1 // SPDX-License-Identifier: GPL-2.0-or-later
3 * Budget Fair Queueing (BFQ) I/O scheduler.
5 * Based on ideas and code from CFQ:
6 * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
8 * Copyright (C) 2008 Fabio Checconi <fabio@gandalf.sssup.it>
9 * Paolo Valente <paolo.valente@unimore.it>
11 * Copyright (C) 2010 Paolo Valente <paolo.valente@unimore.it>
12 * Arianna Avanzini <avanzini@google.com>
14 * Copyright (C) 2017 Paolo Valente <paolo.valente@linaro.org>
16 * BFQ is a proportional-share I/O scheduler, with some extra
17 * low-latency capabilities. BFQ also supports full hierarchical
18 * scheduling through cgroups. Next paragraphs provide an introduction
19 * on BFQ inner workings. Details on BFQ benefits, usage and
20 * limitations can be found in Documentation/block/bfq-iosched.rst.
22 * BFQ is a proportional-share storage-I/O scheduling algorithm based
23 * on the slice-by-slice service scheme of CFQ. But BFQ assigns
24 * budgets, measured in number of sectors, to processes instead of
25 * time slices. The device is not granted to the in-service process
26 * for a given time slice, but until it has exhausted its assigned
27 * budget. This change from the time to the service domain enables BFQ
28 * to distribute the device throughput among processes as desired,
29 * without any distortion due to throughput fluctuations, or to device
30 * internal queueing. BFQ uses an ad hoc internal scheduler, called
31 * B-WF2Q+, to schedule processes according to their budgets. More
32 * precisely, BFQ schedules queues associated with processes. Each
33 * process/queue is assigned a user-configurable weight, and B-WF2Q+
34 * guarantees that each queue receives a fraction of the throughput
35 * proportional to its weight. Thanks to the accurate policy of
36 * B-WF2Q+, BFQ can afford to assign high budgets to I/O-bound
37 * processes issuing sequential requests (to boost the throughput),
38 * and yet guarantee a low latency to interactive and soft real-time
41 * In particular, to provide these low-latency guarantees, BFQ
42 * explicitly privileges the I/O of two classes of time-sensitive
43 * applications: interactive and soft real-time. In more detail, BFQ
44 * behaves this way if the low_latency parameter is set (default
45 * configuration). This feature enables BFQ to provide applications in
46 * these classes with a very low latency.
48 * To implement this feature, BFQ constantly tries to detect whether
49 * the I/O requests in a bfq_queue come from an interactive or a soft
50 * real-time application. For brevity, in these cases, the queue is
51 * said to be interactive or soft real-time. In both cases, BFQ
52 * privileges the service of the queue, over that of non-interactive
53 * and non-soft-real-time queues. This privileging is performed,
54 * mainly, by raising the weight of the queue. So, for brevity, we
55 * call just weight-raising periods the time periods during which a
56 * queue is privileged, because deemed interactive or soft real-time.
58 * The detection of soft real-time queues/applications is described in
59 * detail in the comments on the function
60 * bfq_bfqq_softrt_next_start. On the other hand, the detection of an
61 * interactive queue works as follows: a queue is deemed interactive
62 * if it is constantly non empty only for a limited time interval,
63 * after which it does become empty. The queue may be deemed
64 * interactive again (for a limited time), if it restarts being
65 * constantly non empty, provided that this happens only after the
66 * queue has remained empty for a given minimum idle time.
68 * By default, BFQ computes automatically the above maximum time
69 * interval, i.e., the time interval after which a constantly
70 * non-empty queue stops being deemed interactive. Since a queue is
71 * weight-raised while it is deemed interactive, this maximum time
72 * interval happens to coincide with the (maximum) duration of the
73 * weight-raising for interactive queues.
75 * Finally, BFQ also features additional heuristics for
76 * preserving both a low latency and a high throughput on NCQ-capable,
77 * rotational or flash-based devices, and to get the job done quickly
78 * for applications consisting in many I/O-bound processes.
80 * NOTE: if the main or only goal, with a given device, is to achieve
81 * the maximum-possible throughput at all times, then do switch off
82 * all low-latency heuristics for that device, by setting low_latency
85 * BFQ is described in [1], where also a reference to the initial,
86 * more theoretical paper on BFQ can be found. The interested reader
87 * can find in the latter paper full details on the main algorithm, as
88 * well as formulas of the guarantees and formal proofs of all the
89 * properties. With respect to the version of BFQ presented in these
90 * papers, this implementation adds a few more heuristics, such as the
91 * ones that guarantee a low latency to interactive and soft real-time
92 * applications, and a hierarchical extension based on H-WF2Q+.
94 * B-WF2Q+ is based on WF2Q+, which is described in [2], together with
95 * H-WF2Q+, while the augmented tree used here to implement B-WF2Q+
96 * with O(log N) complexity derives from the one introduced with EEVDF
99 * [1] P. Valente, A. Avanzini, "Evolution of the BFQ Storage I/O
100 * Scheduler", Proceedings of the First Workshop on Mobile System
101 * Technologies (MST-2015), May 2015.
102 * http://algogroup.unimore.it/people/paolo/disk_sched/mst-2015.pdf
104 * [2] Jon C.R. Bennett and H. Zhang, "Hierarchical Packet Fair Queueing
105 * Algorithms", IEEE/ACM Transactions on Networking, 5(5):675-689,
108 * http://www.cs.cmu.edu/~hzhang/papers/TON-97-Oct.ps.gz
110 * [3] I. Stoica and H. Abdel-Wahab, "Earliest Eligible Virtual Deadline
111 * First: A Flexible and Accurate Mechanism for Proportional Share
112 * Resource Allocation", technical report.
114 * http://www.cs.berkeley.edu/~istoica/papers/eevdf-tr-95.pdf
116 #include <linux/module.h>
117 #include <linux/slab.h>
118 #include <linux/blkdev.h>
119 #include <linux/cgroup.h>
120 #include <linux/elevator.h>
121 #include <linux/ktime.h>
122 #include <linux/rbtree.h>
123 #include <linux/ioprio.h>
124 #include <linux/sbitmap.h>
125 #include <linux/delay.h>
129 #include "blk-mq-tag.h"
130 #include "blk-mq-sched.h"
131 #include "bfq-iosched.h"
134 #define BFQ_BFQQ_FNS(name) \
135 void bfq_mark_bfqq_##name(struct bfq_queue *bfqq) \
137 __set_bit(BFQQF_##name, &(bfqq)->flags); \
139 void bfq_clear_bfqq_##name(struct bfq_queue *bfqq) \
141 __clear_bit(BFQQF_##name, &(bfqq)->flags); \
143 int bfq_bfqq_##name(const struct bfq_queue *bfqq) \
145 return test_bit(BFQQF_##name, &(bfqq)->flags); \
148 BFQ_BFQQ_FNS(just_created
);
150 BFQ_BFQQ_FNS(wait_request
);
151 BFQ_BFQQ_FNS(non_blocking_wait_rq
);
152 BFQ_BFQQ_FNS(fifo_expire
);
153 BFQ_BFQQ_FNS(has_short_ttime
);
155 BFQ_BFQQ_FNS(IO_bound
);
156 BFQ_BFQQ_FNS(in_large_burst
);
158 BFQ_BFQQ_FNS(split_coop
);
159 BFQ_BFQQ_FNS(softrt_update
);
160 BFQ_BFQQ_FNS(has_waker
);
161 #undef BFQ_BFQQ_FNS \
163 /* Expiration time of sync (0) and async (1) requests, in ns. */
164 static const u64 bfq_fifo_expire
[2] = { NSEC_PER_SEC
/ 4, NSEC_PER_SEC
/ 8 };
166 /* Maximum backwards seek (magic number lifted from CFQ), in KiB. */
167 static const int bfq_back_max
= 16 * 1024;
169 /* Penalty of a backwards seek, in number of sectors. */
170 static const int bfq_back_penalty
= 2;
172 /* Idling period duration, in ns. */
173 static u64 bfq_slice_idle
= NSEC_PER_SEC
/ 125;
175 /* Minimum number of assigned budgets for which stats are safe to compute. */
176 static const int bfq_stats_min_budgets
= 194;
178 /* Default maximum budget values, in sectors and number of requests. */
179 static const int bfq_default_max_budget
= 16 * 1024;
182 * When a sync request is dispatched, the queue that contains that
183 * request, and all the ancestor entities of that queue, are charged
184 * with the number of sectors of the request. In contrast, if the
185 * request is async, then the queue and its ancestor entities are
186 * charged with the number of sectors of the request, multiplied by
187 * the factor below. This throttles the bandwidth for async I/O,
188 * w.r.t. to sync I/O, and it is done to counter the tendency of async
189 * writes to steal I/O throughput to reads.
191 * The current value of this parameter is the result of a tuning with
192 * several hardware and software configurations. We tried to find the
193 * lowest value for which writes do not cause noticeable problems to
194 * reads. In fact, the lower this parameter, the stabler I/O control,
195 * in the following respect. The lower this parameter is, the less
196 * the bandwidth enjoyed by a group decreases
197 * - when the group does writes, w.r.t. to when it does reads;
198 * - when other groups do reads, w.r.t. to when they do writes.
200 static const int bfq_async_charge_factor
= 3;
202 /* Default timeout values, in jiffies, approximating CFQ defaults. */
203 const int bfq_timeout
= HZ
/ 8;
206 * Time limit for merging (see comments in bfq_setup_cooperator). Set
207 * to the slowest value that, in our tests, proved to be effective in
208 * removing false positives, while not causing true positives to miss
211 * As can be deduced from the low time limit below, queue merging, if
212 * successful, happens at the very beginning of the I/O of the involved
213 * cooperating processes, as a consequence of the arrival of the very
214 * first requests from each cooperator. After that, there is very
215 * little chance to find cooperators.
217 static const unsigned long bfq_merge_time_limit
= HZ
/10;
219 static struct kmem_cache
*bfq_pool
;
221 /* Below this threshold (in ns), we consider thinktime immediate. */
222 #define BFQ_MIN_TT (2 * NSEC_PER_MSEC)
224 /* hw_tag detection: parallel requests threshold and min samples needed. */
225 #define BFQ_HW_QUEUE_THRESHOLD 3
226 #define BFQ_HW_QUEUE_SAMPLES 32
228 #define BFQQ_SEEK_THR (sector_t)(8 * 100)
229 #define BFQQ_SECT_THR_NONROT (sector_t)(2 * 32)
230 #define BFQ_RQ_SEEKY(bfqd, last_pos, rq) \
231 (get_sdist(last_pos, rq) > \
233 (!blk_queue_nonrot(bfqd->queue) || \
234 blk_rq_sectors(rq) < BFQQ_SECT_THR_NONROT))
235 #define BFQQ_CLOSE_THR (sector_t)(8 * 1024)
236 #define BFQQ_SEEKY(bfqq) (hweight32(bfqq->seek_history) > 19)
238 * Sync random I/O is likely to be confused with soft real-time I/O,
239 * because it is characterized by limited throughput and apparently
240 * isochronous arrival pattern. To avoid false positives, queues
241 * containing only random (seeky) I/O are prevented from being tagged
244 #define BFQQ_TOTALLY_SEEKY(bfqq) (bfqq->seek_history == -1)
246 /* Min number of samples required to perform peak-rate update */
247 #define BFQ_RATE_MIN_SAMPLES 32
248 /* Min observation time interval required to perform a peak-rate update (ns) */
249 #define BFQ_RATE_MIN_INTERVAL (300*NSEC_PER_MSEC)
250 /* Target observation time interval for a peak-rate update (ns) */
251 #define BFQ_RATE_REF_INTERVAL NSEC_PER_SEC
254 * Shift used for peak-rate fixed precision calculations.
256 * - the current shift: 16 positions
257 * - the current type used to store rate: u32
258 * - the current unit of measure for rate: [sectors/usec], or, more precisely,
259 * [(sectors/usec) / 2^BFQ_RATE_SHIFT] to take into account the shift,
260 * the range of rates that can be stored is
261 * [1 / 2^BFQ_RATE_SHIFT, 2^(32 - BFQ_RATE_SHIFT)] sectors/usec =
262 * [1 / 2^16, 2^16] sectors/usec = [15e-6, 65536] sectors/usec =
263 * [15, 65G] sectors/sec
264 * Which, assuming a sector size of 512B, corresponds to a range of
267 #define BFQ_RATE_SHIFT 16
270 * When configured for computing the duration of the weight-raising
271 * for interactive queues automatically (see the comments at the
272 * beginning of this file), BFQ does it using the following formula:
273 * duration = (ref_rate / r) * ref_wr_duration,
274 * where r is the peak rate of the device, and ref_rate and
275 * ref_wr_duration are two reference parameters. In particular,
276 * ref_rate is the peak rate of the reference storage device (see
277 * below), and ref_wr_duration is about the maximum time needed, with
278 * BFQ and while reading two files in parallel, to load typical large
279 * applications on the reference device (see the comments on
280 * max_service_from_wr below, for more details on how ref_wr_duration
281 * is obtained). In practice, the slower/faster the device at hand
282 * is, the more/less it takes to load applications with respect to the
283 * reference device. Accordingly, the longer/shorter BFQ grants
284 * weight raising to interactive applications.
286 * BFQ uses two different reference pairs (ref_rate, ref_wr_duration),
287 * depending on whether the device is rotational or non-rotational.
289 * In the following definitions, ref_rate[0] and ref_wr_duration[0]
290 * are the reference values for a rotational device, whereas
291 * ref_rate[1] and ref_wr_duration[1] are the reference values for a
292 * non-rotational device. The reference rates are not the actual peak
293 * rates of the devices used as a reference, but slightly lower
294 * values. The reason for using slightly lower values is that the
295 * peak-rate estimator tends to yield slightly lower values than the
296 * actual peak rate (it can yield the actual peak rate only if there
297 * is only one process doing I/O, and the process does sequential
300 * The reference peak rates are measured in sectors/usec, left-shifted
303 static int ref_rate
[2] = {14000, 33000};
305 * To improve readability, a conversion function is used to initialize
306 * the following array, which entails that the array can be
307 * initialized only in a function.
309 static int ref_wr_duration
[2];
312 * BFQ uses the above-detailed, time-based weight-raising mechanism to
313 * privilege interactive tasks. This mechanism is vulnerable to the
314 * following false positives: I/O-bound applications that will go on
315 * doing I/O for much longer than the duration of weight
316 * raising. These applications have basically no benefit from being
317 * weight-raised at the beginning of their I/O. On the opposite end,
318 * while being weight-raised, these applications
319 * a) unjustly steal throughput to applications that may actually need
321 * b) make BFQ uselessly perform device idling; device idling results
322 * in loss of device throughput with most flash-based storage, and may
323 * increase latencies when used purposelessly.
325 * BFQ tries to reduce these problems, by adopting the following
326 * countermeasure. To introduce this countermeasure, we need first to
327 * finish explaining how the duration of weight-raising for
328 * interactive tasks is computed.
330 * For a bfq_queue deemed as interactive, the duration of weight
331 * raising is dynamically adjusted, as a function of the estimated
332 * peak rate of the device, so as to be equal to the time needed to
333 * execute the 'largest' interactive task we benchmarked so far. By
334 * largest task, we mean the task for which each involved process has
335 * to do more I/O than for any of the other tasks we benchmarked. This
336 * reference interactive task is the start-up of LibreOffice Writer,
337 * and in this task each process/bfq_queue needs to have at most ~110K
338 * sectors transferred.
340 * This last piece of information enables BFQ to reduce the actual
341 * duration of weight-raising for at least one class of I/O-bound
342 * applications: those doing sequential or quasi-sequential I/O. An
343 * example is file copy. In fact, once started, the main I/O-bound
344 * processes of these applications usually consume the above 110K
345 * sectors in much less time than the processes of an application that
346 * is starting, because these I/O-bound processes will greedily devote
347 * almost all their CPU cycles only to their target,
348 * throughput-friendly I/O operations. This is even more true if BFQ
349 * happens to be underestimating the device peak rate, and thus
350 * overestimating the duration of weight raising. But, according to
351 * our measurements, once transferred 110K sectors, these processes
352 * have no right to be weight-raised any longer.
354 * Basing on the last consideration, BFQ ends weight-raising for a
355 * bfq_queue if the latter happens to have received an amount of
356 * service at least equal to the following constant. The constant is
357 * set to slightly more than 110K, to have a minimum safety margin.
359 * This early ending of weight-raising reduces the amount of time
360 * during which interactive false positives cause the two problems
361 * described at the beginning of these comments.
363 static const unsigned long max_service_from_wr
= 120000;
365 #define RQ_BIC(rq) icq_to_bic((rq)->elv.priv[0])
366 #define RQ_BFQQ(rq) ((rq)->elv.priv[1])
368 struct bfq_queue
*bic_to_bfqq(struct bfq_io_cq
*bic
, bool is_sync
)
370 return bic
->bfqq
[is_sync
];
373 void bic_set_bfqq(struct bfq_io_cq
*bic
, struct bfq_queue
*bfqq
, bool is_sync
)
375 bic
->bfqq
[is_sync
] = bfqq
;
378 struct bfq_data
*bic_to_bfqd(struct bfq_io_cq
*bic
)
380 return bic
->icq
.q
->elevator
->elevator_data
;
384 * icq_to_bic - convert iocontext queue structure to bfq_io_cq.
385 * @icq: the iocontext queue.
387 static struct bfq_io_cq
*icq_to_bic(struct io_cq
*icq
)
389 /* bic->icq is the first member, %NULL will convert to %NULL */
390 return container_of(icq
, struct bfq_io_cq
, icq
);
394 * bfq_bic_lookup - search into @ioc a bic associated to @bfqd.
395 * @bfqd: the lookup key.
396 * @ioc: the io_context of the process doing I/O.
397 * @q: the request queue.
399 static struct bfq_io_cq
*bfq_bic_lookup(struct bfq_data
*bfqd
,
400 struct io_context
*ioc
,
401 struct request_queue
*q
)
405 struct bfq_io_cq
*icq
;
407 spin_lock_irqsave(&q
->queue_lock
, flags
);
408 icq
= icq_to_bic(ioc_lookup_icq(ioc
, q
));
409 spin_unlock_irqrestore(&q
->queue_lock
, flags
);
418 * Scheduler run of queue, if there are requests pending and no one in the
419 * driver that will restart queueing.
421 void bfq_schedule_dispatch(struct bfq_data
*bfqd
)
423 if (bfqd
->queued
!= 0) {
424 bfq_log(bfqd
, "schedule dispatch");
425 blk_mq_run_hw_queues(bfqd
->queue
, true);
429 #define bfq_class_idle(bfqq) ((bfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
431 #define bfq_sample_valid(samples) ((samples) > 80)
434 * Lifted from AS - choose which of rq1 and rq2 that is best served now.
435 * We choose the request that is closer to the head right now. Distance
436 * behind the head is penalized and only allowed to a certain extent.
438 static struct request
*bfq_choose_req(struct bfq_data
*bfqd
,
443 sector_t s1
, s2
, d1
= 0, d2
= 0;
444 unsigned long back_max
;
445 #define BFQ_RQ1_WRAP 0x01 /* request 1 wraps */
446 #define BFQ_RQ2_WRAP 0x02 /* request 2 wraps */
447 unsigned int wrap
= 0; /* bit mask: requests behind the disk head? */
449 if (!rq1
|| rq1
== rq2
)
454 if (rq_is_sync(rq1
) && !rq_is_sync(rq2
))
456 else if (rq_is_sync(rq2
) && !rq_is_sync(rq1
))
458 if ((rq1
->cmd_flags
& REQ_META
) && !(rq2
->cmd_flags
& REQ_META
))
460 else if ((rq2
->cmd_flags
& REQ_META
) && !(rq1
->cmd_flags
& REQ_META
))
463 s1
= blk_rq_pos(rq1
);
464 s2
= blk_rq_pos(rq2
);
467 * By definition, 1KiB is 2 sectors.
469 back_max
= bfqd
->bfq_back_max
* 2;
472 * Strict one way elevator _except_ in the case where we allow
473 * short backward seeks which are biased as twice the cost of a
474 * similar forward seek.
478 else if (s1
+ back_max
>= last
)
479 d1
= (last
- s1
) * bfqd
->bfq_back_penalty
;
481 wrap
|= BFQ_RQ1_WRAP
;
485 else if (s2
+ back_max
>= last
)
486 d2
= (last
- s2
) * bfqd
->bfq_back_penalty
;
488 wrap
|= BFQ_RQ2_WRAP
;
490 /* Found required data */
493 * By doing switch() on the bit mask "wrap" we avoid having to
494 * check two variables for all permutations: --> faster!
497 case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
512 case BFQ_RQ1_WRAP
|BFQ_RQ2_WRAP
: /* both rqs wrapped */
515 * Since both rqs are wrapped,
516 * start with the one that's further behind head
517 * (--> only *one* back seek required),
518 * since back seek takes more time than forward.
528 * Async I/O can easily starve sync I/O (both sync reads and sync
529 * writes), by consuming all tags. Similarly, storms of sync writes,
530 * such as those that sync(2) may trigger, can starve sync reads.
531 * Limit depths of async I/O and sync writes so as to counter both
534 static void bfq_limit_depth(unsigned int op
, struct blk_mq_alloc_data
*data
)
536 struct bfq_data
*bfqd
= data
->q
->elevator
->elevator_data
;
538 if (op_is_sync(op
) && !op_is_write(op
))
541 data
->shallow_depth
=
542 bfqd
->word_depths
[!!bfqd
->wr_busy_queues
][op_is_sync(op
)];
544 bfq_log(bfqd
, "[%s] wr_busy %d sync %d depth %u",
545 __func__
, bfqd
->wr_busy_queues
, op_is_sync(op
),
546 data
->shallow_depth
);
549 static struct bfq_queue
*
550 bfq_rq_pos_tree_lookup(struct bfq_data
*bfqd
, struct rb_root
*root
,
551 sector_t sector
, struct rb_node
**ret_parent
,
552 struct rb_node
***rb_link
)
554 struct rb_node
**p
, *parent
;
555 struct bfq_queue
*bfqq
= NULL
;
563 bfqq
= rb_entry(parent
, struct bfq_queue
, pos_node
);
566 * Sort strictly based on sector. Smallest to the left,
567 * largest to the right.
569 if (sector
> blk_rq_pos(bfqq
->next_rq
))
571 else if (sector
< blk_rq_pos(bfqq
->next_rq
))
579 *ret_parent
= parent
;
583 bfq_log(bfqd
, "rq_pos_tree_lookup %llu: returning %d",
584 (unsigned long long)sector
,
585 bfqq
? bfqq
->pid
: 0);
590 static bool bfq_too_late_for_merging(struct bfq_queue
*bfqq
)
592 return bfqq
->service_from_backlogged
> 0 &&
593 time_is_before_jiffies(bfqq
->first_IO_time
+
594 bfq_merge_time_limit
);
598 * The following function is not marked as __cold because it is
599 * actually cold, but for the same performance goal described in the
600 * comments on the likely() at the beginning of
601 * bfq_setup_cooperator(). Unexpectedly, to reach an even lower
602 * execution time for the case where this function is not invoked, we
603 * had to add an unlikely() in each involved if().
606 bfq_pos_tree_add_move(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
)
608 struct rb_node
**p
, *parent
;
609 struct bfq_queue
*__bfqq
;
611 if (bfqq
->pos_root
) {
612 rb_erase(&bfqq
->pos_node
, bfqq
->pos_root
);
613 bfqq
->pos_root
= NULL
;
616 /* oom_bfqq does not participate in queue merging */
617 if (bfqq
== &bfqd
->oom_bfqq
)
621 * bfqq cannot be merged any longer (see comments in
622 * bfq_setup_cooperator): no point in adding bfqq into the
625 if (bfq_too_late_for_merging(bfqq
))
628 if (bfq_class_idle(bfqq
))
633 bfqq
->pos_root
= &bfq_bfqq_to_bfqg(bfqq
)->rq_pos_tree
;
634 __bfqq
= bfq_rq_pos_tree_lookup(bfqd
, bfqq
->pos_root
,
635 blk_rq_pos(bfqq
->next_rq
), &parent
, &p
);
637 rb_link_node(&bfqq
->pos_node
, parent
, p
);
638 rb_insert_color(&bfqq
->pos_node
, bfqq
->pos_root
);
640 bfqq
->pos_root
= NULL
;
644 * The following function returns false either if every active queue
645 * must receive the same share of the throughput (symmetric scenario),
646 * or, as a special case, if bfqq must receive a share of the
647 * throughput lower than or equal to the share that every other active
648 * queue must receive. If bfqq does sync I/O, then these are the only
649 * two cases where bfqq happens to be guaranteed its share of the
650 * throughput even if I/O dispatching is not plugged when bfqq remains
651 * temporarily empty (for more details, see the comments in the
652 * function bfq_better_to_idle()). For this reason, the return value
653 * of this function is used to check whether I/O-dispatch plugging can
656 * The above first case (symmetric scenario) occurs when:
657 * 1) all active queues have the same weight,
658 * 2) all active queues belong to the same I/O-priority class,
659 * 3) all active groups at the same level in the groups tree have the same
661 * 4) all active groups at the same level in the groups tree have the same
662 * number of children.
664 * Unfortunately, keeping the necessary state for evaluating exactly
665 * the last two symmetry sub-conditions above would be quite complex
666 * and time consuming. Therefore this function evaluates, instead,
667 * only the following stronger three sub-conditions, for which it is
668 * much easier to maintain the needed state:
669 * 1) all active queues have the same weight,
670 * 2) all active queues belong to the same I/O-priority class,
671 * 3) there are no active groups.
672 * In particular, the last condition is always true if hierarchical
673 * support or the cgroups interface are not enabled, thus no state
674 * needs to be maintained in this case.
676 static bool bfq_asymmetric_scenario(struct bfq_data
*bfqd
,
677 struct bfq_queue
*bfqq
)
679 bool smallest_weight
= bfqq
&&
680 bfqq
->weight_counter
&&
681 bfqq
->weight_counter
==
683 rb_first_cached(&bfqd
->queue_weights_tree
),
684 struct bfq_weight_counter
,
688 * For queue weights to differ, queue_weights_tree must contain
689 * at least two nodes.
691 bool varied_queue_weights
= !smallest_weight
&&
692 !RB_EMPTY_ROOT(&bfqd
->queue_weights_tree
.rb_root
) &&
693 (bfqd
->queue_weights_tree
.rb_root
.rb_node
->rb_left
||
694 bfqd
->queue_weights_tree
.rb_root
.rb_node
->rb_right
);
696 bool multiple_classes_busy
=
697 (bfqd
->busy_queues
[0] && bfqd
->busy_queues
[1]) ||
698 (bfqd
->busy_queues
[0] && bfqd
->busy_queues
[2]) ||
699 (bfqd
->busy_queues
[1] && bfqd
->busy_queues
[2]);
701 return varied_queue_weights
|| multiple_classes_busy
702 #ifdef CONFIG_BFQ_GROUP_IOSCHED
703 || bfqd
->num_groups_with_pending_reqs
> 0
709 * If the weight-counter tree passed as input contains no counter for
710 * the weight of the input queue, then add that counter; otherwise just
711 * increment the existing counter.
713 * Note that weight-counter trees contain few nodes in mostly symmetric
714 * scenarios. For example, if all queues have the same weight, then the
715 * weight-counter tree for the queues may contain at most one node.
716 * This holds even if low_latency is on, because weight-raised queues
717 * are not inserted in the tree.
718 * In most scenarios, the rate at which nodes are created/destroyed
721 void bfq_weights_tree_add(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
,
722 struct rb_root_cached
*root
)
724 struct bfq_entity
*entity
= &bfqq
->entity
;
725 struct rb_node
**new = &(root
->rb_root
.rb_node
), *parent
= NULL
;
726 bool leftmost
= true;
729 * Do not insert if the queue is already associated with a
730 * counter, which happens if:
731 * 1) a request arrival has caused the queue to become both
732 * non-weight-raised, and hence change its weight, and
733 * backlogged; in this respect, each of the two events
734 * causes an invocation of this function,
735 * 2) this is the invocation of this function caused by the
736 * second event. This second invocation is actually useless,
737 * and we handle this fact by exiting immediately. More
738 * efficient or clearer solutions might possibly be adopted.
740 if (bfqq
->weight_counter
)
744 struct bfq_weight_counter
*__counter
= container_of(*new,
745 struct bfq_weight_counter
,
749 if (entity
->weight
== __counter
->weight
) {
750 bfqq
->weight_counter
= __counter
;
753 if (entity
->weight
< __counter
->weight
)
754 new = &((*new)->rb_left
);
756 new = &((*new)->rb_right
);
761 bfqq
->weight_counter
= kzalloc(sizeof(struct bfq_weight_counter
),
765 * In the unlucky event of an allocation failure, we just
766 * exit. This will cause the weight of queue to not be
767 * considered in bfq_asymmetric_scenario, which, in its turn,
768 * causes the scenario to be deemed wrongly symmetric in case
769 * bfqq's weight would have been the only weight making the
770 * scenario asymmetric. On the bright side, no unbalance will
771 * however occur when bfqq becomes inactive again (the
772 * invocation of this function is triggered by an activation
773 * of queue). In fact, bfq_weights_tree_remove does nothing
774 * if !bfqq->weight_counter.
776 if (unlikely(!bfqq
->weight_counter
))
779 bfqq
->weight_counter
->weight
= entity
->weight
;
780 rb_link_node(&bfqq
->weight_counter
->weights_node
, parent
, new);
781 rb_insert_color_cached(&bfqq
->weight_counter
->weights_node
, root
,
785 bfqq
->weight_counter
->num_active
++;
790 * Decrement the weight counter associated with the queue, and, if the
791 * counter reaches 0, remove the counter from the tree.
792 * See the comments to the function bfq_weights_tree_add() for considerations
795 void __bfq_weights_tree_remove(struct bfq_data
*bfqd
,
796 struct bfq_queue
*bfqq
,
797 struct rb_root_cached
*root
)
799 if (!bfqq
->weight_counter
)
802 bfqq
->weight_counter
->num_active
--;
803 if (bfqq
->weight_counter
->num_active
> 0)
804 goto reset_entity_pointer
;
806 rb_erase_cached(&bfqq
->weight_counter
->weights_node
, root
);
807 kfree(bfqq
->weight_counter
);
809 reset_entity_pointer
:
810 bfqq
->weight_counter
= NULL
;
815 * Invoke __bfq_weights_tree_remove on bfqq and decrement the number
816 * of active groups for each queue's inactive parent entity.
818 void bfq_weights_tree_remove(struct bfq_data
*bfqd
,
819 struct bfq_queue
*bfqq
)
821 struct bfq_entity
*entity
= bfqq
->entity
.parent
;
823 for_each_entity(entity
) {
824 struct bfq_sched_data
*sd
= entity
->my_sched_data
;
826 if (sd
->next_in_service
|| sd
->in_service_entity
) {
828 * entity is still active, because either
829 * next_in_service or in_service_entity is not
830 * NULL (see the comments on the definition of
831 * next_in_service for details on why
832 * in_service_entity must be checked too).
834 * As a consequence, its parent entities are
835 * active as well, and thus this loop must
842 * The decrement of num_groups_with_pending_reqs is
843 * not performed immediately upon the deactivation of
844 * entity, but it is delayed to when it also happens
845 * that the first leaf descendant bfqq of entity gets
846 * all its pending requests completed. The following
847 * instructions perform this delayed decrement, if
848 * needed. See the comments on
849 * num_groups_with_pending_reqs for details.
851 if (entity
->in_groups_with_pending_reqs
) {
852 entity
->in_groups_with_pending_reqs
= false;
853 bfqd
->num_groups_with_pending_reqs
--;
858 * Next function is invoked last, because it causes bfqq to be
859 * freed if the following holds: bfqq is not in service and
860 * has no dispatched request. DO NOT use bfqq after the next
861 * function invocation.
863 __bfq_weights_tree_remove(bfqd
, bfqq
,
864 &bfqd
->queue_weights_tree
);
868 * Return expired entry, or NULL to just start from scratch in rbtree.
870 static struct request
*bfq_check_fifo(struct bfq_queue
*bfqq
,
871 struct request
*last
)
875 if (bfq_bfqq_fifo_expire(bfqq
))
878 bfq_mark_bfqq_fifo_expire(bfqq
);
880 rq
= rq_entry_fifo(bfqq
->fifo
.next
);
882 if (rq
== last
|| ktime_get_ns() < rq
->fifo_time
)
885 bfq_log_bfqq(bfqq
->bfqd
, bfqq
, "check_fifo: returned %p", rq
);
889 static struct request
*bfq_find_next_rq(struct bfq_data
*bfqd
,
890 struct bfq_queue
*bfqq
,
891 struct request
*last
)
893 struct rb_node
*rbnext
= rb_next(&last
->rb_node
);
894 struct rb_node
*rbprev
= rb_prev(&last
->rb_node
);
895 struct request
*next
, *prev
= NULL
;
897 /* Follow expired path, else get first next available. */
898 next
= bfq_check_fifo(bfqq
, last
);
903 prev
= rb_entry_rq(rbprev
);
906 next
= rb_entry_rq(rbnext
);
908 rbnext
= rb_first(&bfqq
->sort_list
);
909 if (rbnext
&& rbnext
!= &last
->rb_node
)
910 next
= rb_entry_rq(rbnext
);
913 return bfq_choose_req(bfqd
, next
, prev
, blk_rq_pos(last
));
916 /* see the definition of bfq_async_charge_factor for details */
917 static unsigned long bfq_serv_to_charge(struct request
*rq
,
918 struct bfq_queue
*bfqq
)
920 if (bfq_bfqq_sync(bfqq
) || bfqq
->wr_coeff
> 1 ||
921 bfq_asymmetric_scenario(bfqq
->bfqd
, bfqq
))
922 return blk_rq_sectors(rq
);
924 return blk_rq_sectors(rq
) * bfq_async_charge_factor
;
928 * bfq_updated_next_req - update the queue after a new next_rq selection.
929 * @bfqd: the device data the queue belongs to.
930 * @bfqq: the queue to update.
932 * If the first request of a queue changes we make sure that the queue
933 * has enough budget to serve at least its first request (if the
934 * request has grown). We do this because if the queue has not enough
935 * budget for its first request, it has to go through two dispatch
936 * rounds to actually get it dispatched.
938 static void bfq_updated_next_req(struct bfq_data
*bfqd
,
939 struct bfq_queue
*bfqq
)
941 struct bfq_entity
*entity
= &bfqq
->entity
;
942 struct request
*next_rq
= bfqq
->next_rq
;
943 unsigned long new_budget
;
948 if (bfqq
== bfqd
->in_service_queue
)
950 * In order not to break guarantees, budgets cannot be
951 * changed after an entity has been selected.
955 new_budget
= max_t(unsigned long,
956 max_t(unsigned long, bfqq
->max_budget
,
957 bfq_serv_to_charge(next_rq
, bfqq
)),
959 if (entity
->budget
!= new_budget
) {
960 entity
->budget
= new_budget
;
961 bfq_log_bfqq(bfqd
, bfqq
, "updated next rq: new budget %lu",
963 bfq_requeue_bfqq(bfqd
, bfqq
, false);
967 static unsigned int bfq_wr_duration(struct bfq_data
*bfqd
)
971 if (bfqd
->bfq_wr_max_time
> 0)
972 return bfqd
->bfq_wr_max_time
;
974 dur
= bfqd
->rate_dur_prod
;
975 do_div(dur
, bfqd
->peak_rate
);
978 * Limit duration between 3 and 25 seconds. The upper limit
979 * has been conservatively set after the following worst case:
980 * on a QEMU/KVM virtual machine
981 * - running in a slow PC
982 * - with a virtual disk stacked on a slow low-end 5400rpm HDD
983 * - serving a heavy I/O workload, such as the sequential reading
985 * mplayer took 23 seconds to start, if constantly weight-raised.
987 * As for higher values than that accommodating the above bad
988 * scenario, tests show that higher values would often yield
989 * the opposite of the desired result, i.e., would worsen
990 * responsiveness by allowing non-interactive applications to
991 * preserve weight raising for too long.
993 * On the other end, lower values than 3 seconds make it
994 * difficult for most interactive tasks to complete their jobs
995 * before weight-raising finishes.
997 return clamp_val(dur
, msecs_to_jiffies(3000), msecs_to_jiffies(25000));
1000 /* switch back from soft real-time to interactive weight raising */
1001 static void switch_back_to_interactive_wr(struct bfq_queue
*bfqq
,
1002 struct bfq_data
*bfqd
)
1004 bfqq
->wr_coeff
= bfqd
->bfq_wr_coeff
;
1005 bfqq
->wr_cur_max_time
= bfq_wr_duration(bfqd
);
1006 bfqq
->last_wr_start_finish
= bfqq
->wr_start_at_switch_to_srt
;
1010 bfq_bfqq_resume_state(struct bfq_queue
*bfqq
, struct bfq_data
*bfqd
,
1011 struct bfq_io_cq
*bic
, bool bfq_already_existing
)
1013 unsigned int old_wr_coeff
= bfqq
->wr_coeff
;
1014 bool busy
= bfq_already_existing
&& bfq_bfqq_busy(bfqq
);
1016 if (bic
->saved_has_short_ttime
)
1017 bfq_mark_bfqq_has_short_ttime(bfqq
);
1019 bfq_clear_bfqq_has_short_ttime(bfqq
);
1021 if (bic
->saved_IO_bound
)
1022 bfq_mark_bfqq_IO_bound(bfqq
);
1024 bfq_clear_bfqq_IO_bound(bfqq
);
1026 bfqq
->entity
.new_weight
= bic
->saved_weight
;
1027 bfqq
->ttime
= bic
->saved_ttime
;
1028 bfqq
->wr_coeff
= bic
->saved_wr_coeff
;
1029 bfqq
->wr_start_at_switch_to_srt
= bic
->saved_wr_start_at_switch_to_srt
;
1030 bfqq
->last_wr_start_finish
= bic
->saved_last_wr_start_finish
;
1031 bfqq
->wr_cur_max_time
= bic
->saved_wr_cur_max_time
;
1033 if (bfqq
->wr_coeff
> 1 && (bfq_bfqq_in_large_burst(bfqq
) ||
1034 time_is_before_jiffies(bfqq
->last_wr_start_finish
+
1035 bfqq
->wr_cur_max_time
))) {
1036 if (bfqq
->wr_cur_max_time
== bfqd
->bfq_wr_rt_max_time
&&
1037 !bfq_bfqq_in_large_burst(bfqq
) &&
1038 time_is_after_eq_jiffies(bfqq
->wr_start_at_switch_to_srt
+
1039 bfq_wr_duration(bfqd
))) {
1040 switch_back_to_interactive_wr(bfqq
, bfqd
);
1043 bfq_log_bfqq(bfqq
->bfqd
, bfqq
,
1044 "resume state: switching off wr");
1048 /* make sure weight will be updated, however we got here */
1049 bfqq
->entity
.prio_changed
= 1;
1054 if (old_wr_coeff
== 1 && bfqq
->wr_coeff
> 1)
1055 bfqd
->wr_busy_queues
++;
1056 else if (old_wr_coeff
> 1 && bfqq
->wr_coeff
== 1)
1057 bfqd
->wr_busy_queues
--;
1060 static int bfqq_process_refs(struct bfq_queue
*bfqq
)
1062 return bfqq
->ref
- bfqq
->allocated
- bfqq
->entity
.on_st_or_in_serv
-
1063 (bfqq
->weight_counter
!= NULL
);
1066 /* Empty burst list and add just bfqq (see comments on bfq_handle_burst) */
1067 static void bfq_reset_burst_list(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
)
1069 struct bfq_queue
*item
;
1070 struct hlist_node
*n
;
1072 hlist_for_each_entry_safe(item
, n
, &bfqd
->burst_list
, burst_list_node
)
1073 hlist_del_init(&item
->burst_list_node
);
1076 * Start the creation of a new burst list only if there is no
1077 * active queue. See comments on the conditional invocation of
1078 * bfq_handle_burst().
1080 if (bfq_tot_busy_queues(bfqd
) == 0) {
1081 hlist_add_head(&bfqq
->burst_list_node
, &bfqd
->burst_list
);
1082 bfqd
->burst_size
= 1;
1084 bfqd
->burst_size
= 0;
1086 bfqd
->burst_parent_entity
= bfqq
->entity
.parent
;
1089 /* Add bfqq to the list of queues in current burst (see bfq_handle_burst) */
1090 static void bfq_add_to_burst(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
)
1092 /* Increment burst size to take into account also bfqq */
1095 if (bfqd
->burst_size
== bfqd
->bfq_large_burst_thresh
) {
1096 struct bfq_queue
*pos
, *bfqq_item
;
1097 struct hlist_node
*n
;
1100 * Enough queues have been activated shortly after each
1101 * other to consider this burst as large.
1103 bfqd
->large_burst
= true;
1106 * We can now mark all queues in the burst list as
1107 * belonging to a large burst.
1109 hlist_for_each_entry(bfqq_item
, &bfqd
->burst_list
,
1111 bfq_mark_bfqq_in_large_burst(bfqq_item
);
1112 bfq_mark_bfqq_in_large_burst(bfqq
);
1115 * From now on, and until the current burst finishes, any
1116 * new queue being activated shortly after the last queue
1117 * was inserted in the burst can be immediately marked as
1118 * belonging to a large burst. So the burst list is not
1119 * needed any more. Remove it.
1121 hlist_for_each_entry_safe(pos
, n
, &bfqd
->burst_list
,
1123 hlist_del_init(&pos
->burst_list_node
);
1125 * Burst not yet large: add bfqq to the burst list. Do
1126 * not increment the ref counter for bfqq, because bfqq
1127 * is removed from the burst list before freeing bfqq
1130 hlist_add_head(&bfqq
->burst_list_node
, &bfqd
->burst_list
);
1134 * If many queues belonging to the same group happen to be created
1135 * shortly after each other, then the processes associated with these
1136 * queues have typically a common goal. In particular, bursts of queue
1137 * creations are usually caused by services or applications that spawn
1138 * many parallel threads/processes. Examples are systemd during boot,
1139 * or git grep. To help these processes get their job done as soon as
1140 * possible, it is usually better to not grant either weight-raising
1141 * or device idling to their queues, unless these queues must be
1142 * protected from the I/O flowing through other active queues.
1144 * In this comment we describe, firstly, the reasons why this fact
1145 * holds, and, secondly, the next function, which implements the main
1146 * steps needed to properly mark these queues so that they can then be
1147 * treated in a different way.
1149 * The above services or applications benefit mostly from a high
1150 * throughput: the quicker the requests of the activated queues are
1151 * cumulatively served, the sooner the target job of these queues gets
1152 * completed. As a consequence, weight-raising any of these queues,
1153 * which also implies idling the device for it, is almost always
1154 * counterproductive, unless there are other active queues to isolate
1155 * these new queues from. If there no other active queues, then
1156 * weight-raising these new queues just lowers throughput in most
1159 * On the other hand, a burst of queue creations may be caused also by
1160 * the start of an application that does not consist of a lot of
1161 * parallel I/O-bound threads. In fact, with a complex application,
1162 * several short processes may need to be executed to start-up the
1163 * application. In this respect, to start an application as quickly as
1164 * possible, the best thing to do is in any case to privilege the I/O
1165 * related to the application with respect to all other
1166 * I/O. Therefore, the best strategy to start as quickly as possible
1167 * an application that causes a burst of queue creations is to
1168 * weight-raise all the queues created during the burst. This is the
1169 * exact opposite of the best strategy for the other type of bursts.
1171 * In the end, to take the best action for each of the two cases, the
1172 * two types of bursts need to be distinguished. Fortunately, this
1173 * seems relatively easy, by looking at the sizes of the bursts. In
1174 * particular, we found a threshold such that only bursts with a
1175 * larger size than that threshold are apparently caused by
1176 * services or commands such as systemd or git grep. For brevity,
1177 * hereafter we call just 'large' these bursts. BFQ *does not*
1178 * weight-raise queues whose creation occurs in a large burst. In
1179 * addition, for each of these queues BFQ performs or does not perform
1180 * idling depending on which choice boosts the throughput more. The
1181 * exact choice depends on the device and request pattern at
1184 * Unfortunately, false positives may occur while an interactive task
1185 * is starting (e.g., an application is being started). The
1186 * consequence is that the queues associated with the task do not
1187 * enjoy weight raising as expected. Fortunately these false positives
1188 * are very rare. They typically occur if some service happens to
1189 * start doing I/O exactly when the interactive task starts.
1191 * Turning back to the next function, it is invoked only if there are
1192 * no active queues (apart from active queues that would belong to the
1193 * same, possible burst bfqq would belong to), and it implements all
1194 * the steps needed to detect the occurrence of a large burst and to
1195 * properly mark all the queues belonging to it (so that they can then
1196 * be treated in a different way). This goal is achieved by
1197 * maintaining a "burst list" that holds, temporarily, the queues that
1198 * belong to the burst in progress. The list is then used to mark
1199 * these queues as belonging to a large burst if the burst does become
1200 * large. The main steps are the following.
1202 * . when the very first queue is created, the queue is inserted into the
1203 * list (as it could be the first queue in a possible burst)
1205 * . if the current burst has not yet become large, and a queue Q that does
1206 * not yet belong to the burst is activated shortly after the last time
1207 * at which a new queue entered the burst list, then the function appends
1208 * Q to the burst list
1210 * . if, as a consequence of the previous step, the burst size reaches
1211 * the large-burst threshold, then
1213 * . all the queues in the burst list are marked as belonging to a
1216 * . the burst list is deleted; in fact, the burst list already served
1217 * its purpose (keeping temporarily track of the queues in a burst,
1218 * so as to be able to mark them as belonging to a large burst in the
1219 * previous sub-step), and now is not needed any more
1221 * . the device enters a large-burst mode
1223 * . if a queue Q that does not belong to the burst is created while
1224 * the device is in large-burst mode and shortly after the last time
1225 * at which a queue either entered the burst list or was marked as
1226 * belonging to the current large burst, then Q is immediately marked
1227 * as belonging to a large burst.
1229 * . if a queue Q that does not belong to the burst is created a while
1230 * later, i.e., not shortly after, than the last time at which a queue
1231 * either entered the burst list or was marked as belonging to the
1232 * current large burst, then the current burst is deemed as finished and:
1234 * . the large-burst mode is reset if set
1236 * . the burst list is emptied
1238 * . Q is inserted in the burst list, as Q may be the first queue
1239 * in a possible new burst (then the burst list contains just Q
1242 static void bfq_handle_burst(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
)
1245 * If bfqq is already in the burst list or is part of a large
1246 * burst, or finally has just been split, then there is
1247 * nothing else to do.
1249 if (!hlist_unhashed(&bfqq
->burst_list_node
) ||
1250 bfq_bfqq_in_large_burst(bfqq
) ||
1251 time_is_after_eq_jiffies(bfqq
->split_time
+
1252 msecs_to_jiffies(10)))
1256 * If bfqq's creation happens late enough, or bfqq belongs to
1257 * a different group than the burst group, then the current
1258 * burst is finished, and related data structures must be
1261 * In this respect, consider the special case where bfqq is
1262 * the very first queue created after BFQ is selected for this
1263 * device. In this case, last_ins_in_burst and
1264 * burst_parent_entity are not yet significant when we get
1265 * here. But it is easy to verify that, whether or not the
1266 * following condition is true, bfqq will end up being
1267 * inserted into the burst list. In particular the list will
1268 * happen to contain only bfqq. And this is exactly what has
1269 * to happen, as bfqq may be the first queue of the first
1272 if (time_is_before_jiffies(bfqd
->last_ins_in_burst
+
1273 bfqd
->bfq_burst_interval
) ||
1274 bfqq
->entity
.parent
!= bfqd
->burst_parent_entity
) {
1275 bfqd
->large_burst
= false;
1276 bfq_reset_burst_list(bfqd
, bfqq
);
1281 * If we get here, then bfqq is being activated shortly after the
1282 * last queue. So, if the current burst is also large, we can mark
1283 * bfqq as belonging to this large burst immediately.
1285 if (bfqd
->large_burst
) {
1286 bfq_mark_bfqq_in_large_burst(bfqq
);
1291 * If we get here, then a large-burst state has not yet been
1292 * reached, but bfqq is being activated shortly after the last
1293 * queue. Then we add bfqq to the burst.
1295 bfq_add_to_burst(bfqd
, bfqq
);
1298 * At this point, bfqq either has been added to the current
1299 * burst or has caused the current burst to terminate and a
1300 * possible new burst to start. In particular, in the second
1301 * case, bfqq has become the first queue in the possible new
1302 * burst. In both cases last_ins_in_burst needs to be moved
1305 bfqd
->last_ins_in_burst
= jiffies
;
1308 static int bfq_bfqq_budget_left(struct bfq_queue
*bfqq
)
1310 struct bfq_entity
*entity
= &bfqq
->entity
;
1312 return entity
->budget
- entity
->service
;
1316 * If enough samples have been computed, return the current max budget
1317 * stored in bfqd, which is dynamically updated according to the
1318 * estimated disk peak rate; otherwise return the default max budget
1320 static int bfq_max_budget(struct bfq_data
*bfqd
)
1322 if (bfqd
->budgets_assigned
< bfq_stats_min_budgets
)
1323 return bfq_default_max_budget
;
1325 return bfqd
->bfq_max_budget
;
1329 * Return min budget, which is a fraction of the current or default
1330 * max budget (trying with 1/32)
1332 static int bfq_min_budget(struct bfq_data
*bfqd
)
1334 if (bfqd
->budgets_assigned
< bfq_stats_min_budgets
)
1335 return bfq_default_max_budget
/ 32;
1337 return bfqd
->bfq_max_budget
/ 32;
1341 * The next function, invoked after the input queue bfqq switches from
1342 * idle to busy, updates the budget of bfqq. The function also tells
1343 * whether the in-service queue should be expired, by returning
1344 * true. The purpose of expiring the in-service queue is to give bfqq
1345 * the chance to possibly preempt the in-service queue, and the reason
1346 * for preempting the in-service queue is to achieve one of the two
1349 * 1. Guarantee to bfqq its reserved bandwidth even if bfqq has
1350 * expired because it has remained idle. In particular, bfqq may have
1351 * expired for one of the following two reasons:
1353 * - BFQQE_NO_MORE_REQUESTS bfqq did not enjoy any device idling
1354 * and did not make it to issue a new request before its last
1355 * request was served;
1357 * - BFQQE_TOO_IDLE bfqq did enjoy device idling, but did not issue
1358 * a new request before the expiration of the idling-time.
1360 * Even if bfqq has expired for one of the above reasons, the process
1361 * associated with the queue may be however issuing requests greedily,
1362 * and thus be sensitive to the bandwidth it receives (bfqq may have
1363 * remained idle for other reasons: CPU high load, bfqq not enjoying
1364 * idling, I/O throttling somewhere in the path from the process to
1365 * the I/O scheduler, ...). But if, after every expiration for one of
1366 * the above two reasons, bfqq has to wait for the service of at least
1367 * one full budget of another queue before being served again, then
1368 * bfqq is likely to get a much lower bandwidth or resource time than
1369 * its reserved ones. To address this issue, two countermeasures need
1372 * First, the budget and the timestamps of bfqq need to be updated in
1373 * a special way on bfqq reactivation: they need to be updated as if
1374 * bfqq did not remain idle and did not expire. In fact, if they are
1375 * computed as if bfqq expired and remained idle until reactivation,
1376 * then the process associated with bfqq is treated as if, instead of
1377 * being greedy, it stopped issuing requests when bfqq remained idle,
1378 * and restarts issuing requests only on this reactivation. In other
1379 * words, the scheduler does not help the process recover the "service
1380 * hole" between bfqq expiration and reactivation. As a consequence,
1381 * the process receives a lower bandwidth than its reserved one. In
1382 * contrast, to recover this hole, the budget must be updated as if
1383 * bfqq was not expired at all before this reactivation, i.e., it must
1384 * be set to the value of the remaining budget when bfqq was
1385 * expired. Along the same line, timestamps need to be assigned the
1386 * value they had the last time bfqq was selected for service, i.e.,
1387 * before last expiration. Thus timestamps need to be back-shifted
1388 * with respect to their normal computation (see [1] for more details
1389 * on this tricky aspect).
1391 * Secondly, to allow the process to recover the hole, the in-service
1392 * queue must be expired too, to give bfqq the chance to preempt it
1393 * immediately. In fact, if bfqq has to wait for a full budget of the
1394 * in-service queue to be completed, then it may become impossible to
1395 * let the process recover the hole, even if the back-shifted
1396 * timestamps of bfqq are lower than those of the in-service queue. If
1397 * this happens for most or all of the holes, then the process may not
1398 * receive its reserved bandwidth. In this respect, it is worth noting
1399 * that, being the service of outstanding requests unpreemptible, a
1400 * little fraction of the holes may however be unrecoverable, thereby
1401 * causing a little loss of bandwidth.
1403 * The last important point is detecting whether bfqq does need this
1404 * bandwidth recovery. In this respect, the next function deems the
1405 * process associated with bfqq greedy, and thus allows it to recover
1406 * the hole, if: 1) the process is waiting for the arrival of a new
1407 * request (which implies that bfqq expired for one of the above two
1408 * reasons), and 2) such a request has arrived soon. The first
1409 * condition is controlled through the flag non_blocking_wait_rq,
1410 * while the second through the flag arrived_in_time. If both
1411 * conditions hold, then the function computes the budget in the
1412 * above-described special way, and signals that the in-service queue
1413 * should be expired. Timestamp back-shifting is done later in
1414 * __bfq_activate_entity.
1416 * 2. Reduce latency. Even if timestamps are not backshifted to let
1417 * the process associated with bfqq recover a service hole, bfqq may
1418 * however happen to have, after being (re)activated, a lower finish
1419 * timestamp than the in-service queue. That is, the next budget of
1420 * bfqq may have to be completed before the one of the in-service
1421 * queue. If this is the case, then preempting the in-service queue
1422 * allows this goal to be achieved, apart from the unpreemptible,
1423 * outstanding requests mentioned above.
1425 * Unfortunately, regardless of which of the above two goals one wants
1426 * to achieve, service trees need first to be updated to know whether
1427 * the in-service queue must be preempted. To have service trees
1428 * correctly updated, the in-service queue must be expired and
1429 * rescheduled, and bfqq must be scheduled too. This is one of the
1430 * most costly operations (in future versions, the scheduling
1431 * mechanism may be re-designed in such a way to make it possible to
1432 * know whether preemption is needed without needing to update service
1433 * trees). In addition, queue preemptions almost always cause random
1434 * I/O, which may in turn cause loss of throughput. Finally, there may
1435 * even be no in-service queue when the next function is invoked (so,
1436 * no queue to compare timestamps with). Because of these facts, the
1437 * next function adopts the following simple scheme to avoid costly
1438 * operations, too frequent preemptions and too many dependencies on
1439 * the state of the scheduler: it requests the expiration of the
1440 * in-service queue (unconditionally) only for queues that need to
1441 * recover a hole. Then it delegates to other parts of the code the
1442 * responsibility of handling the above case 2.
1444 static bool bfq_bfqq_update_budg_for_activation(struct bfq_data
*bfqd
,
1445 struct bfq_queue
*bfqq
,
1446 bool arrived_in_time
)
1448 struct bfq_entity
*entity
= &bfqq
->entity
;
1451 * In the next compound condition, we check also whether there
1452 * is some budget left, because otherwise there is no point in
1453 * trying to go on serving bfqq with this same budget: bfqq
1454 * would be expired immediately after being selected for
1455 * service. This would only cause useless overhead.
1457 if (bfq_bfqq_non_blocking_wait_rq(bfqq
) && arrived_in_time
&&
1458 bfq_bfqq_budget_left(bfqq
) > 0) {
1460 * We do not clear the flag non_blocking_wait_rq here, as
1461 * the latter is used in bfq_activate_bfqq to signal
1462 * that timestamps need to be back-shifted (and is
1463 * cleared right after).
1467 * In next assignment we rely on that either
1468 * entity->service or entity->budget are not updated
1469 * on expiration if bfqq is empty (see
1470 * __bfq_bfqq_recalc_budget). Thus both quantities
1471 * remain unchanged after such an expiration, and the
1472 * following statement therefore assigns to
1473 * entity->budget the remaining budget on such an
1476 entity
->budget
= min_t(unsigned long,
1477 bfq_bfqq_budget_left(bfqq
),
1481 * At this point, we have used entity->service to get
1482 * the budget left (needed for updating
1483 * entity->budget). Thus we finally can, and have to,
1484 * reset entity->service. The latter must be reset
1485 * because bfqq would otherwise be charged again for
1486 * the service it has received during its previous
1489 entity
->service
= 0;
1495 * We can finally complete expiration, by setting service to 0.
1497 entity
->service
= 0;
1498 entity
->budget
= max_t(unsigned long, bfqq
->max_budget
,
1499 bfq_serv_to_charge(bfqq
->next_rq
, bfqq
));
1500 bfq_clear_bfqq_non_blocking_wait_rq(bfqq
);
1505 * Return the farthest past time instant according to jiffies
1508 static unsigned long bfq_smallest_from_now(void)
1510 return jiffies
- MAX_JIFFY_OFFSET
;
1513 static void bfq_update_bfqq_wr_on_rq_arrival(struct bfq_data
*bfqd
,
1514 struct bfq_queue
*bfqq
,
1515 unsigned int old_wr_coeff
,
1516 bool wr_or_deserves_wr
,
1521 if (old_wr_coeff
== 1 && wr_or_deserves_wr
) {
1522 /* start a weight-raising period */
1524 bfqq
->service_from_wr
= 0;
1525 bfqq
->wr_coeff
= bfqd
->bfq_wr_coeff
;
1526 bfqq
->wr_cur_max_time
= bfq_wr_duration(bfqd
);
1529 * No interactive weight raising in progress
1530 * here: assign minus infinity to
1531 * wr_start_at_switch_to_srt, to make sure
1532 * that, at the end of the soft-real-time
1533 * weight raising periods that is starting
1534 * now, no interactive weight-raising period
1535 * may be wrongly considered as still in
1536 * progress (and thus actually started by
1539 bfqq
->wr_start_at_switch_to_srt
=
1540 bfq_smallest_from_now();
1541 bfqq
->wr_coeff
= bfqd
->bfq_wr_coeff
*
1542 BFQ_SOFTRT_WEIGHT_FACTOR
;
1543 bfqq
->wr_cur_max_time
=
1544 bfqd
->bfq_wr_rt_max_time
;
1548 * If needed, further reduce budget to make sure it is
1549 * close to bfqq's backlog, so as to reduce the
1550 * scheduling-error component due to a too large
1551 * budget. Do not care about throughput consequences,
1552 * but only about latency. Finally, do not assign a
1553 * too small budget either, to avoid increasing
1554 * latency by causing too frequent expirations.
1556 bfqq
->entity
.budget
= min_t(unsigned long,
1557 bfqq
->entity
.budget
,
1558 2 * bfq_min_budget(bfqd
));
1559 } else if (old_wr_coeff
> 1) {
1560 if (interactive
) { /* update wr coeff and duration */
1561 bfqq
->wr_coeff
= bfqd
->bfq_wr_coeff
;
1562 bfqq
->wr_cur_max_time
= bfq_wr_duration(bfqd
);
1563 } else if (in_burst
)
1567 * The application is now or still meeting the
1568 * requirements for being deemed soft rt. We
1569 * can then correctly and safely (re)charge
1570 * the weight-raising duration for the
1571 * application with the weight-raising
1572 * duration for soft rt applications.
1574 * In particular, doing this recharge now, i.e.,
1575 * before the weight-raising period for the
1576 * application finishes, reduces the probability
1577 * of the following negative scenario:
1578 * 1) the weight of a soft rt application is
1579 * raised at startup (as for any newly
1580 * created application),
1581 * 2) since the application is not interactive,
1582 * at a certain time weight-raising is
1583 * stopped for the application,
1584 * 3) at that time the application happens to
1585 * still have pending requests, and hence
1586 * is destined to not have a chance to be
1587 * deemed soft rt before these requests are
1588 * completed (see the comments to the
1589 * function bfq_bfqq_softrt_next_start()
1590 * for details on soft rt detection),
1591 * 4) these pending requests experience a high
1592 * latency because the application is not
1593 * weight-raised while they are pending.
1595 if (bfqq
->wr_cur_max_time
!=
1596 bfqd
->bfq_wr_rt_max_time
) {
1597 bfqq
->wr_start_at_switch_to_srt
=
1598 bfqq
->last_wr_start_finish
;
1600 bfqq
->wr_cur_max_time
=
1601 bfqd
->bfq_wr_rt_max_time
;
1602 bfqq
->wr_coeff
= bfqd
->bfq_wr_coeff
*
1603 BFQ_SOFTRT_WEIGHT_FACTOR
;
1605 bfqq
->last_wr_start_finish
= jiffies
;
1610 static bool bfq_bfqq_idle_for_long_time(struct bfq_data
*bfqd
,
1611 struct bfq_queue
*bfqq
)
1613 return bfqq
->dispatched
== 0 &&
1614 time_is_before_jiffies(
1615 bfqq
->budget_timeout
+
1616 bfqd
->bfq_wr_min_idle_time
);
1621 * Return true if bfqq is in a higher priority class, or has a higher
1622 * weight than the in-service queue.
1624 static bool bfq_bfqq_higher_class_or_weight(struct bfq_queue
*bfqq
,
1625 struct bfq_queue
*in_serv_bfqq
)
1627 int bfqq_weight
, in_serv_weight
;
1629 if (bfqq
->ioprio_class
< in_serv_bfqq
->ioprio_class
)
1632 if (in_serv_bfqq
->entity
.parent
== bfqq
->entity
.parent
) {
1633 bfqq_weight
= bfqq
->entity
.weight
;
1634 in_serv_weight
= in_serv_bfqq
->entity
.weight
;
1636 if (bfqq
->entity
.parent
)
1637 bfqq_weight
= bfqq
->entity
.parent
->weight
;
1639 bfqq_weight
= bfqq
->entity
.weight
;
1640 if (in_serv_bfqq
->entity
.parent
)
1641 in_serv_weight
= in_serv_bfqq
->entity
.parent
->weight
;
1643 in_serv_weight
= in_serv_bfqq
->entity
.weight
;
1646 return bfqq_weight
> in_serv_weight
;
1649 static void bfq_bfqq_handle_idle_busy_switch(struct bfq_data
*bfqd
,
1650 struct bfq_queue
*bfqq
,
1655 bool soft_rt
, in_burst
, wr_or_deserves_wr
,
1656 bfqq_wants_to_preempt
,
1657 idle_for_long_time
= bfq_bfqq_idle_for_long_time(bfqd
, bfqq
),
1659 * See the comments on
1660 * bfq_bfqq_update_budg_for_activation for
1661 * details on the usage of the next variable.
1663 arrived_in_time
= ktime_get_ns() <=
1664 bfqq
->ttime
.last_end_request
+
1665 bfqd
->bfq_slice_idle
* 3;
1669 * bfqq deserves to be weight-raised if:
1671 * - it does not belong to a large burst,
1672 * - it has been idle for enough time or is soft real-time,
1673 * - is linked to a bfq_io_cq (it is not shared in any sense).
1675 in_burst
= bfq_bfqq_in_large_burst(bfqq
);
1676 soft_rt
= bfqd
->bfq_wr_max_softrt_rate
> 0 &&
1677 !BFQQ_TOTALLY_SEEKY(bfqq
) &&
1679 time_is_before_jiffies(bfqq
->soft_rt_next_start
) &&
1680 bfqq
->dispatched
== 0;
1681 *interactive
= !in_burst
&& idle_for_long_time
;
1682 wr_or_deserves_wr
= bfqd
->low_latency
&&
1683 (bfqq
->wr_coeff
> 1 ||
1684 (bfq_bfqq_sync(bfqq
) &&
1685 bfqq
->bic
&& (*interactive
|| soft_rt
)));
1688 * Using the last flag, update budget and check whether bfqq
1689 * may want to preempt the in-service queue.
1691 bfqq_wants_to_preempt
=
1692 bfq_bfqq_update_budg_for_activation(bfqd
, bfqq
,
1696 * If bfqq happened to be activated in a burst, but has been
1697 * idle for much more than an interactive queue, then we
1698 * assume that, in the overall I/O initiated in the burst, the
1699 * I/O associated with bfqq is finished. So bfqq does not need
1700 * to be treated as a queue belonging to a burst
1701 * anymore. Accordingly, we reset bfqq's in_large_burst flag
1702 * if set, and remove bfqq from the burst list if it's
1703 * there. We do not decrement burst_size, because the fact
1704 * that bfqq does not need to belong to the burst list any
1705 * more does not invalidate the fact that bfqq was created in
1708 if (likely(!bfq_bfqq_just_created(bfqq
)) &&
1709 idle_for_long_time
&&
1710 time_is_before_jiffies(
1711 bfqq
->budget_timeout
+
1712 msecs_to_jiffies(10000))) {
1713 hlist_del_init(&bfqq
->burst_list_node
);
1714 bfq_clear_bfqq_in_large_burst(bfqq
);
1717 bfq_clear_bfqq_just_created(bfqq
);
1720 if (!bfq_bfqq_IO_bound(bfqq
)) {
1721 if (arrived_in_time
) {
1722 bfqq
->requests_within_timer
++;
1723 if (bfqq
->requests_within_timer
>=
1724 bfqd
->bfq_requests_within_timer
)
1725 bfq_mark_bfqq_IO_bound(bfqq
);
1727 bfqq
->requests_within_timer
= 0;
1730 if (bfqd
->low_latency
) {
1731 if (unlikely(time_is_after_jiffies(bfqq
->split_time
)))
1734 jiffies
- bfqd
->bfq_wr_min_idle_time
- 1;
1736 if (time_is_before_jiffies(bfqq
->split_time
+
1737 bfqd
->bfq_wr_min_idle_time
)) {
1738 bfq_update_bfqq_wr_on_rq_arrival(bfqd
, bfqq
,
1745 if (old_wr_coeff
!= bfqq
->wr_coeff
)
1746 bfqq
->entity
.prio_changed
= 1;
1750 bfqq
->last_idle_bklogged
= jiffies
;
1751 bfqq
->service_from_backlogged
= 0;
1752 bfq_clear_bfqq_softrt_update(bfqq
);
1754 bfq_add_bfqq_busy(bfqd
, bfqq
);
1757 * Expire in-service queue only if preemption may be needed
1758 * for guarantees. In particular, we care only about two
1759 * cases. The first is that bfqq has to recover a service
1760 * hole, as explained in the comments on
1761 * bfq_bfqq_update_budg_for_activation(), i.e., that
1762 * bfqq_wants_to_preempt is true. However, if bfqq does not
1763 * carry time-critical I/O, then bfqq's bandwidth is less
1764 * important than that of queues that carry time-critical I/O.
1765 * So, as a further constraint, we consider this case only if
1766 * bfqq is at least as weight-raised, i.e., at least as time
1767 * critical, as the in-service queue.
1769 * The second case is that bfqq is in a higher priority class,
1770 * or has a higher weight than the in-service queue. If this
1771 * condition does not hold, we don't care because, even if
1772 * bfqq does not start to be served immediately, the resulting
1773 * delay for bfqq's I/O is however lower or much lower than
1774 * the ideal completion time to be guaranteed to bfqq's I/O.
1776 * In both cases, preemption is needed only if, according to
1777 * the timestamps of both bfqq and of the in-service queue,
1778 * bfqq actually is the next queue to serve. So, to reduce
1779 * useless preemptions, the return value of
1780 * next_queue_may_preempt() is considered in the next compound
1781 * condition too. Yet next_queue_may_preempt() just checks a
1782 * simple, necessary condition for bfqq to be the next queue
1783 * to serve. In fact, to evaluate a sufficient condition, the
1784 * timestamps of the in-service queue would need to be
1785 * updated, and this operation is quite costly (see the
1786 * comments on bfq_bfqq_update_budg_for_activation()).
1788 if (bfqd
->in_service_queue
&&
1789 ((bfqq_wants_to_preempt
&&
1790 bfqq
->wr_coeff
>= bfqd
->in_service_queue
->wr_coeff
) ||
1791 bfq_bfqq_higher_class_or_weight(bfqq
, bfqd
->in_service_queue
)) &&
1792 next_queue_may_preempt(bfqd
))
1793 bfq_bfqq_expire(bfqd
, bfqd
->in_service_queue
,
1794 false, BFQQE_PREEMPTED
);
1797 static void bfq_reset_inject_limit(struct bfq_data
*bfqd
,
1798 struct bfq_queue
*bfqq
)
1800 /* invalidate baseline total service time */
1801 bfqq
->last_serv_time_ns
= 0;
1804 * Reset pointer in case we are waiting for
1805 * some request completion.
1807 bfqd
->waited_rq
= NULL
;
1810 * If bfqq has a short think time, then start by setting the
1811 * inject limit to 0 prudentially, because the service time of
1812 * an injected I/O request may be higher than the think time
1813 * of bfqq, and therefore, if one request was injected when
1814 * bfqq remains empty, this injected request might delay the
1815 * service of the next I/O request for bfqq significantly. In
1816 * case bfqq can actually tolerate some injection, then the
1817 * adaptive update will however raise the limit soon. This
1818 * lucky circumstance holds exactly because bfqq has a short
1819 * think time, and thus, after remaining empty, is likely to
1820 * get new I/O enqueued---and then completed---before being
1821 * expired. This is the very pattern that gives the
1822 * limit-update algorithm the chance to measure the effect of
1823 * injection on request service times, and then to update the
1824 * limit accordingly.
1826 * However, in the following special case, the inject limit is
1827 * left to 1 even if the think time is short: bfqq's I/O is
1828 * synchronized with that of some other queue, i.e., bfqq may
1829 * receive new I/O only after the I/O of the other queue is
1830 * completed. Keeping the inject limit to 1 allows the
1831 * blocking I/O to be served while bfqq is in service. And
1832 * this is very convenient both for bfqq and for overall
1833 * throughput, as explained in detail in the comments in
1834 * bfq_update_has_short_ttime().
1836 * On the opposite end, if bfqq has a long think time, then
1837 * start directly by 1, because:
1838 * a) on the bright side, keeping at most one request in
1839 * service in the drive is unlikely to cause any harm to the
1840 * latency of bfqq's requests, as the service time of a single
1841 * request is likely to be lower than the think time of bfqq;
1842 * b) on the downside, after becoming empty, bfqq is likely to
1843 * expire before getting its next request. With this request
1844 * arrival pattern, it is very hard to sample total service
1845 * times and update the inject limit accordingly (see comments
1846 * on bfq_update_inject_limit()). So the limit is likely to be
1847 * never, or at least seldom, updated. As a consequence, by
1848 * setting the limit to 1, we avoid that no injection ever
1849 * occurs with bfqq. On the downside, this proactive step
1850 * further reduces chances to actually compute the baseline
1851 * total service time. Thus it reduces chances to execute the
1852 * limit-update algorithm and possibly raise the limit to more
1855 if (bfq_bfqq_has_short_ttime(bfqq
))
1856 bfqq
->inject_limit
= 0;
1858 bfqq
->inject_limit
= 1;
1860 bfqq
->decrease_time_jif
= jiffies
;
1863 static void bfq_add_request(struct request
*rq
)
1865 struct bfq_queue
*bfqq
= RQ_BFQQ(rq
);
1866 struct bfq_data
*bfqd
= bfqq
->bfqd
;
1867 struct request
*next_rq
, *prev
;
1868 unsigned int old_wr_coeff
= bfqq
->wr_coeff
;
1869 bool interactive
= false;
1871 bfq_log_bfqq(bfqd
, bfqq
, "add_request %d", rq_is_sync(rq
));
1872 bfqq
->queued
[rq_is_sync(rq
)]++;
1875 if (RB_EMPTY_ROOT(&bfqq
->sort_list
) && bfq_bfqq_sync(bfqq
)) {
1877 * Detect whether bfqq's I/O seems synchronized with
1878 * that of some other queue, i.e., whether bfqq, after
1879 * remaining empty, happens to receive new I/O only
1880 * right after some I/O request of the other queue has
1881 * been completed. We call waker queue the other
1882 * queue, and we assume, for simplicity, that bfqq may
1883 * have at most one waker queue.
1885 * A remarkable throughput boost can be reached by
1886 * unconditionally injecting the I/O of the waker
1887 * queue, every time a new bfq_dispatch_request
1888 * happens to be invoked while I/O is being plugged
1889 * for bfqq. In addition to boosting throughput, this
1890 * unblocks bfqq's I/O, thereby improving bandwidth
1891 * and latency for bfqq. Note that these same results
1892 * may be achieved with the general injection
1893 * mechanism, but less effectively. For details on
1894 * this aspect, see the comments on the choice of the
1895 * queue for injection in bfq_select_queue().
1897 * Turning back to the detection of a waker queue, a
1898 * queue Q is deemed as a waker queue for bfqq if, for
1899 * two consecutive times, bfqq happens to become non
1900 * empty right after a request of Q has been
1901 * completed. In particular, on the first time, Q is
1902 * tentatively set as a candidate waker queue, while
1903 * on the second time, the flag
1904 * bfq_bfqq_has_waker(bfqq) is set to confirm that Q
1905 * is a waker queue for bfqq. These detection steps
1906 * are performed only if bfqq has a long think time,
1907 * so as to make it more likely that bfqq's I/O is
1908 * actually being blocked by a synchronization. This
1909 * last filter, plus the above two-times requirement,
1910 * make false positives less likely.
1914 * The sooner a waker queue is detected, the sooner
1915 * throughput can be boosted by injecting I/O from the
1916 * waker queue. Fortunately, detection is likely to be
1917 * actually fast, for the following reasons. While
1918 * blocked by synchronization, bfqq has a long think
1919 * time. This implies that bfqq's inject limit is at
1920 * least equal to 1 (see the comments in
1921 * bfq_update_inject_limit()). So, thanks to
1922 * injection, the waker queue is likely to be served
1923 * during the very first I/O-plugging time interval
1924 * for bfqq. This triggers the first step of the
1925 * detection mechanism. Thanks again to injection, the
1926 * candidate waker queue is then likely to be
1927 * confirmed no later than during the next
1928 * I/O-plugging interval for bfqq.
1930 if (bfqd
->last_completed_rq_bfqq
&&
1931 !bfq_bfqq_has_short_ttime(bfqq
) &&
1932 ktime_get_ns() - bfqd
->last_completion
<
1933 200 * NSEC_PER_USEC
) {
1934 if (bfqd
->last_completed_rq_bfqq
!= bfqq
&&
1935 bfqd
->last_completed_rq_bfqq
!=
1938 * First synchronization detected with
1939 * a candidate waker queue, or with a
1940 * different candidate waker queue
1941 * from the current one.
1943 bfqq
->waker_bfqq
= bfqd
->last_completed_rq_bfqq
;
1946 * If the waker queue disappears, then
1947 * bfqq->waker_bfqq must be reset. To
1948 * this goal, we maintain in each
1949 * waker queue a list, woken_list, of
1950 * all the queues that reference the
1951 * waker queue through their
1952 * waker_bfqq pointer. When the waker
1953 * queue exits, the waker_bfqq pointer
1954 * of all the queues in the woken_list
1957 * In addition, if bfqq is already in
1958 * the woken_list of a waker queue,
1959 * then, before being inserted into
1960 * the woken_list of a new waker
1961 * queue, bfqq must be removed from
1962 * the woken_list of the old waker
1965 if (!hlist_unhashed(&bfqq
->woken_list_node
))
1966 hlist_del_init(&bfqq
->woken_list_node
);
1967 hlist_add_head(&bfqq
->woken_list_node
,
1968 &bfqd
->last_completed_rq_bfqq
->woken_list
);
1970 bfq_clear_bfqq_has_waker(bfqq
);
1971 } else if (bfqd
->last_completed_rq_bfqq
==
1973 !bfq_bfqq_has_waker(bfqq
)) {
1975 * synchronization with waker_bfqq
1976 * seen for the second time
1978 bfq_mark_bfqq_has_waker(bfqq
);
1983 * Periodically reset inject limit, to make sure that
1984 * the latter eventually drops in case workload
1985 * changes, see step (3) in the comments on
1986 * bfq_update_inject_limit().
1988 if (time_is_before_eq_jiffies(bfqq
->decrease_time_jif
+
1989 msecs_to_jiffies(1000)))
1990 bfq_reset_inject_limit(bfqd
, bfqq
);
1993 * The following conditions must hold to setup a new
1994 * sampling of total service time, and then a new
1995 * update of the inject limit:
1996 * - bfqq is in service, because the total service
1997 * time is evaluated only for the I/O requests of
1998 * the queues in service;
1999 * - this is the right occasion to compute or to
2000 * lower the baseline total service time, because
2001 * there are actually no requests in the drive,
2003 * the baseline total service time is available, and
2004 * this is the right occasion to compute the other
2005 * quantity needed to update the inject limit, i.e.,
2006 * the total service time caused by the amount of
2007 * injection allowed by the current value of the
2008 * limit. It is the right occasion because injection
2009 * has actually been performed during the service
2010 * hole, and there are still in-flight requests,
2011 * which are very likely to be exactly the injected
2012 * requests, or part of them;
2013 * - the minimum interval for sampling the total
2014 * service time and updating the inject limit has
2017 if (bfqq
== bfqd
->in_service_queue
&&
2018 (bfqd
->rq_in_driver
== 0 ||
2019 (bfqq
->last_serv_time_ns
> 0 &&
2020 bfqd
->rqs_injected
&& bfqd
->rq_in_driver
> 0)) &&
2021 time_is_before_eq_jiffies(bfqq
->decrease_time_jif
+
2022 msecs_to_jiffies(10))) {
2023 bfqd
->last_empty_occupied_ns
= ktime_get_ns();
2025 * Start the state machine for measuring the
2026 * total service time of rq: setting
2027 * wait_dispatch will cause bfqd->waited_rq to
2028 * be set when rq will be dispatched.
2030 bfqd
->wait_dispatch
= true;
2032 * If there is no I/O in service in the drive,
2033 * then possible injection occurred before the
2034 * arrival of rq will not affect the total
2035 * service time of rq. So the injection limit
2036 * must not be updated as a function of such
2037 * total service time, unless new injection
2038 * occurs before rq is completed. To have the
2039 * injection limit updated only in the latter
2040 * case, reset rqs_injected here (rqs_injected
2041 * will be set in case injection is performed
2042 * on bfqq before rq is completed).
2044 if (bfqd
->rq_in_driver
== 0)
2045 bfqd
->rqs_injected
= false;
2049 elv_rb_add(&bfqq
->sort_list
, rq
);
2052 * Check if this request is a better next-serve candidate.
2054 prev
= bfqq
->next_rq
;
2055 next_rq
= bfq_choose_req(bfqd
, bfqq
->next_rq
, rq
, bfqd
->last_position
);
2056 bfqq
->next_rq
= next_rq
;
2059 * Adjust priority tree position, if next_rq changes.
2060 * See comments on bfq_pos_tree_add_move() for the unlikely().
2062 if (unlikely(!bfqd
->nonrot_with_queueing
&& prev
!= bfqq
->next_rq
))
2063 bfq_pos_tree_add_move(bfqd
, bfqq
);
2065 if (!bfq_bfqq_busy(bfqq
)) /* switching to busy ... */
2066 bfq_bfqq_handle_idle_busy_switch(bfqd
, bfqq
, old_wr_coeff
,
2069 if (bfqd
->low_latency
&& old_wr_coeff
== 1 && !rq_is_sync(rq
) &&
2070 time_is_before_jiffies(
2071 bfqq
->last_wr_start_finish
+
2072 bfqd
->bfq_wr_min_inter_arr_async
)) {
2073 bfqq
->wr_coeff
= bfqd
->bfq_wr_coeff
;
2074 bfqq
->wr_cur_max_time
= bfq_wr_duration(bfqd
);
2076 bfqd
->wr_busy_queues
++;
2077 bfqq
->entity
.prio_changed
= 1;
2079 if (prev
!= bfqq
->next_rq
)
2080 bfq_updated_next_req(bfqd
, bfqq
);
2084 * Assign jiffies to last_wr_start_finish in the following
2087 * . if bfqq is not going to be weight-raised, because, for
2088 * non weight-raised queues, last_wr_start_finish stores the
2089 * arrival time of the last request; as of now, this piece
2090 * of information is used only for deciding whether to
2091 * weight-raise async queues
2093 * . if bfqq is not weight-raised, because, if bfqq is now
2094 * switching to weight-raised, then last_wr_start_finish
2095 * stores the time when weight-raising starts
2097 * . if bfqq is interactive, because, regardless of whether
2098 * bfqq is currently weight-raised, the weight-raising
2099 * period must start or restart (this case is considered
2100 * separately because it is not detected by the above
2101 * conditions, if bfqq is already weight-raised)
2103 * last_wr_start_finish has to be updated also if bfqq is soft
2104 * real-time, because the weight-raising period is constantly
2105 * restarted on idle-to-busy transitions for these queues, but
2106 * this is already done in bfq_bfqq_handle_idle_busy_switch if
2109 if (bfqd
->low_latency
&&
2110 (old_wr_coeff
== 1 || bfqq
->wr_coeff
== 1 || interactive
))
2111 bfqq
->last_wr_start_finish
= jiffies
;
2114 static struct request
*bfq_find_rq_fmerge(struct bfq_data
*bfqd
,
2116 struct request_queue
*q
)
2118 struct bfq_queue
*bfqq
= bfqd
->bio_bfqq
;
2122 return elv_rb_find(&bfqq
->sort_list
, bio_end_sector(bio
));
2127 static sector_t
get_sdist(sector_t last_pos
, struct request
*rq
)
2130 return abs(blk_rq_pos(rq
) - last_pos
);
2135 #if 0 /* Still not clear if we can do without next two functions */
2136 static void bfq_activate_request(struct request_queue
*q
, struct request
*rq
)
2138 struct bfq_data
*bfqd
= q
->elevator
->elevator_data
;
2140 bfqd
->rq_in_driver
++;
2143 static void bfq_deactivate_request(struct request_queue
*q
, struct request
*rq
)
2145 struct bfq_data
*bfqd
= q
->elevator
->elevator_data
;
2147 bfqd
->rq_in_driver
--;
2151 static void bfq_remove_request(struct request_queue
*q
,
2154 struct bfq_queue
*bfqq
= RQ_BFQQ(rq
);
2155 struct bfq_data
*bfqd
= bfqq
->bfqd
;
2156 const int sync
= rq_is_sync(rq
);
2158 if (bfqq
->next_rq
== rq
) {
2159 bfqq
->next_rq
= bfq_find_next_rq(bfqd
, bfqq
, rq
);
2160 bfq_updated_next_req(bfqd
, bfqq
);
2163 if (rq
->queuelist
.prev
!= &rq
->queuelist
)
2164 list_del_init(&rq
->queuelist
);
2165 bfqq
->queued
[sync
]--;
2167 elv_rb_del(&bfqq
->sort_list
, rq
);
2169 elv_rqhash_del(q
, rq
);
2170 if (q
->last_merge
== rq
)
2171 q
->last_merge
= NULL
;
2173 if (RB_EMPTY_ROOT(&bfqq
->sort_list
)) {
2174 bfqq
->next_rq
= NULL
;
2176 if (bfq_bfqq_busy(bfqq
) && bfqq
!= bfqd
->in_service_queue
) {
2177 bfq_del_bfqq_busy(bfqd
, bfqq
, false);
2179 * bfqq emptied. In normal operation, when
2180 * bfqq is empty, bfqq->entity.service and
2181 * bfqq->entity.budget must contain,
2182 * respectively, the service received and the
2183 * budget used last time bfqq emptied. These
2184 * facts do not hold in this case, as at least
2185 * this last removal occurred while bfqq is
2186 * not in service. To avoid inconsistencies,
2187 * reset both bfqq->entity.service and
2188 * bfqq->entity.budget, if bfqq has still a
2189 * process that may issue I/O requests to it.
2191 bfqq
->entity
.budget
= bfqq
->entity
.service
= 0;
2195 * Remove queue from request-position tree as it is empty.
2197 if (bfqq
->pos_root
) {
2198 rb_erase(&bfqq
->pos_node
, bfqq
->pos_root
);
2199 bfqq
->pos_root
= NULL
;
2202 /* see comments on bfq_pos_tree_add_move() for the unlikely() */
2203 if (unlikely(!bfqd
->nonrot_with_queueing
))
2204 bfq_pos_tree_add_move(bfqd
, bfqq
);
2207 if (rq
->cmd_flags
& REQ_META
)
2208 bfqq
->meta_pending
--;
2212 static bool bfq_bio_merge(struct blk_mq_hw_ctx
*hctx
, struct bio
*bio
,
2213 unsigned int nr_segs
)
2215 struct request_queue
*q
= hctx
->queue
;
2216 struct bfq_data
*bfqd
= q
->elevator
->elevator_data
;
2217 struct request
*free
= NULL
;
2219 * bfq_bic_lookup grabs the queue_lock: invoke it now and
2220 * store its return value for later use, to avoid nesting
2221 * queue_lock inside the bfqd->lock. We assume that the bic
2222 * returned by bfq_bic_lookup does not go away before
2223 * bfqd->lock is taken.
2225 struct bfq_io_cq
*bic
= bfq_bic_lookup(bfqd
, current
->io_context
, q
);
2228 spin_lock_irq(&bfqd
->lock
);
2231 bfqd
->bio_bfqq
= bic_to_bfqq(bic
, op_is_sync(bio
->bi_opf
));
2233 bfqd
->bio_bfqq
= NULL
;
2234 bfqd
->bio_bic
= bic
;
2236 ret
= blk_mq_sched_try_merge(q
, bio
, nr_segs
, &free
);
2239 blk_mq_free_request(free
);
2240 spin_unlock_irq(&bfqd
->lock
);
2245 static int bfq_request_merge(struct request_queue
*q
, struct request
**req
,
2248 struct bfq_data
*bfqd
= q
->elevator
->elevator_data
;
2249 struct request
*__rq
;
2251 __rq
= bfq_find_rq_fmerge(bfqd
, bio
, q
);
2252 if (__rq
&& elv_bio_merge_ok(__rq
, bio
)) {
2254 return ELEVATOR_FRONT_MERGE
;
2257 return ELEVATOR_NO_MERGE
;
2260 static struct bfq_queue
*bfq_init_rq(struct request
*rq
);
2262 static void bfq_request_merged(struct request_queue
*q
, struct request
*req
,
2263 enum elv_merge type
)
2265 if (type
== ELEVATOR_FRONT_MERGE
&&
2266 rb_prev(&req
->rb_node
) &&
2268 blk_rq_pos(container_of(rb_prev(&req
->rb_node
),
2269 struct request
, rb_node
))) {
2270 struct bfq_queue
*bfqq
= bfq_init_rq(req
);
2271 struct bfq_data
*bfqd
;
2272 struct request
*prev
, *next_rq
;
2279 /* Reposition request in its sort_list */
2280 elv_rb_del(&bfqq
->sort_list
, req
);
2281 elv_rb_add(&bfqq
->sort_list
, req
);
2283 /* Choose next request to be served for bfqq */
2284 prev
= bfqq
->next_rq
;
2285 next_rq
= bfq_choose_req(bfqd
, bfqq
->next_rq
, req
,
2286 bfqd
->last_position
);
2287 bfqq
->next_rq
= next_rq
;
2289 * If next_rq changes, update both the queue's budget to
2290 * fit the new request and the queue's position in its
2293 if (prev
!= bfqq
->next_rq
) {
2294 bfq_updated_next_req(bfqd
, bfqq
);
2296 * See comments on bfq_pos_tree_add_move() for
2299 if (unlikely(!bfqd
->nonrot_with_queueing
))
2300 bfq_pos_tree_add_move(bfqd
, bfqq
);
2306 * This function is called to notify the scheduler that the requests
2307 * rq and 'next' have been merged, with 'next' going away. BFQ
2308 * exploits this hook to address the following issue: if 'next' has a
2309 * fifo_time lower that rq, then the fifo_time of rq must be set to
2310 * the value of 'next', to not forget the greater age of 'next'.
2312 * NOTE: in this function we assume that rq is in a bfq_queue, basing
2313 * on that rq is picked from the hash table q->elevator->hash, which,
2314 * in its turn, is filled only with I/O requests present in
2315 * bfq_queues, while BFQ is in use for the request queue q. In fact,
2316 * the function that fills this hash table (elv_rqhash_add) is called
2317 * only by bfq_insert_request.
2319 static void bfq_requests_merged(struct request_queue
*q
, struct request
*rq
,
2320 struct request
*next
)
2322 struct bfq_queue
*bfqq
= bfq_init_rq(rq
),
2323 *next_bfqq
= bfq_init_rq(next
);
2329 * If next and rq belong to the same bfq_queue and next is older
2330 * than rq, then reposition rq in the fifo (by substituting next
2331 * with rq). Otherwise, if next and rq belong to different
2332 * bfq_queues, never reposition rq: in fact, we would have to
2333 * reposition it with respect to next's position in its own fifo,
2334 * which would most certainly be too expensive with respect to
2337 if (bfqq
== next_bfqq
&&
2338 !list_empty(&rq
->queuelist
) && !list_empty(&next
->queuelist
) &&
2339 next
->fifo_time
< rq
->fifo_time
) {
2340 list_del_init(&rq
->queuelist
);
2341 list_replace_init(&next
->queuelist
, &rq
->queuelist
);
2342 rq
->fifo_time
= next
->fifo_time
;
2345 if (bfqq
->next_rq
== next
)
2348 bfqg_stats_update_io_merged(bfqq_group(bfqq
), next
->cmd_flags
);
2351 /* Must be called with bfqq != NULL */
2352 static void bfq_bfqq_end_wr(struct bfq_queue
*bfqq
)
2354 if (bfq_bfqq_busy(bfqq
))
2355 bfqq
->bfqd
->wr_busy_queues
--;
2357 bfqq
->wr_cur_max_time
= 0;
2358 bfqq
->last_wr_start_finish
= jiffies
;
2360 * Trigger a weight change on the next invocation of
2361 * __bfq_entity_update_weight_prio.
2363 bfqq
->entity
.prio_changed
= 1;
2366 void bfq_end_wr_async_queues(struct bfq_data
*bfqd
,
2367 struct bfq_group
*bfqg
)
2371 for (i
= 0; i
< 2; i
++)
2372 for (j
= 0; j
< IOPRIO_BE_NR
; j
++)
2373 if (bfqg
->async_bfqq
[i
][j
])
2374 bfq_bfqq_end_wr(bfqg
->async_bfqq
[i
][j
]);
2375 if (bfqg
->async_idle_bfqq
)
2376 bfq_bfqq_end_wr(bfqg
->async_idle_bfqq
);
2379 static void bfq_end_wr(struct bfq_data
*bfqd
)
2381 struct bfq_queue
*bfqq
;
2383 spin_lock_irq(&bfqd
->lock
);
2385 list_for_each_entry(bfqq
, &bfqd
->active_list
, bfqq_list
)
2386 bfq_bfqq_end_wr(bfqq
);
2387 list_for_each_entry(bfqq
, &bfqd
->idle_list
, bfqq_list
)
2388 bfq_bfqq_end_wr(bfqq
);
2389 bfq_end_wr_async(bfqd
);
2391 spin_unlock_irq(&bfqd
->lock
);
2394 static sector_t
bfq_io_struct_pos(void *io_struct
, bool request
)
2397 return blk_rq_pos(io_struct
);
2399 return ((struct bio
*)io_struct
)->bi_iter
.bi_sector
;
2402 static int bfq_rq_close_to_sector(void *io_struct
, bool request
,
2405 return abs(bfq_io_struct_pos(io_struct
, request
) - sector
) <=
2409 static struct bfq_queue
*bfqq_find_close(struct bfq_data
*bfqd
,
2410 struct bfq_queue
*bfqq
,
2413 struct rb_root
*root
= &bfq_bfqq_to_bfqg(bfqq
)->rq_pos_tree
;
2414 struct rb_node
*parent
, *node
;
2415 struct bfq_queue
*__bfqq
;
2417 if (RB_EMPTY_ROOT(root
))
2421 * First, if we find a request starting at the end of the last
2422 * request, choose it.
2424 __bfqq
= bfq_rq_pos_tree_lookup(bfqd
, root
, sector
, &parent
, NULL
);
2429 * If the exact sector wasn't found, the parent of the NULL leaf
2430 * will contain the closest sector (rq_pos_tree sorted by
2431 * next_request position).
2433 __bfqq
= rb_entry(parent
, struct bfq_queue
, pos_node
);
2434 if (bfq_rq_close_to_sector(__bfqq
->next_rq
, true, sector
))
2437 if (blk_rq_pos(__bfqq
->next_rq
) < sector
)
2438 node
= rb_next(&__bfqq
->pos_node
);
2440 node
= rb_prev(&__bfqq
->pos_node
);
2444 __bfqq
= rb_entry(node
, struct bfq_queue
, pos_node
);
2445 if (bfq_rq_close_to_sector(__bfqq
->next_rq
, true, sector
))
2451 static struct bfq_queue
*bfq_find_close_cooperator(struct bfq_data
*bfqd
,
2452 struct bfq_queue
*cur_bfqq
,
2455 struct bfq_queue
*bfqq
;
2458 * We shall notice if some of the queues are cooperating,
2459 * e.g., working closely on the same area of the device. In
2460 * that case, we can group them together and: 1) don't waste
2461 * time idling, and 2) serve the union of their requests in
2462 * the best possible order for throughput.
2464 bfqq
= bfqq_find_close(bfqd
, cur_bfqq
, sector
);
2465 if (!bfqq
|| bfqq
== cur_bfqq
)
2471 static struct bfq_queue
*
2472 bfq_setup_merge(struct bfq_queue
*bfqq
, struct bfq_queue
*new_bfqq
)
2474 int process_refs
, new_process_refs
;
2475 struct bfq_queue
*__bfqq
;
2478 * If there are no process references on the new_bfqq, then it is
2479 * unsafe to follow the ->new_bfqq chain as other bfqq's in the chain
2480 * may have dropped their last reference (not just their last process
2483 if (!bfqq_process_refs(new_bfqq
))
2486 /* Avoid a circular list and skip interim queue merges. */
2487 while ((__bfqq
= new_bfqq
->new_bfqq
)) {
2493 process_refs
= bfqq_process_refs(bfqq
);
2494 new_process_refs
= bfqq_process_refs(new_bfqq
);
2496 * If the process for the bfqq has gone away, there is no
2497 * sense in merging the queues.
2499 if (process_refs
== 0 || new_process_refs
== 0)
2502 bfq_log_bfqq(bfqq
->bfqd
, bfqq
, "scheduling merge with queue %d",
2506 * Merging is just a redirection: the requests of the process
2507 * owning one of the two queues are redirected to the other queue.
2508 * The latter queue, in its turn, is set as shared if this is the
2509 * first time that the requests of some process are redirected to
2512 * We redirect bfqq to new_bfqq and not the opposite, because
2513 * we are in the context of the process owning bfqq, thus we
2514 * have the io_cq of this process. So we can immediately
2515 * configure this io_cq to redirect the requests of the
2516 * process to new_bfqq. In contrast, the io_cq of new_bfqq is
2517 * not available any more (new_bfqq->bic == NULL).
2519 * Anyway, even in case new_bfqq coincides with the in-service
2520 * queue, redirecting requests the in-service queue is the
2521 * best option, as we feed the in-service queue with new
2522 * requests close to the last request served and, by doing so,
2523 * are likely to increase the throughput.
2525 bfqq
->new_bfqq
= new_bfqq
;
2526 new_bfqq
->ref
+= process_refs
;
2530 static bool bfq_may_be_close_cooperator(struct bfq_queue
*bfqq
,
2531 struct bfq_queue
*new_bfqq
)
2533 if (bfq_too_late_for_merging(new_bfqq
))
2536 if (bfq_class_idle(bfqq
) || bfq_class_idle(new_bfqq
) ||
2537 (bfqq
->ioprio_class
!= new_bfqq
->ioprio_class
))
2541 * If either of the queues has already been detected as seeky,
2542 * then merging it with the other queue is unlikely to lead to
2545 if (BFQQ_SEEKY(bfqq
) || BFQQ_SEEKY(new_bfqq
))
2549 * Interleaved I/O is known to be done by (some) applications
2550 * only for reads, so it does not make sense to merge async
2553 if (!bfq_bfqq_sync(bfqq
) || !bfq_bfqq_sync(new_bfqq
))
2560 * Attempt to schedule a merge of bfqq with the currently in-service
2561 * queue or with a close queue among the scheduled queues. Return
2562 * NULL if no merge was scheduled, a pointer to the shared bfq_queue
2563 * structure otherwise.
2565 * The OOM queue is not allowed to participate to cooperation: in fact, since
2566 * the requests temporarily redirected to the OOM queue could be redirected
2567 * again to dedicated queues at any time, the state needed to correctly
2568 * handle merging with the OOM queue would be quite complex and expensive
2569 * to maintain. Besides, in such a critical condition as an out of memory,
2570 * the benefits of queue merging may be little relevant, or even negligible.
2572 * WARNING: queue merging may impair fairness among non-weight raised
2573 * queues, for at least two reasons: 1) the original weight of a
2574 * merged queue may change during the merged state, 2) even being the
2575 * weight the same, a merged queue may be bloated with many more
2576 * requests than the ones produced by its originally-associated
2579 static struct bfq_queue
*
2580 bfq_setup_cooperator(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
,
2581 void *io_struct
, bool request
)
2583 struct bfq_queue
*in_service_bfqq
, *new_bfqq
;
2586 * Do not perform queue merging if the device is non
2587 * rotational and performs internal queueing. In fact, such a
2588 * device reaches a high speed through internal parallelism
2589 * and pipelining. This means that, to reach a high
2590 * throughput, it must have many requests enqueued at the same
2591 * time. But, in this configuration, the internal scheduling
2592 * algorithm of the device does exactly the job of queue
2593 * merging: it reorders requests so as to obtain as much as
2594 * possible a sequential I/O pattern. As a consequence, with
2595 * the workload generated by processes doing interleaved I/O,
2596 * the throughput reached by the device is likely to be the
2597 * same, with and without queue merging.
2599 * Disabling merging also provides a remarkable benefit in
2600 * terms of throughput. Merging tends to make many workloads
2601 * artificially more uneven, because of shared queues
2602 * remaining non empty for incomparably more time than
2603 * non-merged queues. This may accentuate workload
2604 * asymmetries. For example, if one of the queues in a set of
2605 * merged queues has a higher weight than a normal queue, then
2606 * the shared queue may inherit such a high weight and, by
2607 * staying almost always active, may force BFQ to perform I/O
2608 * plugging most of the time. This evidently makes it harder
2609 * for BFQ to let the device reach a high throughput.
2611 * Finally, the likely() macro below is not used because one
2612 * of the two branches is more likely than the other, but to
2613 * have the code path after the following if() executed as
2614 * fast as possible for the case of a non rotational device
2615 * with queueing. We want it because this is the fastest kind
2616 * of device. On the opposite end, the likely() may lengthen
2617 * the execution time of BFQ for the case of slower devices
2618 * (rotational or at least without queueing). But in this case
2619 * the execution time of BFQ matters very little, if not at
2622 if (likely(bfqd
->nonrot_with_queueing
))
2626 * Prevent bfqq from being merged if it has been created too
2627 * long ago. The idea is that true cooperating processes, and
2628 * thus their associated bfq_queues, are supposed to be
2629 * created shortly after each other. This is the case, e.g.,
2630 * for KVM/QEMU and dump I/O threads. Basing on this
2631 * assumption, the following filtering greatly reduces the
2632 * probability that two non-cooperating processes, which just
2633 * happen to do close I/O for some short time interval, have
2634 * their queues merged by mistake.
2636 if (bfq_too_late_for_merging(bfqq
))
2640 return bfqq
->new_bfqq
;
2642 if (!io_struct
|| unlikely(bfqq
== &bfqd
->oom_bfqq
))
2645 /* If there is only one backlogged queue, don't search. */
2646 if (bfq_tot_busy_queues(bfqd
) == 1)
2649 in_service_bfqq
= bfqd
->in_service_queue
;
2651 if (in_service_bfqq
&& in_service_bfqq
!= bfqq
&&
2652 likely(in_service_bfqq
!= &bfqd
->oom_bfqq
) &&
2653 bfq_rq_close_to_sector(io_struct
, request
,
2654 bfqd
->in_serv_last_pos
) &&
2655 bfqq
->entity
.parent
== in_service_bfqq
->entity
.parent
&&
2656 bfq_may_be_close_cooperator(bfqq
, in_service_bfqq
)) {
2657 new_bfqq
= bfq_setup_merge(bfqq
, in_service_bfqq
);
2662 * Check whether there is a cooperator among currently scheduled
2663 * queues. The only thing we need is that the bio/request is not
2664 * NULL, as we need it to establish whether a cooperator exists.
2666 new_bfqq
= bfq_find_close_cooperator(bfqd
, bfqq
,
2667 bfq_io_struct_pos(io_struct
, request
));
2669 if (new_bfqq
&& likely(new_bfqq
!= &bfqd
->oom_bfqq
) &&
2670 bfq_may_be_close_cooperator(bfqq
, new_bfqq
))
2671 return bfq_setup_merge(bfqq
, new_bfqq
);
2676 static void bfq_bfqq_save_state(struct bfq_queue
*bfqq
)
2678 struct bfq_io_cq
*bic
= bfqq
->bic
;
2681 * If !bfqq->bic, the queue is already shared or its requests
2682 * have already been redirected to a shared queue; both idle window
2683 * and weight raising state have already been saved. Do nothing.
2688 bic
->saved_weight
= bfqq
->entity
.orig_weight
;
2689 bic
->saved_ttime
= bfqq
->ttime
;
2690 bic
->saved_has_short_ttime
= bfq_bfqq_has_short_ttime(bfqq
);
2691 bic
->saved_IO_bound
= bfq_bfqq_IO_bound(bfqq
);
2692 bic
->saved_in_large_burst
= bfq_bfqq_in_large_burst(bfqq
);
2693 bic
->was_in_burst_list
= !hlist_unhashed(&bfqq
->burst_list_node
);
2694 if (unlikely(bfq_bfqq_just_created(bfqq
) &&
2695 !bfq_bfqq_in_large_burst(bfqq
) &&
2696 bfqq
->bfqd
->low_latency
)) {
2698 * bfqq being merged right after being created: bfqq
2699 * would have deserved interactive weight raising, but
2700 * did not make it to be set in a weight-raised state,
2701 * because of this early merge. Store directly the
2702 * weight-raising state that would have been assigned
2703 * to bfqq, so that to avoid that bfqq unjustly fails
2704 * to enjoy weight raising if split soon.
2706 bic
->saved_wr_coeff
= bfqq
->bfqd
->bfq_wr_coeff
;
2707 bic
->saved_wr_start_at_switch_to_srt
= bfq_smallest_from_now();
2708 bic
->saved_wr_cur_max_time
= bfq_wr_duration(bfqq
->bfqd
);
2709 bic
->saved_last_wr_start_finish
= jiffies
;
2711 bic
->saved_wr_coeff
= bfqq
->wr_coeff
;
2712 bic
->saved_wr_start_at_switch_to_srt
=
2713 bfqq
->wr_start_at_switch_to_srt
;
2714 bic
->saved_last_wr_start_finish
= bfqq
->last_wr_start_finish
;
2715 bic
->saved_wr_cur_max_time
= bfqq
->wr_cur_max_time
;
2719 void bfq_release_process_ref(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
)
2722 * To prevent bfqq's service guarantees from being violated,
2723 * bfqq may be left busy, i.e., queued for service, even if
2724 * empty (see comments in __bfq_bfqq_expire() for
2725 * details). But, if no process will send requests to bfqq any
2726 * longer, then there is no point in keeping bfqq queued for
2727 * service. In addition, keeping bfqq queued for service, but
2728 * with no process ref any longer, may have caused bfqq to be
2729 * freed when dequeued from service. But this is assumed to
2732 if (bfq_bfqq_busy(bfqq
) && RB_EMPTY_ROOT(&bfqq
->sort_list
) &&
2733 bfqq
!= bfqd
->in_service_queue
)
2734 bfq_del_bfqq_busy(bfqd
, bfqq
, false);
2736 bfq_put_queue(bfqq
);
2740 bfq_merge_bfqqs(struct bfq_data
*bfqd
, struct bfq_io_cq
*bic
,
2741 struct bfq_queue
*bfqq
, struct bfq_queue
*new_bfqq
)
2743 bfq_log_bfqq(bfqd
, bfqq
, "merging with queue %lu",
2744 (unsigned long)new_bfqq
->pid
);
2745 /* Save weight raising and idle window of the merged queues */
2746 bfq_bfqq_save_state(bfqq
);
2747 bfq_bfqq_save_state(new_bfqq
);
2748 if (bfq_bfqq_IO_bound(bfqq
))
2749 bfq_mark_bfqq_IO_bound(new_bfqq
);
2750 bfq_clear_bfqq_IO_bound(bfqq
);
2753 * If bfqq is weight-raised, then let new_bfqq inherit
2754 * weight-raising. To reduce false positives, neglect the case
2755 * where bfqq has just been created, but has not yet made it
2756 * to be weight-raised (which may happen because EQM may merge
2757 * bfqq even before bfq_add_request is executed for the first
2758 * time for bfqq). Handling this case would however be very
2759 * easy, thanks to the flag just_created.
2761 if (new_bfqq
->wr_coeff
== 1 && bfqq
->wr_coeff
> 1) {
2762 new_bfqq
->wr_coeff
= bfqq
->wr_coeff
;
2763 new_bfqq
->wr_cur_max_time
= bfqq
->wr_cur_max_time
;
2764 new_bfqq
->last_wr_start_finish
= bfqq
->last_wr_start_finish
;
2765 new_bfqq
->wr_start_at_switch_to_srt
=
2766 bfqq
->wr_start_at_switch_to_srt
;
2767 if (bfq_bfqq_busy(new_bfqq
))
2768 bfqd
->wr_busy_queues
++;
2769 new_bfqq
->entity
.prio_changed
= 1;
2772 if (bfqq
->wr_coeff
> 1) { /* bfqq has given its wr to new_bfqq */
2774 bfqq
->entity
.prio_changed
= 1;
2775 if (bfq_bfqq_busy(bfqq
))
2776 bfqd
->wr_busy_queues
--;
2779 bfq_log_bfqq(bfqd
, new_bfqq
, "merge_bfqqs: wr_busy %d",
2780 bfqd
->wr_busy_queues
);
2783 * Merge queues (that is, let bic redirect its requests to new_bfqq)
2785 bic_set_bfqq(bic
, new_bfqq
, 1);
2786 bfq_mark_bfqq_coop(new_bfqq
);
2788 * new_bfqq now belongs to at least two bics (it is a shared queue):
2789 * set new_bfqq->bic to NULL. bfqq either:
2790 * - does not belong to any bic any more, and hence bfqq->bic must
2791 * be set to NULL, or
2792 * - is a queue whose owning bics have already been redirected to a
2793 * different queue, hence the queue is destined to not belong to
2794 * any bic soon and bfqq->bic is already NULL (therefore the next
2795 * assignment causes no harm).
2797 new_bfqq
->bic
= NULL
;
2799 * If the queue is shared, the pid is the pid of one of the associated
2800 * processes. Which pid depends on the exact sequence of merge events
2801 * the queue underwent. So printing such a pid is useless and confusing
2802 * because it reports a random pid between those of the associated
2804 * We mark such a queue with a pid -1, and then print SHARED instead of
2805 * a pid in logging messages.
2809 bfq_release_process_ref(bfqd
, bfqq
);
2812 static bool bfq_allow_bio_merge(struct request_queue
*q
, struct request
*rq
,
2815 struct bfq_data
*bfqd
= q
->elevator
->elevator_data
;
2816 bool is_sync
= op_is_sync(bio
->bi_opf
);
2817 struct bfq_queue
*bfqq
= bfqd
->bio_bfqq
, *new_bfqq
;
2820 * Disallow merge of a sync bio into an async request.
2822 if (is_sync
&& !rq_is_sync(rq
))
2826 * Lookup the bfqq that this bio will be queued with. Allow
2827 * merge only if rq is queued there.
2833 * We take advantage of this function to perform an early merge
2834 * of the queues of possible cooperating processes.
2836 new_bfqq
= bfq_setup_cooperator(bfqd
, bfqq
, bio
, false);
2839 * bic still points to bfqq, then it has not yet been
2840 * redirected to some other bfq_queue, and a queue
2841 * merge between bfqq and new_bfqq can be safely
2842 * fulfilled, i.e., bic can be redirected to new_bfqq
2843 * and bfqq can be put.
2845 bfq_merge_bfqqs(bfqd
, bfqd
->bio_bic
, bfqq
,
2848 * If we get here, bio will be queued into new_queue,
2849 * so use new_bfqq to decide whether bio and rq can be
2855 * Change also bqfd->bio_bfqq, as
2856 * bfqd->bio_bic now points to new_bfqq, and
2857 * this function may be invoked again (and then may
2858 * use again bqfd->bio_bfqq).
2860 bfqd
->bio_bfqq
= bfqq
;
2863 return bfqq
== RQ_BFQQ(rq
);
2867 * Set the maximum time for the in-service queue to consume its
2868 * budget. This prevents seeky processes from lowering the throughput.
2869 * In practice, a time-slice service scheme is used with seeky
2872 static void bfq_set_budget_timeout(struct bfq_data
*bfqd
,
2873 struct bfq_queue
*bfqq
)
2875 unsigned int timeout_coeff
;
2877 if (bfqq
->wr_cur_max_time
== bfqd
->bfq_wr_rt_max_time
)
2880 timeout_coeff
= bfqq
->entity
.weight
/ bfqq
->entity
.orig_weight
;
2882 bfqd
->last_budget_start
= ktime_get();
2884 bfqq
->budget_timeout
= jiffies
+
2885 bfqd
->bfq_timeout
* timeout_coeff
;
2888 static void __bfq_set_in_service_queue(struct bfq_data
*bfqd
,
2889 struct bfq_queue
*bfqq
)
2892 bfq_clear_bfqq_fifo_expire(bfqq
);
2894 bfqd
->budgets_assigned
= (bfqd
->budgets_assigned
* 7 + 256) / 8;
2896 if (time_is_before_jiffies(bfqq
->last_wr_start_finish
) &&
2897 bfqq
->wr_coeff
> 1 &&
2898 bfqq
->wr_cur_max_time
== bfqd
->bfq_wr_rt_max_time
&&
2899 time_is_before_jiffies(bfqq
->budget_timeout
)) {
2901 * For soft real-time queues, move the start
2902 * of the weight-raising period forward by the
2903 * time the queue has not received any
2904 * service. Otherwise, a relatively long
2905 * service delay is likely to cause the
2906 * weight-raising period of the queue to end,
2907 * because of the short duration of the
2908 * weight-raising period of a soft real-time
2909 * queue. It is worth noting that this move
2910 * is not so dangerous for the other queues,
2911 * because soft real-time queues are not
2914 * To not add a further variable, we use the
2915 * overloaded field budget_timeout to
2916 * determine for how long the queue has not
2917 * received service, i.e., how much time has
2918 * elapsed since the queue expired. However,
2919 * this is a little imprecise, because
2920 * budget_timeout is set to jiffies if bfqq
2921 * not only expires, but also remains with no
2924 if (time_after(bfqq
->budget_timeout
,
2925 bfqq
->last_wr_start_finish
))
2926 bfqq
->last_wr_start_finish
+=
2927 jiffies
- bfqq
->budget_timeout
;
2929 bfqq
->last_wr_start_finish
= jiffies
;
2932 bfq_set_budget_timeout(bfqd
, bfqq
);
2933 bfq_log_bfqq(bfqd
, bfqq
,
2934 "set_in_service_queue, cur-budget = %d",
2935 bfqq
->entity
.budget
);
2938 bfqd
->in_service_queue
= bfqq
;
2942 * Get and set a new queue for service.
2944 static struct bfq_queue
*bfq_set_in_service_queue(struct bfq_data
*bfqd
)
2946 struct bfq_queue
*bfqq
= bfq_get_next_queue(bfqd
);
2948 __bfq_set_in_service_queue(bfqd
, bfqq
);
2952 static void bfq_arm_slice_timer(struct bfq_data
*bfqd
)
2954 struct bfq_queue
*bfqq
= bfqd
->in_service_queue
;
2957 bfq_mark_bfqq_wait_request(bfqq
);
2960 * We don't want to idle for seeks, but we do want to allow
2961 * fair distribution of slice time for a process doing back-to-back
2962 * seeks. So allow a little bit of time for him to submit a new rq.
2964 sl
= bfqd
->bfq_slice_idle
;
2966 * Unless the queue is being weight-raised or the scenario is
2967 * asymmetric, grant only minimum idle time if the queue
2968 * is seeky. A long idling is preserved for a weight-raised
2969 * queue, or, more in general, in an asymmetric scenario,
2970 * because a long idling is needed for guaranteeing to a queue
2971 * its reserved share of the throughput (in particular, it is
2972 * needed if the queue has a higher weight than some other
2975 if (BFQQ_SEEKY(bfqq
) && bfqq
->wr_coeff
== 1 &&
2976 !bfq_asymmetric_scenario(bfqd
, bfqq
))
2977 sl
= min_t(u64
, sl
, BFQ_MIN_TT
);
2978 else if (bfqq
->wr_coeff
> 1)
2979 sl
= max_t(u32
, sl
, 20ULL * NSEC_PER_MSEC
);
2981 bfqd
->last_idling_start
= ktime_get();
2982 bfqd
->last_idling_start_jiffies
= jiffies
;
2984 hrtimer_start(&bfqd
->idle_slice_timer
, ns_to_ktime(sl
),
2986 bfqg_stats_set_start_idle_time(bfqq_group(bfqq
));
2990 * In autotuning mode, max_budget is dynamically recomputed as the
2991 * amount of sectors transferred in timeout at the estimated peak
2992 * rate. This enables BFQ to utilize a full timeslice with a full
2993 * budget, even if the in-service queue is served at peak rate. And
2994 * this maximises throughput with sequential workloads.
2996 static unsigned long bfq_calc_max_budget(struct bfq_data
*bfqd
)
2998 return (u64
)bfqd
->peak_rate
* USEC_PER_MSEC
*
2999 jiffies_to_msecs(bfqd
->bfq_timeout
)>>BFQ_RATE_SHIFT
;
3003 * Update parameters related to throughput and responsiveness, as a
3004 * function of the estimated peak rate. See comments on
3005 * bfq_calc_max_budget(), and on the ref_wr_duration array.
3007 static void update_thr_responsiveness_params(struct bfq_data
*bfqd
)
3009 if (bfqd
->bfq_user_max_budget
== 0) {
3010 bfqd
->bfq_max_budget
=
3011 bfq_calc_max_budget(bfqd
);
3012 bfq_log(bfqd
, "new max_budget = %d", bfqd
->bfq_max_budget
);
3016 static void bfq_reset_rate_computation(struct bfq_data
*bfqd
,
3019 if (rq
!= NULL
) { /* new rq dispatch now, reset accordingly */
3020 bfqd
->last_dispatch
= bfqd
->first_dispatch
= ktime_get_ns();
3021 bfqd
->peak_rate_samples
= 1;
3022 bfqd
->sequential_samples
= 0;
3023 bfqd
->tot_sectors_dispatched
= bfqd
->last_rq_max_size
=
3025 } else /* no new rq dispatched, just reset the number of samples */
3026 bfqd
->peak_rate_samples
= 0; /* full re-init on next disp. */
3029 "reset_rate_computation at end, sample %u/%u tot_sects %llu",
3030 bfqd
->peak_rate_samples
, bfqd
->sequential_samples
,
3031 bfqd
->tot_sectors_dispatched
);
3034 static void bfq_update_rate_reset(struct bfq_data
*bfqd
, struct request
*rq
)
3036 u32 rate
, weight
, divisor
;
3039 * For the convergence property to hold (see comments on
3040 * bfq_update_peak_rate()) and for the assessment to be
3041 * reliable, a minimum number of samples must be present, and
3042 * a minimum amount of time must have elapsed. If not so, do
3043 * not compute new rate. Just reset parameters, to get ready
3044 * for a new evaluation attempt.
3046 if (bfqd
->peak_rate_samples
< BFQ_RATE_MIN_SAMPLES
||
3047 bfqd
->delta_from_first
< BFQ_RATE_MIN_INTERVAL
)
3048 goto reset_computation
;
3051 * If a new request completion has occurred after last
3052 * dispatch, then, to approximate the rate at which requests
3053 * have been served by the device, it is more precise to
3054 * extend the observation interval to the last completion.
3056 bfqd
->delta_from_first
=
3057 max_t(u64
, bfqd
->delta_from_first
,
3058 bfqd
->last_completion
- bfqd
->first_dispatch
);
3061 * Rate computed in sects/usec, and not sects/nsec, for
3064 rate
= div64_ul(bfqd
->tot_sectors_dispatched
<<BFQ_RATE_SHIFT
,
3065 div_u64(bfqd
->delta_from_first
, NSEC_PER_USEC
));
3068 * Peak rate not updated if:
3069 * - the percentage of sequential dispatches is below 3/4 of the
3070 * total, and rate is below the current estimated peak rate
3071 * - rate is unreasonably high (> 20M sectors/sec)
3073 if ((bfqd
->sequential_samples
< (3 * bfqd
->peak_rate_samples
)>>2 &&
3074 rate
<= bfqd
->peak_rate
) ||
3075 rate
> 20<<BFQ_RATE_SHIFT
)
3076 goto reset_computation
;
3079 * We have to update the peak rate, at last! To this purpose,
3080 * we use a low-pass filter. We compute the smoothing constant
3081 * of the filter as a function of the 'weight' of the new
3084 * As can be seen in next formulas, we define this weight as a
3085 * quantity proportional to how sequential the workload is,
3086 * and to how long the observation time interval is.
3088 * The weight runs from 0 to 8. The maximum value of the
3089 * weight, 8, yields the minimum value for the smoothing
3090 * constant. At this minimum value for the smoothing constant,
3091 * the measured rate contributes for half of the next value of
3092 * the estimated peak rate.
3094 * So, the first step is to compute the weight as a function
3095 * of how sequential the workload is. Note that the weight
3096 * cannot reach 9, because bfqd->sequential_samples cannot
3097 * become equal to bfqd->peak_rate_samples, which, in its
3098 * turn, holds true because bfqd->sequential_samples is not
3099 * incremented for the first sample.
3101 weight
= (9 * bfqd
->sequential_samples
) / bfqd
->peak_rate_samples
;
3104 * Second step: further refine the weight as a function of the
3105 * duration of the observation interval.
3107 weight
= min_t(u32
, 8,
3108 div_u64(weight
* bfqd
->delta_from_first
,
3109 BFQ_RATE_REF_INTERVAL
));
3112 * Divisor ranging from 10, for minimum weight, to 2, for
3115 divisor
= 10 - weight
;
3118 * Finally, update peak rate:
3120 * peak_rate = peak_rate * (divisor-1) / divisor + rate / divisor
3122 bfqd
->peak_rate
*= divisor
-1;
3123 bfqd
->peak_rate
/= divisor
;
3124 rate
/= divisor
; /* smoothing constant alpha = 1/divisor */
3126 bfqd
->peak_rate
+= rate
;
3129 * For a very slow device, bfqd->peak_rate can reach 0 (see
3130 * the minimum representable values reported in the comments
3131 * on BFQ_RATE_SHIFT). Push to 1 if this happens, to avoid
3132 * divisions by zero where bfqd->peak_rate is used as a
3135 bfqd
->peak_rate
= max_t(u32
, 1, bfqd
->peak_rate
);
3137 update_thr_responsiveness_params(bfqd
);
3140 bfq_reset_rate_computation(bfqd
, rq
);
3144 * Update the read/write peak rate (the main quantity used for
3145 * auto-tuning, see update_thr_responsiveness_params()).
3147 * It is not trivial to estimate the peak rate (correctly): because of
3148 * the presence of sw and hw queues between the scheduler and the
3149 * device components that finally serve I/O requests, it is hard to
3150 * say exactly when a given dispatched request is served inside the
3151 * device, and for how long. As a consequence, it is hard to know
3152 * precisely at what rate a given set of requests is actually served
3155 * On the opposite end, the dispatch time of any request is trivially
3156 * available, and, from this piece of information, the "dispatch rate"
3157 * of requests can be immediately computed. So, the idea in the next
3158 * function is to use what is known, namely request dispatch times
3159 * (plus, when useful, request completion times), to estimate what is
3160 * unknown, namely in-device request service rate.
3162 * The main issue is that, because of the above facts, the rate at
3163 * which a certain set of requests is dispatched over a certain time
3164 * interval can vary greatly with respect to the rate at which the
3165 * same requests are then served. But, since the size of any
3166 * intermediate queue is limited, and the service scheme is lossless
3167 * (no request is silently dropped), the following obvious convergence
3168 * property holds: the number of requests dispatched MUST become
3169 * closer and closer to the number of requests completed as the
3170 * observation interval grows. This is the key property used in
3171 * the next function to estimate the peak service rate as a function
3172 * of the observed dispatch rate. The function assumes to be invoked
3173 * on every request dispatch.
3175 static void bfq_update_peak_rate(struct bfq_data
*bfqd
, struct request
*rq
)
3177 u64 now_ns
= ktime_get_ns();
3179 if (bfqd
->peak_rate_samples
== 0) { /* first dispatch */
3180 bfq_log(bfqd
, "update_peak_rate: goto reset, samples %d",
3181 bfqd
->peak_rate_samples
);
3182 bfq_reset_rate_computation(bfqd
, rq
);
3183 goto update_last_values
; /* will add one sample */
3187 * Device idle for very long: the observation interval lasting
3188 * up to this dispatch cannot be a valid observation interval
3189 * for computing a new peak rate (similarly to the late-
3190 * completion event in bfq_completed_request()). Go to
3191 * update_rate_and_reset to have the following three steps
3193 * - close the observation interval at the last (previous)
3194 * request dispatch or completion
3195 * - compute rate, if possible, for that observation interval
3196 * - start a new observation interval with this dispatch
3198 if (now_ns
- bfqd
->last_dispatch
> 100*NSEC_PER_MSEC
&&
3199 bfqd
->rq_in_driver
== 0)
3200 goto update_rate_and_reset
;
3202 /* Update sampling information */
3203 bfqd
->peak_rate_samples
++;
3205 if ((bfqd
->rq_in_driver
> 0 ||
3206 now_ns
- bfqd
->last_completion
< BFQ_MIN_TT
)
3207 && !BFQ_RQ_SEEKY(bfqd
, bfqd
->last_position
, rq
))
3208 bfqd
->sequential_samples
++;
3210 bfqd
->tot_sectors_dispatched
+= blk_rq_sectors(rq
);
3212 /* Reset max observed rq size every 32 dispatches */
3213 if (likely(bfqd
->peak_rate_samples
% 32))
3214 bfqd
->last_rq_max_size
=
3215 max_t(u32
, blk_rq_sectors(rq
), bfqd
->last_rq_max_size
);
3217 bfqd
->last_rq_max_size
= blk_rq_sectors(rq
);
3219 bfqd
->delta_from_first
= now_ns
- bfqd
->first_dispatch
;
3221 /* Target observation interval not yet reached, go on sampling */
3222 if (bfqd
->delta_from_first
< BFQ_RATE_REF_INTERVAL
)
3223 goto update_last_values
;
3225 update_rate_and_reset
:
3226 bfq_update_rate_reset(bfqd
, rq
);
3228 bfqd
->last_position
= blk_rq_pos(rq
) + blk_rq_sectors(rq
);
3229 if (RQ_BFQQ(rq
) == bfqd
->in_service_queue
)
3230 bfqd
->in_serv_last_pos
= bfqd
->last_position
;
3231 bfqd
->last_dispatch
= now_ns
;
3235 * Remove request from internal lists.
3237 static void bfq_dispatch_remove(struct request_queue
*q
, struct request
*rq
)
3239 struct bfq_queue
*bfqq
= RQ_BFQQ(rq
);
3242 * For consistency, the next instruction should have been
3243 * executed after removing the request from the queue and
3244 * dispatching it. We execute instead this instruction before
3245 * bfq_remove_request() (and hence introduce a temporary
3246 * inconsistency), for efficiency. In fact, should this
3247 * dispatch occur for a non in-service bfqq, this anticipated
3248 * increment prevents two counters related to bfqq->dispatched
3249 * from risking to be, first, uselessly decremented, and then
3250 * incremented again when the (new) value of bfqq->dispatched
3251 * happens to be taken into account.
3254 bfq_update_peak_rate(q
->elevator
->elevator_data
, rq
);
3256 bfq_remove_request(q
, rq
);
3260 * There is a case where idling does not have to be performed for
3261 * throughput concerns, but to preserve the throughput share of
3262 * the process associated with bfqq.
3264 * To introduce this case, we can note that allowing the drive
3265 * to enqueue more than one request at a time, and hence
3266 * delegating de facto final scheduling decisions to the
3267 * drive's internal scheduler, entails loss of control on the
3268 * actual request service order. In particular, the critical
3269 * situation is when requests from different processes happen
3270 * to be present, at the same time, in the internal queue(s)
3271 * of the drive. In such a situation, the drive, by deciding
3272 * the service order of the internally-queued requests, does
3273 * determine also the actual throughput distribution among
3274 * these processes. But the drive typically has no notion or
3275 * concern about per-process throughput distribution, and
3276 * makes its decisions only on a per-request basis. Therefore,
3277 * the service distribution enforced by the drive's internal
3278 * scheduler is likely to coincide with the desired throughput
3279 * distribution only in a completely symmetric, or favorably
3280 * skewed scenario where:
3281 * (i-a) each of these processes must get the same throughput as
3283 * (i-b) in case (i-a) does not hold, it holds that the process
3284 * associated with bfqq must receive a lower or equal
3285 * throughput than any of the other processes;
3286 * (ii) the I/O of each process has the same properties, in
3287 * terms of locality (sequential or random), direction
3288 * (reads or writes), request sizes, greediness
3289 * (from I/O-bound to sporadic), and so on;
3291 * In fact, in such a scenario, the drive tends to treat the requests
3292 * of each process in about the same way as the requests of the
3293 * others, and thus to provide each of these processes with about the
3294 * same throughput. This is exactly the desired throughput
3295 * distribution if (i-a) holds, or, if (i-b) holds instead, this is an
3296 * even more convenient distribution for (the process associated with)
3299 * In contrast, in any asymmetric or unfavorable scenario, device
3300 * idling (I/O-dispatch plugging) is certainly needed to guarantee
3301 * that bfqq receives its assigned fraction of the device throughput
3302 * (see [1] for details).
3304 * The problem is that idling may significantly reduce throughput with
3305 * certain combinations of types of I/O and devices. An important
3306 * example is sync random I/O on flash storage with command
3307 * queueing. So, unless bfqq falls in cases where idling also boosts
3308 * throughput, it is important to check conditions (i-a), i(-b) and
3309 * (ii) accurately, so as to avoid idling when not strictly needed for
3310 * service guarantees.
3312 * Unfortunately, it is extremely difficult to thoroughly check
3313 * condition (ii). And, in case there are active groups, it becomes
3314 * very difficult to check conditions (i-a) and (i-b) too. In fact,
3315 * if there are active groups, then, for conditions (i-a) or (i-b) to
3316 * become false 'indirectly', it is enough that an active group
3317 * contains more active processes or sub-groups than some other active
3318 * group. More precisely, for conditions (i-a) or (i-b) to become
3319 * false because of such a group, it is not even necessary that the
3320 * group is (still) active: it is sufficient that, even if the group
3321 * has become inactive, some of its descendant processes still have
3322 * some request already dispatched but still waiting for
3323 * completion. In fact, requests have still to be guaranteed their
3324 * share of the throughput even after being dispatched. In this
3325 * respect, it is easy to show that, if a group frequently becomes
3326 * inactive while still having in-flight requests, and if, when this
3327 * happens, the group is not considered in the calculation of whether
3328 * the scenario is asymmetric, then the group may fail to be
3329 * guaranteed its fair share of the throughput (basically because
3330 * idling may not be performed for the descendant processes of the
3331 * group, but it had to be). We address this issue with the following
3332 * bi-modal behavior, implemented in the function
3333 * bfq_asymmetric_scenario().
3335 * If there are groups with requests waiting for completion
3336 * (as commented above, some of these groups may even be
3337 * already inactive), then the scenario is tagged as
3338 * asymmetric, conservatively, without checking any of the
3339 * conditions (i-a), (i-b) or (ii). So the device is idled for bfqq.
3340 * This behavior matches also the fact that groups are created
3341 * exactly if controlling I/O is a primary concern (to
3342 * preserve bandwidth and latency guarantees).
3344 * On the opposite end, if there are no groups with requests waiting
3345 * for completion, then only conditions (i-a) and (i-b) are actually
3346 * controlled, i.e., provided that conditions (i-a) or (i-b) holds,
3347 * idling is not performed, regardless of whether condition (ii)
3348 * holds. In other words, only if conditions (i-a) and (i-b) do not
3349 * hold, then idling is allowed, and the device tends to be prevented
3350 * from queueing many requests, possibly of several processes. Since
3351 * there are no groups with requests waiting for completion, then, to
3352 * control conditions (i-a) and (i-b) it is enough to check just
3353 * whether all the queues with requests waiting for completion also
3354 * have the same weight.
3356 * Not checking condition (ii) evidently exposes bfqq to the
3357 * risk of getting less throughput than its fair share.
3358 * However, for queues with the same weight, a further
3359 * mechanism, preemption, mitigates or even eliminates this
3360 * problem. And it does so without consequences on overall
3361 * throughput. This mechanism and its benefits are explained
3362 * in the next three paragraphs.
3364 * Even if a queue, say Q, is expired when it remains idle, Q
3365 * can still preempt the new in-service queue if the next
3366 * request of Q arrives soon (see the comments on
3367 * bfq_bfqq_update_budg_for_activation). If all queues and
3368 * groups have the same weight, this form of preemption,
3369 * combined with the hole-recovery heuristic described in the
3370 * comments on function bfq_bfqq_update_budg_for_activation,
3371 * are enough to preserve a correct bandwidth distribution in
3372 * the mid term, even without idling. In fact, even if not
3373 * idling allows the internal queues of the device to contain
3374 * many requests, and thus to reorder requests, we can rather
3375 * safely assume that the internal scheduler still preserves a
3376 * minimum of mid-term fairness.
3378 * More precisely, this preemption-based, idleless approach
3379 * provides fairness in terms of IOPS, and not sectors per
3380 * second. This can be seen with a simple example. Suppose
3381 * that there are two queues with the same weight, but that
3382 * the first queue receives requests of 8 sectors, while the
3383 * second queue receives requests of 1024 sectors. In
3384 * addition, suppose that each of the two queues contains at
3385 * most one request at a time, which implies that each queue
3386 * always remains idle after it is served. Finally, after
3387 * remaining idle, each queue receives very quickly a new
3388 * request. It follows that the two queues are served
3389 * alternatively, preempting each other if needed. This
3390 * implies that, although both queues have the same weight,
3391 * the queue with large requests receives a service that is
3392 * 1024/8 times as high as the service received by the other
3395 * The motivation for using preemption instead of idling (for
3396 * queues with the same weight) is that, by not idling,
3397 * service guarantees are preserved (completely or at least in
3398 * part) without minimally sacrificing throughput. And, if
3399 * there is no active group, then the primary expectation for
3400 * this device is probably a high throughput.
3402 * We are now left only with explaining the two sub-conditions in the
3403 * additional compound condition that is checked below for deciding
3404 * whether the scenario is asymmetric. To explain the first
3405 * sub-condition, we need to add that the function
3406 * bfq_asymmetric_scenario checks the weights of only
3407 * non-weight-raised queues, for efficiency reasons (see comments on
3408 * bfq_weights_tree_add()). Then the fact that bfqq is weight-raised
3409 * is checked explicitly here. More precisely, the compound condition
3410 * below takes into account also the fact that, even if bfqq is being
3411 * weight-raised, the scenario is still symmetric if all queues with
3412 * requests waiting for completion happen to be
3413 * weight-raised. Actually, we should be even more precise here, and
3414 * differentiate between interactive weight raising and soft real-time
3417 * The second sub-condition checked in the compound condition is
3418 * whether there is a fair amount of already in-flight I/O not
3419 * belonging to bfqq. If so, I/O dispatching is to be plugged, for the
3420 * following reason. The drive may decide to serve in-flight
3421 * non-bfqq's I/O requests before bfqq's ones, thereby delaying the
3422 * arrival of new I/O requests for bfqq (recall that bfqq is sync). If
3423 * I/O-dispatching is not plugged, then, while bfqq remains empty, a
3424 * basically uncontrolled amount of I/O from other queues may be
3425 * dispatched too, possibly causing the service of bfqq's I/O to be
3426 * delayed even longer in the drive. This problem gets more and more
3427 * serious as the speed and the queue depth of the drive grow,
3428 * because, as these two quantities grow, the probability to find no
3429 * queue busy but many requests in flight grows too. By contrast,
3430 * plugging I/O dispatching minimizes the delay induced by already
3431 * in-flight I/O, and enables bfqq to recover the bandwidth it may
3432 * lose because of this delay.
3434 * As a side note, it is worth considering that the above
3435 * device-idling countermeasures may however fail in the following
3436 * unlucky scenario: if I/O-dispatch plugging is (correctly) disabled
3437 * in a time period during which all symmetry sub-conditions hold, and
3438 * therefore the device is allowed to enqueue many requests, but at
3439 * some later point in time some sub-condition stops to hold, then it
3440 * may become impossible to make requests be served in the desired
3441 * order until all the requests already queued in the device have been
3442 * served. The last sub-condition commented above somewhat mitigates
3443 * this problem for weight-raised queues.
3445 static bool idling_needed_for_service_guarantees(struct bfq_data
*bfqd
,
3446 struct bfq_queue
*bfqq
)
3448 /* No point in idling for bfqq if it won't get requests any longer */
3449 if (unlikely(!bfqq_process_refs(bfqq
)))
3452 return (bfqq
->wr_coeff
> 1 &&
3453 (bfqd
->wr_busy_queues
<
3454 bfq_tot_busy_queues(bfqd
) ||
3455 bfqd
->rq_in_driver
>=
3456 bfqq
->dispatched
+ 4)) ||
3457 bfq_asymmetric_scenario(bfqd
, bfqq
);
3460 static bool __bfq_bfqq_expire(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
,
3461 enum bfqq_expiration reason
)
3464 * If this bfqq is shared between multiple processes, check
3465 * to make sure that those processes are still issuing I/Os
3466 * within the mean seek distance. If not, it may be time to
3467 * break the queues apart again.
3469 if (bfq_bfqq_coop(bfqq
) && BFQQ_SEEKY(bfqq
))
3470 bfq_mark_bfqq_split_coop(bfqq
);
3473 * Consider queues with a higher finish virtual time than
3474 * bfqq. If idling_needed_for_service_guarantees(bfqq) returns
3475 * true, then bfqq's bandwidth would be violated if an
3476 * uncontrolled amount of I/O from these queues were
3477 * dispatched while bfqq is waiting for its new I/O to
3478 * arrive. This is exactly what may happen if this is a forced
3479 * expiration caused by a preemption attempt, and if bfqq is
3480 * not re-scheduled. To prevent this from happening, re-queue
3481 * bfqq if it needs I/O-dispatch plugging, even if it is
3482 * empty. By doing so, bfqq is granted to be served before the
3483 * above queues (provided that bfqq is of course eligible).
3485 if (RB_EMPTY_ROOT(&bfqq
->sort_list
) &&
3486 !(reason
== BFQQE_PREEMPTED
&&
3487 idling_needed_for_service_guarantees(bfqd
, bfqq
))) {
3488 if (bfqq
->dispatched
== 0)
3490 * Overloading budget_timeout field to store
3491 * the time at which the queue remains with no
3492 * backlog and no outstanding request; used by
3493 * the weight-raising mechanism.
3495 bfqq
->budget_timeout
= jiffies
;
3497 bfq_del_bfqq_busy(bfqd
, bfqq
, true);
3499 bfq_requeue_bfqq(bfqd
, bfqq
, true);
3501 * Resort priority tree of potential close cooperators.
3502 * See comments on bfq_pos_tree_add_move() for the unlikely().
3504 if (unlikely(!bfqd
->nonrot_with_queueing
&&
3505 !RB_EMPTY_ROOT(&bfqq
->sort_list
)))
3506 bfq_pos_tree_add_move(bfqd
, bfqq
);
3510 * All in-service entities must have been properly deactivated
3511 * or requeued before executing the next function, which
3512 * resets all in-service entities as no more in service. This
3513 * may cause bfqq to be freed. If this happens, the next
3514 * function returns true.
3516 return __bfq_bfqd_reset_in_service(bfqd
);
3520 * __bfq_bfqq_recalc_budget - try to adapt the budget to the @bfqq behavior.
3521 * @bfqd: device data.
3522 * @bfqq: queue to update.
3523 * @reason: reason for expiration.
3525 * Handle the feedback on @bfqq budget at queue expiration.
3526 * See the body for detailed comments.
3528 static void __bfq_bfqq_recalc_budget(struct bfq_data
*bfqd
,
3529 struct bfq_queue
*bfqq
,
3530 enum bfqq_expiration reason
)
3532 struct request
*next_rq
;
3533 int budget
, min_budget
;
3535 min_budget
= bfq_min_budget(bfqd
);
3537 if (bfqq
->wr_coeff
== 1)
3538 budget
= bfqq
->max_budget
;
3540 * Use a constant, low budget for weight-raised queues,
3541 * to help achieve a low latency. Keep it slightly higher
3542 * than the minimum possible budget, to cause a little
3543 * bit fewer expirations.
3545 budget
= 2 * min_budget
;
3547 bfq_log_bfqq(bfqd
, bfqq
, "recalc_budg: last budg %d, budg left %d",
3548 bfqq
->entity
.budget
, bfq_bfqq_budget_left(bfqq
));
3549 bfq_log_bfqq(bfqd
, bfqq
, "recalc_budg: last max_budg %d, min budg %d",
3550 budget
, bfq_min_budget(bfqd
));
3551 bfq_log_bfqq(bfqd
, bfqq
, "recalc_budg: sync %d, seeky %d",
3552 bfq_bfqq_sync(bfqq
), BFQQ_SEEKY(bfqd
->in_service_queue
));
3554 if (bfq_bfqq_sync(bfqq
) && bfqq
->wr_coeff
== 1) {
3557 * Caveat: in all the following cases we trade latency
3560 case BFQQE_TOO_IDLE
:
3562 * This is the only case where we may reduce
3563 * the budget: if there is no request of the
3564 * process still waiting for completion, then
3565 * we assume (tentatively) that the timer has
3566 * expired because the batch of requests of
3567 * the process could have been served with a
3568 * smaller budget. Hence, betting that
3569 * process will behave in the same way when it
3570 * becomes backlogged again, we reduce its
3571 * next budget. As long as we guess right,
3572 * this budget cut reduces the latency
3573 * experienced by the process.
3575 * However, if there are still outstanding
3576 * requests, then the process may have not yet
3577 * issued its next request just because it is
3578 * still waiting for the completion of some of
3579 * the still outstanding ones. So in this
3580 * subcase we do not reduce its budget, on the
3581 * contrary we increase it to possibly boost
3582 * the throughput, as discussed in the
3583 * comments to the BUDGET_TIMEOUT case.
3585 if (bfqq
->dispatched
> 0) /* still outstanding reqs */
3586 budget
= min(budget
* 2, bfqd
->bfq_max_budget
);
3588 if (budget
> 5 * min_budget
)
3589 budget
-= 4 * min_budget
;
3591 budget
= min_budget
;
3594 case BFQQE_BUDGET_TIMEOUT
:
3596 * We double the budget here because it gives
3597 * the chance to boost the throughput if this
3598 * is not a seeky process (and has bumped into
3599 * this timeout because of, e.g., ZBR).
3601 budget
= min(budget
* 2, bfqd
->bfq_max_budget
);
3603 case BFQQE_BUDGET_EXHAUSTED
:
3605 * The process still has backlog, and did not
3606 * let either the budget timeout or the disk
3607 * idling timeout expire. Hence it is not
3608 * seeky, has a short thinktime and may be
3609 * happy with a higher budget too. So
3610 * definitely increase the budget of this good
3611 * candidate to boost the disk throughput.
3613 budget
= min(budget
* 4, bfqd
->bfq_max_budget
);
3615 case BFQQE_NO_MORE_REQUESTS
:
3617 * For queues that expire for this reason, it
3618 * is particularly important to keep the
3619 * budget close to the actual service they
3620 * need. Doing so reduces the timestamp
3621 * misalignment problem described in the
3622 * comments in the body of
3623 * __bfq_activate_entity. In fact, suppose
3624 * that a queue systematically expires for
3625 * BFQQE_NO_MORE_REQUESTS and presents a
3626 * new request in time to enjoy timestamp
3627 * back-shifting. The larger the budget of the
3628 * queue is with respect to the service the
3629 * queue actually requests in each service
3630 * slot, the more times the queue can be
3631 * reactivated with the same virtual finish
3632 * time. It follows that, even if this finish
3633 * time is pushed to the system virtual time
3634 * to reduce the consequent timestamp
3635 * misalignment, the queue unjustly enjoys for
3636 * many re-activations a lower finish time
3637 * than all newly activated queues.
3639 * The service needed by bfqq is measured
3640 * quite precisely by bfqq->entity.service.
3641 * Since bfqq does not enjoy device idling,
3642 * bfqq->entity.service is equal to the number
3643 * of sectors that the process associated with
3644 * bfqq requested to read/write before waiting
3645 * for request completions, or blocking for
3648 budget
= max_t(int, bfqq
->entity
.service
, min_budget
);
3653 } else if (!bfq_bfqq_sync(bfqq
)) {
3655 * Async queues get always the maximum possible
3656 * budget, as for them we do not care about latency
3657 * (in addition, their ability to dispatch is limited
3658 * by the charging factor).
3660 budget
= bfqd
->bfq_max_budget
;
3663 bfqq
->max_budget
= budget
;
3665 if (bfqd
->budgets_assigned
>= bfq_stats_min_budgets
&&
3666 !bfqd
->bfq_user_max_budget
)
3667 bfqq
->max_budget
= min(bfqq
->max_budget
, bfqd
->bfq_max_budget
);
3670 * If there is still backlog, then assign a new budget, making
3671 * sure that it is large enough for the next request. Since
3672 * the finish time of bfqq must be kept in sync with the
3673 * budget, be sure to call __bfq_bfqq_expire() *after* this
3676 * If there is no backlog, then no need to update the budget;
3677 * it will be updated on the arrival of a new request.
3679 next_rq
= bfqq
->next_rq
;
3681 bfqq
->entity
.budget
= max_t(unsigned long, bfqq
->max_budget
,
3682 bfq_serv_to_charge(next_rq
, bfqq
));
3684 bfq_log_bfqq(bfqd
, bfqq
, "head sect: %u, new budget %d",
3685 next_rq
? blk_rq_sectors(next_rq
) : 0,
3686 bfqq
->entity
.budget
);
3690 * Return true if the process associated with bfqq is "slow". The slow
3691 * flag is used, in addition to the budget timeout, to reduce the
3692 * amount of service provided to seeky processes, and thus reduce
3693 * their chances to lower the throughput. More details in the comments
3694 * on the function bfq_bfqq_expire().
3696 * An important observation is in order: as discussed in the comments
3697 * on the function bfq_update_peak_rate(), with devices with internal
3698 * queues, it is hard if ever possible to know when and for how long
3699 * an I/O request is processed by the device (apart from the trivial
3700 * I/O pattern where a new request is dispatched only after the
3701 * previous one has been completed). This makes it hard to evaluate
3702 * the real rate at which the I/O requests of each bfq_queue are
3703 * served. In fact, for an I/O scheduler like BFQ, serving a
3704 * bfq_queue means just dispatching its requests during its service
3705 * slot (i.e., until the budget of the queue is exhausted, or the
3706 * queue remains idle, or, finally, a timeout fires). But, during the
3707 * service slot of a bfq_queue, around 100 ms at most, the device may
3708 * be even still processing requests of bfq_queues served in previous
3709 * service slots. On the opposite end, the requests of the in-service
3710 * bfq_queue may be completed after the service slot of the queue
3713 * Anyway, unless more sophisticated solutions are used
3714 * (where possible), the sum of the sizes of the requests dispatched
3715 * during the service slot of a bfq_queue is probably the only
3716 * approximation available for the service received by the bfq_queue
3717 * during its service slot. And this sum is the quantity used in this
3718 * function to evaluate the I/O speed of a process.
3720 static bool bfq_bfqq_is_slow(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
,
3721 bool compensate
, enum bfqq_expiration reason
,
3722 unsigned long *delta_ms
)
3724 ktime_t delta_ktime
;
3726 bool slow
= BFQQ_SEEKY(bfqq
); /* if delta too short, use seekyness */
3728 if (!bfq_bfqq_sync(bfqq
))
3732 delta_ktime
= bfqd
->last_idling_start
;
3734 delta_ktime
= ktime_get();
3735 delta_ktime
= ktime_sub(delta_ktime
, bfqd
->last_budget_start
);
3736 delta_usecs
= ktime_to_us(delta_ktime
);
3738 /* don't use too short time intervals */
3739 if (delta_usecs
< 1000) {
3740 if (blk_queue_nonrot(bfqd
->queue
))
3742 * give same worst-case guarantees as idling
3745 *delta_ms
= BFQ_MIN_TT
/ NSEC_PER_MSEC
;
3746 else /* charge at least one seek */
3747 *delta_ms
= bfq_slice_idle
/ NSEC_PER_MSEC
;
3752 *delta_ms
= delta_usecs
/ USEC_PER_MSEC
;
3755 * Use only long (> 20ms) intervals to filter out excessive
3756 * spikes in service rate estimation.
3758 if (delta_usecs
> 20000) {
3760 * Caveat for rotational devices: processes doing I/O
3761 * in the slower disk zones tend to be slow(er) even
3762 * if not seeky. In this respect, the estimated peak
3763 * rate is likely to be an average over the disk
3764 * surface. Accordingly, to not be too harsh with
3765 * unlucky processes, a process is deemed slow only if
3766 * its rate has been lower than half of the estimated
3769 slow
= bfqq
->entity
.service
< bfqd
->bfq_max_budget
/ 2;
3772 bfq_log_bfqq(bfqd
, bfqq
, "bfq_bfqq_is_slow: slow %d", slow
);
3778 * To be deemed as soft real-time, an application must meet two
3779 * requirements. First, the application must not require an average
3780 * bandwidth higher than the approximate bandwidth required to playback or
3781 * record a compressed high-definition video.
3782 * The next function is invoked on the completion of the last request of a
3783 * batch, to compute the next-start time instant, soft_rt_next_start, such
3784 * that, if the next request of the application does not arrive before
3785 * soft_rt_next_start, then the above requirement on the bandwidth is met.
3787 * The second requirement is that the request pattern of the application is
3788 * isochronous, i.e., that, after issuing a request or a batch of requests,
3789 * the application stops issuing new requests until all its pending requests
3790 * have been completed. After that, the application may issue a new batch,
3792 * For this reason the next function is invoked to compute
3793 * soft_rt_next_start only for applications that meet this requirement,
3794 * whereas soft_rt_next_start is set to infinity for applications that do
3797 * Unfortunately, even a greedy (i.e., I/O-bound) application may
3798 * happen to meet, occasionally or systematically, both the above
3799 * bandwidth and isochrony requirements. This may happen at least in
3800 * the following circumstances. First, if the CPU load is high. The
3801 * application may stop issuing requests while the CPUs are busy
3802 * serving other processes, then restart, then stop again for a while,
3803 * and so on. The other circumstances are related to the storage
3804 * device: the storage device is highly loaded or reaches a low-enough
3805 * throughput with the I/O of the application (e.g., because the I/O
3806 * is random and/or the device is slow). In all these cases, the
3807 * I/O of the application may be simply slowed down enough to meet
3808 * the bandwidth and isochrony requirements. To reduce the probability
3809 * that greedy applications are deemed as soft real-time in these
3810 * corner cases, a further rule is used in the computation of
3811 * soft_rt_next_start: the return value of this function is forced to
3812 * be higher than the maximum between the following two quantities.
3814 * (a) Current time plus: (1) the maximum time for which the arrival
3815 * of a request is waited for when a sync queue becomes idle,
3816 * namely bfqd->bfq_slice_idle, and (2) a few extra jiffies. We
3817 * postpone for a moment the reason for adding a few extra
3818 * jiffies; we get back to it after next item (b). Lower-bounding
3819 * the return value of this function with the current time plus
3820 * bfqd->bfq_slice_idle tends to filter out greedy applications,
3821 * because the latter issue their next request as soon as possible
3822 * after the last one has been completed. In contrast, a soft
3823 * real-time application spends some time processing data, after a
3824 * batch of its requests has been completed.
3826 * (b) Current value of bfqq->soft_rt_next_start. As pointed out
3827 * above, greedy applications may happen to meet both the
3828 * bandwidth and isochrony requirements under heavy CPU or
3829 * storage-device load. In more detail, in these scenarios, these
3830 * applications happen, only for limited time periods, to do I/O
3831 * slowly enough to meet all the requirements described so far,
3832 * including the filtering in above item (a). These slow-speed
3833 * time intervals are usually interspersed between other time
3834 * intervals during which these applications do I/O at a very high
3835 * speed. Fortunately, exactly because of the high speed of the
3836 * I/O in the high-speed intervals, the values returned by this
3837 * function happen to be so high, near the end of any such
3838 * high-speed interval, to be likely to fall *after* the end of
3839 * the low-speed time interval that follows. These high values are
3840 * stored in bfqq->soft_rt_next_start after each invocation of
3841 * this function. As a consequence, if the last value of
3842 * bfqq->soft_rt_next_start is constantly used to lower-bound the
3843 * next value that this function may return, then, from the very
3844 * beginning of a low-speed interval, bfqq->soft_rt_next_start is
3845 * likely to be constantly kept so high that any I/O request
3846 * issued during the low-speed interval is considered as arriving
3847 * to soon for the application to be deemed as soft
3848 * real-time. Then, in the high-speed interval that follows, the
3849 * application will not be deemed as soft real-time, just because
3850 * it will do I/O at a high speed. And so on.
3852 * Getting back to the filtering in item (a), in the following two
3853 * cases this filtering might be easily passed by a greedy
3854 * application, if the reference quantity was just
3855 * bfqd->bfq_slice_idle:
3856 * 1) HZ is so low that the duration of a jiffy is comparable to or
3857 * higher than bfqd->bfq_slice_idle. This happens, e.g., on slow
3858 * devices with HZ=100. The time granularity may be so coarse
3859 * that the approximation, in jiffies, of bfqd->bfq_slice_idle
3860 * is rather lower than the exact value.
3861 * 2) jiffies, instead of increasing at a constant rate, may stop increasing
3862 * for a while, then suddenly 'jump' by several units to recover the lost
3863 * increments. This seems to happen, e.g., inside virtual machines.
3864 * To address this issue, in the filtering in (a) we do not use as a
3865 * reference time interval just bfqd->bfq_slice_idle, but
3866 * bfqd->bfq_slice_idle plus a few jiffies. In particular, we add the
3867 * minimum number of jiffies for which the filter seems to be quite
3868 * precise also in embedded systems and KVM/QEMU virtual machines.
3870 static unsigned long bfq_bfqq_softrt_next_start(struct bfq_data
*bfqd
,
3871 struct bfq_queue
*bfqq
)
3873 return max3(bfqq
->soft_rt_next_start
,
3874 bfqq
->last_idle_bklogged
+
3875 HZ
* bfqq
->service_from_backlogged
/
3876 bfqd
->bfq_wr_max_softrt_rate
,
3877 jiffies
+ nsecs_to_jiffies(bfqq
->bfqd
->bfq_slice_idle
) + 4);
3881 * bfq_bfqq_expire - expire a queue.
3882 * @bfqd: device owning the queue.
3883 * @bfqq: the queue to expire.
3884 * @compensate: if true, compensate for the time spent idling.
3885 * @reason: the reason causing the expiration.
3887 * If the process associated with bfqq does slow I/O (e.g., because it
3888 * issues random requests), we charge bfqq with the time it has been
3889 * in service instead of the service it has received (see
3890 * bfq_bfqq_charge_time for details on how this goal is achieved). As
3891 * a consequence, bfqq will typically get higher timestamps upon
3892 * reactivation, and hence it will be rescheduled as if it had
3893 * received more service than what it has actually received. In the
3894 * end, bfqq receives less service in proportion to how slowly its
3895 * associated process consumes its budgets (and hence how seriously it
3896 * tends to lower the throughput). In addition, this time-charging
3897 * strategy guarantees time fairness among slow processes. In
3898 * contrast, if the process associated with bfqq is not slow, we
3899 * charge bfqq exactly with the service it has received.
3901 * Charging time to the first type of queues and the exact service to
3902 * the other has the effect of using the WF2Q+ policy to schedule the
3903 * former on a timeslice basis, without violating service domain
3904 * guarantees among the latter.
3906 void bfq_bfqq_expire(struct bfq_data
*bfqd
,
3907 struct bfq_queue
*bfqq
,
3909 enum bfqq_expiration reason
)
3912 unsigned long delta
= 0;
3913 struct bfq_entity
*entity
= &bfqq
->entity
;
3916 * Check whether the process is slow (see bfq_bfqq_is_slow).
3918 slow
= bfq_bfqq_is_slow(bfqd
, bfqq
, compensate
, reason
, &delta
);
3921 * As above explained, charge slow (typically seeky) and
3922 * timed-out queues with the time and not the service
3923 * received, to favor sequential workloads.
3925 * Processes doing I/O in the slower disk zones will tend to
3926 * be slow(er) even if not seeky. Therefore, since the
3927 * estimated peak rate is actually an average over the disk
3928 * surface, these processes may timeout just for bad luck. To
3929 * avoid punishing them, do not charge time to processes that
3930 * succeeded in consuming at least 2/3 of their budget. This
3931 * allows BFQ to preserve enough elasticity to still perform
3932 * bandwidth, and not time, distribution with little unlucky
3933 * or quasi-sequential processes.
3935 if (bfqq
->wr_coeff
== 1 &&
3937 (reason
== BFQQE_BUDGET_TIMEOUT
&&
3938 bfq_bfqq_budget_left(bfqq
) >= entity
->budget
/ 3)))
3939 bfq_bfqq_charge_time(bfqd
, bfqq
, delta
);
3941 if (reason
== BFQQE_TOO_IDLE
&&
3942 entity
->service
<= 2 * entity
->budget
/ 10)
3943 bfq_clear_bfqq_IO_bound(bfqq
);
3945 if (bfqd
->low_latency
&& bfqq
->wr_coeff
== 1)
3946 bfqq
->last_wr_start_finish
= jiffies
;
3948 if (bfqd
->low_latency
&& bfqd
->bfq_wr_max_softrt_rate
> 0 &&
3949 RB_EMPTY_ROOT(&bfqq
->sort_list
)) {
3951 * If we get here, and there are no outstanding
3952 * requests, then the request pattern is isochronous
3953 * (see the comments on the function
3954 * bfq_bfqq_softrt_next_start()). Thus we can compute
3955 * soft_rt_next_start. And we do it, unless bfqq is in
3956 * interactive weight raising. We do not do it in the
3957 * latter subcase, for the following reason. bfqq may
3958 * be conveying the I/O needed to load a soft
3959 * real-time application. Such an application will
3960 * actually exhibit a soft real-time I/O pattern after
3961 * it finally starts doing its job. But, if
3962 * soft_rt_next_start is computed here for an
3963 * interactive bfqq, and bfqq had received a lot of
3964 * service before remaining with no outstanding
3965 * request (likely to happen on a fast device), then
3966 * soft_rt_next_start would be assigned such a high
3967 * value that, for a very long time, bfqq would be
3968 * prevented from being possibly considered as soft
3971 * If, instead, the queue still has outstanding
3972 * requests, then we have to wait for the completion
3973 * of all the outstanding requests to discover whether
3974 * the request pattern is actually isochronous.
3976 if (bfqq
->dispatched
== 0 &&
3977 bfqq
->wr_coeff
!= bfqd
->bfq_wr_coeff
)
3978 bfqq
->soft_rt_next_start
=
3979 bfq_bfqq_softrt_next_start(bfqd
, bfqq
);
3980 else if (bfqq
->dispatched
> 0) {
3982 * Schedule an update of soft_rt_next_start to when
3983 * the task may be discovered to be isochronous.
3985 bfq_mark_bfqq_softrt_update(bfqq
);
3989 bfq_log_bfqq(bfqd
, bfqq
,
3990 "expire (%d, slow %d, num_disp %d, short_ttime %d)", reason
,
3991 slow
, bfqq
->dispatched
, bfq_bfqq_has_short_ttime(bfqq
));
3994 * bfqq expired, so no total service time needs to be computed
3995 * any longer: reset state machine for measuring total service
3998 bfqd
->rqs_injected
= bfqd
->wait_dispatch
= false;
3999 bfqd
->waited_rq
= NULL
;
4002 * Increase, decrease or leave budget unchanged according to
4005 __bfq_bfqq_recalc_budget(bfqd
, bfqq
, reason
);
4006 if (__bfq_bfqq_expire(bfqd
, bfqq
, reason
))
4007 /* bfqq is gone, no more actions on it */
4010 /* mark bfqq as waiting a request only if a bic still points to it */
4011 if (!bfq_bfqq_busy(bfqq
) &&
4012 reason
!= BFQQE_BUDGET_TIMEOUT
&&
4013 reason
!= BFQQE_BUDGET_EXHAUSTED
) {
4014 bfq_mark_bfqq_non_blocking_wait_rq(bfqq
);
4016 * Not setting service to 0, because, if the next rq
4017 * arrives in time, the queue will go on receiving
4018 * service with this same budget (as if it never expired)
4021 entity
->service
= 0;
4024 * Reset the received-service counter for every parent entity.
4025 * Differently from what happens with bfqq->entity.service,
4026 * the resetting of this counter never needs to be postponed
4027 * for parent entities. In fact, in case bfqq may have a
4028 * chance to go on being served using the last, partially
4029 * consumed budget, bfqq->entity.service needs to be kept,
4030 * because if bfqq then actually goes on being served using
4031 * the same budget, the last value of bfqq->entity.service is
4032 * needed to properly decrement bfqq->entity.budget by the
4033 * portion already consumed. In contrast, it is not necessary
4034 * to keep entity->service for parent entities too, because
4035 * the bubble up of the new value of bfqq->entity.budget will
4036 * make sure that the budgets of parent entities are correct,
4037 * even in case bfqq and thus parent entities go on receiving
4038 * service with the same budget.
4040 entity
= entity
->parent
;
4041 for_each_entity(entity
)
4042 entity
->service
= 0;
4046 * Budget timeout is not implemented through a dedicated timer, but
4047 * just checked on request arrivals and completions, as well as on
4048 * idle timer expirations.
4050 static bool bfq_bfqq_budget_timeout(struct bfq_queue
*bfqq
)
4052 return time_is_before_eq_jiffies(bfqq
->budget_timeout
);
4056 * If we expire a queue that is actively waiting (i.e., with the
4057 * device idled) for the arrival of a new request, then we may incur
4058 * the timestamp misalignment problem described in the body of the
4059 * function __bfq_activate_entity. Hence we return true only if this
4060 * condition does not hold, or if the queue is slow enough to deserve
4061 * only to be kicked off for preserving a high throughput.
4063 static bool bfq_may_expire_for_budg_timeout(struct bfq_queue
*bfqq
)
4065 bfq_log_bfqq(bfqq
->bfqd
, bfqq
,
4066 "may_budget_timeout: wait_request %d left %d timeout %d",
4067 bfq_bfqq_wait_request(bfqq
),
4068 bfq_bfqq_budget_left(bfqq
) >= bfqq
->entity
.budget
/ 3,
4069 bfq_bfqq_budget_timeout(bfqq
));
4071 return (!bfq_bfqq_wait_request(bfqq
) ||
4072 bfq_bfqq_budget_left(bfqq
) >= bfqq
->entity
.budget
/ 3)
4074 bfq_bfqq_budget_timeout(bfqq
);
4077 static bool idling_boosts_thr_without_issues(struct bfq_data
*bfqd
,
4078 struct bfq_queue
*bfqq
)
4080 bool rot_without_queueing
=
4081 !blk_queue_nonrot(bfqd
->queue
) && !bfqd
->hw_tag
,
4082 bfqq_sequential_and_IO_bound
,
4085 /* No point in idling for bfqq if it won't get requests any longer */
4086 if (unlikely(!bfqq_process_refs(bfqq
)))
4089 bfqq_sequential_and_IO_bound
= !BFQQ_SEEKY(bfqq
) &&
4090 bfq_bfqq_IO_bound(bfqq
) && bfq_bfqq_has_short_ttime(bfqq
);
4093 * The next variable takes into account the cases where idling
4094 * boosts the throughput.
4096 * The value of the variable is computed considering, first, that
4097 * idling is virtually always beneficial for the throughput if:
4098 * (a) the device is not NCQ-capable and rotational, or
4099 * (b) regardless of the presence of NCQ, the device is rotational and
4100 * the request pattern for bfqq is I/O-bound and sequential, or
4101 * (c) regardless of whether it is rotational, the device is
4102 * not NCQ-capable and the request pattern for bfqq is
4103 * I/O-bound and sequential.
4105 * Secondly, and in contrast to the above item (b), idling an
4106 * NCQ-capable flash-based device would not boost the
4107 * throughput even with sequential I/O; rather it would lower
4108 * the throughput in proportion to how fast the device
4109 * is. Accordingly, the next variable is true if any of the
4110 * above conditions (a), (b) or (c) is true, and, in
4111 * particular, happens to be false if bfqd is an NCQ-capable
4112 * flash-based device.
4114 idling_boosts_thr
= rot_without_queueing
||
4115 ((!blk_queue_nonrot(bfqd
->queue
) || !bfqd
->hw_tag
) &&
4116 bfqq_sequential_and_IO_bound
);
4119 * The return value of this function is equal to that of
4120 * idling_boosts_thr, unless a special case holds. In this
4121 * special case, described below, idling may cause problems to
4122 * weight-raised queues.
4124 * When the request pool is saturated (e.g., in the presence
4125 * of write hogs), if the processes associated with
4126 * non-weight-raised queues ask for requests at a lower rate,
4127 * then processes associated with weight-raised queues have a
4128 * higher probability to get a request from the pool
4129 * immediately (or at least soon) when they need one. Thus
4130 * they have a higher probability to actually get a fraction
4131 * of the device throughput proportional to their high
4132 * weight. This is especially true with NCQ-capable drives,
4133 * which enqueue several requests in advance, and further
4134 * reorder internally-queued requests.
4136 * For this reason, we force to false the return value if
4137 * there are weight-raised busy queues. In this case, and if
4138 * bfqq is not weight-raised, this guarantees that the device
4139 * is not idled for bfqq (if, instead, bfqq is weight-raised,
4140 * then idling will be guaranteed by another variable, see
4141 * below). Combined with the timestamping rules of BFQ (see
4142 * [1] for details), this behavior causes bfqq, and hence any
4143 * sync non-weight-raised queue, to get a lower number of
4144 * requests served, and thus to ask for a lower number of
4145 * requests from the request pool, before the busy
4146 * weight-raised queues get served again. This often mitigates
4147 * starvation problems in the presence of heavy write
4148 * workloads and NCQ, thereby guaranteeing a higher
4149 * application and system responsiveness in these hostile
4152 return idling_boosts_thr
&&
4153 bfqd
->wr_busy_queues
== 0;
4157 * For a queue that becomes empty, device idling is allowed only if
4158 * this function returns true for that queue. As a consequence, since
4159 * device idling plays a critical role for both throughput boosting
4160 * and service guarantees, the return value of this function plays a
4161 * critical role as well.
4163 * In a nutshell, this function returns true only if idling is
4164 * beneficial for throughput or, even if detrimental for throughput,
4165 * idling is however necessary to preserve service guarantees (low
4166 * latency, desired throughput distribution, ...). In particular, on
4167 * NCQ-capable devices, this function tries to return false, so as to
4168 * help keep the drives' internal queues full, whenever this helps the
4169 * device boost the throughput without causing any service-guarantee
4172 * Most of the issues taken into account to get the return value of
4173 * this function are not trivial. We discuss these issues in the two
4174 * functions providing the main pieces of information needed by this
4177 static bool bfq_better_to_idle(struct bfq_queue
*bfqq
)
4179 struct bfq_data
*bfqd
= bfqq
->bfqd
;
4180 bool idling_boosts_thr_with_no_issue
, idling_needed_for_service_guar
;
4182 /* No point in idling for bfqq if it won't get requests any longer */
4183 if (unlikely(!bfqq_process_refs(bfqq
)))
4186 if (unlikely(bfqd
->strict_guarantees
))
4190 * Idling is performed only if slice_idle > 0. In addition, we
4193 * (b) bfqq is in the idle io prio class: in this case we do
4194 * not idle because we want to minimize the bandwidth that
4195 * queues in this class can steal to higher-priority queues
4197 if (bfqd
->bfq_slice_idle
== 0 || !bfq_bfqq_sync(bfqq
) ||
4198 bfq_class_idle(bfqq
))
4201 idling_boosts_thr_with_no_issue
=
4202 idling_boosts_thr_without_issues(bfqd
, bfqq
);
4204 idling_needed_for_service_guar
=
4205 idling_needed_for_service_guarantees(bfqd
, bfqq
);
4208 * We have now the two components we need to compute the
4209 * return value of the function, which is true only if idling
4210 * either boosts the throughput (without issues), or is
4211 * necessary to preserve service guarantees.
4213 return idling_boosts_thr_with_no_issue
||
4214 idling_needed_for_service_guar
;
4218 * If the in-service queue is empty but the function bfq_better_to_idle
4219 * returns true, then:
4220 * 1) the queue must remain in service and cannot be expired, and
4221 * 2) the device must be idled to wait for the possible arrival of a new
4222 * request for the queue.
4223 * See the comments on the function bfq_better_to_idle for the reasons
4224 * why performing device idling is the best choice to boost the throughput
4225 * and preserve service guarantees when bfq_better_to_idle itself
4228 static bool bfq_bfqq_must_idle(struct bfq_queue
*bfqq
)
4230 return RB_EMPTY_ROOT(&bfqq
->sort_list
) && bfq_better_to_idle(bfqq
);
4234 * This function chooses the queue from which to pick the next extra
4235 * I/O request to inject, if it finds a compatible queue. See the
4236 * comments on bfq_update_inject_limit() for details on the injection
4237 * mechanism, and for the definitions of the quantities mentioned
4240 static struct bfq_queue
*
4241 bfq_choose_bfqq_for_injection(struct bfq_data
*bfqd
)
4243 struct bfq_queue
*bfqq
, *in_serv_bfqq
= bfqd
->in_service_queue
;
4244 unsigned int limit
= in_serv_bfqq
->inject_limit
;
4247 * - bfqq is not weight-raised and therefore does not carry
4248 * time-critical I/O,
4250 * - regardless of whether bfqq is weight-raised, bfqq has
4251 * however a long think time, during which it can absorb the
4252 * effect of an appropriate number of extra I/O requests
4253 * from other queues (see bfq_update_inject_limit for
4254 * details on the computation of this number);
4255 * then injection can be performed without restrictions.
4257 bool in_serv_always_inject
= in_serv_bfqq
->wr_coeff
== 1 ||
4258 !bfq_bfqq_has_short_ttime(in_serv_bfqq
);
4262 * - the baseline total service time could not be sampled yet,
4263 * so the inject limit happens to be still 0, and
4264 * - a lot of time has elapsed since the plugging of I/O
4265 * dispatching started, so drive speed is being wasted
4267 * then temporarily raise inject limit to one request.
4269 if (limit
== 0 && in_serv_bfqq
->last_serv_time_ns
== 0 &&
4270 bfq_bfqq_wait_request(in_serv_bfqq
) &&
4271 time_is_before_eq_jiffies(bfqd
->last_idling_start_jiffies
+
4272 bfqd
->bfq_slice_idle
)
4276 if (bfqd
->rq_in_driver
>= limit
)
4280 * Linear search of the source queue for injection; but, with
4281 * a high probability, very few steps are needed to find a
4282 * candidate queue, i.e., a queue with enough budget left for
4283 * its next request. In fact:
4284 * - BFQ dynamically updates the budget of every queue so as
4285 * to accommodate the expected backlog of the queue;
4286 * - if a queue gets all its requests dispatched as injected
4287 * service, then the queue is removed from the active list
4288 * (and re-added only if it gets new requests, but then it
4289 * is assigned again enough budget for its new backlog).
4291 list_for_each_entry(bfqq
, &bfqd
->active_list
, bfqq_list
)
4292 if (!RB_EMPTY_ROOT(&bfqq
->sort_list
) &&
4293 (in_serv_always_inject
|| bfqq
->wr_coeff
> 1) &&
4294 bfq_serv_to_charge(bfqq
->next_rq
, bfqq
) <=
4295 bfq_bfqq_budget_left(bfqq
)) {
4297 * Allow for only one large in-flight request
4298 * on non-rotational devices, for the
4299 * following reason. On non-rotationl drives,
4300 * large requests take much longer than
4301 * smaller requests to be served. In addition,
4302 * the drive prefers to serve large requests
4303 * w.r.t. to small ones, if it can choose. So,
4304 * having more than one large requests queued
4305 * in the drive may easily make the next first
4306 * request of the in-service queue wait for so
4307 * long to break bfqq's service guarantees. On
4308 * the bright side, large requests let the
4309 * drive reach a very high throughput, even if
4310 * there is only one in-flight large request
4313 if (blk_queue_nonrot(bfqd
->queue
) &&
4314 blk_rq_sectors(bfqq
->next_rq
) >=
4315 BFQQ_SECT_THR_NONROT
)
4316 limit
= min_t(unsigned int, 1, limit
);
4318 limit
= in_serv_bfqq
->inject_limit
;
4320 if (bfqd
->rq_in_driver
< limit
) {
4321 bfqd
->rqs_injected
= true;
4330 * Select a queue for service. If we have a current queue in service,
4331 * check whether to continue servicing it, or retrieve and set a new one.
4333 static struct bfq_queue
*bfq_select_queue(struct bfq_data
*bfqd
)
4335 struct bfq_queue
*bfqq
;
4336 struct request
*next_rq
;
4337 enum bfqq_expiration reason
= BFQQE_BUDGET_TIMEOUT
;
4339 bfqq
= bfqd
->in_service_queue
;
4343 bfq_log_bfqq(bfqd
, bfqq
, "select_queue: already in-service queue");
4346 * Do not expire bfqq for budget timeout if bfqq may be about
4347 * to enjoy device idling. The reason why, in this case, we
4348 * prevent bfqq from expiring is the same as in the comments
4349 * on the case where bfq_bfqq_must_idle() returns true, in
4350 * bfq_completed_request().
4352 if (bfq_may_expire_for_budg_timeout(bfqq
) &&
4353 !bfq_bfqq_must_idle(bfqq
))
4358 * This loop is rarely executed more than once. Even when it
4359 * happens, it is much more convenient to re-execute this loop
4360 * than to return NULL and trigger a new dispatch to get a
4363 next_rq
= bfqq
->next_rq
;
4365 * If bfqq has requests queued and it has enough budget left to
4366 * serve them, keep the queue, otherwise expire it.
4369 if (bfq_serv_to_charge(next_rq
, bfqq
) >
4370 bfq_bfqq_budget_left(bfqq
)) {
4372 * Expire the queue for budget exhaustion,
4373 * which makes sure that the next budget is
4374 * enough to serve the next request, even if
4375 * it comes from the fifo expired path.
4377 reason
= BFQQE_BUDGET_EXHAUSTED
;
4381 * The idle timer may be pending because we may
4382 * not disable disk idling even when a new request
4385 if (bfq_bfqq_wait_request(bfqq
)) {
4387 * If we get here: 1) at least a new request
4388 * has arrived but we have not disabled the
4389 * timer because the request was too small,
4390 * 2) then the block layer has unplugged
4391 * the device, causing the dispatch to be
4394 * Since the device is unplugged, now the
4395 * requests are probably large enough to
4396 * provide a reasonable throughput.
4397 * So we disable idling.
4399 bfq_clear_bfqq_wait_request(bfqq
);
4400 hrtimer_try_to_cancel(&bfqd
->idle_slice_timer
);
4407 * No requests pending. However, if the in-service queue is idling
4408 * for a new request, or has requests waiting for a completion and
4409 * may idle after their completion, then keep it anyway.
4411 * Yet, inject service from other queues if it boosts
4412 * throughput and is possible.
4414 if (bfq_bfqq_wait_request(bfqq
) ||
4415 (bfqq
->dispatched
!= 0 && bfq_better_to_idle(bfqq
))) {
4416 struct bfq_queue
*async_bfqq
=
4417 bfqq
->bic
&& bfqq
->bic
->bfqq
[0] &&
4418 bfq_bfqq_busy(bfqq
->bic
->bfqq
[0]) &&
4419 bfqq
->bic
->bfqq
[0]->next_rq
?
4420 bfqq
->bic
->bfqq
[0] : NULL
;
4423 * The next three mutually-exclusive ifs decide
4424 * whether to try injection, and choose the queue to
4425 * pick an I/O request from.
4427 * The first if checks whether the process associated
4428 * with bfqq has also async I/O pending. If so, it
4429 * injects such I/O unconditionally. Injecting async
4430 * I/O from the same process can cause no harm to the
4431 * process. On the contrary, it can only increase
4432 * bandwidth and reduce latency for the process.
4434 * The second if checks whether there happens to be a
4435 * non-empty waker queue for bfqq, i.e., a queue whose
4436 * I/O needs to be completed for bfqq to receive new
4437 * I/O. This happens, e.g., if bfqq is associated with
4438 * a process that does some sync. A sync generates
4439 * extra blocking I/O, which must be completed before
4440 * the process associated with bfqq can go on with its
4441 * I/O. If the I/O of the waker queue is not served,
4442 * then bfqq remains empty, and no I/O is dispatched,
4443 * until the idle timeout fires for bfqq. This is
4444 * likely to result in lower bandwidth and higher
4445 * latencies for bfqq, and in a severe loss of total
4446 * throughput. The best action to take is therefore to
4447 * serve the waker queue as soon as possible. So do it
4448 * (without relying on the third alternative below for
4449 * eventually serving waker_bfqq's I/O; see the last
4450 * paragraph for further details). This systematic
4451 * injection of I/O from the waker queue does not
4452 * cause any delay to bfqq's I/O. On the contrary,
4453 * next bfqq's I/O is brought forward dramatically,
4454 * for it is not blocked for milliseconds.
4456 * The third if checks whether bfqq is a queue for
4457 * which it is better to avoid injection. It is so if
4458 * bfqq delivers more throughput when served without
4459 * any further I/O from other queues in the middle, or
4460 * if the service times of bfqq's I/O requests both
4461 * count more than overall throughput, and may be
4462 * easily increased by injection (this happens if bfqq
4463 * has a short think time). If none of these
4464 * conditions holds, then a candidate queue for
4465 * injection is looked for through
4466 * bfq_choose_bfqq_for_injection(). Note that the
4467 * latter may return NULL (for example if the inject
4468 * limit for bfqq is currently 0).
4470 * NOTE: motivation for the second alternative
4472 * Thanks to the way the inject limit is updated in
4473 * bfq_update_has_short_ttime(), it is rather likely
4474 * that, if I/O is being plugged for bfqq and the
4475 * waker queue has pending I/O requests that are
4476 * blocking bfqq's I/O, then the third alternative
4477 * above lets the waker queue get served before the
4478 * I/O-plugging timeout fires. So one may deem the
4479 * second alternative superfluous. It is not, because
4480 * the third alternative may be way less effective in
4481 * case of a synchronization. For two main
4482 * reasons. First, throughput may be low because the
4483 * inject limit may be too low to guarantee the same
4484 * amount of injected I/O, from the waker queue or
4485 * other queues, that the second alternative
4486 * guarantees (the second alternative unconditionally
4487 * injects a pending I/O request of the waker queue
4488 * for each bfq_dispatch_request()). Second, with the
4489 * third alternative, the duration of the plugging,
4490 * i.e., the time before bfqq finally receives new I/O,
4491 * may not be minimized, because the waker queue may
4492 * happen to be served only after other queues.
4495 icq_to_bic(async_bfqq
->next_rq
->elv
.icq
) == bfqq
->bic
&&
4496 bfq_serv_to_charge(async_bfqq
->next_rq
, async_bfqq
) <=
4497 bfq_bfqq_budget_left(async_bfqq
))
4498 bfqq
= bfqq
->bic
->bfqq
[0];
4499 else if (bfq_bfqq_has_waker(bfqq
) &&
4500 bfq_bfqq_busy(bfqq
->waker_bfqq
) &&
4502 bfq_serv_to_charge(bfqq
->waker_bfqq
->next_rq
,
4503 bfqq
->waker_bfqq
) <=
4504 bfq_bfqq_budget_left(bfqq
->waker_bfqq
)
4506 bfqq
= bfqq
->waker_bfqq
;
4507 else if (!idling_boosts_thr_without_issues(bfqd
, bfqq
) &&
4508 (bfqq
->wr_coeff
== 1 || bfqd
->wr_busy_queues
> 1 ||
4509 !bfq_bfqq_has_short_ttime(bfqq
)))
4510 bfqq
= bfq_choose_bfqq_for_injection(bfqd
);
4517 reason
= BFQQE_NO_MORE_REQUESTS
;
4519 bfq_bfqq_expire(bfqd
, bfqq
, false, reason
);
4521 bfqq
= bfq_set_in_service_queue(bfqd
);
4523 bfq_log_bfqq(bfqd
, bfqq
, "select_queue: checking new queue");
4528 bfq_log_bfqq(bfqd
, bfqq
, "select_queue: returned this queue");
4530 bfq_log(bfqd
, "select_queue: no queue returned");
4535 static void bfq_update_wr_data(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
)
4537 struct bfq_entity
*entity
= &bfqq
->entity
;
4539 if (bfqq
->wr_coeff
> 1) { /* queue is being weight-raised */
4540 bfq_log_bfqq(bfqd
, bfqq
,
4541 "raising period dur %u/%u msec, old coeff %u, w %d(%d)",
4542 jiffies_to_msecs(jiffies
- bfqq
->last_wr_start_finish
),
4543 jiffies_to_msecs(bfqq
->wr_cur_max_time
),
4545 bfqq
->entity
.weight
, bfqq
->entity
.orig_weight
);
4547 if (entity
->prio_changed
)
4548 bfq_log_bfqq(bfqd
, bfqq
, "WARN: pending prio change");
4551 * If the queue was activated in a burst, or too much
4552 * time has elapsed from the beginning of this
4553 * weight-raising period, then end weight raising.
4555 if (bfq_bfqq_in_large_burst(bfqq
))
4556 bfq_bfqq_end_wr(bfqq
);
4557 else if (time_is_before_jiffies(bfqq
->last_wr_start_finish
+
4558 bfqq
->wr_cur_max_time
)) {
4559 if (bfqq
->wr_cur_max_time
!= bfqd
->bfq_wr_rt_max_time
||
4560 time_is_before_jiffies(bfqq
->wr_start_at_switch_to_srt
+
4561 bfq_wr_duration(bfqd
)))
4562 bfq_bfqq_end_wr(bfqq
);
4564 switch_back_to_interactive_wr(bfqq
, bfqd
);
4565 bfqq
->entity
.prio_changed
= 1;
4568 if (bfqq
->wr_coeff
> 1 &&
4569 bfqq
->wr_cur_max_time
!= bfqd
->bfq_wr_rt_max_time
&&
4570 bfqq
->service_from_wr
> max_service_from_wr
) {
4571 /* see comments on max_service_from_wr */
4572 bfq_bfqq_end_wr(bfqq
);
4576 * To improve latency (for this or other queues), immediately
4577 * update weight both if it must be raised and if it must be
4578 * lowered. Since, entity may be on some active tree here, and
4579 * might have a pending change of its ioprio class, invoke
4580 * next function with the last parameter unset (see the
4581 * comments on the function).
4583 if ((entity
->weight
> entity
->orig_weight
) != (bfqq
->wr_coeff
> 1))
4584 __bfq_entity_update_weight_prio(bfq_entity_service_tree(entity
),
4589 * Dispatch next request from bfqq.
4591 static struct request
*bfq_dispatch_rq_from_bfqq(struct bfq_data
*bfqd
,
4592 struct bfq_queue
*bfqq
)
4594 struct request
*rq
= bfqq
->next_rq
;
4595 unsigned long service_to_charge
;
4597 service_to_charge
= bfq_serv_to_charge(rq
, bfqq
);
4599 bfq_bfqq_served(bfqq
, service_to_charge
);
4601 if (bfqq
== bfqd
->in_service_queue
&& bfqd
->wait_dispatch
) {
4602 bfqd
->wait_dispatch
= false;
4603 bfqd
->waited_rq
= rq
;
4606 bfq_dispatch_remove(bfqd
->queue
, rq
);
4608 if (bfqq
!= bfqd
->in_service_queue
)
4612 * If weight raising has to terminate for bfqq, then next
4613 * function causes an immediate update of bfqq's weight,
4614 * without waiting for next activation. As a consequence, on
4615 * expiration, bfqq will be timestamped as if has never been
4616 * weight-raised during this service slot, even if it has
4617 * received part or even most of the service as a
4618 * weight-raised queue. This inflates bfqq's timestamps, which
4619 * is beneficial, as bfqq is then more willing to leave the
4620 * device immediately to possible other weight-raised queues.
4622 bfq_update_wr_data(bfqd
, bfqq
);
4625 * Expire bfqq, pretending that its budget expired, if bfqq
4626 * belongs to CLASS_IDLE and other queues are waiting for
4629 if (!(bfq_tot_busy_queues(bfqd
) > 1 && bfq_class_idle(bfqq
)))
4632 bfq_bfqq_expire(bfqd
, bfqq
, false, BFQQE_BUDGET_EXHAUSTED
);
4638 static bool bfq_has_work(struct blk_mq_hw_ctx
*hctx
)
4640 struct bfq_data
*bfqd
= hctx
->queue
->elevator
->elevator_data
;
4643 * Avoiding lock: a race on bfqd->busy_queues should cause at
4644 * most a call to dispatch for nothing
4646 return !list_empty_careful(&bfqd
->dispatch
) ||
4647 bfq_tot_busy_queues(bfqd
) > 0;
4650 static struct request
*__bfq_dispatch_request(struct blk_mq_hw_ctx
*hctx
)
4652 struct bfq_data
*bfqd
= hctx
->queue
->elevator
->elevator_data
;
4653 struct request
*rq
= NULL
;
4654 struct bfq_queue
*bfqq
= NULL
;
4656 if (!list_empty(&bfqd
->dispatch
)) {
4657 rq
= list_first_entry(&bfqd
->dispatch
, struct request
,
4659 list_del_init(&rq
->queuelist
);
4665 * Increment counters here, because this
4666 * dispatch does not follow the standard
4667 * dispatch flow (where counters are
4672 goto inc_in_driver_start_rq
;
4676 * We exploit the bfq_finish_requeue_request hook to
4677 * decrement rq_in_driver, but
4678 * bfq_finish_requeue_request will not be invoked on
4679 * this request. So, to avoid unbalance, just start
4680 * this request, without incrementing rq_in_driver. As
4681 * a negative consequence, rq_in_driver is deceptively
4682 * lower than it should be while this request is in
4683 * service. This may cause bfq_schedule_dispatch to be
4684 * invoked uselessly.
4686 * As for implementing an exact solution, the
4687 * bfq_finish_requeue_request hook, if defined, is
4688 * probably invoked also on this request. So, by
4689 * exploiting this hook, we could 1) increment
4690 * rq_in_driver here, and 2) decrement it in
4691 * bfq_finish_requeue_request. Such a solution would
4692 * let the value of the counter be always accurate,
4693 * but it would entail using an extra interface
4694 * function. This cost seems higher than the benefit,
4695 * being the frequency of non-elevator-private
4696 * requests very low.
4701 bfq_log(bfqd
, "dispatch requests: %d busy queues",
4702 bfq_tot_busy_queues(bfqd
));
4704 if (bfq_tot_busy_queues(bfqd
) == 0)
4708 * Force device to serve one request at a time if
4709 * strict_guarantees is true. Forcing this service scheme is
4710 * currently the ONLY way to guarantee that the request
4711 * service order enforced by the scheduler is respected by a
4712 * queueing device. Otherwise the device is free even to make
4713 * some unlucky request wait for as long as the device
4716 * Of course, serving one request at at time may cause loss of
4719 if (bfqd
->strict_guarantees
&& bfqd
->rq_in_driver
> 0)
4722 bfqq
= bfq_select_queue(bfqd
);
4726 rq
= bfq_dispatch_rq_from_bfqq(bfqd
, bfqq
);
4729 inc_in_driver_start_rq
:
4730 bfqd
->rq_in_driver
++;
4732 rq
->rq_flags
|= RQF_STARTED
;
4738 #ifdef CONFIG_BFQ_CGROUP_DEBUG
4739 static void bfq_update_dispatch_stats(struct request_queue
*q
,
4741 struct bfq_queue
*in_serv_queue
,
4742 bool idle_timer_disabled
)
4744 struct bfq_queue
*bfqq
= rq
? RQ_BFQQ(rq
) : NULL
;
4746 if (!idle_timer_disabled
&& !bfqq
)
4750 * rq and bfqq are guaranteed to exist until this function
4751 * ends, for the following reasons. First, rq can be
4752 * dispatched to the device, and then can be completed and
4753 * freed, only after this function ends. Second, rq cannot be
4754 * merged (and thus freed because of a merge) any longer,
4755 * because it has already started. Thus rq cannot be freed
4756 * before this function ends, and, since rq has a reference to
4757 * bfqq, the same guarantee holds for bfqq too.
4759 * In addition, the following queue lock guarantees that
4760 * bfqq_group(bfqq) exists as well.
4762 spin_lock_irq(&q
->queue_lock
);
4763 if (idle_timer_disabled
)
4765 * Since the idle timer has been disabled,
4766 * in_serv_queue contained some request when
4767 * __bfq_dispatch_request was invoked above, which
4768 * implies that rq was picked exactly from
4769 * in_serv_queue. Thus in_serv_queue == bfqq, and is
4770 * therefore guaranteed to exist because of the above
4773 bfqg_stats_update_idle_time(bfqq_group(in_serv_queue
));
4775 struct bfq_group
*bfqg
= bfqq_group(bfqq
);
4777 bfqg_stats_update_avg_queue_size(bfqg
);
4778 bfqg_stats_set_start_empty_time(bfqg
);
4779 bfqg_stats_update_io_remove(bfqg
, rq
->cmd_flags
);
4781 spin_unlock_irq(&q
->queue_lock
);
4784 static inline void bfq_update_dispatch_stats(struct request_queue
*q
,
4786 struct bfq_queue
*in_serv_queue
,
4787 bool idle_timer_disabled
) {}
4788 #endif /* CONFIG_BFQ_CGROUP_DEBUG */
4790 static struct request
*bfq_dispatch_request(struct blk_mq_hw_ctx
*hctx
)
4792 struct bfq_data
*bfqd
= hctx
->queue
->elevator
->elevator_data
;
4794 struct bfq_queue
*in_serv_queue
;
4795 bool waiting_rq
, idle_timer_disabled
;
4797 spin_lock_irq(&bfqd
->lock
);
4799 in_serv_queue
= bfqd
->in_service_queue
;
4800 waiting_rq
= in_serv_queue
&& bfq_bfqq_wait_request(in_serv_queue
);
4802 rq
= __bfq_dispatch_request(hctx
);
4804 idle_timer_disabled
=
4805 waiting_rq
&& !bfq_bfqq_wait_request(in_serv_queue
);
4807 spin_unlock_irq(&bfqd
->lock
);
4809 bfq_update_dispatch_stats(hctx
->queue
, rq
, in_serv_queue
,
4810 idle_timer_disabled
);
4816 * Task holds one reference to the queue, dropped when task exits. Each rq
4817 * in-flight on this queue also holds a reference, dropped when rq is freed.
4819 * Scheduler lock must be held here. Recall not to use bfqq after calling
4820 * this function on it.
4822 void bfq_put_queue(struct bfq_queue
*bfqq
)
4824 struct bfq_queue
*item
;
4825 struct hlist_node
*n
;
4826 struct bfq_group
*bfqg
= bfqq_group(bfqq
);
4829 bfq_log_bfqq(bfqq
->bfqd
, bfqq
, "put_queue: %p %d",
4836 if (!hlist_unhashed(&bfqq
->burst_list_node
)) {
4837 hlist_del_init(&bfqq
->burst_list_node
);
4839 * Decrement also burst size after the removal, if the
4840 * process associated with bfqq is exiting, and thus
4841 * does not contribute to the burst any longer. This
4842 * decrement helps filter out false positives of large
4843 * bursts, when some short-lived process (often due to
4844 * the execution of commands by some service) happens
4845 * to start and exit while a complex application is
4846 * starting, and thus spawning several processes that
4847 * do I/O (and that *must not* be treated as a large
4848 * burst, see comments on bfq_handle_burst).
4850 * In particular, the decrement is performed only if:
4851 * 1) bfqq is not a merged queue, because, if it is,
4852 * then this free of bfqq is not triggered by the exit
4853 * of the process bfqq is associated with, but exactly
4854 * by the fact that bfqq has just been merged.
4855 * 2) burst_size is greater than 0, to handle
4856 * unbalanced decrements. Unbalanced decrements may
4857 * happen in te following case: bfqq is inserted into
4858 * the current burst list--without incrementing
4859 * bust_size--because of a split, but the current
4860 * burst list is not the burst list bfqq belonged to
4861 * (see comments on the case of a split in
4864 if (bfqq
->bic
&& bfqq
->bfqd
->burst_size
> 0)
4865 bfqq
->bfqd
->burst_size
--;
4869 * bfqq does not exist any longer, so it cannot be woken by
4870 * any other queue, and cannot wake any other queue. Then bfqq
4871 * must be removed from the woken list of its possible waker
4872 * queue, and all queues in the woken list of bfqq must stop
4873 * having a waker queue. Strictly speaking, these updates
4874 * should be performed when bfqq remains with no I/O source
4875 * attached to it, which happens before bfqq gets freed. In
4876 * particular, this happens when the last process associated
4877 * with bfqq exits or gets associated with a different
4878 * queue. However, both events lead to bfqq being freed soon,
4879 * and dangling references would come out only after bfqq gets
4880 * freed. So these updates are done here, as a simple and safe
4881 * way to handle all cases.
4883 /* remove bfqq from woken list */
4884 if (!hlist_unhashed(&bfqq
->woken_list_node
))
4885 hlist_del_init(&bfqq
->woken_list_node
);
4887 /* reset waker for all queues in woken list */
4888 hlist_for_each_entry_safe(item
, n
, &bfqq
->woken_list
,
4890 item
->waker_bfqq
= NULL
;
4891 bfq_clear_bfqq_has_waker(item
);
4892 hlist_del_init(&item
->woken_list_node
);
4895 if (bfqq
->bfqd
&& bfqq
->bfqd
->last_completed_rq_bfqq
== bfqq
)
4896 bfqq
->bfqd
->last_completed_rq_bfqq
= NULL
;
4898 kmem_cache_free(bfq_pool
, bfqq
);
4899 bfqg_and_blkg_put(bfqg
);
4902 static void bfq_put_cooperator(struct bfq_queue
*bfqq
)
4904 struct bfq_queue
*__bfqq
, *next
;
4907 * If this queue was scheduled to merge with another queue, be
4908 * sure to drop the reference taken on that queue (and others in
4909 * the merge chain). See bfq_setup_merge and bfq_merge_bfqqs.
4911 __bfqq
= bfqq
->new_bfqq
;
4915 next
= __bfqq
->new_bfqq
;
4916 bfq_put_queue(__bfqq
);
4921 static void bfq_exit_bfqq(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
)
4923 if (bfqq
== bfqd
->in_service_queue
) {
4924 __bfq_bfqq_expire(bfqd
, bfqq
, BFQQE_BUDGET_TIMEOUT
);
4925 bfq_schedule_dispatch(bfqd
);
4928 bfq_log_bfqq(bfqd
, bfqq
, "exit_bfqq: %p, %d", bfqq
, bfqq
->ref
);
4930 bfq_put_cooperator(bfqq
);
4932 bfq_release_process_ref(bfqd
, bfqq
);
4935 static void bfq_exit_icq_bfqq(struct bfq_io_cq
*bic
, bool is_sync
)
4937 struct bfq_queue
*bfqq
= bic_to_bfqq(bic
, is_sync
);
4938 struct bfq_data
*bfqd
;
4941 bfqd
= bfqq
->bfqd
; /* NULL if scheduler already exited */
4944 unsigned long flags
;
4946 spin_lock_irqsave(&bfqd
->lock
, flags
);
4948 bfq_exit_bfqq(bfqd
, bfqq
);
4949 bic_set_bfqq(bic
, NULL
, is_sync
);
4950 spin_unlock_irqrestore(&bfqd
->lock
, flags
);
4954 static void bfq_exit_icq(struct io_cq
*icq
)
4956 struct bfq_io_cq
*bic
= icq_to_bic(icq
);
4958 bfq_exit_icq_bfqq(bic
, true);
4959 bfq_exit_icq_bfqq(bic
, false);
4963 * Update the entity prio values; note that the new values will not
4964 * be used until the next (re)activation.
4967 bfq_set_next_ioprio_data(struct bfq_queue
*bfqq
, struct bfq_io_cq
*bic
)
4969 struct task_struct
*tsk
= current
;
4971 struct bfq_data
*bfqd
= bfqq
->bfqd
;
4976 ioprio_class
= IOPRIO_PRIO_CLASS(bic
->ioprio
);
4977 switch (ioprio_class
) {
4979 dev_err(bfqq
->bfqd
->queue
->backing_dev_info
->dev
,
4980 "bfq: bad prio class %d\n", ioprio_class
);
4982 case IOPRIO_CLASS_NONE
:
4984 * No prio set, inherit CPU scheduling settings.
4986 bfqq
->new_ioprio
= task_nice_ioprio(tsk
);
4987 bfqq
->new_ioprio_class
= task_nice_ioclass(tsk
);
4989 case IOPRIO_CLASS_RT
:
4990 bfqq
->new_ioprio
= IOPRIO_PRIO_DATA(bic
->ioprio
);
4991 bfqq
->new_ioprio_class
= IOPRIO_CLASS_RT
;
4993 case IOPRIO_CLASS_BE
:
4994 bfqq
->new_ioprio
= IOPRIO_PRIO_DATA(bic
->ioprio
);
4995 bfqq
->new_ioprio_class
= IOPRIO_CLASS_BE
;
4997 case IOPRIO_CLASS_IDLE
:
4998 bfqq
->new_ioprio_class
= IOPRIO_CLASS_IDLE
;
4999 bfqq
->new_ioprio
= 7;
5003 if (bfqq
->new_ioprio
>= IOPRIO_BE_NR
) {
5004 pr_crit("bfq_set_next_ioprio_data: new_ioprio %d\n",
5006 bfqq
->new_ioprio
= IOPRIO_BE_NR
;
5009 bfqq
->entity
.new_weight
= bfq_ioprio_to_weight(bfqq
->new_ioprio
);
5010 bfqq
->entity
.prio_changed
= 1;
5013 static struct bfq_queue
*bfq_get_queue(struct bfq_data
*bfqd
,
5014 struct bio
*bio
, bool is_sync
,
5015 struct bfq_io_cq
*bic
);
5017 static void bfq_check_ioprio_change(struct bfq_io_cq
*bic
, struct bio
*bio
)
5019 struct bfq_data
*bfqd
= bic_to_bfqd(bic
);
5020 struct bfq_queue
*bfqq
;
5021 int ioprio
= bic
->icq
.ioc
->ioprio
;
5024 * This condition may trigger on a newly created bic, be sure to
5025 * drop the lock before returning.
5027 if (unlikely(!bfqd
) || likely(bic
->ioprio
== ioprio
))
5030 bic
->ioprio
= ioprio
;
5032 bfqq
= bic_to_bfqq(bic
, false);
5034 bfq_release_process_ref(bfqd
, bfqq
);
5035 bfqq
= bfq_get_queue(bfqd
, bio
, BLK_RW_ASYNC
, bic
);
5036 bic_set_bfqq(bic
, bfqq
, false);
5039 bfqq
= bic_to_bfqq(bic
, true);
5041 bfq_set_next_ioprio_data(bfqq
, bic
);
5044 static void bfq_init_bfqq(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
,
5045 struct bfq_io_cq
*bic
, pid_t pid
, int is_sync
)
5047 RB_CLEAR_NODE(&bfqq
->entity
.rb_node
);
5048 INIT_LIST_HEAD(&bfqq
->fifo
);
5049 INIT_HLIST_NODE(&bfqq
->burst_list_node
);
5050 INIT_HLIST_NODE(&bfqq
->woken_list_node
);
5051 INIT_HLIST_HEAD(&bfqq
->woken_list
);
5057 bfq_set_next_ioprio_data(bfqq
, bic
);
5061 * No need to mark as has_short_ttime if in
5062 * idle_class, because no device idling is performed
5063 * for queues in idle class
5065 if (!bfq_class_idle(bfqq
))
5066 /* tentatively mark as has_short_ttime */
5067 bfq_mark_bfqq_has_short_ttime(bfqq
);
5068 bfq_mark_bfqq_sync(bfqq
);
5069 bfq_mark_bfqq_just_created(bfqq
);
5071 bfq_clear_bfqq_sync(bfqq
);
5073 /* set end request to minus infinity from now */
5074 bfqq
->ttime
.last_end_request
= ktime_get_ns() + 1;
5076 bfq_mark_bfqq_IO_bound(bfqq
);
5080 /* Tentative initial value to trade off between thr and lat */
5081 bfqq
->max_budget
= (2 * bfq_max_budget(bfqd
)) / 3;
5082 bfqq
->budget_timeout
= bfq_smallest_from_now();
5085 bfqq
->last_wr_start_finish
= jiffies
;
5086 bfqq
->wr_start_at_switch_to_srt
= bfq_smallest_from_now();
5087 bfqq
->split_time
= bfq_smallest_from_now();
5090 * To not forget the possibly high bandwidth consumed by a
5091 * process/queue in the recent past,
5092 * bfq_bfqq_softrt_next_start() returns a value at least equal
5093 * to the current value of bfqq->soft_rt_next_start (see
5094 * comments on bfq_bfqq_softrt_next_start). Set
5095 * soft_rt_next_start to now, to mean that bfqq has consumed
5096 * no bandwidth so far.
5098 bfqq
->soft_rt_next_start
= jiffies
;
5100 /* first request is almost certainly seeky */
5101 bfqq
->seek_history
= 1;
5104 static struct bfq_queue
**bfq_async_queue_prio(struct bfq_data
*bfqd
,
5105 struct bfq_group
*bfqg
,
5106 int ioprio_class
, int ioprio
)
5108 switch (ioprio_class
) {
5109 case IOPRIO_CLASS_RT
:
5110 return &bfqg
->async_bfqq
[0][ioprio
];
5111 case IOPRIO_CLASS_NONE
:
5112 ioprio
= IOPRIO_NORM
;
5114 case IOPRIO_CLASS_BE
:
5115 return &bfqg
->async_bfqq
[1][ioprio
];
5116 case IOPRIO_CLASS_IDLE
:
5117 return &bfqg
->async_idle_bfqq
;
5123 static struct bfq_queue
*bfq_get_queue(struct bfq_data
*bfqd
,
5124 struct bio
*bio
, bool is_sync
,
5125 struct bfq_io_cq
*bic
)
5127 const int ioprio
= IOPRIO_PRIO_DATA(bic
->ioprio
);
5128 const int ioprio_class
= IOPRIO_PRIO_CLASS(bic
->ioprio
);
5129 struct bfq_queue
**async_bfqq
= NULL
;
5130 struct bfq_queue
*bfqq
;
5131 struct bfq_group
*bfqg
;
5135 bfqg
= bfq_find_set_group(bfqd
, __bio_blkcg(bio
));
5137 bfqq
= &bfqd
->oom_bfqq
;
5142 async_bfqq
= bfq_async_queue_prio(bfqd
, bfqg
, ioprio_class
,
5149 bfqq
= kmem_cache_alloc_node(bfq_pool
,
5150 GFP_NOWAIT
| __GFP_ZERO
| __GFP_NOWARN
,
5154 bfq_init_bfqq(bfqd
, bfqq
, bic
, current
->pid
,
5156 bfq_init_entity(&bfqq
->entity
, bfqg
);
5157 bfq_log_bfqq(bfqd
, bfqq
, "allocated");
5159 bfqq
= &bfqd
->oom_bfqq
;
5160 bfq_log_bfqq(bfqd
, bfqq
, "using oom bfqq");
5165 * Pin the queue now that it's allocated, scheduler exit will
5170 * Extra group reference, w.r.t. sync
5171 * queue. This extra reference is removed
5172 * only if bfqq->bfqg disappears, to
5173 * guarantee that this queue is not freed
5174 * until its group goes away.
5176 bfq_log_bfqq(bfqd
, bfqq
, "get_queue, bfqq not in async: %p, %d",
5182 bfqq
->ref
++; /* get a process reference to this queue */
5183 bfq_log_bfqq(bfqd
, bfqq
, "get_queue, at end: %p, %d", bfqq
, bfqq
->ref
);
5188 static void bfq_update_io_thinktime(struct bfq_data
*bfqd
,
5189 struct bfq_queue
*bfqq
)
5191 struct bfq_ttime
*ttime
= &bfqq
->ttime
;
5192 u64 elapsed
= ktime_get_ns() - bfqq
->ttime
.last_end_request
;
5194 elapsed
= min_t(u64
, elapsed
, 2ULL * bfqd
->bfq_slice_idle
);
5196 ttime
->ttime_samples
= (7*bfqq
->ttime
.ttime_samples
+ 256) / 8;
5197 ttime
->ttime_total
= div_u64(7*ttime
->ttime_total
+ 256*elapsed
, 8);
5198 ttime
->ttime_mean
= div64_ul(ttime
->ttime_total
+ 128,
5199 ttime
->ttime_samples
);
5203 bfq_update_io_seektime(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
,
5206 bfqq
->seek_history
<<= 1;
5207 bfqq
->seek_history
|= BFQ_RQ_SEEKY(bfqd
, bfqq
->last_request_pos
, rq
);
5209 if (bfqq
->wr_coeff
> 1 &&
5210 bfqq
->wr_cur_max_time
== bfqd
->bfq_wr_rt_max_time
&&
5211 BFQQ_TOTALLY_SEEKY(bfqq
))
5212 bfq_bfqq_end_wr(bfqq
);
5215 static void bfq_update_has_short_ttime(struct bfq_data
*bfqd
,
5216 struct bfq_queue
*bfqq
,
5217 struct bfq_io_cq
*bic
)
5219 bool has_short_ttime
= true, state_changed
;
5222 * No need to update has_short_ttime if bfqq is async or in
5223 * idle io prio class, or if bfq_slice_idle is zero, because
5224 * no device idling is performed for bfqq in this case.
5226 if (!bfq_bfqq_sync(bfqq
) || bfq_class_idle(bfqq
) ||
5227 bfqd
->bfq_slice_idle
== 0)
5230 /* Idle window just restored, statistics are meaningless. */
5231 if (time_is_after_eq_jiffies(bfqq
->split_time
+
5232 bfqd
->bfq_wr_min_idle_time
))
5235 /* Think time is infinite if no process is linked to
5236 * bfqq. Otherwise check average think time to
5237 * decide whether to mark as has_short_ttime
5239 if (atomic_read(&bic
->icq
.ioc
->active_ref
) == 0 ||
5240 (bfq_sample_valid(bfqq
->ttime
.ttime_samples
) &&
5241 bfqq
->ttime
.ttime_mean
> bfqd
->bfq_slice_idle
))
5242 has_short_ttime
= false;
5244 state_changed
= has_short_ttime
!= bfq_bfqq_has_short_ttime(bfqq
);
5246 if (has_short_ttime
)
5247 bfq_mark_bfqq_has_short_ttime(bfqq
);
5249 bfq_clear_bfqq_has_short_ttime(bfqq
);
5252 * Until the base value for the total service time gets
5253 * finally computed for bfqq, the inject limit does depend on
5254 * the think-time state (short|long). In particular, the limit
5255 * is 0 or 1 if the think time is deemed, respectively, as
5256 * short or long (details in the comments in
5257 * bfq_update_inject_limit()). Accordingly, the next
5258 * instructions reset the inject limit if the think-time state
5259 * has changed and the above base value is still to be
5262 * However, the reset is performed only if more than 100 ms
5263 * have elapsed since the last update of the inject limit, or
5264 * (inclusive) if the change is from short to long think
5265 * time. The reason for this waiting is as follows.
5267 * bfqq may have a long think time because of a
5268 * synchronization with some other queue, i.e., because the
5269 * I/O of some other queue may need to be completed for bfqq
5270 * to receive new I/O. Details in the comments on the choice
5271 * of the queue for injection in bfq_select_queue().
5273 * As stressed in those comments, if such a synchronization is
5274 * actually in place, then, without injection on bfqq, the
5275 * blocking I/O cannot happen to served while bfqq is in
5276 * service. As a consequence, if bfqq is granted
5277 * I/O-dispatch-plugging, then bfqq remains empty, and no I/O
5278 * is dispatched, until the idle timeout fires. This is likely
5279 * to result in lower bandwidth and higher latencies for bfqq,
5280 * and in a severe loss of total throughput.
5282 * On the opposite end, a non-zero inject limit may allow the
5283 * I/O that blocks bfqq to be executed soon, and therefore
5284 * bfqq to receive new I/O soon.
5286 * But, if the blocking gets actually eliminated, then the
5287 * next think-time sample for bfqq may be very low. This in
5288 * turn may cause bfqq's think time to be deemed
5289 * short. Without the 100 ms barrier, this new state change
5290 * would cause the body of the next if to be executed
5291 * immediately. But this would set to 0 the inject
5292 * limit. Without injection, the blocking I/O would cause the
5293 * think time of bfqq to become long again, and therefore the
5294 * inject limit to be raised again, and so on. The only effect
5295 * of such a steady oscillation between the two think-time
5296 * states would be to prevent effective injection on bfqq.
5298 * In contrast, if the inject limit is not reset during such a
5299 * long time interval as 100 ms, then the number of short
5300 * think time samples can grow significantly before the reset
5301 * is performed. As a consequence, the think time state can
5302 * become stable before the reset. Therefore there will be no
5303 * state change when the 100 ms elapse, and no reset of the
5304 * inject limit. The inject limit remains steadily equal to 1
5305 * both during and after the 100 ms. So injection can be
5306 * performed at all times, and throughput gets boosted.
5308 * An inject limit equal to 1 is however in conflict, in
5309 * general, with the fact that the think time of bfqq is
5310 * short, because injection may be likely to delay bfqq's I/O
5311 * (as explained in the comments in
5312 * bfq_update_inject_limit()). But this does not happen in
5313 * this special case, because bfqq's low think time is due to
5314 * an effective handling of a synchronization, through
5315 * injection. In this special case, bfqq's I/O does not get
5316 * delayed by injection; on the contrary, bfqq's I/O is
5317 * brought forward, because it is not blocked for
5320 * In addition, serving the blocking I/O much sooner, and much
5321 * more frequently than once per I/O-plugging timeout, makes
5322 * it much quicker to detect a waker queue (the concept of
5323 * waker queue is defined in the comments in
5324 * bfq_add_request()). This makes it possible to start sooner
5325 * to boost throughput more effectively, by injecting the I/O
5326 * of the waker queue unconditionally on every
5327 * bfq_dispatch_request().
5329 * One last, important benefit of not resetting the inject
5330 * limit before 100 ms is that, during this time interval, the
5331 * base value for the total service time is likely to get
5332 * finally computed for bfqq, freeing the inject limit from
5333 * its relation with the think time.
5335 if (state_changed
&& bfqq
->last_serv_time_ns
== 0 &&
5336 (time_is_before_eq_jiffies(bfqq
->decrease_time_jif
+
5337 msecs_to_jiffies(100)) ||
5339 bfq_reset_inject_limit(bfqd
, bfqq
);
5343 * Called when a new fs request (rq) is added to bfqq. Check if there's
5344 * something we should do about it.
5346 static void bfq_rq_enqueued(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
,
5349 if (rq
->cmd_flags
& REQ_META
)
5350 bfqq
->meta_pending
++;
5352 bfqq
->last_request_pos
= blk_rq_pos(rq
) + blk_rq_sectors(rq
);
5354 if (bfqq
== bfqd
->in_service_queue
&& bfq_bfqq_wait_request(bfqq
)) {
5355 bool small_req
= bfqq
->queued
[rq_is_sync(rq
)] == 1 &&
5356 blk_rq_sectors(rq
) < 32;
5357 bool budget_timeout
= bfq_bfqq_budget_timeout(bfqq
);
5360 * There is just this request queued: if
5361 * - the request is small, and
5362 * - we are idling to boost throughput, and
5363 * - the queue is not to be expired,
5366 * In this way, if the device is being idled to wait
5367 * for a new request from the in-service queue, we
5368 * avoid unplugging the device and committing the
5369 * device to serve just a small request. In contrast
5370 * we wait for the block layer to decide when to
5371 * unplug the device: hopefully, new requests will be
5372 * merged to this one quickly, then the device will be
5373 * unplugged and larger requests will be dispatched.
5375 if (small_req
&& idling_boosts_thr_without_issues(bfqd
, bfqq
) &&
5380 * A large enough request arrived, or idling is being
5381 * performed to preserve service guarantees, or
5382 * finally the queue is to be expired: in all these
5383 * cases disk idling is to be stopped, so clear
5384 * wait_request flag and reset timer.
5386 bfq_clear_bfqq_wait_request(bfqq
);
5387 hrtimer_try_to_cancel(&bfqd
->idle_slice_timer
);
5390 * The queue is not empty, because a new request just
5391 * arrived. Hence we can safely expire the queue, in
5392 * case of budget timeout, without risking that the
5393 * timestamps of the queue are not updated correctly.
5394 * See [1] for more details.
5397 bfq_bfqq_expire(bfqd
, bfqq
, false,
5398 BFQQE_BUDGET_TIMEOUT
);
5402 /* returns true if it causes the idle timer to be disabled */
5403 static bool __bfq_insert_request(struct bfq_data
*bfqd
, struct request
*rq
)
5405 struct bfq_queue
*bfqq
= RQ_BFQQ(rq
),
5406 *new_bfqq
= bfq_setup_cooperator(bfqd
, bfqq
, rq
, true);
5407 bool waiting
, idle_timer_disabled
= false;
5411 * Release the request's reference to the old bfqq
5412 * and make sure one is taken to the shared queue.
5414 new_bfqq
->allocated
++;
5418 * If the bic associated with the process
5419 * issuing this request still points to bfqq
5420 * (and thus has not been already redirected
5421 * to new_bfqq or even some other bfq_queue),
5422 * then complete the merge and redirect it to
5425 if (bic_to_bfqq(RQ_BIC(rq
), 1) == bfqq
)
5426 bfq_merge_bfqqs(bfqd
, RQ_BIC(rq
),
5429 bfq_clear_bfqq_just_created(bfqq
);
5431 * rq is about to be enqueued into new_bfqq,
5432 * release rq reference on bfqq
5434 bfq_put_queue(bfqq
);
5435 rq
->elv
.priv
[1] = new_bfqq
;
5439 bfq_update_io_thinktime(bfqd
, bfqq
);
5440 bfq_update_has_short_ttime(bfqd
, bfqq
, RQ_BIC(rq
));
5441 bfq_update_io_seektime(bfqd
, bfqq
, rq
);
5443 waiting
= bfqq
&& bfq_bfqq_wait_request(bfqq
);
5444 bfq_add_request(rq
);
5445 idle_timer_disabled
= waiting
&& !bfq_bfqq_wait_request(bfqq
);
5447 rq
->fifo_time
= ktime_get_ns() + bfqd
->bfq_fifo_expire
[rq_is_sync(rq
)];
5448 list_add_tail(&rq
->queuelist
, &bfqq
->fifo
);
5450 bfq_rq_enqueued(bfqd
, bfqq
, rq
);
5452 return idle_timer_disabled
;
5455 #ifdef CONFIG_BFQ_CGROUP_DEBUG
5456 static void bfq_update_insert_stats(struct request_queue
*q
,
5457 struct bfq_queue
*bfqq
,
5458 bool idle_timer_disabled
,
5459 unsigned int cmd_flags
)
5465 * bfqq still exists, because it can disappear only after
5466 * either it is merged with another queue, or the process it
5467 * is associated with exits. But both actions must be taken by
5468 * the same process currently executing this flow of
5471 * In addition, the following queue lock guarantees that
5472 * bfqq_group(bfqq) exists as well.
5474 spin_lock_irq(&q
->queue_lock
);
5475 bfqg_stats_update_io_add(bfqq_group(bfqq
), bfqq
, cmd_flags
);
5476 if (idle_timer_disabled
)
5477 bfqg_stats_update_idle_time(bfqq_group(bfqq
));
5478 spin_unlock_irq(&q
->queue_lock
);
5481 static inline void bfq_update_insert_stats(struct request_queue
*q
,
5482 struct bfq_queue
*bfqq
,
5483 bool idle_timer_disabled
,
5484 unsigned int cmd_flags
) {}
5485 #endif /* CONFIG_BFQ_CGROUP_DEBUG */
5487 static void bfq_insert_request(struct blk_mq_hw_ctx
*hctx
, struct request
*rq
,
5490 struct request_queue
*q
= hctx
->queue
;
5491 struct bfq_data
*bfqd
= q
->elevator
->elevator_data
;
5492 struct bfq_queue
*bfqq
;
5493 bool idle_timer_disabled
= false;
5494 unsigned int cmd_flags
;
5496 #ifdef CONFIG_BFQ_GROUP_IOSCHED
5497 if (!cgroup_subsys_on_dfl(io_cgrp_subsys
) && rq
->bio
)
5498 bfqg_stats_update_legacy_io(q
, rq
);
5500 spin_lock_irq(&bfqd
->lock
);
5501 if (blk_mq_sched_try_insert_merge(q
, rq
)) {
5502 spin_unlock_irq(&bfqd
->lock
);
5506 spin_unlock_irq(&bfqd
->lock
);
5508 blk_mq_sched_request_inserted(rq
);
5510 spin_lock_irq(&bfqd
->lock
);
5511 bfqq
= bfq_init_rq(rq
);
5512 if (!bfqq
|| at_head
|| blk_rq_is_passthrough(rq
)) {
5514 list_add(&rq
->queuelist
, &bfqd
->dispatch
);
5516 list_add_tail(&rq
->queuelist
, &bfqd
->dispatch
);
5518 idle_timer_disabled
= __bfq_insert_request(bfqd
, rq
);
5520 * Update bfqq, because, if a queue merge has occurred
5521 * in __bfq_insert_request, then rq has been
5522 * redirected into a new queue.
5526 if (rq_mergeable(rq
)) {
5527 elv_rqhash_add(q
, rq
);
5534 * Cache cmd_flags before releasing scheduler lock, because rq
5535 * may disappear afterwards (for example, because of a request
5538 cmd_flags
= rq
->cmd_flags
;
5540 spin_unlock_irq(&bfqd
->lock
);
5542 bfq_update_insert_stats(q
, bfqq
, idle_timer_disabled
,
5546 static void bfq_insert_requests(struct blk_mq_hw_ctx
*hctx
,
5547 struct list_head
*list
, bool at_head
)
5549 while (!list_empty(list
)) {
5552 rq
= list_first_entry(list
, struct request
, queuelist
);
5553 list_del_init(&rq
->queuelist
);
5554 bfq_insert_request(hctx
, rq
, at_head
);
5558 static void bfq_update_hw_tag(struct bfq_data
*bfqd
)
5560 struct bfq_queue
*bfqq
= bfqd
->in_service_queue
;
5562 bfqd
->max_rq_in_driver
= max_t(int, bfqd
->max_rq_in_driver
,
5563 bfqd
->rq_in_driver
);
5565 if (bfqd
->hw_tag
== 1)
5569 * This sample is valid if the number of outstanding requests
5570 * is large enough to allow a queueing behavior. Note that the
5571 * sum is not exact, as it's not taking into account deactivated
5574 if (bfqd
->rq_in_driver
+ bfqd
->queued
<= BFQ_HW_QUEUE_THRESHOLD
)
5578 * If active queue hasn't enough requests and can idle, bfq might not
5579 * dispatch sufficient requests to hardware. Don't zero hw_tag in this
5582 if (bfqq
&& bfq_bfqq_has_short_ttime(bfqq
) &&
5583 bfqq
->dispatched
+ bfqq
->queued
[0] + bfqq
->queued
[1] <
5584 BFQ_HW_QUEUE_THRESHOLD
&&
5585 bfqd
->rq_in_driver
< BFQ_HW_QUEUE_THRESHOLD
)
5588 if (bfqd
->hw_tag_samples
++ < BFQ_HW_QUEUE_SAMPLES
)
5591 bfqd
->hw_tag
= bfqd
->max_rq_in_driver
> BFQ_HW_QUEUE_THRESHOLD
;
5592 bfqd
->max_rq_in_driver
= 0;
5593 bfqd
->hw_tag_samples
= 0;
5595 bfqd
->nonrot_with_queueing
=
5596 blk_queue_nonrot(bfqd
->queue
) && bfqd
->hw_tag
;
5599 static void bfq_completed_request(struct bfq_queue
*bfqq
, struct bfq_data
*bfqd
)
5604 bfq_update_hw_tag(bfqd
);
5606 bfqd
->rq_in_driver
--;
5609 if (!bfqq
->dispatched
&& !bfq_bfqq_busy(bfqq
)) {
5611 * Set budget_timeout (which we overload to store the
5612 * time at which the queue remains with no backlog and
5613 * no outstanding request; used by the weight-raising
5616 bfqq
->budget_timeout
= jiffies
;
5618 bfq_weights_tree_remove(bfqd
, bfqq
);
5621 now_ns
= ktime_get_ns();
5623 bfqq
->ttime
.last_end_request
= now_ns
;
5626 * Using us instead of ns, to get a reasonable precision in
5627 * computing rate in next check.
5629 delta_us
= div_u64(now_ns
- bfqd
->last_completion
, NSEC_PER_USEC
);
5632 * If the request took rather long to complete, and, according
5633 * to the maximum request size recorded, this completion latency
5634 * implies that the request was certainly served at a very low
5635 * rate (less than 1M sectors/sec), then the whole observation
5636 * interval that lasts up to this time instant cannot be a
5637 * valid time interval for computing a new peak rate. Invoke
5638 * bfq_update_rate_reset to have the following three steps
5640 * - close the observation interval at the last (previous)
5641 * request dispatch or completion
5642 * - compute rate, if possible, for that observation interval
5643 * - reset to zero samples, which will trigger a proper
5644 * re-initialization of the observation interval on next
5647 if (delta_us
> BFQ_MIN_TT
/NSEC_PER_USEC
&&
5648 (bfqd
->last_rq_max_size
<<BFQ_RATE_SHIFT
)/delta_us
<
5649 1UL<<(BFQ_RATE_SHIFT
- 10))
5650 bfq_update_rate_reset(bfqd
, NULL
);
5651 bfqd
->last_completion
= now_ns
;
5652 bfqd
->last_completed_rq_bfqq
= bfqq
;
5655 * If we are waiting to discover whether the request pattern
5656 * of the task associated with the queue is actually
5657 * isochronous, and both requisites for this condition to hold
5658 * are now satisfied, then compute soft_rt_next_start (see the
5659 * comments on the function bfq_bfqq_softrt_next_start()). We
5660 * do not compute soft_rt_next_start if bfqq is in interactive
5661 * weight raising (see the comments in bfq_bfqq_expire() for
5662 * an explanation). We schedule this delayed update when bfqq
5663 * expires, if it still has in-flight requests.
5665 if (bfq_bfqq_softrt_update(bfqq
) && bfqq
->dispatched
== 0 &&
5666 RB_EMPTY_ROOT(&bfqq
->sort_list
) &&
5667 bfqq
->wr_coeff
!= bfqd
->bfq_wr_coeff
)
5668 bfqq
->soft_rt_next_start
=
5669 bfq_bfqq_softrt_next_start(bfqd
, bfqq
);
5672 * If this is the in-service queue, check if it needs to be expired,
5673 * or if we want to idle in case it has no pending requests.
5675 if (bfqd
->in_service_queue
== bfqq
) {
5676 if (bfq_bfqq_must_idle(bfqq
)) {
5677 if (bfqq
->dispatched
== 0)
5678 bfq_arm_slice_timer(bfqd
);
5680 * If we get here, we do not expire bfqq, even
5681 * if bfqq was in budget timeout or had no
5682 * more requests (as controlled in the next
5683 * conditional instructions). The reason for
5684 * not expiring bfqq is as follows.
5686 * Here bfqq->dispatched > 0 holds, but
5687 * bfq_bfqq_must_idle() returned true. This
5688 * implies that, even if no request arrives
5689 * for bfqq before bfqq->dispatched reaches 0,
5690 * bfqq will, however, not be expired on the
5691 * completion event that causes bfqq->dispatch
5692 * to reach zero. In contrast, on this event,
5693 * bfqq will start enjoying device idling
5694 * (I/O-dispatch plugging).
5696 * But, if we expired bfqq here, bfqq would
5697 * not have the chance to enjoy device idling
5698 * when bfqq->dispatched finally reaches
5699 * zero. This would expose bfqq to violation
5700 * of its reserved service guarantees.
5703 } else if (bfq_may_expire_for_budg_timeout(bfqq
))
5704 bfq_bfqq_expire(bfqd
, bfqq
, false,
5705 BFQQE_BUDGET_TIMEOUT
);
5706 else if (RB_EMPTY_ROOT(&bfqq
->sort_list
) &&
5707 (bfqq
->dispatched
== 0 ||
5708 !bfq_better_to_idle(bfqq
)))
5709 bfq_bfqq_expire(bfqd
, bfqq
, false,
5710 BFQQE_NO_MORE_REQUESTS
);
5713 if (!bfqd
->rq_in_driver
)
5714 bfq_schedule_dispatch(bfqd
);
5717 static void bfq_finish_requeue_request_body(struct bfq_queue
*bfqq
)
5721 bfq_put_queue(bfqq
);
5725 * The processes associated with bfqq may happen to generate their
5726 * cumulative I/O at a lower rate than the rate at which the device
5727 * could serve the same I/O. This is rather probable, e.g., if only
5728 * one process is associated with bfqq and the device is an SSD. It
5729 * results in bfqq becoming often empty while in service. In this
5730 * respect, if BFQ is allowed to switch to another queue when bfqq
5731 * remains empty, then the device goes on being fed with I/O requests,
5732 * and the throughput is not affected. In contrast, if BFQ is not
5733 * allowed to switch to another queue---because bfqq is sync and
5734 * I/O-dispatch needs to be plugged while bfqq is temporarily
5735 * empty---then, during the service of bfqq, there will be frequent
5736 * "service holes", i.e., time intervals during which bfqq gets empty
5737 * and the device can only consume the I/O already queued in its
5738 * hardware queues. During service holes, the device may even get to
5739 * remaining idle. In the end, during the service of bfqq, the device
5740 * is driven at a lower speed than the one it can reach with the kind
5741 * of I/O flowing through bfqq.
5743 * To counter this loss of throughput, BFQ implements a "request
5744 * injection mechanism", which tries to fill the above service holes
5745 * with I/O requests taken from other queues. The hard part in this
5746 * mechanism is finding the right amount of I/O to inject, so as to
5747 * both boost throughput and not break bfqq's bandwidth and latency
5748 * guarantees. In this respect, the mechanism maintains a per-queue
5749 * inject limit, computed as below. While bfqq is empty, the injection
5750 * mechanism dispatches extra I/O requests only until the total number
5751 * of I/O requests in flight---i.e., already dispatched but not yet
5752 * completed---remains lower than this limit.
5754 * A first definition comes in handy to introduce the algorithm by
5755 * which the inject limit is computed. We define as first request for
5756 * bfqq, an I/O request for bfqq that arrives while bfqq is in
5757 * service, and causes bfqq to switch from empty to non-empty. The
5758 * algorithm updates the limit as a function of the effect of
5759 * injection on the service times of only the first requests of
5760 * bfqq. The reason for this restriction is that these are the
5761 * requests whose service time is affected most, because they are the
5762 * first to arrive after injection possibly occurred.
5764 * To evaluate the effect of injection, the algorithm measures the
5765 * "total service time" of first requests. We define as total service
5766 * time of an I/O request, the time that elapses since when the
5767 * request is enqueued into bfqq, to when it is completed. This
5768 * quantity allows the whole effect of injection to be measured. It is
5769 * easy to see why. Suppose that some requests of other queues are
5770 * actually injected while bfqq is empty, and that a new request R
5771 * then arrives for bfqq. If the device does start to serve all or
5772 * part of the injected requests during the service hole, then,
5773 * because of this extra service, it may delay the next invocation of
5774 * the dispatch hook of BFQ. Then, even after R gets eventually
5775 * dispatched, the device may delay the actual service of R if it is
5776 * still busy serving the extra requests, or if it decides to serve,
5777 * before R, some extra request still present in its queues. As a
5778 * conclusion, the cumulative extra delay caused by injection can be
5779 * easily evaluated by just comparing the total service time of first
5780 * requests with and without injection.
5782 * The limit-update algorithm works as follows. On the arrival of a
5783 * first request of bfqq, the algorithm measures the total time of the
5784 * request only if one of the three cases below holds, and, for each
5785 * case, it updates the limit as described below:
5787 * (1) If there is no in-flight request. This gives a baseline for the
5788 * total service time of the requests of bfqq. If the baseline has
5789 * not been computed yet, then, after computing it, the limit is
5790 * set to 1, to start boosting throughput, and to prepare the
5791 * ground for the next case. If the baseline has already been
5792 * computed, then it is updated, in case it results to be lower
5793 * than the previous value.
5795 * (2) If the limit is higher than 0 and there are in-flight
5796 * requests. By comparing the total service time in this case with
5797 * the above baseline, it is possible to know at which extent the
5798 * current value of the limit is inflating the total service
5799 * time. If the inflation is below a certain threshold, then bfqq
5800 * is assumed to be suffering from no perceivable loss of its
5801 * service guarantees, and the limit is even tentatively
5802 * increased. If the inflation is above the threshold, then the
5803 * limit is decreased. Due to the lack of any hysteresis, this
5804 * logic makes the limit oscillate even in steady workload
5805 * conditions. Yet we opted for it, because it is fast in reaching
5806 * the best value for the limit, as a function of the current I/O
5807 * workload. To reduce oscillations, this step is disabled for a
5808 * short time interval after the limit happens to be decreased.
5810 * (3) Periodically, after resetting the limit, to make sure that the
5811 * limit eventually drops in case the workload changes. This is
5812 * needed because, after the limit has gone safely up for a
5813 * certain workload, it is impossible to guess whether the
5814 * baseline total service time may have changed, without measuring
5815 * it again without injection. A more effective version of this
5816 * step might be to just sample the baseline, by interrupting
5817 * injection only once, and then to reset/lower the limit only if
5818 * the total service time with the current limit does happen to be
5821 * More details on each step are provided in the comments on the
5822 * pieces of code that implement these steps: the branch handling the
5823 * transition from empty to non empty in bfq_add_request(), the branch
5824 * handling injection in bfq_select_queue(), and the function
5825 * bfq_choose_bfqq_for_injection(). These comments also explain some
5826 * exceptions, made by the injection mechanism in some special cases.
5828 static void bfq_update_inject_limit(struct bfq_data
*bfqd
,
5829 struct bfq_queue
*bfqq
)
5831 u64 tot_time_ns
= ktime_get_ns() - bfqd
->last_empty_occupied_ns
;
5832 unsigned int old_limit
= bfqq
->inject_limit
;
5834 if (bfqq
->last_serv_time_ns
> 0 && bfqd
->rqs_injected
) {
5835 u64 threshold
= (bfqq
->last_serv_time_ns
* 3)>>1;
5837 if (tot_time_ns
>= threshold
&& old_limit
> 0) {
5838 bfqq
->inject_limit
--;
5839 bfqq
->decrease_time_jif
= jiffies
;
5840 } else if (tot_time_ns
< threshold
&&
5841 old_limit
<= bfqd
->max_rq_in_driver
)
5842 bfqq
->inject_limit
++;
5846 * Either we still have to compute the base value for the
5847 * total service time, and there seem to be the right
5848 * conditions to do it, or we can lower the last base value
5851 * NOTE: (bfqd->rq_in_driver == 1) means that there is no I/O
5852 * request in flight, because this function is in the code
5853 * path that handles the completion of a request of bfqq, and,
5854 * in particular, this function is executed before
5855 * bfqd->rq_in_driver is decremented in such a code path.
5857 if ((bfqq
->last_serv_time_ns
== 0 && bfqd
->rq_in_driver
== 1) ||
5858 tot_time_ns
< bfqq
->last_serv_time_ns
) {
5859 if (bfqq
->last_serv_time_ns
== 0) {
5861 * Now we certainly have a base value: make sure we
5862 * start trying injection.
5864 bfqq
->inject_limit
= max_t(unsigned int, 1, old_limit
);
5866 bfqq
->last_serv_time_ns
= tot_time_ns
;
5867 } else if (!bfqd
->rqs_injected
&& bfqd
->rq_in_driver
== 1)
5869 * No I/O injected and no request still in service in
5870 * the drive: these are the exact conditions for
5871 * computing the base value of the total service time
5872 * for bfqq. So let's update this value, because it is
5873 * rather variable. For example, it varies if the size
5874 * or the spatial locality of the I/O requests in bfqq
5877 bfqq
->last_serv_time_ns
= tot_time_ns
;
5880 /* update complete, not waiting for any request completion any longer */
5881 bfqd
->waited_rq
= NULL
;
5882 bfqd
->rqs_injected
= false;
5886 * Handle either a requeue or a finish for rq. The things to do are
5887 * the same in both cases: all references to rq are to be dropped. In
5888 * particular, rq is considered completed from the point of view of
5891 static void bfq_finish_requeue_request(struct request
*rq
)
5893 struct bfq_queue
*bfqq
= RQ_BFQQ(rq
);
5894 struct bfq_data
*bfqd
;
5897 * Requeue and finish hooks are invoked in blk-mq without
5898 * checking whether the involved request is actually still
5899 * referenced in the scheduler. To handle this fact, the
5900 * following two checks make this function exit in case of
5901 * spurious invocations, for which there is nothing to do.
5903 * First, check whether rq has nothing to do with an elevator.
5905 if (unlikely(!(rq
->rq_flags
& RQF_ELVPRIV
)))
5909 * rq either is not associated with any icq, or is an already
5910 * requeued request that has not (yet) been re-inserted into
5913 if (!rq
->elv
.icq
|| !bfqq
)
5918 if (rq
->rq_flags
& RQF_STARTED
)
5919 bfqg_stats_update_completion(bfqq_group(bfqq
),
5921 rq
->io_start_time_ns
,
5924 if (likely(rq
->rq_flags
& RQF_STARTED
)) {
5925 unsigned long flags
;
5927 spin_lock_irqsave(&bfqd
->lock
, flags
);
5929 if (rq
== bfqd
->waited_rq
)
5930 bfq_update_inject_limit(bfqd
, bfqq
);
5932 bfq_completed_request(bfqq
, bfqd
);
5933 bfq_finish_requeue_request_body(bfqq
);
5935 spin_unlock_irqrestore(&bfqd
->lock
, flags
);
5938 * Request rq may be still/already in the scheduler,
5939 * in which case we need to remove it (this should
5940 * never happen in case of requeue). And we cannot
5941 * defer such a check and removal, to avoid
5942 * inconsistencies in the time interval from the end
5943 * of this function to the start of the deferred work.
5944 * This situation seems to occur only in process
5945 * context, as a consequence of a merge. In the
5946 * current version of the code, this implies that the
5950 if (!RB_EMPTY_NODE(&rq
->rb_node
)) {
5951 bfq_remove_request(rq
->q
, rq
);
5952 bfqg_stats_update_io_remove(bfqq_group(bfqq
),
5955 bfq_finish_requeue_request_body(bfqq
);
5959 * Reset private fields. In case of a requeue, this allows
5960 * this function to correctly do nothing if it is spuriously
5961 * invoked again on this same request (see the check at the
5962 * beginning of the function). Probably, a better general
5963 * design would be to prevent blk-mq from invoking the requeue
5964 * or finish hooks of an elevator, for a request that is not
5965 * referred by that elevator.
5967 * Resetting the following fields would break the
5968 * request-insertion logic if rq is re-inserted into a bfq
5969 * internal queue, without a re-preparation. Here we assume
5970 * that re-insertions of requeued requests, without
5971 * re-preparation, can happen only for pass_through or at_head
5972 * requests (which are not re-inserted into bfq internal
5975 rq
->elv
.priv
[0] = NULL
;
5976 rq
->elv
.priv
[1] = NULL
;
5980 * Removes the association between the current task and bfqq, assuming
5981 * that bic points to the bfq iocontext of the task.
5982 * Returns NULL if a new bfqq should be allocated, or the old bfqq if this
5983 * was the last process referring to that bfqq.
5985 static struct bfq_queue
*
5986 bfq_split_bfqq(struct bfq_io_cq
*bic
, struct bfq_queue
*bfqq
)
5988 bfq_log_bfqq(bfqq
->bfqd
, bfqq
, "splitting queue");
5990 if (bfqq_process_refs(bfqq
) == 1) {
5991 bfqq
->pid
= current
->pid
;
5992 bfq_clear_bfqq_coop(bfqq
);
5993 bfq_clear_bfqq_split_coop(bfqq
);
5997 bic_set_bfqq(bic
, NULL
, 1);
5999 bfq_put_cooperator(bfqq
);
6001 bfq_release_process_ref(bfqq
->bfqd
, bfqq
);
6005 static struct bfq_queue
*bfq_get_bfqq_handle_split(struct bfq_data
*bfqd
,
6006 struct bfq_io_cq
*bic
,
6008 bool split
, bool is_sync
,
6011 struct bfq_queue
*bfqq
= bic_to_bfqq(bic
, is_sync
);
6013 if (likely(bfqq
&& bfqq
!= &bfqd
->oom_bfqq
))
6020 bfq_put_queue(bfqq
);
6021 bfqq
= bfq_get_queue(bfqd
, bio
, is_sync
, bic
);
6023 bic_set_bfqq(bic
, bfqq
, is_sync
);
6024 if (split
&& is_sync
) {
6025 if ((bic
->was_in_burst_list
&& bfqd
->large_burst
) ||
6026 bic
->saved_in_large_burst
)
6027 bfq_mark_bfqq_in_large_burst(bfqq
);
6029 bfq_clear_bfqq_in_large_burst(bfqq
);
6030 if (bic
->was_in_burst_list
)
6032 * If bfqq was in the current
6033 * burst list before being
6034 * merged, then we have to add
6035 * it back. And we do not need
6036 * to increase burst_size, as
6037 * we did not decrement
6038 * burst_size when we removed
6039 * bfqq from the burst list as
6040 * a consequence of a merge
6042 * bfq_put_queue). In this
6043 * respect, it would be rather
6044 * costly to know whether the
6045 * current burst list is still
6046 * the same burst list from
6047 * which bfqq was removed on
6048 * the merge. To avoid this
6049 * cost, if bfqq was in a
6050 * burst list, then we add
6051 * bfqq to the current burst
6052 * list without any further
6053 * check. This can cause
6054 * inappropriate insertions,
6055 * but rarely enough to not
6056 * harm the detection of large
6057 * bursts significantly.
6059 hlist_add_head(&bfqq
->burst_list_node
,
6062 bfqq
->split_time
= jiffies
;
6069 * Only reset private fields. The actual request preparation will be
6070 * performed by bfq_init_rq, when rq is either inserted or merged. See
6071 * comments on bfq_init_rq for the reason behind this delayed
6074 static void bfq_prepare_request(struct request
*rq
, struct bio
*bio
)
6077 * Regardless of whether we have an icq attached, we have to
6078 * clear the scheduler pointers, as they might point to
6079 * previously allocated bic/bfqq structs.
6081 rq
->elv
.priv
[0] = rq
->elv
.priv
[1] = NULL
;
6085 * If needed, init rq, allocate bfq data structures associated with
6086 * rq, and increment reference counters in the destination bfq_queue
6087 * for rq. Return the destination bfq_queue for rq, or NULL is rq is
6088 * not associated with any bfq_queue.
6090 * This function is invoked by the functions that perform rq insertion
6091 * or merging. One may have expected the above preparation operations
6092 * to be performed in bfq_prepare_request, and not delayed to when rq
6093 * is inserted or merged. The rationale behind this delayed
6094 * preparation is that, after the prepare_request hook is invoked for
6095 * rq, rq may still be transformed into a request with no icq, i.e., a
6096 * request not associated with any queue. No bfq hook is invoked to
6097 * signal this transformation. As a consequence, should these
6098 * preparation operations be performed when the prepare_request hook
6099 * is invoked, and should rq be transformed one moment later, bfq
6100 * would end up in an inconsistent state, because it would have
6101 * incremented some queue counters for an rq destined to
6102 * transformation, without any chance to correctly lower these
6103 * counters back. In contrast, no transformation can still happen for
6104 * rq after rq has been inserted or merged. So, it is safe to execute
6105 * these preparation operations when rq is finally inserted or merged.
6107 static struct bfq_queue
*bfq_init_rq(struct request
*rq
)
6109 struct request_queue
*q
= rq
->q
;
6110 struct bio
*bio
= rq
->bio
;
6111 struct bfq_data
*bfqd
= q
->elevator
->elevator_data
;
6112 struct bfq_io_cq
*bic
;
6113 const int is_sync
= rq_is_sync(rq
);
6114 struct bfq_queue
*bfqq
;
6115 bool new_queue
= false;
6116 bool bfqq_already_existing
= false, split
= false;
6118 if (unlikely(!rq
->elv
.icq
))
6122 * Assuming that elv.priv[1] is set only if everything is set
6123 * for this rq. This holds true, because this function is
6124 * invoked only for insertion or merging, and, after such
6125 * events, a request cannot be manipulated any longer before
6126 * being removed from bfq.
6128 if (rq
->elv
.priv
[1])
6129 return rq
->elv
.priv
[1];
6131 bic
= icq_to_bic(rq
->elv
.icq
);
6133 bfq_check_ioprio_change(bic
, bio
);
6135 bfq_bic_update_cgroup(bic
, bio
);
6137 bfqq
= bfq_get_bfqq_handle_split(bfqd
, bic
, bio
, false, is_sync
,
6140 if (likely(!new_queue
)) {
6141 /* If the queue was seeky for too long, break it apart. */
6142 if (bfq_bfqq_coop(bfqq
) && bfq_bfqq_split_coop(bfqq
)) {
6143 bfq_log_bfqq(bfqd
, bfqq
, "breaking apart bfqq");
6145 /* Update bic before losing reference to bfqq */
6146 if (bfq_bfqq_in_large_burst(bfqq
))
6147 bic
->saved_in_large_burst
= true;
6149 bfqq
= bfq_split_bfqq(bic
, bfqq
);
6153 bfqq
= bfq_get_bfqq_handle_split(bfqd
, bic
, bio
,
6157 bfqq_already_existing
= true;
6163 bfq_log_bfqq(bfqd
, bfqq
, "get_request %p: bfqq %p, %d",
6164 rq
, bfqq
, bfqq
->ref
);
6166 rq
->elv
.priv
[0] = bic
;
6167 rq
->elv
.priv
[1] = bfqq
;
6170 * If a bfq_queue has only one process reference, it is owned
6171 * by only this bic: we can then set bfqq->bic = bic. in
6172 * addition, if the queue has also just been split, we have to
6175 if (likely(bfqq
!= &bfqd
->oom_bfqq
) && bfqq_process_refs(bfqq
) == 1) {
6179 * The queue has just been split from a shared
6180 * queue: restore the idle window and the
6181 * possible weight raising period.
6183 bfq_bfqq_resume_state(bfqq
, bfqd
, bic
,
6184 bfqq_already_existing
);
6189 * Consider bfqq as possibly belonging to a burst of newly
6190 * created queues only if:
6191 * 1) A burst is actually happening (bfqd->burst_size > 0)
6193 * 2) There is no other active queue. In fact, if, in
6194 * contrast, there are active queues not belonging to the
6195 * possible burst bfqq may belong to, then there is no gain
6196 * in considering bfqq as belonging to a burst, and
6197 * therefore in not weight-raising bfqq. See comments on
6198 * bfq_handle_burst().
6200 * This filtering also helps eliminating false positives,
6201 * occurring when bfqq does not belong to an actual large
6202 * burst, but some background task (e.g., a service) happens
6203 * to trigger the creation of new queues very close to when
6204 * bfqq and its possible companion queues are created. See
6205 * comments on bfq_handle_burst() for further details also on
6208 if (unlikely(bfq_bfqq_just_created(bfqq
) &&
6209 (bfqd
->burst_size
> 0 ||
6210 bfq_tot_busy_queues(bfqd
) == 0)))
6211 bfq_handle_burst(bfqd
, bfqq
);
6217 bfq_idle_slice_timer_body(struct bfq_data
*bfqd
, struct bfq_queue
*bfqq
)
6219 enum bfqq_expiration reason
;
6220 unsigned long flags
;
6222 spin_lock_irqsave(&bfqd
->lock
, flags
);
6225 * Considering that bfqq may be in race, we should firstly check
6226 * whether bfqq is in service before doing something on it. If
6227 * the bfqq in race is not in service, it has already been expired
6228 * through __bfq_bfqq_expire func and its wait_request flags has
6229 * been cleared in __bfq_bfqd_reset_in_service func.
6231 if (bfqq
!= bfqd
->in_service_queue
) {
6232 spin_unlock_irqrestore(&bfqd
->lock
, flags
);
6236 bfq_clear_bfqq_wait_request(bfqq
);
6238 if (bfq_bfqq_budget_timeout(bfqq
))
6240 * Also here the queue can be safely expired
6241 * for budget timeout without wasting
6244 reason
= BFQQE_BUDGET_TIMEOUT
;
6245 else if (bfqq
->queued
[0] == 0 && bfqq
->queued
[1] == 0)
6247 * The queue may not be empty upon timer expiration,
6248 * because we may not disable the timer when the
6249 * first request of the in-service queue arrives
6250 * during disk idling.
6252 reason
= BFQQE_TOO_IDLE
;
6254 goto schedule_dispatch
;
6256 bfq_bfqq_expire(bfqd
, bfqq
, true, reason
);
6259 spin_unlock_irqrestore(&bfqd
->lock
, flags
);
6260 bfq_schedule_dispatch(bfqd
);
6264 * Handler of the expiration of the timer running if the in-service queue
6265 * is idling inside its time slice.
6267 static enum hrtimer_restart
bfq_idle_slice_timer(struct hrtimer
*timer
)
6269 struct bfq_data
*bfqd
= container_of(timer
, struct bfq_data
,
6271 struct bfq_queue
*bfqq
= bfqd
->in_service_queue
;
6274 * Theoretical race here: the in-service queue can be NULL or
6275 * different from the queue that was idling if a new request
6276 * arrives for the current queue and there is a full dispatch
6277 * cycle that changes the in-service queue. This can hardly
6278 * happen, but in the worst case we just expire a queue too
6282 bfq_idle_slice_timer_body(bfqd
, bfqq
);
6284 return HRTIMER_NORESTART
;
6287 static void __bfq_put_async_bfqq(struct bfq_data
*bfqd
,
6288 struct bfq_queue
**bfqq_ptr
)
6290 struct bfq_queue
*bfqq
= *bfqq_ptr
;
6292 bfq_log(bfqd
, "put_async_bfqq: %p", bfqq
);
6294 bfq_bfqq_move(bfqd
, bfqq
, bfqd
->root_group
);
6296 bfq_log_bfqq(bfqd
, bfqq
, "put_async_bfqq: putting %p, %d",
6298 bfq_put_queue(bfqq
);
6304 * Release all the bfqg references to its async queues. If we are
6305 * deallocating the group these queues may still contain requests, so
6306 * we reparent them to the root cgroup (i.e., the only one that will
6307 * exist for sure until all the requests on a device are gone).
6309 void bfq_put_async_queues(struct bfq_data
*bfqd
, struct bfq_group
*bfqg
)
6313 for (i
= 0; i
< 2; i
++)
6314 for (j
= 0; j
< IOPRIO_BE_NR
; j
++)
6315 __bfq_put_async_bfqq(bfqd
, &bfqg
->async_bfqq
[i
][j
]);
6317 __bfq_put_async_bfqq(bfqd
, &bfqg
->async_idle_bfqq
);
6321 * See the comments on bfq_limit_depth for the purpose of
6322 * the depths set in the function. Return minimum shallow depth we'll use.
6324 static unsigned int bfq_update_depths(struct bfq_data
*bfqd
,
6325 struct sbitmap_queue
*bt
)
6327 unsigned int i
, j
, min_shallow
= UINT_MAX
;
6330 * In-word depths if no bfq_queue is being weight-raised:
6331 * leaving 25% of tags only for sync reads.
6333 * In next formulas, right-shift the value
6334 * (1U<<bt->sb.shift), instead of computing directly
6335 * (1U<<(bt->sb.shift - something)), to be robust against
6336 * any possible value of bt->sb.shift, without having to
6337 * limit 'something'.
6339 /* no more than 50% of tags for async I/O */
6340 bfqd
->word_depths
[0][0] = max((1U << bt
->sb
.shift
) >> 1, 1U);
6342 * no more than 75% of tags for sync writes (25% extra tags
6343 * w.r.t. async I/O, to prevent async I/O from starving sync
6346 bfqd
->word_depths
[0][1] = max(((1U << bt
->sb
.shift
) * 3) >> 2, 1U);
6349 * In-word depths in case some bfq_queue is being weight-
6350 * raised: leaving ~63% of tags for sync reads. This is the
6351 * highest percentage for which, in our tests, application
6352 * start-up times didn't suffer from any regression due to tag
6355 /* no more than ~18% of tags for async I/O */
6356 bfqd
->word_depths
[1][0] = max(((1U << bt
->sb
.shift
) * 3) >> 4, 1U);
6357 /* no more than ~37% of tags for sync writes (~20% extra tags) */
6358 bfqd
->word_depths
[1][1] = max(((1U << bt
->sb
.shift
) * 6) >> 4, 1U);
6360 for (i
= 0; i
< 2; i
++)
6361 for (j
= 0; j
< 2; j
++)
6362 min_shallow
= min(min_shallow
, bfqd
->word_depths
[i
][j
]);
6367 static void bfq_depth_updated(struct blk_mq_hw_ctx
*hctx
)
6369 struct bfq_data
*bfqd
= hctx
->queue
->elevator
->elevator_data
;
6370 struct blk_mq_tags
*tags
= hctx
->sched_tags
;
6371 unsigned int min_shallow
;
6373 min_shallow
= bfq_update_depths(bfqd
, &tags
->bitmap_tags
);
6374 sbitmap_queue_min_shallow_depth(&tags
->bitmap_tags
, min_shallow
);
6377 static int bfq_init_hctx(struct blk_mq_hw_ctx
*hctx
, unsigned int index
)
6379 bfq_depth_updated(hctx
);
6383 static void bfq_exit_queue(struct elevator_queue
*e
)
6385 struct bfq_data
*bfqd
= e
->elevator_data
;
6386 struct bfq_queue
*bfqq
, *n
;
6388 hrtimer_cancel(&bfqd
->idle_slice_timer
);
6390 spin_lock_irq(&bfqd
->lock
);
6391 list_for_each_entry_safe(bfqq
, n
, &bfqd
->idle_list
, bfqq_list
)
6392 bfq_deactivate_bfqq(bfqd
, bfqq
, false, false);
6393 spin_unlock_irq(&bfqd
->lock
);
6395 hrtimer_cancel(&bfqd
->idle_slice_timer
);
6397 /* release oom-queue reference to root group */
6398 bfqg_and_blkg_put(bfqd
->root_group
);
6400 #ifdef CONFIG_BFQ_GROUP_IOSCHED
6401 blkcg_deactivate_policy(bfqd
->queue
, &blkcg_policy_bfq
);
6403 spin_lock_irq(&bfqd
->lock
);
6404 bfq_put_async_queues(bfqd
, bfqd
->root_group
);
6405 kfree(bfqd
->root_group
);
6406 spin_unlock_irq(&bfqd
->lock
);
6412 static void bfq_init_root_group(struct bfq_group
*root_group
,
6413 struct bfq_data
*bfqd
)
6417 #ifdef CONFIG_BFQ_GROUP_IOSCHED
6418 root_group
->entity
.parent
= NULL
;
6419 root_group
->my_entity
= NULL
;
6420 root_group
->bfqd
= bfqd
;
6422 root_group
->rq_pos_tree
= RB_ROOT
;
6423 for (i
= 0; i
< BFQ_IOPRIO_CLASSES
; i
++)
6424 root_group
->sched_data
.service_tree
[i
] = BFQ_SERVICE_TREE_INIT
;
6425 root_group
->sched_data
.bfq_class_idle_last_service
= jiffies
;
6428 static int bfq_init_queue(struct request_queue
*q
, struct elevator_type
*e
)
6430 struct bfq_data
*bfqd
;
6431 struct elevator_queue
*eq
;
6433 eq
= elevator_alloc(q
, e
);
6437 bfqd
= kzalloc_node(sizeof(*bfqd
), GFP_KERNEL
, q
->node
);
6439 kobject_put(&eq
->kobj
);
6442 eq
->elevator_data
= bfqd
;
6444 spin_lock_irq(&q
->queue_lock
);
6446 spin_unlock_irq(&q
->queue_lock
);
6449 * Our fallback bfqq if bfq_find_alloc_queue() runs into OOM issues.
6450 * Grab a permanent reference to it, so that the normal code flow
6451 * will not attempt to free it.
6453 bfq_init_bfqq(bfqd
, &bfqd
->oom_bfqq
, NULL
, 1, 0);
6454 bfqd
->oom_bfqq
.ref
++;
6455 bfqd
->oom_bfqq
.new_ioprio
= BFQ_DEFAULT_QUEUE_IOPRIO
;
6456 bfqd
->oom_bfqq
.new_ioprio_class
= IOPRIO_CLASS_BE
;
6457 bfqd
->oom_bfqq
.entity
.new_weight
=
6458 bfq_ioprio_to_weight(bfqd
->oom_bfqq
.new_ioprio
);
6460 /* oom_bfqq does not participate to bursts */
6461 bfq_clear_bfqq_just_created(&bfqd
->oom_bfqq
);
6464 * Trigger weight initialization, according to ioprio, at the
6465 * oom_bfqq's first activation. The oom_bfqq's ioprio and ioprio
6466 * class won't be changed any more.
6468 bfqd
->oom_bfqq
.entity
.prio_changed
= 1;
6472 INIT_LIST_HEAD(&bfqd
->dispatch
);
6474 hrtimer_init(&bfqd
->idle_slice_timer
, CLOCK_MONOTONIC
,
6476 bfqd
->idle_slice_timer
.function
= bfq_idle_slice_timer
;
6478 bfqd
->queue_weights_tree
= RB_ROOT_CACHED
;
6479 bfqd
->num_groups_with_pending_reqs
= 0;
6481 INIT_LIST_HEAD(&bfqd
->active_list
);
6482 INIT_LIST_HEAD(&bfqd
->idle_list
);
6483 INIT_HLIST_HEAD(&bfqd
->burst_list
);
6486 bfqd
->nonrot_with_queueing
= blk_queue_nonrot(bfqd
->queue
);
6488 bfqd
->bfq_max_budget
= bfq_default_max_budget
;
6490 bfqd
->bfq_fifo_expire
[0] = bfq_fifo_expire
[0];
6491 bfqd
->bfq_fifo_expire
[1] = bfq_fifo_expire
[1];
6492 bfqd
->bfq_back_max
= bfq_back_max
;
6493 bfqd
->bfq_back_penalty
= bfq_back_penalty
;
6494 bfqd
->bfq_slice_idle
= bfq_slice_idle
;
6495 bfqd
->bfq_timeout
= bfq_timeout
;
6497 bfqd
->bfq_requests_within_timer
= 120;
6499 bfqd
->bfq_large_burst_thresh
= 8;
6500 bfqd
->bfq_burst_interval
= msecs_to_jiffies(180);
6502 bfqd
->low_latency
= true;
6505 * Trade-off between responsiveness and fairness.
6507 bfqd
->bfq_wr_coeff
= 30;
6508 bfqd
->bfq_wr_rt_max_time
= msecs_to_jiffies(300);
6509 bfqd
->bfq_wr_max_time
= 0;
6510 bfqd
->bfq_wr_min_idle_time
= msecs_to_jiffies(2000);
6511 bfqd
->bfq_wr_min_inter_arr_async
= msecs_to_jiffies(500);
6512 bfqd
->bfq_wr_max_softrt_rate
= 7000; /*
6513 * Approximate rate required
6514 * to playback or record a
6515 * high-definition compressed
6518 bfqd
->wr_busy_queues
= 0;
6521 * Begin by assuming, optimistically, that the device peak
6522 * rate is equal to 2/3 of the highest reference rate.
6524 bfqd
->rate_dur_prod
= ref_rate
[blk_queue_nonrot(bfqd
->queue
)] *
6525 ref_wr_duration
[blk_queue_nonrot(bfqd
->queue
)];
6526 bfqd
->peak_rate
= ref_rate
[blk_queue_nonrot(bfqd
->queue
)] * 2 / 3;
6528 spin_lock_init(&bfqd
->lock
);
6531 * The invocation of the next bfq_create_group_hierarchy
6532 * function is the head of a chain of function calls
6533 * (bfq_create_group_hierarchy->blkcg_activate_policy->
6534 * blk_mq_freeze_queue) that may lead to the invocation of the
6535 * has_work hook function. For this reason,
6536 * bfq_create_group_hierarchy is invoked only after all
6537 * scheduler data has been initialized, apart from the fields
6538 * that can be initialized only after invoking
6539 * bfq_create_group_hierarchy. This, in particular, enables
6540 * has_work to correctly return false. Of course, to avoid
6541 * other inconsistencies, the blk-mq stack must then refrain
6542 * from invoking further scheduler hooks before this init
6543 * function is finished.
6545 bfqd
->root_group
= bfq_create_group_hierarchy(bfqd
, q
->node
);
6546 if (!bfqd
->root_group
)
6548 bfq_init_root_group(bfqd
->root_group
, bfqd
);
6549 bfq_init_entity(&bfqd
->oom_bfqq
.entity
, bfqd
->root_group
);
6551 wbt_disable_default(q
);
6556 kobject_put(&eq
->kobj
);
6560 static void bfq_slab_kill(void)
6562 kmem_cache_destroy(bfq_pool
);
6565 static int __init
bfq_slab_setup(void)
6567 bfq_pool
= KMEM_CACHE(bfq_queue
, 0);
6573 static ssize_t
bfq_var_show(unsigned int var
, char *page
)
6575 return sprintf(page
, "%u\n", var
);
6578 static int bfq_var_store(unsigned long *var
, const char *page
)
6580 unsigned long new_val
;
6581 int ret
= kstrtoul(page
, 10, &new_val
);
6589 #define SHOW_FUNCTION(__FUNC, __VAR, __CONV) \
6590 static ssize_t __FUNC(struct elevator_queue *e, char *page) \
6592 struct bfq_data *bfqd = e->elevator_data; \
6593 u64 __data = __VAR; \
6595 __data = jiffies_to_msecs(__data); \
6596 else if (__CONV == 2) \
6597 __data = div_u64(__data, NSEC_PER_MSEC); \
6598 return bfq_var_show(__data, (page)); \
6600 SHOW_FUNCTION(bfq_fifo_expire_sync_show
, bfqd
->bfq_fifo_expire
[1], 2);
6601 SHOW_FUNCTION(bfq_fifo_expire_async_show
, bfqd
->bfq_fifo_expire
[0], 2);
6602 SHOW_FUNCTION(bfq_back_seek_max_show
, bfqd
->bfq_back_max
, 0);
6603 SHOW_FUNCTION(bfq_back_seek_penalty_show
, bfqd
->bfq_back_penalty
, 0);
6604 SHOW_FUNCTION(bfq_slice_idle_show
, bfqd
->bfq_slice_idle
, 2);
6605 SHOW_FUNCTION(bfq_max_budget_show
, bfqd
->bfq_user_max_budget
, 0);
6606 SHOW_FUNCTION(bfq_timeout_sync_show
, bfqd
->bfq_timeout
, 1);
6607 SHOW_FUNCTION(bfq_strict_guarantees_show
, bfqd
->strict_guarantees
, 0);
6608 SHOW_FUNCTION(bfq_low_latency_show
, bfqd
->low_latency
, 0);
6609 #undef SHOW_FUNCTION
6611 #define USEC_SHOW_FUNCTION(__FUNC, __VAR) \
6612 static ssize_t __FUNC(struct elevator_queue *e, char *page) \
6614 struct bfq_data *bfqd = e->elevator_data; \
6615 u64 __data = __VAR; \
6616 __data = div_u64(__data, NSEC_PER_USEC); \
6617 return bfq_var_show(__data, (page)); \
6619 USEC_SHOW_FUNCTION(bfq_slice_idle_us_show
, bfqd
->bfq_slice_idle
);
6620 #undef USEC_SHOW_FUNCTION
6622 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV) \
6624 __FUNC(struct elevator_queue *e, const char *page, size_t count) \
6626 struct bfq_data *bfqd = e->elevator_data; \
6627 unsigned long __data, __min = (MIN), __max = (MAX); \
6630 ret = bfq_var_store(&__data, (page)); \
6633 if (__data < __min) \
6635 else if (__data > __max) \
6638 *(__PTR) = msecs_to_jiffies(__data); \
6639 else if (__CONV == 2) \
6640 *(__PTR) = (u64)__data * NSEC_PER_MSEC; \
6642 *(__PTR) = __data; \
6645 STORE_FUNCTION(bfq_fifo_expire_sync_store
, &bfqd
->bfq_fifo_expire
[1], 1,
6647 STORE_FUNCTION(bfq_fifo_expire_async_store
, &bfqd
->bfq_fifo_expire
[0], 1,
6649 STORE_FUNCTION(bfq_back_seek_max_store
, &bfqd
->bfq_back_max
, 0, INT_MAX
, 0);
6650 STORE_FUNCTION(bfq_back_seek_penalty_store
, &bfqd
->bfq_back_penalty
, 1,
6652 STORE_FUNCTION(bfq_slice_idle_store
, &bfqd
->bfq_slice_idle
, 0, INT_MAX
, 2);
6653 #undef STORE_FUNCTION
6655 #define USEC_STORE_FUNCTION(__FUNC, __PTR, MIN, MAX) \
6656 static ssize_t __FUNC(struct elevator_queue *e, const char *page, size_t count)\
6658 struct bfq_data *bfqd = e->elevator_data; \
6659 unsigned long __data, __min = (MIN), __max = (MAX); \
6662 ret = bfq_var_store(&__data, (page)); \
6665 if (__data < __min) \
6667 else if (__data > __max) \
6669 *(__PTR) = (u64)__data * NSEC_PER_USEC; \
6672 USEC_STORE_FUNCTION(bfq_slice_idle_us_store
, &bfqd
->bfq_slice_idle
, 0,
6674 #undef USEC_STORE_FUNCTION
6676 static ssize_t
bfq_max_budget_store(struct elevator_queue
*e
,
6677 const char *page
, size_t count
)
6679 struct bfq_data
*bfqd
= e
->elevator_data
;
6680 unsigned long __data
;
6683 ret
= bfq_var_store(&__data
, (page
));
6688 bfqd
->bfq_max_budget
= bfq_calc_max_budget(bfqd
);
6690 if (__data
> INT_MAX
)
6692 bfqd
->bfq_max_budget
= __data
;
6695 bfqd
->bfq_user_max_budget
= __data
;
6701 * Leaving this name to preserve name compatibility with cfq
6702 * parameters, but this timeout is used for both sync and async.
6704 static ssize_t
bfq_timeout_sync_store(struct elevator_queue
*e
,
6705 const char *page
, size_t count
)
6707 struct bfq_data
*bfqd
= e
->elevator_data
;
6708 unsigned long __data
;
6711 ret
= bfq_var_store(&__data
, (page
));
6717 else if (__data
> INT_MAX
)
6720 bfqd
->bfq_timeout
= msecs_to_jiffies(__data
);
6721 if (bfqd
->bfq_user_max_budget
== 0)
6722 bfqd
->bfq_max_budget
= bfq_calc_max_budget(bfqd
);
6727 static ssize_t
bfq_strict_guarantees_store(struct elevator_queue
*e
,
6728 const char *page
, size_t count
)
6730 struct bfq_data
*bfqd
= e
->elevator_data
;
6731 unsigned long __data
;
6734 ret
= bfq_var_store(&__data
, (page
));
6740 if (!bfqd
->strict_guarantees
&& __data
== 1
6741 && bfqd
->bfq_slice_idle
< 8 * NSEC_PER_MSEC
)
6742 bfqd
->bfq_slice_idle
= 8 * NSEC_PER_MSEC
;
6744 bfqd
->strict_guarantees
= __data
;
6749 static ssize_t
bfq_low_latency_store(struct elevator_queue
*e
,
6750 const char *page
, size_t count
)
6752 struct bfq_data
*bfqd
= e
->elevator_data
;
6753 unsigned long __data
;
6756 ret
= bfq_var_store(&__data
, (page
));
6762 if (__data
== 0 && bfqd
->low_latency
!= 0)
6764 bfqd
->low_latency
= __data
;
6769 #define BFQ_ATTR(name) \
6770 __ATTR(name, 0644, bfq_##name##_show, bfq_##name##_store)
6772 static struct elv_fs_entry bfq_attrs
[] = {
6773 BFQ_ATTR(fifo_expire_sync
),
6774 BFQ_ATTR(fifo_expire_async
),
6775 BFQ_ATTR(back_seek_max
),
6776 BFQ_ATTR(back_seek_penalty
),
6777 BFQ_ATTR(slice_idle
),
6778 BFQ_ATTR(slice_idle_us
),
6779 BFQ_ATTR(max_budget
),
6780 BFQ_ATTR(timeout_sync
),
6781 BFQ_ATTR(strict_guarantees
),
6782 BFQ_ATTR(low_latency
),
6786 static struct elevator_type iosched_bfq_mq
= {
6788 .limit_depth
= bfq_limit_depth
,
6789 .prepare_request
= bfq_prepare_request
,
6790 .requeue_request
= bfq_finish_requeue_request
,
6791 .finish_request
= bfq_finish_requeue_request
,
6792 .exit_icq
= bfq_exit_icq
,
6793 .insert_requests
= bfq_insert_requests
,
6794 .dispatch_request
= bfq_dispatch_request
,
6795 .next_request
= elv_rb_latter_request
,
6796 .former_request
= elv_rb_former_request
,
6797 .allow_merge
= bfq_allow_bio_merge
,
6798 .bio_merge
= bfq_bio_merge
,
6799 .request_merge
= bfq_request_merge
,
6800 .requests_merged
= bfq_requests_merged
,
6801 .request_merged
= bfq_request_merged
,
6802 .has_work
= bfq_has_work
,
6803 .depth_updated
= bfq_depth_updated
,
6804 .init_hctx
= bfq_init_hctx
,
6805 .init_sched
= bfq_init_queue
,
6806 .exit_sched
= bfq_exit_queue
,
6809 .icq_size
= sizeof(struct bfq_io_cq
),
6810 .icq_align
= __alignof__(struct bfq_io_cq
),
6811 .elevator_attrs
= bfq_attrs
,
6812 .elevator_name
= "bfq",
6813 .elevator_owner
= THIS_MODULE
,
6815 MODULE_ALIAS("bfq-iosched");
6817 static int __init
bfq_init(void)
6821 #ifdef CONFIG_BFQ_GROUP_IOSCHED
6822 ret
= blkcg_policy_register(&blkcg_policy_bfq
);
6828 if (bfq_slab_setup())
6832 * Times to load large popular applications for the typical
6833 * systems installed on the reference devices (see the
6834 * comments before the definition of the next
6835 * array). Actually, we use slightly lower values, as the
6836 * estimated peak rate tends to be smaller than the actual
6837 * peak rate. The reason for this last fact is that estimates
6838 * are computed over much shorter time intervals than the long
6839 * intervals typically used for benchmarking. Why? First, to
6840 * adapt more quickly to variations. Second, because an I/O
6841 * scheduler cannot rely on a peak-rate-evaluation workload to
6842 * be run for a long time.
6844 ref_wr_duration
[0] = msecs_to_jiffies(7000); /* actually 8 sec */
6845 ref_wr_duration
[1] = msecs_to_jiffies(2500); /* actually 3 sec */
6847 ret
= elv_register(&iosched_bfq_mq
);
6856 #ifdef CONFIG_BFQ_GROUP_IOSCHED
6857 blkcg_policy_unregister(&blkcg_policy_bfq
);
6862 static void __exit
bfq_exit(void)
6864 elv_unregister(&iosched_bfq_mq
);
6865 #ifdef CONFIG_BFQ_GROUP_IOSCHED
6866 blkcg_policy_unregister(&blkcg_policy_bfq
);
6871 module_init(bfq_init
);
6872 module_exit(bfq_exit
);
6874 MODULE_AUTHOR("Paolo Valente");
6875 MODULE_LICENSE("GPL");
6876 MODULE_DESCRIPTION("MQ Budget Fair Queueing I/O Scheduler");