c Linux-2.6.12-rc2 |
1291 1292 1293 1294 1295 |
ad->fifo_expire[REQ_SYNC] = default_read_expire; ad->fifo_expire[REQ_ASYNC] = default_write_expire; ad->antic_expire = default_antic_expire; ad->batch_expire[REQ_SYNC] = default_read_batch_expire; ad->batch_expire[REQ_ASYNC] = default_write_batch_expire; |
1da177e4c Linux-2.6.12-rc2 |
1296 1297 1298 1299 1300 |
ad->current_batch_expires = jiffies + ad->batch_expire[REQ_SYNC]; ad->write_batch_count = ad->batch_expire[REQ_ASYNC] / 10; if (ad->write_batch_count < 2) ad->write_batch_count = 2; |
bc1c11697 [PATCH] elevator ... |
1301 |
return ad; |
1da177e4c Linux-2.6.12-rc2 |
1302 1303 1304 1305 1306 |
} /* * sysfs parts below */ |
1da177e4c Linux-2.6.12-rc2 |
1307 1308 1309 1310 |
static ssize_t as_var_show(unsigned int var, char *page) { |
1da177e4c Linux-2.6.12-rc2 |
1311 1312 1313 1314 1315 1316 1317 |
return sprintf(page, "%d ", var); } static ssize_t as_var_store(unsigned long *var, const char *page, size_t count) { |
1da177e4c Linux-2.6.12-rc2 |
1318 |
char *p = (char *) page; |
c9b3ad673 [PATCH] as-iosche... |
1319 |
*var = simple_strtoul(p, &p, 10); |
1da177e4c Linux-2.6.12-rc2 |
1320 1321 |
return count; } |
e572ec7e4 [PATCH] fix rmmod... |
1322 |
static ssize_t est_time_show(elevator_t *e, char *page) |
1da177e4c Linux-2.6.12-rc2 |
1323 |
{ |
3d1ab40f4 [PATCH] elevator_... |
1324 |
struct as_data *ad = e->elevator_data; |
1da177e4c Linux-2.6.12-rc2 |
1325 |
int pos = 0; |
f5b3db001 [PATCH] as: coope... |
1326 1327 1328 1329 1330 1331 1332 |
pos += sprintf(page+pos, "%lu %% exit probability ", 100*ad->exit_prob/256); pos += sprintf(page+pos, "%lu %% probability of exiting without a " "cooperating process submitting IO ", 100*ad->exit_no_coop/256); |
1da177e4c Linux-2.6.12-rc2 |
1333 1334 |
pos += sprintf(page+pos, "%lu ms new thinktime ", ad->new_ttime_mean); |
f5b3db001 [PATCH] as: coope... |
1335 1336 1337 |
pos += sprintf(page+pos, "%llu sectors new seek distance ", (unsigned long long)ad->new_seek_mean); |
1da177e4c Linux-2.6.12-rc2 |
1338 1339 1340 1341 1342 |
return pos; } #define SHOW_FUNCTION(__FUNC, __VAR) \ |
3d1ab40f4 [PATCH] elevator_... |
1343 |
static ssize_t __FUNC(elevator_t *e, char *page) \ |
1da177e4c Linux-2.6.12-rc2 |
1344 |
{ \ |
3d1ab40f4 [PATCH] elevator_... |
1345 |
struct as_data *ad = e->elevator_data; \ |
1da177e4c Linux-2.6.12-rc2 |
1346 1347 |
return as_var_show(jiffies_to_msecs((__VAR)), (page)); \ } |
e572ec7e4 [PATCH] fix rmmod... |
1348 1349 1350 1351 1352 |
SHOW_FUNCTION(as_read_expire_show, ad->fifo_expire[REQ_SYNC]); SHOW_FUNCTION(as_write_expire_show, ad->fifo_expire[REQ_ASYNC]); SHOW_FUNCTION(as_antic_expire_show, ad->antic_expire); SHOW_FUNCTION(as_read_batch_expire_show, ad->batch_expire[REQ_SYNC]); SHOW_FUNCTION(as_write_batch_expire_show, ad->batch_expire[REQ_ASYNC]); |
1da177e4c Linux-2.6.12-rc2 |
1353 1354 1355 |
#undef SHOW_FUNCTION #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX) \ |
3d1ab40f4 [PATCH] elevator_... |
1356 |
static ssize_t __FUNC(elevator_t *e, const char *page, size_t count) \ |
1da177e4c Linux-2.6.12-rc2 |
1357 |
{ \ |
3d1ab40f4 [PATCH] elevator_... |
1358 1359 |
struct as_data *ad = e->elevator_data; \ int ret = as_var_store(__PTR, (page), count); \ |
1da177e4c Linux-2.6.12-rc2 |
1360 1361 1362 1363 1364 1365 1366 |
if (*(__PTR) < (MIN)) \ *(__PTR) = (MIN); \ else if (*(__PTR) > (MAX)) \ *(__PTR) = (MAX); \ *(__PTR) = msecs_to_jiffies(*(__PTR)); \ return ret; \ } |
e572ec7e4 [PATCH] fix rmmod... |
1367 1368 1369 1370 |
STORE_FUNCTION(as_read_expire_store, &ad->fifo_expire[REQ_SYNC], 0, INT_MAX); STORE_FUNCTION(as_write_expire_store, &ad->fifo_expire[REQ_ASYNC], 0, INT_MAX); STORE_FUNCTION(as_antic_expire_store, &ad->antic_expire, 0, INT_MAX); STORE_FUNCTION(as_read_batch_expire_store, |
1da177e4c Linux-2.6.12-rc2 |
1371 |
&ad->batch_expire[REQ_SYNC], 0, INT_MAX); |
e572ec7e4 [PATCH] fix rmmod... |
1372 |
STORE_FUNCTION(as_write_batch_expire_store, |
1da177e4c Linux-2.6.12-rc2 |
1373 1374 |
&ad->batch_expire[REQ_ASYNC], 0, INT_MAX); #undef STORE_FUNCTION |
e572ec7e4 [PATCH] fix rmmod... |
1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 |
#define AS_ATTR(name) \ __ATTR(name, S_IRUGO|S_IWUSR, as_##name##_show, as_##name##_store) static struct elv_fs_entry as_attrs[] = { __ATTR_RO(est_time), AS_ATTR(read_expire), AS_ATTR(write_expire), AS_ATTR(antic_expire), AS_ATTR(read_batch_expire), AS_ATTR(write_batch_expire), __ATTR_NULL |
1da177e4c Linux-2.6.12-rc2 |
1386 |
}; |
1da177e4c Linux-2.6.12-rc2 |
1387 1388 1389 1390 1391 |
static struct elevator_type iosched_as = { .ops = { .elevator_merge_fn = as_merge, .elevator_merged_fn = as_merged_request, .elevator_merge_req_fn = as_merged_requests, |
b4878f245 [PATCH] 02/05: up... |
1392 1393 1394 |
.elevator_dispatch_fn = as_dispatch_request, .elevator_add_req_fn = as_add_request, .elevator_activate_req_fn = as_activate_request, |
1da177e4c Linux-2.6.12-rc2 |
1395 1396 1397 |
.elevator_deactivate_req_fn = as_deactivate_request, .elevator_queue_empty_fn = as_queue_empty, .elevator_completed_req_fn = as_completed_request, |
e37f346e3 [PATCH] as-iosche... |
1398 1399 |
.elevator_former_req_fn = elv_rb_former_request, .elevator_latter_req_fn = elv_rb_latter_request, |
1da177e4c Linux-2.6.12-rc2 |
1400 1401 1402 |
.elevator_may_queue_fn = as_may_queue, .elevator_init_fn = as_init_queue, .elevator_exit_fn = as_exit_queue, |
e17a9489b [PATCH] stop elv_... |
1403 |
.trim = as_trim, |
1da177e4c Linux-2.6.12-rc2 |
1404 |
}, |
3d1ab40f4 [PATCH] elevator_... |
1405 |
.elevator_attrs = as_attrs, |
1da177e4c Linux-2.6.12-rc2 |
1406 1407 1408 1409 1410 1411 |
.elevator_name = "anticipatory", .elevator_owner = THIS_MODULE, }; static int __init as_init(void) { |
c65fb61b3 [PATCH] Allow as-... |
1412 |
return elv_register(&iosched_as); |
1da177e4c Linux-2.6.12-rc2 |
1413 1414 1415 1416 |
} static void __exit as_exit(void) { |
6e9a4738c [PATCH] completio... |
1417 |
DECLARE_COMPLETION_ONSTACK(all_gone); |
1da177e4c Linux-2.6.12-rc2 |
1418 |
elv_unregister(&iosched_as); |
334e94de9 [PATCH] deal with... |
1419 |
ioc_gone = &all_gone; |
fba822722 [PATCH 1/2] iosch... |
1420 1421 |
/* ioc_gone's update must be visible before reading ioc_count */ smp_wmb(); |
e4313dd42 [PATCH] as-iosche... |
1422 |
if (elv_ioc_count_read(ioc_count)) |
fba822722 [PATCH 1/2] iosch... |
1423 |
wait_for_completion(ioc_gone); |
334e94de9 [PATCH] deal with... |
1424 |
synchronize_rcu(); |
1da177e4c Linux-2.6.12-rc2 |
1425 1426 1427 1428 1429 1430 1431 1432 |
} module_init(as_init); module_exit(as_exit); MODULE_AUTHOR("Nick Piggin"); MODULE_LICENSE("GPL"); MODULE_DESCRIPTION("anticipatory IO scheduler"); |