block/deadline-iosched.c

   1 /*
   2  *  Deadline i/o scheduler.
   3  *
   4  *  Copyright (C) 2002 Jens Axboe <axboe@kernel.dk>
   5  */
   6 #include <linux/kernel.h>
   7 #include <linux/fs.h>
   8 #include <linux/blkdev.h>
   9 #include <linux/elevator.h>
  10 #include <linux/bio.h>
  11 #include <linux/module.h>
  12 #include <linux/slab.h>
  13 #include <linux/init.h>
  14 #include <linux/compiler.h>
  15 #include <linux/rbtree.h>
  16
  17 /*
  18  * See Documentation/block/deadline-iosched.txt
  19  */
  20 static const int read_expire = HZ / 2;  /* max time before a read is submitted. */
  21 static const int write_expire = 5 * HZ; /* ditto for writes, these limits are SOFT! */
  22 static const int writes_starved = 2;    /* max times reads can starve a write */
  23 static const int fifo_batch = 16;       /* # of sequential requests treated as one
  24                                      by the above parameters. For throughput. */
  25
  26 struct deadline_data {
  27         /*
  28          * run time data
  29          */
  30
  31         /*
  32          * requests (deadline_rq s) are present on both sort_list and fifo_list
  33          */
  34         struct rb_root sort_list[2];
  35         struct list_head fifo_list[2];
  36
  37         /*
  38          * next in sort order. read, write or both are NULL
  39          */
  40         struct request *next_rq[2];
  41         unsigned int batching;          /* number of sequential requests made */
  42         sector_t last_sector;           /* head position */
  43         unsigned int starved;           /* times reads have starved writes */
  44
  45         /*
  46          * settings that change how the i/o scheduler behaves
  47          */
  48         int fifo_expire[2];
  49         int fifo_batch;
  50         int writes_starved;
  51         int front_merges;
  52 };
  53
  54 static void deadline_move_request(struct deadline_data *, struct request *);
  55
  56 static inline struct rb_root *
  57 deadline_rb_root(struct deadline_data *dd, struct request *rq)
  58 {
  59         return &dd->sort_list[rq_data_dir(rq)];
  60 }
  61
  62 /*
  63  * get the request after `rq' in sector-sorted order
  64  */
  65 static inline struct request *
  66 deadline_latter_request(struct request *rq)
  67 {
  68         struct rb_node *node = rb_next(&rq->rb_node);
  69
  70         if (node)
  71                 return rb_entry_rq(node);
  72
  73         return NULL;
  74 }
  75
  76 static void
  77 deadline_add_rq_rb(struct deadline_data *dd, struct request *rq)
  78 {
  79         struct rb_root *root = deadline_rb_root(dd, rq);
  80
  81         elv_rb_add(root, rq);
  82 }
  83
  84 static inline void
  85 deadline_del_rq_rb(struct deadline_data *dd, struct request *rq)
  86 {
  87         const int data_dir = rq_data_dir(rq);
  88
  89         if (dd->next_rq[data_dir] == rq)
  90                 dd->next_rq[data_dir] = deadline_latter_request(rq);
  91
  92         elv_rb_del(deadline_rb_root(dd, rq), rq);
  93 }
  94
  95 /*
  96  * add rq to rbtree and fifo
  97  */
  98 static void
  99 deadline_add_request(struct request_queue *q, struct request *rq)
 100 {
 101         struct deadline_data *dd = q->elevator->elevator_data;
 102         const int data_dir = rq_data_dir(rq);
 103
 104         deadline_add_rq_rb(dd, rq);
 105
 106         /*
 107          * set expire time and add to fifo list
 108          */
 109         rq_set_fifo_time(rq, jiffies + dd->fifo_expire[data_dir]);
 110         list_add_tail(&rq->queuelist, &dd->fifo_list[data_dir]);
 111 }
 112
 113 /*
 114  * remove rq from rbtree and fifo.
 115  */
 116 static void deadline_remove_request(struct request_queue *q, struct request *rq)
 117 {
 118         struct deadline_data *dd = q->elevator->elevator_data;
 119
 120         rq_fifo_clear(rq);
 121         deadline_del_rq_rb(dd, rq);
 122 }
 123
 124 static int
 125 deadline_merge(struct request_queue *q, struct request **req, struct bio *bio)
 126 {
 127         struct deadline_data *dd = q->elevator->elevator_data;
 128         struct request *__rq;
 129         int ret;
 130
 131         /*
 132          * check for front merge
 133          */
 134         if (dd->front_merges) {
 135                 sector_t sector = bio->bi_sector + bio_sectors(bio);
 136
 137                 __rq = elv_rb_find(&dd->sort_list[bio_data_dir(bio)], sector);
 138                 if (__rq) {
 139                         BUG_ON(sector != blk_rq_pos(__rq));
 140
 141                         if (elv_rq_merge_ok(__rq, bio)) {
 142                                 ret = ELEVATOR_FRONT_MERGE;
 143                                 goto out;
 144                         }
 145                 }
 146         }
 147
 148         return ELEVATOR_NO_MERGE;
 149 out:
 150         *req = __rq;
 151         return ret;
 152 }
 153
 154 static void deadline_merged_request(struct request_queue *q,
 155                                     struct request *req, int type)
 156 {
 157         struct deadline_data *dd = q->elevator->elevator_data;
 158
 159         /*
 160          * if the merge was a front merge, we need to reposition request
 161          */
 162         if (type == ELEVATOR_FRONT_MERGE) {
 163                 elv_rb_del(deadline_rb_root(dd, req), req);
 164                 deadline_add_rq_rb(dd, req);
 165         }
 166 }
 167
 168 static void
 169 deadline_merged_requests(struct request_queue *q, struct request *req,
 170                          struct request *next)
 171 {
 172         /*
 173          * if next expires before rq, assign its expire time to rq
 174          * and move into next position (next will be deleted) in fifo
 175          */
 176         if (!list_empty(&req->queuelist) && !list_empty(&next->queuelist)) {
 177                 if (time_before(rq_fifo_time(next), rq_fifo_time(req))) {
 178                         list_move(&req->queuelist, &next->queuelist);
 179                         rq_set_fifo_time(req, rq_fifo_time(next));
 180                 }
 181         }
 182
 183         /*
 184          * kill knowledge of next, this one is a goner
 185          */
 186         deadline_remove_request(q, next);
 187 }
 188
 189 /*
 190  * move request from sort list to dispatch queue.
 191  */
 192 static inline void
 193 deadline_move_to_dispatch(struct deadline_data *dd, struct request *rq)
 194 {
 195         struct request_queue *q = rq->q;
 196
 197         deadline_remove_request(q, rq);
 198         elv_dispatch_add_tail(q, rq);
 199 }
 200
 201 /*
 202  * move an entry to dispatch queue
 203  */
 204 static void
 205 deadline_move_request(struct deadline_data *dd, struct request *rq)
 206 {
 207         const int data_dir = rq_data_dir(rq);
 208
 209         dd->next_rq[READ] = NULL;
 210         dd->next_rq[WRITE] = NULL;
 211         dd->next_rq[data_dir] = deadline_latter_request(rq);
 212
 213         dd->last_sector = rq_end_sector(rq);
 214
 215         /*
 216          * take it off the sort and fifo list, move
 217          * to dispatch queue
 218          */
 219         deadline_move_to_dispatch(dd, rq);
 220 }
 221
 222 /*
 223  * deadline_check_fifo returns 0 if there are no expired requests on the fifo,
 224  * 1 otherwise. Requires !list_empty(&dd->fifo_list[data_dir])
 225  */
 226 static inline int deadline_check_fifo(struct deadline_data *dd, int ddir)
 227 {
 228         struct request *rq = rq_entry_fifo(dd->fifo_list[ddir].next);
 229
 230         /*
 231          * rq is expired!
 232          */
 233         if (time_after(jiffies, rq_fifo_time(rq)))
 234                 return 1;
 235
 236         return 0;
 237 }
 238
 239 /*
 240  * deadline_dispatch_requests selects the best request according to
 241  * read/write expire, fifo_batch, etc
 242  */
 243 static int deadline_dispatch_requests(struct request_queue *q, int force)
 244 {
 245         struct deadline_data *dd = q->elevator->elevator_data;
 246         const int reads = !list_empty(&dd->fifo_list[READ]);
 247         const int writes = !list_empty(&dd->fifo_list[WRITE]);
 248         struct request *rq;
 249         int data_dir;
 250
 251         /*
 252          * batches are currently reads XOR writes
 253          */
 254         if (dd->next_rq[WRITE])
 255                 rq = dd->next_rq[WRITE];
 256         else
 257                 rq = dd->next_rq[READ];
 258
 259         if (rq && dd->batching < dd->fifo_batch)
 260                 /* we have a next request are still entitled to batch */
 261                 goto dispatch_request;
 262
 263         /*
 264          * at this point we are not running a batch. select the appropriate
 265          * data direction (read / write)
 266          */
 267
 268         if (reads) {
 269                 BUG_ON(RB_EMPTY_ROOT(&dd->sort_list[READ]));
 270
 271                 if (writes && (dd->starved++ >= dd->writes_starved))
 272                         goto dispatch_writes;
 273
 274                 data_dir = READ;
 275
 276                 goto dispatch_find_request;
 277         }
 278
 279         /*
 280          * there are either no reads or writes have been starved
 281          */
 282
 283         if (writes) {
 284 dispatch_writes:
 285                 BUG_ON(RB_EMPTY_ROOT(&dd->sort_list[WRITE]));
 286
 287                 dd->starved = 0;
 288
 289                 data_dir = WRITE;
 290
 291                 goto dispatch_find_request;
 292         }
 293
 294         return 0;
 295
 296 dispatch_find_request:
 297         /*
 298          * we are not running a batch, find best request for selected data_dir
 299          */
 300         if (deadline_check_fifo(dd, data_dir) || !dd->next_rq[data_dir]) {
 301                 /*
 302                  * A deadline has expired, the last request was in the other
 303                  * direction, or we have run out of higher-sectored requests.
 304                  * Start again from the request with the earliest expiry time.
 305                  */
 306                 rq = rq_entry_fifo(dd->fifo_list[data_dir].next);
 307         } else {
 308                 /*
 309                  * The last req was the same dir and we have a next request in
 310                  * sort order. No expired requests so continue on from here.
 311                  */
 312                 rq = dd->next_rq[data_dir];
 313         }
 314
 315         dd->batching = 0;
 316
 317 dispatch_request:
 318         /*
 319          * rq is the selected appropriate request.
 320          */
 321         dd->batching++;
 322         deadline_move_request(dd, rq);
 323
 324         return 1;
 325 }
 326
 327 static void deadline_exit_queue(struct elevator_queue *e)
 328 {
 329         struct deadline_data *dd = e->elevator_data;
 330
 331         BUG_ON(!list_empty(&dd->fifo_list[READ]));
 332         BUG_ON(!list_empty(&dd->fifo_list[WRITE]));
 333
 334         kfree(dd);
 335 }
 336
 337 /*
 338  * initialize elevator private data (deadline_data).
 339  */
 340 static int deadline_init_queue(struct request_queue *q)
 341 {
 342         struct deadline_data *dd;
 343
 344         dd = kmalloc_node(sizeof(*dd), GFP_KERNEL | __GFP_ZERO, q->node);
 345         if (!dd)
 346                 return -ENOMEM;
 347
 348         INIT_LIST_HEAD(&dd->fifo_list[READ]);
 349         INIT_LIST_HEAD(&dd->fifo_list[WRITE]);
 350         dd->sort_list[READ] = RB_ROOT;
 351         dd->sort_list[WRITE] = RB_ROOT;
 352         dd->fifo_expire[READ] = read_expire;
 353         dd->fifo_expire[WRITE] = write_expire;
 354         dd->writes_starved = writes_starved;
 355         dd->front_merges = 1;
 356         dd->fifo_batch = fifo_batch;
 357
 358         q->elevator->elevator_data = dd;
 359         return 0;
 360 }
 361
 362 /*
 363  * sysfs parts below
 364  */
 365
 366 static ssize_t
 367 deadline_var_show(int var, char *page)
 368 {
 369         return sprintf(page, "%d\n", var);
 370 }
 371
 372 static ssize_t
 373 deadline_var_store(int *var, const char *page, size_t count)
 374 {
 375         char *p = (char *) page;
 376
 377         *var = simple_strtol(p, &p, 10);
 378         return count;
 379 }
 380
 381 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV)                            \
 382 static ssize_t __FUNC(struct elevator_queue *e, char *page)             \
 383 {                                                                       \
 384         struct deadline_data *dd = e->elevator_data;                    \
 385         int __data = __VAR;                                             \
 386         if (__CONV)                                                     \
 387                 __data = jiffies_to_msecs(__data);                      \
 388         return deadline_var_show(__data, (page));                       \
 389 }
 390 SHOW_FUNCTION(deadline_read_expire_show, dd->fifo_expire[READ], 1);
 391 SHOW_FUNCTION(deadline_write_expire_show, dd->fifo_expire[WRITE], 1);
 392 SHOW_FUNCTION(deadline_writes_starved_show, dd->writes_starved, 0);
 393 SHOW_FUNCTION(deadline_front_merges_show, dd->front_merges, 0);
 394 SHOW_FUNCTION(deadline_fifo_batch_show, dd->fifo_batch, 0);
 395 #undef SHOW_FUNCTION
 396
 397 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)                 \
 398 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
 399 {                                                                       \
 400         struct deadline_data *dd = e->elevator_data;                    \
 401         int __data;                                                     \
 402         int ret = deadline_var_store(&__data, (page), count);           \
 403         if (__data < (MIN))                                             \
 404                 __data = (MIN);                                         \
 405         else if (__data > (MAX))                                        \
 406                 __data = (MAX);                                         \
 407         if (__CONV)                                                     \
 408                 *(__PTR) = msecs_to_jiffies(__data);                    \
 409         else                                                            \
 410                 *(__PTR) = __data;                                      \
 411         return ret;                                                     \
 412 }
 413 STORE_FUNCTION(deadline_read_expire_store, &dd->fifo_expire[READ], 0, INT_MAX, 1);
 414 STORE_FUNCTION(deadline_write_expire_store, &dd->fifo_expire[WRITE], 0, INT_MAX, 1);
 415 STORE_FUNCTION(deadline_writes_starved_store, &dd->writes_starved, INT_MIN, INT_MAX, 0);
 416 STORE_FUNCTION(deadline_front_merges_store, &dd->front_merges, 0, 1, 0);
 417 STORE_FUNCTION(deadline_fifo_batch_store, &dd->fifo_batch, 0, INT_MAX, 0);
 418 #undef STORE_FUNCTION
 419
 420 #define DD_ATTR(name) \
 421         __ATTR(name, S_IRUGO|S_IWUSR, deadline_##name##_show, \
 422                                       deadline_##name##_store)
 423
 424 static struct elv_fs_entry deadline_attrs[] = {
 425         DD_ATTR(read_expire),
 426         DD_ATTR(write_expire),
 427         DD_ATTR(writes_starved),
 428         DD_ATTR(front_merges),
 429         DD_ATTR(fifo_batch),
 430         __ATTR_NULL
 431 };
 432
 433 static struct elevator_type iosched_deadline = {
 434         .ops = {
 435                 .elevator_merge_fn =            deadline_merge,
 436                 .elevator_merged_fn =           deadline_merged_request,
 437                 .elevator_merge_req_fn =        deadline_merged_requests,
 438                 .elevator_dispatch_fn =         deadline_dispatch_requests,
 439                 .elevator_add_req_fn =          deadline_add_request,
 440                 .elevator_former_req_fn =       elv_rb_former_request,
 441                 .elevator_latter_req_fn =       elv_rb_latter_request,
 442                 .elevator_init_fn =             deadline_init_queue,
 443                 .elevator_exit_fn =             deadline_exit_queue,
 444         },
 445
 446         .elevator_attrs = deadline_attrs,
 447         .elevator_name = "deadline",
 448         .elevator_owner = THIS_MODULE,
 449 };
 450
 451 static int __init deadline_init(void)
 452 {
 453         return elv_register(&iosched_deadline);
 454 }
 455
 456 static void __exit deadline_exit(void)
 457 {
 458         elv_unregister(&iosched_deadline);
 459 }
 460
 461 module_init(deadline_init);
 462 module_exit(deadline_exit);
 463
 464 MODULE_AUTHOR("Jens Axboe");
 465 MODULE_LICENSE("GPL");
 466 MODULE_DESCRIPTION("deadline IO scheduler");