2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* CFQ, or complete fairness queueing, disk scheduler.
|
|
|
|
*
|
|
|
|
* Based on ideas from a previously unfinished io
|
|
|
|
* scheduler (round robin per-process disk scheduling) and Andrea Arcangeli.
|
|
|
|
*
|
2006-09-04 15:41:16 +02:00
|
|
|
* Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
#include <linux/module.h>
|
2006-03-18 18:29:52 +01:00
|
|
|
#include <linux/blkdev.h>
|
|
|
|
#include <linux/elevator.h>
|
2005-04-17 00:20:36 +02:00
|
|
|
#include <linux/hash.h>
|
|
|
|
#include <linux/rbtree.h>
|
2005-06-27 10:55:12 +02:00
|
|
|
#include <linux/ioprio.h>
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* tunables
|
|
|
|
*/
|
2006-01-06 09:46:02 +01:00
|
|
|
static const int cfq_quantum = 4; /* max queue in one round of service */
|
|
|
|
static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
|
|
|
|
static const int cfq_back_max = 16 * 1024; /* maximum backwards seek, in KiB */
|
|
|
|
static const int cfq_back_penalty = 2; /* penalty of a backwards seek */
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-01-06 09:46:02 +01:00
|
|
|
static const int cfq_slice_sync = HZ / 10;
|
2005-06-27 10:56:24 +02:00
|
|
|
static int cfq_slice_async = HZ / 25;
|
2006-01-06 09:46:02 +01:00
|
|
|
static const int cfq_slice_async_rq = 2;
|
2006-06-16 11:23:00 +02:00
|
|
|
static int cfq_slice_idle = HZ / 125;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
#define CFQ_IDLE_GRACE (HZ / 10)
|
|
|
|
#define CFQ_SLICE_SCALE (5)
|
|
|
|
|
|
|
|
#define CFQ_KEY_ASYNC (0)
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* for the hash of cfqq inside the cfqd
|
|
|
|
*/
|
|
|
|
#define CFQ_QHASH_SHIFT 6
|
|
|
|
#define CFQ_QHASH_ENTRIES (1 << CFQ_QHASH_SHIFT)
|
|
|
|
#define list_entry_qhash(entry) hlist_entry((entry), struct cfq_queue, cfq_hash)
|
|
|
|
|
|
|
|
#define list_entry_cfqq(ptr) list_entry((ptr), struct cfq_queue, cfq_list)
|
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
#define RQ_CIC(rq) ((struct cfq_io_context*)(rq)->elevator_private)
|
|
|
|
#define RQ_CFQQ(rq) ((rq)->elevator_private2)
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
static kmem_cache_t *cfq_pool;
|
|
|
|
static kmem_cache_t *cfq_ioc_pool;
|
|
|
|
|
2006-07-19 05:07:12 +02:00
|
|
|
static DEFINE_PER_CPU(unsigned long, ioc_count);
|
2006-03-18 21:05:53 +01:00
|
|
|
static struct completion *ioc_gone;
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
#define CFQ_PRIO_LISTS IOPRIO_BE_NR
|
|
|
|
#define cfq_class_idle(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
|
|
|
|
#define cfq_class_rt(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
#define ASYNC (0)
|
|
|
|
#define SYNC (1)
|
|
|
|
|
|
|
|
#define cfq_cfqq_dispatched(cfqq) \
|
|
|
|
((cfqq)->on_dispatch[ASYNC] + (cfqq)->on_dispatch[SYNC])
|
|
|
|
|
|
|
|
#define cfq_cfqq_class_sync(cfqq) ((cfqq)->key != CFQ_KEY_ASYNC)
|
|
|
|
|
|
|
|
#define cfq_cfqq_sync(cfqq) \
|
|
|
|
(cfq_cfqq_class_sync(cfqq) || (cfqq)->on_dispatch[SYNC])
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2006-03-28 13:03:44 +02:00
|
|
|
#define sample_valid(samples) ((samples) > 80)
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* Per block device queue structure
|
|
|
|
*/
|
2005-04-17 00:20:36 +02:00
|
|
|
struct cfq_data {
|
2005-06-27 10:55:12 +02:00
|
|
|
request_queue_t *queue;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* rr list of queues with requests and the count of them
|
|
|
|
*/
|
|
|
|
struct list_head rr_list[CFQ_PRIO_LISTS];
|
|
|
|
struct list_head busy_rr;
|
|
|
|
struct list_head cur_rr;
|
|
|
|
struct list_head idle_rr;
|
|
|
|
unsigned int busy_queues;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* cfqq lookup hash
|
|
|
|
*/
|
2005-04-17 00:20:36 +02:00
|
|
|
struct hlist_head *cfq_hash;
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
int rq_in_driver;
|
2006-06-01 10:12:26 +02:00
|
|
|
int hw_tag;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* idle window management
|
|
|
|
*/
|
|
|
|
struct timer_list idle_slice_timer;
|
|
|
|
struct work_struct unplug_work;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
struct cfq_queue *active_queue;
|
|
|
|
struct cfq_io_context *active_cic;
|
|
|
|
int cur_prio, cur_end_prio;
|
|
|
|
unsigned int dispatch_slice;
|
|
|
|
|
|
|
|
struct timer_list idle_class_timer;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
sector_t last_sector;
|
2005-06-27 10:55:12 +02:00
|
|
|
unsigned long last_end_request;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* tunables, see top of file
|
|
|
|
*/
|
|
|
|
unsigned int cfq_quantum;
|
2005-06-27 10:55:12 +02:00
|
|
|
unsigned int cfq_fifo_expire[2];
|
2005-04-17 00:20:36 +02:00
|
|
|
unsigned int cfq_back_penalty;
|
|
|
|
unsigned int cfq_back_max;
|
2005-06-27 10:55:12 +02:00
|
|
|
unsigned int cfq_slice[2];
|
|
|
|
unsigned int cfq_slice_async_rq;
|
|
|
|
unsigned int cfq_slice_idle;
|
2006-03-18 19:51:22 +01:00
|
|
|
|
|
|
|
struct list_head cic_list;
|
2005-04-17 00:20:36 +02:00
|
|
|
};
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* Per process-grouping structure
|
|
|
|
*/
|
2005-04-17 00:20:36 +02:00
|
|
|
struct cfq_queue {
|
|
|
|
/* reference count */
|
|
|
|
atomic_t ref;
|
|
|
|
/* parent cfq_data */
|
|
|
|
struct cfq_data *cfqd;
|
2005-06-27 10:55:12 +02:00
|
|
|
/* cfqq lookup hash */
|
2005-04-17 00:20:36 +02:00
|
|
|
struct hlist_node cfq_hash;
|
|
|
|
/* hash key */
|
2005-06-27 10:55:12 +02:00
|
|
|
unsigned int key;
|
2006-07-19 14:56:28 +02:00
|
|
|
/* member of the rr/busy/cur/idle cfqd list */
|
2005-04-17 00:20:36 +02:00
|
|
|
struct list_head cfq_list;
|
|
|
|
/* sorted list of pending requests */
|
|
|
|
struct rb_root sort_list;
|
|
|
|
/* if fifo isn't expired, next request to serve */
|
2006-07-13 12:39:25 +02:00
|
|
|
struct request *next_rq;
|
2005-04-17 00:20:36 +02:00
|
|
|
/* requests queued in sort_list */
|
|
|
|
int queued[2];
|
|
|
|
/* currently allocated requests */
|
|
|
|
int allocated[2];
|
2006-07-23 01:42:19 +02:00
|
|
|
/* pending metadata requests */
|
|
|
|
int meta_pending;
|
2005-04-17 00:20:36 +02:00
|
|
|
/* fifo list of requests in sort_list */
|
2005-06-27 10:55:12 +02:00
|
|
|
struct list_head fifo;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
unsigned long slice_start;
|
|
|
|
unsigned long slice_end;
|
|
|
|
unsigned long slice_left;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
/* number of requests that are on the dispatch list */
|
|
|
|
int on_dispatch[2];
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
/* io prio of this group */
|
|
|
|
unsigned short ioprio, org_ioprio;
|
|
|
|
unsigned short ioprio_class, org_ioprio_class;
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
/* various state flags, see below */
|
|
|
|
unsigned int flags;
|
2005-04-17 00:20:36 +02:00
|
|
|
};
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
enum cfqq_state_flags {
|
|
|
|
CFQ_CFQQ_FLAG_on_rr = 0,
|
|
|
|
CFQ_CFQQ_FLAG_wait_request,
|
|
|
|
CFQ_CFQQ_FLAG_must_alloc,
|
|
|
|
CFQ_CFQQ_FLAG_must_alloc_slice,
|
|
|
|
CFQ_CFQQ_FLAG_must_dispatch,
|
|
|
|
CFQ_CFQQ_FLAG_fifo_expire,
|
|
|
|
CFQ_CFQQ_FLAG_idle_window,
|
|
|
|
CFQ_CFQQ_FLAG_prio_changed,
|
2006-07-28 09:48:51 +02:00
|
|
|
CFQ_CFQQ_FLAG_queue_new,
|
2005-06-27 10:56:24 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
#define CFQ_CFQQ_FNS(name) \
|
|
|
|
static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq) \
|
|
|
|
{ \
|
|
|
|
cfqq->flags |= (1 << CFQ_CFQQ_FLAG_##name); \
|
|
|
|
} \
|
|
|
|
static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq) \
|
|
|
|
{ \
|
|
|
|
cfqq->flags &= ~(1 << CFQ_CFQQ_FLAG_##name); \
|
|
|
|
} \
|
|
|
|
static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq) \
|
|
|
|
{ \
|
|
|
|
return (cfqq->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0; \
|
|
|
|
}
|
|
|
|
|
|
|
|
CFQ_CFQQ_FNS(on_rr);
|
|
|
|
CFQ_CFQQ_FNS(wait_request);
|
|
|
|
CFQ_CFQQ_FNS(must_alloc);
|
|
|
|
CFQ_CFQQ_FNS(must_alloc_slice);
|
|
|
|
CFQ_CFQQ_FNS(must_dispatch);
|
|
|
|
CFQ_CFQQ_FNS(fifo_expire);
|
|
|
|
CFQ_CFQQ_FNS(idle_window);
|
|
|
|
CFQ_CFQQ_FNS(prio_changed);
|
2006-07-28 09:48:51 +02:00
|
|
|
CFQ_CFQQ_FNS(queue_new);
|
2005-06-27 10:56:24 +02:00
|
|
|
#undef CFQ_CFQQ_FNS
|
|
|
|
|
|
|
|
static struct cfq_queue *cfq_find_cfq_hash(struct cfq_data *, unsigned int, unsigned short);
|
2006-07-13 12:39:25 +02:00
|
|
|
static void cfq_dispatch_insert(request_queue_t *, struct request *);
|
2006-03-18 20:58:37 +01:00
|
|
|
static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, unsigned int key, struct task_struct *tsk, gfp_t gfp_mask);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-28 05:14:05 +02:00
|
|
|
/*
|
|
|
|
* scheduler run of queue, if there are requests pending and no one in the
|
|
|
|
* driver that will restart queueing
|
|
|
|
*/
|
|
|
|
static inline void cfq_schedule_dispatch(struct cfq_data *cfqd)
|
|
|
|
{
|
2006-02-28 09:35:11 +01:00
|
|
|
if (cfqd->busy_queues)
|
2005-06-28 05:14:05 +02:00
|
|
|
kblockd_schedule_work(&cfqd->unplug_work);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int cfq_queue_empty(request_queue_t *q)
|
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
return !cfqd->busy_queues;
|
2005-06-28 05:14:05 +02:00
|
|
|
}
|
|
|
|
|
2006-03-28 13:03:44 +02:00
|
|
|
static inline pid_t cfq_queue_pid(struct task_struct *task, int rw)
|
|
|
|
{
|
2006-06-13 08:26:10 +02:00
|
|
|
if (rw == READ || rw == WRITE_SYNC)
|
2006-03-28 13:03:44 +02:00
|
|
|
return task->pid;
|
|
|
|
|
|
|
|
return CFQ_KEY_ASYNC;
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
2006-07-13 12:39:25 +02:00
|
|
|
* Lifted from AS - choose which of rq1 and rq2 that is best served now.
|
2005-04-17 00:20:36 +02:00
|
|
|
* We choose the request that is closest to the head right now. Distance
|
2006-03-28 08:59:49 +02:00
|
|
|
* behind the head is penalized and only allowed to a certain extent.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
static struct request *
|
|
|
|
cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
sector_t last, s1, s2, d1 = 0, d2 = 0;
|
|
|
|
unsigned long back_max;
|
2006-03-28 08:59:49 +02:00
|
|
|
#define CFQ_RQ1_WRAP 0x01 /* request 1 wraps */
|
|
|
|
#define CFQ_RQ2_WRAP 0x02 /* request 2 wraps */
|
|
|
|
unsigned wrap = 0; /* bit mask: requests behind the disk head? */
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
if (rq1 == NULL || rq1 == rq2)
|
|
|
|
return rq2;
|
|
|
|
if (rq2 == NULL)
|
|
|
|
return rq1;
|
2005-08-24 14:57:54 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
if (rq_is_sync(rq1) && !rq_is_sync(rq2))
|
|
|
|
return rq1;
|
|
|
|
else if (rq_is_sync(rq2) && !rq_is_sync(rq1))
|
|
|
|
return rq2;
|
2006-07-23 01:42:19 +02:00
|
|
|
if (rq_is_meta(rq1) && !rq_is_meta(rq2))
|
|
|
|
return rq1;
|
|
|
|
else if (rq_is_meta(rq2) && !rq_is_meta(rq1))
|
|
|
|
return rq2;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
s1 = rq1->sector;
|
|
|
|
s2 = rq2->sector;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
last = cfqd->last_sector;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* by definition, 1KiB is 2 sectors
|
|
|
|
*/
|
|
|
|
back_max = cfqd->cfq_back_max * 2;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Strict one way elevator _except_ in the case where we allow
|
|
|
|
* short backward seeks which are biased as twice the cost of a
|
|
|
|
* similar forward seek.
|
|
|
|
*/
|
|
|
|
if (s1 >= last)
|
|
|
|
d1 = s1 - last;
|
|
|
|
else if (s1 + back_max >= last)
|
|
|
|
d1 = (last - s1) * cfqd->cfq_back_penalty;
|
|
|
|
else
|
2006-03-28 08:59:49 +02:00
|
|
|
wrap |= CFQ_RQ1_WRAP;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (s2 >= last)
|
|
|
|
d2 = s2 - last;
|
|
|
|
else if (s2 + back_max >= last)
|
|
|
|
d2 = (last - s2) * cfqd->cfq_back_penalty;
|
|
|
|
else
|
2006-03-28 08:59:49 +02:00
|
|
|
wrap |= CFQ_RQ2_WRAP;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/* Found required data */
|
2006-03-28 08:59:49 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* By doing switch() on the bit mask "wrap" we avoid having to
|
|
|
|
* check two variables for all permutations: --> faster!
|
|
|
|
*/
|
|
|
|
switch (wrap) {
|
2006-07-13 12:39:25 +02:00
|
|
|
case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
|
2006-03-28 08:59:49 +02:00
|
|
|
if (d1 < d2)
|
2006-07-13 12:39:25 +02:00
|
|
|
return rq1;
|
2006-03-28 08:59:49 +02:00
|
|
|
else if (d2 < d1)
|
2006-07-13 12:39:25 +02:00
|
|
|
return rq2;
|
2006-03-28 08:59:49 +02:00
|
|
|
else {
|
|
|
|
if (s1 >= s2)
|
2006-07-13 12:39:25 +02:00
|
|
|
return rq1;
|
2006-03-28 08:59:49 +02:00
|
|
|
else
|
2006-07-13 12:39:25 +02:00
|
|
|
return rq2;
|
2006-03-28 08:59:49 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-28 08:59:49 +02:00
|
|
|
case CFQ_RQ2_WRAP:
|
2006-07-13 12:39:25 +02:00
|
|
|
return rq1;
|
2006-03-28 08:59:49 +02:00
|
|
|
case CFQ_RQ1_WRAP:
|
2006-07-13 12:39:25 +02:00
|
|
|
return rq2;
|
|
|
|
case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
|
2006-03-28 08:59:49 +02:00
|
|
|
default:
|
|
|
|
/*
|
|
|
|
* Since both rqs are wrapped,
|
|
|
|
* start with the one that's further behind head
|
|
|
|
* (--> only *one* back seek required),
|
|
|
|
* since back seek takes more time than forward.
|
|
|
|
*/
|
|
|
|
if (s1 <= s2)
|
2006-07-13 12:39:25 +02:00
|
|
|
return rq1;
|
2005-04-17 00:20:36 +02:00
|
|
|
else
|
2006-07-13 12:39:25 +02:00
|
|
|
return rq2;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* would be nice to take fifo expire time into account as well
|
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
static struct request *
|
|
|
|
cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
|
|
|
|
struct request *last)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-07-13 12:33:14 +02:00
|
|
|
struct rb_node *rbnext = rb_next(&last->rb_node);
|
|
|
|
struct rb_node *rbprev = rb_prev(&last->rb_node);
|
2006-07-13 12:39:25 +02:00
|
|
|
struct request *next = NULL, *prev = NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:33:14 +02:00
|
|
|
BUG_ON(RB_EMPTY_NODE(&last->rb_node));
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (rbprev)
|
2006-07-13 12:39:25 +02:00
|
|
|
prev = rb_entry_rq(rbprev);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:33:14 +02:00
|
|
|
if (rbnext)
|
2006-07-13 12:39:25 +02:00
|
|
|
next = rb_entry_rq(rbnext);
|
2006-07-13 12:33:14 +02:00
|
|
|
else {
|
|
|
|
rbnext = rb_first(&cfqq->sort_list);
|
|
|
|
if (rbnext && rbnext != &last->rb_node)
|
2006-07-13 12:39:25 +02:00
|
|
|
next = rb_entry_rq(rbnext);
|
2006-07-13 12:33:14 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:33:14 +02:00
|
|
|
return cfq_choose_req(cfqd, next, prev);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
static void cfq_resort_rr_list(struct cfq_queue *cfqq, int preempted)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
struct cfq_data *cfqd = cfqq->cfqd;
|
2006-07-28 09:48:51 +02:00
|
|
|
struct list_head *list;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
BUG_ON(!cfq_cfqq_on_rr(cfqq));
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
list_del(&cfqq->cfq_list);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
if (cfq_class_rt(cfqq))
|
|
|
|
list = &cfqd->cur_rr;
|
|
|
|
else if (cfq_class_idle(cfqq))
|
|
|
|
list = &cfqd->idle_rr;
|
|
|
|
else {
|
|
|
|
/*
|
|
|
|
* if cfqq has requests in flight, don't allow it to be
|
|
|
|
* found in cfq_set_active_queue before it has finished them.
|
|
|
|
* this is done to increase fairness between a process that
|
|
|
|
* has lots of io pending vs one that only generates one
|
|
|
|
* sporadically or synchronously
|
|
|
|
*/
|
2005-06-27 10:56:24 +02:00
|
|
|
if (cfq_cfqq_dispatched(cfqq))
|
2005-06-27 10:55:12 +02:00
|
|
|
list = &cfqd->busy_rr;
|
|
|
|
else
|
|
|
|
list = &cfqd->rr_list[cfqq->ioprio];
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
2006-07-28 09:48:51 +02:00
|
|
|
* If this queue was preempted or is new (never been serviced), let
|
|
|
|
* it be added first for fairness but beind other new queues.
|
|
|
|
* Otherwise, just add to the back of the list.
|
2005-06-27 10:55:12 +02:00
|
|
|
*/
|
2006-07-28 09:48:51 +02:00
|
|
|
if (preempted || cfq_cfqq_queue_new(cfqq)) {
|
|
|
|
struct list_head *n = list;
|
|
|
|
struct cfq_queue *__cfqq;
|
2006-06-01 18:53:43 +02:00
|
|
|
|
2006-07-28 09:48:51 +02:00
|
|
|
while (n->next != list) {
|
|
|
|
__cfqq = list_entry_cfqq(n->next);
|
|
|
|
if (!cfq_cfqq_queue_new(__cfqq))
|
|
|
|
break;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-28 09:48:51 +02:00
|
|
|
n = n->next;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-28 09:48:51 +02:00
|
|
|
list = n;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-07-28 09:48:51 +02:00
|
|
|
list_add_tail(&cfqq->cfq_list, list);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* add to busy list of queues for service, trying to be fair in ordering
|
2005-06-27 10:55:12 +02:00
|
|
|
* the pending list according to last request service
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
static inline void
|
2005-10-20 16:42:29 +02:00
|
|
|
cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:56:24 +02:00
|
|
|
BUG_ON(cfq_cfqq_on_rr(cfqq));
|
|
|
|
cfq_mark_cfqq_on_rr(cfqq);
|
2005-04-17 00:20:36 +02:00
|
|
|
cfqd->busy_queues++;
|
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
cfq_resort_rr_list(cfqq, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
|
|
|
{
|
2005-06-27 10:56:24 +02:00
|
|
|
BUG_ON(!cfq_cfqq_on_rr(cfqq));
|
|
|
|
cfq_clear_cfqq_on_rr(cfqq);
|
2006-07-19 14:56:28 +02:00
|
|
|
list_del_init(&cfqq->cfq_list);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
BUG_ON(!cfqd->busy_queues);
|
|
|
|
cfqd->busy_queues--;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* rb tree support functions
|
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
static inline void cfq_del_rq_rb(struct request *rq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_queue *cfqq = RQ_CFQQ(rq);
|
2005-10-20 16:42:29 +02:00
|
|
|
struct cfq_data *cfqd = cfqq->cfqd;
|
2006-07-13 12:39:25 +02:00
|
|
|
const int sync = rq_is_sync(rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
BUG_ON(!cfqq->queued[sync]);
|
|
|
|
cfqq->queued[sync]--;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
elv_rb_del(&cfqq->sort_list, rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-06-21 09:36:18 +02:00
|
|
|
if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
|
2005-10-20 16:42:29 +02:00
|
|
|
cfq_del_cfqq_rr(cfqd, cfqq);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
static void cfq_add_rq_rb(struct request *rq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_queue *cfqq = RQ_CFQQ(rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
struct cfq_data *cfqd = cfqq->cfqd;
|
2006-07-13 12:33:14 +02:00
|
|
|
struct request *__alias;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:37:56 +02:00
|
|
|
cfqq->queued[rq_is_sync(rq)]++;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* looks a little odd, but the first insert might return an alias.
|
|
|
|
* if that happens, put the alias on the dispatch list
|
|
|
|
*/
|
2006-07-13 12:33:14 +02:00
|
|
|
while ((__alias = elv_rb_add(&cfqq->sort_list, rq)) != NULL)
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_dispatch_insert(cfqd->queue, __alias);
|
2006-10-31 14:21:55 +01:00
|
|
|
|
|
|
|
if (!cfq_cfqq_on_rr(cfqq))
|
|
|
|
cfq_add_cfqq_rr(cfqd, cfqq);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-07-13 12:37:56 +02:00
|
|
|
elv_rb_del(&cfqq->sort_list, rq);
|
|
|
|
cfqq->queued[rq_is_sync(rq)]--;
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_add_rq_rb(rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-03-28 13:03:44 +02:00
|
|
|
static struct request *
|
|
|
|
cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-03-28 13:03:44 +02:00
|
|
|
struct task_struct *tsk = current;
|
|
|
|
pid_t key = cfq_queue_pid(tsk, bio_data_dir(bio));
|
|
|
|
struct cfq_queue *cfqq;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-28 13:03:44 +02:00
|
|
|
cfqq = cfq_find_cfq_hash(cfqd, key, tsk->ioprio);
|
2006-07-22 16:48:31 +02:00
|
|
|
if (cfqq) {
|
|
|
|
sector_t sector = bio->bi_sector + bio_sectors(bio);
|
|
|
|
|
2006-07-13 12:33:14 +02:00
|
|
|
return elv_rb_find(&cfqq->sort_list, sector);
|
2006-07-22 16:48:31 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
static void cfq_activate_request(request_queue_t *q, struct request *rq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
2005-06-27 10:56:24 +02:00
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
cfqd->rq_in_driver++;
|
2006-06-01 10:12:26 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If the depth is larger 1, it really could be queueing. But lets
|
|
|
|
* make the mark a little higher - idling could still be good for
|
|
|
|
* low queueing, and a low queueing number could also just indicate
|
|
|
|
* a SCSI mid layer like behaviour where limit+1 is often seen.
|
|
|
|
*/
|
|
|
|
if (!cfqd->hw_tag && cfqd->rq_in_driver > 4)
|
|
|
|
cfqd->hw_tag = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
static void cfq_deactivate_request(request_queue_t *q, struct request *rq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-10-20 16:42:29 +02:00
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
|
|
|
|
|
|
|
WARN_ON(!cfqd->rq_in_driver);
|
|
|
|
cfqd->rq_in_driver--;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
static void cfq_remove_request(struct request *rq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_queue *cfqq = RQ_CFQQ(rq);
|
2006-07-13 12:33:14 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
if (cfqq->next_rq == rq)
|
|
|
|
cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
list_del_init(&rq->queuelist);
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_del_rq_rb(rq);
|
2006-07-23 01:42:19 +02:00
|
|
|
|
|
|
|
if (rq_is_meta(rq)) {
|
|
|
|
WARN_ON(!cfqq->meta_pending);
|
|
|
|
cfqq->meta_pending--;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
cfq_merge(request_queue_t *q, struct request **req, struct bio *bio)
|
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
|
|
|
struct request *__rq;
|
|
|
|
|
2006-03-28 13:03:44 +02:00
|
|
|
__rq = cfq_find_rq_fmerge(cfqd, bio);
|
2005-06-27 10:55:12 +02:00
|
|
|
if (__rq && elv_rq_merge_ok(__rq, bio)) {
|
2006-07-28 09:23:08 +02:00
|
|
|
*req = __rq;
|
|
|
|
return ELEVATOR_FRONT_MERGE;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return ELEVATOR_NO_MERGE;
|
|
|
|
}
|
|
|
|
|
2006-07-13 12:33:14 +02:00
|
|
|
static void cfq_merged_request(request_queue_t *q, struct request *req,
|
|
|
|
int type)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-07-13 12:33:14 +02:00
|
|
|
if (type == ELEVATOR_FRONT_MERGE) {
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_queue *cfqq = RQ_CFQQ(req);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_reposition_rq_rb(cfqq, req);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
cfq_merged_requests(request_queue_t *q, struct request *rq,
|
|
|
|
struct request *next)
|
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* reposition in fifo if next is older than rq
|
|
|
|
*/
|
|
|
|
if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
|
|
|
|
time_before(next->start_time, rq->start_time))
|
|
|
|
list_move(&rq->queuelist, &next->queuelist);
|
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
cfq_remove_request(next);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
__cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
|
|
|
{
|
|
|
|
if (cfqq) {
|
|
|
|
/*
|
|
|
|
* stop potential idle class queues waiting service
|
|
|
|
*/
|
|
|
|
del_timer(&cfqd->idle_class_timer);
|
|
|
|
|
|
|
|
cfqq->slice_start = jiffies;
|
|
|
|
cfqq->slice_end = 0;
|
|
|
|
cfqq->slice_left = 0;
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_clear_cfqq_must_alloc_slice(cfqq);
|
|
|
|
cfq_clear_cfqq_fifo_expire(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
cfqd->active_queue = cfqq;
|
|
|
|
}
|
|
|
|
|
2006-02-28 09:35:11 +01:00
|
|
|
/*
|
|
|
|
* current cfqq expired its slice (or was too idle), select new one
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
__cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
|
|
|
|
int preempted)
|
|
|
|
{
|
|
|
|
unsigned long now = jiffies;
|
|
|
|
|
|
|
|
if (cfq_cfqq_wait_request(cfqq))
|
|
|
|
del_timer(&cfqd->idle_slice_timer);
|
|
|
|
|
2006-07-28 09:48:51 +02:00
|
|
|
if (!preempted && !cfq_cfqq_dispatched(cfqq))
|
2006-02-28 09:35:11 +01:00
|
|
|
cfq_schedule_dispatch(cfqd);
|
|
|
|
|
|
|
|
cfq_clear_cfqq_must_dispatch(cfqq);
|
|
|
|
cfq_clear_cfqq_wait_request(cfqq);
|
2006-07-28 09:48:51 +02:00
|
|
|
cfq_clear_cfqq_queue_new(cfqq);
|
2006-02-28 09:35:11 +01:00
|
|
|
|
|
|
|
/*
|
|
|
|
* store what was left of this slice, if the queue idled out
|
|
|
|
* or was preempted
|
|
|
|
*/
|
|
|
|
if (time_after(cfqq->slice_end, now))
|
|
|
|
cfqq->slice_left = cfqq->slice_end - now;
|
|
|
|
else
|
|
|
|
cfqq->slice_left = 0;
|
|
|
|
|
|
|
|
if (cfq_cfqq_on_rr(cfqq))
|
|
|
|
cfq_resort_rr_list(cfqq, preempted);
|
|
|
|
|
|
|
|
if (cfqq == cfqd->active_queue)
|
|
|
|
cfqd->active_queue = NULL;
|
|
|
|
|
|
|
|
if (cfqd->active_cic) {
|
|
|
|
put_io_context(cfqd->active_cic->ioc);
|
|
|
|
cfqd->active_cic = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
cfqd->dispatch_slice = 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void cfq_slice_expired(struct cfq_data *cfqd, int preempted)
|
|
|
|
{
|
|
|
|
struct cfq_queue *cfqq = cfqd->active_queue;
|
|
|
|
|
|
|
|
if (cfqq)
|
|
|
|
__cfq_slice_expired(cfqd, cfqq, preempted);
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* 0
|
|
|
|
* 0,1
|
|
|
|
* 0,1,2
|
|
|
|
* 0,1,2,3
|
|
|
|
* 0,1,2,3,4
|
|
|
|
* 0,1,2,3,4,5
|
|
|
|
* 0,1,2,3,4,5,6
|
|
|
|
* 0,1,2,3,4,5,6,7
|
|
|
|
*/
|
|
|
|
static int cfq_get_next_prio_level(struct cfq_data *cfqd)
|
|
|
|
{
|
|
|
|
int prio, wrap;
|
|
|
|
|
|
|
|
prio = -1;
|
|
|
|
wrap = 0;
|
|
|
|
do {
|
|
|
|
int p;
|
|
|
|
|
|
|
|
for (p = cfqd->cur_prio; p <= cfqd->cur_end_prio; p++) {
|
|
|
|
if (!list_empty(&cfqd->rr_list[p])) {
|
|
|
|
prio = p;
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (prio != -1)
|
|
|
|
break;
|
|
|
|
cfqd->cur_prio = 0;
|
|
|
|
if (++cfqd->cur_end_prio == CFQ_PRIO_LISTS) {
|
|
|
|
cfqd->cur_end_prio = 0;
|
|
|
|
if (wrap)
|
|
|
|
break;
|
|
|
|
wrap = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
2005-06-27 10:55:12 +02:00
|
|
|
} while (1);
|
|
|
|
|
|
|
|
if (unlikely(prio == -1))
|
|
|
|
return -1;
|
|
|
|
|
|
|
|
BUG_ON(prio >= CFQ_PRIO_LISTS);
|
|
|
|
|
|
|
|
list_splice_init(&cfqd->rr_list[prio], &cfqd->cur_rr);
|
|
|
|
|
|
|
|
cfqd->cur_prio = prio + 1;
|
|
|
|
if (cfqd->cur_prio > cfqd->cur_end_prio) {
|
|
|
|
cfqd->cur_end_prio = cfqd->cur_prio;
|
|
|
|
cfqd->cur_prio = 0;
|
|
|
|
}
|
|
|
|
if (cfqd->cur_end_prio == CFQ_PRIO_LISTS) {
|
|
|
|
cfqd->cur_prio = 0;
|
|
|
|
cfqd->cur_end_prio = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
return prio;
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
2006-02-28 09:35:11 +01:00
|
|
|
struct cfq_queue *cfqq = NULL;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2006-07-22 16:48:31 +02:00
|
|
|
if (!list_empty(&cfqd->cur_rr) || cfq_get_next_prio_level(cfqd) != -1) {
|
|
|
|
/*
|
|
|
|
* if current list is non-empty, grab first entry. if it is
|
|
|
|
* empty, get next prio level and grab first entry then if any
|
|
|
|
* are spliced
|
|
|
|
*/
|
2005-06-27 10:55:12 +02:00
|
|
|
cfqq = list_entry_cfqq(cfqd->cur_rr.next);
|
2006-07-22 16:48:31 +02:00
|
|
|
} else if (!list_empty(&cfqd->busy_rr)) {
|
|
|
|
/*
|
|
|
|
* If no new queues are available, check if the busy list has
|
|
|
|
* some before falling back to idle io.
|
|
|
|
*/
|
2006-06-01 10:07:26 +02:00
|
|
|
cfqq = list_entry_cfqq(cfqd->busy_rr.next);
|
2006-07-22 16:48:31 +02:00
|
|
|
} else if (!list_empty(&cfqd->idle_rr)) {
|
|
|
|
/*
|
|
|
|
* if we have idle queues and no rt or be queues had pending
|
|
|
|
* requests, either allow immediate service if the grace period
|
|
|
|
* has passed or arm the idle grace timer
|
|
|
|
*/
|
2005-06-27 10:55:12 +02:00
|
|
|
unsigned long end = cfqd->last_end_request + CFQ_IDLE_GRACE;
|
|
|
|
|
|
|
|
if (time_after_eq(jiffies, end))
|
|
|
|
cfqq = list_entry_cfqq(cfqd->idle_rr.next);
|
|
|
|
else
|
|
|
|
mod_timer(&cfqd->idle_class_timer, end);
|
|
|
|
}
|
|
|
|
|
|
|
|
__cfq_set_active_queue(cfqd, cfqq);
|
2005-06-27 10:56:24 +02:00
|
|
|
return cfqq;
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
#define CIC_SEEKY(cic) ((cic)->seek_mean > (128 * 1024))
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
static int cfq_arm_slice_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
|
|
|
|
|
|
|
{
|
2006-03-28 13:03:44 +02:00
|
|
|
struct cfq_io_context *cic;
|
2006-02-28 09:35:11 +01:00
|
|
|
unsigned long sl;
|
|
|
|
|
2006-06-21 09:36:18 +02:00
|
|
|
WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
|
2005-06-27 10:55:12 +02:00
|
|
|
WARN_ON(cfqq != cfqd->active_queue);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* idle is disabled, either manually or by past process history
|
|
|
|
*/
|
|
|
|
if (!cfqd->cfq_slice_idle)
|
|
|
|
return 0;
|
2005-06-27 10:56:24 +02:00
|
|
|
if (!cfq_cfqq_idle_window(cfqq))
|
2005-06-27 10:55:12 +02:00
|
|
|
return 0;
|
|
|
|
/*
|
|
|
|
* task has exited, don't wait
|
|
|
|
*/
|
2006-03-28 13:03:44 +02:00
|
|
|
cic = cfqd->active_cic;
|
|
|
|
if (!cic || !cic->ioc->task)
|
2005-06-27 10:55:12 +02:00
|
|
|
return 0;
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_mark_cfqq_must_dispatch(cfqq);
|
|
|
|
cfq_mark_cfqq_wait_request(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2006-02-28 09:35:11 +01:00
|
|
|
sl = min(cfqq->slice_end - 1, (unsigned long) cfqd->cfq_slice_idle);
|
2006-03-28 13:03:44 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* we don't want to idle for seeks, but we do want to allow
|
|
|
|
* fair distribution of slice time for a process doing back-to-back
|
|
|
|
* seeks. so allow a little bit of time for him to submit a new rq
|
|
|
|
*/
|
2006-06-16 11:23:00 +02:00
|
|
|
if (sample_valid(cic->seek_samples) && CIC_SEEKY(cic))
|
2006-07-25 15:05:21 +02:00
|
|
|
sl = min(sl, msecs_to_jiffies(2));
|
2006-03-28 13:03:44 +02:00
|
|
|
|
2006-02-28 09:35:11 +01:00
|
|
|
mod_timer(&cfqd->idle_slice_timer, jiffies + sl);
|
2005-06-27 10:55:12 +02:00
|
|
|
return 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
static void cfq_dispatch_insert(request_queue_t *q, struct request *rq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_queue *cfqq = RQ_CFQQ(rq);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2006-07-13 12:37:56 +02:00
|
|
|
cfq_remove_request(rq);
|
|
|
|
cfqq->on_dispatch[rq_is_sync(rq)]++;
|
|
|
|
elv_dispatch_sort(q, rq);
|
2006-06-16 15:35:39 +02:00
|
|
|
|
|
|
|
rq = list_entry(q->queue_head.prev, struct request, queuelist);
|
|
|
|
cfqd->last_sector = rq->sector + rq->nr_sectors;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* return expired entry, or NULL to just start from scratch in rbtree
|
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
static inline struct request *cfq_check_fifo(struct cfq_queue *cfqq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = cfqq->cfqd;
|
2005-06-27 10:55:12 +02:00
|
|
|
struct request *rq;
|
2006-07-22 16:48:31 +02:00
|
|
|
int fifo;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
if (cfq_cfqq_fifo_expire(cfqq))
|
2005-04-17 00:20:36 +02:00
|
|
|
return NULL;
|
2006-07-22 16:48:31 +02:00
|
|
|
if (list_empty(&cfqq->fifo))
|
|
|
|
return NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-22 16:48:31 +02:00
|
|
|
fifo = cfq_cfqq_class_sync(cfqq);
|
|
|
|
rq = rq_entry_fifo(cfqq->fifo.next);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-22 16:48:31 +02:00
|
|
|
if (time_after(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo])) {
|
|
|
|
cfq_mark_cfqq_fifo_expire(cfqq);
|
|
|
|
return rq;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2005-06-27 10:56:24 +02:00
|
|
|
* Scale schedule slice based on io priority. Use the sync time slice only
|
|
|
|
* if a queue is marked sync and has sync io queued. A sync queue with async
|
|
|
|
* io only, should not get full sync slice length.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2005-06-27 10:55:12 +02:00
|
|
|
static inline int
|
|
|
|
cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
|
|
|
{
|
|
|
|
const int base_slice = cfqd->cfq_slice[cfq_cfqq_sync(cfqq)];
|
|
|
|
|
|
|
|
WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
|
|
|
|
|
|
|
|
return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - cfqq->ioprio));
|
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
static inline void
|
2005-06-27 10:55:12 +02:00
|
|
|
cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
static inline int
|
|
|
|
cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
|
|
|
{
|
|
|
|
const int base_rq = cfqd->cfq_slice_async_rq;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
return 2 * (base_rq + base_rq * (CFQ_PRIO_LISTS - 1 - cfqq->ioprio));
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* get next queue for service
|
|
|
|
*/
|
2005-11-10 08:49:19 +01:00
|
|
|
static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
unsigned long now = jiffies;
|
2005-04-17 00:20:36 +02:00
|
|
|
struct cfq_queue *cfqq;
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
cfqq = cfqd->active_queue;
|
|
|
|
if (!cfqq)
|
|
|
|
goto new_queue;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* slice has expired
|
|
|
|
*/
|
2005-06-27 10:56:24 +02:00
|
|
|
if (!cfq_cfqq_must_dispatch(cfqq) && time_after(now, cfqq->slice_end))
|
|
|
|
goto expire;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* if queue has requests, dispatch one. if not, check if
|
|
|
|
* enough slice is left to wait for one
|
|
|
|
*/
|
2006-06-21 09:36:18 +02:00
|
|
|
if (!RB_EMPTY_ROOT(&cfqq->sort_list))
|
2005-06-27 10:55:12 +02:00
|
|
|
goto keep_queue;
|
2006-06-16 11:23:00 +02:00
|
|
|
else if (cfq_cfqq_dispatched(cfqq)) {
|
|
|
|
cfqq = NULL;
|
|
|
|
goto keep_queue;
|
|
|
|
} else if (cfq_cfqq_class_sync(cfqq)) {
|
2005-06-27 10:55:12 +02:00
|
|
|
if (cfq_arm_slice_timer(cfqd, cfqq))
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
expire:
|
2005-06-27 10:55:12 +02:00
|
|
|
cfq_slice_expired(cfqd, 0);
|
2005-06-27 10:56:24 +02:00
|
|
|
new_queue:
|
|
|
|
cfqq = cfq_set_active_queue(cfqd);
|
2005-06-27 10:55:12 +02:00
|
|
|
keep_queue:
|
2005-06-27 10:56:24 +02:00
|
|
|
return cfqq;
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
__cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
|
|
|
|
int max_dispatch)
|
|
|
|
{
|
|
|
|
int dispatched = 0;
|
|
|
|
|
2006-06-21 09:36:18 +02:00
|
|
|
BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
do {
|
2006-07-13 12:39:25 +02:00
|
|
|
struct request *rq;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
/*
|
2005-06-27 10:55:12 +02:00
|
|
|
* follow expired path, else get first next available
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
if ((rq = cfq_check_fifo(cfqq)) == NULL)
|
|
|
|
rq = cfqq->next_rq;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* finally, insert request into driver dispatch list
|
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_dispatch_insert(cfqd->queue, rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
cfqd->dispatch_slice++;
|
|
|
|
dispatched++;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
if (!cfqd->active_cic) {
|
2006-07-13 12:39:25 +02:00
|
|
|
atomic_inc(&RQ_CIC(rq)->ioc->refcount);
|
|
|
|
cfqd->active_cic = RQ_CIC(rq);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-06-21 09:36:18 +02:00
|
|
|
if (RB_EMPTY_ROOT(&cfqq->sort_list))
|
2005-06-27 10:55:12 +02:00
|
|
|
break;
|
|
|
|
|
|
|
|
} while (dispatched < max_dispatch);
|
|
|
|
|
|
|
|
/*
|
2006-06-16 11:23:00 +02:00
|
|
|
* if slice end isn't set yet, set it.
|
2005-06-27 10:55:12 +02:00
|
|
|
*/
|
|
|
|
if (!cfqq->slice_end)
|
|
|
|
cfq_set_prio_slice(cfqd, cfqq);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* expire an async queue immediately if it has used up its slice. idle
|
|
|
|
* queue always expire after 1 dispatch round.
|
|
|
|
*/
|
|
|
|
if ((!cfq_cfqq_sync(cfqq) &&
|
|
|
|
cfqd->dispatch_slice >= cfq_prio_to_maxrq(cfqd, cfqq)) ||
|
2006-06-16 11:23:00 +02:00
|
|
|
cfq_class_idle(cfqq) ||
|
|
|
|
!cfq_cfqq_idle_window(cfqq))
|
2005-06-27 10:55:12 +02:00
|
|
|
cfq_slice_expired(cfqd, 0);
|
|
|
|
|
|
|
|
return dispatched;
|
|
|
|
}
|
|
|
|
|
2005-11-10 08:49:19 +01:00
|
|
|
static int
|
|
|
|
cfq_forced_dispatch_cfqqs(struct list_head *list)
|
|
|
|
{
|
|
|
|
struct cfq_queue *cfqq, *next;
|
2006-06-16 11:23:00 +02:00
|
|
|
int dispatched;
|
2005-11-10 08:49:19 +01:00
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
dispatched = 0;
|
2005-11-10 08:49:19 +01:00
|
|
|
list_for_each_entry_safe(cfqq, next, list, cfq_list) {
|
2006-07-13 12:39:25 +02:00
|
|
|
while (cfqq->next_rq) {
|
|
|
|
cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
|
2005-11-10 08:49:19 +01:00
|
|
|
dispatched++;
|
|
|
|
}
|
|
|
|
BUG_ON(!list_empty(&cfqq->fifo));
|
|
|
|
}
|
2006-06-16 11:23:00 +02:00
|
|
|
|
2005-11-10 08:49:19 +01:00
|
|
|
return dispatched;
|
|
|
|
}
|
|
|
|
|
|
|
|
static int
|
|
|
|
cfq_forced_dispatch(struct cfq_data *cfqd)
|
|
|
|
{
|
|
|
|
int i, dispatched = 0;
|
|
|
|
|
|
|
|
for (i = 0; i < CFQ_PRIO_LISTS; i++)
|
|
|
|
dispatched += cfq_forced_dispatch_cfqqs(&cfqd->rr_list[i]);
|
|
|
|
|
|
|
|
dispatched += cfq_forced_dispatch_cfqqs(&cfqd->busy_rr);
|
|
|
|
dispatched += cfq_forced_dispatch_cfqqs(&cfqd->cur_rr);
|
|
|
|
dispatched += cfq_forced_dispatch_cfqqs(&cfqd->idle_rr);
|
|
|
|
|
|
|
|
cfq_slice_expired(cfqd, 0);
|
|
|
|
|
|
|
|
BUG_ON(cfqd->busy_queues);
|
|
|
|
|
|
|
|
return dispatched;
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
static int
|
2005-10-20 16:42:29 +02:00
|
|
|
cfq_dispatch_requests(request_queue_t *q, int force)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
2006-06-16 11:23:00 +02:00
|
|
|
struct cfq_queue *cfqq, *prev_cfqq;
|
|
|
|
int dispatched;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
if (!cfqd->busy_queues)
|
|
|
|
return 0;
|
|
|
|
|
2005-11-10 08:49:19 +01:00
|
|
|
if (unlikely(force))
|
|
|
|
return cfq_forced_dispatch(cfqd);
|
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
dispatched = 0;
|
|
|
|
prev_cfqq = NULL;
|
|
|
|
while ((cfqq = cfq_select_queue(cfqd)) != NULL) {
|
2005-10-20 16:42:29 +02:00
|
|
|
int max_dispatch;
|
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
/*
|
|
|
|
* Don't repeat dispatch from the previous queue.
|
|
|
|
*/
|
|
|
|
if (prev_cfqq == cfqq)
|
|
|
|
break;
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_clear_cfqq_must_dispatch(cfqq);
|
|
|
|
cfq_clear_cfqq_wait_request(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
del_timer(&cfqd->idle_slice_timer);
|
|
|
|
|
2005-11-10 08:49:19 +01:00
|
|
|
max_dispatch = cfqd->cfq_quantum;
|
|
|
|
if (cfq_class_idle(cfqq))
|
|
|
|
max_dispatch = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
dispatched += __cfq_dispatch_requests(cfqd, cfqq, max_dispatch);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* If the dispatch cfqq has idling enabled and is still
|
|
|
|
* the active queue, break out.
|
|
|
|
*/
|
|
|
|
if (cfq_cfqq_idle_window(cfqq) && cfqd->active_queue)
|
|
|
|
break;
|
|
|
|
|
|
|
|
prev_cfqq = cfqq;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
return dispatched;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-07-13 12:39:25 +02:00
|
|
|
* task holds one reference to the queue, dropped when task exits. each rq
|
|
|
|
* in-flight on this queue also holds a reference, dropped when rq is freed.
|
2005-04-17 00:20:36 +02:00
|
|
|
*
|
|
|
|
* queue lock must be held here.
|
|
|
|
*/
|
|
|
|
static void cfq_put_queue(struct cfq_queue *cfqq)
|
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
struct cfq_data *cfqd = cfqq->cfqd;
|
|
|
|
|
|
|
|
BUG_ON(atomic_read(&cfqq->ref) <= 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
if (!atomic_dec_and_test(&cfqq->ref))
|
|
|
|
return;
|
|
|
|
|
|
|
|
BUG_ON(rb_first(&cfqq->sort_list));
|
2005-06-27 10:55:12 +02:00
|
|
|
BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
|
2005-06-27 10:56:24 +02:00
|
|
|
BUG_ON(cfq_cfqq_on_rr(cfqq));
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-02-28 09:35:11 +01:00
|
|
|
if (unlikely(cfqd->active_queue == cfqq))
|
2005-06-27 10:56:24 +02:00
|
|
|
__cfq_slice_expired(cfqd, cfqq, 0);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* it's on the empty list and still hashed
|
|
|
|
*/
|
|
|
|
list_del(&cfqq->cfq_list);
|
|
|
|
hlist_del(&cfqq->cfq_hash);
|
|
|
|
kmem_cache_free(cfq_pool, cfqq);
|
|
|
|
}
|
|
|
|
|
2006-07-18 22:24:11 +02:00
|
|
|
static struct cfq_queue *
|
2005-06-27 10:56:24 +02:00
|
|
|
__cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned int prio,
|
|
|
|
const int hashval)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
struct hlist_head *hash_list = &cfqd->cfq_hash[hashval];
|
2006-03-28 13:03:44 +02:00
|
|
|
struct hlist_node *entry;
|
|
|
|
struct cfq_queue *__cfqq;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-28 13:03:44 +02:00
|
|
|
hlist_for_each_entry(__cfqq, entry, hash_list, cfq_hash) {
|
2006-03-14 21:32:50 +01:00
|
|
|
const unsigned short __p = IOPRIO_PRIO_VALUE(__cfqq->org_ioprio_class, __cfqq->org_ioprio);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-28 13:03:44 +02:00
|
|
|
if (__cfqq->key == key && (__p == prio || !prio))
|
2005-04-17 00:20:36 +02:00
|
|
|
return __cfqq;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static struct cfq_queue *
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned short prio)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:56:24 +02:00
|
|
|
return __cfq_find_cfq_hash(cfqd, key, prio, hash_long(key, CFQ_QHASH_SHIFT));
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
static void cfq_free_io_context(struct io_context *ioc)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
struct cfq_io_context *__cic;
|
2006-03-28 08:59:01 +02:00
|
|
|
struct rb_node *n;
|
|
|
|
int freed = 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
while ((n = rb_first(&ioc->cic_root)) != NULL) {
|
|
|
|
__cic = rb_entry(n, struct cfq_io_context, rb_node);
|
|
|
|
rb_erase(&__cic->rb_node, &ioc->cic_root);
|
2005-06-27 10:55:12 +02:00
|
|
|
kmem_cache_free(cfq_ioc_pool, __cic);
|
2006-03-18 21:05:53 +01:00
|
|
|
freed++;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-07-19 05:07:12 +02:00
|
|
|
elv_ioc_count_mod(ioc_count, -freed);
|
|
|
|
|
|
|
|
if (ioc_gone && !elv_ioc_count_read(ioc_count))
|
2006-03-18 21:05:53 +01:00
|
|
|
complete(ioc_gone);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-07-22 16:48:31 +02:00
|
|
|
static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-07-22 16:48:31 +02:00
|
|
|
if (unlikely(cfqq == cfqd->active_queue))
|
|
|
|
__cfq_slice_expired(cfqd, cfqq, 0);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2006-07-22 16:48:31 +02:00
|
|
|
cfq_put_queue(cfqq);
|
|
|
|
}
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2006-07-22 16:48:31 +02:00
|
|
|
static void __cfq_exit_single_io_context(struct cfq_data *cfqd,
|
|
|
|
struct cfq_io_context *cic)
|
|
|
|
{
|
2006-08-29 09:05:44 +02:00
|
|
|
list_del_init(&cic->queue_list);
|
|
|
|
smp_wmb();
|
|
|
|
cic->key = NULL;
|
|
|
|
|
2006-03-18 19:38:01 +01:00
|
|
|
if (cic->cfqq[ASYNC]) {
|
2006-07-22 16:48:31 +02:00
|
|
|
cfq_exit_cfqq(cfqd, cic->cfqq[ASYNC]);
|
2006-03-18 19:38:01 +01:00
|
|
|
cic->cfqq[ASYNC] = NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (cic->cfqq[SYNC]) {
|
2006-07-22 16:48:31 +02:00
|
|
|
cfq_exit_cfqq(cfqd, cic->cfqq[SYNC]);
|
2006-03-18 19:38:01 +01:00
|
|
|
cic->cfqq[SYNC] = NULL;
|
|
|
|
}
|
2006-07-22 16:48:31 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Called with interrupts disabled
|
|
|
|
*/
|
|
|
|
static void cfq_exit_single_io_context(struct cfq_io_context *cic)
|
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = cic->key;
|
|
|
|
|
|
|
|
if (cfqd) {
|
|
|
|
request_queue_t *q = cfqd->queue;
|
|
|
|
|
2006-08-29 09:05:44 +02:00
|
|
|
spin_lock_irq(q->queue_lock);
|
2006-07-22 16:48:31 +02:00
|
|
|
__cfq_exit_single_io_context(cfqd, cic);
|
2006-08-29 09:05:44 +02:00
|
|
|
spin_unlock_irq(q->queue_lock);
|
2006-07-22 16:48:31 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
static void cfq_exit_io_context(struct io_context *ioc)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
struct cfq_io_context *__cic;
|
2006-03-28 08:59:01 +02:00
|
|
|
struct rb_node *n;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* put the reference this task is holding to the various queues
|
|
|
|
*/
|
2006-03-28 08:59:01 +02:00
|
|
|
|
|
|
|
n = rb_first(&ioc->cic_root);
|
|
|
|
while (n != NULL) {
|
|
|
|
__cic = rb_entry(n, struct cfq_io_context, rb_node);
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
cfq_exit_single_io_context(__cic);
|
2006-03-28 08:59:01 +02:00
|
|
|
n = rb_next(n);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
static struct cfq_io_context *
|
2005-10-21 09:20:53 +02:00
|
|
|
cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2006-07-19 23:39:40 +02:00
|
|
|
struct cfq_io_context *cic;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-19 23:39:40 +02:00
|
|
|
cic = kmem_cache_alloc_node(cfq_ioc_pool, gfp_mask, cfqd->queue->node);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (cic) {
|
2006-06-14 19:11:57 +02:00
|
|
|
memset(cic, 0, sizeof(*cic));
|
2005-06-27 10:55:12 +02:00
|
|
|
cic->last_end_request = jiffies;
|
2006-06-14 19:11:57 +02:00
|
|
|
INIT_LIST_HEAD(&cic->queue_list);
|
2005-06-27 10:55:12 +02:00
|
|
|
cic->dtor = cfq_free_io_context;
|
|
|
|
cic->exit = cfq_exit_io_context;
|
2006-07-19 05:07:12 +02:00
|
|
|
elv_ioc_count_inc(ioc_count);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return cic;
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
static void cfq_init_prio_data(struct cfq_queue *cfqq)
|
|
|
|
{
|
|
|
|
struct task_struct *tsk = current;
|
|
|
|
int ioprio_class;
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
if (!cfq_cfqq_prio_changed(cfqq))
|
2005-06-27 10:55:12 +02:00
|
|
|
return;
|
|
|
|
|
|
|
|
ioprio_class = IOPRIO_PRIO_CLASS(tsk->ioprio);
|
|
|
|
switch (ioprio_class) {
|
|
|
|
default:
|
|
|
|
printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
|
|
|
|
case IOPRIO_CLASS_NONE:
|
|
|
|
/*
|
|
|
|
* no prio set, place us in the middle of the BE classes
|
|
|
|
*/
|
|
|
|
cfqq->ioprio = task_nice_ioprio(tsk);
|
|
|
|
cfqq->ioprio_class = IOPRIO_CLASS_BE;
|
|
|
|
break;
|
|
|
|
case IOPRIO_CLASS_RT:
|
|
|
|
cfqq->ioprio = task_ioprio(tsk);
|
|
|
|
cfqq->ioprio_class = IOPRIO_CLASS_RT;
|
|
|
|
break;
|
|
|
|
case IOPRIO_CLASS_BE:
|
|
|
|
cfqq->ioprio = task_ioprio(tsk);
|
|
|
|
cfqq->ioprio_class = IOPRIO_CLASS_BE;
|
|
|
|
break;
|
|
|
|
case IOPRIO_CLASS_IDLE:
|
|
|
|
cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
|
|
|
|
cfqq->ioprio = 7;
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_clear_cfqq_idle_window(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
break;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* keep track of original prio settings in case we have to temporarily
|
|
|
|
* elevate the priority of this queue
|
|
|
|
*/
|
|
|
|
cfqq->org_ioprio = cfqq->ioprio;
|
|
|
|
cfqq->org_ioprio_class = cfqq->ioprio_class;
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
if (cfq_cfqq_on_rr(cfqq))
|
2005-06-27 10:55:12 +02:00
|
|
|
cfq_resort_rr_list(cfqq, 0);
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_clear_cfqq_prio_changed(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
2006-03-18 19:25:24 +01:00
|
|
|
static inline void changed_ioprio(struct cfq_io_context *cic)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
2006-03-18 19:25:24 +01:00
|
|
|
struct cfq_data *cfqd = cic->key;
|
|
|
|
struct cfq_queue *cfqq;
|
2006-10-30 19:54:23 +01:00
|
|
|
unsigned long flags;
|
2006-06-14 09:10:45 +02:00
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
if (unlikely(!cfqd))
|
|
|
|
return;
|
|
|
|
|
2006-10-30 19:54:23 +01:00
|
|
|
spin_lock_irqsave(cfqd->queue->queue_lock, flags);
|
2006-06-16 11:23:00 +02:00
|
|
|
|
|
|
|
cfqq = cic->cfqq[ASYNC];
|
|
|
|
if (cfqq) {
|
|
|
|
struct cfq_queue *new_cfqq;
|
|
|
|
new_cfqq = cfq_get_queue(cfqd, CFQ_KEY_ASYNC, cic->ioc->task,
|
|
|
|
GFP_ATOMIC);
|
|
|
|
if (new_cfqq) {
|
|
|
|
cic->cfqq[ASYNC] = new_cfqq;
|
|
|
|
cfq_put_queue(cfqq);
|
|
|
|
}
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
2006-06-16 11:23:00 +02:00
|
|
|
|
|
|
|
cfqq = cic->cfqq[SYNC];
|
|
|
|
if (cfqq)
|
|
|
|
cfq_mark_cfqq_prio_changed(cfqq);
|
|
|
|
|
2006-10-30 19:54:23 +01:00
|
|
|
spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
2006-08-29 09:05:44 +02:00
|
|
|
static void cfq_ioc_set_ioprio(struct io_context *ioc)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
2006-03-18 19:26:44 +01:00
|
|
|
struct cfq_io_context *cic;
|
2006-03-28 08:59:01 +02:00
|
|
|
struct rb_node *n;
|
2006-03-18 19:26:44 +01:00
|
|
|
|
2006-08-29 09:05:44 +02:00
|
|
|
ioc->ioprio_changed = 0;
|
2006-03-18 19:26:44 +01:00
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
n = rb_first(&ioc->cic_root);
|
|
|
|
while (n != NULL) {
|
|
|
|
cic = rb_entry(n, struct cfq_io_context, rb_node);
|
2006-05-30 21:11:04 +02:00
|
|
|
|
2006-03-18 19:25:24 +01:00
|
|
|
changed_ioprio(cic);
|
2006-03-28 08:59:01 +02:00
|
|
|
n = rb_next(n);
|
|
|
|
}
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static struct cfq_queue *
|
2006-03-18 20:58:37 +01:00
|
|
|
cfq_get_queue(struct cfq_data *cfqd, unsigned int key, struct task_struct *tsk,
|
2005-10-21 09:20:53 +02:00
|
|
|
gfp_t gfp_mask)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
|
|
|
const int hashval = hash_long(key, CFQ_QHASH_SHIFT);
|
|
|
|
struct cfq_queue *cfqq, *new_cfqq = NULL;
|
2006-03-18 20:58:37 +01:00
|
|
|
unsigned short ioprio;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
retry:
|
2006-03-18 20:58:37 +01:00
|
|
|
ioprio = tsk->ioprio;
|
2005-06-27 10:56:24 +02:00
|
|
|
cfqq = __cfq_find_cfq_hash(cfqd, key, ioprio, hashval);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
if (!cfqq) {
|
|
|
|
if (new_cfqq) {
|
|
|
|
cfqq = new_cfqq;
|
|
|
|
new_cfqq = NULL;
|
|
|
|
} else if (gfp_mask & __GFP_WAIT) {
|
2006-07-22 16:48:31 +02:00
|
|
|
/*
|
|
|
|
* Inform the allocator of the fact that we will
|
|
|
|
* just repeat this allocation if it fails, to allow
|
|
|
|
* the allocator to do whatever it needs to attempt to
|
|
|
|
* free memory.
|
|
|
|
*/
|
2005-06-27 10:55:12 +02:00
|
|
|
spin_unlock_irq(cfqd->queue->queue_lock);
|
2006-07-19 23:39:40 +02:00
|
|
|
new_cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask|__GFP_NOFAIL, cfqd->queue->node);
|
2005-06-27 10:55:12 +02:00
|
|
|
spin_lock_irq(cfqd->queue->queue_lock);
|
|
|
|
goto retry;
|
|
|
|
} else {
|
2006-07-19 23:39:40 +02:00
|
|
|
cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask, cfqd->queue->node);
|
2005-06-27 10:55:12 +02:00
|
|
|
if (!cfqq)
|
|
|
|
goto out;
|
|
|
|
}
|
|
|
|
|
|
|
|
memset(cfqq, 0, sizeof(*cfqq));
|
|
|
|
|
|
|
|
INIT_HLIST_NODE(&cfqq->cfq_hash);
|
|
|
|
INIT_LIST_HEAD(&cfqq->cfq_list);
|
|
|
|
INIT_LIST_HEAD(&cfqq->fifo);
|
|
|
|
|
|
|
|
cfqq->key = key;
|
|
|
|
hlist_add_head(&cfqq->cfq_hash, &cfqd->cfq_hash[hashval]);
|
|
|
|
atomic_set(&cfqq->ref, 0);
|
|
|
|
cfqq->cfqd = cfqd;
|
|
|
|
/*
|
|
|
|
* set ->slice_left to allow preemption for a new process
|
|
|
|
*/
|
|
|
|
cfqq->slice_left = 2 * cfqd->cfq_slice_idle;
|
2006-06-16 11:23:00 +02:00
|
|
|
cfq_mark_cfqq_idle_window(cfqq);
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_mark_cfqq_prio_changed(cfqq);
|
2006-07-28 09:48:51 +02:00
|
|
|
cfq_mark_cfqq_queue_new(cfqq);
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_init_prio_data(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
if (new_cfqq)
|
|
|
|
kmem_cache_free(cfq_pool, new_cfqq);
|
|
|
|
|
|
|
|
atomic_inc(&cfqq->ref);
|
|
|
|
out:
|
|
|
|
WARN_ON((gfp_mask & __GFP_WAIT) && !cfqq);
|
|
|
|
return cfqq;
|
|
|
|
}
|
|
|
|
|
2006-04-18 09:45:18 +02:00
|
|
|
static void
|
|
|
|
cfq_drop_dead_cic(struct io_context *ioc, struct cfq_io_context *cic)
|
|
|
|
{
|
2006-08-29 09:05:44 +02:00
|
|
|
WARN_ON(!list_empty(&cic->queue_list));
|
2006-04-18 09:45:18 +02:00
|
|
|
rb_erase(&cic->rb_node, &ioc->cic_root);
|
|
|
|
kmem_cache_free(cfq_ioc_pool, cic);
|
2006-07-19 05:07:12 +02:00
|
|
|
elv_ioc_count_dec(ioc_count);
|
2006-04-18 09:45:18 +02:00
|
|
|
}
|
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
static struct cfq_io_context *
|
|
|
|
cfq_cic_rb_lookup(struct cfq_data *cfqd, struct io_context *ioc)
|
|
|
|
{
|
2006-04-18 09:45:18 +02:00
|
|
|
struct rb_node *n;
|
2006-03-28 08:59:01 +02:00
|
|
|
struct cfq_io_context *cic;
|
2006-04-18 19:18:31 +02:00
|
|
|
void *k, *key = cfqd;
|
2006-03-28 08:59:01 +02:00
|
|
|
|
2006-04-18 09:45:18 +02:00
|
|
|
restart:
|
|
|
|
n = ioc->cic_root.rb_node;
|
2006-03-28 08:59:01 +02:00
|
|
|
while (n) {
|
|
|
|
cic = rb_entry(n, struct cfq_io_context, rb_node);
|
2006-04-18 19:18:31 +02:00
|
|
|
/* ->key must be copied to avoid race with cfq_exit_queue() */
|
|
|
|
k = cic->key;
|
|
|
|
if (unlikely(!k)) {
|
2006-04-18 09:45:18 +02:00
|
|
|
cfq_drop_dead_cic(ioc, cic);
|
|
|
|
goto restart;
|
|
|
|
}
|
2006-03-28 08:59:01 +02:00
|
|
|
|
2006-04-18 19:18:31 +02:00
|
|
|
if (key < k)
|
2006-03-28 08:59:01 +02:00
|
|
|
n = n->rb_left;
|
2006-04-18 19:18:31 +02:00
|
|
|
else if (key > k)
|
2006-03-28 08:59:01 +02:00
|
|
|
n = n->rb_right;
|
|
|
|
else
|
|
|
|
return cic;
|
|
|
|
}
|
|
|
|
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
|
|
|
static inline void
|
|
|
|
cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc,
|
|
|
|
struct cfq_io_context *cic)
|
|
|
|
{
|
2006-04-18 09:45:18 +02:00
|
|
|
struct rb_node **p;
|
|
|
|
struct rb_node *parent;
|
2006-03-28 08:59:01 +02:00
|
|
|
struct cfq_io_context *__cic;
|
2006-10-30 19:07:48 +01:00
|
|
|
unsigned long flags;
|
2006-04-18 19:18:31 +02:00
|
|
|
void *k;
|
2006-03-28 08:59:01 +02:00
|
|
|
|
|
|
|
cic->ioc = ioc;
|
|
|
|
cic->key = cfqd;
|
|
|
|
|
2006-04-18 09:45:18 +02:00
|
|
|
restart:
|
|
|
|
parent = NULL;
|
|
|
|
p = &ioc->cic_root.rb_node;
|
2006-03-28 08:59:01 +02:00
|
|
|
while (*p) {
|
|
|
|
parent = *p;
|
|
|
|
__cic = rb_entry(parent, struct cfq_io_context, rb_node);
|
2006-04-18 19:18:31 +02:00
|
|
|
/* ->key must be copied to avoid race with cfq_exit_queue() */
|
|
|
|
k = __cic->key;
|
|
|
|
if (unlikely(!k)) {
|
2006-08-21 08:36:12 +02:00
|
|
|
cfq_drop_dead_cic(ioc, __cic);
|
2006-04-18 09:45:18 +02:00
|
|
|
goto restart;
|
|
|
|
}
|
2006-03-28 08:59:01 +02:00
|
|
|
|
2006-04-18 19:18:31 +02:00
|
|
|
if (cic->key < k)
|
2006-03-28 08:59:01 +02:00
|
|
|
p = &(*p)->rb_left;
|
2006-04-18 19:18:31 +02:00
|
|
|
else if (cic->key > k)
|
2006-03-28 08:59:01 +02:00
|
|
|
p = &(*p)->rb_right;
|
|
|
|
else
|
|
|
|
BUG();
|
|
|
|
}
|
|
|
|
|
|
|
|
rb_link_node(&cic->rb_node, parent, p);
|
|
|
|
rb_insert_color(&cic->rb_node, &ioc->cic_root);
|
2006-08-29 09:05:44 +02:00
|
|
|
|
2006-10-30 19:07:48 +01:00
|
|
|
spin_lock_irqsave(cfqd->queue->queue_lock, flags);
|
2006-03-28 08:59:01 +02:00
|
|
|
list_add(&cic->queue_list, &cfqd->cic_list);
|
2006-10-30 19:07:48 +01:00
|
|
|
spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
|
2006-03-28 08:59:01 +02:00
|
|
|
}
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
/*
|
|
|
|
* Setup general io context and cfq io context. There can be several cfq
|
|
|
|
* io contexts per general io context, if this process is doing io to more
|
2006-03-28 08:59:01 +02:00
|
|
|
* than one device managed by cfq.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
|
|
|
static struct cfq_io_context *
|
2006-03-28 08:59:01 +02:00
|
|
|
cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
struct io_context *ioc = NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
struct cfq_io_context *cic;
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
might_sleep_if(gfp_mask & __GFP_WAIT);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-19 23:39:40 +02:00
|
|
|
ioc = get_io_context(gfp_mask, cfqd->queue->node);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (!ioc)
|
|
|
|
return NULL;
|
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
cic = cfq_cic_rb_lookup(cfqd, ioc);
|
|
|
|
if (cic)
|
|
|
|
goto out;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
cic = cfq_alloc_io_context(cfqd, gfp_mask);
|
|
|
|
if (cic == NULL)
|
|
|
|
goto err;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
cfq_cic_link(cfqd, ioc, cic);
|
2005-04-17 00:20:36 +02:00
|
|
|
out:
|
2006-08-29 09:05:44 +02:00
|
|
|
smp_read_barrier_depends();
|
|
|
|
if (unlikely(ioc->ioprio_changed))
|
|
|
|
cfq_ioc_set_ioprio(ioc);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
return cic;
|
|
|
|
err:
|
|
|
|
put_io_context(ioc);
|
|
|
|
return NULL;
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
static void
|
|
|
|
cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_io_context *cic)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
unsigned long elapsed, ttime;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* if this context already has stuff queued, thinktime is from
|
|
|
|
* last queue not last end
|
|
|
|
*/
|
|
|
|
#if 0
|
|
|
|
if (time_after(cic->last_end_request, cic->last_queue))
|
|
|
|
elapsed = jiffies - cic->last_end_request;
|
|
|
|
else
|
|
|
|
elapsed = jiffies - cic->last_queue;
|
|
|
|
#else
|
|
|
|
elapsed = jiffies - cic->last_end_request;
|
|
|
|
#endif
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
ttime = min(elapsed, 2UL * cfqd->cfq_slice_idle);
|
2005-06-17 16:15:10 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
cic->ttime_samples = (7*cic->ttime_samples + 256) / 8;
|
|
|
|
cic->ttime_total = (7*cic->ttime_total + 256*ttime) / 8;
|
|
|
|
cic->ttime_mean = (cic->ttime_total + 128) / cic->ttime_samples;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-28 13:03:44 +02:00
|
|
|
static void
|
|
|
|
cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_io_context *cic,
|
2006-07-13 12:39:25 +02:00
|
|
|
struct request *rq)
|
2006-03-28 13:03:44 +02:00
|
|
|
{
|
|
|
|
sector_t sdist;
|
|
|
|
u64 total;
|
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
if (cic->last_request_pos < rq->sector)
|
|
|
|
sdist = rq->sector - cic->last_request_pos;
|
2006-03-28 13:03:44 +02:00
|
|
|
else
|
2006-07-13 12:39:25 +02:00
|
|
|
sdist = cic->last_request_pos - rq->sector;
|
2006-03-28 13:03:44 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Don't allow the seek distance to get too large from the
|
|
|
|
* odd fragment, pagein, etc
|
|
|
|
*/
|
|
|
|
if (cic->seek_samples <= 60) /* second&third seek */
|
|
|
|
sdist = min(sdist, (cic->seek_mean * 4) + 2*1024*1024);
|
|
|
|
else
|
|
|
|
sdist = min(sdist, (cic->seek_mean * 4) + 2*1024*64);
|
|
|
|
|
|
|
|
cic->seek_samples = (7*cic->seek_samples + 256) / 8;
|
|
|
|
cic->seek_total = (7*cic->seek_total + (u64)256*sdist) / 8;
|
|
|
|
total = cic->seek_total + (cic->seek_samples/2);
|
|
|
|
do_div(total, cic->seek_samples);
|
|
|
|
cic->seek_mean = (sector_t)total;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* Disable idle window if the process thinks too long or seeks so much that
|
|
|
|
* it doesn't matter
|
|
|
|
*/
|
|
|
|
static void
|
|
|
|
cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
|
|
|
|
struct cfq_io_context *cic)
|
|
|
|
{
|
2005-06-27 10:56:24 +02:00
|
|
|
int enable_idle = cfq_cfqq_idle_window(cfqq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
if (!cic->ioc->task || !cfqd->cfq_slice_idle ||
|
|
|
|
(cfqd->hw_tag && CIC_SEEKY(cic)))
|
2005-06-27 10:55:12 +02:00
|
|
|
enable_idle = 0;
|
|
|
|
else if (sample_valid(cic->ttime_samples)) {
|
|
|
|
if (cic->ttime_mean > cfqd->cfq_slice_idle)
|
|
|
|
enable_idle = 0;
|
|
|
|
else
|
|
|
|
enable_idle = 1;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
if (enable_idle)
|
|
|
|
cfq_mark_cfqq_idle_window(cfqq);
|
|
|
|
else
|
|
|
|
cfq_clear_cfqq_idle_window(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* Check if new_cfqq should preempt the currently active queue. Return 0 for
|
|
|
|
* no or if we aren't sure, a 1 will cause a preempt.
|
|
|
|
*/
|
|
|
|
static int
|
|
|
|
cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq,
|
2006-07-13 12:39:25 +02:00
|
|
|
struct request *rq)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
|
|
|
struct cfq_queue *cfqq = cfqd->active_queue;
|
|
|
|
|
|
|
|
if (cfq_class_idle(new_cfqq))
|
|
|
|
return 0;
|
|
|
|
|
|
|
|
if (!cfqq)
|
2006-06-16 11:23:00 +02:00
|
|
|
return 0;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
if (cfq_class_idle(cfqq))
|
|
|
|
return 1;
|
2005-06-27 10:56:24 +02:00
|
|
|
if (!cfq_cfqq_wait_request(new_cfqq))
|
2005-06-27 10:55:12 +02:00
|
|
|
return 0;
|
|
|
|
/*
|
|
|
|
* if it doesn't have slice left, forget it
|
|
|
|
*/
|
|
|
|
if (new_cfqq->slice_left < cfqd->cfq_slice_idle)
|
|
|
|
return 0;
|
2006-07-23 01:42:19 +02:00
|
|
|
/*
|
|
|
|
* if the new request is sync, but the currently running queue is
|
|
|
|
* not, let the sync request have priority.
|
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
|
2005-06-27 10:55:12 +02:00
|
|
|
return 1;
|
2006-07-23 01:42:19 +02:00
|
|
|
/*
|
|
|
|
* So both queues are sync. Let the new request get disk time if
|
|
|
|
* it's a metadata request and the current queue is doing regular IO.
|
|
|
|
*/
|
|
|
|
if (rq_is_meta(rq) && !cfqq->meta_pending)
|
|
|
|
return 1;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
return 0;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* cfqq preempts the active queue. if we allowed preempt with no slice left,
|
|
|
|
* let it have half of its nominal slice.
|
|
|
|
*/
|
|
|
|
static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
|
|
|
|
{
|
2006-07-19 20:29:12 +02:00
|
|
|
cfq_slice_expired(cfqd, 1);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
if (!cfqq->slice_left)
|
|
|
|
cfqq->slice_left = cfq_prio_to_slice(cfqd, cfqq) / 2;
|
|
|
|
|
2006-07-19 20:29:12 +02:00
|
|
|
/*
|
|
|
|
* Put the new queue at the front of the of the current list,
|
|
|
|
* so we know that it will be selected next.
|
|
|
|
*/
|
|
|
|
BUG_ON(!cfq_cfqq_on_rr(cfqq));
|
|
|
|
list_move(&cfqq->cfq_list, &cfqd->cur_rr);
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
cfqq->slice_end = cfqq->slice_left + jiffies;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2006-07-13 12:39:25 +02:00
|
|
|
* Called when a new fs request (rq) is added (to cfqq). Check if there's
|
2005-06-27 10:55:12 +02:00
|
|
|
* something we should do about it
|
|
|
|
*/
|
|
|
|
static void
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
|
|
|
|
struct request *rq)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_io_context *cic = RQ_CIC(rq);
|
2006-06-01 10:09:56 +02:00
|
|
|
|
2006-07-23 01:42:19 +02:00
|
|
|
if (rq_is_meta(rq))
|
|
|
|
cfqq->meta_pending++;
|
|
|
|
|
2006-07-13 12:33:14 +02:00
|
|
|
/*
|
2006-07-13 12:37:56 +02:00
|
|
|
* check if this request is a better next-serve candidate)) {
|
2006-07-13 12:33:14 +02:00
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq);
|
|
|
|
BUG_ON(!cfqq->next_rq);
|
2006-07-13 12:33:14 +02:00
|
|
|
|
2005-08-24 14:57:54 +02:00
|
|
|
/*
|
|
|
|
* we never wait for an async request and we don't allow preemption
|
|
|
|
* of an async request. so just return early
|
|
|
|
*/
|
2006-07-13 12:39:25 +02:00
|
|
|
if (!rq_is_sync(rq)) {
|
2006-06-01 10:09:56 +02:00
|
|
|
/*
|
|
|
|
* sync process issued an async request, if it's waiting
|
|
|
|
* then expire it and kick rq handling.
|
|
|
|
*/
|
|
|
|
if (cic == cfqd->active_cic &&
|
|
|
|
del_timer(&cfqd->idle_slice_timer)) {
|
|
|
|
cfq_slice_expired(cfqd, 0);
|
2006-07-20 14:54:05 +02:00
|
|
|
blk_start_queueing(cfqd->queue);
|
2006-06-01 10:09:56 +02:00
|
|
|
}
|
2005-08-24 14:57:54 +02:00
|
|
|
return;
|
2006-06-01 10:09:56 +02:00
|
|
|
}
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2005-08-24 14:57:54 +02:00
|
|
|
cfq_update_io_thinktime(cfqd, cic);
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_update_io_seektime(cfqd, cic, rq);
|
2005-08-24 14:57:54 +02:00
|
|
|
cfq_update_idle_window(cfqd, cfqq, cic);
|
|
|
|
|
|
|
|
cic->last_queue = jiffies;
|
2006-07-13 12:39:25 +02:00
|
|
|
cic->last_request_pos = rq->sector + rq->nr_sectors;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
if (cfqq == cfqd->active_queue) {
|
|
|
|
/*
|
|
|
|
* if we are waiting for a request for this queue, let it rip
|
|
|
|
* immediately and flag that we must not expire this queue
|
|
|
|
* just now
|
|
|
|
*/
|
2005-06-27 10:56:24 +02:00
|
|
|
if (cfq_cfqq_wait_request(cfqq)) {
|
|
|
|
cfq_mark_cfqq_must_dispatch(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
del_timer(&cfqd->idle_slice_timer);
|
2006-07-20 14:54:05 +02:00
|
|
|
blk_start_queueing(cfqd->queue);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
2006-07-13 12:39:25 +02:00
|
|
|
} else if (cfq_should_preempt(cfqd, cfqq, rq)) {
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* not the active queue - expire current slice if it is
|
|
|
|
* idle and has expired it's mean thinktime or this new queue
|
|
|
|
* has some old slice time left and is of higher priority
|
|
|
|
*/
|
|
|
|
cfq_preempt_queue(cfqd, cfqq);
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_mark_cfqq_must_dispatch(cfqq);
|
2006-07-20 14:54:05 +02:00
|
|
|
blk_start_queueing(cfqd->queue);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
static void cfq_insert_request(request_queue_t *q, struct request *rq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-10-20 16:42:29 +02:00
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_queue *cfqq = RQ_CFQQ(rq);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
cfq_init_prio_data(cfqq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_add_rq_rb(rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
list_add_tail(&rq->queuelist, &cfqq->fifo);
|
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
cfq_rq_enqueued(cfqd, cfqq, rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void cfq_completed_request(request_queue_t *q, struct request *rq)
|
|
|
|
{
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_queue *cfqq = RQ_CFQQ(rq);
|
2005-10-20 16:42:29 +02:00
|
|
|
struct cfq_data *cfqd = cfqq->cfqd;
|
2006-07-13 12:37:56 +02:00
|
|
|
const int sync = rq_is_sync(rq);
|
2005-10-20 16:42:29 +02:00
|
|
|
unsigned long now;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
now = jiffies;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
WARN_ON(!cfqd->rq_in_driver);
|
|
|
|
WARN_ON(!cfqq->on_dispatch[sync]);
|
|
|
|
cfqd->rq_in_driver--;
|
|
|
|
cfqq->on_dispatch[sync]--;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-10-20 16:42:29 +02:00
|
|
|
if (!cfq_class_idle(cfqq))
|
|
|
|
cfqd->last_end_request = now;
|
2005-06-27 10:56:24 +02:00
|
|
|
|
2006-07-28 09:48:51 +02:00
|
|
|
if (!cfq_cfqq_dispatched(cfqq) && cfq_cfqq_on_rr(cfqq))
|
|
|
|
cfq_resort_rr_list(cfqq, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-06-16 11:23:00 +02:00
|
|
|
if (sync)
|
2006-07-13 12:39:25 +02:00
|
|
|
RQ_CIC(rq)->last_end_request = now;
|
2006-06-16 11:23:00 +02:00
|
|
|
|
|
|
|
/*
|
|
|
|
* If this is the active queue, check if it needs to be expired,
|
|
|
|
* or if we want to idle in case it has no pending requests.
|
|
|
|
*/
|
|
|
|
if (cfqd->active_queue == cfqq) {
|
|
|
|
if (time_after(now, cfqq->slice_end))
|
|
|
|
cfq_slice_expired(cfqd, 0);
|
2006-06-21 09:36:18 +02:00
|
|
|
else if (sync && RB_EMPTY_ROOT(&cfqq->sort_list)) {
|
2006-06-16 11:23:00 +02:00
|
|
|
if (!cfq_arm_slice_timer(cfqd, cfqq))
|
|
|
|
cfq_schedule_dispatch(cfqd);
|
|
|
|
}
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* we temporarily boost lower priority queues if they are holding fs exclusive
|
|
|
|
* resources. they are boosted to normal prio (CLASS_BE/4)
|
|
|
|
*/
|
|
|
|
static void cfq_prio_boost(struct cfq_queue *cfqq)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
const int ioprio_class = cfqq->ioprio_class;
|
|
|
|
const int ioprio = cfqq->ioprio;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
if (has_fs_excl()) {
|
|
|
|
/*
|
|
|
|
* boost idle prio on transactions that would lock out other
|
|
|
|
* users of the filesystem
|
|
|
|
*/
|
|
|
|
if (cfq_class_idle(cfqq))
|
|
|
|
cfqq->ioprio_class = IOPRIO_CLASS_BE;
|
|
|
|
if (cfqq->ioprio > IOPRIO_NORM)
|
|
|
|
cfqq->ioprio = IOPRIO_NORM;
|
|
|
|
} else {
|
|
|
|
/*
|
|
|
|
* check if we need to unboost the queue
|
|
|
|
*/
|
|
|
|
if (cfqq->ioprio_class != cfqq->org_ioprio_class)
|
|
|
|
cfqq->ioprio_class = cfqq->org_ioprio_class;
|
|
|
|
if (cfqq->ioprio != cfqq->org_ioprio)
|
|
|
|
cfqq->ioprio = cfqq->org_ioprio;
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* refile between round-robin lists if we moved the priority class
|
|
|
|
*/
|
|
|
|
if ((ioprio_class != cfqq->ioprio_class || ioprio != cfqq->ioprio) &&
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_cfqq_on_rr(cfqq))
|
2005-06-27 10:55:12 +02:00
|
|
|
cfq_resort_rr_list(cfqq, 0);
|
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-22 16:48:31 +02:00
|
|
|
static inline int __cfq_may_queue(struct cfq_queue *cfqq)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
2005-06-27 10:56:24 +02:00
|
|
|
if ((cfq_cfqq_wait_request(cfqq) || cfq_cfqq_must_alloc(cfqq)) &&
|
2005-06-28 05:14:05 +02:00
|
|
|
!cfq_cfqq_must_alloc_slice(cfqq)) {
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_mark_cfqq_must_alloc_slice(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
return ELV_MQUEUE_MUST;
|
2005-06-27 10:56:24 +02:00
|
|
|
}
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
return ELV_MQUEUE_MAY;
|
|
|
|
}
|
|
|
|
|
2006-07-28 09:32:57 +02:00
|
|
|
static int cfq_may_queue(request_queue_t *q, int rw)
|
2005-06-27 10:55:12 +02:00
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
|
|
|
struct task_struct *tsk = current;
|
|
|
|
struct cfq_queue *cfqq;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* don't force setup of a queue from here, as a call to may_queue
|
|
|
|
* does not necessarily imply that a request actually will be queued.
|
|
|
|
* so just lookup a possibly existing queue, or return 'may queue'
|
|
|
|
* if that fails
|
|
|
|
*/
|
2005-06-27 10:56:24 +02:00
|
|
|
cfqq = cfq_find_cfq_hash(cfqd, cfq_queue_pid(tsk, rw), tsk->ioprio);
|
2005-06-27 10:55:12 +02:00
|
|
|
if (cfqq) {
|
|
|
|
cfq_init_prio_data(cfqq);
|
|
|
|
cfq_prio_boost(cfqq);
|
|
|
|
|
2006-07-22 16:48:31 +02:00
|
|
|
return __cfq_may_queue(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return ELV_MQUEUE_MAY;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* queue lock held here
|
|
|
|
*/
|
|
|
|
static void cfq_put_request(request_queue_t *q, struct request *rq)
|
|
|
|
{
|
2006-07-13 12:39:25 +02:00
|
|
|
struct cfq_queue *cfqq = RQ_CFQQ(rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
if (cfqq) {
|
2005-06-27 10:55:12 +02:00
|
|
|
const int rw = rq_data_dir(rq);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
BUG_ON(!cfqq->allocated[rw]);
|
|
|
|
cfqq->allocated[rw]--;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
put_io_context(RQ_CIC(rq)->ioc);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
rq->elevator_private = NULL;
|
2006-07-13 12:39:25 +02:00
|
|
|
rq->elevator_private2 = NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
cfq_put_queue(cfqq);
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
2005-06-27 10:55:12 +02:00
|
|
|
* Allocate cfq data structures associated with this request.
|
2005-04-17 00:20:36 +02:00
|
|
|
*/
|
2005-06-27 10:55:12 +02:00
|
|
|
static int
|
2006-07-28 09:32:57 +02:00
|
|
|
cfq_set_request(request_queue_t *q, struct request *rq, gfp_t gfp_mask)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = q->elevator->elevator_data;
|
2005-06-27 10:56:24 +02:00
|
|
|
struct task_struct *tsk = current;
|
2005-04-17 00:20:36 +02:00
|
|
|
struct cfq_io_context *cic;
|
|
|
|
const int rw = rq_data_dir(rq);
|
2005-06-27 10:56:24 +02:00
|
|
|
pid_t key = cfq_queue_pid(tsk, rw);
|
2005-06-27 10:55:12 +02:00
|
|
|
struct cfq_queue *cfqq;
|
2005-04-17 00:20:36 +02:00
|
|
|
unsigned long flags;
|
2006-03-18 19:38:01 +01:00
|
|
|
int is_sync = key != CFQ_KEY_ASYNC;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
might_sleep_if(gfp_mask & __GFP_WAIT);
|
|
|
|
|
2006-03-28 08:59:01 +02:00
|
|
|
cic = cfq_get_io_context(cfqd, gfp_mask);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
spin_lock_irqsave(q->queue_lock, flags);
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
if (!cic)
|
|
|
|
goto queue_fail;
|
|
|
|
|
2006-03-18 19:38:01 +01:00
|
|
|
if (!cic->cfqq[is_sync]) {
|
2006-03-18 20:58:37 +01:00
|
|
|
cfqq = cfq_get_queue(cfqd, key, tsk, gfp_mask);
|
2005-06-27 10:55:12 +02:00
|
|
|
if (!cfqq)
|
|
|
|
goto queue_fail;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-03-18 19:38:01 +01:00
|
|
|
cic->cfqq[is_sync] = cfqq;
|
2005-06-27 10:55:12 +02:00
|
|
|
} else
|
2006-03-18 19:38:01 +01:00
|
|
|
cfqq = cic->cfqq[is_sync];
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
cfqq->allocated[rw]++;
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_clear_cfqq_must_alloc(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
atomic_inc(&cfqq->ref);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
spin_unlock_irqrestore(q->queue_lock, flags);
|
2005-06-27 10:56:24 +02:00
|
|
|
|
2006-07-13 12:39:25 +02:00
|
|
|
rq->elevator_private = cic;
|
|
|
|
rq->elevator_private2 = cfqq;
|
|
|
|
return 0;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
queue_fail:
|
|
|
|
if (cic)
|
|
|
|
put_io_context(cic->ioc);
|
2006-07-22 16:48:31 +02:00
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_schedule_dispatch(cfqd);
|
2005-04-17 00:20:36 +02:00
|
|
|
spin_unlock_irqrestore(q->queue_lock, flags);
|
|
|
|
return 1;
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
static void cfq_kick_queue(void *data)
|
|
|
|
{
|
|
|
|
request_queue_t *q = data;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(q->queue_lock, flags);
|
2006-07-20 14:54:05 +02:00
|
|
|
blk_start_queueing(q);
|
2005-06-27 10:55:12 +02:00
|
|
|
spin_unlock_irqrestore(q->queue_lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Timer running if the active_queue is currently idling inside its time slice
|
|
|
|
*/
|
|
|
|
static void cfq_idle_slice_timer(unsigned long data)
|
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = (struct cfq_data *) data;
|
|
|
|
struct cfq_queue *cfqq;
|
|
|
|
unsigned long flags;
|
|
|
|
|
|
|
|
spin_lock_irqsave(cfqd->queue->queue_lock, flags);
|
|
|
|
|
|
|
|
if ((cfqq = cfqd->active_queue) != NULL) {
|
|
|
|
unsigned long now = jiffies;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* expired
|
|
|
|
*/
|
|
|
|
if (time_after(now, cfqq->slice_end))
|
|
|
|
goto expire;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* only expire and reinvoke request handler, if there are
|
|
|
|
* other queues with pending requests
|
|
|
|
*/
|
2006-06-16 11:23:00 +02:00
|
|
|
if (!cfqd->busy_queues)
|
2005-06-27 10:55:12 +02:00
|
|
|
goto out_cont;
|
|
|
|
|
|
|
|
/*
|
|
|
|
* not expired and it has a request pending, let it dispatch
|
|
|
|
*/
|
2006-06-21 09:36:18 +02:00
|
|
|
if (!RB_EMPTY_ROOT(&cfqq->sort_list)) {
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_mark_cfqq_must_dispatch(cfqq);
|
2005-06-27 10:55:12 +02:00
|
|
|
goto out_kick;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
expire:
|
|
|
|
cfq_slice_expired(cfqd, 0);
|
|
|
|
out_kick:
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_schedule_dispatch(cfqd);
|
2005-06-27 10:55:12 +02:00
|
|
|
out_cont:
|
|
|
|
spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* Timer running if an idle class queue is waiting for service
|
|
|
|
*/
|
|
|
|
static void cfq_idle_class_timer(unsigned long data)
|
|
|
|
{
|
|
|
|
struct cfq_data *cfqd = (struct cfq_data *) data;
|
|
|
|
unsigned long flags, end;
|
|
|
|
|
|
|
|
spin_lock_irqsave(cfqd->queue->queue_lock, flags);
|
|
|
|
|
|
|
|
/*
|
|
|
|
* race with a non-idle queue, reset timer
|
|
|
|
*/
|
|
|
|
end = cfqd->last_end_request + CFQ_IDLE_GRACE;
|
2006-06-01 10:13:43 +02:00
|
|
|
if (!time_after_eq(jiffies, end))
|
|
|
|
mod_timer(&cfqd->idle_class_timer, end);
|
|
|
|
else
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_schedule_dispatch(cfqd);
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
|
|
|
|
}
|
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
|
|
|
|
{
|
|
|
|
del_timer_sync(&cfqd->idle_slice_timer);
|
|
|
|
del_timer_sync(&cfqd->idle_class_timer);
|
|
|
|
blk_sync_queue(cfqd->queue);
|
|
|
|
}
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
static void cfq_exit_queue(elevator_t *e)
|
|
|
|
{
|
2005-06-27 10:55:12 +02:00
|
|
|
struct cfq_data *cfqd = e->elevator_data;
|
2006-03-18 19:51:22 +01:00
|
|
|
request_queue_t *q = cfqd->queue;
|
2005-06-27 10:55:12 +02:00
|
|
|
|
2005-06-27 10:56:24 +02:00
|
|
|
cfq_shutdown_timer_wq(cfqd);
|
2006-03-28 08:59:01 +02:00
|
|
|
|
2006-03-18 19:51:22 +01:00
|
|
|
spin_lock_irq(q->queue_lock);
|
2006-03-28 08:59:01 +02:00
|
|
|
|
2006-03-18 19:51:22 +01:00
|
|
|
if (cfqd->active_queue)
|
|
|
|
__cfq_slice_expired(cfqd, cfqd->active_queue, 0);
|
2006-03-28 08:59:01 +02:00
|
|
|
|
|
|
|
while (!list_empty(&cfqd->cic_list)) {
|
2006-03-18 19:51:22 +01:00
|
|
|
struct cfq_io_context *cic = list_entry(cfqd->cic_list.next,
|
|
|
|
struct cfq_io_context,
|
|
|
|
queue_list);
|
2006-07-22 16:48:31 +02:00
|
|
|
|
|
|
|
__cfq_exit_single_io_context(cfqd, cic);
|
2006-03-18 19:51:22 +01:00
|
|
|
}
|
2006-03-28 08:59:01 +02:00
|
|
|
|
2006-03-18 19:51:22 +01:00
|
|
|
spin_unlock_irq(q->queue_lock);
|
2006-03-18 18:05:37 +01:00
|
|
|
|
|
|
|
cfq_shutdown_timer_wq(cfqd);
|
|
|
|
|
|
|
|
kfree(cfqd->cfq_hash);
|
|
|
|
kfree(cfqd);
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
2006-06-08 08:49:06 +02:00
|
|
|
static void *cfq_init_queue(request_queue_t *q, elevator_t *e)
|
2005-04-17 00:20:36 +02:00
|
|
|
{
|
|
|
|
struct cfq_data *cfqd;
|
|
|
|
int i;
|
|
|
|
|
2006-07-19 23:39:40 +02:00
|
|
|
cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (!cfqd)
|
2006-06-08 08:49:06 +02:00
|
|
|
return NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
memset(cfqd, 0, sizeof(*cfqd));
|
2005-06-27 10:55:12 +02:00
|
|
|
|
|
|
|
for (i = 0; i < CFQ_PRIO_LISTS; i++)
|
|
|
|
INIT_LIST_HEAD(&cfqd->rr_list[i]);
|
|
|
|
|
|
|
|
INIT_LIST_HEAD(&cfqd->busy_rr);
|
|
|
|
INIT_LIST_HEAD(&cfqd->cur_rr);
|
|
|
|
INIT_LIST_HEAD(&cfqd->idle_rr);
|
2006-03-18 19:51:22 +01:00
|
|
|
INIT_LIST_HEAD(&cfqd->cic_list);
|
2005-04-17 00:20:36 +02:00
|
|
|
|
2006-07-19 23:39:40 +02:00
|
|
|
cfqd->cfq_hash = kmalloc_node(sizeof(struct hlist_head) * CFQ_QHASH_ENTRIES, GFP_KERNEL, q->node);
|
2005-04-17 00:20:36 +02:00
|
|
|
if (!cfqd->cfq_hash)
|
2006-07-13 12:39:25 +02:00
|
|
|
goto out_free;
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
for (i = 0; i < CFQ_QHASH_ENTRIES; i++)
|
|
|
|
INIT_HLIST_HEAD(&cfqd->cfq_hash[i]);
|
|
|
|
|
|
|
|
cfqd->queue = q;
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
init_timer(&cfqd->idle_slice_timer);
|
|
|
|
cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
|
|
|
|
cfqd->idle_slice_timer.data = (unsigned long) cfqd;
|
|
|
|
|
|
|
|
init_timer(&cfqd->idle_class_timer);
|
|
|
|
cfqd->idle_class_timer.function = cfq_idle_class_timer;
|
|
|
|
cfqd->idle_class_timer.data = (unsigned long) cfqd;
|
|
|
|
|
|
|
|
INIT_WORK(&cfqd->unplug_work, cfq_kick_queue, q);
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
cfqd->cfq_quantum = cfq_quantum;
|
2005-06-27 10:55:12 +02:00
|
|
|
cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
|
|
|
|
cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
|
2005-04-17 00:20:36 +02:00
|
|
|
cfqd->cfq_back_max = cfq_back_max;
|
|
|
|
cfqd->cfq_back_penalty = cfq_back_penalty;
|
2005-06-27 10:55:12 +02:00
|
|
|
cfqd->cfq_slice[0] = cfq_slice_async;
|
|
|
|
cfqd->cfq_slice[1] = cfq_slice_sync;
|
|
|
|
cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
|
|
|
|
cfqd->cfq_slice_idle = cfq_slice_idle;
|
2005-06-27 10:56:24 +02:00
|
|
|
|
2006-06-08 08:49:06 +02:00
|
|
|
return cfqd;
|
2006-07-13 12:39:25 +02:00
|
|
|
out_free:
|
2005-04-17 00:20:36 +02:00
|
|
|
kfree(cfqd);
|
2006-06-08 08:49:06 +02:00
|
|
|
return NULL;
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
static void cfq_slab_kill(void)
|
|
|
|
{
|
|
|
|
if (cfq_pool)
|
|
|
|
kmem_cache_destroy(cfq_pool);
|
|
|
|
if (cfq_ioc_pool)
|
|
|
|
kmem_cache_destroy(cfq_ioc_pool);
|
|
|
|
}
|
|
|
|
|
|
|
|
static int __init cfq_slab_setup(void)
|
|
|
|
{
|
|
|
|
cfq_pool = kmem_cache_create("cfq_pool", sizeof(struct cfq_queue), 0, 0,
|
|
|
|
NULL, NULL);
|
|
|
|
if (!cfq_pool)
|
|
|
|
goto fail;
|
|
|
|
|
|
|
|
cfq_ioc_pool = kmem_cache_create("cfq_ioc_pool",
|
|
|
|
sizeof(struct cfq_io_context), 0, 0, NULL, NULL);
|
|
|
|
if (!cfq_ioc_pool)
|
|
|
|
goto fail;
|
|
|
|
|
|
|
|
return 0;
|
|
|
|
fail:
|
|
|
|
cfq_slab_kill();
|
|
|
|
return -ENOMEM;
|
|
|
|
}
|
|
|
|
|
|
|
|
/*
|
|
|
|
* sysfs parts below -->
|
|
|
|
*/
|
|
|
|
|
|
|
|
static ssize_t
|
|
|
|
cfq_var_show(unsigned int var, char *page)
|
|
|
|
{
|
|
|
|
return sprintf(page, "%d\n", var);
|
|
|
|
}
|
|
|
|
|
|
|
|
static ssize_t
|
|
|
|
cfq_var_store(unsigned int *var, const char *page, size_t count)
|
|
|
|
{
|
|
|
|
char *p = (char *) page;
|
|
|
|
|
|
|
|
*var = simple_strtoul(p, &p, 10);
|
|
|
|
return count;
|
|
|
|
}
|
|
|
|
|
|
|
|
#define SHOW_FUNCTION(__FUNC, __VAR, __CONV) \
|
2006-03-19 00:35:43 +01:00
|
|
|
static ssize_t __FUNC(elevator_t *e, char *page) \
|
2005-04-17 00:20:36 +02:00
|
|
|
{ \
|
2006-03-19 00:35:43 +01:00
|
|
|
struct cfq_data *cfqd = e->elevator_data; \
|
2005-04-17 00:20:36 +02:00
|
|
|
unsigned int __data = __VAR; \
|
|
|
|
if (__CONV) \
|
|
|
|
__data = jiffies_to_msecs(__data); \
|
|
|
|
return cfq_var_show(__data, (page)); \
|
|
|
|
}
|
|
|
|
SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
|
2005-06-27 10:55:12 +02:00
|
|
|
SHOW_FUNCTION(cfq_fifo_expire_sync_show, cfqd->cfq_fifo_expire[1], 1);
|
|
|
|
SHOW_FUNCTION(cfq_fifo_expire_async_show, cfqd->cfq_fifo_expire[0], 1);
|
2006-03-19 04:27:18 +01:00
|
|
|
SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
|
|
|
|
SHOW_FUNCTION(cfq_back_seek_penalty_show, cfqd->cfq_back_penalty, 0);
|
2005-06-27 10:55:12 +02:00
|
|
|
SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
|
|
|
|
SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
|
|
|
|
SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
|
|
|
|
SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
#undef SHOW_FUNCTION
|
|
|
|
|
|
|
|
#define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV) \
|
2006-03-19 00:35:43 +01:00
|
|
|
static ssize_t __FUNC(elevator_t *e, const char *page, size_t count) \
|
2005-04-17 00:20:36 +02:00
|
|
|
{ \
|
2006-03-19 00:35:43 +01:00
|
|
|
struct cfq_data *cfqd = e->elevator_data; \
|
2005-04-17 00:20:36 +02:00
|
|
|
unsigned int __data; \
|
|
|
|
int ret = cfq_var_store(&__data, (page), count); \
|
|
|
|
if (__data < (MIN)) \
|
|
|
|
__data = (MIN); \
|
|
|
|
else if (__data > (MAX)) \
|
|
|
|
__data = (MAX); \
|
|
|
|
if (__CONV) \
|
|
|
|
*(__PTR) = msecs_to_jiffies(__data); \
|
|
|
|
else \
|
|
|
|
*(__PTR) = __data; \
|
|
|
|
return ret; \
|
|
|
|
}
|
|
|
|
STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
|
2005-06-27 10:55:12 +02:00
|
|
|
STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1, UINT_MAX, 1);
|
|
|
|
STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1, UINT_MAX, 1);
|
2006-03-19 04:27:18 +01:00
|
|
|
STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
|
|
|
|
STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1, UINT_MAX, 0);
|
2005-06-27 10:55:12 +02:00
|
|
|
STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
|
|
|
|
STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
|
|
|
|
STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
|
|
|
|
STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1, UINT_MAX, 0);
|
2005-04-17 00:20:36 +02:00
|
|
|
#undef STORE_FUNCTION
|
|
|
|
|
2006-03-19 04:27:18 +01:00
|
|
|
#define CFQ_ATTR(name) \
|
|
|
|
__ATTR(name, S_IRUGO|S_IWUSR, cfq_##name##_show, cfq_##name##_store)
|
|
|
|
|
|
|
|
static struct elv_fs_entry cfq_attrs[] = {
|
|
|
|
CFQ_ATTR(quantum),
|
|
|
|
CFQ_ATTR(fifo_expire_sync),
|
|
|
|
CFQ_ATTR(fifo_expire_async),
|
|
|
|
CFQ_ATTR(back_seek_max),
|
|
|
|
CFQ_ATTR(back_seek_penalty),
|
|
|
|
CFQ_ATTR(slice_sync),
|
|
|
|
CFQ_ATTR(slice_async),
|
|
|
|
CFQ_ATTR(slice_async_rq),
|
|
|
|
CFQ_ATTR(slice_idle),
|
|
|
|
__ATTR_NULL
|
2005-04-17 00:20:36 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
static struct elevator_type iosched_cfq = {
|
|
|
|
.ops = {
|
|
|
|
.elevator_merge_fn = cfq_merge,
|
|
|
|
.elevator_merged_fn = cfq_merged_request,
|
|
|
|
.elevator_merge_req_fn = cfq_merged_requests,
|
2005-10-20 16:42:29 +02:00
|
|
|
.elevator_dispatch_fn = cfq_dispatch_requests,
|
2005-04-17 00:20:36 +02:00
|
|
|
.elevator_add_req_fn = cfq_insert_request,
|
2005-10-20 16:42:29 +02:00
|
|
|
.elevator_activate_req_fn = cfq_activate_request,
|
2005-04-17 00:20:36 +02:00
|
|
|
.elevator_deactivate_req_fn = cfq_deactivate_request,
|
|
|
|
.elevator_queue_empty_fn = cfq_queue_empty,
|
|
|
|
.elevator_completed_req_fn = cfq_completed_request,
|
2006-07-13 12:33:14 +02:00
|
|
|
.elevator_former_req_fn = elv_rb_former_request,
|
|
|
|
.elevator_latter_req_fn = elv_rb_latter_request,
|
2005-04-17 00:20:36 +02:00
|
|
|
.elevator_set_req_fn = cfq_set_request,
|
|
|
|
.elevator_put_req_fn = cfq_put_request,
|
|
|
|
.elevator_may_queue_fn = cfq_may_queue,
|
|
|
|
.elevator_init_fn = cfq_init_queue,
|
|
|
|
.elevator_exit_fn = cfq_exit_queue,
|
2006-08-29 09:05:44 +02:00
|
|
|
.trim = cfq_free_io_context,
|
2005-04-17 00:20:36 +02:00
|
|
|
},
|
2006-03-19 00:35:43 +01:00
|
|
|
.elevator_attrs = cfq_attrs,
|
2005-04-17 00:20:36 +02:00
|
|
|
.elevator_name = "cfq",
|
|
|
|
.elevator_owner = THIS_MODULE,
|
|
|
|
};
|
|
|
|
|
|
|
|
static int __init cfq_init(void)
|
|
|
|
{
|
|
|
|
int ret;
|
|
|
|
|
2005-06-27 10:55:12 +02:00
|
|
|
/*
|
|
|
|
* could be 0 on HZ < 1000 setups
|
|
|
|
*/
|
|
|
|
if (!cfq_slice_async)
|
|
|
|
cfq_slice_async = 1;
|
|
|
|
if (!cfq_slice_idle)
|
|
|
|
cfq_slice_idle = 1;
|
|
|
|
|
2005-04-17 00:20:36 +02:00
|
|
|
if (cfq_slab_setup())
|
|
|
|
return -ENOMEM;
|
|
|
|
|
|
|
|
ret = elv_register(&iosched_cfq);
|
2005-06-27 10:55:12 +02:00
|
|
|
if (ret)
|
|
|
|
cfq_slab_kill();
|
2005-04-17 00:20:36 +02:00
|
|
|
|
|
|
|
return ret;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void __exit cfq_exit(void)
|
|
|
|
{
|
2006-10-01 08:28:10 +02:00
|
|
|
DECLARE_COMPLETION_ONSTACK(all_gone);
|
2005-04-17 00:20:36 +02:00
|
|
|
elv_unregister(&iosched_cfq);
|
2006-03-18 21:05:53 +01:00
|
|
|
ioc_gone = &all_gone;
|
[PATCH 1/2] iosched: fix typo and barrier()
On rmmod path, cfq/as waits to make sure all io-contexts was
freed. However, it's using complete(), not wait_for_completion().
I think barrier() is not enough in here. To avoid the following case,
this patch replaces barrier() with smb_wmb().
cpu0 visibility cpu1
[ioc_gnone=NULL,ioc_count=1]
ioc_gnone = &all_gone NULL,ioc_count=1
atomic_read(&ioc_count) NULL,ioc_count=1
wait_for_completion() NULL,ioc_count=0 atomic_sub_and_test()
NULL,ioc_count=0 if ( && ioc_gone)
[ioc_gone==NULL,
so doesn't call complete()]
&all_gone,ioc_count=0
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Jens Axboe <axboe@suse.de>
2006-04-18 09:44:06 +02:00
|
|
|
/* ioc_gone's update must be visible before reading ioc_count */
|
|
|
|
smp_wmb();
|
2006-07-19 05:07:12 +02:00
|
|
|
if (elv_ioc_count_read(ioc_count))
|
[PATCH 1/2] iosched: fix typo and barrier()
On rmmod path, cfq/as waits to make sure all io-contexts was
freed. However, it's using complete(), not wait_for_completion().
I think barrier() is not enough in here. To avoid the following case,
this patch replaces barrier() with smb_wmb().
cpu0 visibility cpu1
[ioc_gnone=NULL,ioc_count=1]
ioc_gnone = &all_gone NULL,ioc_count=1
atomic_read(&ioc_count) NULL,ioc_count=1
wait_for_completion() NULL,ioc_count=0 atomic_sub_and_test()
NULL,ioc_count=0 if ( && ioc_gone)
[ioc_gone==NULL,
so doesn't call complete()]
&all_gone,ioc_count=0
Signed-off-by: OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>
Signed-off-by: Jens Axboe <axboe@suse.de>
2006-04-18 09:44:06 +02:00
|
|
|
wait_for_completion(ioc_gone);
|
2006-03-18 21:05:53 +01:00
|
|
|
synchronize_rcu();
|
2005-10-31 00:01:39 +01:00
|
|
|
cfq_slab_kill();
|
2005-04-17 00:20:36 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
module_init(cfq_init);
|
|
|
|
module_exit(cfq_exit);
|
|
|
|
|
|
|
|
MODULE_AUTHOR("Jens Axboe");
|
|
|
|
MODULE_LICENSE("GPL");
|
|
|
|
MODULE_DESCRIPTION("Completely Fair Queueing IO scheduler");
|