* Based on ideas from a previously unfinished io
* scheduler (round robin per-process disk scheduling) and Andrea Arcangeli.
*
- * Copyright (C) 2003 Jens Axboe <axboe@suse.de>
+ * Copyright (C) 2003 Jens Axboe <axboe@kernel.dk>
*/
#include <linux/module.h>
#include <linux/blkdev.h>
* tunables
*/
static const int cfq_quantum = 4; /* max queue in one round of service */
-static const int cfq_queued = 8; /* minimum rq allocate limit per-queue*/
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 */
#define CFQ_KEY_ASYNC (0)
-static DEFINE_SPINLOCK(cfq_exit_lock);
-
/*
* for the hash of cfqq inside the cfqd
*/
#define list_entry_cfqq(ptr) list_entry((ptr), struct cfq_queue, cfq_list)
-#define RQ_DATA(rq) (rq)->elevator_private
+#define RQ_CIC(rq) ((struct cfq_io_context*)(rq)->elevator_private)
+#define RQ_CFQQ(rq) ((rq)->elevator_private2)
-static kmem_cache_t *crq_pool;
-static kmem_cache_t *cfq_pool;
-static kmem_cache_t *cfq_ioc_pool;
+static struct kmem_cache *cfq_pool;
+static struct kmem_cache *cfq_ioc_pool;
-static atomic_t ioc_count = ATOMIC_INIT(0);
+static DEFINE_PER_CPU(unsigned long, ioc_count);
static struct completion *ioc_gone;
#define CFQ_PRIO_LISTS IOPRIO_BE_NR
#define cfq_class_idle(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
-#define cfq_class_be(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_BE)
#define cfq_class_rt(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
#define ASYNC (0)
unsigned int busy_queues;
/*
- * non-ordered list of empty cfqq's
- */
- struct list_head empty_list;
-
- /*
* cfqq lookup hash
*/
struct hlist_head *cfq_hash;
- mempool_t *crq_pool;
-
int rq_in_driver;
int hw_tag;
/*
- * schedule slice state info
- */
- /*
* idle window management
*/
struct timer_list idle_slice_timer;
sector_t last_sector;
unsigned long last_end_request;
- unsigned int rq_starved;
-
/*
* tunables, see top of file
*/
unsigned int cfq_quantum;
- unsigned int cfq_queued;
unsigned int cfq_fifo_expire[2];
unsigned int cfq_back_penalty;
unsigned int cfq_back_max;
struct hlist_node cfq_hash;
/* hash key */
unsigned int key;
- /* on either rr or empty list of cfqd */
+ /* member of the rr/busy/cur/idle cfqd list */
struct list_head cfq_list;
/* sorted list of pending requests */
struct rb_root sort_list;
/* if fifo isn't expired, next request to serve */
- struct cfq_rq *next_crq;
+ struct request *next_rq;
/* requests queued in sort_list */
int queued[2];
/* currently allocated requests */
int allocated[2];
+ /* pending metadata requests */
+ int meta_pending;
/* fifo list of requests in sort_list */
struct list_head fifo;
- unsigned long slice_start;
unsigned long slice_end;
- unsigned long slice_left;
unsigned long service_last;
+ long slice_resid;
/* number of requests that are on the dispatch list */
int on_dispatch[2];
unsigned int flags;
};
-struct cfq_rq {
- struct request *request;
-
- struct cfq_queue *cfq_queue;
- struct cfq_io_context *io_context;
-};
-
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,
+ CFQ_CFQQ_FLAG_on_rr = 0, /* on round-robin busy list */
+ CFQ_CFQQ_FLAG_wait_request, /* waiting for a request */
+ CFQ_CFQQ_FLAG_must_alloc, /* must be allowed rq alloc */
+ CFQ_CFQQ_FLAG_must_alloc_slice, /* per-slice must_alloc flag */
+ CFQ_CFQQ_FLAG_must_dispatch, /* must dispatch, even if expired */
+ CFQ_CFQQ_FLAG_fifo_expire, /* FIFO checked in this slice */
+ CFQ_CFQQ_FLAG_idle_window, /* slice idling enabled */
+ CFQ_CFQQ_FLAG_prio_changed, /* task priority has changed */
+ CFQ_CFQQ_FLAG_queue_new, /* queue never been serviced */
+ CFQ_CFQQ_FLAG_slice_new, /* no requests dispatched in slice */
};
#define CFQ_CFQQ_FNS(name) \
CFQ_CFQQ_FNS(fifo_expire);
CFQ_CFQQ_FNS(idle_window);
CFQ_CFQQ_FNS(prio_changed);
+CFQ_CFQQ_FNS(queue_new);
+CFQ_CFQQ_FNS(slice_new);
#undef CFQ_CFQQ_FNS
static struct cfq_queue *cfq_find_cfq_hash(struct cfq_data *, unsigned int, unsigned short);
-static void cfq_dispatch_insert(request_queue_t *, struct cfq_rq *);
+static void cfq_dispatch_insert(request_queue_t *, struct request *);
static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, unsigned int key, struct task_struct *tsk, gfp_t gfp_mask);
/*
return !cfqd->busy_queues;
}
-static inline pid_t cfq_queue_pid(struct task_struct *task, int rw)
+static inline pid_t cfq_queue_pid(struct task_struct *task, int rw, int is_sync)
{
- if (rw == READ || rw == WRITE_SYNC)
+ /*
+ * Use the per-process queue, for read requests and syncronous writes
+ */
+ if (!(rw & REQ_RW) || is_sync)
return task->pid;
return CFQ_KEY_ASYNC;
}
/*
- * Lifted from AS - choose which of crq1 and crq2 that is best served now.
+ * 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.
+ */
+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));
+}
+
+static inline void
+cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+{
+ cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies;
+ cfqq->slice_end += cfqq->slice_resid;
+
+ /*
+ * Don't carry over residual for more than one slice, we only want
+ * to slightly correct the fairness. Carrying over forever would
+ * easily introduce oscillations.
+ */
+ cfqq->slice_resid = 0;
+}
+
+/*
+ * We need to wrap this check in cfq_cfqq_slice_new(), since ->slice_end
+ * isn't valid until the first request from the dispatch is activated
+ * and the slice time set.
+ */
+static inline int cfq_slice_used(struct cfq_queue *cfqq)
+{
+ if (cfq_cfqq_slice_new(cfqq))
+ return 0;
+ if (time_before(jiffies, cfqq->slice_end))
+ return 0;
+
+ return 1;
+}
+
+/*
+ * Lifted from AS - choose which of rq1 and rq2 that is best served now.
* We choose the request that is closest to the head right now. Distance
* behind the head is penalized and only allowed to a certain extent.
*/
-static struct cfq_rq *
-cfq_choose_req(struct cfq_data *cfqd, struct cfq_rq *crq1, struct cfq_rq *crq2)
+static struct request *
+cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2)
{
sector_t last, s1, s2, d1 = 0, d2 = 0;
unsigned long back_max;
#define CFQ_RQ2_WRAP 0x02 /* request 2 wraps */
unsigned wrap = 0; /* bit mask: requests behind the disk head? */
- if (crq1 == NULL || crq1 == crq2)
- return crq2;
- if (crq2 == NULL)
- return crq1;
+ if (rq1 == NULL || rq1 == rq2)
+ return rq2;
+ if (rq2 == NULL)
+ return rq1;
- if (rq_is_sync(crq1->request) && !rq_is_sync(crq2->request))
- return crq1;
- else if (rq_is_sync(crq2->request) && !rq_is_sync(crq1->request))
- return crq2;
+ if (rq_is_sync(rq1) && !rq_is_sync(rq2))
+ return rq1;
+ else if (rq_is_sync(rq2) && !rq_is_sync(rq1))
+ return rq2;
+ if (rq_is_meta(rq1) && !rq_is_meta(rq2))
+ return rq1;
+ else if (rq_is_meta(rq2) && !rq_is_meta(rq1))
+ return rq2;
- s1 = crq1->request->sector;
- s2 = crq2->request->sector;
+ s1 = rq1->sector;
+ s2 = rq2->sector;
last = cfqd->last_sector;
* check two variables for all permutations: --> faster!
*/
switch (wrap) {
- case 0: /* common case for CFQ: crq1 and crq2 not wrapped */
+ case 0: /* common case for CFQ: rq1 and rq2 not wrapped */
if (d1 < d2)
- return crq1;
+ return rq1;
else if (d2 < d1)
- return crq2;
+ return rq2;
else {
if (s1 >= s2)
- return crq1;
+ return rq1;
else
- return crq2;
+ return rq2;
}
case CFQ_RQ2_WRAP:
- return crq1;
+ return rq1;
case CFQ_RQ1_WRAP:
- return crq2;
- case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both crqs wrapped */
+ return rq2;
+ case (CFQ_RQ1_WRAP|CFQ_RQ2_WRAP): /* both rqs wrapped */
default:
/*
* Since both rqs are wrapped,
* since back seek takes more time than forward.
*/
if (s1 <= s2)
- return crq1;
+ return rq1;
else
- return crq2;
+ return rq2;
}
}
/*
* would be nice to take fifo expire time into account as well
*/
-static struct cfq_rq *
-cfq_find_next_crq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
- struct cfq_rq *last_crq)
+static struct request *
+cfq_find_next_rq(struct cfq_data *cfqd, struct cfq_queue *cfqq,
+ struct request *last)
{
- struct request *last = last_crq->request;
struct rb_node *rbnext = rb_next(&last->rb_node);
struct rb_node *rbprev = rb_prev(&last->rb_node);
- struct cfq_rq *next = NULL, *prev = NULL;
+ struct request *next = NULL, *prev = NULL;
BUG_ON(RB_EMPTY_NODE(&last->rb_node));
if (rbprev)
- prev = RQ_DATA(rb_entry_rq(rbprev));
+ prev = rb_entry_rq(rbprev);
if (rbnext)
- next = RQ_DATA(rb_entry_rq(rbnext));
+ next = rb_entry_rq(rbnext);
else {
rbnext = rb_first(&cfqq->sort_list);
if (rbnext && rbnext != &last->rb_node)
- next = RQ_DATA(rb_entry_rq(rbnext));
+ next = rb_entry_rq(rbnext);
}
return cfq_choose_req(cfqd, next, prev);
static void cfq_resort_rr_list(struct cfq_queue *cfqq, int preempted)
{
struct cfq_data *cfqd = cfqq->cfqd;
- struct list_head *list, *entry;
+ struct list_head *list, *n;
+ struct cfq_queue *__cfqq;
- BUG_ON(!cfq_cfqq_on_rr(cfqq));
+ /*
+ * Resorting requires the cfqq to be on the RR list already.
+ */
+ if (!cfq_cfqq_on_rr(cfqq))
+ return;
list_del(&cfqq->cfq_list);
list = &cfqd->rr_list[cfqq->ioprio];
}
- /*
- * if queue was preempted, just add to front to be fair. busy_rr
- * isn't sorted, but insert at the back for fairness.
- */
- if (preempted || list == &cfqd->busy_rr) {
- if (preempted)
- list = list->prev;
+ if (preempted || cfq_cfqq_queue_new(cfqq)) {
+ /*
+ * If this queue was preempted or is new (never been serviced),
+ * let it be added first for fairness but beind other new
+ * queues.
+ */
+ n = list;
+ while (n->next != list) {
+ __cfqq = list_entry_cfqq(n->next);
+ if (!cfq_cfqq_queue_new(__cfqq))
+ break;
+ n = n->next;
+ }
+ list_add_tail(&cfqq->cfq_list, n);
+ } else if (!cfq_cfqq_class_sync(cfqq)) {
+ /*
+ * async queue always goes to the end. this wont be overly
+ * unfair to writes, as the sort of the sync queue wont be
+ * allowed to pass the async queue again.
+ */
list_add_tail(&cfqq->cfq_list, list);
- return;
- }
-
- /*
- * sort by when queue was last serviced
- */
- entry = list;
- while ((entry = entry->prev) != list) {
- struct cfq_queue *__cfqq = list_entry_cfqq(entry);
+ } else {
+ /*
+ * sort by last service, but don't cross a new or async
+ * queue. we don't cross a new queue because it hasn't been
+ * service before, and we don't cross an async queue because
+ * it gets added to the end on expire.
+ */
+ n = list;
+ while ((n = n->prev) != list) {
+ struct cfq_queue *__cfqq = list_entry_cfqq(n);
- if (!__cfqq->service_last)
- break;
- if (time_before(__cfqq->service_last, cfqq->service_last))
- break;
+ if (!cfq_cfqq_class_sync(cfqq) || !__cfqq->service_last)
+ break;
+ if (time_before(__cfqq->service_last, cfqq->service_last))
+ break;
+ }
+ list_add(&cfqq->cfq_list, n);
}
-
- list_add(&cfqq->cfq_list, entry);
}
/*
{
BUG_ON(!cfq_cfqq_on_rr(cfqq));
cfq_clear_cfqq_on_rr(cfqq);
- list_move(&cfqq->cfq_list, &cfqd->empty_list);
+ list_del_init(&cfqq->cfq_list);
BUG_ON(!cfqd->busy_queues);
cfqd->busy_queues--;
/*
* rb tree support functions
*/
-static inline void cfq_del_crq_rb(struct cfq_rq *crq)
+static inline void cfq_del_rq_rb(struct request *rq)
{
- struct cfq_queue *cfqq = crq->cfq_queue;
+ struct cfq_queue *cfqq = RQ_CFQQ(rq);
struct cfq_data *cfqd = cfqq->cfqd;
- const int sync = rq_is_sync(crq->request);
+ const int sync = rq_is_sync(rq);
BUG_ON(!cfqq->queued[sync]);
cfqq->queued[sync]--;
- elv_rb_del(&cfqq->sort_list, crq->request);
+ elv_rb_del(&cfqq->sort_list, rq);
if (cfq_cfqq_on_rr(cfqq) && RB_EMPTY_ROOT(&cfqq->sort_list))
cfq_del_cfqq_rr(cfqd, cfqq);
}
-static void cfq_add_crq_rb(struct cfq_rq *crq)
+static void cfq_add_rq_rb(struct request *rq)
{
- struct cfq_queue *cfqq = crq->cfq_queue;
+ struct cfq_queue *cfqq = RQ_CFQQ(rq);
struct cfq_data *cfqd = cfqq->cfqd;
- struct request *rq = crq->request;
struct request *__alias;
cfqq->queued[rq_is_sync(rq)]++;
* if that happens, put the alias on the dispatch list
*/
while ((__alias = elv_rb_add(&cfqq->sort_list, rq)) != NULL)
- cfq_dispatch_insert(cfqd->queue, RQ_DATA(__alias));
+ cfq_dispatch_insert(cfqd->queue, __alias);
+
+ if (!cfq_cfqq_on_rr(cfqq))
+ cfq_add_cfqq_rr(cfqd, cfqq);
}
static inline void
-cfq_reposition_crq_rb(struct cfq_queue *cfqq, struct cfq_rq *crq)
+cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
{
- struct request *rq = crq->request;
-
elv_rb_del(&cfqq->sort_list, rq);
cfqq->queued[rq_is_sync(rq)]--;
- cfq_add_crq_rb(crq);
+ cfq_add_rq_rb(rq);
}
static struct request *
cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
{
struct task_struct *tsk = current;
- pid_t key = cfq_queue_pid(tsk, bio_data_dir(bio));
- sector_t sector = bio->bi_sector + bio_sectors(bio);
+ pid_t key = cfq_queue_pid(tsk, bio_data_dir(bio), bio_sync(bio));
struct cfq_queue *cfqq;
cfqq = cfq_find_cfq_hash(cfqd, key, tsk->ioprio);
- if (cfqq)
+ if (cfqq) {
+ sector_t sector = bio->bi_sector + bio_sectors(bio);
+
return elv_rb_find(&cfqq->sort_list, sector);
+ }
return NULL;
}
static void cfq_remove_request(struct request *rq)
{
- struct cfq_rq *crq = RQ_DATA(rq);
- struct cfq_queue *cfqq = crq->cfq_queue;
+ struct cfq_queue *cfqq = RQ_CFQQ(rq);
- if (cfqq->next_crq == crq)
- cfqq->next_crq = cfq_find_next_crq(cfqq->cfqd, cfqq, crq);
+ if (cfqq->next_rq == rq)
+ cfqq->next_rq = cfq_find_next_rq(cfqq->cfqd, cfqq, rq);
list_del_init(&rq->queuelist);
- cfq_del_crq_rb(crq);
+ cfq_del_rq_rb(rq);
+
+ if (rq_is_meta(rq)) {
+ WARN_ON(!cfqq->meta_pending);
+ cfqq->meta_pending--;
+ }
}
static int
static void cfq_merged_request(request_queue_t *q, struct request *req,
int type)
{
- struct cfq_rq *crq = RQ_DATA(req);
-
if (type == ELEVATOR_FRONT_MERGE) {
- struct cfq_queue *cfqq = crq->cfq_queue;
+ struct cfq_queue *cfqq = RQ_CFQQ(req);
- cfq_reposition_crq_rb(cfqq, crq);
+ cfq_reposition_rq_rb(cfqq, req);
}
}
cfq_remove_request(next);
}
+static int cfq_allow_merge(request_queue_t *q, struct request *rq,
+ struct bio *bio)
+{
+ struct cfq_data *cfqd = q->elevator->elevator_data;
+ const int rw = bio_data_dir(bio);
+ struct cfq_queue *cfqq;
+ pid_t key;
+
+ /*
+ * Disallow merge of a sync bio into an async request.
+ */
+ if ((bio_data_dir(bio) == READ || bio_sync(bio)) && !rq_is_sync(rq))
+ return 0;
+
+ /*
+ * Lookup the cfqq that this bio will be queued with. Allow
+ * merge only if rq is queued there.
+ */
+ key = cfq_queue_pid(current, rw, bio_sync(bio));
+ cfqq = cfq_find_cfq_hash(cfqd, key, current->ioprio);
+
+ if (cfqq == RQ_CFQQ(rq))
+ return 1;
+
+ return 0;
+}
+
static inline void
__cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
*/
del_timer(&cfqd->idle_class_timer);
- cfqq->slice_start = jiffies;
cfqq->slice_end = 0;
- cfqq->slice_left = 0;
cfq_clear_cfqq_must_alloc_slice(cfqq);
cfq_clear_cfqq_fifo_expire(cfqq);
+ cfq_mark_cfqq_slice_new(cfqq);
}
cfqd->active_queue = cfqq;
*/
static void
__cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq,
- int preempted)
+ int preempted, int timed_out)
{
- unsigned long now = jiffies;
-
if (cfq_cfqq_wait_request(cfqq))
del_timer(&cfqd->idle_slice_timer);
- if (!preempted && !cfq_cfqq_dispatched(cfqq)) {
- cfqq->service_last = now;
- cfq_schedule_dispatch(cfqd);
- }
-
cfq_clear_cfqq_must_dispatch(cfqq);
cfq_clear_cfqq_wait_request(cfqq);
+ cfq_clear_cfqq_queue_new(cfqq);
/*
* 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 (timed_out && !cfq_cfqq_slice_new(cfqq))
+ cfqq->slice_resid = cfqq->slice_end - jiffies;
- if (cfq_cfqq_on_rr(cfqq))
- cfq_resort_rr_list(cfqq, preempted);
+ cfq_resort_rr_list(cfqq, preempted);
if (cfqq == cfqd->active_queue)
cfqd->active_queue = NULL;
cfqd->dispatch_slice = 0;
}
-static inline void cfq_slice_expired(struct cfq_data *cfqd, int preempted)
+static inline void cfq_slice_expired(struct cfq_data *cfqd, int preempted,
+ int timed_out)
{
struct cfq_queue *cfqq = cfqd->active_queue;
if (cfqq)
- __cfq_slice_expired(cfqd, cfqq, preempted);
+ __cfq_slice_expired(cfqd, cfqq, preempted, timed_out);
}
/*
{
struct cfq_queue *cfqq = NULL;
- /*
- * 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
- */
- if (!list_empty(&cfqd->cur_rr) || cfq_get_next_prio_level(cfqd) != -1)
+ 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
+ */
cfqq = list_entry_cfqq(cfqd->cur_rr.next);
-
- /*
- * If no new queues are available, check if the busy list has some
- * before falling back to idle io.
- */
- if (!cfqq && !list_empty(&cfqd->busy_rr))
+ } 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.
+ */
cfqq = list_entry_cfqq(cfqd->busy_rr.next);
-
- /*
- * 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
- */
- if (!cfqq && !list_empty(&cfqd->idle_rr)) {
+ } 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
+ */
unsigned long end = cfqd->last_end_request + CFQ_IDLE_GRACE;
if (time_after_eq(jiffies, end))
#define CIC_SEEKY(cic) ((cic)->seek_mean > (128 * 1024))
-static int cfq_arm_slice_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
-
+static int cfq_arm_slice_timer(struct cfq_data *cfqd)
{
+ struct cfq_queue *cfqq = cfqd->active_queue;
struct cfq_io_context *cic;
unsigned long sl;
WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list));
- WARN_ON(cfqq != cfqd->active_queue);
/*
* idle is disabled, either manually or by past process history
return 1;
}
-static void cfq_dispatch_insert(request_queue_t *q, struct cfq_rq *crq)
+static void cfq_dispatch_insert(request_queue_t *q, struct request *rq)
{
struct cfq_data *cfqd = q->elevator->elevator_data;
- struct cfq_queue *cfqq = crq->cfq_queue;
- struct request *rq = crq->request;
+ struct cfq_queue *cfqq = RQ_CFQQ(rq);
cfq_remove_request(rq);
cfqq->on_dispatch[rq_is_sync(rq)]++;
/*
* return expired entry, or NULL to just start from scratch in rbtree
*/
-static inline struct cfq_rq *cfq_check_fifo(struct cfq_queue *cfqq)
+static inline struct request *cfq_check_fifo(struct cfq_queue *cfqq)
{
struct cfq_data *cfqd = cfqq->cfqd;
struct request *rq;
- struct cfq_rq *crq;
+ int fifo;
if (cfq_cfqq_fifo_expire(cfqq))
return NULL;
- if (!list_empty(&cfqq->fifo)) {
- int fifo = cfq_cfqq_class_sync(cfqq);
-
- crq = RQ_DATA(rq_entry_fifo(cfqq->fifo.next));
- rq = crq->request;
- if (time_after(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo])) {
- cfq_mark_cfqq_fifo_expire(cfqq);
- return crq;
- }
- }
-
- return NULL;
-}
+ cfq_mark_cfqq_fifo_expire(cfqq);
-/*
- * 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.
- */
-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)];
+ if (list_empty(&cfqq->fifo))
+ return NULL;
- WARN_ON(cfqq->ioprio >= IOPRIO_BE_NR);
+ fifo = cfq_cfqq_class_sync(cfqq);
+ rq = rq_entry_fifo(cfqq->fifo.next);
- return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - cfqq->ioprio));
-}
+ if (time_after(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo]))
+ return rq;
-static inline void
-cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq)
-{
- cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies;
+ return NULL;
}
static inline int
*/
static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
{
- unsigned long now = jiffies;
struct cfq_queue *cfqq;
cfqq = cfqd->active_queue;
/*
* slice has expired
*/
- if (!cfq_cfqq_must_dispatch(cfqq) && time_after(now, cfqq->slice_end))
+ if (!cfq_cfqq_must_dispatch(cfqq) && cfq_slice_used(cfqq))
goto expire;
/*
*/
if (!RB_EMPTY_ROOT(&cfqq->sort_list))
goto keep_queue;
- else if (cfq_cfqq_dispatched(cfqq)) {
+ else if (cfq_cfqq_slice_new(cfqq) || cfq_cfqq_dispatched(cfqq)) {
cfqq = NULL;
goto keep_queue;
} else if (cfq_cfqq_class_sync(cfqq)) {
- if (cfq_arm_slice_timer(cfqd, cfqq))
+ if (cfq_arm_slice_timer(cfqd))
return NULL;
}
expire:
- cfq_slice_expired(cfqd, 0);
+ cfq_slice_expired(cfqd, 0, 0);
new_queue:
cfqq = cfq_set_active_queue(cfqd);
keep_queue:
BUG_ON(RB_EMPTY_ROOT(&cfqq->sort_list));
do {
- struct cfq_rq *crq;
+ struct request *rq;
/*
* follow expired path, else get first next available
*/
- if ((crq = cfq_check_fifo(cfqq)) == NULL)
- crq = cfqq->next_crq;
+ if ((rq = cfq_check_fifo(cfqq)) == NULL)
+ rq = cfqq->next_rq;
/*
* finally, insert request into driver dispatch list
*/
- cfq_dispatch_insert(cfqd->queue, crq);
+ cfq_dispatch_insert(cfqd->queue, rq);
cfqd->dispatch_slice++;
dispatched++;
if (!cfqd->active_cic) {
- atomic_inc(&crq->io_context->ioc->refcount);
- cfqd->active_cic = crq->io_context;
+ atomic_inc(&RQ_CIC(rq)->ioc->refcount);
+ cfqd->active_cic = RQ_CIC(rq);
}
if (RB_EMPTY_ROOT(&cfqq->sort_list))
} while (dispatched < max_dispatch);
/*
- * if slice end isn't set yet, set it.
- */
- 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)) ||
- cfq_class_idle(cfqq) ||
- !cfq_cfqq_idle_window(cfqq))
- cfq_slice_expired(cfqd, 0);
+ cfq_class_idle(cfqq)) {
+ cfqq->slice_end = jiffies + 1;
+ cfq_slice_expired(cfqd, 0, 0);
+ }
return dispatched;
}
cfq_forced_dispatch_cfqqs(struct list_head *list)
{
struct cfq_queue *cfqq, *next;
- struct cfq_rq *crq;
int dispatched;
dispatched = 0;
list_for_each_entry_safe(cfqq, next, list, cfq_list) {
- while ((crq = cfqq->next_crq)) {
- cfq_dispatch_insert(cfqq->cfqd->queue, crq);
+ while (cfqq->next_rq) {
+ cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq);
dispatched++;
}
BUG_ON(!list_empty(&cfqq->fifo));
dispatched += cfq_forced_dispatch_cfqqs(&cfqd->cur_rr);
dispatched += cfq_forced_dispatch_cfqqs(&cfqd->idle_rr);
- cfq_slice_expired(cfqd, 0);
+ cfq_slice_expired(cfqd, 0, 0);
BUG_ON(cfqd->busy_queues);
if (prev_cfqq == cfqq)
break;
+ /*
+ * So we have dispatched before in this round, if the
+ * next queue has idling enabled (must be sync), don't
+ * allow it service until the previous have continued.
+ */
+ if (cfqd->rq_in_driver && cfq_cfqq_idle_window(cfqq))
+ break;
+
cfq_clear_cfqq_must_dispatch(cfqq);
cfq_clear_cfqq_wait_request(cfqq);
del_timer(&cfqd->idle_slice_timer);
max_dispatch = 1;
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;
}
}
/*
- * task holds one reference to the queue, dropped when task exits. each crq
- * in-flight on this queue also holds a reference, dropped when crq is freed.
+ * 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.
*
* queue lock must be held here.
*/
BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]);
BUG_ON(cfq_cfqq_on_rr(cfqq));
- if (unlikely(cfqd->active_queue == cfqq))
- __cfq_slice_expired(cfqd, cfqq, 0);
+ if (unlikely(cfqd->active_queue == cfqq)) {
+ __cfq_slice_expired(cfqd, cfqq, 0, 0);
+ cfq_schedule_dispatch(cfqd);
+ }
/*
* it's on the empty list and still hashed
kmem_cache_free(cfq_pool, cfqq);
}
-static inline struct cfq_queue *
+static struct cfq_queue *
__cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned int prio,
const int hashval)
{
freed++;
}
- if (atomic_sub_and_test(freed, &ioc_count) && ioc_gone)
+ elv_ioc_count_mod(ioc_count, -freed);
+
+ if (ioc_gone && !elv_ioc_count_read(ioc_count))
complete(ioc_gone);
}
-static void cfq_trim(struct io_context *ioc)
+static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
- ioc->set_ioprio = NULL;
- cfq_free_io_context(ioc);
+ if (unlikely(cfqq == cfqd->active_queue)) {
+ __cfq_slice_expired(cfqd, cfqq, 0, 0);
+ cfq_schedule_dispatch(cfqd);
+ }
+
+ cfq_put_queue(cfqq);
}
-/*
- * Called with interrupts disabled
- */
-static void cfq_exit_single_io_context(struct cfq_io_context *cic)
+static void __cfq_exit_single_io_context(struct cfq_data *cfqd,
+ struct cfq_io_context *cic)
{
- struct cfq_data *cfqd = cic->key;
- request_queue_t *q;
-
- if (!cfqd)
- return;
-
- q = cfqd->queue;
-
- WARN_ON(!irqs_disabled());
-
- spin_lock(q->queue_lock);
+ list_del_init(&cic->queue_list);
+ smp_wmb();
+ cic->key = NULL;
if (cic->cfqq[ASYNC]) {
- if (unlikely(cic->cfqq[ASYNC] == cfqd->active_queue))
- __cfq_slice_expired(cfqd, cic->cfqq[ASYNC], 0);
- cfq_put_queue(cic->cfqq[ASYNC]);
+ cfq_exit_cfqq(cfqd, cic->cfqq[ASYNC]);
cic->cfqq[ASYNC] = NULL;
}
if (cic->cfqq[SYNC]) {
- if (unlikely(cic->cfqq[SYNC] == cfqd->active_queue))
- __cfq_slice_expired(cfqd, cic->cfqq[SYNC], 0);
- cfq_put_queue(cic->cfqq[SYNC]);
+ cfq_exit_cfqq(cfqd, cic->cfqq[SYNC]);
cic->cfqq[SYNC] = NULL;
}
+}
- cic->key = NULL;
- list_del_init(&cic->queue_list);
- spin_unlock(q->queue_lock);
+
+/*
+ * 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;
+
+ spin_lock_irq(q->queue_lock);
+ __cfq_exit_single_io_context(cfqd, cic);
+ spin_unlock_irq(q->queue_lock);
+ }
}
static void cfq_exit_io_context(struct io_context *ioc)
{
struct cfq_io_context *__cic;
- unsigned long flags;
struct rb_node *n;
/*
* put the reference this task is holding to the various queues
*/
- spin_lock_irqsave(&cfq_exit_lock, flags);
n = rb_first(&ioc->cic_root);
while (n != NULL) {
cfq_exit_single_io_context(__cic);
n = rb_next(n);
}
-
- spin_unlock_irqrestore(&cfq_exit_lock, flags);
}
static struct cfq_io_context *
cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
{
- struct cfq_io_context *cic = kmem_cache_alloc(cfq_ioc_pool, gfp_mask);
+ struct cfq_io_context *cic;
+ cic = kmem_cache_alloc_node(cfq_ioc_pool, gfp_mask, cfqd->queue->node);
if (cic) {
memset(cic, 0, sizeof(*cic));
cic->last_end_request = jiffies;
INIT_LIST_HEAD(&cic->queue_list);
cic->dtor = cfq_free_io_context;
cic->exit = cfq_exit_io_context;
- atomic_inc(&ioc_count);
+ elv_ioc_count_inc(ioc_count);
}
return cic;
cfqq->org_ioprio = cfqq->ioprio;
cfqq->org_ioprio_class = cfqq->ioprio_class;
- if (cfq_cfqq_on_rr(cfqq))
- cfq_resort_rr_list(cfqq, 0);
-
+ cfq_resort_rr_list(cfqq, 0);
cfq_clear_cfqq_prio_changed(cfqq);
}
{
struct cfq_data *cfqd = cic->key;
struct cfq_queue *cfqq;
+ unsigned long flags;
if (unlikely(!cfqd))
return;
- spin_lock(cfqd->queue->queue_lock);
+ spin_lock_irqsave(cfqd->queue->queue_lock, flags);
cfqq = cic->cfqq[ASYNC];
if (cfqq) {
if (cfqq)
cfq_mark_cfqq_prio_changed(cfqq);
- spin_unlock(cfqd->queue->queue_lock);
+ spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
}
-/*
- * callback from sys_ioprio_set, irqs are disabled
- */
-static int cfq_ioc_set_ioprio(struct io_context *ioc, unsigned int ioprio)
+static void cfq_ioc_set_ioprio(struct io_context *ioc)
{
struct cfq_io_context *cic;
struct rb_node *n;
- spin_lock(&cfq_exit_lock);
+ ioc->ioprio_changed = 0;
n = rb_first(&ioc->cic_root);
while (n != NULL) {
changed_ioprio(cic);
n = rb_next(n);
}
-
- spin_unlock(&cfq_exit_lock);
-
- return 0;
}
static struct cfq_queue *
cfqq = new_cfqq;
new_cfqq = NULL;
} else if (gfp_mask & __GFP_WAIT) {
+ /*
+ * 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.
+ */
spin_unlock_irq(cfqd->queue->queue_lock);
- new_cfqq = kmem_cache_alloc(cfq_pool, gfp_mask);
+ new_cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask|__GFP_NOFAIL, cfqd->queue->node);
spin_lock_irq(cfqd->queue->queue_lock);
goto retry;
} else {
- cfqq = kmem_cache_alloc(cfq_pool, gfp_mask);
+ cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask, cfqd->queue->node);
if (!cfqq)
goto out;
}
hlist_add_head(&cfqq->cfq_hash, &cfqd->cfq_hash[hashval]);
atomic_set(&cfqq->ref, 0);
cfqq->cfqd = cfqd;
- cfqq->service_last = 0;
- /*
- * set ->slice_left to allow preemption for a new process
- */
- cfqq->slice_left = 2 * cfqd->cfq_slice_idle;
+
cfq_mark_cfqq_idle_window(cfqq);
cfq_mark_cfqq_prio_changed(cfqq);
+ cfq_mark_cfqq_queue_new(cfqq);
cfq_init_prio_data(cfqq);
}
static void
cfq_drop_dead_cic(struct io_context *ioc, struct cfq_io_context *cic)
{
- spin_lock(&cfq_exit_lock);
+ WARN_ON(!list_empty(&cic->queue_list));
rb_erase(&cic->rb_node, &ioc->cic_root);
- list_del_init(&cic->queue_list);
- spin_unlock(&cfq_exit_lock);
kmem_cache_free(cfq_ioc_pool, cic);
- atomic_dec(&ioc_count);
+ elv_ioc_count_dec(ioc_count);
}
static struct cfq_io_context *
struct rb_node **p;
struct rb_node *parent;
struct cfq_io_context *__cic;
+ unsigned long flags;
void *k;
cic->ioc = ioc;
cic->key = cfqd;
- ioc->set_ioprio = cfq_ioc_set_ioprio;
restart:
parent = NULL;
p = &ioc->cic_root.rb_node;
BUG();
}
- spin_lock(&cfq_exit_lock);
rb_link_node(&cic->rb_node, parent, p);
rb_insert_color(&cic->rb_node, &ioc->cic_root);
+
+ spin_lock_irqsave(cfqd->queue->queue_lock, flags);
list_add(&cic->queue_list, &cfqd->cic_list);
- spin_unlock(&cfq_exit_lock);
+ spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
}
/*
might_sleep_if(gfp_mask & __GFP_WAIT);
- ioc = get_io_context(gfp_mask);
+ ioc = get_io_context(gfp_mask, cfqd->queue->node);
if (!ioc)
return NULL;
cfq_cic_link(cfqd, ioc, cic);
out:
+ smp_read_barrier_depends();
+ if (unlikely(ioc->ioprio_changed))
+ cfq_ioc_set_ioprio(ioc);
+
return cic;
err:
put_io_context(ioc);
static void
cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_io_context *cic)
{
- unsigned long elapsed, ttime;
-
- /*
- * 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
-
- ttime = min(elapsed, 2UL * cfqd->cfq_slice_idle);
+ unsigned long elapsed = jiffies - cic->last_end_request;
+ unsigned long ttime = min(elapsed, 2UL * cfqd->cfq_slice_idle);
cic->ttime_samples = (7*cic->ttime_samples + 256) / 8;
cic->ttime_total = (7*cic->ttime_total + 256*ttime) / 8;
}
static void
-cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_io_context *cic,
- struct cfq_rq *crq)
+cfq_update_io_seektime(struct cfq_io_context *cic, struct request *rq)
{
sector_t sdist;
u64 total;
- if (cic->last_request_pos < crq->request->sector)
- sdist = crq->request->sector - cic->last_request_pos;
+ if (cic->last_request_pos < rq->sector)
+ sdist = rq->sector - cic->last_request_pos;
else
- sdist = cic->last_request_pos - crq->request->sector;
+ sdist = cic->last_request_pos - rq->sector;
/*
* Don't allow the seek distance to get too large from the
cfq_clear_cfqq_idle_window(cfqq);
}
-
/*
* 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,
- struct cfq_rq *crq)
+ struct request *rq)
{
struct cfq_queue *cfqq = cfqd->active_queue;
if (!cfq_cfqq_wait_request(new_cfqq))
return 0;
/*
- * if it doesn't have slice left, forget it
+ * if the new request is sync, but the currently running queue is
+ * not, let the sync request have priority.
*/
- if (new_cfqq->slice_left < cfqd->cfq_slice_idle)
- return 0;
- if (rq_is_sync(crq->request) && !cfq_cfqq_sync(cfqq))
+ if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq))
+ return 1;
+ /*
+ * 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;
return 0;
*/
static void cfq_preempt_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
{
- struct cfq_queue *__cfqq, *next;
-
- list_for_each_entry_safe(__cfqq, next, &cfqd->cur_rr, cfq_list)
- cfq_resort_rr_list(__cfqq, 1);
-
- if (!cfqq->slice_left)
- cfqq->slice_left = cfq_prio_to_slice(cfqd, cfqq) / 2;
+ cfq_slice_expired(cfqd, 1, 1);
- cfqq->slice_end = cfqq->slice_left + jiffies;
- cfq_slice_expired(cfqd, 1);
- __cfq_set_active_queue(cfqd, cfqq);
-}
-
-/*
- * should really be a ll_rw_blk.c helper
- */
-static void cfq_start_queueing(struct cfq_data *cfqd, struct cfq_queue *cfqq)
-{
- request_queue_t *q = cfqd->queue;
+ /*
+ * 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);
- if (!blk_queue_plugged(q))
- q->request_fn(q);
- else
- __generic_unplug_device(q);
+ cfqq->slice_end = 0;
+ cfq_mark_cfqq_slice_new(cfqq);
}
/*
- * Called when a new fs request (crq) is added (to cfqq). Check if there's
+ * Called when a new fs request (rq) is added (to cfqq). Check if there's
* something we should do about it
*/
static void
-cfq_crq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
- struct cfq_rq *crq)
+cfq_rq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq,
+ struct request *rq)
{
- struct cfq_io_context *cic = crq->io_context;
+ struct cfq_io_context *cic = RQ_CIC(rq);
+
+ if (rq_is_meta(rq))
+ cfqq->meta_pending++;
/*
* check if this request is a better next-serve candidate)) {
*/
- cfqq->next_crq = cfq_choose_req(cfqd, cfqq->next_crq, crq);
- BUG_ON(!cfqq->next_crq);
+ cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq);
+ BUG_ON(!cfqq->next_rq);
/*
* we never wait for an async request and we don't allow preemption
* of an async request. so just return early
*/
- if (!rq_is_sync(crq->request)) {
+ if (!rq_is_sync(rq)) {
/*
* 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);
- cfq_start_queueing(cfqd, cfqq);
+ cfq_slice_expired(cfqd, 0, 0);
+ blk_start_queueing(cfqd->queue);
}
return;
}
cfq_update_io_thinktime(cfqd, cic);
- cfq_update_io_seektime(cfqd, cic, crq);
+ cfq_update_io_seektime(cic, rq);
cfq_update_idle_window(cfqd, cfqq, cic);
- cic->last_queue = jiffies;
- cic->last_request_pos = crq->request->sector + crq->request->nr_sectors;
+ cic->last_request_pos = rq->sector + rq->nr_sectors;
if (cfqq == cfqd->active_queue) {
/*
if (cfq_cfqq_wait_request(cfqq)) {
cfq_mark_cfqq_must_dispatch(cfqq);
del_timer(&cfqd->idle_slice_timer);
- cfq_start_queueing(cfqd, cfqq);
+ blk_start_queueing(cfqd->queue);
}
- } else if (cfq_should_preempt(cfqd, cfqq, crq)) {
+ } else if (cfq_should_preempt(cfqd, cfqq, rq)) {
/*
* not the active queue - expire current slice if it is
* idle and has expired it's mean thinktime or this new queue
*/
cfq_preempt_queue(cfqd, cfqq);
cfq_mark_cfqq_must_dispatch(cfqq);
- cfq_start_queueing(cfqd, cfqq);
+ blk_start_queueing(cfqd->queue);
}
}
static void cfq_insert_request(request_queue_t *q, struct request *rq)
{
struct cfq_data *cfqd = q->elevator->elevator_data;
- struct cfq_rq *crq = RQ_DATA(rq);
- struct cfq_queue *cfqq = crq->cfq_queue;
+ struct cfq_queue *cfqq = RQ_CFQQ(rq);
cfq_init_prio_data(cfqq);
- cfq_add_crq_rb(crq);
-
- if (!cfq_cfqq_on_rr(cfqq))
- cfq_add_cfqq_rr(cfqd, cfqq);
+ cfq_add_rq_rb(rq);
list_add_tail(&rq->queuelist, &cfqq->fifo);
- cfq_crq_enqueued(cfqd, cfqq, crq);
+ cfq_rq_enqueued(cfqd, cfqq, rq);
}
static void cfq_completed_request(request_queue_t *q, struct request *rq)
{
- struct cfq_rq *crq = RQ_DATA(rq);
- struct cfq_queue *cfqq = crq->cfq_queue;
+ struct cfq_queue *cfqq = RQ_CFQQ(rq);
struct cfq_data *cfqd = cfqq->cfqd;
const int sync = rq_is_sync(rq);
unsigned long now;
WARN_ON(!cfqq->on_dispatch[sync]);
cfqd->rq_in_driver--;
cfqq->on_dispatch[sync]--;
+ cfqq->service_last = now;
if (!cfq_class_idle(cfqq))
cfqd->last_end_request = now;
- if (!cfq_cfqq_dispatched(cfqq)) {
- if (cfq_cfqq_on_rr(cfqq)) {
- cfqq->service_last = now;
- cfq_resort_rr_list(cfqq, 0);
- }
- }
+ cfq_resort_rr_list(cfqq, 0);
if (sync)
- crq->io_context->last_end_request = now;
+ RQ_CIC(rq)->last_end_request = now;
/*
* 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);
+ if (cfq_cfqq_slice_new(cfqq)) {
+ cfq_set_prio_slice(cfqd, cfqq);
+ cfq_clear_cfqq_slice_new(cfqq);
+ }
+ if (cfq_slice_used(cfqq))
+ cfq_slice_expired(cfqd, 0, 1);
else if (sync && RB_EMPTY_ROOT(&cfqq->sort_list)) {
- if (!cfq_arm_slice_timer(cfqd, cfqq))
+ if (!cfq_arm_slice_timer(cfqd))
cfq_schedule_dispatch(cfqd);
}
}
/*
* refile between round-robin lists if we moved the priority class
*/
- if ((ioprio_class != cfqq->ioprio_class || ioprio != cfqq->ioprio) &&
- cfq_cfqq_on_rr(cfqq))
+ if ((ioprio_class != cfqq->ioprio_class || ioprio != cfqq->ioprio))
cfq_resort_rr_list(cfqq, 0);
}
-static inline int
-__cfq_may_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq,
- struct task_struct *task, int rw)
+static inline int __cfq_may_queue(struct cfq_queue *cfqq)
{
if ((cfq_cfqq_wait_request(cfqq) || cfq_cfqq_must_alloc(cfqq)) &&
!cfq_cfqq_must_alloc_slice(cfqq)) {
return ELV_MQUEUE_MAY;
}
-static int cfq_may_queue(request_queue_t *q, int rw, struct bio *bio)
+static int cfq_may_queue(request_queue_t *q, int rw)
{
struct cfq_data *cfqd = q->elevator->elevator_data;
struct task_struct *tsk = current;
struct cfq_queue *cfqq;
+ unsigned int key;
+
+ key = cfq_queue_pid(tsk, rw, rw & REQ_RW_SYNC);
/*
* don't force setup of a queue from here, as a call to may_queue
* so just lookup a possibly existing queue, or return 'may queue'
* if that fails
*/
- cfqq = cfq_find_cfq_hash(cfqd, cfq_queue_pid(tsk, rw), tsk->ioprio);
+ cfqq = cfq_find_cfq_hash(cfqd, key, tsk->ioprio);
if (cfqq) {
cfq_init_prio_data(cfqq);
cfq_prio_boost(cfqq);
- return __cfq_may_queue(cfqd, cfqq, tsk, rw);
+ return __cfq_may_queue(cfqq);
}
return ELV_MQUEUE_MAY;
}
-static void cfq_check_waiters(request_queue_t *q, struct cfq_queue *cfqq)
-{
- struct cfq_data *cfqd = q->elevator->elevator_data;
-
- if (unlikely(cfqd->rq_starved)) {
- struct request_list *rl = &q->rq;
-
- smp_mb();
- if (waitqueue_active(&rl->wait[READ]))
- wake_up(&rl->wait[READ]);
- if (waitqueue_active(&rl->wait[WRITE]))
- wake_up(&rl->wait[WRITE]);
- }
-}
-
/*
* queue lock held here
*/
-static void cfq_put_request(request_queue_t *q, struct request *rq)
+static void cfq_put_request(struct request *rq)
{
- struct cfq_data *cfqd = q->elevator->elevator_data;
- struct cfq_rq *crq = RQ_DATA(rq);
+ struct cfq_queue *cfqq = RQ_CFQQ(rq);
- if (crq) {
- struct cfq_queue *cfqq = crq->cfq_queue;
+ if (cfqq) {
const int rw = rq_data_dir(rq);
BUG_ON(!cfqq->allocated[rw]);
cfqq->allocated[rw]--;
- put_io_context(crq->io_context->ioc);
+ put_io_context(RQ_CIC(rq)->ioc);
- mempool_free(crq, cfqd->crq_pool);
rq->elevator_private = NULL;
+ rq->elevator_private2 = NULL;
- cfq_check_waiters(q, cfqq);
cfq_put_queue(cfqq);
}
}
* Allocate cfq data structures associated with this request.
*/
static int
-cfq_set_request(request_queue_t *q, struct request *rq, struct bio *bio,
- gfp_t gfp_mask)
+cfq_set_request(request_queue_t *q, struct request *rq, gfp_t gfp_mask)
{
struct cfq_data *cfqd = q->elevator->elevator_data;
struct task_struct *tsk = current;
struct cfq_io_context *cic;
const int rw = rq_data_dir(rq);
- pid_t key = cfq_queue_pid(tsk, rw);
+ const int is_sync = rq_is_sync(rq);
+ pid_t key = cfq_queue_pid(tsk, rw, is_sync);
struct cfq_queue *cfqq;
- struct cfq_rq *crq;
unsigned long flags;
- int is_sync = key != CFQ_KEY_ASYNC;
might_sleep_if(gfp_mask & __GFP_WAIT);
cfqq->allocated[rw]++;
cfq_clear_cfqq_must_alloc(cfqq);
- cfqd->rq_starved = 0;
atomic_inc(&cfqq->ref);
- spin_unlock_irqrestore(q->queue_lock, flags);
- crq = mempool_alloc(cfqd->crq_pool, gfp_mask);
- if (crq) {
- crq->request = rq;
- crq->cfq_queue = cfqq;
- crq->io_context = cic;
+ spin_unlock_irqrestore(q->queue_lock, flags);
- rq->elevator_private = crq;
- return 0;
- }
+ rq->elevator_private = cic;
+ rq->elevator_private2 = cfqq;
+ return 0;
- spin_lock_irqsave(q->queue_lock, flags);
- cfqq->allocated[rw]--;
- if (!(cfqq->allocated[0] + cfqq->allocated[1]))
- cfq_mark_cfqq_must_alloc(cfqq);
- cfq_put_queue(cfqq);
queue_fail:
if (cic)
put_io_context(cic->ioc);
- /*
- * mark us rq allocation starved. we need to kickstart the process
- * ourselves if there are no pending requests that can do it for us.
- * that would be an extremely rare OOM situation
- */
- cfqd->rq_starved = 1;
+
cfq_schedule_dispatch(cfqd);
spin_unlock_irqrestore(q->queue_lock, flags);
return 1;
}
-static void cfq_kick_queue(void *data)
+static void cfq_kick_queue(struct work_struct *work)
{
- request_queue_t *q = data;
- struct cfq_data *cfqd = q->elevator->elevator_data;
+ struct cfq_data *cfqd =
+ container_of(work, struct cfq_data, unplug_work);
+ request_queue_t *q = cfqd->queue;
unsigned long flags;
spin_lock_irqsave(q->queue_lock, flags);
-
- if (cfqd->rq_starved) {
- struct request_list *rl = &q->rq;
-
- /*
- * we aren't guaranteed to get a request after this, but we
- * have to be opportunistic
- */
- smp_mb();
- if (waitqueue_active(&rl->wait[READ]))
- wake_up(&rl->wait[READ]);
- if (waitqueue_active(&rl->wait[WRITE]))
- wake_up(&rl->wait[WRITE]);
- }
-
- blk_remove_plug(q);
- q->request_fn(q);
+ blk_start_queueing(q);
spin_unlock_irqrestore(q->queue_lock, flags);
}
struct cfq_data *cfqd = (struct cfq_data *) data;
struct cfq_queue *cfqq;
unsigned long flags;
+ int timed_out = 1;
spin_lock_irqsave(cfqd->queue->queue_lock, flags);
if ((cfqq = cfqd->active_queue) != NULL) {
- unsigned long now = jiffies;
+ timed_out = 0;
/*
* expired
*/
- if (time_after(now, cfqq->slice_end))
+ if (cfq_slice_used(cfqq))
goto expire;
/*
}
}
expire:
- cfq_slice_expired(cfqd, 0);
+ cfq_slice_expired(cfqd, 0, timed_out);
out_kick:
cfq_schedule_dispatch(cfqd);
out_cont:
cfq_shutdown_timer_wq(cfqd);
- spin_lock(&cfq_exit_lock);
spin_lock_irq(q->queue_lock);
if (cfqd->active_queue)
- __cfq_slice_expired(cfqd, cfqd->active_queue, 0);
+ __cfq_slice_expired(cfqd, cfqd->active_queue, 0, 0);
while (!list_empty(&cfqd->cic_list)) {
struct cfq_io_context *cic = list_entry(cfqd->cic_list.next,
struct cfq_io_context,
queue_list);
- if (cic->cfqq[ASYNC]) {
- cfq_put_queue(cic->cfqq[ASYNC]);
- cic->cfqq[ASYNC] = NULL;
- }
- if (cic->cfqq[SYNC]) {
- cfq_put_queue(cic->cfqq[SYNC]);
- cic->cfqq[SYNC] = NULL;
- }
- cic->key = NULL;
- list_del_init(&cic->queue_list);
+
+ __cfq_exit_single_io_context(cfqd, cic);
}
spin_unlock_irq(q->queue_lock);
- spin_unlock(&cfq_exit_lock);
cfq_shutdown_timer_wq(cfqd);
- mempool_destroy(cfqd->crq_pool);
kfree(cfqd->cfq_hash);
kfree(cfqd);
}
-static void *cfq_init_queue(request_queue_t *q, elevator_t *e)
+static void *cfq_init_queue(request_queue_t *q)
{
struct cfq_data *cfqd;
int i;
- cfqd = kmalloc(sizeof(*cfqd), GFP_KERNEL);
+ cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
if (!cfqd)
return NULL;
INIT_LIST_HEAD(&cfqd->busy_rr);
INIT_LIST_HEAD(&cfqd->cur_rr);
INIT_LIST_HEAD(&cfqd->idle_rr);
- INIT_LIST_HEAD(&cfqd->empty_list);
INIT_LIST_HEAD(&cfqd->cic_list);
- cfqd->cfq_hash = kmalloc(sizeof(struct hlist_head) * CFQ_QHASH_ENTRIES, GFP_KERNEL);
+ cfqd->cfq_hash = kmalloc_node(sizeof(struct hlist_head) * CFQ_QHASH_ENTRIES, GFP_KERNEL, q->node);
if (!cfqd->cfq_hash)
- goto out_crqhash;
-
- cfqd->crq_pool = mempool_create_slab_pool(BLKDEV_MIN_RQ, crq_pool);
- if (!cfqd->crq_pool)
- goto out_crqpool;
+ goto out_free;
for (i = 0; i < CFQ_QHASH_ENTRIES; i++)
INIT_HLIST_HEAD(&cfqd->cfq_hash[i]);
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);
+ INIT_WORK(&cfqd->unplug_work, cfq_kick_queue);
- cfqd->cfq_queued = cfq_queued;
cfqd->cfq_quantum = cfq_quantum;
cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0];
cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1];
cfqd->cfq_slice_idle = cfq_slice_idle;
return cfqd;
-out_crqpool:
- kfree(cfqd->cfq_hash);
-out_crqhash:
+out_free:
kfree(cfqd);
return NULL;
}
static void cfq_slab_kill(void)
{
- if (crq_pool)
- kmem_cache_destroy(crq_pool);
if (cfq_pool)
kmem_cache_destroy(cfq_pool);
if (cfq_ioc_pool)
static int __init cfq_slab_setup(void)
{
- crq_pool = kmem_cache_create("crq_pool", sizeof(struct cfq_rq), 0, 0,
- NULL, NULL);
- if (!crq_pool)
- goto fail;
-
cfq_pool = kmem_cache_create("cfq_pool", sizeof(struct cfq_queue), 0, 0,
NULL, NULL);
if (!cfq_pool)
return cfq_var_show(__data, (page)); \
}
SHOW_FUNCTION(cfq_quantum_show, cfqd->cfq_quantum, 0);
-SHOW_FUNCTION(cfq_queued_show, cfqd->cfq_queued, 0);
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);
SHOW_FUNCTION(cfq_back_seek_max_show, cfqd->cfq_back_max, 0);
return ret; \
}
STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
-STORE_FUNCTION(cfq_queued_store, &cfqd->cfq_queued, 1, UINT_MAX, 0);
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);
STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
static struct elv_fs_entry cfq_attrs[] = {
CFQ_ATTR(quantum),
- CFQ_ATTR(queued),
CFQ_ATTR(fifo_expire_sync),
CFQ_ATTR(fifo_expire_async),
CFQ_ATTR(back_seek_max),
.elevator_merge_fn = cfq_merge,
.elevator_merged_fn = cfq_merged_request,
.elevator_merge_req_fn = cfq_merged_requests,
+ .elevator_allow_merge_fn = cfq_allow_merge,
.elevator_dispatch_fn = cfq_dispatch_requests,
.elevator_add_req_fn = cfq_insert_request,
.elevator_activate_req_fn = cfq_activate_request,
.elevator_may_queue_fn = cfq_may_queue,
.elevator_init_fn = cfq_init_queue,
.elevator_exit_fn = cfq_exit_queue,
- .trim = cfq_trim,
+ .trim = cfq_free_io_context,
},
.elevator_attrs = cfq_attrs,
.elevator_name = "cfq",
static void __exit cfq_exit(void)
{
- DECLARE_COMPLETION(all_gone);
+ DECLARE_COMPLETION_ONSTACK(all_gone);
elv_unregister(&iosched_cfq);
ioc_gone = &all_gone;
/* ioc_gone's update must be visible before reading ioc_count */
smp_wmb();
- if (atomic_read(&ioc_count))
+ if (elv_ioc_count_read(ioc_count))
wait_for_completion(ioc_gone);
synchronize_rcu();
cfq_slab_kill();