===== arch/i386/kernel/entry.S 1.88 vs edited =====
--- 1.88/arch/i386/kernel/entry.S	2004-12-12 19:37:55 +01:00
+++ edited/arch/i386/kernel/entry.S	2004-12-13 18:40:46 +01:00
@@ -865,5 +865,7 @@ ENTRY(sys_call_table)
 	.long sys_add_key
 	.long sys_request_key
 	.long sys_keyctl
+	.long sys_ioprio_set
+	.long sys_ioprio_get		/* 290 */
 
 syscall_table_size=(.-sys_call_table)
===== arch/ia64/kernel/entry.S 1.69 vs edited =====
--- 1.69/arch/ia64/kernel/entry.S	2004-11-23 21:18:30 +01:00
+++ edited/arch/ia64/kernel/entry.S	2004-12-13 18:45:57 +01:00
@@ -1531,8 +1531,8 @@ sys_call_table:
 	data8 sys_add_key
 	data8 sys_request_key
 	data8 sys_keyctl
-	data8 sys_ni_syscall
-	data8 sys_ni_syscall			// 1275
+	data8 sys_ioprio_set
+	data8 sys_ioprio_get			// 1275
 	data8 sys_ni_syscall
 	data8 sys_ni_syscall
 	data8 sys_ni_syscall
===== arch/ppc/kernel/misc.S 1.63 vs edited =====
--- 1.63/arch/ppc/kernel/misc.S	2004-10-22 11:27:40 +02:00
+++ edited/arch/ppc/kernel/misc.S	2004-12-13 18:41:48 +01:00
@@ -1450,3 +1450,5 @@ _GLOBAL(sys_call_table)
 	.long sys_add_key
 	.long sys_request_key		/* 270 */
 	.long sys_keyctl
+	.long sys_ioprio_set
+	.long sys_ioprio_get
===== drivers/block/cfq-iosched.c 1.16 vs edited =====
--- 1.16/drivers/block/cfq-iosched.c	2004-12-07 10:00:14 +01:00
+++ edited/drivers/block/cfq-iosched.c	2004-12-21 15:36:11 +01:00
@@ -21,22 +21,32 @@
 #include <linux/hash.h>
 #include <linux/rbtree.h>
 #include <linux/mempool.h>
-
-static unsigned long max_elapsed_crq;
-static unsigned long max_elapsed_dispatch;
+#include <linux/ioprio.h>
 
 /*
  * tunables
  */
 static int cfq_quantum = 4;		/* max queue in one round of service */
 static int cfq_queued = 8;		/* minimum rq allocate limit per-queue*/
-static int cfq_service = HZ;		/* period over which service is avg */
 static int cfq_fifo_expire_r = HZ / 2;	/* fifo timeout for sync requests */
 static int cfq_fifo_expire_w = 5 * HZ;	/* fifo timeout for async requests */
 static int cfq_fifo_rate = HZ / 8;	/* fifo expiry rate */
 static int cfq_back_max = 16 * 1024;	/* maximum backwards seek, in KiB */
 static int cfq_back_penalty = 2;	/* penalty of a backwards seek */
 
+static int cfq_slice_sync = HZ / 45;
+static int cfq_slice_async = HZ / 175;
+static int cfq_slice_async_rq = 16;
+static int cfq_slice_idle = HZ / 100;
+
+#define CFQ_MAX_SEEK_MEAN	((sector_t) -1)
+#define CFQ_IDLE_GRACE		(HZ / 4)
+
+/*
+ * disable queueing at the driver/hardware level
+ */
+static int cfq_max_depth = 1;
+
 /*
  * for the hash of cfqq inside the cfqd
  */
@@ -55,6 +65,7 @@ static int cfq_back_penalty = 2;	/* pena
 #define list_entry_hash(ptr)	hlist_entry((ptr), struct cfq_rq, hash)
 
 #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 RQ_DATA(rq)		(rq)->elevator_private
 
@@ -76,55 +87,75 @@ static int cfq_back_penalty = 2;	/* pena
 #define rq_rb_key(rq)		(rq)->sector
 
 /*
- * threshold for switching off non-tag accounting
- */
-#define CFQ_MAX_TAG		(4)
-
-/*
  * sort key types and names
  */
 enum {
 	CFQ_KEY_PGID,
 	CFQ_KEY_TGID,
+	CFQ_KEY_PID,
 	CFQ_KEY_UID,
 	CFQ_KEY_GID,
 	CFQ_KEY_LAST,
 };
 
-static char *cfq_key_types[] = { "pgid", "tgid", "uid", "gid", NULL };
-
-/*
- * spare queue
- */
-#define CFQ_KEY_SPARE		(~0UL)
+static char *cfq_key_types[] = { "pgid", "tgid", "pid", "uid", "gid", NULL };
 
 static kmem_cache_t *crq_pool;
 static kmem_cache_t *cfq_pool;
 static kmem_cache_t *cfq_ioc_pool;
 
+#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)
+
+/*
+ * Per block device queue structure
+ */
 struct cfq_data {
-	struct list_head rr_list;
+	atomic_t ref;
+	request_queue_t *queue;
+
+	/*
+	 * rr list of queues with requests and the count of them
+	 */
+	struct list_head rr_list[CFQ_PRIO_LISTS];
+	struct list_head cur_rr;
+	struct list_head idle_rr;
+	unsigned int busy_queues;
+
 	struct list_head empty_list;
 
 	struct hlist_head *cfq_hash;
 	struct hlist_head *crq_hash;
 
-	/* queues on rr_list (ie they have pending requests */
-	unsigned int busy_queues;
-
 	unsigned int max_queued;
 
-	atomic_t ref;
-
 	int key_type;
 
 	mempool_t *crq_pool;
 
-	request_queue_t *queue;
+	int rq_in_driver;
 
-	sector_t last_sector;
+	/*
+	 * schedule slice state info
+	 */
+	/*
+	 * idle window management
+	 */
+	struct timer_list idle_slice_timer;
+	struct work_struct unplug_work;
+	unsigned long idle_start;
 
-	int rq_in_driver;
+	struct cfq_queue *active_queue;
+	struct cfq_io_context *active_cic;
+	int cur_prio;
+	unsigned int dispatch_slice;
+
+	struct timer_list idle_class_timer;
+
+	sector_t last_sector;
+	unsigned long last_end_request;
 
 	/*
 	 * tunables, see top of file
@@ -136,11 +167,15 @@ struct cfq_data {
 	unsigned int cfq_fifo_batch_expire;
 	unsigned int cfq_back_penalty;
 	unsigned int cfq_back_max;
-	unsigned int find_best_crq;
-
-	unsigned int cfq_tagged;
+	unsigned int cfq_slice[2];
+	unsigned int cfq_slice_async_rq;
+	unsigned int cfq_slice_idle;
+	unsigned int cfq_max_depth;
 };
 
+/*
+ * Per process-grouping structure
+ */
 struct cfq_queue {
 	/* reference count */
 	atomic_t ref;
@@ -150,8 +185,6 @@ struct cfq_queue {
 	struct hlist_node cfq_hash;
 	/* hash key */
 	unsigned long key;
-	/* whether queue is on rr (or empty) list */
-	int on_rr;
 	/* on either rr or empty list of cfqd */
 	struct list_head cfq_list;
 	/* sorted list of pending requests */
@@ -169,15 +202,34 @@ struct cfq_queue {
 
 	int key_type;
 
-	unsigned long service_start;
-	unsigned long service_used;
-
-	unsigned int max_rate;
+	unsigned long slice_start;
+	unsigned long slice_end;
+	unsigned long slice_left;
+	unsigned long service_last;
 
 	/* number of requests that have been handed to the driver */
 	int in_flight;
-	/* number of currently allocated requests */
-	int alloc_limit[2];
+
+	/* io prio of this group */
+	int ioprio;
+	int org_ioprio;
+	short ioprio_class;
+	short org_ioprio_class;
+
+	unsigned int idle_hit, idle_miss;
+
+	/* whether queue is on rr (or empty) list */
+	unsigned int on_rr : 1;
+	/* idle slice, waiting for new request submission */
+	unsigned int wait_request : 1;
+	/* set when wait_request gets set, reset on first rq alloc */
+	unsigned int must_alloc : 1;
+	/* idle slice, request added, now waiting to dispatch it */
+	unsigned int must_dispatch : 1;
+
+	unsigned int idle_window : 1;
+	unsigned int doing_sync : 1;
+	unsigned int prio_changed : 1;
 };
 
 struct cfq_rq {
@@ -189,19 +241,19 @@ struct cfq_rq {
 	struct cfq_queue *cfq_queue;
 	struct cfq_io_context *io_context;
 
-	unsigned long service_start;
-	unsigned long queue_start;
+	sector_t end_pos;
 
 	unsigned int in_flight : 1;
 	unsigned int accounted : 1;
 	unsigned int is_sync   : 1;
-	unsigned int is_write  : 1;
+	unsigned int requeued  : 1;
 };
 
 static struct cfq_queue *cfq_find_cfq_hash(struct cfq_data *, unsigned long);
 static void cfq_dispatch_sort(request_queue_t *, struct cfq_rq *);
 static void cfq_update_next_crq(struct cfq_rq *);
 static void cfq_put_cfqd(struct cfq_data *cfqd);
+static struct cfq_io_context *cfq_get_io_context(struct cfq_queue **, int);
 
 /*
  * what the fairness is based on (ie how processes are grouped and
@@ -219,6 +271,8 @@ cfq_hash_key(struct cfq_data *cfqd, stru
 		default:
 		case CFQ_KEY_TGID:
 			return tsk->tgid;
+		case CFQ_KEY_PID:
+			return tsk->pid;
 		case CFQ_KEY_UID:
 			return tsk->uid;
 		case CFQ_KEY_GID:
@@ -292,6 +346,10 @@ cfq_choose_req(struct cfq_data *cfqd, st
 		return crq2;
 	if (crq2 == NULL)
 		return crq1;
+	if (crq1->requeued)
+		return crq1;
+	if (crq2->requeued)
+		return crq2;
 
 	s1 = crq1->request->sector;
 	s2 = crq2->request->sector;
@@ -406,67 +464,40 @@ static void cfq_update_next_crq(struct c
 		cfqq->next_crq = cfq_find_next_crq(cfqq->cfqd, cfqq, crq);
 }
 
-static int cfq_check_sort_rr_list(struct cfq_queue *cfqq)
+static void cfq_resort_rr_list(struct cfq_queue *cfqq, int preempted)
 {
-	struct list_head *head = &cfqq->cfqd->rr_list;
-	struct list_head *next, *prev;
+	struct list_head *list, *entry;
 
-	/*
-	 * list might still be ordered
-	 */
-	next = cfqq->cfq_list.next;
-	if (next != head) {
-		struct cfq_queue *cnext = list_entry_cfqq(next);
+	BUG_ON(!cfqq->on_rr);
 
-		if (cfqq->service_used > cnext->service_used)
-			return 1;
-	}
-
-	prev = cfqq->cfq_list.prev;
-	if (prev != head) {
-		struct cfq_queue *cprev = list_entry_cfqq(prev);
-
-		if (cfqq->service_used < cprev->service_used)
-			return 1;
-	}
-
-	return 0;
-}
+	list_del(&cfqq->cfq_list);
 
-static void cfq_sort_rr_list(struct cfq_queue *cfqq, int new_queue)
-{
-	struct list_head *entry = &cfqq->cfqd->rr_list;
+	if (cfq_class_rt(cfqq))
+		list = &cfqq->cfqd->cur_rr;
+	else if (cfq_class_be(cfqq))
+		list = &cfqq->cfqd->rr_list[cfqq->ioprio];
+	else
+		list = &cfqq->cfqd->idle_rr;
 
-	if (!cfqq->on_rr)
-		return;
-	if (!new_queue && !cfq_check_sort_rr_list(cfqq))
+	/*
+	 * if queue was preempted, just add to front to be fair
+	 */
+	if (unlikely(preempted)) {
+		list_add(&cfqq->cfq_list, list);
 		return;
-
-	list_del(&cfqq->cfq_list);
+	}
 
 	/*
-	 * sort by our mean service_used, sub-sort by in-flight requests
+	 * sort by when queue was last serviced
 	 */
-	while ((entry = entry->prev) != &cfqq->cfqd->rr_list) {
+	entry = list;
+	while ((entry = entry->prev) != list) {
 		struct cfq_queue *__cfqq = list_entry_cfqq(entry);
 
-		if (cfqq->service_used > __cfqq->service_used)
+		if (!__cfqq->service_last)
+			break;
+		if (time_before(__cfqq->service_last, cfqq->service_last))
 			break;
-		else if (cfqq->service_used == __cfqq->service_used) {
-			struct list_head *prv;
-
-			while ((prv = entry->prev) != &cfqq->cfqd->rr_list) {
-				__cfqq = list_entry_cfqq(prv);
-
-				WARN_ON(__cfqq->service_used > cfqq->service_used);
-				if (cfqq->service_used != __cfqq->service_used)
-					break;
-				if (cfqq->in_flight > __cfqq->in_flight)
-					break;
-
-				entry = prv;
-			}
-		}
 	}
 
 	list_add(&cfqq->cfq_list, entry);
@@ -474,21 +505,32 @@ static void cfq_sort_rr_list(struct cfq_
 
 /*
  * add to busy list of queues for service, trying to be fair in ordering
- * the pending list according to requests serviced
+ * the pending list according to last request service
  */
 static inline void
-cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq, int requeue)
 {
-	/*
-	 * it's currently on the empty list
-	 */
+	const int idle = cfq_class_idle(cfqq);
+
+	BUG_ON(cfqq->on_rr);
 	cfqq->on_rr = 1;
 	cfqd->busy_queues++;
 
-	if (time_after(jiffies, cfqq->service_start + cfq_service))
-		cfqq->service_used >>= 3;
+	/*
+	 * idle io should not end up on cur_rr and get premium service
+	 */
+	if (!requeue || idle)
+		cfq_resort_rr_list(cfqq, 0);
+	else {
+		list_del(&cfqq->cfq_list);
+		list_add_tail(&cfqq->cfq_list, &cfqd->cur_rr);
+	}
 
-	cfq_sort_rr_list(cfqq, 1);
+	/*
+	 * if this is higher prio, adjust cur_prio so we are served next
+	 */
+	if (cfq_class_be(cfqq) && cfqd->cur_prio > cfqq->ioprio)
+		cfqd->cur_prio = cfqq->ioprio;
 }
 
 static inline void
@@ -512,10 +554,10 @@ static inline void cfq_del_crq_rb(struct
 		struct cfq_data *cfqd = cfqq->cfqd;
 
 		BUG_ON(!cfqq->queued[crq->is_sync]);
+		cfqq->queued[crq->is_sync]--;
 
 		cfq_update_next_crq(crq);
 
-		cfqq->queued[crq->is_sync]--;
 		rb_erase(&crq->rb_node, &cfqq->sort_list);
 		RB_CLEAR_COLOR(&crq->rb_node);
 
@@ -567,7 +609,7 @@ static void cfq_add_crq_rb(struct cfq_rq
 	rb_insert_color(&crq->rb_node, &cfqq->sort_list);
 
 	if (!cfqq->on_rr)
-		cfq_add_cfqq_rr(cfqd, cfqq);
+		cfq_add_cfqq_rr(cfqd, cfqq, crq->requeued);
 
 	/*
 	 * check if this request is a better next-serve candidate
@@ -615,22 +657,33 @@ out:
 /*
  * make sure the service time gets corrected on reissue of this request
  */
+static void cfq_enqueue(struct cfq_data *cfqd, struct request *rq);
 static void cfq_requeue_request(request_queue_t *q, struct request *rq)
 {
+	struct cfq_data *cfqd = q->elevator->elevator_data;
 	struct cfq_rq *crq = RQ_DATA(rq);
 
 	if (crq) {
 		struct cfq_queue *cfqq = crq->cfq_queue;
 
-		if (cfqq->cfqd->cfq_tagged) {
-			cfqq->service_used--;
-			cfq_sort_rr_list(cfqq, 0);
+		if (crq->accounted) {
+			crq->accounted = 0;
+			WARN_ON(!cfqq->cfqd->rq_in_driver);
+			cfqq->cfqd->rq_in_driver--;
+		}
+		if (crq->in_flight) {
+			crq->in_flight = 0;
+			WARN_ON(!cfqq->in_flight);
+			cfqq->in_flight--;
 		}
-
-		crq->accounted = 0;
-		cfqq->cfqd->rq_in_driver--;
 	}
-	list_add(&rq->queuelist, &q->queue_head);
+
+	if (blk_fs_request(rq)) {
+		crq->cfq_queue->next_crq = crq;
+		crq->requeued = 1;
+		cfq_enqueue(cfqd, rq);
+	} else
+		list_add(&rq->queuelist, &q->queue_head);
 }
 
 static void cfq_remove_request(request_queue_t *q, struct request *rq)
@@ -640,9 +693,8 @@ static void cfq_remove_request(request_q
 	if (crq) {
 		cfq_remove_merge_hints(q, crq);
 		list_del_init(&rq->queuelist);
+		cfq_del_crq_rb(crq);
 
-		if (crq->cfq_queue)
-			cfq_del_crq_rb(crq);
 	}
 }
 
@@ -707,22 +759,155 @@ static void
 cfq_merged_requests(request_queue_t *q, struct request *rq,
 		    struct request *next)
 {
-	struct cfq_rq *crq = RQ_DATA(rq);
 	struct cfq_rq *cnext = RQ_DATA(next);
 
 	cfq_merged_request(q, rq);
 
-	if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist)) {
-		if (time_before(cnext->queue_start, crq->queue_start)) {
-			list_move(&rq->queuelist, &next->queuelist);
-			crq->queue_start = cnext->queue_start;
-		}
-	}
+	/*
+	 * reposition in fifo if next is older than rq
+	 */
+	if (!list_empty(&rq->queuelist) && !list_empty(&next->queuelist) &&
+	    time_before(next->start_time, rq->start_time))
+		list_move(&rq->queuelist, &next->queuelist);
 
 	cfq_update_next_crq(cnext);
 	cfq_remove_request(q, next);
 }
 
+static inline void
+__cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+{
+	if (cfqq) {
+		cfqq->slice_start = jiffies;
+		cfqq->slice_end = 0;
+		cfqq->slice_left = 0;
+		cfqd->cur_prio = cfqq->ioprio;
+		del_timer(&cfqd->idle_class_timer);
+	}
+
+	cfqd->active_queue = cfqq;
+}
+
+static void cfq_set_active_queue(struct cfq_data *cfqd)
+{
+	struct cfq_queue *cfqq = NULL;
+	int prio = cfqd->cur_prio;
+
+	do {
+		/*
+		 * to prevent starvation, we splice the prio list when we
+		 * first start serving that prio level.
+		 */
+		if (!list_empty(&cfqd->cur_rr)) {
+			cfqq = list_entry_cfqq(cfqd->cur_rr.next);
+			break;
+		} else if (!list_empty(&cfqd->rr_list[prio])) {
+			list_splice_init(&cfqd->rr_list[prio], &cfqd->cur_rr);
+			cfqq = list_entry_cfqq(cfqd->cur_rr.next);
+			break;
+		}
+
+		if (++prio == CFQ_PRIO_LISTS)
+			prio = 0;
+
+	} while (prio != cfqd->cur_prio);
+
+	/*
+	 * 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)) {
+		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 if (!timer_pending(&cfqd->idle_class_timer)) {
+			cfqd->idle_class_timer.expires = end;
+			add_timer(&cfqd->idle_class_timer);
+		}
+	}
+
+	__cfq_set_active_queue(cfqd, cfqq);
+}
+
+/*
+ * current cfqq expired its slice (or was too idle), select new one
+ */
+static inline void cfq_slice_expired(struct cfq_data *cfqd, int preempted)
+{
+	struct cfq_queue *cfqq = cfqd->active_queue;
+
+	if (cfqq) {
+		unsigned long now = jiffies;
+
+		if (cfqq->wait_request)
+			del_timer(&cfqd->idle_slice_timer);
+
+		if (!preempted)
+			cfqq->service_last = now;
+
+		cfqq->must_dispatch = 0;
+		cfqq->wait_request = 0;
+
+		/*
+		 * store what was left of this slice, if the queue idled out
+		 * or was preempted
+		 */
+		if (time_before(now, cfqq->slice_end))
+			cfqq->slice_left = now - cfqq->slice_end;
+		else
+			cfqq->slice_left = 0;
+
+		if (cfqq->on_rr)
+			cfq_resort_rr_list(cfqq, 1);
+
+		/*
+		 * check whether we should move the current prio service level.
+		 * we do so once we have served all queues at this level once
+		 */
+		if (list_empty(&cfqd->cur_rr)) {
+			if (++cfqd->cur_prio == CFQ_PRIO_LISTS)
+				cfqd->cur_prio = 0;
+		}
+
+		cfqd->active_queue = NULL;
+
+		if (cfqd->active_cic) {
+			put_io_context(cfqd->active_cic->ioc);
+			cfqd->active_cic = NULL;
+		}
+	}
+
+	cfqd->dispatch_slice = 0;
+}
+
+static int
+cfq_arm_slice_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+			
+{
+	WARN_ON(!RB_EMPTY(&cfqq->sort_list));
+	WARN_ON(cfqq != cfqd->active_queue);
+
+	if (!cfqd->cfq_slice_idle)
+		return 0;
+	if (!cfqq->idle_window)
+		return 0;
+
+	cfqq->wait_request = 1;
+	cfqq->must_alloc = 1;
+
+	if (!timer_pending(&cfqd->idle_slice_timer)) {
+		unsigned long slice_left = cfqq->slice_end - 1;
+
+		cfqd->idle_start = jiffies;
+		cfqd->idle_slice_timer.expires = min(jiffies + cfqd->cfq_slice_idle, slice_left);
+		add_timer(&cfqd->idle_slice_timer);
+	}
+
+	return 1;
+}
+
 /*
  * we dispatch cfqd->cfq_quantum requests in total from the rr_list queues,
  * this function sector sorts the selected request to minimize seeks. we start
@@ -736,31 +921,34 @@ static void cfq_dispatch_sort(request_qu
 	struct request *__rq;
 	sector_t last;
 
-	cfq_del_crq_rb(crq);
-	cfq_remove_merge_hints(q, crq);
 	list_del(&crq->request->queuelist);
 
 	last = cfqd->last_sector;
-	while ((entry = entry->prev) != head) {
-		__rq = list_entry_rq(entry);
-
-		if (blk_barrier_rq(crq->request))
+	list_for_each_entry_reverse(__rq, head, queuelist) {
+		if (blk_barrier_rq(__rq))
 			break;
-		if (!blk_fs_request(crq->request))
+		if (!blk_fs_request(__rq))
 			break;
 
-		if (crq->request->sector > __rq->sector)
+		if (__rq->sector <= crq->request->sector)
 			break;
 		if (__rq->sector > last && crq->request->sector < last) {
-			last = crq->request->sector;
+			last = crq->request->sector + crq->request->nr_sectors;
 			break;
 		}
+		entry = &__rq->queuelist;
 	}
 
 	cfqd->last_sector = last;
+
+	cfq_del_crq_rb(crq);
+	cfq_remove_merge_hints(q, crq);
+
+	crq->end_pos = crq->request->sector + crq->request->nr_sectors;
 	crq->in_flight = 1;
+	crq->requeued = 0;
 	cfqq->in_flight++;
-	list_add(&crq->request->queuelist, entry);
+	list_add_tail(&crq->request->queuelist, entry);
 }
 
 /*
@@ -772,102 +960,163 @@ static inline struct cfq_rq *cfq_check_f
 	const int reads = !list_empty(&cfqq->fifo[0]);
 	const int writes = !list_empty(&cfqq->fifo[1]);
 	unsigned long now = jiffies;
+	struct request *rq;
 	struct cfq_rq *crq;
 
 	if (time_before(now, cfqq->last_fifo_expire + cfqd->cfq_fifo_batch_expire))
 		return NULL;
 
-	crq = RQ_DATA(list_entry(cfqq->fifo[0].next, struct request, queuelist));
-	if (reads && time_after(now, crq->queue_start + cfqd->cfq_fifo_expire_r)) {
-		cfqq->last_fifo_expire = now;
-		return crq;
+	if (reads) {
+		crq = RQ_DATA(list_entry_fifo(cfqq->fifo[READ].next));
+		rq = crq->request;
+		if (time_after(now, rq->start_time + cfqd->cfq_fifo_expire_r)) {
+			cfqq->last_fifo_expire = now;
+			return crq;
+		}
 	}
 
-	crq = RQ_DATA(list_entry(cfqq->fifo[1].next, struct request, queuelist));
-	if (writes && time_after(now, crq->queue_start + cfqd->cfq_fifo_expire_w)) {
-		cfqq->last_fifo_expire = now;
-		return crq;
+	if (writes) {
+		crq = RQ_DATA(list_entry_fifo(cfqq->fifo[WRITE].next));
+		rq = crq->request;
+		if (time_after(now, rq->start_time + cfqd->cfq_fifo_expire_w)) {
+			cfqq->last_fifo_expire = now;
+			return crq;
+		}
 	}
 
 	return NULL;
 }
 
 /*
- * dispatch a single request from given queue
+ * Scale schedule slice based on io priority
  */
-static inline void
-cfq_dispatch_request(request_queue_t *q, struct cfq_data *cfqd,
-		     struct cfq_queue *cfqq)
+static inline int cfq_prio_to_slice(struct cfq_data *cfqd, int prio, int sync)
 {
-	struct cfq_rq *crq;
+	int base_slice = cfqd->cfq_slice[sync];
+
+	return base_slice + base_slice * (CFQ_PRIO_LISTS - 1 - prio);
+}
+
+/*
+ * get next queue for service
+ */
+static struct cfq_queue *cfq_select_queue(struct cfq_data *cfqd)
+{
+	unsigned long now = jiffies;
+	struct cfq_queue *cfqq;
+
+	cfqq = cfqd->active_queue;
+	if (!cfqq)
+		goto new_queue;
 
 	/*
-	 * follow expired path, else get first next available
+	 * slice has expired
 	 */
-	if ((crq = cfq_check_fifo(cfqq)) == NULL) {
-		if (cfqd->find_best_crq)
-			crq = cfqq->next_crq;
-		else
-			crq = rb_entry_crq(rb_first(&cfqq->sort_list));
-	}
-
-	cfqd->last_sector = crq->request->sector + crq->request->nr_sectors;
+	if (!cfqq->must_dispatch && time_after(jiffies, cfqq->slice_end))
+		goto new_queue;
 
 	/*
-	 * finally, insert request into driver list
+	 * if queue has requests, dispatch one. if not, check if
+	 * enough slice is left to wait for one
 	 */
-	cfq_dispatch_sort(q, crq);
+	if (!RB_EMPTY(&cfqq->sort_list))
+		goto keep_queue;
+	else if (time_before(now, cfqq->slice_end)) {
+		if (cfq_arm_slice_timer(cfqd, cfqq))
+			return NULL;
+	}
+
+new_queue:
+	cfq_slice_expired(cfqd, 0);
+	cfq_set_active_queue(cfqd);
+keep_queue:
+	return cfqd->active_queue;
 }
 
-static int cfq_dispatch_requests(request_queue_t *q, int max_dispatch)
+static int
+__cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
+			int max_dispatch)
 {
-	struct cfq_data *cfqd = q->elevator->elevator_data;
-	struct cfq_queue *cfqq;
-	struct list_head *entry, *tmp;
-	int queued, busy_queues, first_round;
+	int dispatched = 0, sync = 0;
 
-	if (list_empty(&cfqd->rr_list))
-		return 0;
+	BUG_ON(RB_EMPTY(&cfqq->sort_list));
 
-	queued = 0;
-	first_round = 1;
-restart:
-	busy_queues = 0;
-	list_for_each_safe(entry, tmp, &cfqd->rr_list) {
-		cfqq = list_entry_cfqq(entry);
+	do {
+		struct cfq_rq *crq;
 
-		BUG_ON(RB_EMPTY(&cfqq->sort_list));
+		/*
+		 * follow expired path, else get first next available
+		 */
+		if ((crq = cfq_check_fifo(cfqq)) == NULL)
+			crq = cfqq->next_crq;
 
 		/*
-		 * first round of queueing, only select from queues that
-		 * don't already have io in-flight
+		 * finally, insert request into driver list
 		 */
-		if (first_round && cfqq->in_flight)
-			continue;
+		cfq_dispatch_sort(cfqd->queue, crq);
+
+		cfqd->dispatch_slice++;
+		dispatched++;
+		sync += crq->is_sync;
+
+		if (!cfqd->active_cic) {
+			atomic_inc(&crq->io_context->ioc->refcount);
+			cfqd->active_cic = crq->io_context;
+		}
 
-		cfq_dispatch_request(q, cfqd, cfqq);
+		if (RB_EMPTY(&cfqq->sort_list))
+			break;
 
-		if (!RB_EMPTY(&cfqq->sort_list))
-			busy_queues++;
+	} while (dispatched < max_dispatch);
 
-		queued++;
-	}
+	/*
+	 * if slice end isn't set yet, set it. if at least one request was
+	 * sync, use the sync time slice value
+	 */
+	if (!cfqq->slice_end)
+		cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq->ioprio, !!sync) + jiffies;
 
-	if ((queued < max_dispatch) && (busy_queues || first_round)) {
-		first_round = 0;
-		goto restart;
+	/*
+	 * expire an async queue immediately if it has used up its slice. idle
+	 * queue always expire after 1 dispatch round.
+	 */
+	if ((!sync && cfqd->dispatch_slice >= cfqd->cfq_slice_async_rq) ||
+	    cfq_class_idle(cfqq))
+		cfq_slice_expired(cfqd, 0);
+
+	cfqq->doing_sync = !!sync;
+	return dispatched;
+}
+
+static int cfq_dispatch_requests(request_queue_t *q, int max_dispatch)
+{
+	struct cfq_data *cfqd = q->elevator->elevator_data;
+	struct cfq_queue *cfqq;
+
+	if (!cfqd->busy_queues)
+		return 0;
+
+	cfqq = cfq_select_queue(cfqd);
+	if (cfqq) {
+		cfqq->wait_request = 0;
+		cfqq->must_dispatch = 0;
+		del_timer(&cfqd->idle_slice_timer);
+
+		if (cfq_class_idle(cfqq))
+			max_dispatch = 1;
+
+		return __cfq_dispatch_requests(cfqd, cfqq, max_dispatch);
 	}
 
-	return queued;
+	return 0;
 }
 
 static inline void cfq_account_dispatch(struct cfq_rq *crq)
 {
 	struct cfq_queue *cfqq = crq->cfq_queue;
 	struct cfq_data *cfqd = cfqq->cfqd;
-	unsigned long now, elapsed;
 
-	if (!blk_fs_request(crq->request))
+	if (unlikely(!blk_fs_request(crq->request)))
 		return;
 
 	/*
@@ -877,37 +1126,8 @@ static inline void cfq_account_dispatch(
 	if (crq->accounted)
 		return;
 
-	now = jiffies;
-	if (cfqq->service_start == ~0UL)
-		cfqq->service_start = now;
-
-	/*
-	 * on drives with tagged command queueing, command turn-around time
-	 * doesn't necessarily reflect the time spent processing this very
-	 * command inside the drive. so do the accounting differently there,
-	 * by just sorting on the number of requests
-	 */
-	if (cfqd->cfq_tagged) {
-		if (time_after(now, cfqq->service_start + cfq_service)) {
-			cfqq->service_start = now;
-			cfqq->service_used /= 10;
-		}
-
-		cfqq->service_used++;
-		cfq_sort_rr_list(cfqq, 0);
-	}
-
-	elapsed = now - crq->queue_start;
-	if (elapsed > max_elapsed_dispatch)
-		max_elapsed_dispatch = elapsed;
-
 	crq->accounted = 1;
-	crq->service_start = now;
-
-	if (++cfqd->rq_in_driver >= CFQ_MAX_TAG && !cfqd->cfq_tagged) {
-		cfqq->cfqd->cfq_tagged = 1;
-		printk("cfq: depth %d reached, tagging now on\n", CFQ_MAX_TAG);
-	}
+	cfqd->rq_in_driver++;
 }
 
 static inline void
@@ -921,20 +1141,12 @@ cfq_account_completion(struct cfq_queue 
 	WARN_ON(!cfqd->rq_in_driver);
 	cfqd->rq_in_driver--;
 
-	if (!cfqd->cfq_tagged) {
-		unsigned long now = jiffies;
-		unsigned long duration = now - crq->service_start;
-
-		if (time_after(now, cfqq->service_start + cfq_service)) {
-			cfqq->service_start = now;
-			cfqq->service_used >>= 3;
-		}
-
-		cfqq->service_used += duration;
-		cfq_sort_rr_list(cfqq, 0);
+	if (!cfq_class_idle(cfqq))
+		cfqd->last_end_request = jiffies;
 
-		if (duration > max_elapsed_crq)
-			max_elapsed_crq = duration;
+	if (crq->is_sync) {
+		crq->io_context->last_end_request = jiffies;
+		crq->io_context->last_request_pos = crq->end_pos;
 	}
 }
 
@@ -948,7 +1160,15 @@ static struct request *cfq_next_request(
 dispatch:
 		rq = list_entry_rq(q->queue_head.next);
 
-		if ((crq = RQ_DATA(rq)) != NULL) {
+		crq = RQ_DATA(rq);
+		if (crq) {
+			/*
+			 * if idle window is disabled, allow queue buildup
+			 */
+			if (!crq->in_flight && !crq->cfq_queue->idle_window &&
+			    cfqd->rq_in_driver >= cfqd->cfq_max_depth)
+				return NULL;
+
 			cfq_remove_merge_hints(q, crq);
 			cfq_account_dispatch(crq);
 		}
@@ -970,6 +1190,8 @@ dispatch:
  */
 static void cfq_put_queue(struct cfq_queue *cfqq)
 {
+	struct cfq_data *cfqd = cfqq->cfqd;
+
 	BUG_ON(!atomic_read(&cfqq->ref));
 
 	if (!atomic_dec_and_test(&cfqq->ref))
@@ -978,6 +1200,9 @@ static void cfq_put_queue(struct cfq_que
 	BUG_ON(rb_first(&cfqq->sort_list));
 	BUG_ON(cfqq->on_rr);
 
+	if (unlikely(cfqd->active_queue == cfqq))
+		cfq_slice_expired(cfqd, 0);
+
 	cfq_put_cfqd(cfqq->cfqd);
 
 	/*
@@ -1044,55 +1269,95 @@ static void cfq_free_io_context(struct c
 	kmem_cache_free(cfq_ioc_pool, cic);
 }
 
+static void cfq_exit_single_io_context(struct cfq_io_context *cic)
+{
+	struct cfq_data *cfqd = cic->cfqq->cfqd;
+	request_queue_t *q = cfqd->queue;
+	unsigned long flags;
+
+	spin_lock_irqsave(q->queue_lock, flags);
+
+	if (cic->cfqq == cfqd->active_queue) {
+		cfq_slice_expired(cfqd, 0);
+
+		if (cfqd->busy_queues)
+			kblockd_schedule_work(&cfqd->unplug_work);
+	}
+
+	cfq_put_queue(cic->cfqq);
+	cic->cfqq = NULL;
+	spin_unlock_irqrestore(q->queue_lock, flags);
+	put_io_context(cic->ioc);
+}
+
 /*
- * locking hierarchy is: io_context lock -> queue locks
+ * Only the task itself ever browses or updates the cic list, so no locking
+ * is needed
  */
 static void cfq_exit_io_context(struct cfq_io_context *cic)
 {
-	struct cfq_queue *cfqq = cic->cfqq;
-	struct list_head *entry = &cic->list;
-	request_queue_t *q;
-	unsigned long flags;
+	struct cfq_io_context *__cic, *nxt;
 
 	/*
 	 * put the reference this task is holding to the various queues
 	 */
-	spin_lock_irqsave(&cic->ioc->lock, flags);
-	while ((entry = cic->list.next) != &cic->list) {
-		struct cfq_io_context *__cic;
-
-		__cic = list_entry(entry, struct cfq_io_context, list);
-		list_del(entry);
-
-		q = __cic->cfqq->cfqd->queue;
-		spin_lock(q->queue_lock);
-		cfq_put_queue(__cic->cfqq);
-		spin_unlock(q->queue_lock);
+	list_for_each_entry_safe(__cic, nxt, &cic->list, list) {
+		list_del(&__cic->list);
+		cfq_exit_single_io_context(__cic);
 	}
 
-	q = cfqq->cfqd->queue;
-	spin_lock(q->queue_lock);
-	cfq_put_queue(cfqq);
-	spin_unlock(q->queue_lock);
-
-	cic->cfqq = NULL;
-	spin_unlock_irqrestore(&cic->ioc->lock, flags);
+	cfq_exit_single_io_context(cic);
 }
 
-static struct cfq_io_context *cfq_alloc_io_context(int gfp_flags)
+static struct cfq_io_context *
+cfq_alloc_io_context(struct cfq_data *cfqd, int gfp_flags)
 {
 	struct cfq_io_context *cic = kmem_cache_alloc(cfq_ioc_pool, gfp_flags);
 
 	if (cic) {
-		cic->dtor = cfq_free_io_context;
-		cic->exit = cfq_exit_io_context;
 		INIT_LIST_HEAD(&cic->list);
 		cic->cfqq = NULL;
+		cic->last_end_request = jiffies;
+		cic->ttime_total = 0;
+		cic->ttime_samples = 0;
+		cic->ttime_mean = 0;
+		cic->seek_samples = 0;
+		cic->last_request_pos = cfqd->last_sector;
+		cic->seek_total = 0;
+		cic->seek_mean = 0;
+		cic->dtor = cfq_free_io_context;
+		cic->exit = cfq_exit_io_context;
 	}
 
 	return cic;
 }
 
+static inline void changed_ioprio(struct cfq_queue *cfqq)
+{
+	if (cfqq) {
+		struct cfq_data *cfqd = cfqq->cfqd;
+
+		spin_lock(cfqd->queue->queue_lock);
+		cfqq->prio_changed = 1;
+		spin_unlock(cfqd->queue->queue_lock);
+	}
+}
+
+/*
+ * callback from sys_ioprio_set, irqs are disabled
+ */
+static int cfq_ioc_set_ioprio(struct io_context *ioc, unsigned int ioprio)
+{
+	struct cfq_io_context *cic = ioc->cic;
+
+	changed_ioprio(cic->cfqq);
+
+	list_for_each_entry(cic, &cic->list, list)
+		changed_ioprio(cic->cfqq);
+
+	return 0;
+}
+
 /*
  * Setup general io context and cfq io context. There can be several cfq
  * io contexts per general io context, if this process is doing io to more
@@ -1114,18 +1379,25 @@ cfq_get_io_context(struct cfq_queue **cf
 		return NULL;
 
 	if ((cic = ioc->cic) == NULL) {
-		cic = cfq_alloc_io_context(gfp_flags);
+		cic = cfq_alloc_io_context(cfqd, gfp_flags);
 
 		if (cic == NULL)
 			goto err;
 
+		/*
+		 * manually increment generic io_context usage count, it
+		 * cannot go away since we are already holding one ref to it
+		 */
 		ioc->cic = cic;
+		ioc->set_ioprio = cfq_ioc_set_ioprio;
 		cic->ioc = ioc;
+		atomic_inc(&ioc->refcount);
+
 		cic->cfqq = __cfqq;
 		atomic_inc(&__cfqq->ref);
+		atomic_inc(&cfqd->ref);
 	} else {
 		struct cfq_io_context *__cic;
-		unsigned long flags;
 
 		/*
 		 * since the first cic on the list is actually the head
@@ -1140,35 +1412,37 @@ cfq_get_io_context(struct cfq_queue **cf
 		 * should be ok here, the list will usually not be more than
 		 * 1 or a few entries long
 		 */
-		spin_lock_irqsave(&ioc->lock, flags);
 		list_for_each_entry(__cic, &cic->list, list) {
 			/*
 			 * this process is already holding a reference to
 			 * this queue, so no need to get one more
 			 */
 			if (__cic->cfqq == __cfqq) {
+				/*
+				 * move hot entries to the front
+				 */
+				list_del(&__cic->list);
+				list_add(&__cic->list, &cic->list);
 				cic = __cic;
-				spin_unlock_irqrestore(&ioc->lock, flags);
 				goto out;
 			}
 		}
-		spin_unlock_irqrestore(&ioc->lock, flags);
 
 		/*
 		 * nope, process doesn't have a cic assoicated with this
 		 * cfqq yet. get a new one and add to list
 		 */
-		__cic = cfq_alloc_io_context(gfp_flags);
+		__cic = cfq_alloc_io_context(cfqd, gfp_flags);
 		if (__cic == NULL)
 			goto err;
 
 		__cic->ioc = ioc;
+		atomic_inc(&ioc->refcount);
+
 		__cic->cfqq = __cfqq;
 		atomic_inc(&__cfqq->ref);
-		spin_lock_irqsave(&ioc->lock, flags);
+		atomic_inc(&cfqd->ref);
 		list_add(&__cic->list, &cic->list);
-		spin_unlock_irqrestore(&ioc->lock, flags);
-
 		cic = __cic;
 		*cfqq = __cfqq;
 	}
@@ -1187,8 +1461,54 @@ err:
 	return NULL;
 }
 
+static void cfq_init_prio_data(struct cfq_queue *cfqq)
+{
+	int ioprio_class;
+
+	if (!cfqq->prio_changed)
+		return;
+
+	ioprio_class = IOPRIO_PRIO_CLASS(current->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_class = IOPRIO_CLASS_BE;
+			cfqq->ioprio = 4;
+			break;
+		case IOPRIO_CLASS_RT:
+			cfqq->ioprio_class = IOPRIO_CLASS_RT;
+			cfqq->ioprio = task_ioprio(current);
+			break;
+		case IOPRIO_CLASS_BE:
+			cfqq->ioprio_class = IOPRIO_CLASS_BE;
+			cfqq->ioprio = task_ioprio(current);
+			break;
+		case IOPRIO_CLASS_IDLE:
+			cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
+			cfqq->ioprio = 7;
+			cfqq->idle_window = 0;
+			break;
+	}
+
+	/*
+	 * keep track of original prio settings in case we have to temporarily
+	 * elevate the priority of this queue
+	 */
+	cfqq->org_ioprio = cfqq->ioprio;
+	cfqq->org_ioprio_class = cfqq->ioprio_class;
+
+	if (cfqq->on_rr)
+		cfq_resort_rr_list(cfqq, 0);
+
+	cfqq->prio_changed = 0;
+}
+
 static struct cfq_queue *
-__cfq_get_queue(struct cfq_data *cfqd, unsigned long key, int gfp_mask)
+cfq_get_queue(struct cfq_data *cfqd, unsigned long key, int gfp_mask)
 {
 	const int hashval = hash_long(key, CFQ_QHASH_SHIFT);
 	struct cfq_queue *cfqq, *new_cfqq = NULL;
@@ -1205,8 +1525,11 @@ retry:
 			new_cfqq = kmem_cache_alloc(cfq_pool, gfp_mask);
 			spin_lock_irq(cfqd->queue->queue_lock);
 			goto retry;
-		} else
-			goto out;
+		} else {
+			cfqq = kmem_cache_alloc(cfq_pool, gfp_mask);
+			if (!cfqq)
+				goto out;
+		}
 
 		memset(cfqq, 0, sizeof(*cfqq));
 
@@ -1222,7 +1545,15 @@ retry:
 		cfqq->cfqd = cfqd;
 		atomic_inc(&cfqd->ref);
 		cfqq->key_type = cfqd->key_type;
-		cfqq->service_start = ~0UL;
+		cfqq->service_last = 0;
+		/*
+		 * set ->slice_left to allow preemption for a new process
+		 */
+		cfqq->slice_left = 2 * cfqd->cfq_slice_idle;
+		cfqq->idle_window = 1;
+		cfqq->ioprio = -1;
+		cfqq->ioprio_class = -1;
+		cfqq->prio_changed = 1;
 	}
 
 	if (new_cfqq)
@@ -1234,36 +1565,224 @@ out:
 	return cfqq;
 }
 
-static struct cfq_queue *
-cfq_get_queue(struct cfq_data *cfqd, unsigned long key, int gfp_mask)
+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);
+
+	cic->ttime_samples = (7*cic->ttime_samples + 256) / 8;
+	cic->ttime_total = (7*cic->ttime_total + 256*ttime) / 8;
+	cic->ttime_mean = (cic->ttime_total + 128) / cic->ttime_samples;
+}
+
+static void
+cfq_update_io_seektime(struct cfq_data *cfqd, struct cfq_rq *crq,
+		       struct cfq_io_context *cic)
+{
+	sector_t seek;
+	u64 total;
+
+	if (!cic->last_request_pos)
+		seek = 0;
+	else if (cic->last_request_pos < crq->request->sector)
+		seek = crq->request->sector - cic->last_request_pos;
+	else
+		seek = cic->last_request_pos - crq->request->sector;
+
+	/*
+	 * Don't allow the seek distance to get too large from the
+	 * odd fragment, pagein, etc
+	 */
+	if (cic->seek_samples <= 60) /* second&third seek */
+		seek = min(seek, (cic->seek_mean * 4) + 2*1024*1024);
+	else
+		seek = min(seek, (cic->seek_mean * 4)	+ 2*1024*64);
+
+	cic->seek_samples = (7*cic->seek_samples + 256) / 8;
+	cic->seek_total = (7*cic->seek_total + (u64)256*seek) / 8;
+	total = cic->seek_total + (cic->seek_samples/2);
+	do_div(total, cic->seek_samples);
+	cic->seek_mean = total;
+}
+
+#define sample_valid(samples)	((samples) > 80)
+
+/*
+ * Disable idle window if the process thinks too long or seeks so much that
+ * it doesn't matter
+ */
+static void
+cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
+		       struct cfq_io_context *cic)
+{
+	int enable_idle = 0;
+
+	if (!cic->ioc->task)
+		goto disable;
+
+	if (sample_valid(cic->seek_samples) &&
+	    cic->seek_mean > CFQ_MAX_SEEK_MEAN)
+		goto disable;
+
+	if (sample_valid(cic->ttime_samples) && 
+	    cic->ttime_mean > cfqd->cfq_slice_idle)
+		goto disable;
+
+	if (!cfqd->cfq_slice_idle)
+		goto disable;
+
+	enable_idle = 1;
+disable:
+	cfqq->idle_window = enable_idle;
+}
+
+
+/*
+ * 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 cfq_queue *cfqq = cfqd->active_queue;
+
+	if (!cfqq)
+		return 1;
+
+	if (cfq_class_idle(new_cfqq))
+		return 0;
+	if (cfq_class_idle(cfqq))
+		return 1;
+	if (!new_cfqq->wait_request)
+		return 0;
+
+	if (new_cfqq->ioprio > cfqq->ioprio)
+		return 0;
+	if (crq->is_sync && !cfqq->doing_sync)
+		return 1;
+	if (new_cfqq->ioprio == cfqq->ioprio)
+		return 0;
+
+	/*
+	 * if it doesn't have slice left, forget it
+	 */
+	if (new_cfqq->slice_left >= 2 * cfqd->cfq_slice_idle)
+		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->ioprio, 1)/2;
+	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;
-	struct cfq_queue *cfqq;
 
-	spin_lock_irq(q->queue_lock);
-	cfqq = __cfq_get_queue(cfqd, key, gfp_mask);
-	spin_unlock_irq(q->queue_lock);
+	if (!blk_queue_plugged(q))
+		q->request_fn(q);
+	else
+		__generic_unplug_device(q);
+}
 
-	return cfqq;
+/*
+ * Called when a new fs request (crq) 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)
+{
+
+	if (crq->is_sync) {
+		struct cfq_io_context *cic = crq->io_context;
+
+		cfq_update_io_thinktime(cfqd, cic);
+		cfq_update_io_seektime(cfqd, crq, cic);
+		cfq_update_idle_window(cfqd, cfqq, cic);
+
+		cic->last_queue = jiffies;
+	}
+
+	if (cfqq == cfqd->active_queue) {
+		/*
+		 * if we are waiting for a request for this queue, let it rip
+		 * immediately and flag that we must not expire this queue
+		 * just now
+		 */
+		if (cfqq->wait_request) {
+			cfqq->idle_hit++;
+			cfqq->must_dispatch = 1;
+			del_timer(&cfqd->idle_slice_timer);
+			cfq_start_queueing(cfqd, cfqq);
+		}
+	} else if (cfq_should_preempt(cfqd, cfqq, crq)) {
+		/*
+		 * not the active queue - expire current slice if it is
+		 * idle and has expired it's mean thinktime or this new queue
+		 * has some old slice time left and is of higher priority
+		 */
+		cfq_preempt_queue(cfqd, cfqq);
+		cfqq->must_dispatch = 1;
+		cfq_start_queueing(cfqd, cfqq);
+	}
 }
 
-static void cfq_enqueue(struct cfq_data *cfqd, struct cfq_rq *crq)
+static void cfq_enqueue(struct cfq_data *cfqd, struct request *rq)
 {
-	crq->is_sync = 0;
-	if (rq_data_dir(crq->request) == READ || current->flags & PF_SYNCWRITE)
-		crq->is_sync = 1;
+	struct cfq_rq *crq = RQ_DATA(rq);
+	struct cfq_queue *cfqq = crq->cfq_queue;
+
+	cfq_init_prio_data(cfqq);
 
 	cfq_add_crq_rb(crq);
-	crq->queue_start = jiffies;
 
-	list_add_tail(&crq->request->queuelist, &crq->cfq_queue->fifo[crq->is_sync]);
+	list_add_tail(&rq->queuelist, &cfqq->fifo[crq->is_sync]);
+
+	if (rq_mergeable(rq)) {
+		cfq_add_crq_hash(cfqd, crq);
+
+		if (!cfqd->queue->last_merge)
+			cfqd->queue->last_merge = rq;
+	}
+
+	cfq_crq_enqueued(cfqd, cfqq, crq);
 }
 
 static void
 cfq_insert_request(request_queue_t *q, struct request *rq, int where)
 {
 	struct cfq_data *cfqd = q->elevator->elevator_data;
-	struct cfq_rq *crq = RQ_DATA(rq);
 
 	switch (where) {
 		case ELEVATOR_INSERT_BACK:
@@ -1276,26 +1795,19 @@ cfq_insert_request(request_queue_t *q, s
 			break;
 		case ELEVATOR_INSERT_SORT:
 			BUG_ON(!blk_fs_request(rq));
-			cfq_enqueue(cfqd, crq);
+			cfq_enqueue(cfqd, rq);
 			break;
 		default:
 			printk("%s: bad insert point %d\n", __FUNCTION__,where);
 			return;
 	}
-
-	if (rq_mergeable(rq)) {
-		cfq_add_crq_hash(cfqd, crq);
-
-		if (!q->last_merge)
-			q->last_merge = rq;
-	}
 }
 
 static int cfq_queue_empty(request_queue_t *q)
 {
 	struct cfq_data *cfqd = q->elevator->elevator_data;
 
-	return list_empty(&q->queue_head) && list_empty(&cfqd->rr_list);
+	return list_empty(&q->queue_head) && !cfqd->busy_queues;
 }
 
 static void cfq_completed_request(request_queue_t *q, struct request *rq)
@@ -1340,50 +1852,77 @@ cfq_latter_request(request_queue_t *q, s
 	return NULL;
 }
 
+/*
+ * we temporarily boost lower priority queues if they are holding fs exclusive
+ * resources. they are boosted to normal prio (CLASS_BE/4)
+ */
+static void cfq_prio_boost(struct cfq_queue *cfqq)
+{
+	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
+		 * users of the filesystem
+		 */
+		if (cfq_class_idle(cfqq))
+			cfqq->ioprio_class = IOPRIO_CLASS_BE;
+		if (cfqq->ioprio > 4)
+			cfqq->ioprio = 4;
+	} else {
+		/*
+		 * check if we need to unboost the queue
+		 */
+		if (cfqq->ioprio_class != cfqq->org_ioprio_class)
+			cfqq->ioprio_class = cfqq->org_ioprio_class;
+		if (cfqq->ioprio != cfqq->org_ioprio)
+			cfqq->ioprio = cfqq->org_ioprio;
+	}
+
+	/*
+	 * refile between round-robin lists if we moved the priority class
+	 */
+	if ((ioprio_class != cfqq->ioprio_class || ioprio != cfqq->ioprio) &&
+	    cfqq->on_rr)
+		cfq_resort_rr_list(cfqq, 0);
+}
+
 static int cfq_may_queue(request_queue_t *q, int rw)
 {
 	struct cfq_data *cfqd = q->elevator->elevator_data;
 	struct cfq_queue *cfqq;
 	int ret = ELV_MQUEUE_MAY;
 
-	if (current->flags & PF_MEMALLOC)
-		return ELV_MQUEUE_MAY;
-
 	cfqq = cfq_find_cfq_hash(cfqd, cfq_hash_key(cfqd, current));
 	if (cfqq) {
-		int limit = cfqd->max_queued;
-
-		if (cfqq->allocated[rw] < cfqd->cfq_queued)
-			return ELV_MQUEUE_MUST;
-
-		if (cfqd->busy_queues)
-			limit = q->nr_requests / cfqd->busy_queues;
+		cfq_prio_boost(cfqq);
 
-		if (limit < cfqd->cfq_queued)
-			limit = cfqd->cfq_queued;
-		else if (limit > cfqd->max_queued)
-			limit = cfqd->max_queued;
-
-		if (cfqq->allocated[rw] >= limit) {
-			if (limit > cfqq->alloc_limit[rw])
-				cfqq->alloc_limit[rw] = limit;
+		if (!cfq_class_idle(cfqq))
+			ret = ELV_MQUEUE_MUST;
 
+		if (cfqq->must_alloc)
+			goto out;
+		if (cfqq->allocated[rw] < cfqd->cfq_queued)
+			goto out;
+		if (cfqq->allocated[rw] >= cfqd->max_queued)
 			ret = ELV_MQUEUE_NO;
-		}
 	}
 
+out:
 	return ret;
 }
 
 static void cfq_check_waiters(request_queue_t *q, struct cfq_queue *cfqq)
 {
 	struct request_list *rl = &q->rq;
-	const int write = waitqueue_active(&rl->wait[WRITE]);
-	const int read = waitqueue_active(&rl->wait[READ]);
+	const int writes = waitqueue_active(&rl->wait[WRITE]);
+	const int reads = waitqueue_active(&rl->wait[READ]);
+	struct cfq_data *cfqd = q->elevator->elevator_data;
 
-	if (read && cfqq->allocated[READ] < cfqq->alloc_limit[READ])
+	if (reads && cfqq->allocated[READ] == cfqd->max_queued)
 		wake_up(&rl->wait[READ]);
-	if (write && cfqq->allocated[WRITE] < cfqq->alloc_limit[WRITE])
+	if (writes && cfqq->allocated[WRITE] == cfqd->max_queued)
 		wake_up(&rl->wait[WRITE]);
 }
 
@@ -1397,15 +1936,15 @@ static void cfq_put_request(request_queu
 
 	if (crq) {
 		struct cfq_queue *cfqq = crq->cfq_queue;
+		const int rw = rq_data_dir(rq);
 
 		BUG_ON(q->last_merge == rq);
 		BUG_ON(!hlist_unhashed(&crq->hash));
 
-		if (crq->io_context)
-			put_io_context(crq->io_context->ioc);
+		BUG_ON(!cfqq->allocated[rw]);
+		cfqq->allocated[rw]--;
 
-		BUG_ON(!cfqq->allocated[crq->is_write]);
-		cfqq->allocated[crq->is_write]--;
+		put_io_context(crq->io_context->ioc);
 
 		mempool_free(crq, cfqd->crq_pool);
 		rq->elevator_private = NULL;
@@ -1422,9 +1961,9 @@ static void cfq_put_request(request_queu
 static int cfq_set_request(request_queue_t *q, struct request *rq, int gfp_mask)
 {
 	struct cfq_data *cfqd = q->elevator->elevator_data;
-	struct cfq_io_context *cic;
 	const int rw = rq_data_dir(rq);
 	struct cfq_queue *cfqq, *saved_cfqq;
+	struct cfq_io_context *cic;
 	struct cfq_rq *crq;
 	unsigned long flags;
 
@@ -1432,21 +1971,13 @@ static int cfq_set_request(request_queue
 
 	spin_lock_irqsave(q->queue_lock, flags);
 
-	cfqq = __cfq_get_queue(cfqd, cfq_hash_key(cfqd, current), gfp_mask);
-	if (!cfqq) {
-#if 0
-		cfqq = cfq_get_queue(cfqd, CFQ_KEY_SPARE, gfp_mask);
-		printk("%s: got spare queue\n", current->comm);
-#else
+	cfqq = cfq_get_queue(cfqd, cfq_hash_key(cfqd, current), gfp_mask);
+	if (!cfqq)
 		goto out_lock;
-#endif
-	}
 
 repeat:
-	if (cfqq->allocated[rw] >= cfqd->max_queued)
-		goto out_lock;
-
 	cfqq->allocated[rw]++;
+	cfqq->must_alloc = 0;
 	spin_unlock_irqrestore(q->queue_lock, flags);
 
 	/*
@@ -1461,6 +1992,7 @@ repeat:
 	 * repeat allocation checks on queue change
 	 */
 	if (unlikely(saved_cfqq != cfqq)) {
+		put_io_context(cic->ioc);
 		spin_lock_irqsave(q->queue_lock, flags);
 		saved_cfqq->allocated[rw]--;
 		goto repeat;
@@ -1474,11 +2006,10 @@ repeat:
 		INIT_HLIST_NODE(&crq->hash);
 		crq->cfq_queue = cfqq;
 		crq->io_context = cic;
-		crq->service_start = crq->queue_start = 0;
-		crq->in_flight = crq->accounted = crq->is_sync = 0;
-		crq->is_write = rw;
+		crq->in_flight = crq->accounted = 0;
+		crq->is_sync = (rw == READ || current->flags & PF_SYNCWRITE);
+		crq->requeued = 0;
 		rq->elevator_private = crq;
-		cfqq->alloc_limit[rw] = 0;
 		return 0;
 	}
 
@@ -1486,28 +2017,103 @@ repeat:
 err:
 	spin_lock_irqsave(q->queue_lock, flags);
 	cfqq->allocated[rw]--;
+	cfqq->must_alloc = 1;
 	cfq_put_queue(cfqq);
 out_lock:
 	spin_unlock_irqrestore(q->queue_lock, flags);
 	return 1;
 }
 
+static void cfq_kick_queue(void *data)
+{
+	request_queue_t *q = data;
+
+	blk_run_queue(q);
+}
+
+/*
+ * Timer running if the active_queue is currently idling inside its time slice
+ */
+static void cfq_idle_slice_timer(unsigned long data)
+{
+	struct cfq_data *cfqd = (struct cfq_data *) data;
+	struct cfq_queue *cfqq;
+	unsigned long flags;
+
+	spin_lock_irqsave(cfqd->queue->queue_lock, flags);
+
+	if ((cfqq = cfqd->active_queue) != NULL) {
+		/*
+		 * expired
+		 */
+		if (time_after(jiffies, cfqq->slice_end))
+			goto miss;
+
+		/*
+		 * only expire and reinvoke request handler, if there are
+		 * other queues with pending requests
+		 */
+		if (!cfqd->busy_queues) {
+			cfqd->idle_slice_timer.expires = min(jiffies + cfqd->cfq_slice_idle, cfqq->slice_end);
+			add_timer(&cfqd->idle_slice_timer);
+			goto out_cont;
+		}
+
+		/*
+		 * not expired and it has a request pending, let it dispatch
+		 */
+		if (!RB_EMPTY(&cfqq->sort_list)) {
+			cfqq->must_dispatch = 1;
+			goto out_kick;
+		}
+miss:
+		cfqq->idle_miss++;
+	}
+
+	cfq_slice_expired(cfqd, 0);
+out_kick:
+	if (cfqd->busy_queues)
+		kblockd_schedule_work(&cfqd->unplug_work);
+out_cont:
+	spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
+}
+
+/*
+ * Timer running if an idle class queue is waiting for service
+ */
+static void cfq_idle_class_timer(unsigned long data)
+{
+	struct cfq_data *cfqd = (struct cfq_data *) data;
+	unsigned long flags;
+
+	spin_lock_irqsave(cfqd->queue->queue_lock, flags);
+
+	/*
+	 * race with a non-idle queue, reset timer
+	 */
+	if (!time_after_eq(jiffies, cfqd->last_end_request + CFQ_IDLE_GRACE)) {
+		cfqd->idle_class_timer.expires = cfqd->last_end_request + CFQ_IDLE_GRACE;
+		add_timer(&cfqd->idle_class_timer);
+	} else
+		kblockd_schedule_work(&cfqd->unplug_work);
+
+	spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
+}
+
+
 static void cfq_put_cfqd(struct cfq_data *cfqd)
 {
 	request_queue_t *q = cfqd->queue;
-	struct cfq_queue *cfqq;
 
 	if (!atomic_dec_and_test(&cfqd->ref))
 		return;
 
+	blk_sync_queue(q);
+
 	/*
-	 * kill spare queue, getting it means we have two refences to it.
-	 * drop both
+	 * restore ->nr_requests
 	 */
-	spin_lock_irq(q->queue_lock);
-	cfqq = __cfq_get_queue(cfqd, CFQ_KEY_SPARE, GFP_ATOMIC);
-	cfq_put_queue(cfqq);
-	cfq_put_queue(cfqq);
+	q->nr_requests = BLKDEV_MAX_RQ;
 	spin_unlock_irq(q->queue_lock);
 
 	blk_put_queue(q);
@@ -1520,13 +2126,16 @@ static void cfq_put_cfqd(struct cfq_data
 
 static void cfq_exit_queue(elevator_t *e)
 {
-	cfq_put_cfqd(e->elevator_data);
+	struct cfq_data *cfqd = e->elevator_data;
+
+	del_timer_sync(&cfqd->idle_slice_timer);
+	del_timer_sync(&cfqd->idle_class_timer);
+	cfq_put_cfqd(cfqd);
 }
 
 static int cfq_init_queue(request_queue_t *q, elevator_t *e)
 {
 	struct cfq_data *cfqd;
-	struct cfq_queue *cfqq;
 	int i;
 
 	cfqd = kmalloc(sizeof(*cfqd), GFP_KERNEL);
@@ -1534,7 +2143,12 @@ static int cfq_init_queue(request_queue_
 		return -ENOMEM;
 
 	memset(cfqd, 0, sizeof(*cfqd));
-	INIT_LIST_HEAD(&cfqd->rr_list);
+
+	for (i = 0; i < CFQ_PRIO_LISTS; i++)
+		INIT_LIST_HEAD(&cfqd->rr_list[i]);
+
+	INIT_LIST_HEAD(&cfqd->cur_rr);
+	INIT_LIST_HEAD(&cfqd->idle_rr);
 	INIT_LIST_HEAD(&cfqd->empty_list);
 
 	cfqd->crq_hash = kmalloc(sizeof(struct hlist_head) * CFQ_MHASH_ENTRIES, GFP_KERNEL);
@@ -1559,22 +2173,20 @@ static int cfq_init_queue(request_queue_
 	cfqd->queue = q;
 	atomic_inc(&q->refcnt);
 
-	/*
-	 * setup spare failure queue
-	 */
-	cfqq = cfq_get_queue(cfqd, CFQ_KEY_SPARE, GFP_KERNEL);
-	if (!cfqq)
-		goto out_spare;
-
-	/*
-	 * just set it to some high value, we want anyone to be able to queue
-	 * some requests. fairness is handled differently
-	 */
-	q->nr_requests = 1024;
-	cfqd->max_queued = q->nr_requests / 16;
+	cfqd->max_queued = q->nr_requests / 4;
 	q->nr_batching = cfq_queued;
-	cfqd->key_type = CFQ_KEY_TGID;
-	cfqd->find_best_crq = 1;
+	cfqd->key_type = CFQ_KEY_PID;
+
+	init_timer(&cfqd->idle_slice_timer);
+	cfqd->idle_slice_timer.function = cfq_idle_slice_timer;
+	cfqd->idle_slice_timer.data = (unsigned long) cfqd;
+
+	init_timer(&cfqd->idle_class_timer);
+	cfqd->idle_class_timer.function = cfq_idle_class_timer;
+	cfqd->idle_class_timer.data = (unsigned long) cfqd;
+
+	INIT_WORK(&cfqd->unplug_work, cfq_kick_queue, q);
+
 	atomic_set(&cfqd->ref, 1);
 
 	cfqd->cfq_queued = cfq_queued;
@@ -1584,10 +2196,13 @@ static int cfq_init_queue(request_queue_
 	cfqd->cfq_fifo_batch_expire = cfq_fifo_rate;
 	cfqd->cfq_back_max = cfq_back_max;
 	cfqd->cfq_back_penalty = cfq_back_penalty;
+	cfqd->cfq_slice[0] = cfq_slice_async;
+	cfqd->cfq_slice[1] = cfq_slice_sync;
+	cfqd->cfq_slice_async_rq = cfq_slice_async_rq;
+	cfqd->cfq_slice_idle = cfq_slice_idle;
+	cfqd->cfq_max_depth = cfq_max_depth;
 
 	return 0;
-out_spare:
-	mempool_destroy(cfqd->crq_pool);
 out_crqpool:
 	kfree(cfqd->cfq_hash);
 out_cfqhash:
@@ -1630,7 +2245,6 @@ fail:
 	return -ENOMEM;
 }
 
-
 /*
  * sysfs parts below -->
  */
@@ -1656,13 +2270,6 @@ cfq_var_store(unsigned int *var, const c
 }
 
 static ssize_t
-cfq_clear_elapsed(struct cfq_data *cfqd, const char *page, size_t count)
-{
-	max_elapsed_dispatch = max_elapsed_crq = 0;
-	return count;
-}
-
-static ssize_t
 cfq_set_key_type(struct cfq_data *cfqd, const char *page, size_t count)
 {
 	spin_lock_irq(cfqd->queue->queue_lock);
@@ -1670,6 +2277,8 @@ cfq_set_key_type(struct cfq_data *cfqd, 
 		cfqd->key_type = CFQ_KEY_PGID;
 	else if (!strncmp(page, "tgid", 4))
 		cfqd->key_type = CFQ_KEY_TGID;
+	else if (!strncmp(page, "pid", 3))
+		cfqd->key_type = CFQ_KEY_PID;
 	else if (!strncmp(page, "uid", 3))
 		cfqd->key_type = CFQ_KEY_UID;
 	else if (!strncmp(page, "gid", 3))
@@ -1707,9 +2316,13 @@ SHOW_FUNCTION(cfq_queued_show, cfqd->cfq
 SHOW_FUNCTION(cfq_fifo_expire_r_show, cfqd->cfq_fifo_expire_r, 1);
 SHOW_FUNCTION(cfq_fifo_expire_w_show, cfqd->cfq_fifo_expire_w, 1);
 SHOW_FUNCTION(cfq_fifo_batch_expire_show, cfqd->cfq_fifo_batch_expire, 1);
-SHOW_FUNCTION(cfq_find_best_show, cfqd->find_best_crq, 0);
 SHOW_FUNCTION(cfq_back_max_show, cfqd->cfq_back_max, 0);
 SHOW_FUNCTION(cfq_back_penalty_show, cfqd->cfq_back_penalty, 0);
+SHOW_FUNCTION(cfq_slice_idle_show, cfqd->cfq_slice_idle, 1);
+SHOW_FUNCTION(cfq_slice_sync_show, cfqd->cfq_slice[1], 1);
+SHOW_FUNCTION(cfq_slice_async_show, cfqd->cfq_slice[0], 1);
+SHOW_FUNCTION(cfq_slice_async_rq_show, cfqd->cfq_slice_async_rq, 0);
+SHOW_FUNCTION(cfq_max_depth_show, cfqd->cfq_max_depth, 0);
 #undef SHOW_FUNCTION
 
 #define STORE_FUNCTION(__FUNC, __PTR, MIN, MAX, __CONV)			\
@@ -1732,9 +2345,13 @@ STORE_FUNCTION(cfq_queued_store, &cfqd->
 STORE_FUNCTION(cfq_fifo_expire_r_store, &cfqd->cfq_fifo_expire_r, 1, UINT_MAX, 1);
 STORE_FUNCTION(cfq_fifo_expire_w_store, &cfqd->cfq_fifo_expire_w, 1, UINT_MAX, 1);
 STORE_FUNCTION(cfq_fifo_batch_expire_store, &cfqd->cfq_fifo_batch_expire, 0, UINT_MAX, 1);
-STORE_FUNCTION(cfq_find_best_store, &cfqd->find_best_crq, 0, 1, 0);
 STORE_FUNCTION(cfq_back_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
 STORE_FUNCTION(cfq_back_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_max_depth_store, &cfqd->cfq_max_depth, 1, UINT_MAX, 0);
 #undef STORE_FUNCTION
 
 static struct cfq_fs_entry cfq_quantum_entry = {
@@ -1762,11 +2379,6 @@ static struct cfq_fs_entry cfq_fifo_batc
 	.show = cfq_fifo_batch_expire_show,
 	.store = cfq_fifo_batch_expire_store,
 };
-static struct cfq_fs_entry cfq_find_best_entry = {
-	.attr = {.name = "find_best_crq", .mode = S_IRUGO | S_IWUSR },
-	.show = cfq_find_best_show,
-	.store = cfq_find_best_store,
-};
 static struct cfq_fs_entry cfq_back_max_entry = {
 	.attr = {.name = "back_seek_max", .mode = S_IRUGO | S_IWUSR },
 	.show = cfq_back_max_show,
@@ -1777,15 +2389,36 @@ static struct cfq_fs_entry cfq_back_pena
 	.show = cfq_back_penalty_show,
 	.store = cfq_back_penalty_store,
 };
-static struct cfq_fs_entry cfq_clear_elapsed_entry = {
-	.attr = {.name = "clear_elapsed", .mode = S_IWUSR },
-	.store = cfq_clear_elapsed,
+static struct cfq_fs_entry cfq_slice_sync_entry = {
+	.attr = {.name = "slice_sync", .mode = S_IRUGO | S_IWUSR },
+	.show = cfq_slice_sync_show,
+	.store = cfq_slice_sync_store,
+};
+static struct cfq_fs_entry cfq_slice_async_entry = {
+	.attr = {.name = "slice_async", .mode = S_IRUGO | S_IWUSR },
+	.show = cfq_slice_async_show,
+	.store = cfq_slice_async_store,
+};
+static struct cfq_fs_entry cfq_slice_async_rq_entry = {
+	.attr = {.name = "slice_async_rq", .mode = S_IRUGO | S_IWUSR },
+	.show = cfq_slice_async_rq_show,
+	.store = cfq_slice_async_rq_store,
+};
+static struct cfq_fs_entry cfq_slice_idle_entry = {
+	.attr = {.name = "slice_idle", .mode = S_IRUGO | S_IWUSR },
+	.show = cfq_slice_idle_show,
+	.store = cfq_slice_idle_store,
 };
 static struct cfq_fs_entry cfq_key_type_entry = {
 	.attr = {.name = "key_type", .mode = S_IRUGO | S_IWUSR },
 	.show = cfq_read_key_type,
 	.store = cfq_set_key_type,
 };
+static struct cfq_fs_entry cfq_max_depth_entry = {
+	.attr = {.name = "max_depth", .mode = S_IRUGO | S_IWUSR },
+	.show = cfq_max_depth_show,
+	.store = cfq_max_depth_store,
+};
 
 static struct attribute *default_attrs[] = {
 	&cfq_quantum_entry.attr,
@@ -1794,10 +2427,13 @@ static struct attribute *default_attrs[]
 	&cfq_fifo_expire_w_entry.attr,
 	&cfq_fifo_batch_expire_entry.attr,
 	&cfq_key_type_entry.attr,
-	&cfq_find_best_entry.attr,
 	&cfq_back_max_entry.attr,
 	&cfq_back_penalty_entry.attr,
-	&cfq_clear_elapsed_entry.attr,
+	&cfq_slice_sync_entry.attr,
+	&cfq_slice_async_entry.attr,
+	&cfq_slice_async_rq_entry.attr,
+	&cfq_slice_idle_entry.attr,
+	&cfq_max_depth_entry.attr,
 	NULL,
 };
 
@@ -1862,25 +2498,50 @@ static struct elevator_type iosched_cfq 
 	.elevator_owner =	THIS_MODULE,
 };
 
-int cfq_init(void)
+static int __init cfq_init(void)
 {
 	int ret;
 
+	/*
+	 * could be 0 on HZ < 1000 setups
+	 */
+	if (!cfq_slice_async)
+		cfq_slice_async = 1;
+	if (!cfq_slice_idle)
+		cfq_slice_idle = 1;
+
 	if (cfq_slab_setup())
 		return -ENOMEM;
 
 	ret = elv_register(&iosched_cfq);
-	if (!ret) {
-		__module_get(THIS_MODULE);
-		return 0;
-	}
+	if (ret)
+		cfq_slab_kill();
 
-	cfq_slab_kill();
 	return ret;
 }
 
 static void __exit cfq_exit(void)
 {
+	struct task_struct *g, *p;
+	unsigned long flags;
+
+	read_lock_irqsave(&tasklist_lock, flags);
+
+	/*
+	 * iterate each process in the system, removing our io_context
+	 */
+	do_each_thread(g, p) {
+		struct io_context *ioc = p->io_context;
+
+		if (ioc && ioc->cic) {
+			ioc->cic->exit(ioc->cic);
+			cfq_free_io_context(ioc->cic);
+			ioc->cic = NULL;
+		}
+	} while_each_thread(g, p);
+
+	read_unlock_irqrestore(&tasklist_lock, flags);
+
 	cfq_slab_kill();
 	elv_unregister(&iosched_cfq);
 }
===== drivers/block/ll_rw_blk.c 1.281 vs edited =====
--- 1.281/drivers/block/ll_rw_blk.c	2004-12-01 09:13:57 +01:00
+++ edited/drivers/block/ll_rw_blk.c	2004-12-21 15:33:52 +01:00
@@ -1257,11 +1257,7 @@ void __generic_unplug_device(request_que
 	if (!blk_remove_plug(q))
 		return;
 
-	/*
-	 * was plugged, fire request_fn if queue has stuff to do
-	 */
-	if (elv_next_request(q))
-		q->request_fn(q);
+	q->request_fn(q);
 }
 EXPORT_SYMBOL(__generic_unplug_device);
 
@@ -2152,7 +2148,6 @@ void __blk_put_request(request_queue_t *
 		return;
 
 	req->rq_status = RQ_INACTIVE;
-	req->q = NULL;
 	req->rl = NULL;
 
 	/*
@@ -2502,6 +2497,7 @@ void blk_finish_queue_drain(request_queu
 {
 	struct request_list *rl = &q->rq;
 	struct request *rq;
+	int requeued = 0;
 
 	spin_lock_irq(q->queue_lock);
 	clear_bit(QUEUE_FLAG_DRAIN, &q->queue_flags);
@@ -2510,9 +2506,13 @@ void blk_finish_queue_drain(request_queu
 		rq = list_entry_rq(q->drain_list.next);
 
 		list_del_init(&rq->queuelist);
-		__elv_add_request(q, rq, ELEVATOR_INSERT_BACK, 1);
+		elv_requeue_request(q, rq);
+		requeued++;
 	}
 
+	if (requeued)
+		q->request_fn(q);
+
 	spin_unlock_irq(q->queue_lock);
 
 	wake_up(&rl->wait[0]);
@@ -3069,6 +3069,7 @@ void exit_io_context(void)
 	local_irq_save(flags);
 	ioc = current->io_context;
 	current->io_context = NULL;
+	ioc->task = NULL;
 	local_irq_restore(flags);
 
 	if (ioc->aic && ioc->aic->exit)
@@ -3103,12 +3104,12 @@ struct io_context *get_io_context(int gf
 	ret = kmem_cache_alloc(iocontext_cachep, gfp_flags);
 	if (ret) {
 		atomic_set(&ret->refcount, 1);
-		ret->pid = tsk->pid;
+		ret->task = current;
+		ret->set_ioprio = NULL;
 		ret->last_waited = jiffies; /* doesn't matter... */
 		ret->nr_batch_requests = 0; /* because this is 0 */
 		ret->aic = NULL;
 		ret->cic = NULL;
-		spin_lock_init(&ret->lock);
 
 		local_irq_save(flags);
 
@@ -3405,3 +3406,4 @@ void blk_unregister_queue(struct gendisk
 		kobject_put(&disk->kobj);
 	}
 }
+
===== fs/Makefile 1.65 vs edited =====
--- 1.65/fs/Makefile	2004-10-29 10:14:03 +02:00
+++ edited/fs/Makefile	2004-12-20 14:26:20 +01:00
@@ -10,6 +10,7 @@ obj-y :=	open.o read_write.o file_table.
 		ioctl.o readdir.o select.o fifo.o locks.o dcache.o inode.o \
 		attr.o bad_inode.o file.o filesystems.o namespace.o aio.o \
 		seq_file.o xattr.o libfs.o fs-writeback.o mpage.o direct-io.o \
+		ioprio.o
 
 obj-$(CONFIG_EPOLL)		+= eventpoll.o
 obj-$(CONFIG_COMPAT)		+= compat.o
===== fs/reiserfs/journal.c 1.98 vs edited =====
--- 1.98/fs/reiserfs/journal.c	2004-10-19 07:26:49 +02:00
+++ edited/fs/reiserfs/journal.c	2004-12-20 14:26:20 +01:00
@@ -659,18 +659,22 @@ struct buffer_chunk {
 
 static void write_chunk(struct buffer_chunk *chunk) {
     int i;
+    get_fs_excl();
     for (i = 0; i < chunk->nr ; i++) {
 	submit_logged_buffer(chunk->bh[i]) ;
     }
     chunk->nr = 0;
+    put_fs_excl();
 }
 
 static void write_ordered_chunk(struct buffer_chunk *chunk) {
     int i;
+    get_fs_excl();
     for (i = 0; i < chunk->nr ; i++) {
 	submit_ordered_buffer(chunk->bh[i]) ;
     }
     chunk->nr = 0;
+    put_fs_excl();
 }
 
 static int add_to_chunk(struct buffer_chunk *chunk, struct buffer_head *bh,
@@ -932,6 +936,8 @@ static int flush_commit_list(struct supe
     return 0 ;
   }
 
+  get_fs_excl();
+
   /* before we can put our commit blocks on disk, we have to make sure everyone older than
   ** us is on disk too
   */
@@ -1069,6 +1075,7 @@ put_jl:
 
   if (retval)
     reiserfs_abort (s, retval, "Journal write error in %s", __FUNCTION__);
+  put_fs_excl();
   return retval;
 }
 
@@ -1243,6 +1250,7 @@ static int flush_journal_list(struct sup
   unsigned long j_len_saved = jl->j_len ;
   struct reiserfs_journal *journal = SB_JOURNAL (s);
   int err = 0;
+  int had_super;
 
   BUG_ON (j_len_saved <= 0);
 
@@ -1265,6 +1273,8 @@ static int flush_journal_list(struct sup
     return 0 ;
   }
 
+  get_fs_excl();
+
   /* if all the work is already done, get out of here */
   if (atomic_read(&(jl->j_nonzerolen)) <= 0 && 
       atomic_read(&(jl->j_commit_left)) <= 0) {
@@ -1464,6 +1474,7 @@ flush_older_and_return:
   put_journal_list(s, jl);
   if (flushall)
     up(&journal->j_flush_sem);
+  put_fs_excl();
   return err ;
 } 
 
@@ -2733,6 +2744,7 @@ relock:
   th->t_trans_id = journal->j_trans_id ;
   unlock_journal(p_s_sb) ;
   INIT_LIST_HEAD (&th->t_list);
+  get_fs_excl();
   return 0 ;
 
 out_fail:
@@ -3536,6 +3548,7 @@ static int do_journal_end(struct reiserf
   BUG_ON (th->t_refcount > 1);
   BUG_ON (!th->t_trans_id);
 
+  put_fs_excl();
   current->journal_info = th->t_handle_save;
   reiserfs_check_lock_depth(p_s_sb, "journal end");
   if (journal->j_len == 0) {
===== include/asm-i386/unistd.h 1.43 vs edited =====
--- 1.43/include/asm-i386/unistd.h	2004-10-24 12:32:46 +02:00
+++ edited/include/asm-i386/unistd.h	2004-12-13 18:42:50 +01:00
@@ -294,8 +294,10 @@
 #define __NR_add_key		286
 #define __NR_request_key	287
 #define __NR_keyctl		288
+#define __NR_ioprio_set		289
+#define __NR_ioprio_get		290
 
-#define NR_syscalls 289
+#define NR_syscalls 291
 
 /*
  * user-visible error numbers are in the range -1 - -128: see
===== include/asm-ia64/unistd.h 1.53 vs edited =====
--- 1.53/include/asm-ia64/unistd.h	2004-11-23 21:18:59 +01:00
+++ edited/include/asm-ia64/unistd.h	2004-12-13 18:45:58 +01:00
@@ -263,6 +263,8 @@
 #define __NR_add_key			1271
 #define __NR_request_key		1272
 #define __NR_keyctl			1273
+#define __NR_ioprio_set			1274
+#define __NR_ioprio_get			1275
 
 #ifdef __KERNEL__
 
===== include/asm-ppc/unistd.h 1.34 vs edited =====
--- 1.34/include/asm-ppc/unistd.h	2004-10-22 11:27:40 +02:00
+++ edited/include/asm-ppc/unistd.h	2004-12-13 18:44:02 +01:00
@@ -276,8 +276,10 @@
 #define __NR_add_key		269
 #define __NR_request_key	270
 #define __NR_keyctl		271
+#define __NR_ioprio_set		272
+#define __NR_ioprio_get		273
 
-#define __NR_syscalls		272
+#define __NR_syscalls		274
 
 #define __NR(n)	#n
 
===== include/asm-x86_64/unistd.h 1.31 vs edited =====
--- 1.31/include/asm-x86_64/unistd.h	2004-09-03 11:08:14 +02:00
+++ edited/include/asm-x86_64/unistd.h	2004-12-13 18:45:13 +01:00
@@ -556,8 +556,12 @@ __SYSCALL(__NR_mq_getsetattr, sys_mq_get
 __SYSCALL(__NR_kexec_load, sys_ni_syscall)
 #define __NR_waitid		247
 __SYSCALL(__NR_waitid, sys_waitid)
+#define __NR_ioprio_set		248
+__SYSCALL(__NR_ioprio_set, sys_ioprio_set)
+#define __NR_ioprio_get		249
+__SYSCALL(__NR_ioprio_get, sys_ioprio_get)
 
-#define __NR_syscall_max __NR_waitid
+#define __NR_syscall_max __NR_ioprio_get
 #ifndef __NO_STUBS
 
 /* user-visible error numbers are in the range -1 - -4095 */
===== include/linux/blkdev.h 1.157 vs edited =====
--- 1.157/include/linux/blkdev.h	2004-11-11 09:39:16 +01:00
+++ edited/include/linux/blkdev.h	2004-12-21 15:33:52 +01:00
@@ -54,16 +54,27 @@ struct as_io_context {
 
 struct cfq_queue;
 struct cfq_io_context {
-	void (*dtor)(struct cfq_io_context *);
-	void (*exit)(struct cfq_io_context *);
-
-	struct io_context *ioc;
-
 	/*
 	 * circular list of cfq_io_contexts belonging to a process io context
 	 */
 	struct list_head list;
 	struct cfq_queue *cfqq;
+
+	struct io_context *ioc;
+
+	unsigned long last_end_request;
+	unsigned long last_queue;
+	unsigned long ttime_total;
+	unsigned long ttime_samples;
+	unsigned long ttime_mean;
+
+	unsigned int seek_samples;
+	sector_t last_request_pos;
+	u64 seek_total;
+	sector_t seek_mean;
+
+	void (*dtor)(struct cfq_io_context *);
+	void (*exit)(struct cfq_io_context *);
 };
 
 /*
@@ -73,7 +84,9 @@ struct cfq_io_context {
  */
 struct io_context {
 	atomic_t refcount;
-	pid_t pid;
+	struct task_struct *task;
+
+	int (*set_ioprio)(struct io_context *, unsigned int);
 
 	/*
 	 * For request batching
@@ -81,8 +94,6 @@ struct io_context {
 	unsigned long last_waited; /* Time last woken after wait for request */
 	int nr_batch_requests;     /* Number of requests left in the batch */
 
-	spinlock_t lock;
-
 	struct as_io_context *aic;
 	struct cfq_io_context *cic;
 };
@@ -722,6 +733,5 @@ void kblockd_flush(void);
 	MODULE_ALIAS("block-major-" __stringify(major) "-" __stringify(minor))
 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
-
 
 #endif
===== include/linux/fs.h 1.362 vs edited =====
--- 1.362/include/linux/fs.h	2004-10-29 10:14:03 +02:00
+++ edited/include/linux/fs.h	2004-12-20 14:26:20 +01:00
@@ -18,6 +18,7 @@
 #include <linux/cache.h>
 #include <linux/prio_tree.h>
 #include <linux/kobject.h>
+#include <linux/sched.h>
 #include <asm/atomic.h>
 
 struct iovec;
@@ -811,16 +812,34 @@ enum {
 #define vfs_check_frozen(sb, level) \
 	wait_event((sb)->s_wait_unfrozen, ((sb)->s_frozen < (level)))
 
+static inline void get_fs_excl(void)
+{
+	atomic_inc(&current->fs_excl);
+}
+
+static inline void put_fs_excl(void)
+{
+	atomic_dec(&current->fs_excl);
+}
+
+static inline int has_fs_excl(void)
+{
+	return atomic_read(&current->fs_excl);
+}
+
+
 /*
  * Superblock locking.
  */
 static inline void lock_super(struct super_block * sb)
 {
+	get_fs_excl();
 	down(&sb->s_lock);
 }
 
 static inline void unlock_super(struct super_block * sb)
 {
+	put_fs_excl();
 	up(&sb->s_lock);
 }
 
===== include/linux/init_task.h 1.32 vs edited =====
--- 1.32/include/linux/init_task.h	2004-10-19 07:26:36 +02:00
+++ edited/include/linux/init_task.h	2004-12-21 15:33:07 +01:00
@@ -79,6 +79,7 @@ extern struct group_info init_groups;
 	.mm		= NULL,						\
 	.active_mm	= &init_mm,					\
 	.run_list	= LIST_HEAD_INIT(tsk.run_list),			\
+	.ioprio		= 0,						\
 	.time_slice	= HZ,						\
 	.tasks		= LIST_HEAD_INIT(tsk.tasks),			\
 	.ptrace_children= LIST_HEAD_INIT(tsk.ptrace_children),		\
@@ -112,6 +113,7 @@ extern struct group_info init_groups;
 	.proc_lock	= SPIN_LOCK_UNLOCKED,				\
 	.switch_lock	= SPIN_LOCK_UNLOCKED,				\
 	.journal_info	= NULL,						\
+	.fs_excl	= ATOMIC_INIT(0),				\
 }
 
 
===== include/linux/sched.h 1.271 vs edited =====
--- 1.271/include/linux/sched.h	2004-12-03 08:31:57 +01:00
+++ edited/include/linux/sched.h	2004-12-20 14:38:46 +01:00
@@ -523,6 +523,8 @@ struct task_struct {
 	struct list_head run_list;
 	prio_array_t *array;
 
+	unsigned int ioprio;
+
 	unsigned long sleep_avg;
 	long interactive_credit;
 	unsigned long long timestamp, last_ran;
@@ -664,6 +666,7 @@ struct task_struct {
   	struct mempolicy *mempolicy;
   	short il_next;		/* could be shared with used_math */
 #endif
+	atomic_t fs_excl;
 };
 
 static inline pid_t process_group(struct task_struct *tsk)
===== kernel/exit.c 1.170 vs edited =====
--- 1.170/kernel/exit.c	2004-12-17 09:09:08 +01:00
+++ edited/kernel/exit.c	2004-12-20 14:35:24 +01:00
@@ -786,6 +786,8 @@ fastcall NORET_TYPE void do_exit(long co
 
 	profile_task_exit(tsk);
 
+	WARN_ON(atomic_read(&tsk->fs_excl));
+
 	if (unlikely(in_interrupt()))
 		panic("Aiee, killing interrupt handler!");
 	if (unlikely(!tsk->pid))
===== kernel/fork.c 1.223 vs edited =====
--- 1.223/kernel/fork.c	2004-12-03 08:25:48 +01:00
+++ edited/kernel/fork.c	2004-12-20 14:26:20 +01:00
@@ -1006,6 +1006,11 @@ static task_t *copy_process(unsigned lon
 		spin_unlock(&current->sighand->siglock);
 	}
 
+	/*
+	 * inherit ioprio
+	 */
+	p->ioprio = current->ioprio;
+
 	SET_LINKS(p);
 	if (unlikely(p->ptrace & PT_PTRACED))
 		__ptrace_link(p, current->parent);
--- /dev/null	2004-10-02 05:17:35.000000000 +0200
+++ linux-2.6/fs/ioprio.c	2004-12-21 15:32:43.000000000 +0100
@@ -0,0 +1,46 @@
+#include <linux/kernel.h>
+#include <linux/ioprio.h>
+#include <linux/blkdev.h>
+
+asmlinkage int sys_ioprio_set(int ioprio)
+{
+	int class = IOPRIO_PRIO_CLASS(ioprio);
+	int data = IOPRIO_PRIO_DATA(ioprio);
+	struct io_context *ioc;
+	unsigned long flags;
+
+	switch (class) {
+		case IOPRIO_CLASS_RT:
+			if (!capable(CAP_SYS_ADMIN))
+				return -EPERM;
+			/* fall through, rt has prio field too */
+		case IOPRIO_CLASS_BE:
+			if (data >= IOPRIO_BE_NR)
+				return -EINVAL;
+
+			break;
+		case IOPRIO_CLASS_IDLE:
+			break;
+		default:
+			return -EINVAL;
+	}
+
+	ioc = get_io_context(GFP_KERNEL);
+	if (!ioc)
+		return -ENOMEM;
+
+	local_irq_save(flags);
+	if (ioc->set_ioprio)
+		ioc->set_ioprio(ioc, ioprio);
+	current->ioprio = ioprio;
+	local_irq_restore(flags);
+
+	put_io_context(ioc);
+	return 0;
+}
+
+asmlinkage int sys_ioprio_get(void)
+{
+	return current->ioprio;
+}
+
--- /dev/null	2004-10-02 05:17:35.000000000 +0200
+++ linux-2.6/include/linux/ioprio.h	2004-12-21 15:33:52.000000000 +0100
@@ -0,0 +1,42 @@
+#ifndef IOPRIO_H
+#define IOPRIO_H
+
+#include <linux/sched.h>
+
+/*
+ * Gives us 8 prio classes with 24-bits of data for each class
+ */
+#define IOPRIO_CLASS_SHIFT	(24)
+#define IOPRIO_PRIO_MASK	((1UL << IOPRIO_CLASS_SHIFT) - 1)
+
+#define IOPRIO_PRIO_CLASS(mask)	((mask) >> IOPRIO_CLASS_SHIFT)
+#define IOPRIO_PRIO_DATA(mask)	((mask) & IOPRIO_PRIO_MASK)
+
+enum {
+	IOPRIO_CLASS_NONE,
+	IOPRIO_CLASS_RT,
+	IOPRIO_CLASS_BE,
+	IOPRIO_CLASS_IDLE,
+};
+
+/*
+ * 8 best effort priority levels are supported
+ */
+#define IOPRIO_BE_NR	(8)
+
+asmlinkage int sys_ioprio_set(int);
+asmlinkage int sys_ioprio_get(void);
+
+/*
+ * if process has set io priority explicitly, use that. if not, convert
+ * the cpu scheduler nice value to an io priority
+ */
+static inline int task_ioprio(struct task_struct *task)
+{
+	if (!task->ioprio)
+		return (task_nice(task) + 20) / 5;
+	else
+		return IOPRIO_PRIO_DATA(task->ioprio);
+}
+
+#endif