===== 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 @@
 	.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 @@
 	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 @@
 	.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	2004-12-13 18:45:57 +01:00
@@ -22,21 +22,27 @@
 #include <linux/rbtree.h>
 #include <linux/mempool.h>
 
-static unsigned long max_elapsed_crq;
-static unsigned long max_elapsed_dispatch;
-
 /*
  * 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 / 50;
+static int cfq_slice_async = HZ / 175;
+static int cfq_slice_async_rq = 16;
+static int cfq_slice_idle = HZ / 100;
+
+/*
+ * disable queueing at the driver/hardware level
+ */
+static int cfq_max_depth = 1;
+
 /*
  * for the hash of cfqq inside the cfqd
  */
@@ -55,6 +61,7 @@
 #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
 
@@ -76,22 +83,18 @@
 #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 };
+static char *cfq_key_types[] = { "pgid", "tgid", "pid", "uid", "gid", NULL };
 
 /*
  * spare queue
@@ -102,8 +105,13 @@
 static kmem_cache_t *cfq_pool;
 static kmem_cache_t *cfq_ioc_pool;
 
+#define CFQ_PRIO_NR		IOPRIO_NR
+
 struct cfq_data {
-	struct list_head rr_list;
+	atomic_t ref;
+
+	struct list_head rr_list[CFQ_PRIO_NR];
+	struct list_head cur_rr;
 	struct list_head empty_list;
 
 	struct hlist_head *cfq_hash;
@@ -114,8 +122,6 @@
 
 	unsigned int max_queued;
 
-	atomic_t ref;
-
 	int key_type;
 
 	mempool_t *crq_pool;
@@ -127,6 +133,20 @@
 	int rq_in_driver;
 
 	/*
+	 * schedule slice state info
+	 */
+	struct timer_list timer;
+	struct work_struct unplug_work;
+	struct cfq_queue *active_queue;
+	struct cfq_io_context *active_cic;
+	int cur_prio;
+	unsigned int dispatch_slice;
+	unsigned long idle_start;
+
+	unsigned long new_ttime_total;
+	unsigned long new_ttime_mean;
+
+	/*
 	 * tunables, see top of file
 	 */
 	unsigned int cfq_quantum;
@@ -136,9 +156,10 @@
 	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;
 };
 
 struct cfq_queue {
@@ -150,8 +171,6 @@
 	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 +188,27 @@
 
 	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;
+
+	/* 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;
+	/* idle slice, request added, now waiting to dispatch it */
+	unsigned int must_dispatch : 1;
+
+	unsigned int idle_hit, idle_miss;
+
+	char name[16];
 };
 
 struct cfq_rq {
@@ -195,13 +226,13 @@
 	unsigned int in_flight : 1;
 	unsigned int accounted : 1;
 	unsigned int is_sync   : 1;
-	unsigned int is_write  : 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 +250,8 @@
 		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:
@@ -406,67 +439,23 @@
 		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)
 {
-	struct list_head *head = &cfqq->cfqd->rr_list;
-	struct list_head *next, *prev;
-
-	/*
-	 * list might still be ordered
-	 */
-	next = cfqq->cfq_list.next;
-	if (next != head) {
-		struct cfq_queue *cnext = list_entry_cfqq(next);
-
-		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;
-}
-
-static void cfq_sort_rr_list(struct cfq_queue *cfqq, int new_queue)
-{
-	struct list_head *entry = &cfqq->cfqd->rr_list;
-
-	if (!cfqq->on_rr)
-		return;
-	if (!new_queue && !cfq_check_sort_rr_list(cfqq))
-		return;
+	struct list_head *list = &cfqq->cfqd->rr_list[cfqq->ioprio];
+	struct list_head *entry = list;
 
 	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) {
+	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);
@@ -479,16 +468,12 @@
 static inline void
 cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 {
-	/*
-	 * it's currently on the empty list
-	 */
-	cfqq->on_rr = 1;
-	cfqd->busy_queues++;
+	BUG_ON(cfqq->on_rr);
 
-	if (time_after(jiffies, cfqq->service_start + cfq_service))
-		cfqq->service_used >>= 3;
+	cfqd->busy_queues++;
+	cfqq->on_rr = 1;
 
-	cfq_sort_rr_list(cfqq, 1);
+	cfq_resort_rr_list(cfqq);
 }
 
 static inline void
@@ -512,10 +497,10 @@
 		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);
 
@@ -619,17 +604,13 @@
 {
 	struct cfq_rq *crq = RQ_DATA(rq);
 
-	if (crq) {
+	if (crq && crq->accounted) {
 		struct cfq_queue *cfqq = crq->cfq_queue;
 
-		if (cfqq->cfqd->cfq_tagged) {
-			cfqq->service_used--;
-			cfq_sort_rr_list(cfqq, 0);
-		}
-
 		crq->accounted = 0;
 		cfqq->cfqd->rq_in_driver--;
 	}
+
 	list_add(&rq->queuelist, &q->queue_head);
 }
 
@@ -640,9 +621,8 @@
 	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);
 	}
 }
 
@@ -723,6 +703,185 @@
 	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;
+		cfqq->wait_request = 0;
+
+		cfqd->active_cic = cfq_get_io_context(&cfqq, GFP_ATOMIC);
+		cfqd->cur_prio = cfqq->ioprio;
+	}
+
+	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_NR)
+			prio = 0;
+
+	} while (prio != cfqd->cur_prio);
+
+	__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->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
+		 */
+		if (time_before(now, cfqq->slice_end))
+			cfqq->slice_left = now - cfqq->slice_end;
+		else
+			cfqq->slice_left = 0;
+
+		if (cfqq->on_rr) {
+			/*
+			 * try to be fair to the preempted queue - it gets
+			 * first service next time this prio level is served
+			 */
+			if (preempted)
+				list_move(&cfqq->cfq_list, &cfqd->rr_list[cfqq->ioprio]);
+			else
+				cfq_resort_rr_list(cfqq);
+		}
+
+		/*
+		 * 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_NR)
+				cfqd->cur_prio = 0;
+		}
+
+		cfqd->active_queue = NULL;
+
+		if (cfqd->active_cic)
+			put_io_context(cfqd->active_cic->ioc);
+	}
+
+	cfqd->dispatch_slice = 0;
+}
+
+static int
+cfq_arm_slice_timer(struct cfq_data *cfqd, struct cfq_queue *cfqq,
+		    struct cfq_io_context *cic)
+			
+{
+	WARN_ON(!RB_EMPTY(&cfqq->sort_list));
+
+	cfqq->wait_request = 1;
+
+	if (!cfqd->cfq_slice_idle)
+		return 0;
+
+	/*
+	 * don't arm a slice timer if we don't expect the task to run
+	 * after the slice has ended
+	 */
+	if (cic) {
+		unsigned long sched_time;
+
+		if (!cic->ioc->task)
+			return 0;
+
+		if (cic->ttime_samples == 0 &&
+		    cfqd->new_ttime_mean > cfqd->cfq_slice_idle)
+			return 0;
+		else if (cic->ttime_mean > cfqd->cfq_slice_idle)
+			return 0;
+
+		sched_time = task_will_schedule_at(cic->ioc->task);
+		if (sched_time && time_after(sched_time, cfqq->slice_end))
+			return 0;
+	}
+
+	if (!timer_pending(&cfqd->timer)) {
+		unsigned long slice_left = cfqq->slice_end - 1;
+
+		cfqd->idle_start = jiffies;
+		cfqd->timer.expires = min(jiffies + cfqd->cfq_slice_idle, slice_left);
+		add_timer(&cfqd->timer);
+	}
+
+	return 1;
+}
+
+/*
+ * 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;
+
+	/*
+	 * slice has expired
+	 */
+	if (!cfqq->must_dispatch && time_after(jiffies, cfqq->slice_end))
+		goto new_queue;
+
+	/*
+	 * if queue has requests, dispatch one. if not, check if
+	 * enough slice is left to wait for one
+	 */
+	if (!RB_EMPTY(&cfqq->sort_list))
+		goto keep_queue;
+	else if (time_before(now, cfqq->slice_end)) {
+		if (cfq_arm_slice_timer(cfqd, cfqq, NULL))
+			return NULL;
+	}
+
+new_queue:
+	cfq_slice_expired(cfqd, 0);
+	cfq_set_active_queue(cfqd);
+keep_queue:
+	return cfqd->active_queue;
+}
+
 /*
  * 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
@@ -741,26 +900,25 @@
 	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;
 			break;
 		}
+		entry = &__rq->queuelist;
 	}
 
 	cfqd->last_sector = last;
 	crq->in_flight = 1;
 	cfqq->in_flight++;
-	list_add(&crq->request->queuelist, entry);
+	list_add_tail(&crq->request->queuelist, entry);
 }
 
 /*
@@ -777,97 +935,109 @@
 	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));
+		if (time_after(now, crq->queue_start + 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));
+		if (time_after(now, crq->queue_start + 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];
 
-	/*
-	 * follow expired path, else get first next available
-	 */
-	if ((crq = cfq_check_fifo(cfqq)) == NULL) {
-		if (cfqd->find_best_crq)
+	return base_slice + base_slice * (CFQ_PRIO_NR - 1 - prio);
+}
+
+static int
+__cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
+			int max_dispatch)
+{
+	int dispatched = 0, sync = 0;
+
+	BUG_ON(RB_EMPTY(&cfqq->sort_list));
+
+	do {
+		struct cfq_rq *crq;
+
+		/*
+		 * follow expired path, else get first next available
+		 */
+		if ((crq = cfq_check_fifo(cfqq)) == NULL)
 			crq = cfqq->next_crq;
-		else
-			crq = rb_entry_crq(rb_first(&cfqq->sort_list));
-	}
 
-	cfqd->last_sector = crq->request->sector + crq->request->nr_sectors;
+		cfqd->last_sector = crq->request->sector + crq->request->nr_sectors;
+
+		/*
+		 * finally, insert request into driver list
+		 */
+		cfq_dispatch_sort(cfqd->queue, crq);
+
+		cfqd->dispatch_slice++;
+		dispatched++;
+		sync += crq->is_sync;
+
+		if (RB_EMPTY(&cfqq->sort_list))
+			break;
+
+	} while (dispatched < max_dispatch);
 
 	/*
-	 * finally, insert request into driver list
+	 * if slice end isn't set yet, set it. if at least one request was
+	 * sync, use the sync time slice value
 	 */
-	cfq_dispatch_sort(q, crq);
+	if (!cfqq->slice_end)
+		cfqq->slice_end = cfq_prio_to_slice(cfqd, cfqq->ioprio, !!sync) + jiffies;
+
+	/*
+	 * expire an async queue immediately if it has used up its slice
+	 */
+	if (!sync && cfqd->dispatch_slice >= cfqd->cfq_slice_async_rq)
+		cfq_slice_expired(cfqd, 0);
+
+	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;
-	struct list_head *entry, *tmp;
-	int queued, busy_queues, first_round;
 
-	if (list_empty(&cfqd->rr_list))
+	if (!cfqd->busy_queues)
 		return 0;
 
-	queued = 0;
-	first_round = 1;
-restart:
-	busy_queues = 0;
-	list_for_each_safe(entry, tmp, &cfqd->rr_list) {
-		cfqq = list_entry_cfqq(entry);
-
-		BUG_ON(RB_EMPTY(&cfqq->sort_list));
-
-		/*
-		 * first round of queueing, only select from queues that
-		 * don't already have io in-flight
-		 */
-		if (first_round && cfqq->in_flight)
-			continue;
-
-		cfq_dispatch_request(q, cfqd, cfqq);
-
-		if (!RB_EMPTY(&cfqq->sort_list))
-			busy_queues++;
-
-		queued++;
-	}
-
-	if ((queued < max_dispatch) && (busy_queues || first_round)) {
-		first_round = 0;
-		goto restart;
+	cfqq = cfq_select_queue(cfqd);
+	if (cfqq) {
+		cfqq->wait_request = 0;
+		cfqq->must_dispatch = 0;
+		del_timer(&cfqd->timer);
+		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 +1047,9 @@
 	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);
-	}
+	crq->service_start = jiffies;
+	cfqd->rq_in_driver++;
 }
 
 static inline void
@@ -915,27 +1057,27 @@
 {
 	struct cfq_data *cfqd = cfqq->cfqd;
 
-	if (!crq->accounted)
-		return;
-
-	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 (crq->accounted) {
+		WARN_ON(!cfqd->rq_in_driver);
+		cfqd->rq_in_driver--;
+	}
 
-		if (time_after(now, cfqq->service_start + cfq_service)) {
-			cfqq->service_start = now;
-			cfqq->service_used >>= 3;
-		}
+	/*
+	 * queue was preempted while this request was servicing
+	 */
+	if (cfqd->active_queue != cfqq)
+		return;
 
-		cfqq->service_used += duration;
-		cfq_sort_rr_list(cfqq, 0);
+	if (crq->is_sync)
+		crq->io_context->last_end_request = jiffies;
 
-		if (duration > max_elapsed_crq)
-			max_elapsed_crq = duration;
-	}
+	/*
+	 * this is still the active queue. if we have nothing to do and no
+	 * more pending requests in flight, wait for a new sync request if
+	 * this request was sync itself
+	 */
+	if (RB_EMPTY(&cfqq->sort_list) && crq->is_sync && !cfqq->in_flight)
+		cfq_arm_slice_timer(cfqd, cfqq, crq->io_context);
 }
 
 static struct request *cfq_next_request(request_queue_t *q)
@@ -943,6 +1085,9 @@
 	struct cfq_data *cfqd = q->elevator->elevator_data;
 	struct request *rq;
 
+	if (cfqd->rq_in_driver >= cfqd->cfq_max_depth)
+		return NULL;
+
 	if (!list_empty(&q->queue_head)) {
 		struct cfq_rq *crq;
 dispatch:
@@ -970,6 +1115,8 @@
  */
 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 +1125,9 @@
 	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);
 
 	/*
@@ -1039,44 +1189,63 @@
 	spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
 }
 
+static void cfq_queue_remove(struct cfq_queue *cfqq)
+{
+	struct cfq_data *cfqd = cfqq->cfqd;
+
+	/*
+	 * don't waste time waiting for the timer to expire for
+	 * a dead process
+	 */
+	if (cfqq == cfqd->active_queue && del_timer(&cfqd->timer))
+		kblockd_schedule_work(&cfqd->unplug_work);
+}
+
+/*
+ * A note on the RCU usage in CFQ: All updates on the cic list are being
+ * done from only one task, so no additional locking is needed. Updates are
+ * done in two locations: when a process sets up a new per-queue context and
+ * when a process exits.
+ */
+static void cfq_rcu_queue_put(struct rcu_head *h)
+{
+	struct cfq_io_context *cic = container_of(h, struct cfq_io_context,rcu);
+	request_queue_t *q = cic->cfqq->cfqd->queue;
+	unsigned long flags;
+
+	spin_lock_irqsave(q->queue_lock, flags);
+	cfq_queue_remove(cic->cfqq);
+	cfq_put_queue(cic->cfqq);
+	cic->cfqq = NULL;
+	spin_unlock_irqrestore(q->queue_lock, flags);
+
+	put_io_context(cic->ioc);
+}
+
 static void cfq_free_io_context(struct cfq_io_context *cic)
 {
 	kmem_cache_free(cfq_ioc_pool, cic);
 }
 
 /*
- * locking hierarchy is: io_context lock -> queue locks
+ * See cfq_rcu_queue_put
  */
 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 list_head *entry, *nxt;
+	struct cfq_io_context *__cic;
 
 	/*
 	 * 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;
-
+	list_for_each_safe_rcu(entry, nxt, &cic->list) {
 		__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_del_rcu(&__cic->list);
+		call_rcu(&__cic->rcu, cfq_rcu_queue_put);
 	}
 
-	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);
+	call_rcu(&cic->rcu, cfq_rcu_queue_put);
 }
 
 static struct cfq_io_context *cfq_alloc_io_context(int gfp_flags)
@@ -1084,10 +1253,15 @@
 	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;
+		INIT_RCU_HEAD(&cic->rcu);
+		cic->last_end_request = 0;
+		cic->ttime_total = 0;
+		cic->ttime_samples = 0;
+		cic->ttime_mean = 0;
+		cic->dtor = cfq_free_io_context;
+		cic->exit = cfq_exit_io_context;
 	}
 
 	return cic;
@@ -1119,13 +1293,19 @@
 		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;
 		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
@@ -1140,19 +1320,24 @@
 		 * 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) {
+		rcu_read_lock();
+		list_for_each_entry_rcu(__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) {
+				rcu_read_unlock();
+				/*
+				 * move hot entries to the front
+				 */
+				list_del_rcu(&__cic->list);
+				list_add_rcu(&__cic->list, &cic->list);
 				cic = __cic;
-				spin_unlock_irqrestore(&ioc->lock, flags);
 				goto out;
 			}
 		}
-		spin_unlock_irqrestore(&ioc->lock, flags);
+		rcu_read_unlock();
 
 		/*
 		 * nope, process doesn't have a cic assoicated with this
@@ -1163,12 +1348,12 @@
 			goto err;
 
 		__cic->ioc = ioc;
+		atomic_inc(&ioc->refcount);
+
 		__cic->cfqq = __cfqq;
 		atomic_inc(&__cfqq->ref);
-		spin_lock_irqsave(&ioc->lock, flags);
-		list_add(&__cic->list, &cic->list);
-		spin_unlock_irqrestore(&ioc->lock, flags);
-
+		atomic_inc(&cfqd->ref);
+		list_add_rcu(&__cic->list, &cic->list);
 		cic = __cic;
 		*cfqq = __cfqq;
 	}
@@ -1205,8 +1390,11 @@
 			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 +1410,10 @@
 		cfqq->cfqd = cfqd;
 		atomic_inc(&cfqd->ref);
 		cfqq->key_type = cfqd->key_type;
-		cfqq->service_start = ~0UL;
+		cfqq->service_last = 0;
+		cfqq->slice_left = 2 * cfqd->cfq_slice_idle;
+		cfqq->ioprio = task_ioprio(current);
+		strncpy(cfqq->name, current->comm, sizeof(cfqq->name) - 1);
 	}
 
 	if (new_cfqq)
@@ -1247,23 +1438,150 @@
 	return cfqq;
 }
 
-static void cfq_enqueue(struct cfq_data *cfqd, struct cfq_rq *crq)
+static void
+cfq_update_iostat(struct cfq_data *cfqd, struct cfq_io_context *cic,
+		  struct cfq_rq *crq)
 {
-	crq->is_sync = 0;
-	if (rq_data_dir(crq->request) == READ || current->flags & PF_SYNCWRITE)
-		crq->is_sync = 1;
+	unsigned long elapsed = jiffies - cic->last_end_request;
+	unsigned long ttime = min(elapsed, 2UL * cfqd->cfq_slice_idle);
+
+	if (cic->ttime_samples == 0) {
+		cfqd->new_ttime_total = (7*cfqd->new_ttime_total+256*ttime)/8;
+		cfqd->new_ttime_mean = cfqd->new_ttime_total / 256;
+	}
+
+	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;
+}
+
+/*
+ * Check if new_cfqq should preempt the currently active queue
+ */
+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;
+	struct cfq_io_context *cic;
+
+	if (!cfqq)
+		return 0;
+	if ((cic = cfqd->active_cic) == NULL)
+		return 0;
+
+	/*
+	 * if ioprio of new queue is less than active queue, never preempt
+	 */
+	if (new_cfqq->ioprio < cfqq->ioprio)
+		return 0;
+	else if (new_cfqq->ioprio == cfqq->ioprio) {
+		if (!crq->is_sync &&
+		    (cfqq->wait_request || cfqq->must_dispatch))
+			return 0;
+		if (cic->ttime_samples == 0)
+			return 0;
+		if (cfqd->idle_start - jiffies < cic->ttime_mean)
+			return 0;
+	} else {
+		if (!crq->is_sync &&
+		    (cfqq->wait_request || cfqq->must_dispatch))
+			return 0;
+
+		/*
+		 * new_cfqq is higher prio, allow preempt if it has slice left
+		 */
+		if (new_cfqq->slice_left < 2 * cfqd->cfq_slice_idle)
+			return 0;
+	}
+
+	return 1;
+}
+
+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);
+
+	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;
+
+	if (!blk_queue_plugged(q))
+		q->request_fn(q);
+	else
+		__generic_unplug_device(q);
+}
+
+/*
+ * 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->io_context)
+		cfq_update_iostat(cfqd, crq->io_context, crq);
+
+	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->timer);
+			cfq_start_queueing(cfqd, cfqq);
+		}
+	} else if (cfqd->active_queue && 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 request *rq)
+{
+	struct cfq_rq *crq = RQ_DATA(rq);
+	struct cfq_queue *cfqq = crq->cfq_queue;
 
 	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 +1594,19 @@
 			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)
@@ -1345,32 +1656,31 @@
 	struct cfq_data *cfqd = q->elevator->elevator_data;
 	struct cfq_queue *cfqq;
 	int ret = ELV_MQUEUE_MAY;
+	int limit;
 
 	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;
-
-		if (limit < cfqd->cfq_queued)
-			limit = cfqd->cfq_queued;
-		else if (limit > cfqd->max_queued)
-			limit = cfqd->max_queued;
+	if (unlikely(!cfqq))
+		return ELV_MQUEUE_MAY;
 
-		if (cfqq->allocated[rw] >= limit) {
-			if (limit > cfqq->alloc_limit[rw])
-				cfqq->alloc_limit[rw] = limit;
+	if (cfqq->allocated[rw] < cfqd->cfq_queued)
+		return ELV_MQUEUE_MUST;
+	if (cfqq->wait_request)
+		return ELV_MQUEUE_MUST;
+
+	limit = cfqd->max_queued;
+	if (cfqd->busy_queues)
+		limit = q->nr_requests / cfqd->busy_queues;
+
+	if (limit < cfqd->cfq_queued)
+		limit = cfqd->cfq_queued;
+	else if (limit > cfqd->max_queued)
+		limit = cfqd->max_queued;
 
-			ret = ELV_MQUEUE_NO;
-		}
-	}
+	if (cfqq->allocated[rw] >= limit)
+		ret = ELV_MQUEUE_NO;
 
 	return ret;
 }
@@ -1378,12 +1688,13 @@
 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]);
 }
 
@@ -1397,16 +1708,17 @@
 
 	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));
 
+		BUG_ON(!cfqq->allocated[rw]);
+		cfqq->allocated[rw]--;
+
 		if (crq->io_context)
 			put_io_context(crq->io_context->ioc);
 
-		BUG_ON(!cfqq->allocated[crq->is_write]);
-		cfqq->allocated[crq->is_write]--;
-
 		mempool_free(crq, cfqd->crq_pool);
 		rq->elevator_private = NULL;
 
@@ -1422,9 +1734,9 @@
 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;
 
@@ -1461,6 +1773,7 @@
 	 * 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;
@@ -1475,10 +1788,9 @@
 		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);
 		rq->elevator_private = crq;
-		cfqq->alloc_limit[rw] = 0;
 		return 0;
 	}
 
@@ -1492,6 +1804,60 @@
 	return 1;
 }
 
+static void cfq_kick_queue(void *data)
+{
+	request_queue_t *q = data;
+
+	blk_run_queue(q);
+}
+
+static void cfq_schedule_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) {
+			unsigned long slice_left = cfqq->slice_end - 1;
+	
+			cfqd->timer.expires = min(jiffies + cfqd->cfq_slice_idle, slice_left);
+			add_timer(&cfqd->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_cont;
+		}
+miss:
+		cfqq->idle_miss++;
+	}
+
+	cfq_slice_expired(cfqd, 0);
+out_cont:
+	if (cfqd->busy_queues)
+		kblockd_schedule_work(&cfqd->unplug_work);
+	else
+		WARN_ON(!list_empty(&cfqd->cur_rr));
+	spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
+}
+
 static void cfq_put_cfqd(struct cfq_data *cfqd)
 {
 	request_queue_t *q = cfqd->queue;
@@ -1500,6 +1866,8 @@
 	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
@@ -1508,6 +1876,11 @@
 	cfqq = __cfq_get_queue(cfqd, CFQ_KEY_SPARE, GFP_ATOMIC);
 	cfq_put_queue(cfqq);
 	cfq_put_queue(cfqq);
+
+	/*
+	 * restore ->nr_requests
+	 */
+	q->nr_requests = BLKDEV_MAX_RQ;
 	spin_unlock_irq(q->queue_lock);
 
 	blk_put_queue(q);
@@ -1534,7 +1907,11 @@
 		return -ENOMEM;
 
 	memset(cfqd, 0, sizeof(*cfqd));
-	INIT_LIST_HEAD(&cfqd->rr_list);
+
+	for (i = 0; i < CFQ_PRIO_NR; i++)
+		INIT_LIST_HEAD(&cfqd->rr_list[i]);
+
+	INIT_LIST_HEAD(&cfqd->cur_rr);
 	INIT_LIST_HEAD(&cfqd->empty_list);
 
 	cfqd->crq_hash = kmalloc(sizeof(struct hlist_head) * CFQ_MHASH_ENTRIES, GFP_KERNEL);
@@ -1571,10 +1948,16 @@
 	 * some requests. fairness is handled differently
 	 */
 	q->nr_requests = 1024;
-	cfqd->max_queued = q->nr_requests / 16;
+	cfqd->max_queued = q->nr_requests / 8;
 	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->timer);
+	cfqd->timer.function = cfq_schedule_timer;
+	cfqd->timer.data = (unsigned long) cfqd;
+
+	INIT_WORK(&cfqd->unplug_work, cfq_kick_queue, q);
+
 	atomic_set(&cfqd->ref, 1);
 
 	cfqd->cfq_queued = cfq_queued;
@@ -1584,6 +1967,11 @@
 	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:
@@ -1630,7 +2018,6 @@
 	return -ENOMEM;
 }
 
-
 /*
  * sysfs parts below -->
  */
@@ -1656,13 +2043,6 @@
 }
 
 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 +2050,8 @@
 		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 +2089,13 @@
 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 +2118,13 @@
 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 +2152,6 @@
 	.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 +2162,36 @@
 	.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 +2200,13 @@
 	&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 +2271,50 @@
 	.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	2004-12-13 18:45:57 +01:00
@@ -1257,11 +1257,7 @@
 	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 @@
 		return;
 
 	req->rq_status = RQ_INACTIVE;
-	req->q = NULL;
 	req->rl = NULL;
 
 	/*
@@ -2502,6 +2497,7 @@
 {
 	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 @@
 		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 @@
 	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 +3104,11 @@
 	ret = kmem_cache_alloc(iocontext_cachep, gfp_flags);
 	if (ret) {
 		atomic_set(&ret->refcount, 1);
-		ret->pid = tsk->pid;
+		ret->task = current;
 		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);
 
@@ -3404,4 +3404,18 @@
 		kobject_unregister(&q->kobj);
 		kobject_put(&disk->kobj);
 	}
+}
+
+asmlinkage int sys_ioprio_set(int ioprio)
+{
+	if (ioprio < 0 || ioprio >= IOPRIO_NR)
+		return -EINVAL;
+
+	current->ioprio = ioprio;
+	return 0;
+}
+
+asmlinkage int sys_ioprio_get(void)
+{
+	return task_ioprio(current);
 }
===== 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_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	2004-12-13 18:45:58 +01:00
@@ -54,16 +54,22 @@
 
 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 rcu_head rcu;
+	struct io_context *ioc;
+
+	unsigned long last_end_request;
+	unsigned long ttime_total;
+	unsigned long ttime_samples;
+	unsigned long ttime_mean;
+
+	void (*dtor)(struct cfq_io_context *);
+	void (*exit)(struct cfq_io_context *);
 };
 
 /*
@@ -73,7 +79,7 @@
  */
 struct io_context {
 	atomic_t refcount;
-	pid_t pid;
+	struct task_struct *task;
 
 	/*
 	 * For request batching
@@ -81,8 +87,6 @@
 	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;
 };
@@ -723,5 +727,24 @@
 #define MODULE_ALIAS_BLOCKDEV_MAJOR(major) \
 	MODULE_ALIAS("block-major-" __stringify(major) "-*")
 
+/*
+ * io prio stuff
+ */
+#define IOPRIO_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 != -1)
+		return task->ioprio;
+	else
+		return (task_nice(task) + 20) / 5;
+}
 
 #endif
===== 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	2004-12-13 18:45:45 +01:00
@@ -112,6 +112,7 @@
 	.proc_lock	= SPIN_LOCK_UNLOCKED,				\
 	.switch_lock	= SPIN_LOCK_UNLOCKED,				\
 	.journal_info	= NULL,						\
+	.ioprio		= -1,						\
 }
 
 
===== 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	2004-12-13 18:45:58 +01:00
@@ -523,6 +523,8 @@
 	struct list_head run_list;
 	prio_array_t *array;
 
+	int ioprio;
+
 	unsigned long sleep_avg;
 	long interactive_credit;
 	unsigned long long timestamp, last_ran;
@@ -740,6 +742,7 @@
 extern int task_nice(const task_t *p);
 extern int task_curr(const task_t *p);
 extern int idle_cpu(int cpu);
+extern unsigned long task_will_schedule_at(const task_t *p);
 
 void yield(void);
 
===== kernel/fork.c 1.223 vs edited =====
--- 1.223/kernel/fork.c	2004-12-03 08:25:48 +01:00
+++ edited/kernel/fork.c	2004-12-13 18:45:58 +01:00
@@ -948,6 +948,11 @@
 	INIT_LIST_HEAD(&p->ptrace_children);
 	INIT_LIST_HEAD(&p->ptrace_list);
 
+	/*
+	 * inherit ioprio
+	 */
+	p->ioprio = current->ioprio;
+
 	/* Need tasklist lock for parent etc handling! */
 	write_lock_irq(&tasklist_lock);
 
===== kernel/sched.c 1.383 vs edited =====
--- 1.383/kernel/sched.c	2004-12-08 09:53:23 +01:00
+++ edited/kernel/sched.c	2004-12-13 18:45:58 +01:00
@@ -813,6 +813,41 @@
 	return cpu_curr(task_cpu(p)) == p;
 }
 
+/**
+ * task_will_schedule_at - in how many ticks will the task run, most likely
+ * @p: the task in question
+ */
+unsigned long task_will_schedule_at(const task_t *p)
+{
+	/*
+	 * Task is executing right now
+	 */
+	if (task_curr(p))
+		return 0;
+
+	/*
+	 * Task is not executing but on a runqueue - try to guess
+	 * how much time it will take for it to run again, but using
+	 * the current task's remaining ticks. This is not accurate,
+	 * but a good guess. (We use the min() to avoid the small race
+	 * that is due to us dereferencing the current task without
+	 * locking)
+	 */
+	if (p->array)
+		return min(cpu_curr(task_cpu(p))->time_slice,
+					(unsigned int)MAX_SLEEP_AVG);
+
+	/*
+	 * for blocked tasks, return half of the average sleep time.
+	 * (because this is the average sleep-time we'll see if we
+	 * sample the period randomly.)
+	 */
+	return NS_TO_JIFFIES(p->sleep_avg) / 2;
+}
+
+EXPORT_SYMBOL_GPL(task_will_schedule_at);
+
+
 #ifdef CONFIG_SMP
 enum request_type {
 	REQ_MOVE_TASK,