X-Git-Url: http://ftp.safe.ca/?a=blobdiff_plain;f=block%2Fcfq-iosched.c;h=6a062eebbd15301320e7491b5dd45d17f2204a3c;hb=0c032ab889e7b20b8a5a7d09313e4aca214a15f7;hp=b2fd8cac2147ef6c9df249853c50025a32259f99;hpb=21183b07ee4be405362af8454f3647781c77df1b;p=safe%2Fjmp%2Flinux-2.6 diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c index b2fd8ca..6a062ee 100644 --- a/block/cfq-iosched.c +++ b/block/cfq-iosched.c @@ -4,107 +4,99 @@ * Based on ideas from a previously unfinished io * scheduler (round robin per-process disk scheduling) and Andrea Arcangeli. * - * Copyright (C) 2003 Jens Axboe + * Copyright (C) 2003 Jens Axboe */ #include #include #include -#include #include #include +#include /* * 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*/ +/* max queue in one round of service */ +static const int cfq_quantum = 4; 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 */ - +/* maximum backwards seek, in KiB */ +static const int cfq_back_max = 16 * 1024; +/* penalty of a backwards seek */ +static const int cfq_back_penalty = 2; static const int cfq_slice_sync = HZ / 10; static int cfq_slice_async = HZ / 25; static const int cfq_slice_async_rq = 2; static int cfq_slice_idle = HZ / 125; -#define CFQ_IDLE_GRACE (HZ / 10) -#define CFQ_SLICE_SCALE (5) - -#define CFQ_KEY_ASYNC (0) - -static DEFINE_SPINLOCK(cfq_exit_lock); +/* + * offset from end of service tree + */ +#define CFQ_IDLE_DELAY (HZ / 5) /* - * for the hash of cfqq inside the cfqd + * below this threshold, we consider thinktime immediate */ -#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 CFQ_MIN_TT (2) -#define list_entry_cfqq(ptr) list_entry((ptr), struct cfq_queue, cfq_list) -#define list_entry_fifo(ptr) list_entry((ptr), struct request, queuelist) +#define CFQ_SLICE_SCALE (5) +#define CFQ_HW_QUEUE_MIN (5) -#define RQ_DATA(rq) (rq)->elevator_private +#define RQ_CIC(rq) \ + ((struct cfq_io_context *) (rq)->elevator_private) +#define RQ_CFQQ(rq) (struct cfq_queue *) ((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; +static DEFINE_SPINLOCK(ioc_gone_lock); #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) #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]) - #define sample_valid(samples) ((samples) > 80) /* + * Most of our rbtree usage is for sorting with min extraction, so + * if we cache the leftmost node we don't have to walk down the tree + * to find it. Idea borrowed from Ingo Molnars CFS scheduler. We should + * move this into the elevator for the rq sorting as well. + */ +struct cfq_rb_root { + struct rb_root rb; + struct rb_node *left; +}; +#define CFQ_RB_ROOT (struct cfq_rb_root) { RB_ROOT, NULL, } + +/* * Per block device queue structure */ struct cfq_data { - request_queue_t *queue; + struct request_queue *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; + struct cfq_rb_root service_tree; unsigned int busy_queues; - /* - * non-ordered list of empty cfqq's - */ - struct list_head empty_list; + int rq_in_driver; + int sync_flight; /* - * cfqq lookup hash + * queue-depth detection */ - struct hlist_head *cfq_hash; - - mempool_t *crq_pool; - - int rq_in_driver; + int rq_queued; int hw_tag; + int hw_tag_samples; + int rq_in_driver_peak; /* - * schedule slice state info - */ - /* * idle window management */ struct timer_list idle_slice_timer; @@ -112,21 +104,20 @@ struct cfq_data { 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; + /* + * async queue for each priority case + */ + struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR]; + struct cfq_queue *async_idle_cfqq; - sector_t last_sector; + sector_t last_position; 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; @@ -143,18 +134,18 @@ struct cfq_data { struct cfq_queue { /* reference count */ atomic_t ref; + /* various state flags, see below */ + unsigned int flags; /* parent cfq_data */ struct cfq_data *cfqd; - /* cfqq lookup hash */ - struct hlist_node cfq_hash; - /* hash key */ - unsigned int key; - /* on either rr or empty list of cfqd */ - struct list_head cfq_list; + /* service_tree member */ + struct rb_node rb_node; + /* service_tree key */ + unsigned long rb_key; /* 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 */ @@ -162,54 +153,47 @@ struct cfq_queue { /* 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]; + /* pending metadata requests */ + int meta_pending; + /* number of requests that are on the dispatch list or inside driver */ + int dispatched; /* io prio of this group */ unsigned short ioprio, org_ioprio; unsigned short ioprio_class, org_ioprio_class; - /* various state flags, see below */ - unsigned int flags; -}; - -struct cfq_rq { - struct request *request; - - struct cfq_queue *cfq_queue; - struct cfq_io_context *io_context; - - unsigned int crq_flags; + pid_t pid; }; 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 */ + CFQ_CFQQ_FLAG_sync, /* synchronous queue */ }; #define CFQ_CFQQ_FNS(name) \ static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq) \ { \ - cfqq->flags |= (1 << CFQ_CFQQ_FLAG_##name); \ + (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); \ + (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; \ + return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0; \ } CFQ_CFQQ_FNS(on_rr); @@ -220,32 +204,45 @@ CFQ_CFQQ_FNS(must_dispatch); 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); +CFQ_CFQQ_FNS(sync); #undef CFQ_CFQQ_FNS -enum cfq_rq_state_flags { - CFQ_CRQ_FLAG_is_sync = 0, -}; +#define cfq_log_cfqq(cfqd, cfqq, fmt, args...) \ + blk_add_trace_msg((cfqd)->queue, "cfq%d " fmt, (cfqq)->pid, ##args) +#define cfq_log(cfqd, fmt, args...) \ + blk_add_trace_msg((cfqd)->queue, "cfq " fmt, ##args) -#define CFQ_CRQ_FNS(name) \ -static inline void cfq_mark_crq_##name(struct cfq_rq *crq) \ -{ \ - crq->crq_flags |= (1 << CFQ_CRQ_FLAG_##name); \ -} \ -static inline void cfq_clear_crq_##name(struct cfq_rq *crq) \ -{ \ - crq->crq_flags &= ~(1 << CFQ_CRQ_FLAG_##name); \ -} \ -static inline int cfq_crq_##name(const struct cfq_rq *crq) \ -{ \ - return (crq->crq_flags & (1 << CFQ_CRQ_FLAG_##name)) != 0; \ +static void cfq_dispatch_insert(struct request_queue *, struct request *); +static struct cfq_queue *cfq_get_queue(struct cfq_data *, int, + struct io_context *, gfp_t); +static struct cfq_io_context *cfq_cic_lookup(struct cfq_data *, + struct io_context *); + +static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_context *cic, + int is_sync) +{ + return cic->cfqq[!!is_sync]; } -CFQ_CRQ_FNS(is_sync); -#undef CFQ_CRQ_FNS +static inline void cic_set_cfqq(struct cfq_io_context *cic, + struct cfq_queue *cfqq, int is_sync) +{ + cic->cfqq[!!is_sync] = cfqq; +} -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 struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, unsigned int key, struct task_struct *tsk, gfp_t gfp_mask); +/* + * We regard a request as SYNC, if it's either a read or has the SYNC bit + * set (in which case it could also be direct WRITE). + */ +static inline int cfq_bio_sync(struct bio *bio) +{ + if (bio_data_dir(bio) == READ || bio_sync(bio)) + return 1; + + return 0; +} /* * scheduler run of queue, if there are requests pending and no one in the @@ -253,32 +250,69 @@ static struct cfq_queue *cfq_get_queue(struct cfq_data *cfqd, unsigned int key, */ static inline void cfq_schedule_dispatch(struct cfq_data *cfqd) { - if (cfqd->busy_queues) - kblockd_schedule_work(&cfqd->unplug_work); + if (cfqd->busy_queues) { + cfq_log(cfqd, "schedule dispatch"); + kblockd_schedule_work(cfqd->queue, &cfqd->unplug_work); + } } -static int cfq_queue_empty(request_queue_t *q) +static int cfq_queue_empty(struct request_queue *q) { struct cfq_data *cfqd = q->elevator->elevator_data; return !cfqd->busy_queues; } -static inline pid_t cfq_queue_pid(struct task_struct *task, int rw) +/* + * 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_slice(struct cfq_data *cfqd, int sync, + unsigned short prio) +{ + const int base_slice = cfqd->cfq_slice[sync]; + + WARN_ON(prio >= IOPRIO_BE_NR); + + return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - prio)); +} + +static inline int +cfq_prio_to_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq) +{ + return cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio); +} + +static inline void +cfq_set_prio_slice(struct cfq_data *cfqd, struct cfq_queue *cfqq) { - if (rw == READ || rw == WRITE_SYNC) - return task->pid; + cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq) + jiffies; + cfq_log_cfqq(cfqd, cfqq, "set_slice=%lu", cfqq->slice_end - jiffies); +} - return CFQ_KEY_ASYNC; +/* + * 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 crq1 and crq2 that is best served now. + * 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; @@ -286,20 +320,24 @@ cfq_choose_req(struct cfq_data *cfqd, struct cfq_rq *crq1, struct cfq_rq *crq2) #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 (cfq_crq_is_sync(crq1) && !cfq_crq_is_sync(crq2)) - return crq1; - else if (cfq_crq_is_sync(crq2) && !cfq_crq_is_sync(crq1)) - 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; + last = cfqd->last_position; /* * by definition, 1KiB is 2 sectors @@ -332,23 +370,23 @@ cfq_choose_req(struct cfq_data *cfqd, struct cfq_rq *crq1, struct cfq_rq *crq2) * 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, @@ -357,115 +395,189 @@ cfq_choose_req(struct cfq_data *cfqd, struct cfq_rq *crq1, struct cfq_rq *crq2) * since back seek takes more time than forward. */ if (s1 <= s2) - return crq1; + return rq1; else - return crq2; + return rq2; } } /* + * The below is leftmost cache rbtree addon + */ +static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root) +{ + if (!root->left) + root->left = rb_first(&root->rb); + + if (root->left) + return rb_entry(root->left, struct cfq_queue, rb_node); + + return NULL; +} + +static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root) +{ + if (root->left == n) + root->left = NULL; + + rb_erase(n, &root->rb); + RB_CLEAR_NODE(n); +} + +/* * 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) +static unsigned long cfq_slice_offset(struct cfq_data *cfqd, + struct cfq_queue *cfqq) { - struct cfq_data *cfqd = cfqq->cfqd; - struct list_head *list, *entry; + /* + * just an approximation, should be ok. + */ + return (cfqd->busy_queues - 1) * (cfq_prio_slice(cfqd, 1, 0) - + cfq_prio_slice(cfqd, cfq_cfqq_sync(cfqq), cfqq->ioprio)); +} - BUG_ON(!cfq_cfqq_on_rr(cfqq)); +/* + * The cfqd->service_tree holds all pending cfq_queue's that have + * requests waiting to be processed. It is sorted in the order that + * we will service the queues. + */ +static void cfq_service_tree_add(struct cfq_data *cfqd, + struct cfq_queue *cfqq, int add_front) +{ + struct rb_node **p, *parent; + struct cfq_queue *__cfqq; + unsigned long rb_key; + int left; + + if (cfq_class_idle(cfqq)) { + rb_key = CFQ_IDLE_DELAY; + parent = rb_last(&cfqd->service_tree.rb); + if (parent && parent != &cfqq->rb_node) { + __cfqq = rb_entry(parent, struct cfq_queue, rb_node); + rb_key += __cfqq->rb_key; + } else + rb_key += jiffies; + } else if (!add_front) { + rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies; + rb_key += cfqq->slice_resid; + cfqq->slice_resid = 0; + } else + rb_key = 0; + + if (!RB_EMPTY_NODE(&cfqq->rb_node)) { + /* + * same position, nothing more to do + */ + if (rb_key == cfqq->rb_key) + return; - list_del(&cfqq->cfq_list); + cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree); + } + + left = 1; + parent = NULL; + p = &cfqd->service_tree.rb.rb_node; + while (*p) { + struct rb_node **n; + + parent = *p; + __cfqq = rb_entry(parent, struct cfq_queue, rb_node); - 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 + * sort RT queues first, we always want to give + * preference to them. IDLE queues goes to the back. + * after that, sort on the next service time. */ - if (cfq_cfqq_dispatched(cfqq)) - list = &cfqd->busy_rr; + if (cfq_class_rt(cfqq) > cfq_class_rt(__cfqq)) + n = &(*p)->rb_left; + else if (cfq_class_rt(cfqq) < cfq_class_rt(__cfqq)) + n = &(*p)->rb_right; + else if (cfq_class_idle(cfqq) < cfq_class_idle(__cfqq)) + n = &(*p)->rb_left; + else if (cfq_class_idle(cfqq) > cfq_class_idle(__cfqq)) + n = &(*p)->rb_right; + else if (rb_key < __cfqq->rb_key) + n = &(*p)->rb_left; else - list = &cfqd->rr_list[cfqq->ioprio]; - } + n = &(*p)->rb_right; - /* - * 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 (n == &(*p)->rb_right) + left = 0; - list_add_tail(&cfqq->cfq_list, list); - return; + p = n; } - /* - * sort by when queue was last serviced - */ - entry = list; - while ((entry = entry->prev) != list) { - struct cfq_queue *__cfqq = list_entry_cfqq(entry); + if (left) + cfqd->service_tree.left = &cfqq->rb_node; - if (!__cfqq->service_last) - break; - if (time_before(__cfqq->service_last, cfqq->service_last)) - break; - } + cfqq->rb_key = rb_key; + rb_link_node(&cfqq->rb_node, parent, p); + rb_insert_color(&cfqq->rb_node, &cfqd->service_tree.rb); +} - list_add(&cfqq->cfq_list, entry); +/* + * Update cfqq's position in the service tree. + */ +static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq) +{ + /* + * Resorting requires the cfqq to be on the RR list already. + */ + if (cfq_cfqq_on_rr(cfqq)) + cfq_service_tree_add(cfqd, cfqq, 0); } /* * add to busy list of queues for service, trying to be fair in ordering * the pending list according to last request service */ -static inline void -cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) +static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) { + cfq_log_cfqq(cfqd, cfqq, "add_to_rr"); BUG_ON(cfq_cfqq_on_rr(cfqq)); cfq_mark_cfqq_on_rr(cfqq); cfqd->busy_queues++; - cfq_resort_rr_list(cfqq, 0); + cfq_resort_rr_list(cfqd, cfqq); } -static inline void -cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) +/* + * Called when the cfqq no longer has requests pending, remove it from + * the service tree. + */ +static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) { + cfq_log_cfqq(cfqd, cfqq, "del_from_rr"); BUG_ON(!cfq_cfqq_on_rr(cfqq)); cfq_clear_cfqq_on_rr(cfqq); - list_move(&cfqq->cfq_list, &cfqd->empty_list); + + if (!RB_EMPTY_NODE(&cfqq->rb_node)) + cfq_rb_erase(&cfqq->rb_node, &cfqd->service_tree); BUG_ON(!cfqd->busy_queues); cfqd->busy_queues--; @@ -474,99 +586,114 @@ cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq) /* * rb tree support functions */ -static inline void cfq_del_crq_rb(struct cfq_rq *crq) +static 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 = cfq_crq_is_sync(crq); + 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[cfq_crq_is_sync(crq)]++; + cfqq->queued[rq_is_sync(rq)]++; /* * looks a little odd, but the first insert might return an alias. * 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); + + /* + * check if this request is a better next-serve candidate + */ + cfqq->next_rq = cfq_choose_req(cfqd, cfqq->next_rq, rq); + BUG_ON(!cfqq->next_rq); } -static inline void -cfq_reposition_crq_rb(struct cfq_queue *cfqq, struct cfq_rq *crq) +static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq) { - elv_rb_del(&cfqq->sort_list, crq->request); - cfqq->queued[cfq_crq_is_sync(crq)]--; - cfq_add_crq_rb(crq); + elv_rb_del(&cfqq->sort_list, rq); + cfqq->queued[rq_is_sync(rq)]--; + 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); + struct cfq_io_context *cic; struct cfq_queue *cfqq; - cfqq = cfq_find_cfq_hash(cfqd, key, tsk->ioprio); - if (cfqq) + cic = cfq_cic_lookup(cfqd, tsk->io_context); + if (!cic) + return NULL; + + cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio)); + if (cfqq) { + sector_t sector = bio->bi_sector + bio_sectors(bio); + return elv_rb_find(&cfqq->sort_list, sector); + } return NULL; } -static void cfq_activate_request(request_queue_t *q, struct request *rq) +static void cfq_activate_request(struct request_queue *q, struct request *rq) { struct cfq_data *cfqd = q->elevator->elevator_data; cfqd->rq_in_driver++; + cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "activate rq, drv=%d", + cfqd->rq_in_driver); - /* - * 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; + cfqd->last_position = rq->hard_sector + rq->hard_nr_sectors; } -static void cfq_deactivate_request(request_queue_t *q, struct request *rq) +static void cfq_deactivate_request(struct request_queue *q, struct request *rq) { struct cfq_data *cfqd = q->elevator->elevator_data; WARN_ON(!cfqd->rq_in_driver); cfqd->rq_in_driver--; + cfq_log_cfqq(cfqd, RQ_CFQQ(rq), "deactivate rq, drv=%d", + cfqd->rq_in_driver); } 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); + + cfqq->cfqd->rq_queued--; + if (rq_is_meta(rq)) { + WARN_ON(!cfqq->meta_pending); + cfqq->meta_pending--; + } } -static int -cfq_merge(request_queue_t *q, struct request **req, struct bio *bio) +static int cfq_merge(struct request_queue *q, struct request **req, + struct bio *bio) { struct cfq_data *cfqd = q->elevator->elevator_data; struct request *__rq; @@ -580,20 +707,18 @@ cfq_merge(request_queue_t *q, struct request **req, struct bio *bio) return ELEVATOR_NO_MERGE; } -static void cfq_merged_request(request_queue_t *q, struct request *req, +static void cfq_merged_request(struct request_queue *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); } } static void -cfq_merged_requests(request_queue_t *q, struct request *rq, +cfq_merged_requests(struct request_queue *q, struct request *rq, struct request *next) { /* @@ -606,20 +731,44 @@ cfq_merged_requests(request_queue_t *q, struct request *rq, cfq_remove_request(next); } -static inline void -__cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq) +static int cfq_allow_merge(struct request_queue *q, struct request *rq, + struct bio *bio) { - if (cfqq) { - /* - * stop potential idle class queues waiting service - */ - del_timer(&cfqd->idle_class_timer); + struct cfq_data *cfqd = q->elevator->elevator_data; + struct cfq_io_context *cic; + struct cfq_queue *cfqq; + + /* + * Disallow merge of a sync bio into an async request. + */ + if (cfq_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. + */ + cic = cfq_cic_lookup(cfqd, current->io_context); + if (!cic) + return 0; - cfqq->slice_start = jiffies; + cfqq = cic_to_cfqq(cic, cfq_bio_sync(bio)); + if (cfqq == RQ_CFQQ(rq)) + return 1; + + return 0; +} + +static void __cfq_set_active_queue(struct cfq_data *cfqd, + struct cfq_queue *cfqq) +{ + if (cfqq) { + cfq_log_cfqq(cfqd, cfqq, "set_active"); 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); + cfq_clear_cfqq_queue_new(cfqq); } cfqd->active_queue = cfqq; @@ -630,32 +779,25 @@ __cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq) */ static void __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq, - int preempted) + int timed_out) { - unsigned long now = jiffies; + cfq_log_cfqq(cfqd, cfqq, "slice expired t=%d", timed_out); 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); /* - * store what was left of this slice, if the queue idled out - * or was preempted + * store what was left of this slice, if the queue idled/timed out */ - 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; + cfq_log_cfqq(cfqd, cfqq, "resid=%ld", cfqq->slice_resid); + } - if (cfq_cfqq_on_rr(cfqq)) - cfq_resort_rr_list(cfqq, preempted); + cfq_resort_rr_list(cfqd, cfqq); if (cfqq == cfqd->active_queue) cfqd->active_queue = NULL; @@ -664,212 +806,174 @@ __cfq_slice_expired(struct cfq_data *cfqd, struct cfq_queue *cfqq, 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) +static inline void cfq_slice_expired(struct cfq_data *cfqd, int timed_out) { struct cfq_queue *cfqq = cfqd->active_queue; if (cfqq) - __cfq_slice_expired(cfqd, cfqq, preempted); + __cfq_slice_expired(cfqd, cfqq, timed_out); } /* - * 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 + * Get next queue for service. Unless we have a queue preemption, + * we'll simply select the first cfqq in the service tree. */ -static int cfq_get_next_prio_level(struct cfq_data *cfqd) +static struct cfq_queue *cfq_get_next_queue(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; - } - } while (1); - - if (unlikely(prio == -1)) - return -1; + if (RB_EMPTY_ROOT(&cfqd->service_tree.rb)) + return NULL; - BUG_ON(prio >= CFQ_PRIO_LISTS); + return cfq_rb_first(&cfqd->service_tree); +} - list_splice_init(&cfqd->rr_list[prio], &cfqd->cur_rr); +/* + * Get and set a new active queue for service. + */ +static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd) +{ + struct cfq_queue *cfqq; - 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; - } + cfqq = cfq_get_next_queue(cfqd); + __cfq_set_active_queue(cfqd, cfqq); + return cfqq; +} - return prio; +static inline sector_t cfq_dist_from_last(struct cfq_data *cfqd, + struct request *rq) +{ + if (rq->sector >= cfqd->last_position) + return rq->sector - cfqd->last_position; + else + return cfqd->last_position - rq->sector; } -static struct cfq_queue *cfq_set_active_queue(struct cfq_data *cfqd) +static inline int cfq_rq_close(struct cfq_data *cfqd, struct request *rq) { - struct cfq_queue *cfqq = NULL; + struct cfq_io_context *cic = cfqd->active_cic; - /* - * 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) - cfqq = list_entry_cfqq(cfqd->cur_rr.next); + if (!sample_valid(cic->seek_samples)) + return 0; - /* - * 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)) - cfqq = list_entry_cfqq(cfqd->busy_rr.next); + return cfq_dist_from_last(cfqd, rq) <= cic->seek_mean; +} +static int cfq_close_cooperator(struct cfq_data *cfq_data, + struct cfq_queue *cfqq) +{ /* - * 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 + * We should notice if some of the queues are cooperating, eg + * working closely on the same area of the disk. In that case, + * we can group them together and don't waste time idling. */ - if (!cfqq && !list_empty(&cfqd->idle_rr)) { - 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); - return cfqq; + return 0; } -#define CIC_SEEKY(cic) ((cic)->seek_mean > (128 * 1024)) - -static int cfq_arm_slice_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq) +#define CIC_SEEKY(cic) ((cic)->seek_mean > (8 * 1024)) +static void cfq_arm_slice_timer(struct cfq_data *cfqd) { + struct cfq_queue *cfqq = cfqd->active_queue; struct cfq_io_context *cic; unsigned long sl; + /* + * SSD device without seek penalty, disable idling. But only do so + * for devices that support queuing, otherwise we still have a problem + * with sync vs async workloads. + */ + if (blk_queue_nonrot(cfqd->queue) && cfqd->hw_tag) + return; + WARN_ON(!RB_EMPTY_ROOT(&cfqq->sort_list)); - WARN_ON(cfqq != cfqd->active_queue); + WARN_ON(cfq_cfqq_slice_new(cfqq)); /* * idle is disabled, either manually or by past process history */ - if (!cfqd->cfq_slice_idle) - return 0; - if (!cfq_cfqq_idle_window(cfqq)) - return 0; + if (!cfqd->cfq_slice_idle || !cfq_cfqq_idle_window(cfqq)) + return; + + /* + * still requests with the driver, don't idle + */ + if (cfqd->rq_in_driver) + return; + /* * task has exited, don't wait */ cic = cfqd->active_cic; - if (!cic || !cic->ioc->task) - return 0; + if (!cic || !atomic_read(&cic->ioc->nr_tasks)) + return; + + /* + * See if this prio level has a good candidate + */ + if (cfq_close_cooperator(cfqd, cfqq) && + (sample_valid(cic->ttime_samples) && cic->ttime_mean > 2)) + return; cfq_mark_cfqq_must_dispatch(cfqq); cfq_mark_cfqq_wait_request(cfqq); - sl = min(cfqq->slice_end - 1, (unsigned long) cfqd->cfq_slice_idle); - /* * 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 */ + sl = cfqd->cfq_slice_idle; if (sample_valid(cic->seek_samples) && CIC_SEEKY(cic)) - sl = min(sl, msecs_to_jiffies(2)); + sl = min(sl, msecs_to_jiffies(CFQ_MIN_TT)); mod_timer(&cfqd->idle_slice_timer, jiffies + sl); - return 1; + cfq_log(cfqd, "arm_idle: %lu", sl); } -static void cfq_dispatch_insert(request_queue_t *q, struct cfq_rq *crq) +/* + * Move request from internal lists to the request queue dispatch list. + */ +static void cfq_dispatch_insert(struct request_queue *q, struct request *rq) { struct cfq_data *cfqd = q->elevator->elevator_data; - struct cfq_queue *cfqq = crq->cfq_queue; - struct request *rq; + struct cfq_queue *cfqq = RQ_CFQQ(rq); + + cfq_log_cfqq(cfqd, cfqq, "dispatch_insert"); - cfq_remove_request(crq->request); - cfqq->on_dispatch[cfq_crq_is_sync(crq)]++; - elv_dispatch_sort(q, crq->request); + cfq_remove_request(rq); + cfqq->dispatched++; + elv_dispatch_sort(q, rq); - rq = list_entry(q->queue_head.prev, struct request, queuelist); - cfqd->last_sector = rq->sector + rq->nr_sectors; + if (cfq_cfqq_sync(cfqq)) + cfqd->sync_flight++; } /* * 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 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); + cfq_mark_cfqq_fifo_expire(cfqq); - crq = RQ_DATA(list_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; -} - -/* - * 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_sync(cfqq); + rq = rq_entry_fifo(cfqq->fifo.next); - return base_slice + (base_slice/CFQ_SLICE_SCALE * (4 - cfqq->ioprio)); -} + if (time_before(jiffies, rq->start_time + cfqd->cfq_fifo_expire[fifo])) + rq = NULL; -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; + cfq_log_cfqq(cfqd, cfqq, "fifo=%p", rq); + return rq; } static inline int @@ -883,11 +987,11 @@ cfq_prio_to_maxrq(struct cfq_data *cfqd, struct cfq_queue *cfqq) } /* - * get next queue for service + * Select a queue for service. If we have a current active queue, + * check whether to continue servicing it, or retrieve and set a new one. */ static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd) { - unsigned long now = jiffies; struct cfq_queue *cfqq; cfqq = cfqd->active_queue; @@ -895,23 +999,27 @@ static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd) goto new_queue; /* - * slice has expired + * The active queue has run out of time, expire it and select new. */ - if (!cfq_cfqq_must_dispatch(cfqq) && time_after(now, cfqq->slice_end)) + if (cfq_slice_used(cfqq)) goto expire; /* - * if queue has requests, dispatch one. if not, check if - * enough slice is left to wait for one + * The active queue has requests and isn't expired, allow it to + * dispatch. */ if (!RB_EMPTY_ROOT(&cfqq->sort_list)) goto keep_queue; - else if (cfq_cfqq_dispatched(cfqq)) { + + /* + * No requests pending. If the active queue still has requests in + * flight or is idling for a new request, allow either of these + * conditions to happen (or time out) before selecting a new queue. + */ + if (timer_pending(&cfqd->idle_slice_timer) || + (cfqq->dispatched && cfq_cfqq_idle_window(cfqq))) { cfqq = NULL; goto keep_queue; - } else if (cfq_cfqq_class_sync(cfqq)) { - if (cfq_arm_slice_timer(cfqd, cfqq)) - return NULL; } expire: @@ -922,6 +1030,10 @@ keep_queue: return cfqq; } +/* + * Dispatch some requests from cfqq, moving them to the request queue + * dispatch list. + */ static int __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq, int max_dispatch) @@ -931,25 +1043,25 @@ __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq, 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; + rq = cfq_check_fifo(cfqq); + if (rq == 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)) @@ -958,67 +1070,56 @@ __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq, } 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)) + if (cfqd->busy_queues > 1 && ((!cfq_cfqq_sync(cfqq) && + dispatched >= cfq_prio_to_maxrq(cfqd, cfqq)) || + cfq_class_idle(cfqq))) { + cfqq->slice_end = jiffies + 1; cfq_slice_expired(cfqd, 0); + } return dispatched; } -static int -cfq_forced_dispatch_cfqqs(struct list_head *list) +static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq) { - struct cfq_queue *cfqq, *next; - struct cfq_rq *crq; - int dispatched; + int dispatched = 0; - dispatched = 0; - list_for_each_entry_safe(cfqq, next, list, cfq_list) { - while ((crq = cfqq->next_crq)) { - cfq_dispatch_insert(cfqq->cfqd->queue, crq); - dispatched++; - } - BUG_ON(!list_empty(&cfqq->fifo)); + while (cfqq->next_rq) { + cfq_dispatch_insert(cfqq->cfqd->queue, cfqq->next_rq); + dispatched++; } + BUG_ON(!list_empty(&cfqq->fifo)); return dispatched; } -static int -cfq_forced_dispatch(struct cfq_data *cfqd) +/* + * Drain our current requests. Used for barriers and when switching + * io schedulers on-the-fly. + */ +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]); + struct cfq_queue *cfqq; + int dispatched = 0; - dispatched += cfq_forced_dispatch_cfqqs(&cfqd->busy_rr); - dispatched += cfq_forced_dispatch_cfqqs(&cfqd->cur_rr); - dispatched += cfq_forced_dispatch_cfqqs(&cfqd->idle_rr); + while ((cfqq = cfq_rb_first(&cfqd->service_tree)) != NULL) + dispatched += __cfq_forced_dispatch_cfqq(cfqq); cfq_slice_expired(cfqd, 0); BUG_ON(cfqd->busy_queues); + cfq_log(cfqd, "forced_dispatch=%d\n", dispatched); return dispatched; } -static int -cfq_dispatch_requests(request_queue_t *q, int force) +static int cfq_dispatch_requests(struct request_queue *q, int force) { struct cfq_data *cfqd = q->elevator->elevator_data; - struct cfq_queue *cfqq, *prev_cfqq; + struct cfq_queue *cfqq; int dispatched; if (!cfqd->busy_queues) @@ -1028,42 +1129,37 @@ cfq_dispatch_requests(request_queue_t *q, int force) return cfq_forced_dispatch(cfqd); dispatched = 0; - prev_cfqq = NULL; while ((cfqq = cfq_select_queue(cfqd)) != NULL) { int max_dispatch; - /* - * Don't repeat dispatch from the previous queue. - */ - if (prev_cfqq == cfqq) - break; - - cfq_clear_cfqq_must_dispatch(cfqq); - cfq_clear_cfqq_wait_request(cfqq); - del_timer(&cfqd->idle_slice_timer); - max_dispatch = cfqd->cfq_quantum; if (cfq_class_idle(cfqq)) max_dispatch = 1; - dispatched += __cfq_dispatch_requests(cfqd, cfqq, max_dispatch); + if (cfqq->dispatched >= max_dispatch) { + if (cfqd->busy_queues > 1) + break; + if (cfqq->dispatched >= 4 * max_dispatch) + break; + } - /* - * If the dispatch cfqq has idling enabled and is still - * the active queue, break out. - */ - if (cfq_cfqq_idle_window(cfqq) && cfqd->active_queue) + if (cfqd->sync_flight && !cfq_cfqq_sync(cfqq)) break; - prev_cfqq = cfqq; + cfq_clear_cfqq_must_dispatch(cfqq); + cfq_clear_cfqq_wait_request(cfqq); + del_timer(&cfqd->idle_slice_timer); + + dispatched += __cfq_dispatch_requests(cfqd, cfqq, max_dispatch); } + cfq_log(cfqd, "dispatched=%d", dispatched); return dispatched; } /* - * 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. */ @@ -1076,144 +1172,186 @@ static void cfq_put_queue(struct cfq_queue *cfqq) if (!atomic_dec_and_test(&cfqq->ref)) return; + cfq_log_cfqq(cfqd, cfqq, "put_queue"); BUG_ON(rb_first(&cfqq->sort_list)); BUG_ON(cfqq->allocated[READ] + cfqq->allocated[WRITE]); BUG_ON(cfq_cfqq_on_rr(cfqq)); - if (unlikely(cfqd->active_queue == cfqq)) + if (unlikely(cfqd->active_queue == cfqq)) { __cfq_slice_expired(cfqd, cfqq, 0); + cfq_schedule_dispatch(cfqd); + } - /* - * 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); } -static inline struct cfq_queue * -__cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned int prio, - const int hashval) +/* + * Must always be called with the rcu_read_lock() held + */ +static void +__call_for_each_cic(struct io_context *ioc, + void (*func)(struct io_context *, struct cfq_io_context *)) { - struct hlist_head *hash_list = &cfqd->cfq_hash[hashval]; - struct hlist_node *entry; - struct cfq_queue *__cfqq; - - hlist_for_each_entry(__cfqq, entry, hash_list, cfq_hash) { - const unsigned short __p = IOPRIO_PRIO_VALUE(__cfqq->org_ioprio_class, __cfqq->org_ioprio); - - if (__cfqq->key == key && (__p == prio || !prio)) - return __cfqq; - } + struct cfq_io_context *cic; + struct hlist_node *n; - return NULL; + hlist_for_each_entry_rcu(cic, n, &ioc->cic_list, cic_list) + func(ioc, cic); } -static struct cfq_queue * -cfq_find_cfq_hash(struct cfq_data *cfqd, unsigned int key, unsigned short prio) +/* + * Call func for each cic attached to this ioc. + */ +static void +call_for_each_cic(struct io_context *ioc, + void (*func)(struct io_context *, struct cfq_io_context *)) { - return __cfq_find_cfq_hash(cfqd, key, prio, hash_long(key, CFQ_QHASH_SHIFT)); + rcu_read_lock(); + __call_for_each_cic(ioc, func); + rcu_read_unlock(); } -static void cfq_free_io_context(struct io_context *ioc) +static void cfq_cic_free_rcu(struct rcu_head *head) { - struct cfq_io_context *__cic; - struct rb_node *n; - int freed = 0; + struct cfq_io_context *cic; + + cic = container_of(head, struct cfq_io_context, rcu_head); + + kmem_cache_free(cfq_ioc_pool, cic); + elv_ioc_count_dec(ioc_count); - 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); - kmem_cache_free(cfq_ioc_pool, __cic); - freed++; + if (ioc_gone) { + /* + * CFQ scheduler is exiting, grab exit lock and check + * the pending io context count. If it hits zero, + * complete ioc_gone and set it back to NULL + */ + spin_lock(&ioc_gone_lock); + if (ioc_gone && !elv_ioc_count_read(ioc_count)) { + complete(ioc_gone); + ioc_gone = NULL; + } + spin_unlock(&ioc_gone_lock); } +} - if (atomic_sub_and_test(freed, &ioc_count) && ioc_gone) - complete(ioc_gone); +static void cfq_cic_free(struct cfq_io_context *cic) +{ + call_rcu(&cic->rcu_head, cfq_cic_free_rcu); } -static void cfq_trim(struct io_context *ioc) +static void cic_free_func(struct io_context *ioc, struct cfq_io_context *cic) { - ioc->set_ioprio = NULL; - cfq_free_io_context(ioc); + unsigned long flags; + + BUG_ON(!cic->dead_key); + + spin_lock_irqsave(&ioc->lock, flags); + radix_tree_delete(&ioc->radix_root, cic->dead_key); + hlist_del_rcu(&cic->cic_list); + spin_unlock_irqrestore(&ioc->lock, flags); + + cfq_cic_free(cic); } /* - * Called with interrupts disabled + * Must be called with rcu_read_lock() held or preemption otherwise disabled. + * Only two callers of this - ->dtor() which is called with the rcu_read_lock(), + * and ->trim() which is called with the task lock held */ -static void cfq_exit_single_io_context(struct cfq_io_context *cic) +static void cfq_free_io_context(struct io_context *ioc) { - struct cfq_data *cfqd = cic->key; - request_queue_t *q; + /* + * ioc->refcount is zero here, or we are called from elv_unregister(), + * so no more cic's are allowed to be linked into this ioc. So it + * should be ok to iterate over the known list, we will see all cic's + * since no new ones are added. + */ + __call_for_each_cic(ioc, cic_free_func); +} - if (!cfqd) - return; +static void cfq_exit_cfqq(struct cfq_data *cfqd, struct cfq_queue *cfqq) +{ + if (unlikely(cfqq == cfqd->active_queue)) { + __cfq_slice_expired(cfqd, cfqq, 0); + cfq_schedule_dispatch(cfqd); + } - q = cfqd->queue; + cfq_put_queue(cfqq); +} - WARN_ON(!irqs_disabled()); +static void __cfq_exit_single_io_context(struct cfq_data *cfqd, + struct cfq_io_context *cic) +{ + struct io_context *ioc = cic->ioc; - spin_lock(q->queue_lock); + list_del_init(&cic->queue_list); + + /* + * Make sure key == NULL is seen for dead queues + */ + smp_wmb(); + cic->dead_key = (unsigned long) cic->key; + cic->key = NULL; + + if (ioc->ioc_data == cic) + rcu_assign_pointer(ioc->ioc_data, 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); } -static void cfq_exit_io_context(struct io_context *ioc) +static void cfq_exit_single_io_context(struct io_context *ioc, + struct cfq_io_context *cic) { - 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); + struct cfq_data *cfqd = cic->key; - n = rb_first(&ioc->cic_root); - while (n != NULL) { - __cic = rb_entry(n, struct cfq_io_context, rb_node); + if (cfqd) { + struct request_queue *q = cfqd->queue; + unsigned long flags; - cfq_exit_single_io_context(__cic); - n = rb_next(n); + spin_lock_irqsave(q->queue_lock, flags); + __cfq_exit_single_io_context(cfqd, cic); + spin_unlock_irqrestore(q->queue_lock, flags); } +} - spin_unlock_irqrestore(&cfq_exit_lock, flags); +/* + * The process that ioc belongs to has exited, we need to clean up + * and put the internal structures we have that belongs to that process. + */ +static void cfq_exit_io_context(struct io_context *ioc) +{ + call_for_each_cic(ioc, cfq_exit_single_io_context); } 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 | __GFP_ZERO, + cfqd->queue->node); if (cic) { - memset(cic, 0, sizeof(*cic)); cic->last_end_request = jiffies; INIT_LIST_HEAD(&cic->queue_list); + INIT_HLIST_NODE(&cic->cic_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; } -static void cfq_init_prio_data(struct cfq_queue *cfqq) +static void cfq_init_prio_data(struct cfq_queue *cfqq, struct io_context *ioc) { struct task_struct *tsk = current; int ioprio_class; @@ -1221,30 +1359,30 @@ static void cfq_init_prio_data(struct cfq_queue *cfqq) if (!cfq_cfqq_prio_changed(cfqq)) return; - ioprio_class = IOPRIO_PRIO_CLASS(tsk->ioprio); + ioprio_class = IOPRIO_PRIO_CLASS(ioc->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; - cfq_clear_cfqq_idle_window(cfqq); - break; + default: + printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class); + case IOPRIO_CLASS_NONE: + /* + * no prio set, inherit CPU scheduling settings + */ + cfqq->ioprio = task_nice_ioprio(tsk); + cfqq->ioprio_class = task_nice_ioclass(tsk); + break; + case IOPRIO_CLASS_RT: + cfqq->ioprio = task_ioprio(ioc); + cfqq->ioprio_class = IOPRIO_CLASS_RT; + break; + case IOPRIO_CLASS_BE: + cfqq->ioprio = task_ioprio(ioc); + cfqq->ioprio_class = IOPRIO_CLASS_BE; + break; + case IOPRIO_CLASS_IDLE: + cfqq->ioprio_class = IOPRIO_CLASS_IDLE; + cfqq->ioprio = 7; + cfq_clear_cfqq_idle_window(cfqq); + break; } /* @@ -1253,28 +1391,24 @@ static void cfq_init_prio_data(struct cfq_queue *cfqq) */ 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_clear_cfqq_prio_changed(cfqq); } -static inline void changed_ioprio(struct cfq_io_context *cic) +static void changed_ioprio(struct io_context *ioc, struct cfq_io_context *cic) { 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) { struct cfq_queue *new_cfqq; - new_cfqq = cfq_get_queue(cfqd, CFQ_KEY_ASYNC, cic->ioc->task, - GFP_ATOMIC); + new_cfqq = cfq_get_queue(cfqd, ASYNC, cic->ioc, GFP_ATOMIC); if (new_cfqq) { cic->cfqq[ASYNC] = new_cfqq; cfq_put_queue(cfqq); @@ -1285,167 +1419,228 @@ static inline void changed_ioprio(struct cfq_io_context *cic) 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); - - n = rb_first(&ioc->cic_root); - while (n != NULL) { - cic = rb_entry(n, struct cfq_io_context, rb_node); - - changed_ioprio(cic); - n = rb_next(n); - } - - spin_unlock(&cfq_exit_lock); - - return 0; + call_for_each_cic(ioc, changed_ioprio); + ioc->ioprio_changed = 0; } static struct cfq_queue * -cfq_get_queue(struct cfq_data *cfqd, unsigned int key, struct task_struct *tsk, - gfp_t gfp_mask) +cfq_find_alloc_queue(struct cfq_data *cfqd, int is_sync, + struct io_context *ioc, gfp_t gfp_mask) { - const int hashval = hash_long(key, CFQ_QHASH_SHIFT); struct cfq_queue *cfqq, *new_cfqq = NULL; - unsigned short ioprio; + struct cfq_io_context *cic; retry: - ioprio = tsk->ioprio; - cfqq = __cfq_find_cfq_hash(cfqd, key, ioprio, hashval); + cic = cfq_cic_lookup(cfqd, ioc); + /* cic always exists here */ + cfqq = cic_to_cfqq(cic, is_sync); if (!cfqq) { if (new_cfqq) { 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 | __GFP_ZERO, + 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 | __GFP_ZERO, + cfqd->queue->node); if (!cfqq) goto out; } - memset(cfqq, 0, sizeof(*cfqq)); - - INIT_HLIST_NODE(&cfqq->cfq_hash); - INIT_LIST_HEAD(&cfqq->cfq_list); + RB_CLEAR_NODE(&cfqq->rb_node); 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; - 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_init_prio_data(cfqq); + cfq_mark_cfqq_queue_new(cfqq); + + cfq_init_prio_data(cfqq, ioc); + + if (is_sync) { + if (!cfq_class_idle(cfqq)) + cfq_mark_cfqq_idle_window(cfqq); + cfq_mark_cfqq_sync(cfqq); + } + cfqq->pid = current->pid; + cfq_log_cfqq(cfqd, cfqq, "alloced"); } if (new_cfqq) kmem_cache_free(cfq_pool, new_cfqq); - atomic_inc(&cfqq->ref); out: WARN_ON((gfp_mask & __GFP_WAIT) && !cfqq); return cfqq; } +static struct cfq_queue ** +cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio) +{ + switch (ioprio_class) { + case IOPRIO_CLASS_RT: + return &cfqd->async_cfqq[0][ioprio]; + case IOPRIO_CLASS_BE: + return &cfqd->async_cfqq[1][ioprio]; + case IOPRIO_CLASS_IDLE: + return &cfqd->async_idle_cfqq; + default: + BUG(); + } +} + +static struct cfq_queue * +cfq_get_queue(struct cfq_data *cfqd, int is_sync, struct io_context *ioc, + gfp_t gfp_mask) +{ + const int ioprio = task_ioprio(ioc); + const int ioprio_class = task_ioprio_class(ioc); + struct cfq_queue **async_cfqq = NULL; + struct cfq_queue *cfqq = NULL; + + if (!is_sync) { + async_cfqq = cfq_async_queue_prio(cfqd, ioprio_class, ioprio); + cfqq = *async_cfqq; + } + + if (!cfqq) { + cfqq = cfq_find_alloc_queue(cfqd, is_sync, ioc, gfp_mask); + if (!cfqq) + return NULL; + } + + /* + * pin the queue now that it's allocated, scheduler exit will prune it + */ + if (!is_sync && !(*async_cfqq)) { + atomic_inc(&cfqq->ref); + *async_cfqq = cfqq; + } + + atomic_inc(&cfqq->ref); + return cfqq; +} + +/* + * We drop cfq io contexts lazily, so we may find a dead one. + */ static void -cfq_drop_dead_cic(struct io_context *ioc, struct cfq_io_context *cic) +cfq_drop_dead_cic(struct cfq_data *cfqd, struct io_context *ioc, + struct cfq_io_context *cic) { - spin_lock(&cfq_exit_lock); - 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); + unsigned long flags; + + WARN_ON(!list_empty(&cic->queue_list)); + + spin_lock_irqsave(&ioc->lock, flags); + + BUG_ON(ioc->ioc_data == cic); + + radix_tree_delete(&ioc->radix_root, (unsigned long) cfqd); + hlist_del_rcu(&cic->cic_list); + spin_unlock_irqrestore(&ioc->lock, flags); + + cfq_cic_free(cic); } static struct cfq_io_context * -cfq_cic_rb_lookup(struct cfq_data *cfqd, struct io_context *ioc) +cfq_cic_lookup(struct cfq_data *cfqd, struct io_context *ioc) { - struct rb_node *n; struct cfq_io_context *cic; - void *k, *key = cfqd; + unsigned long flags; + void *k; -restart: - n = ioc->cic_root.rb_node; - while (n) { - cic = rb_entry(n, struct cfq_io_context, rb_node); + if (unlikely(!ioc)) + return NULL; + + rcu_read_lock(); + + /* + * we maintain a last-hit cache, to avoid browsing over the tree + */ + cic = rcu_dereference(ioc->ioc_data); + if (cic && cic->key == cfqd) { + rcu_read_unlock(); + return cic; + } + + do { + cic = radix_tree_lookup(&ioc->radix_root, (unsigned long) cfqd); + rcu_read_unlock(); + if (!cic) + break; /* ->key must be copied to avoid race with cfq_exit_queue() */ k = cic->key; if (unlikely(!k)) { - cfq_drop_dead_cic(ioc, cic); - goto restart; + cfq_drop_dead_cic(cfqd, ioc, cic); + rcu_read_lock(); + continue; } - if (key < k) - n = n->rb_left; - else if (key > k) - n = n->rb_right; - else - return cic; - } + spin_lock_irqsave(&ioc->lock, flags); + rcu_assign_pointer(ioc->ioc_data, cic); + spin_unlock_irqrestore(&ioc->lock, flags); + break; + } while (1); - return NULL; + return cic; } -static inline void -cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc, - struct cfq_io_context *cic) +/* + * Add cic into ioc, using cfqd as the search key. This enables us to lookup + * the process specific cfq io context when entered from the block layer. + * Also adds the cic to a per-cfqd list, used when this queue is removed. + */ +static int cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc, + struct cfq_io_context *cic, gfp_t gfp_mask) { - struct rb_node **p; - struct rb_node *parent; - struct cfq_io_context *__cic; - void *k; + unsigned long flags; + int ret; - cic->ioc = ioc; - cic->key = cfqd; + ret = radix_tree_preload(gfp_mask); + if (!ret) { + cic->ioc = ioc; + cic->key = cfqd; - ioc->set_ioprio = cfq_ioc_set_ioprio; -restart: - parent = NULL; - p = &ioc->cic_root.rb_node; - while (*p) { - parent = *p; - __cic = rb_entry(parent, struct cfq_io_context, rb_node); - /* ->key must be copied to avoid race with cfq_exit_queue() */ - k = __cic->key; - if (unlikely(!k)) { - cfq_drop_dead_cic(ioc, __cic); - goto restart; - } + spin_lock_irqsave(&ioc->lock, flags); + ret = radix_tree_insert(&ioc->radix_root, + (unsigned long) cfqd, cic); + if (!ret) + hlist_add_head_rcu(&cic->cic_list, &ioc->cic_list); + spin_unlock_irqrestore(&ioc->lock, flags); - if (cic->key < k) - p = &(*p)->rb_left; - else if (cic->key > k) - p = &(*p)->rb_right; - else - BUG(); + radix_tree_preload_end(); + + if (!ret) { + spin_lock_irqsave(cfqd->queue->queue_lock, flags); + list_add(&cic->queue_list, &cfqd->cic_list); + spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); + } } - spin_lock(&cfq_exit_lock); - rb_link_node(&cic->rb_node, parent, p); - rb_insert_color(&cic->rb_node, &ioc->cic_root); - list_add(&cic->queue_list, &cfqd->cic_list); - spin_unlock(&cfq_exit_lock); + if (ret) + printk(KERN_ERR "cfq: cic link failed!\n"); + + return ret; } /* @@ -1461,11 +1656,11 @@ cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask) 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; - cic = cfq_cic_rb_lookup(cfqd, ioc); + cic = cfq_cic_lookup(cfqd, ioc); if (cic) goto out; @@ -1473,9 +1668,17 @@ cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask) if (cic == NULL) goto err; - cfq_cic_link(cfqd, ioc, cic); + if (cfq_cic_link(cfqd, ioc, cic, gfp_mask)) + goto err_free; + out: + smp_read_barrier_depends(); + if (unlikely(ioc->ioprio_changed)) + cfq_ioc_set_ioprio(ioc); + return cic; +err_free: + cfq_cic_free(cic); err: put_io_context(ioc); return NULL; @@ -1484,22 +1687,8 @@ err: 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; @@ -1508,15 +1697,15 @@ cfq_update_io_thinktime(struct cfq_data *cfqd, struct cfq_io_context *cic) static void cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_io_context *cic, - struct cfq_rq *crq) + 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 @@ -1542,9 +1731,17 @@ static void cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq, struct cfq_io_context *cic) { - int enable_idle = cfq_cfqq_idle_window(cfqq); + int old_idle, enable_idle; + + /* + * Don't idle for async or idle io prio class + */ + if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq)) + return; + + enable_idle = old_idle = cfq_cfqq_idle_window(cfqq); - if (!cic->ioc->task || !cfqd->cfq_slice_idle || + if (!atomic_read(&cic->ioc->nr_tasks) || !cfqd->cfq_slice_idle || (cfqd->hw_tag && CIC_SEEKY(cic))) enable_idle = 0; else if (sample_valid(cic->ttime_samples)) { @@ -1554,39 +1751,60 @@ cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq, enable_idle = 1; } - if (enable_idle) - cfq_mark_cfqq_idle_window(cfqq); - else - cfq_clear_cfqq_idle_window(cfqq); + if (old_idle != enable_idle) { + cfq_log_cfqq(cfqd, cfqq, "idle=%d", enable_idle); + if (enable_idle) + cfq_mark_cfqq_idle_window(cfqq); + else + 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; + struct cfq_queue *cfqq; - if (cfq_class_idle(new_cfqq)) + cfqq = cfqd->active_queue; + if (!cfqq) return 0; - if (!cfqq) + if (cfq_slice_used(cfqq)) + return 1; + + if (cfq_class_idle(new_cfqq)) return 0; if (cfq_class_idle(cfqq)) return 1; - if (!cfq_cfqq_wait_request(new_cfqq)) - return 0; + + /* + * if the new request is sync, but the currently running queue is + * not, let the sync request have priority. + */ + if (rq_is_sync(rq) && !cfq_cfqq_sync(cfqq)) + return 1; + /* - * if it doesn't have slice left, forget it + * 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 (new_cfqq->slice_left < cfqd->cfq_slice_idle) + if (rq_is_meta(rq) && !cfqq->meta_pending) + return 1; + + if (!cfqd->active_cic || !cfq_cfqq_wait_request(cfqq)) return 0; - if (cfq_crq_is_sync(crq) && !cfq_cfqq_sync(cfqq)) + + /* + * if this request is as-good as one we would expect from the + * current cfqq, let it preempt + */ + if (cfq_rq_close(cfqd, rq)) return 1; return 0; @@ -1598,71 +1816,40 @@ cfq_should_preempt(struct cfq_data *cfqd, struct cfq_queue *new_cfqq, */ 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; - - cfqq->slice_end = cfqq->slice_left + jiffies; + cfq_log_cfqq(cfqd, cfqq, "preempt"); 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)); - if (!blk_queue_plugged(q)) - q->request_fn(q); - else - __generic_unplug_device(q); + cfq_service_tree_add(cfqd, cfqq, 1); + + 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; - - /* - * 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); + struct cfq_io_context *cic = RQ_CIC(rq); - /* - * we never wait for an async request and we don't allow preemption - * of an async request. so just return early - */ - if (!cfq_crq_is_sync(crq)) { - /* - * 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); - } - return; - } + cfqd->rq_queued++; + if (rq_is_meta(rq)) + cfqq->meta_pending++; cfq_update_io_thinktime(cfqd, cic); - cfq_update_io_seektime(cfqd, cic, crq); + cfq_update_io_seektime(cfqd, 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) { /* @@ -1673,9 +1860,9 @@ cfq_crq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq, 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 @@ -1683,68 +1870,93 @@ cfq_crq_enqueued(struct cfq_data *cfqd, struct cfq_queue *cfqq, */ 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) +static void cfq_insert_request(struct request_queue *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_log_cfqq(cfqd, cfqq, "insert_request"); + cfq_init_prio_data(cfqq, RQ_CIC(rq)->ioc); - 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); +} + +/* + * Update hw_tag based on peak queue depth over 50 samples under + * sufficient load. + */ +static void cfq_update_hw_tag(struct cfq_data *cfqd) +{ + if (cfqd->rq_in_driver > cfqd->rq_in_driver_peak) + cfqd->rq_in_driver_peak = cfqd->rq_in_driver; + + if (cfqd->rq_queued <= CFQ_HW_QUEUE_MIN && + cfqd->rq_in_driver <= CFQ_HW_QUEUE_MIN) + return; + + if (cfqd->hw_tag_samples++ < 50) + return; + + if (cfqd->rq_in_driver_peak >= CFQ_HW_QUEUE_MIN) + cfqd->hw_tag = 1; + else + cfqd->hw_tag = 0; + + cfqd->hw_tag_samples = 0; + cfqd->rq_in_driver_peak = 0; } -static void cfq_completed_request(request_queue_t *q, struct request *rq) +static void cfq_completed_request(struct request_queue *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 = cfq_crq_is_sync(crq); + const int sync = rq_is_sync(rq); unsigned long now; now = jiffies; + cfq_log_cfqq(cfqd, cfqq, "complete"); + + cfq_update_hw_tag(cfqd); WARN_ON(!cfqd->rq_in_driver); - WARN_ON(!cfqq->on_dispatch[sync]); + WARN_ON(!cfqq->dispatched); cfqd->rq_in_driver--; - cfqq->on_dispatch[sync]--; + cfqq->dispatched--; + + if (cfq_cfqq_sync(cfqq)) + cfqd->sync_flight--; 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); - } - } - 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); - else if (sync && RB_EMPTY_ROOT(&cfqq->sort_list)) { - if (!cfq_arm_slice_timer(cfqd, cfqq)) - cfq_schedule_dispatch(cfqd); + if (cfq_cfqq_slice_new(cfqq)) { + cfq_set_prio_slice(cfqd, cfqq); + cfq_clear_cfqq_slice_new(cfqq); } + if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq)) + cfq_slice_expired(cfqd, 1); + else if (sync && RB_EMPTY_ROOT(&cfqq->sort_list)) + cfq_arm_slice_timer(cfqd); } + + if (!cfqd->rq_in_driver) + cfq_schedule_dispatch(cfqd); } /* @@ -1753,9 +1965,6 @@ static void cfq_completed_request(request_queue_t *q, struct request *rq) */ static void cfq_prio_boost(struct cfq_queue *cfqq) { - const int ioprio_class = cfqq->ioprio_class; - const int ioprio = cfqq->ioprio; - if (has_fs_excl()) { /* * boost idle prio on transactions that would lock out other @@ -1774,18 +1983,9 @@ static void cfq_prio_boost(struct cfq_queue *cfqq) if (cfqq->ioprio != cfqq->org_ioprio) cfqq->ioprio = cfqq->org_ioprio; } - - /* - * 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)) - 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)) { @@ -1796,10 +1996,11 @@ __cfq_may_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq, return ELV_MQUEUE_MAY; } -static int cfq_may_queue(request_queue_t *q, int rw, struct bio *bio) +static int cfq_may_queue(struct request_queue *q, int rw) { struct cfq_data *cfqd = q->elevator->elevator_data; struct task_struct *tsk = current; + struct cfq_io_context *cic; struct cfq_queue *cfqq; /* @@ -1808,53 +2009,39 @@ static int cfq_may_queue(request_queue_t *q, int rw, struct bio *bio) * 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); + cic = cfq_cic_lookup(cfqd, tsk->io_context); + if (!cic) + return ELV_MQUEUE_MAY; + + cfqq = cic_to_cfqq(cic, rw & REQ_RW_SYNC); if (cfqq) { - cfq_init_prio_data(cfqq); + cfq_init_prio_data(cfqq, cic->ioc); 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); } } @@ -1863,18 +2050,14 @@ static void cfq_put_request(request_queue_t *q, struct request *rq) * 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(struct request_queue *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); 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); @@ -1885,79 +2068,45 @@ cfq_set_request(request_queue_t *q, struct request *rq, struct bio *bio, if (!cic) goto queue_fail; - if (!cic->cfqq[is_sync]) { - cfqq = cfq_get_queue(cfqd, key, tsk, gfp_mask); + cfqq = cic_to_cfqq(cic, is_sync); + if (!cfqq) { + cfqq = cfq_get_queue(cfqd, is_sync, cic->ioc, gfp_mask); + if (!cfqq) goto queue_fail; - cic->cfqq[is_sync] = cfqq; - } else - cfqq = cic->cfqq[is_sync]; + cic_set_cfqq(cic, cfqq, is_sync); + } 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; - - if (is_sync) - cfq_mark_crq_is_sync(crq); - else - cfq_clear_crq_is_sync(crq); + 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); + cfq_log(cfqd, "set_request fail"); 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); + struct request_queue *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); } @@ -1969,16 +2118,20 @@ static void cfq_idle_slice_timer(unsigned long data) struct cfq_data *cfqd = (struct cfq_data *) data; struct cfq_queue *cfqq; unsigned long flags; + int timed_out = 1; + + cfq_log(cfqd, "idle timer fired"); spin_lock_irqsave(cfqd->queue->queue_lock, flags); - if ((cfqq = cfqd->active_queue) != NULL) { - unsigned long now = jiffies; + cfqq = cfqd->active_queue; + if (cfqq) { + timed_out = 0; /* * expired */ - if (time_after(now, cfqq->slice_end)) + if (cfq_slice_used(cfqq)) goto expire; /* @@ -1997,50 +2150,41 @@ static void cfq_idle_slice_timer(unsigned long data) } } expire: - cfq_slice_expired(cfqd, 0); + cfq_slice_expired(cfqd, timed_out); out_kick: cfq_schedule_dispatch(cfqd); 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) +static void cfq_shutdown_timer_wq(struct cfq_data *cfqd) { - 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; - if (!time_after_eq(jiffies, end)) - mod_timer(&cfqd->idle_class_timer, end); - else - cfq_schedule_dispatch(cfqd); - - spin_unlock_irqrestore(cfqd->queue->queue_lock, flags); + del_timer_sync(&cfqd->idle_slice_timer); + kblockd_flush_work(&cfqd->unplug_work); } -static void cfq_shutdown_timer_wq(struct cfq_data *cfqd) +static void cfq_put_async_queues(struct cfq_data *cfqd) { - del_timer_sync(&cfqd->idle_slice_timer); - del_timer_sync(&cfqd->idle_class_timer); - blk_sync_queue(cfqd->queue); + int i; + + for (i = 0; i < IOPRIO_BE_NR; i++) { + if (cfqd->async_cfqq[0][i]) + cfq_put_queue(cfqd->async_cfqq[0][i]); + if (cfqd->async_cfqq[1][i]) + cfq_put_queue(cfqd->async_cfqq[1][i]); + } + + if (cfqd->async_idle_cfqq) + cfq_put_queue(cfqd->async_idle_cfqq); } static void cfq_exit_queue(elevator_t *e) { struct cfq_data *cfqd = e->elevator_data; - request_queue_t *q = cfqd->queue; + struct request_queue *q = cfqd->queue; cfq_shutdown_timer_wq(cfqd); - spin_lock(&cfq_exit_lock); spin_lock_irq(q->queue_lock); if (cfqd->active_queue) @@ -2050,72 +2194,39 @@ static void cfq_exit_queue(elevator_t *e) 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); } + cfq_put_async_queues(cfqd); + 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(struct request_queue *q) { struct cfq_data *cfqd; - int i; - cfqd = kmalloc(sizeof(*cfqd), GFP_KERNEL); + cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node); if (!cfqd) return NULL; - memset(cfqd, 0, sizeof(*cfqd)); - - 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); - INIT_LIST_HEAD(&cfqd->empty_list); + cfqd->service_tree = CFQ_RB_ROOT; INIT_LIST_HEAD(&cfqd->cic_list); - cfqd->cfq_hash = kmalloc(sizeof(struct hlist_head) * CFQ_QHASH_ENTRIES, GFP_KERNEL); - 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; - - for (i = 0; i < CFQ_QHASH_ENTRIES; i++) - INIT_HLIST_HEAD(&cfqd->cfq_hash[i]); - cfqd->queue = q; 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); - INIT_WORK(&cfqd->unplug_work, cfq_kick_queue, q); - - cfqd->cfq_queued = cfq_queued; + cfqd->last_end_request = jiffies; cfqd->cfq_quantum = cfq_quantum; cfqd->cfq_fifo_expire[0] = cfq_fifo_expire[0]; cfqd->cfq_fifo_expire[1] = cfq_fifo_expire[1]; @@ -2125,19 +2236,17 @@ static void *cfq_init_queue(request_queue_t *q, elevator_t *e) cfqd->cfq_slice[1] = cfq_slice_sync; cfqd->cfq_slice_async_rq = cfq_slice_async_rq; cfqd->cfq_slice_idle = cfq_slice_idle; + cfqd->hw_tag = 1; return cfqd; -out_crqpool: - kfree(cfqd->cfq_hash); -out_crqhash: - kfree(cfqd); - return NULL; } static void cfq_slab_kill(void) { - if (crq_pool) - kmem_cache_destroy(crq_pool); + /* + * Caller already ensured that pending RCU callbacks are completed, + * so we should have no busy allocations at this point. + */ if (cfq_pool) kmem_cache_destroy(cfq_pool); if (cfq_ioc_pool) @@ -2146,18 +2255,11 @@ static void cfq_slab_kill(void) 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); + cfq_pool = KMEM_CACHE(cfq_queue, 0); if (!cfq_pool) goto fail; - cfq_ioc_pool = kmem_cache_create("cfq_ioc_pool", - sizeof(struct cfq_io_context), 0, 0, NULL, NULL); + cfq_ioc_pool = KMEM_CACHE(cfq_io_context, 0); if (!cfq_ioc_pool) goto fail; @@ -2170,7 +2272,6 @@ fail: /* * sysfs parts below --> */ - static ssize_t cfq_var_show(unsigned int var, char *page) { @@ -2196,7 +2297,6 @@ static ssize_t __FUNC(elevator_t *e, char *page) \ 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); @@ -2224,15 +2324,18 @@ static ssize_t __FUNC(elevator_t *e, const char *page, size_t count) \ 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_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); -STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1, UINT_MAX, 0); +STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1, + UINT_MAX, 0); 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); +STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1, + UINT_MAX, 0); #undef STORE_FUNCTION #define CFQ_ATTR(name) \ @@ -2240,7 +2343,6 @@ STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1, UINT_MAX, 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), @@ -2257,6 +2359,7 @@ static struct elevator_type iosched_cfq = { .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, @@ -2270,7 +2373,7 @@ static struct elevator_type iosched_cfq = { .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", @@ -2279,8 +2382,6 @@ static struct elevator_type iosched_cfq = { static int __init cfq_init(void) { - int ret; - /* * could be 0 on HZ < 1000 setups */ @@ -2292,23 +2393,25 @@ static int __init cfq_init(void) if (cfq_slab_setup()) return -ENOMEM; - ret = elv_register(&iosched_cfq); - if (ret) - cfq_slab_kill(); + elv_register(&iosched_cfq); - return ret; + return 0; } 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)) - wait_for_completion(ioc_gone); - synchronize_rcu(); + + /* + * this also protects us from entering cfq_slab_kill() with + * pending RCU callbacks + */ + if (elv_ioc_count_read(ioc_count)) + wait_for_completion(&all_gone); cfq_slab_kill(); }