cfq-iosched: get rid of the need for __GFP_NOFAIL in cfq_find_alloc_queue()
[linux-2.6.git] / block / cfq-iosched.c
1 /*
2  *  CFQ, or complete fairness queueing, disk scheduler.
3  *
4  *  Based on ideas from a previously unfinished io
5  *  scheduler (round robin per-process disk scheduling) and Andrea Arcangeli.
6  *
7  *  Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
8  */
9 #include <linux/module.h>
10 #include <linux/blkdev.h>
11 #include <linux/elevator.h>
12 #include <linux/rbtree.h>
13 #include <linux/ioprio.h>
14 #include <linux/blktrace_api.h>
15
16 /*
17  * tunables
18  */
19 /* max queue in one round of service */
20 static const int cfq_quantum = 4;
21 static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
22 /* maximum backwards seek, in KiB */
23 static const int cfq_back_max = 16 * 1024;
24 /* penalty of a backwards seek */
25 static const int cfq_back_penalty = 2;
26 static const int cfq_slice_sync = HZ / 10;
27 static int cfq_slice_async = HZ / 25;
28 static const int cfq_slice_async_rq = 2;
29 static int cfq_slice_idle = HZ / 125;
30
31 /*
32  * offset from end of service tree
33  */
34 #define CFQ_IDLE_DELAY          (HZ / 5)
35
36 /*
37  * below this threshold, we consider thinktime immediate
38  */
39 #define CFQ_MIN_TT              (2)
40
41 #define CFQ_SLICE_SCALE         (5)
42 #define CFQ_HW_QUEUE_MIN        (5)
43
44 #define RQ_CIC(rq)              \
45         ((struct cfq_io_context *) (rq)->elevator_private)
46 #define RQ_CFQQ(rq)             (struct cfq_queue *) ((rq)->elevator_private2)
47
48 static struct kmem_cache *cfq_pool;
49 static struct kmem_cache *cfq_ioc_pool;
50
51 static DEFINE_PER_CPU(unsigned long, ioc_count);
52 static struct completion *ioc_gone;
53 static DEFINE_SPINLOCK(ioc_gone_lock);
54
55 #define CFQ_PRIO_LISTS          IOPRIO_BE_NR
56 #define cfq_class_idle(cfqq)    ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
57 #define cfq_class_rt(cfqq)      ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
58
59 #define sample_valid(samples)   ((samples) > 80)
60
61 /*
62  * Most of our rbtree usage is for sorting with min extraction, so
63  * if we cache the leftmost node we don't have to walk down the tree
64  * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should
65  * move this into the elevator for the rq sorting as well.
66  */
67 struct cfq_rb_root {
68         struct rb_root rb;
69         struct rb_node *left;
70 };
71 #define CFQ_RB_ROOT     (struct cfq_rb_root) { RB_ROOT, NULL, }
72
73 /*
74  * Per process-grouping structure
75  */
76 struct cfq_queue {
77         /* reference count */
78         atomic_t ref;
79         /* various state flags, see below */
80         unsigned int flags;
81         /* parent cfq_data */
82         struct cfq_data *cfqd;
83         /* service_tree member */
84         struct rb_node rb_node;
85         /* service_tree key */
86         unsigned long rb_key;
87         /* prio tree member */
88         struct rb_node p_node;
89         /* prio tree root we belong to, if any */
90         struct rb_root *p_root;
91         /* sorted list of pending requests */
92         struct rb_root sort_list;
93         /* if fifo isn't expired, next request to serve */
94         struct request *next_rq;
95         /* requests queued in sort_list */
96         int queued[2];
97         /* currently allocated requests */
98         int allocated[2];
99         /* fifo list of requests in sort_list */
100         struct list_head fifo;
101
102         unsigned long slice_end;
103         long slice_resid;
104         unsigned int slice_dispatch;
105
106         /* pending metadata requests */
107         int meta_pending;
108         /* number of requests that are on the dispatch list or inside driver */
109         int dispatched;
110
111         /* io prio of this group */
112         unsigned short ioprio, org_ioprio;
113         unsigned short ioprio_class, org_ioprio_class;
114
115         pid_t pid;
116 };
117
118 /*
119  * Per block device queue structure
120  */
121 struct cfq_data {
122         struct request_queue *queue;
123
124         /*
125          * rr list of queues with requests and the count of them
126          */
127         struct cfq_rb_root service_tree;
128
129         /*
130          * Each priority tree is sorted by next_request position.  These
131          * trees are used when determining if two or more queues are
132          * interleaving requests (see cfq_close_cooperator).
133          */
134         struct rb_root prio_trees[CFQ_PRIO_LISTS];
135
136         unsigned int busy_queues;
137         /*
138          * Used to track any pending rt requests so we can pre-empt current
139          * non-RT cfqq in service when this value is non-zero.
140          */
141         unsigned int busy_rt_queues;
142
143         int rq_in_driver;
144         int sync_flight;
145
146         /*
147          * queue-depth detection
148          */
149         int rq_queued;
150         int hw_tag;
151         int hw_tag_samples;
152         int rq_in_driver_peak;
153
154         /*
155          * idle window management
156          */
157         struct timer_list idle_slice_timer;
158         struct work_struct unplug_work;
159
160         struct cfq_queue *active_queue;
161         struct cfq_io_context *active_cic;
162
163         /*
164          * async queue for each priority case
165          */
166         struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
167         struct cfq_queue *async_idle_cfqq;
168
169         sector_t last_position;
170
171         /*
172          * tunables, see top of file
173          */
174         unsigned int cfq_quantum;
175         unsigned int cfq_fifo_expire[2];
176         unsigned int cfq_back_penalty;
177         unsigned int cfq_back_max;
178         unsigned int cfq_slice[2];
179         unsigned int cfq_slice_async_rq;
180         unsigned int cfq_slice_idle;
181
182         struct list_head cic_list;
183
184         /*
185          * Fallback dummy cfqq for extreme OOM conditions
186          */
187         struct cfq_queue oom_cfqq;
188 };
189
190 enum cfqq_state_flags {
191         CFQ_CFQQ_FLAG_on_rr = 0,        /* on round-robin busy list */
192         CFQ_CFQQ_FLAG_wait_request,     /* waiting for a request */
193         CFQ_CFQQ_FLAG_must_dispatch,    /* must be allowed a dispatch */
194         CFQ_CFQQ_FLAG_must_alloc,       /* must be allowed rq alloc */
195         CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */
196         CFQ_CFQQ_FLAG_fifo_expire,      /* FIFO checked in this slice */
197         CFQ_CFQQ_FLAG_idle_window,      /* slice idling enabled */
198         CFQ_CFQQ_FLAG_prio_changed,     /* task priority has changed */
199         CFQ_CFQQ_FLAG_slice_new,        /* no requests dispatched in slice */
200         CFQ_CFQQ_FLAG_sync,             /* synchronous queue */
201         CFQ_CFQQ_FLAG_coop,             /* has done a coop jump of the queue */
202 };
203
204 #define CFQ_CFQQ_FNS(name)                                              \
205 static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)         \
206 {                                                                       \
207         (cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);                   \
208 }                                                                       \
209 static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)        \
210 {                                                                       \
211         (cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);                  \
212 }                                                                       \
213 static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)         \
214 {                                                                       \
215         return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;      \
216 }
217
218 CFQ_CFQQ_FNS(on_rr);
219 CFQ_CFQQ_FNS(wait_request);
220 CFQ_CFQQ_FNS(must_dispatch);
221 CFQ_CFQQ_FNS(must_alloc);
222 CFQ_CFQQ_FNS(must_alloc_slice);
223 CFQ_CFQQ_FNS(fifo_expire);
224 CFQ_CFQQ_FNS(idle_window);
225 CFQ_CFQQ_FNS(prio_changed);
226 CFQ_CFQQ_FNS(slice_new);
227 CFQ_CFQQ_FNS(sync);
228 CFQ_CFQQ_FNS(coop);
229 #undef CFQ_CFQQ_FNS
230
231 #define cfq_log_cfqq(cfqd, cfqq, fmt, args...)  \
232         blk_add_trace_msg((cfqd)->queue, "cfq%d " fmt, (cfqq)->pid, ##args)
233 #define cfq_log(cfqd, fmt, args...)     \
234         blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args)
235
236 static void cfq_dispatch_insert(struct request_queue *, struct request *);
237 static struct cfq_queue *cfq_get_queue(struct cfq_data *, int,
238                                        struct io_context *, gfp_t);
239 static struct cfq_io_context *cfq_cic_lookup(struct cfq_data *,
240                                                 struct io_context *);
241
242 static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_context *cic,
243                                             int is_sync)
244 {
245         return cic->cfqq[!!is_sync];
246 }
247
248 static inline void cic_set_cfqq(struct cfq_io_context *cic,
249                                 struct cfq_queue *cfqq, int is_sync)
250 {
251         cic->cfqq[!!is_sync] = cfqq;
252 }
253
254 /*
255  * We regard a request as SYNC, if it's either a read or has the SYNC bit
256  * set (in which case it could also be direct WRITE).
257  */
258 static inline int cfq_bio_sync(struct bio *bio)
259 {
260         if (bio_data_dir(bio) == READ || bio_sync(bio))
261                 return 1;
262
263         return 0;
264 }
265
266 /*
267  * scheduler run of queue, if there are requests pending and no one in the
268  * driver that will restart queueing
269  */
270 static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
271 {
272         if (cfqd->busy_queues) {
273                 cfq_log(cfqd, "schedule dispatch");
274                 kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work);
275         }
276 }
277
278 static int cfq_queue_empty(struct request_queue *q)
279 {
280         struct cfq_data *cfqd = q->elevator->elevator_data;
281
282         return !cfqd->busy_queues;
283 }
284
285 /*
286  * Scale schedule slice based on io priority. Use the sync time slice only
287  * if a queue is marked sync and has sync io queued. A sync queue with async
288  * io only, should not get full sync slice length.
289  */
290 static inline int cfq_prio_slice(struct cfq_data *cfqd, int sync,
291                                  unsigned short prio)
292 {
293         const int base_slice = cfqd->cfq_slice[sync];
294
295         WARN_ON(prio >= IOPRIO_BE_NR);
296
297         return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio));
298 }
299
300 static inline int
301 cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
302 {
303         return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio);
304 }
305
306 static inline void
307 cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
308 {
309         cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies;
310         cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies);
311 }
312
313 /*
314  * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
315  * isn't valid until the first request from the dispatch is activated
316  * and the slice time set.
317  */
318 static inline int cfq_slice_used(struct cfq_queue *cfqq)
319 {
320         if (cfq_cfqq_slice_new(cfqq))
321                 return 0;
322         if (time_before(jiffies, cfqq->slice_end))
323                 return 0;
324
325         return 1;
326 }
327
328 /*
329  * Lifted from AS - choose which of rq1 and rq2 that is best served now.
330  * We choose the request that is closest to the head right now. Distance
331  * behind the head is penalized and only allowed to a certain extent.
332  */
333 static struct request *
334 cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2)
335 {
336         sector_t last, s1, s2, d1 = 0, d2 = 0;
337         unsigned long back_max;
338 #define CFQ_RQ1_WRAP    0x01 /* request 1 wraps */
339 #define CFQ_RQ2_WRAP    0x02 /* request 2 wraps */
340         unsigned wrap = 0; /* bit mask: requests behind the disk head? */
341
342         if (rq1 == NULL || rq1 == rq2)
343                 return rq2;
344         if (rq2 == NULL)
345                 return rq1;
346
347         if (rq_is_sync(rq1) && !rq_is_sync(rq2))
348                 return rq1;
349         else if (rq_is_sync(rq2) && !rq_is_sync(rq1))
350                 return rq2;
351         if (rq_is_meta(rq1) && !rq_is_meta(rq2))
352                 return rq1;
353         else if (rq_is_meta(rq2) && !rq_is_meta(rq1))
354                 return rq2;
355
356         s1 = blk_rq_pos(rq1);
357         s2 = blk_rq_pos(rq2);
358
359         last = cfqd->last_position;
360
361         /*
362          * by definition, 1KiB is 2 sectors
363          */
364         back_max = cfqd->cfq_back_max * 2;
365
366         /*
367          * Strict one way elevator _except_ in the case where we allow
368          * short backward seeks which are biased as twice the cost of a
369          * similar forward seek.
370          */
371         if (s1 >= last)
372                 d1 = s1 - last;
373         else if (s1 + back_max >= last)
374                 d1 = (last - s1) * cfqd->cfq_back_penalty;
375         else
376                 wrap |= CFQ_RQ1_WRAP;
377
378         if (s2 >= last)
379                 d2 = s2 - last;
380         else if (s2 + back_max >= last)
381                 d2 = (last - s2) * cfqd->cfq_back_penalty;
382         else
383                 wrap |= CFQ_RQ2_WRAP;
384
385         /* Found required data */
386
387         /*
388          * By doing switch() on the bit mask "wrap" we avoid having to
389          * check two variables for all permutations: --> faster!
390          */
391         switch (wrap) {
392         case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
393                 if (d1 < d2)
394                         return rq1;
395                 else if (d2 < d1)
396                         return rq2;
397                 else {
398                         if (s1 >= s2)
399                                 return rq1;
400                         else
401                                 return rq2;
402                 }
403
404         case CFQ_RQ2_WRAP:
405                 return rq1;
406         case CFQ_RQ1_WRAP:
407                 return rq2;
408         case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
409         default:
410                 /*
411                  * Since both rqs are wrapped,
412                  * start with the one that's further behind head
413                  * (--> only *one* back seek required),
414                  * since back seek takes more time than forward.
415                  */
416                 if (s1 <= s2)
417                         return rq1;
418                 else
419                         return rq2;
420         }
421 }
422
423 /*
424  * The below is leftmost cache rbtree addon
425  */
426 static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
427 {
428         if (!root->left)
429                 root->left = rb_first(&root->rb);
430
431         if (root->left)
432                 return rb_entry(root->left, struct cfq_queue, rb_node);
433
434         return NULL;
435 }
436
437 static void rb_erase_init(struct rb_node *n, struct rb_root *root)
438 {
439         rb_erase(n, root);
440         RB_CLEAR_NODE(n);
441 }
442
443 static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
444 {
445         if (root->left == n)
446                 root->left = NULL;
447         rb_erase_init(n, &root->rb);
448 }
449
450 /*
451  * would be nice to take fifo expire time into account as well
452  */
453 static struct request *
454 cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
455                   struct request *last)
456 {
457         struct rb_node *rbnext = rb_next(&last->rb_node);
458         struct rb_node *rbprev = rb_prev(&last->rb_node);
459         struct request *next = NULL, *prev = NULL;
460
461         BUG_ON(RB_EMPTY_NODE(&last->rb_node));
462
463         if (rbprev)
464                 prev = rb_entry_rq(rbprev);
465
466         if (rbnext)
467                 next = rb_entry_rq(rbnext);
468         else {
469                 rbnext = rb_first(&cfqq->sort_list);
470                 if (rbnext && rbnext != &last->rb_node)
471                         next = rb_entry_rq(rbnext);
472         }
473
474         return cfq_choose_req(cfqd, next, prev);
475 }
476
477 static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
478                                       struct cfq_queue *cfqq)
479 {
480         /*
481          * just an approximation, should be ok.
482          */
483         return (cfqd->busy_queues - 1) * (cfq_prio_slice(cfqd, 1, 0) -
484                        cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio));
485 }
486
487 /*
488  * The cfqd->service_tree holds all pending cfq_queue's that have
489  * requests waiting to be processed. It is sorted in the order that
490  * we will service the queues.
491  */
492 static void cfq_service_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq,
493                                  int add_front)
494 {
495         struct rb_node **p, *parent;
496         struct cfq_queue *__cfqq;
497         unsigned long rb_key;
498         int left;
499
500         if (cfq_class_idle(cfqq)) {
501                 rb_key = CFQ_IDLE_DELAY;
502                 parent = rb_last(&cfqd->service_tree.rb);
503                 if (parent && parent != &cfqq->rb_node) {
504                         __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
505                         rb_key += __cfqq->rb_key;
506                 } else
507                         rb_key += jiffies;
508         } else if (!add_front) {
509                 rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
510                 rb_key += cfqq->slice_resid;
511                 cfqq->slice_resid = 0;
512         } else
513                 rb_key = 0;
514
515         if (!RB_EMPTY_NODE(&cfqq->rb_node)) {
516                 /*
517                  * same position, nothing more to do
518                  */
519                 if (rb_key == cfqq->rb_key)
520                         return;
521
522                 cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree);
523         }
524
525         left = 1;
526         parent = NULL;
527         p = &cfqd->service_tree.rb.rb_node;
528         while (*p) {
529                 struct rb_node **n;
530
531                 parent = *p;
532                 __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
533
534                 /*
535                  * sort RT queues first, we always want to give
536                  * preference to them. IDLE queues goes to the back.
537                  * after that, sort on the next service time.
538                  */
539                 if (cfq_class_rt(cfqq) > cfq_class_rt(__cfqq))
540                         n = &(*p)->rb_left;
541                 else if (cfq_class_rt(cfqq) < cfq_class_rt(__cfqq))
542                         n = &(*p)->rb_right;
543                 else if (cfq_class_idle(cfqq) < cfq_class_idle(__cfqq))
544                         n = &(*p)->rb_left;
545                 else if (cfq_class_idle(cfqq) > cfq_class_idle(__cfqq))
546                         n = &(*p)->rb_right;
547                 else if (rb_key < __cfqq->rb_key)
548                         n = &(*p)->rb_left;
549                 else
550                         n = &(*p)->rb_right;
551
552                 if (n == &(*p)->rb_right)
553                         left = 0;
554
555                 p = n;
556         }
557
558         if (left)
559                 cfqd->service_tree.left = &cfqq->rb_node;
560
561         cfqq->rb_key = rb_key;
562         rb_link_node(&cfqq->rb_node, parent, p);
563         rb_insert_color(&cfqq->rb_node, &cfqd->service_tree.rb);
564 }
565
566 static struct cfq_queue *
567 cfq_prio_tree_lookup(struct cfq_data *cfqd, struct rb_root *root,
568                      sector_t sector, struct rb_node **ret_parent,
569                      struct rb_node ***rb_link)
570 {
571         struct rb_node **p, *parent;
572         struct cfq_queue *cfqq = NULL;
573
574         parent = NULL;
575         p = &root->rb_node;
576         while (*p) {
577                 struct rb_node **n;
578
579                 parent = *p;
580                 cfqq = rb_entry(parent, struct cfq_queue, p_node);
581
582                 /*
583                  * Sort strictly based on sector.  Smallest to the left,
584                  * largest to the right.
585                  */
586                 if (sector > blk_rq_pos(cfqq->next_rq))
587                         n = &(*p)->rb_right;
588                 else if (sector < blk_rq_pos(cfqq->next_rq))
589                         n = &(*p)->rb_left;
590                 else
591                         break;
592                 p = n;
593                 cfqq = NULL;
594         }
595
596         *ret_parent = parent;
597         if (rb_link)
598                 *rb_link = p;
599         return cfqq;
600 }
601
602 static void cfq_prio_tree_add(struct cfq_data *cfqd, struct cfq_queue *cfqq)
603 {
604         struct rb_node **p, *parent;
605         struct cfq_queue *__cfqq;
606
607         if (cfqq->p_root) {
608                 rb_erase(&cfqq->p_node, cfqq->p_root);
609                 cfqq->p_root = NULL;
610         }
611
612         if (cfq_class_idle(cfqq))
613                 return;
614         if (!cfqq->next_rq)
615                 return;
616
617         cfqq->p_root = &cfqd->prio_trees[cfqq->org_ioprio];
618         __cfqq = cfq_prio_tree_lookup(cfqd, cfqq->p_root,
619                                       blk_rq_pos(cfqq->next_rq), &parent, &p);
620         if (!__cfqq) {
621                 rb_link_node(&cfqq->p_node, parent, p);
622                 rb_insert_color(&cfqq->p_node, cfqq->p_root);
623         } else
624                 cfqq->p_root = NULL;
625 }
626
627 /*
628  * Update cfqq's position in the service tree.
629  */
630 static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
631 {
632         /*
633          * Resorting requires the cfqq to be on the RR list already.
634          */
635         if (cfq_cfqq_on_rr(cfqq)) {
636                 cfq_service_tree_add(cfqd, cfqq, 0);
637                 cfq_prio_tree_add(cfqd, cfqq);
638         }
639 }
640
641 /*
642  * add to busy list of queues for service, trying to be fair in ordering
643  * the pending list according to last request service
644  */
645 static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
646 {
647         cfq_log_cfqq(cfqd, cfqq, "add_to_rr");
648         BUG_ON(cfq_cfqq_on_rr(cfqq));
649         cfq_mark_cfqq_on_rr(cfqq);
650         cfqd->busy_queues++;
651         if (cfq_class_rt(cfqq))
652                 cfqd->busy_rt_queues++;
653
654         cfq_resort_rr_list(cfqd, cfqq);
655 }
656
657 /*
658  * Called when the cfqq no longer has requests pending, remove it from
659  * the service tree.
660  */
661 static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
662 {
663         cfq_log_cfqq(cfqd, cfqq, "del_from_rr");
664         BUG_ON(!cfq_cfqq_on_rr(cfqq));
665         cfq_clear_cfqq_on_rr(cfqq);
666
667         if (!RB_EMPTY_NODE(&cfqq->rb_node))
668                 cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree);
669         if (cfqq->p_root) {
670                 rb_erase(&cfqq->p_node, cfqq->p_root);
671                 cfqq->p_root = NULL;
672         }
673
674         BUG_ON(!cfqd->busy_queues);
675         cfqd->busy_queues--;
676         if (cfq_class_rt(cfqq))
677                 cfqd->busy_rt_queues--;
678 }
679
680 /*
681  * rb tree support functions
682  */
683 static void cfq_del_rq_rb(struct request *rq)
684 {
685         struct cfq_queue *cfqq = RQ_CFQQ(rq);
686         struct cfq_data *cfqd = cfqq->cfqd;
687         const int sync = rq_is_sync(rq);
688
689         BUG_ON(!cfqq->queued[sync]);
690         cfqq->queued[sync]--;
691
692         elv_rb_del(&cfqq->sort_list, rq);
693
694         if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
695                 cfq_del_cfqq_rr(cfqd, cfqq);
696 }
697
698 static void cfq_add_rq_rb(struct request *rq)
699 {
700         struct cfq_queue *cfqq = RQ_CFQQ(rq);
701         struct cfq_data *cfqd = cfqq->cfqd;
702         struct request *__alias, *prev;
703
704         cfqq->queued[rq_is_sync(rq)]++;
705
706         /*
707          * looks a little odd, but the first insert might return an alias.
708          * if that happens, put the alias on the dispatch list
709          */
710         while ((__alias = elv_rb_add(&cfqq->sort_list, rq)) != NULL)
711                 cfq_dispatch_insert(cfqd->queue, __alias);
712
713         if (!cfq_cfqq_on_rr(cfqq))
714                 cfq_add_cfqq_rr(cfqd, cfqq);
715
716         /*
717          * check if this request is a better next-serve candidate
718          */
719         prev = cfqq->next_rq;
720         cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq);
721
722         /*
723          * adjust priority tree position, if ->next_rq changes
724          */
725         if (prev != cfqq->next_rq)
726                 cfq_prio_tree_add(cfqd, cfqq);
727
728         BUG_ON(!cfqq->next_rq);
729 }
730
731 static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
732 {
733         elv_rb_del(&cfqq->sort_list, rq);
734         cfqq->queued[rq_is_sync(rq)]--;
735         cfq_add_rq_rb(rq);
736 }
737
738 static struct request *
739 cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
740 {
741         struct task_struct *tsk = current;
742         struct cfq_io_context *cic;
743         struct cfq_queue *cfqq;
744
745         cic = cfq_cic_lookup(cfqd, tsk->io_context);
746         if (!cic)
747                 return NULL;
748
749         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
750         if (cfqq) {
751                 sector_t sector = bio->bi_sector + bio_sectors(bio);
752
753                 return elv_rb_find(&cfqq->sort_list, sector);
754         }
755
756         return NULL;
757 }
758
759 static void cfq_activate_request(struct request_queue *q, struct request *rq)
760 {
761         struct cfq_data *cfqd = q->elevator->elevator_data;
762
763         cfqd->rq_in_driver++;
764         cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d",
765                                                 cfqd->rq_in_driver);
766
767         cfqd->last_position = blk_rq_pos(rq) + blk_rq_sectors(rq);
768 }
769
770 static void cfq_deactivate_request(struct request_queue *q, struct request *rq)
771 {
772         struct cfq_data *cfqd = q->elevator->elevator_data;
773
774         WARN_ON(!cfqd->rq_in_driver);
775         cfqd->rq_in_driver--;
776         cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d",
777                                                 cfqd->rq_in_driver);
778 }
779
780 static void cfq_remove_request(struct request *rq)
781 {
782         struct cfq_queue *cfqq = RQ_CFQQ(rq);
783
784         if (cfqq->next_rq == rq)
785                 cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
786
787         list_del_init(&rq->queuelist);
788         cfq_del_rq_rb(rq);
789
790         cfqq->cfqd->rq_queued--;
791         if (rq_is_meta(rq)) {
792                 WARN_ON(!cfqq->meta_pending);
793                 cfqq->meta_pending--;
794         }
795 }
796
797 static int cfq_merge(struct request_queue *q, struct request **req,
798                      struct bio *bio)
799 {
800         struct cfq_data *cfqd = q->elevator->elevator_data;
801         struct request *__rq;
802
803         __rq = cfq_find_rq_fmerge(cfqd, bio);
804         if (__rq && elv_rq_merge_ok(__rq, bio)) {
805                 *req = __rq;
806                 return ELEVATOR_FRONT_MERGE;
807         }
808
809         return ELEVATOR_NO_MERGE;
810 }
811
812 static void cfq_merged_request(struct request_queue *q, struct request *req,
813                                int type)
814 {
815         if (type == ELEVATOR_FRONT_MERGE) {
816                 struct cfq_queue *cfqq = RQ_CFQQ(req);
817
818                 cfq_reposition_rq_rb(cfqq, req);
819         }
820 }
821
822 static void
823 cfq_merged_requests(struct request_queue *q, struct request *rq,
824                     struct request *next)
825 {
826         /*
827          * reposition in fifo if next is older than rq
828          */
829         if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
830             time_before(next->start_time, rq->start_time))
831                 list_move(&rq->queuelist, &next->queuelist);
832
833         cfq_remove_request(next);
834 }
835
836 static int cfq_allow_merge(struct request_queue *q, struct request *rq,
837                            struct bio *bio)
838 {
839         struct cfq_data *cfqd = q->elevator->elevator_data;
840         struct cfq_io_context *cic;
841         struct cfq_queue *cfqq;
842
843         /*
844          * Disallow merge of a sync bio into an async request.
845          */
846         if (cfq_bio_sync(bio) && !rq_is_sync(rq))
847                 return 0;
848
849         /*
850          * Lookup the cfqq that this bio will be queued with. Allow
851          * merge only if rq is queued there.
852          */
853         cic = cfq_cic_lookup(cfqd, current->io_context);
854         if (!cic)
855                 return 0;
856
857         cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio));
858         if (cfqq == RQ_CFQQ(rq))
859                 return 1;
860
861         return 0;
862 }
863
864 static void __cfq_set_active_queue(struct cfq_data *cfqd,
865                                    struct cfq_queue *cfqq)
866 {
867         if (cfqq) {
868                 cfq_log_cfqq(cfqd, cfqq, "set_active");
869                 cfqq->slice_end = 0;
870                 cfqq->slice_dispatch = 0;
871
872                 cfq_clear_cfqq_wait_request(cfqq);
873                 cfq_clear_cfqq_must_dispatch(cfqq);
874                 cfq_clear_cfqq_must_alloc_slice(cfqq);
875                 cfq_clear_cfqq_fifo_expire(cfqq);
876                 cfq_mark_cfqq_slice_new(cfqq);
877
878                 del_timer(&cfqd->idle_slice_timer);
879         }
880
881         cfqd->active_queue = cfqq;
882 }
883
884 /*
885  * current cfqq expired its slice (or was too idle), select new one
886  */
887 static void
888 __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
889                     int timed_out)
890 {
891         cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out);
892
893         if (cfq_cfqq_wait_request(cfqq))
894                 del_timer(&cfqd->idle_slice_timer);
895
896         cfq_clear_cfqq_wait_request(cfqq);
897
898         /*
899          * store what was left of this slice, if the queue idled/timed out
900          */
901         if (timed_out && !cfq_cfqq_slice_new(cfqq)) {
902                 cfqq->slice_resid = cfqq->slice_end - jiffies;
903                 cfq_log_cfqq(cfqd, cfqq, "resid=%ld", cfqq->slice_resid);
904         }
905
906         cfq_resort_rr_list(cfqd, cfqq);
907
908         if (cfqq == cfqd->active_queue)
909                 cfqd->active_queue = NULL;
910
911         if (cfqd->active_cic) {
912                 put_io_context(cfqd->active_cic->ioc);
913                 cfqd->active_cic = NULL;
914         }
915 }
916
917 static inline void cfq_slice_expired(struct cfq_data *cfqd, int timed_out)
918 {
919         struct cfq_queue *cfqq = cfqd->active_queue;
920
921         if (cfqq)
922                 __cfq_slice_expired(cfqd, cfqq, timed_out);
923 }
924
925 /*
926  * Get next queue for service. Unless we have a queue preemption,
927  * we'll simply select the first cfqq in the service tree.
928  */
929 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
930 {
931         if (RB_EMPTY_ROOT(&cfqd->service_tree.rb))
932                 return NULL;
933
934         return cfq_rb_first(&cfqd->service_tree);
935 }
936
937 /*
938  * Get and set a new active queue for service.
939  */
940 static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd,
941                                               struct cfq_queue *cfqq)
942 {
943         if (!cfqq) {
944                 cfqq = cfq_get_next_queue(cfqd);
945                 if (cfqq)
946                         cfq_clear_cfqq_coop(cfqq);
947         }
948
949         __cfq_set_active_queue(cfqd, cfqq);
950         return cfqq;
951 }
952
953 static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd,
954                                           struct request *rq)
955 {
956         if (blk_rq_pos(rq) >= cfqd->last_position)
957                 return blk_rq_pos(rq) - cfqd->last_position;
958         else
959                 return cfqd->last_position - blk_rq_pos(rq);
960 }
961
962 #define CIC_SEEK_THR    8 * 1024
963 #define CIC_SEEKY(cic)  ((cic)->seek_mean > CIC_SEEK_THR)
964
965 static inline int cfq_rq_close(struct cfq_data *cfqd, struct request *rq)
966 {
967         struct cfq_io_context *cic = cfqd->active_cic;
968         sector_t sdist = cic->seek_mean;
969
970         if (!sample_valid(cic->seek_samples))
971                 sdist = CIC_SEEK_THR;
972
973         return cfq_dist_from_last(cfqd, rq) <= sdist;
974 }
975
976 static struct cfq_queue *cfqq_close(struct cfq_data *cfqd,
977                                     struct cfq_queue *cur_cfqq)
978 {
979         struct rb_root *root = &cfqd->prio_trees[cur_cfqq->org_ioprio];
980         struct rb_node *parent, *node;
981         struct cfq_queue *__cfqq;
982         sector_t sector = cfqd->last_position;
983
984         if (RB_EMPTY_ROOT(root))
985                 return NULL;
986
987         /*
988          * First, if we find a request starting at the end of the last
989          * request, choose it.
990          */
991         __cfqq = cfq_prio_tree_lookup(cfqd, root, sector, &parent, NULL);
992         if (__cfqq)
993                 return __cfqq;
994
995         /*
996          * If the exact sector wasn't found, the parent of the NULL leaf
997          * will contain the closest sector.
998          */
999         __cfqq = rb_entry(parent, struct cfq_queue, p_node);
1000         if (cfq_rq_close(cfqd, __cfqq->next_rq))
1001                 return __cfqq;
1002
1003         if (blk_rq_pos(__cfqq->next_rq) < sector)
1004                 node = rb_next(&__cfqq->p_node);
1005         else
1006                 node = rb_prev(&__cfqq->p_node);
1007         if (!node)
1008                 return NULL;
1009
1010         __cfqq = rb_entry(node, struct cfq_queue, p_node);
1011         if (cfq_rq_close(cfqd, __cfqq->next_rq))
1012                 return __cfqq;
1013
1014         return NULL;
1015 }
1016
1017 /*
1018  * cfqd - obvious
1019  * cur_cfqq - passed in so that we don't decide that the current queue is
1020  *            closely cooperating with itself.
1021  *
1022  * So, basically we're assuming that that cur_cfqq has dispatched at least
1023  * one request, and that cfqd->last_position reflects a position on the disk
1024  * associated with the I/O issued by cur_cfqq.  I'm not sure this is a valid
1025  * assumption.
1026  */
1027 static struct cfq_queue *cfq_close_cooperator(struct cfq_data *cfqd,
1028                                               struct cfq_queue *cur_cfqq,
1029                                               int probe)
1030 {
1031         struct cfq_queue *cfqq;
1032
1033         /*
1034          * A valid cfq_io_context is necessary to compare requests against
1035          * the seek_mean of the current cfqq.
1036          */
1037         if (!cfqd->active_cic)
1038                 return NULL;
1039
1040         /*
1041          * We should notice if some of the queues are cooperating, eg
1042          * working closely on the same area of the disk. In that case,
1043          * we can group them together and don't waste time idling.
1044          */
1045         cfqq = cfqq_close(cfqd, cur_cfqq);
1046         if (!cfqq)
1047                 return NULL;
1048
1049         if (cfq_cfqq_coop(cfqq))
1050                 return NULL;
1051
1052         if (!probe)
1053                 cfq_mark_cfqq_coop(cfqq);
1054         return cfqq;
1055 }
1056
1057 static void cfq_arm_slice_timer(struct cfq_data *cfqd)
1058 {
1059         struct cfq_queue *cfqq = cfqd->active_queue;
1060         struct cfq_io_context *cic;
1061         unsigned long sl;
1062
1063         /*
1064          * SSD device without seek penalty, disable idling. But only do so
1065          * for devices that support queuing, otherwise we still have a problem
1066          * with sync vs async workloads.
1067          */
1068         if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag)
1069                 return;
1070
1071         WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
1072         WARN_ON(cfq_cfqq_slice_new(cfqq));
1073
1074         /*
1075          * idle is disabled, either manually or by past process history
1076          */
1077         if (!cfqd->cfq_slice_idle || !cfq_cfqq_idle_window(cfqq))
1078                 return;
1079
1080         /*
1081          * still requests with the driver, don't idle
1082          */
1083         if (cfqd->rq_in_driver)
1084                 return;
1085
1086         /*
1087          * task has exited, don't wait
1088          */
1089         cic = cfqd->active_cic;
1090         if (!cic || !atomic_read(&cic->ioc->nr_tasks))
1091                 return;
1092
1093         cfq_mark_cfqq_wait_request(cfqq);
1094
1095         /*
1096          * we don't want to idle for seeks, but we do want to allow
1097          * fair distribution of slice time for a process doing back-to-back
1098          * seeks. so allow a little bit of time for him to submit a new rq
1099          */
1100         sl = cfqd->cfq_slice_idle;
1101         if (sample_valid(cic->seek_samples) && CIC_SEEKY(cic))
1102                 sl = min(sl, msecs_to_jiffies(CFQ_MIN_TT));
1103
1104         mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
1105         cfq_log_cfqq(cfqd, cfqq, "arm_idle: %lu", sl);
1106 }
1107
1108 /*
1109  * Move request from internal lists to the request queue dispatch list.
1110  */
1111 static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
1112 {
1113         struct cfq_data *cfqd = q->elevator->elevator_data;
1114         struct cfq_queue *cfqq = RQ_CFQQ(rq);
1115
1116         cfq_log_cfqq(cfqd, cfqq, "dispatch_insert");
1117
1118         cfq_remove_request(rq);
1119         cfqq->dispatched++;
1120         elv_dispatch_sort(q, rq);
1121
1122         if (cfq_cfqq_sync(cfqq))
1123                 cfqd->sync_flight++;
1124 }
1125
1126 /*
1127  * return expired entry, or NULL to just start from scratch in rbtree
1128  */
1129 static struct request *cfq_check_fifo(struct cfq_queue *cfqq)
1130 {
1131         struct cfq_data *cfqd = cfqq->cfqd;
1132         struct request *rq;
1133         int fifo;
1134
1135         if (cfq_cfqq_fifo_expire(cfqq))
1136                 return NULL;
1137
1138         cfq_mark_cfqq_fifo_expire(cfqq);
1139
1140         if (list_empty(&cfqq->fifo))
1141                 return NULL;
1142
1143         fifo = cfq_cfqq_sync(cfqq);
1144         rq = rq_entry_fifo(cfqq->fifo.next);
1145
1146         if (time_before(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo]))
1147                 rq = NULL;
1148
1149         cfq_log_cfqq(cfqd, cfqq, "fifo=%p", rq);
1150         return rq;
1151 }
1152
1153 static inline int
1154 cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1155 {
1156         const int base_rq = cfqd->cfq_slice_async_rq;
1157
1158         WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
1159
1160         return 2 * (base_rq + base_rq * (CFQ_PRIO_LISTS - 1 - cfqq->ioprio));
1161 }
1162
1163 /*
1164  * Select a queue for service. If we have a current active queue,
1165  * check whether to continue servicing it, or retrieve and set a new one.
1166  */
1167 static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
1168 {
1169         struct cfq_queue *cfqq, *new_cfqq = NULL;
1170
1171         cfqq = cfqd->active_queue;
1172         if (!cfqq)
1173                 goto new_queue;
1174
1175         /*
1176          * The active queue has run out of time, expire it and select new.
1177          */
1178         if (cfq_slice_used(cfqq) && !cfq_cfqq_must_dispatch(cfqq))
1179                 goto expire;
1180
1181         /*
1182          * If we have a RT cfqq waiting, then we pre-empt the current non-rt
1183          * cfqq.
1184          */
1185         if (!cfq_class_rt(cfqq) && cfqd->busy_rt_queues) {
1186                 /*
1187                  * We simulate this as cfqq timed out so that it gets to bank
1188                  * the remaining of its time slice.
1189                  */
1190                 cfq_log_cfqq(cfqd, cfqq, "preempt");
1191                 cfq_slice_expired(cfqd, 1);
1192                 goto new_queue;
1193         }
1194
1195         /*
1196          * The active queue has requests and isn't expired, allow it to
1197          * dispatch.
1198          */
1199         if (!RB_EMPTY_ROOT(&cfqq->sort_list))
1200                 goto keep_queue;
1201
1202         /*
1203          * If another queue has a request waiting within our mean seek
1204          * distance, let it run.  The expire code will check for close
1205          * cooperators and put the close queue at the front of the service
1206          * tree.
1207          */
1208         new_cfqq = cfq_close_cooperator(cfqd, cfqq, 0);
1209         if (new_cfqq)
1210                 goto expire;
1211
1212         /*
1213          * No requests pending. If the active queue still has requests in
1214          * flight or is idling for a new request, allow either of these
1215          * conditions to happen (or time out) before selecting a new queue.
1216          */
1217         if (timer_pending(&cfqd->idle_slice_timer) ||
1218             (cfqq->dispatched && cfq_cfqq_idle_window(cfqq))) {
1219                 cfqq = NULL;
1220                 goto keep_queue;
1221         }
1222
1223 expire:
1224         cfq_slice_expired(cfqd, 0);
1225 new_queue:
1226         cfqq = cfq_set_active_queue(cfqd, new_cfqq);
1227 keep_queue:
1228         return cfqq;
1229 }
1230
1231 static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
1232 {
1233         int dispatched = 0;
1234
1235         while (cfqq->next_rq) {
1236                 cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
1237                 dispatched++;
1238         }
1239
1240         BUG_ON(!list_empty(&cfqq->fifo));
1241         return dispatched;
1242 }
1243
1244 /*
1245  * Drain our current requests. Used for barriers and when switching
1246  * io schedulers on-the-fly.
1247  */
1248 static int cfq_forced_dispatch(struct cfq_data *cfqd)
1249 {
1250         struct cfq_queue *cfqq;
1251         int dispatched = 0;
1252
1253         while ((cfqq = cfq_rb_first(&cfqd->service_tree)) != NULL)
1254                 dispatched += __cfq_forced_dispatch_cfqq(cfqq);
1255
1256         cfq_slice_expired(cfqd, 0);
1257
1258         BUG_ON(cfqd->busy_queues);
1259
1260         cfq_log(cfqd, "forced_dispatch=%d", dispatched);
1261         return dispatched;
1262 }
1263
1264 /*
1265  * Dispatch a request from cfqq, moving them to the request queue
1266  * dispatch list.
1267  */
1268 static void cfq_dispatch_request(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1269 {
1270         struct request *rq;
1271
1272         BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
1273
1274         /*
1275          * follow expired path, else get first next available
1276          */
1277         rq = cfq_check_fifo(cfqq);
1278         if (!rq)
1279                 rq = cfqq->next_rq;
1280
1281         /*
1282          * insert request into driver dispatch list
1283          */
1284         cfq_dispatch_insert(cfqd->queue, rq);
1285
1286         if (!cfqd->active_cic) {
1287                 struct cfq_io_context *cic = RQ_CIC(rq);
1288
1289                 atomic_long_inc(&cic->ioc->refcount);
1290                 cfqd->active_cic = cic;
1291         }
1292 }
1293
1294 /*
1295  * Find the cfqq that we need to service and move a request from that to the
1296  * dispatch list
1297  */
1298 static int cfq_dispatch_requests(struct request_queue *q, int force)
1299 {
1300         struct cfq_data *cfqd = q->elevator->elevator_data;
1301         struct cfq_queue *cfqq;
1302         unsigned int max_dispatch;
1303
1304         if (!cfqd->busy_queues)
1305                 return 0;
1306
1307         if (unlikely(force))
1308                 return cfq_forced_dispatch(cfqd);
1309
1310         cfqq = cfq_select_queue(cfqd);
1311         if (!cfqq)
1312                 return 0;
1313
1314         /*
1315          * If this is an async queue and we have sync IO in flight, let it wait
1316          */
1317         if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq))
1318                 return 0;
1319
1320         max_dispatch = cfqd->cfq_quantum;
1321         if (cfq_class_idle(cfqq))
1322                 max_dispatch = 1;
1323
1324         /*
1325          * Does this cfqq already have too much IO in flight?
1326          */
1327         if (cfqq->dispatched >= max_dispatch) {
1328                 /*
1329                  * idle queue must always only have a single IO in flight
1330                  */
1331                 if (cfq_class_idle(cfqq))
1332                         return 0;
1333
1334                 /*
1335                  * We have other queues, don't allow more IO from this one
1336                  */
1337                 if (cfqd->busy_queues > 1)
1338                         return 0;
1339
1340                 /*
1341                  * we are the only queue, allow up to 4 times of 'quantum'
1342                  */
1343                 if (cfqq->dispatched >= 4 * max_dispatch)
1344                         return 0;
1345         }
1346
1347         /*
1348          * Dispatch a request from this cfqq
1349          */
1350         cfq_dispatch_request(cfqd, cfqq);
1351         cfqq->slice_dispatch++;
1352         cfq_clear_cfqq_must_dispatch(cfqq);
1353
1354         /*
1355          * expire an async queue immediately if it has used up its slice. idle
1356          * queue always expire after 1 dispatch round.
1357          */
1358         if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) &&
1359             cfqq->slice_dispatch >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
1360             cfq_class_idle(cfqq))) {
1361                 cfqq->slice_end = jiffies + 1;
1362                 cfq_slice_expired(cfqd, 0);
1363         }
1364
1365         cfq_log(cfqd, "dispatched a request");
1366         return 1;
1367 }
1368
1369 /*
1370  * task holds one reference to the queue, dropped when task exits. each rq
1371  * in-flight on this queue also holds a reference, dropped when rq is freed.
1372  *
1373  * queue lock must be held here.
1374  */
1375 static void cfq_put_queue(struct cfq_queue *cfqq)
1376 {
1377         struct cfq_data *cfqd = cfqq->cfqd;
1378
1379         BUG_ON(atomic_read(&cfqq->ref) <= 0);
1380
1381         if (!atomic_dec_and_test(&cfqq->ref))
1382                 return;
1383
1384         cfq_log_cfqq(cfqd, cfqq, "put_queue");
1385         BUG_ON(rb_first(&cfqq->sort_list));
1386         BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
1387         BUG_ON(cfq_cfqq_on_rr(cfqq));
1388
1389         if (unlikely(cfqd->active_queue == cfqq)) {
1390                 __cfq_slice_expired(cfqd, cfqq, 0);
1391                 cfq_schedule_dispatch(cfqd);
1392         }
1393
1394         kmem_cache_free(cfq_pool, cfqq);
1395 }
1396
1397 /*
1398  * Must always be called with the rcu_read_lock() held
1399  */
1400 static void
1401 __call_for_each_cic(struct io_context *ioc,
1402                     void (*func)(struct io_context *, struct cfq_io_context *))
1403 {
1404         struct cfq_io_context *cic;
1405         struct hlist_node *n;
1406
1407         hlist_for_each_entry_rcu(cic, n, &ioc->cic_list, cic_list)
1408                 func(ioc, cic);
1409 }
1410
1411 /*
1412  * Call func for each cic attached to this ioc.
1413  */
1414 static void
1415 call_for_each_cic(struct io_context *ioc,
1416                   void (*func)(struct io_context *, struct cfq_io_context *))
1417 {
1418         rcu_read_lock();
1419         __call_for_each_cic(ioc, func);
1420         rcu_read_unlock();
1421 }
1422
1423 static void cfq_cic_free_rcu(struct rcu_head *head)
1424 {
1425         struct cfq_io_context *cic;
1426
1427         cic = container_of(head, struct cfq_io_context, rcu_head);
1428
1429         kmem_cache_free(cfq_ioc_pool, cic);
1430         elv_ioc_count_dec(ioc_count);
1431
1432         if (ioc_gone) {
1433                 /*
1434                  * CFQ scheduler is exiting, grab exit lock and check
1435                  * the pending io context count. If it hits zero,
1436                  * complete ioc_gone and set it back to NULL
1437                  */
1438                 spin_lock(&ioc_gone_lock);
1439                 if (ioc_gone && !elv_ioc_count_read(ioc_count)) {
1440                         complete(ioc_gone);
1441                         ioc_gone = NULL;
1442                 }
1443                 spin_unlock(&ioc_gone_lock);
1444         }
1445 }
1446
1447 static void cfq_cic_free(struct cfq_io_context *cic)
1448 {
1449         call_rcu(&cic->rcu_head, cfq_cic_free_rcu);
1450 }
1451
1452 static void cic_free_func(struct io_context *ioc, struct cfq_io_context *cic)
1453 {
1454         unsigned long flags;
1455
1456         BUG_ON(!cic->dead_key);
1457
1458         spin_lock_irqsave(&ioc->lock, flags);
1459         radix_tree_delete(&ioc->radix_root, cic->dead_key);
1460         hlist_del_rcu(&cic->cic_list);
1461         spin_unlock_irqrestore(&ioc->lock, flags);
1462
1463         cfq_cic_free(cic);
1464 }
1465
1466 /*
1467  * Must be called with rcu_read_lock() held or preemption otherwise disabled.
1468  * Only two callers of this - ->dtor() which is called with the rcu_read_lock(),
1469  * and ->trim() which is called with the task lock held
1470  */
1471 static void cfq_free_io_context(struct io_context *ioc)
1472 {
1473         /*
1474          * ioc->refcount is zero here, or we are called from elv_unregister(),
1475          * so no more cic's are allowed to be linked into this ioc.  So it
1476          * should be ok to iterate over the known list, we will see all cic's
1477          * since no new ones are added.
1478          */
1479         __call_for_each_cic(ioc, cic_free_func);
1480 }
1481
1482 static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
1483 {
1484         if (unlikely(cfqq == cfqd->active_queue)) {
1485                 __cfq_slice_expired(cfqd, cfqq, 0);
1486                 cfq_schedule_dispatch(cfqd);
1487         }
1488
1489         cfq_put_queue(cfqq);
1490 }
1491
1492 static void __cfq_exit_single_io_context(struct cfq_data *cfqd,
1493                                          struct cfq_io_context *cic)
1494 {
1495         struct io_context *ioc = cic->ioc;
1496
1497         list_del_init(&cic->queue_list);
1498
1499         /*
1500          * Make sure key == NULL is seen for dead queues
1501          */
1502         smp_wmb();
1503         cic->dead_key = (unsigned long) cic->key;
1504         cic->key = NULL;
1505
1506         if (ioc->ioc_data == cic)
1507                 rcu_assign_pointer(ioc->ioc_data, NULL);
1508
1509         if (cic->cfqq[BLK_RW_ASYNC]) {
1510                 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_ASYNC]);
1511                 cic->cfqq[BLK_RW_ASYNC] = NULL;
1512         }
1513
1514         if (cic->cfqq[BLK_RW_SYNC]) {
1515                 cfq_exit_cfqq(cfqd, cic->cfqq[BLK_RW_SYNC]);
1516                 cic->cfqq[BLK_RW_SYNC] = NULL;
1517         }
1518 }
1519
1520 static void cfq_exit_single_io_context(struct io_context *ioc,
1521                                        struct cfq_io_context *cic)
1522 {
1523         struct cfq_data *cfqd = cic->key;
1524
1525         if (cfqd) {
1526                 struct request_queue *q = cfqd->queue;
1527                 unsigned long flags;
1528
1529                 spin_lock_irqsave(q->queue_lock, flags);
1530
1531                 /*
1532                  * Ensure we get a fresh copy of the ->key to prevent
1533                  * race between exiting task and queue
1534                  */
1535                 smp_read_barrier_depends();
1536                 if (cic->key)
1537                         __cfq_exit_single_io_context(cfqd, cic);
1538
1539                 spin_unlock_irqrestore(q->queue_lock, flags);
1540         }
1541 }
1542
1543 /*
1544  * The process that ioc belongs to has exited, we need to clean up
1545  * and put the internal structures we have that belongs to that process.
1546  */
1547 static void cfq_exit_io_context(struct io_context *ioc)
1548 {
1549         call_for_each_cic(ioc, cfq_exit_single_io_context);
1550 }
1551
1552 static struct cfq_io_context *
1553 cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
1554 {
1555         struct cfq_io_context *cic;
1556
1557         cic = kmem_cache_alloc_node(cfq_ioc_pool, gfp_mask | __GFP_ZERO,
1558                                                         cfqd->queue->node);
1559         if (cic) {
1560                 cic->last_end_request = jiffies;
1561                 INIT_LIST_HEAD(&cic->queue_list);
1562                 INIT_HLIST_NODE(&cic->cic_list);
1563                 cic->dtor = cfq_free_io_context;
1564                 cic->exit = cfq_exit_io_context;
1565                 elv_ioc_count_inc(ioc_count);
1566         }
1567
1568         return cic;
1569 }
1570
1571 static void cfq_init_prio_data(struct cfq_queue *cfqq, struct io_context *ioc)
1572 {
1573         struct task_struct *tsk = current;
1574         int ioprio_class;
1575
1576         if (!cfq_cfqq_prio_changed(cfqq))
1577                 return;
1578
1579         ioprio_class = IOPRIO_PRIO_CLASS(ioc->ioprio);
1580         switch (ioprio_class) {
1581         default:
1582                 printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
1583         case IOPRIO_CLASS_NONE:
1584                 /*
1585                  * no prio set, inherit CPU scheduling settings
1586                  */
1587                 cfqq->ioprio = task_nice_ioprio(tsk);
1588                 cfqq->ioprio_class = task_nice_ioclass(tsk);
1589                 break;
1590         case IOPRIO_CLASS_RT:
1591                 cfqq->ioprio = task_ioprio(ioc);
1592                 cfqq->ioprio_class = IOPRIO_CLASS_RT;
1593                 break;
1594         case IOPRIO_CLASS_BE:
1595                 cfqq->ioprio = task_ioprio(ioc);
1596                 cfqq->ioprio_class = IOPRIO_CLASS_BE;
1597                 break;
1598         case IOPRIO_CLASS_IDLE:
1599                 cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
1600                 cfqq->ioprio = 7;
1601                 cfq_clear_cfqq_idle_window(cfqq);
1602                 break;
1603         }
1604
1605         /*
1606          * keep track of original prio settings in case we have to temporarily
1607          * elevate the priority of this queue
1608          */
1609         cfqq->org_ioprio = cfqq->ioprio;
1610         cfqq->org_ioprio_class = cfqq->ioprio_class;
1611         cfq_clear_cfqq_prio_changed(cfqq);
1612 }
1613
1614 static void changed_ioprio(struct io_context *ioc, struct cfq_io_context *cic)
1615 {
1616         struct cfq_data *cfqd = cic->key;
1617         struct cfq_queue *cfqq;
1618         unsigned long flags;
1619
1620         if (unlikely(!cfqd))
1621                 return;
1622
1623         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
1624
1625         cfqq = cic->cfqq[BLK_RW_ASYNC];
1626         if (cfqq) {
1627                 struct cfq_queue *new_cfqq;
1628                 new_cfqq = cfq_get_queue(cfqd, BLK_RW_ASYNC, cic->ioc,
1629                                                 GFP_ATOMIC);
1630                 if (new_cfqq) {
1631                         cic->cfqq[BLK_RW_ASYNC] = new_cfqq;
1632                         cfq_put_queue(cfqq);
1633                 }
1634         }
1635
1636         cfqq = cic->cfqq[BLK_RW_SYNC];
1637         if (cfqq)
1638                 cfq_mark_cfqq_prio_changed(cfqq);
1639
1640         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
1641 }
1642
1643 static void cfq_ioc_set_ioprio(struct io_context *ioc)
1644 {
1645         call_for_each_cic(ioc, changed_ioprio);
1646         ioc->ioprio_changed = 0;
1647 }
1648
1649 static void cfq_init_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1650                           pid_t pid, int is_sync)
1651 {
1652         RB_CLEAR_NODE(&cfqq->rb_node);
1653         RB_CLEAR_NODE(&cfqq->p_node);
1654         INIT_LIST_HEAD(&cfqq->fifo);
1655
1656         atomic_set(&cfqq->ref, 0);
1657         cfqq->cfqd = cfqd;
1658
1659         cfq_mark_cfqq_prio_changed(cfqq);
1660
1661         if (is_sync) {
1662                 if (!cfq_class_idle(cfqq))
1663                         cfq_mark_cfqq_idle_window(cfqq);
1664                 cfq_mark_cfqq_sync(cfqq);
1665         }
1666         cfqq->pid = pid;
1667 }
1668
1669 static struct cfq_queue *
1670 cfq_find_alloc_queue(struct cfq_data *cfqd, int is_sync,
1671                      struct io_context *ioc, gfp_t gfp_mask)
1672 {
1673         struct cfq_queue *cfqq, *new_cfqq = NULL;
1674         struct cfq_io_context *cic;
1675
1676 retry:
1677         cic = cfq_cic_lookup(cfqd, ioc);
1678         /* cic always exists here */
1679         cfqq = cic_to_cfqq(cic, is_sync);
1680
1681         /*
1682          * Always try a new alloc if we fell back to the OOM cfqq
1683          * originally, since it should just be a temporary situation.
1684          */
1685         if (!cfqq || cfqq == &cfqd->oom_cfqq) {
1686                 cfqq = NULL;
1687                 if (new_cfqq) {
1688                         cfqq = new_cfqq;
1689                         new_cfqq = NULL;
1690                 } else if (gfp_mask & __GFP_WAIT) {
1691                         spin_unlock_irq(cfqd->queue->queue_lock);
1692                         new_cfqq = kmem_cache_alloc_node(cfq_pool,
1693                                         gfp_mask | __GFP_ZERO,
1694                                         cfqd->queue->node);
1695                         spin_lock_irq(cfqd->queue->queue_lock);
1696                         if (new_cfqq)
1697                                 goto retry;
1698                 } else {
1699                         cfqq = kmem_cache_alloc_node(cfq_pool,
1700                                         gfp_mask | __GFP_ZERO,
1701                                         cfqd->queue->node);
1702                 }
1703
1704                 if (cfqq) {
1705                         cfq_init_cfqq(cfqd, cfqq, current->pid, is_sync);
1706                         cfq_init_prio_data(cfqq, ioc);
1707                         cfq_log_cfqq(cfqd, cfqq, "alloced");
1708                 } else
1709                         cfqq = &cfqd->oom_cfqq;
1710         }
1711
1712         if (new_cfqq)
1713                 kmem_cache_free(cfq_pool, new_cfqq);
1714
1715         return cfqq;
1716 }
1717
1718 static struct cfq_queue **
1719 cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio)
1720 {
1721         switch (ioprio_class) {
1722         case IOPRIO_CLASS_RT:
1723                 return &cfqd->async_cfqq[0][ioprio];
1724         case IOPRIO_CLASS_BE:
1725                 return &cfqd->async_cfqq[1][ioprio];
1726         case IOPRIO_CLASS_IDLE:
1727                 return &cfqd->async_idle_cfqq;
1728         default:
1729                 BUG();
1730         }
1731 }
1732
1733 static struct cfq_queue *
1734 cfq_get_queue(struct cfq_data *cfqd, int is_sync, struct io_context *ioc,
1735               gfp_t gfp_mask)
1736 {
1737         const int ioprio = task_ioprio(ioc);
1738         const int ioprio_class = task_ioprio_class(ioc);
1739         struct cfq_queue **async_cfqq = NULL;
1740         struct cfq_queue *cfqq = NULL;
1741
1742         if (!is_sync) {
1743                 async_cfqq = cfq_async_queue_prio(cfqd, ioprio_class, ioprio);
1744                 cfqq = *async_cfqq;
1745         }
1746
1747         if (!cfqq)
1748                 cfqq = cfq_find_alloc_queue(cfqd, is_sync, ioc, gfp_mask);
1749
1750         /*
1751          * pin the queue now that it's allocated, scheduler exit will prune it
1752          */
1753         if (!is_sync && !(*async_cfqq)) {
1754                 atomic_inc(&cfqq->ref);
1755                 *async_cfqq = cfqq;
1756         }
1757
1758         atomic_inc(&cfqq->ref);
1759         return cfqq;
1760 }
1761
1762 /*
1763  * We drop cfq io contexts lazily, so we may find a dead one.
1764  */
1765 static void
1766 cfq_drop_dead_cic(struct cfq_data *cfqd, struct io_context *ioc,
1767                   struct cfq_io_context *cic)
1768 {
1769         unsigned long flags;
1770
1771         WARN_ON(!list_empty(&cic->queue_list));
1772
1773         spin_lock_irqsave(&ioc->lock, flags);
1774
1775         BUG_ON(ioc->ioc_data == cic);
1776
1777         radix_tree_delete(&ioc->radix_root, (unsigned long) cfqd);
1778         hlist_del_rcu(&cic->cic_list);
1779         spin_unlock_irqrestore(&ioc->lock, flags);
1780
1781         cfq_cic_free(cic);
1782 }
1783
1784 static struct cfq_io_context *
1785 cfq_cic_lookup(struct cfq_data *cfqd, struct io_context *ioc)
1786 {
1787         struct cfq_io_context *cic;
1788         unsigned long flags;
1789         void *k;
1790
1791         if (unlikely(!ioc))
1792                 return NULL;
1793
1794         rcu_read_lock();
1795
1796         /*
1797          * we maintain a last-hit cache, to avoid browsing over the tree
1798          */
1799         cic = rcu_dereference(ioc->ioc_data);
1800         if (cic && cic->key == cfqd) {
1801                 rcu_read_unlock();
1802                 return cic;
1803         }
1804
1805         do {
1806                 cic = radix_tree_lookup(&ioc->radix_root, (unsigned long) cfqd);
1807                 rcu_read_unlock();
1808                 if (!cic)
1809                         break;
1810                 /* ->key must be copied to avoid race with cfq_exit_queue() */
1811                 k = cic->key;
1812                 if (unlikely(!k)) {
1813                         cfq_drop_dead_cic(cfqd, ioc, cic);
1814                         rcu_read_lock();
1815                         continue;
1816                 }
1817
1818                 spin_lock_irqsave(&ioc->lock, flags);
1819                 rcu_assign_pointer(ioc->ioc_data, cic);
1820                 spin_unlock_irqrestore(&ioc->lock, flags);
1821                 break;
1822         } while (1);
1823
1824         return cic;
1825 }
1826
1827 /*
1828  * Add cic into ioc, using cfqd as the search key. This enables us to lookup
1829  * the process specific cfq io context when entered from the block layer.
1830  * Also adds the cic to a per-cfqd list, used when this queue is removed.
1831  */
1832 static int cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc,
1833                         struct cfq_io_context *cic, gfp_t gfp_mask)
1834 {
1835         unsigned long flags;
1836         int ret;
1837
1838         ret = radix_tree_preload(gfp_mask);
1839         if (!ret) {
1840                 cic->ioc = ioc;
1841                 cic->key = cfqd;
1842
1843                 spin_lock_irqsave(&ioc->lock, flags);
1844                 ret = radix_tree_insert(&ioc->radix_root,
1845                                                 (unsigned long) cfqd, cic);
1846                 if (!ret)
1847                         hlist_add_head_rcu(&cic->cic_list, &ioc->cic_list);
1848                 spin_unlock_irqrestore(&ioc->lock, flags);
1849
1850                 radix_tree_preload_end();
1851
1852                 if (!ret) {
1853                         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
1854                         list_add(&cic->queue_list, &cfqd->cic_list);
1855                         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
1856                 }
1857         }
1858
1859         if (ret)
1860                 printk(KERN_ERR "cfq: cic link failed!\n");
1861
1862         return ret;
1863 }
1864
1865 /*
1866  * Setup general io context and cfq io context. There can be several cfq
1867  * io contexts per general io context, if this process is doing io to more
1868  * than one device managed by cfq.
1869  */
1870 static struct cfq_io_context *
1871 cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
1872 {
1873         struct io_context *ioc = NULL;
1874         struct cfq_io_context *cic;
1875
1876         might_sleep_if(gfp_mask & __GFP_WAIT);
1877
1878         ioc = get_io_context(gfp_mask, cfqd->queue->node);
1879         if (!ioc)
1880                 return NULL;
1881
1882         cic = cfq_cic_lookup(cfqd, ioc);
1883         if (cic)
1884                 goto out;
1885
1886         cic = cfq_alloc_io_context(cfqd, gfp_mask);
1887         if (cic == NULL)
1888                 goto err;
1889
1890         if (cfq_cic_link(cfqd, ioc, cic, gfp_mask))
1891                 goto err_free;
1892
1893 out:
1894         smp_read_barrier_depends();
1895         if (unlikely(ioc->ioprio_changed))
1896                 cfq_ioc_set_ioprio(ioc);
1897
1898         return cic;
1899 err_free:
1900         cfq_cic_free(cic);
1901 err:
1902         put_io_context(ioc);
1903         return NULL;
1904 }
1905
1906 static void
1907 cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_io_context *cic)
1908 {
1909         unsigned long elapsed = jiffies - cic->last_end_request;
1910         unsigned long ttime = min(elapsed, 2UL * cfqd->cfq_slice_idle);
1911
1912         cic->ttime_samples = (7*cic->ttime_samples + 256) / 8;
1913         cic->ttime_total = (7*cic->ttime_total + 256*ttime) / 8;
1914         cic->ttime_mean = (cic->ttime_total + 128) / cic->ttime_samples;
1915 }
1916
1917 static void
1918 cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_io_context *cic,
1919                        struct request *rq)
1920 {
1921         sector_t sdist;
1922         u64 total;
1923
1924         if (!cic->last_request_pos)
1925                 sdist = 0;
1926         else if (cic->last_request_pos < blk_rq_pos(rq))
1927                 sdist = blk_rq_pos(rq) - cic->last_request_pos;
1928         else
1929                 sdist = cic->last_request_pos - blk_rq_pos(rq);
1930
1931         /*
1932          * Don't allow the seek distance to get too large from the
1933          * odd fragment, pagein, etc
1934          */
1935         if (cic->seek_samples <= 60) /* second&third seek */
1936                 sdist = min(sdist, (cic->seek_mean * 4) + 2*1024*1024);
1937         else
1938                 sdist = min(sdist, (cic->seek_mean * 4) + 2*1024*64);
1939
1940         cic->seek_samples = (7*cic->seek_samples + 256) / 8;
1941         cic->seek_total = (7*cic->seek_total + (u64)256*sdist) / 8;
1942         total = cic->seek_total + (cic->seek_samples/2);
1943         do_div(total, cic->seek_samples);
1944         cic->seek_mean = (sector_t)total;
1945 }
1946
1947 /*
1948  * Disable idle window if the process thinks too long or seeks so much that
1949  * it doesn't matter
1950  */
1951 static void
1952 cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
1953                        struct cfq_io_context *cic)
1954 {
1955         int old_idle, enable_idle;
1956
1957         /*
1958          * Don't idle for async or idle io prio class
1959          */
1960         if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
1961                 return;
1962
1963         enable_idle = old_idle = cfq_cfqq_idle_window(cfqq);
1964
1965         if (!atomic_read(&cic->ioc->nr_tasks) || !cfqd->cfq_slice_idle ||
1966             (cfqd->hw_tag && CIC_SEEKY(cic)))
1967                 enable_idle = 0;
1968         else if (sample_valid(cic->ttime_samples)) {
1969                 if (cic->ttime_mean > cfqd->cfq_slice_idle)
1970                         enable_idle = 0;
1971                 else
1972                         enable_idle = 1;
1973         }
1974
1975         if (old_idle != enable_idle) {
1976                 cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle);
1977                 if (enable_idle)
1978                         cfq_mark_cfqq_idle_window(cfqq);
1979                 else
1980                         cfq_clear_cfqq_idle_window(cfqq);
1981         }
1982 }
1983
1984 /*
1985  * Check if new_cfqq should preempt the currently active queue. Return 0 for
1986  * no or if we aren't sure, a 1 will cause a preempt.
1987  */
1988 static int
1989 cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
1990                    struct request *rq)
1991 {
1992         struct cfq_queue *cfqq;
1993
1994         cfqq = cfqd->active_queue;
1995         if (!cfqq)
1996                 return 0;
1997
1998         if (cfq_slice_used(cfqq))
1999                 return 1;
2000
2001         if (cfq_class_idle(new_cfqq))
2002                 return 0;
2003
2004         if (cfq_class_idle(cfqq))
2005                 return 1;
2006
2007         /*
2008          * if the new request is sync, but the currently running queue is
2009          * not, let the sync request have priority.
2010          */
2011         if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
2012                 return 1;
2013
2014         /*
2015          * So both queues are sync. Let the new request get disk time if
2016          * it's a metadata request and the current queue is doing regular IO.
2017          */
2018         if (rq_is_meta(rq) && !cfqq->meta_pending)
2019                 return 1;
2020
2021         /*
2022          * Allow an RT request to pre-empt an ongoing non-RT cfqq timeslice.
2023          */
2024         if (cfq_class_rt(new_cfqq) && !cfq_class_rt(cfqq))
2025                 return 1;
2026
2027         if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq))
2028                 return 0;
2029
2030         /*
2031          * if this request is as-good as one we would expect from the
2032          * current cfqq, let it preempt
2033          */
2034         if (cfq_rq_close(cfqd, rq))
2035                 return 1;
2036
2037         return 0;
2038 }
2039
2040 /*
2041  * cfqq preempts the active queue. if we allowed preempt with no slice left,
2042  * let it have half of its nominal slice.
2043  */
2044 static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
2045 {
2046         cfq_log_cfqq(cfqd, cfqq, "preempt");
2047         cfq_slice_expired(cfqd, 1);
2048
2049         /*
2050          * Put the new queue at the front of the of the current list,
2051          * so we know that it will be selected next.
2052          */
2053         BUG_ON(!cfq_cfqq_on_rr(cfqq));
2054
2055         cfq_service_tree_add(cfqd, cfqq, 1);
2056
2057         cfqq->slice_end = 0;
2058         cfq_mark_cfqq_slice_new(cfqq);
2059 }
2060
2061 /*
2062  * Called when a new fs request (rq) is added (to cfqq). Check if there's
2063  * something we should do about it
2064  */
2065 static void
2066 cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
2067                 struct request *rq)
2068 {
2069         struct cfq_io_context *cic = RQ_CIC(rq);
2070
2071         cfqd->rq_queued++;
2072         if (rq_is_meta(rq))
2073                 cfqq->meta_pending++;
2074
2075         cfq_update_io_thinktime(cfqd, cic);
2076         cfq_update_io_seektime(cfqd, cic, rq);
2077         cfq_update_idle_window(cfqd, cfqq, cic);
2078
2079         cic->last_request_pos = blk_rq_pos(rq) + blk_rq_sectors(rq);
2080
2081         if (cfqq == cfqd->active_queue) {
2082                 /*
2083                  * Remember that we saw a request from this process, but
2084                  * don't start queuing just yet. Otherwise we risk seeing lots
2085                  * of tiny requests, because we disrupt the normal plugging
2086                  * and merging. If the request is already larger than a single
2087                  * page, let it rip immediately. For that case we assume that
2088                  * merging is already done. Ditto for a busy system that
2089                  * has other work pending, don't risk delaying until the
2090                  * idle timer unplug to continue working.
2091                  */
2092                 if (cfq_cfqq_wait_request(cfqq)) {
2093                         if (blk_rq_bytes(rq) > PAGE_CACHE_SIZE ||
2094                             cfqd->busy_queues > 1) {
2095                                 del_timer(&cfqd->idle_slice_timer);
2096                         __blk_run_queue(cfqd->queue);
2097                         }
2098                         cfq_mark_cfqq_must_dispatch(cfqq);
2099                 }
2100         } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
2101                 /*
2102                  * not the active queue - expire current slice if it is
2103                  * idle and has expired it's mean thinktime or this new queue
2104                  * has some old slice time left and is of higher priority or
2105                  * this new queue is RT and the current one is BE
2106                  */
2107                 cfq_preempt_queue(cfqd, cfqq);
2108                 __blk_run_queue(cfqd->queue);
2109         }
2110 }
2111
2112 static void cfq_insert_request(struct request_queue *q, struct request *rq)
2113 {
2114         struct cfq_data *cfqd = q->elevator->elevator_data;
2115         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2116
2117         cfq_log_cfqq(cfqd, cfqq, "insert_request");
2118         cfq_init_prio_data(cfqq, RQ_CIC(rq)->ioc);
2119
2120         cfq_add_rq_rb(rq);
2121
2122         list_add_tail(&rq->queuelist, &cfqq->fifo);
2123
2124         cfq_rq_enqueued(cfqd, cfqq, rq);
2125 }
2126
2127 /*
2128  * Update hw_tag based on peak queue depth over 50 samples under
2129  * sufficient load.
2130  */
2131 static void cfq_update_hw_tag(struct cfq_data *cfqd)
2132 {
2133         if (cfqd->rq_in_driver > cfqd->rq_in_driver_peak)
2134                 cfqd->rq_in_driver_peak = cfqd->rq_in_driver;
2135
2136         if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN &&
2137             cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN)
2138                 return;
2139
2140         if (cfqd->hw_tag_samples++ < 50)
2141                 return;
2142
2143         if (cfqd->rq_in_driver_peak >= CFQ_HW_QUEUE_MIN)
2144                 cfqd->hw_tag = 1;
2145         else
2146                 cfqd->hw_tag = 0;
2147
2148         cfqd->hw_tag_samples = 0;
2149         cfqd->rq_in_driver_peak = 0;
2150 }
2151
2152 static void cfq_completed_request(struct request_queue *q, struct request *rq)
2153 {
2154         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2155         struct cfq_data *cfqd = cfqq->cfqd;
2156         const int sync = rq_is_sync(rq);
2157         unsigned long now;
2158
2159         now = jiffies;
2160         cfq_log_cfqq(cfqd, cfqq, "complete");
2161
2162         cfq_update_hw_tag(cfqd);
2163
2164         WARN_ON(!cfqd->rq_in_driver);
2165         WARN_ON(!cfqq->dispatched);
2166         cfqd->rq_in_driver--;
2167         cfqq->dispatched--;
2168
2169         if (cfq_cfqq_sync(cfqq))
2170                 cfqd->sync_flight--;
2171
2172         if (sync)
2173                 RQ_CIC(rq)->last_end_request = now;
2174
2175         /*
2176          * If this is the active queue, check if it needs to be expired,
2177          * or if we want to idle in case it has no pending requests.
2178          */
2179         if (cfqd->active_queue == cfqq) {
2180                 const bool cfqq_empty = RB_EMPTY_ROOT(&cfqq->sort_list);
2181
2182                 if (cfq_cfqq_slice_new(cfqq)) {
2183                         cfq_set_prio_slice(cfqd, cfqq);
2184                         cfq_clear_cfqq_slice_new(cfqq);
2185                 }
2186                 /*
2187                  * If there are no requests waiting in this queue, and
2188                  * there are other queues ready to issue requests, AND
2189                  * those other queues are issuing requests within our
2190                  * mean seek distance, give them a chance to run instead
2191                  * of idling.
2192                  */
2193                 if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
2194                         cfq_slice_expired(cfqd, 1);
2195                 else if (cfqq_empty && !cfq_close_cooperator(cfqd, cfqq, 1) &&
2196                          sync && !rq_noidle(rq))
2197                         cfq_arm_slice_timer(cfqd);
2198         }
2199
2200         if (!cfqd->rq_in_driver)
2201                 cfq_schedule_dispatch(cfqd);
2202 }
2203
2204 /*
2205  * we temporarily boost lower priority queues if they are holding fs exclusive
2206  * resources. they are boosted to normal prio (CLASS_BE/4)
2207  */
2208 static void cfq_prio_boost(struct cfq_queue *cfqq)
2209 {
2210         if (has_fs_excl()) {
2211                 /*
2212                  * boost idle prio on transactions that would lock out other
2213                  * users of the filesystem
2214                  */
2215                 if (cfq_class_idle(cfqq))
2216                         cfqq->ioprio_class = IOPRIO_CLASS_BE;
2217                 if (cfqq->ioprio > IOPRIO_NORM)
2218                         cfqq->ioprio = IOPRIO_NORM;
2219         } else {
2220                 /*
2221                  * check if we need to unboost the queue
2222                  */
2223                 if (cfqq->ioprio_class != cfqq->org_ioprio_class)
2224                         cfqq->ioprio_class = cfqq->org_ioprio_class;
2225                 if (cfqq->ioprio != cfqq->org_ioprio)
2226                         cfqq->ioprio = cfqq->org_ioprio;
2227         }
2228 }
2229
2230 static inline int __cfq_may_queue(struct cfq_queue *cfqq)
2231 {
2232         if ((cfq_cfqq_wait_request(cfqq) || cfq_cfqq_must_alloc(cfqq)) &&
2233             !cfq_cfqq_must_alloc_slice(cfqq)) {
2234                 cfq_mark_cfqq_must_alloc_slice(cfqq);
2235                 return ELV_MQUEUE_MUST;
2236         }
2237
2238         return ELV_MQUEUE_MAY;
2239 }
2240
2241 static int cfq_may_queue(struct request_queue *q, int rw)
2242 {
2243         struct cfq_data *cfqd = q->elevator->elevator_data;
2244         struct task_struct *tsk = current;
2245         struct cfq_io_context *cic;
2246         struct cfq_queue *cfqq;
2247
2248         /*
2249          * don't force setup of a queue from here, as a call to may_queue
2250          * does not necessarily imply that a request actually will be queued.
2251          * so just lookup a possibly existing queue, or return 'may queue'
2252          * if that fails
2253          */
2254         cic = cfq_cic_lookup(cfqd, tsk->io_context);
2255         if (!cic)
2256                 return ELV_MQUEUE_MAY;
2257
2258         cfqq = cic_to_cfqq(cic, rw_is_sync(rw));
2259         if (cfqq) {
2260                 cfq_init_prio_data(cfqq, cic->ioc);
2261                 cfq_prio_boost(cfqq);
2262
2263                 return __cfq_may_queue(cfqq);
2264         }
2265
2266         return ELV_MQUEUE_MAY;
2267 }
2268
2269 /*
2270  * queue lock held here
2271  */
2272 static void cfq_put_request(struct request *rq)
2273 {
2274         struct cfq_queue *cfqq = RQ_CFQQ(rq);
2275
2276         if (cfqq) {
2277                 const int rw = rq_data_dir(rq);
2278
2279                 BUG_ON(!cfqq->allocated[rw]);
2280                 cfqq->allocated[rw]--;
2281
2282                 put_io_context(RQ_CIC(rq)->ioc);
2283
2284                 rq->elevator_private = NULL;
2285                 rq->elevator_private2 = NULL;
2286
2287                 cfq_put_queue(cfqq);
2288         }
2289 }
2290
2291 /*
2292  * Allocate cfq data structures associated with this request.
2293  */
2294 static int
2295 cfq_set_request(struct request_queue *q, struct request *rq, gfp_t gfp_mask)
2296 {
2297         struct cfq_data *cfqd = q->elevator->elevator_data;
2298         struct cfq_io_context *cic;
2299         const int rw = rq_data_dir(rq);
2300         const int is_sync = rq_is_sync(rq);
2301         struct cfq_queue *cfqq;
2302         unsigned long flags;
2303
2304         might_sleep_if(gfp_mask & __GFP_WAIT);
2305
2306         cic = cfq_get_io_context(cfqd, gfp_mask);
2307
2308         spin_lock_irqsave(q->queue_lock, flags);
2309
2310         if (!cic)
2311                 goto queue_fail;
2312
2313         cfqq = cic_to_cfqq(cic, is_sync);
2314         if (!cfqq) {
2315                 cfqq = cfq_get_queue(cfqd, is_sync, cic->ioc, gfp_mask);
2316
2317                 if (!cfqq)
2318                         goto queue_fail;
2319
2320                 cic_set_cfqq(cic, cfqq, is_sync);
2321         }
2322
2323         cfqq->allocated[rw]++;
2324         cfq_clear_cfqq_must_alloc(cfqq);
2325         atomic_inc(&cfqq->ref);
2326
2327         spin_unlock_irqrestore(q->queue_lock, flags);
2328
2329         rq->elevator_private = cic;
2330         rq->elevator_private2 = cfqq;
2331         return 0;
2332
2333 queue_fail:
2334         if (cic)
2335                 put_io_context(cic->ioc);
2336
2337         cfq_schedule_dispatch(cfqd);
2338         spin_unlock_irqrestore(q->queue_lock, flags);
2339         cfq_log(cfqd, "set_request fail");
2340         return 1;
2341 }
2342
2343 static void cfq_kick_queue(struct work_struct *work)
2344 {
2345         struct cfq_data *cfqd =
2346                 container_of(work, struct cfq_data, unplug_work);
2347         struct request_queue *q = cfqd->queue;
2348
2349         spin_lock_irq(q->queue_lock);
2350         __blk_run_queue(cfqd->queue);
2351         spin_unlock_irq(q->queue_lock);
2352 }
2353
2354 /*
2355  * Timer running if the active_queue is currently idling inside its time slice
2356  */
2357 static void cfq_idle_slice_timer(unsigned long data)
2358 {
2359         struct cfq_data *cfqd = (struct cfq_data *) data;
2360         struct cfq_queue *cfqq;
2361         unsigned long flags;
2362         int timed_out = 1;
2363
2364         cfq_log(cfqd, "idle timer fired");
2365
2366         spin_lock_irqsave(cfqd->queue->queue_lock, flags);
2367
2368         cfqq = cfqd->active_queue;
2369         if (cfqq) {
2370                 timed_out = 0;
2371
2372                 /*
2373                  * We saw a request before the queue expired, let it through
2374                  */
2375                 if (cfq_cfqq_must_dispatch(cfqq))
2376                         goto out_kick;
2377
2378                 /*
2379                  * expired
2380                  */
2381                 if (cfq_slice_used(cfqq))
2382                         goto expire;
2383
2384                 /*
2385                  * only expire and reinvoke request handler, if there are
2386                  * other queues with pending requests
2387                  */
2388                 if (!cfqd->busy_queues)
2389                         goto out_cont;
2390
2391                 /*
2392                  * not expired and it has a request pending, let it dispatch
2393                  */
2394                 if (!RB_EMPTY_ROOT(&cfqq->sort_list))
2395                         goto out_kick;
2396         }
2397 expire:
2398         cfq_slice_expired(cfqd, timed_out);
2399 out_kick:
2400         cfq_schedule_dispatch(cfqd);
2401 out_cont:
2402         spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
2403 }
2404
2405 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
2406 {
2407         del_timer_sync(&cfqd->idle_slice_timer);
2408         cancel_work_sync(&cfqd->unplug_work);
2409 }
2410
2411 static void cfq_put_async_queues(struct cfq_data *cfqd)
2412 {
2413         int i;
2414
2415         for (i = 0; i < IOPRIO_BE_NR; i++) {
2416                 if (cfqd->async_cfqq[0][i])
2417                         cfq_put_queue(cfqd->async_cfqq[0][i]);
2418                 if (cfqd->async_cfqq[1][i])
2419                         cfq_put_queue(cfqd->async_cfqq[1][i]);
2420         }
2421
2422         if (cfqd->async_idle_cfqq)
2423                 cfq_put_queue(cfqd->async_idle_cfqq);
2424 }
2425
2426 static void cfq_exit_queue(struct elevator_queue *e)
2427 {
2428         struct cfq_data *cfqd = e->elevator_data;
2429         struct request_queue *q = cfqd->queue;
2430
2431         cfq_shutdown_timer_wq(cfqd);
2432
2433         spin_lock_irq(q->queue_lock);
2434
2435         if (cfqd->active_queue)
2436                 __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
2437
2438         while (!list_empty(&cfqd->cic_list)) {
2439                 struct cfq_io_context *cic = list_entry(cfqd->cic_list.next,
2440                                                         struct cfq_io_context,
2441                                                         queue_list);
2442
2443                 __cfq_exit_single_io_context(cfqd, cic);
2444         }
2445
2446         cfq_put_async_queues(cfqd);
2447
2448         spin_unlock_irq(q->queue_lock);
2449
2450         cfq_shutdown_timer_wq(cfqd);
2451
2452         kfree(cfqd);
2453 }
2454
2455 static void *cfq_init_queue(struct request_queue *q)
2456 {
2457         struct cfq_data *cfqd;
2458         int i;
2459
2460         cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node);
2461         if (!cfqd)
2462                 return NULL;
2463
2464         cfqd->service_tree = CFQ_RB_ROOT;
2465
2466         /*
2467          * Not strictly needed (since RB_ROOT just clears the node and we
2468          * zeroed cfqd on alloc), but better be safe in case someone decides
2469          * to add magic to the rb code
2470          */
2471         for (i = 0; i < CFQ_PRIO_LISTS; i++)
2472                 cfqd->prio_trees[i] = RB_ROOT;
2473
2474         /*
2475          * Our fallback cfqq if cfq_find_alloc_queue() runs into OOM issues.
2476          * Grab a permanent reference to it, so that the normal code flow
2477          * will not attempt to free it.
2478          */
2479         cfq_init_cfqq(cfqd, &cfqd->oom_cfqq, 1, 0);
2480         atomic_inc(&cfqd->oom_cfqq.ref);
2481
2482         INIT_LIST_HEAD(&cfqd->cic_list);
2483
2484         cfqd->queue = q;
2485
2486         init_timer(&cfqd->idle_slice_timer);
2487         cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
2488         cfqd->idle_slice_timer.data = (unsigned long) cfqd;
2489
2490         INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
2491
2492         cfqd->cfq_quantum = cfq_quantum;
2493         cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
2494         cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
2495         cfqd->cfq_back_max = cfq_back_max;
2496         cfqd->cfq_back_penalty = cfq_back_penalty;
2497         cfqd->cfq_slice[0] = cfq_slice_async;
2498         cfqd->cfq_slice[1] = cfq_slice_sync;
2499         cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
2500         cfqd->cfq_slice_idle = cfq_slice_idle;
2501         cfqd->hw_tag = 1;
2502
2503         return cfqd;
2504 }
2505
2506 static void cfq_slab_kill(void)
2507 {
2508         /*
2509          * Caller already ensured that pending RCU callbacks are completed,
2510          * so we should have no busy allocations at this point.
2511          */
2512         if (cfq_pool)
2513                 kmem_cache_destroy(cfq_pool);
2514         if (cfq_ioc_pool)
2515                 kmem_cache_destroy(cfq_ioc_pool);
2516 }
2517
2518 static int __init cfq_slab_setup(void)
2519 {
2520         cfq_pool = KMEM_CACHE(cfq_queue, 0);
2521         if (!cfq_pool)
2522                 goto fail;
2523
2524         cfq_ioc_pool = KMEM_CACHE(cfq_io_context, 0);
2525         if (!cfq_ioc_pool)
2526                 goto fail;
2527
2528         return 0;
2529 fail:
2530         cfq_slab_kill();
2531         return -ENOMEM;
2532 }
2533
2534 /*
2535  * sysfs parts below -->
2536  */
2537 static ssize_t
2538 cfq_var_show(unsigned int var, char *page)
2539 {
2540         return sprintf(page, "%d\n", var);
2541 }
2542
2543 static ssize_t
2544 cfq_var_store(unsigned int *var, const char *page, size_t count)
2545 {
2546         char *p = (char *) page;
2547
2548         *var = simple_strtoul(p, &p, 10);
2549         return count;
2550 }
2551
2552 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV)                            \
2553 static ssize_t __FUNC(struct elevator_queue *e, char *page)             \
2554 {                                                                       \
2555         struct cfq_data *cfqd = e->elevator_data;                       \
2556         unsigned int __data = __VAR;                                    \
2557         if (__CONV)                                                     \
2558                 __data = jiffies_to_msecs(__data);                      \
2559         return cfq_var_show(__data, (page));                            \
2560 }
2561 SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
2562 SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
2563 SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
2564 SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
2565 SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
2566 SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
2567 SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
2568 SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
2569 SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
2570 #undef SHOW_FUNCTION
2571
2572 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)                 \
2573 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count) \
2574 {                                                                       \
2575         struct cfq_data *cfqd = e->elevator_data;                       \
2576         unsigned int __data;                                            \
2577         int ret = cfq_var_store(&__data, (page), count);                \
2578         if (__data < (MIN))                                             \
2579                 __data = (MIN);                                         \
2580         else if (__data > (MAX))                                        \
2581                 __data = (MAX);                                         \
2582         if (__CONV)                                                     \
2583                 *(__PTR) = msecs_to_jiffies(__data);                    \
2584         else                                                            \
2585                 *(__PTR) = __data;                                      \
2586         return ret;                                                     \
2587 }
2588 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
2589 STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1,
2590                 UINT_MAX, 1);
2591 STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1,
2592                 UINT_MAX, 1);
2593 STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
2594 STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1,
2595                 UINT_MAX, 0);
2596 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
2597 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
2598 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
2599 STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
2600                 UINT_MAX, 0);
2601 #undef STORE_FUNCTION
2602
2603 #define CFQ_ATTR(name) \
2604         __ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
2605
2606 static struct elv_fs_entry cfq_attrs[] = {
2607         CFQ_ATTR(quantum),
2608         CFQ_ATTR(fifo_expire_sync),
2609         CFQ_ATTR(fifo_expire_async),
2610         CFQ_ATTR(back_seek_max),
2611         CFQ_ATTR(back_seek_penalty),
2612         CFQ_ATTR(slice_sync),
2613         CFQ_ATTR(slice_async),
2614         CFQ_ATTR(slice_async_rq),
2615         CFQ_ATTR(slice_idle),
2616         __ATTR_NULL
2617 };
2618
2619 static struct elevator_type iosched_cfq = {
2620         .ops = {
2621                 .elevator_merge_fn =            cfq_merge,
2622                 .elevator_merged_fn =           cfq_merged_request,
2623                 .elevator_merge_req_fn =        cfq_merged_requests,
2624                 .elevator_allow_merge_fn =      cfq_allow_merge,
2625                 .elevator_dispatch_fn =         cfq_dispatch_requests,
2626                 .elevator_add_req_fn =          cfq_insert_request,
2627                 .elevator_activate_req_fn =     cfq_activate_request,
2628                 .elevator_deactivate_req_fn =   cfq_deactivate_request,
2629                 .elevator_queue_empty_fn =      cfq_queue_empty,
2630                 .elevator_completed_req_fn =    cfq_completed_request,
2631                 .elevator_former_req_fn =       elv_rb_former_request,
2632                 .elevator_latter_req_fn =       elv_rb_latter_request,
2633                 .elevator_set_req_fn =          cfq_set_request,
2634                 .elevator_put_req_fn =          cfq_put_request,
2635                 .elevator_may_queue_fn =        cfq_may_queue,
2636                 .elevator_init_fn =             cfq_init_queue,
2637                 .elevator_exit_fn =             cfq_exit_queue,
2638                 .trim =                         cfq_free_io_context,
2639         },
2640         .elevator_attrs =       cfq_attrs,
2641         .elevator_name =        "cfq",
2642         .elevator_owner =       THIS_MODULE,
2643 };
2644
2645 static int __init cfq_init(void)
2646 {
2647         /*
2648          * could be 0 on HZ < 1000 setups
2649          */
2650         if (!cfq_slice_async)
2651                 cfq_slice_async = 1;
2652         if (!cfq_slice_idle)
2653                 cfq_slice_idle = 1;
2654
2655         if (cfq_slab_setup())
2656                 return -ENOMEM;
2657
2658         elv_register(&iosched_cfq);
2659
2660         return 0;
2661 }
2662
2663 static void __exit cfq_exit(void)
2664 {
2665         DECLARE_COMPLETION_ONSTACK(all_gone);
2666         elv_unregister(&iosched_cfq);
2667         ioc_gone = &all_gone;
2668         /* ioc_gone's update must be visible before reading ioc_count */
2669         smp_wmb();
2670
2671         /*
2672          * this also protects us from entering cfq_slab_kill() with
2673          * pending RCU callbacks
2674          */
2675         if (elv_ioc_count_read(ioc_count))
2676                 wait_for_completion(&all_gone);
2677         cfq_slab_kill();
2678 }
2679
2680 module_init(cfq_init);
2681 module_exit(cfq_exit);
2682
2683 MODULE_AUTHOR("Jens Axboe");
2684 MODULE_LICENSE("GPL");
2685 MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");