drm/radeon: fixup RV550 chip family
[sfrench/cifs-2.6.git] / block / cfq-iosched.c
index 13553e015d7203f244db2e411884c56cbee1bc26..0f962ecae91fb68345fe1d916791a78e06a2f2e8 100644 (file)
 /*
  * tunables
  */
-static const int cfq_quantum = 4;              /* max queue in one round of service */
+/* max queue in one round of service */
+static const int cfq_quantum = 4;
 static const int cfq_fifo_expire[2] = { HZ / 4, HZ / 8 };
-static const int cfq_back_max = 16 * 1024;     /* maximum backwards seek, in KiB */
-static const int cfq_back_penalty = 2;         /* penalty of a backwards seek */
-
+/* maximum backwards seek, in KiB */
+static const int cfq_back_max = 16 * 1024;
+/* penalty of a backwards seek */
+static const int cfq_back_penalty = 2;
 static const int cfq_slice_sync = HZ / 10;
 static int cfq_slice_async = HZ / 25;
 static const int cfq_slice_async_rq = 2;
 static int cfq_slice_idle = HZ / 125;
 
 /*
- * grace period before allowing idle class to get disk access
+ * offset from end of service tree
  */
-#define CFQ_IDLE_GRACE         (HZ / 10)
+#define CFQ_IDLE_DELAY         (HZ / 5)
 
 /*
  * below this threshold, we consider thinktime immediate
@@ -37,7 +39,8 @@ static int cfq_slice_idle = HZ / 125;
 
 #define CFQ_SLICE_SCALE                (5)
 
-#define RQ_CIC(rq)             ((struct cfq_io_context*)(rq)->elevator_private)
+#define RQ_CIC(rq)             \
+       ((struct cfq_io_context *) (rq)->elevator_private)
 #define RQ_CFQQ(rq)            ((rq)->elevator_private2)
 
 static struct kmem_cache *cfq_pool;
@@ -98,8 +101,6 @@ struct cfq_data {
        struct cfq_queue *async_cfqq[2][IOPRIO_BE_NR];
        struct cfq_queue *async_idle_cfqq;
 
-       struct timer_list idle_class_timer;
-
        sector_t last_position;
        unsigned long last_end_request;
 
@@ -173,15 +174,15 @@ enum cfqq_state_flags {
 #define CFQ_CFQQ_FNS(name)                                             \
 static inline void cfq_mark_cfqq_##name(struct cfq_queue *cfqq)                \
 {                                                                      \
-       cfqq->flags |= (1 << CFQ_CFQQ_FLAG_##name);                     \
+       (cfqq)->flags |= (1 << CFQ_CFQQ_FLAG_##name);                   \
 }                                                                      \
 static inline void cfq_clear_cfqq_##name(struct cfq_queue *cfqq)       \
 {                                                                      \
-       cfqq->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);                    \
+       (cfqq)->flags &= ~(1 << CFQ_CFQQ_FLAG_##name);                  \
 }                                                                      \
 static inline int cfq_cfqq_##name(const struct cfq_queue *cfqq)                \
 {                                                                      \
-       return (cfqq->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;        \
+       return ((cfqq)->flags & (1 << CFQ_CFQQ_FLAG_##name)) != 0;      \
 }
 
 CFQ_CFQQ_FNS(on_rr);
@@ -199,8 +200,8 @@ CFQ_CFQQ_FNS(sync);
 
 static void cfq_dispatch_insert(struct request_queue *, struct request *);
 static struct cfq_queue *cfq_get_queue(struct cfq_data *, int,
-                                      struct task_struct *, gfp_t);
-static struct cfq_io_context *cfq_cic_rb_lookup(struct cfq_data *,
+                                      struct io_context *, gfp_t);
+static struct cfq_io_context *cfq_cic_lookup(struct cfq_data *,
                                                struct io_context *);
 
 static inline struct cfq_queue *cic_to_cfqq(struct cfq_io_context *cic,
@@ -384,12 +385,15 @@ cfq_choose_req(struct cfq_data *cfqd, struct request *rq1, struct request *rq2)
 /*
  * The below is leftmost cache rbtree addon
  */
-static struct rb_node *cfq_rb_first(struct cfq_rb_root *root)
+static struct cfq_queue *cfq_rb_first(struct cfq_rb_root *root)
 {
        if (!root->left)
                root->left = rb_first(&root->rb);
 
-       return root->left;
+       if (root->left)
+               return rb_entry(root->left, struct cfq_queue, rb_node);
+
+       return NULL;
 }
 
 static void cfq_rb_erase(struct rb_node *n, struct cfq_rb_root *root)
@@ -446,12 +450,20 @@ static unsigned long cfq_slice_offset(struct cfq_data *cfqd,
 static void cfq_service_tree_add(struct cfq_data *cfqd,
                                    struct cfq_queue *cfqq, int add_front)
 {
-       struct rb_node **p = &cfqd->service_tree.rb.rb_node;
-       struct rb_node *parent = NULL;
+       struct rb_node **p, *parent;
+       struct cfq_queue *__cfqq;
        unsigned long rb_key;
        int left;
 
-       if (!add_front) {
+       if (cfq_class_idle(cfqq)) {
+               rb_key = CFQ_IDLE_DELAY;
+               parent = rb_last(&cfqd->service_tree.rb);
+               if (parent && parent != &cfqq->rb_node) {
+                       __cfqq = rb_entry(parent, struct cfq_queue, rb_node);
+                       rb_key += __cfqq->rb_key;
+               } else
+                       rb_key += jiffies;
+       } else if (!add_front) {
                rb_key = cfq_slice_offset(cfqd, cfqq) + jiffies;
                rb_key += cfqq->slice_resid;
                cfqq->slice_resid = 0;
@@ -469,8 +481,9 @@ static void cfq_service_tree_add(struct cfq_data *cfqd,
        }
 
        left = 1;
+       parent = NULL;
+       p = &cfqd->service_tree.rb.rb_node;
        while (*p) {
-               struct cfq_queue *__cfqq;
                struct rb_node **n;
 
                parent = *p;
@@ -524,8 +537,7 @@ static void cfq_resort_rr_list(struct cfq_data *cfqd, struct cfq_queue *cfqq)
  * add to busy list of queues for service, trying to be fair in ordering
  * the pending list according to last request service
  */
-static inline void
-cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+static void cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 {
        BUG_ON(cfq_cfqq_on_rr(cfqq));
        cfq_mark_cfqq_on_rr(cfqq);
@@ -538,8 +550,7 @@ cfq_add_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
  * Called when the cfqq no longer has requests pending, remove it from
  * the service tree.
  */
-static inline void
-cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+static void cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 {
        BUG_ON(!cfq_cfqq_on_rr(cfqq));
        cfq_clear_cfqq_on_rr(cfqq);
@@ -554,7 +565,7 @@ cfq_del_cfqq_rr(struct cfq_data *cfqd, struct cfq_queue *cfqq)
 /*
  * rb tree support functions
  */
-static inline void cfq_del_rq_rb(struct request *rq)
+static void cfq_del_rq_rb(struct request *rq)
 {
        struct cfq_queue *cfqq = RQ_CFQQ(rq);
        struct cfq_data *cfqd = cfqq->cfqd;
@@ -594,8 +605,7 @@ static void cfq_add_rq_rb(struct request *rq)
        BUG_ON(!cfqq->next_rq);
 }
 
-static inline void
-cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
+static void cfq_reposition_rq_rb(struct cfq_queue *cfqq, struct request *rq)
 {
        elv_rb_del(&cfqq->sort_list, rq);
        cfqq->queued[rq_is_sync(rq)]--;
@@ -609,7 +619,7 @@ cfq_find_rq_fmerge(struct cfq_data *cfqd, struct bio *bio)
        struct cfq_io_context *cic;
        struct cfq_queue *cfqq;
 
-       cic = cfq_cic_rb_lookup(cfqd, tsk->io_context);
+       cic = cfq_cic_lookup(cfqd, tsk->io_context);
        if (!cic)
                return NULL;
 
@@ -721,7 +731,7 @@ static int cfq_allow_merge(struct request_queue *q, struct request *rq,
         * Lookup the cfqq that this bio will be queued with. Allow
         * merge only if rq is queued there.
         */
-       cic = cfq_cic_rb_lookup(cfqd, current->io_context);
+       cic = cfq_cic_lookup(cfqd, current->io_context);
        if (!cic)
                return 0;
 
@@ -732,15 +742,10 @@ static int cfq_allow_merge(struct request_queue *q, struct request *rq,
        return 0;
 }
 
-static inline void
-__cfq_set_active_queue(struct cfq_data *cfqd, struct cfq_queue *cfqq)
+static void __cfq_set_active_queue(struct cfq_data *cfqd,
+                                  struct cfq_queue *cfqq)
 {
        if (cfqq) {
-               /*
-                * stop potential idle class queues waiting service
-                */
-               del_timer(&cfqd->idle_class_timer);
-
                cfqq->slice_end = 0;
                cfq_clear_cfqq_must_alloc_slice(cfqq);
                cfq_clear_cfqq_fifo_expire(cfqq);
@@ -789,47 +794,16 @@ static inline void cfq_slice_expired(struct cfq_data *cfqd, int timed_out)
                __cfq_slice_expired(cfqd, cfqq, timed_out);
 }
 
-static int start_idle_class_timer(struct cfq_data *cfqd)
-{
-       unsigned long end = cfqd->last_end_request + CFQ_IDLE_GRACE;
-       unsigned long now = jiffies;
-
-       if (time_before(now, end) &&
-           time_after_eq(now, cfqd->last_end_request)) {
-               mod_timer(&cfqd->idle_class_timer, end);
-               return 1;
-       }
-
-       return 0;
-}
-
 /*
  * Get next queue for service. Unless we have a queue preemption,
  * we'll simply select the first cfqq in the service tree.
  */
 static struct cfq_queue *cfq_get_next_queue(struct cfq_data *cfqd)
 {
-       struct cfq_queue *cfqq;
-       struct rb_node *n;
-
        if (RB_EMPTY_ROOT(&cfqd->service_tree.rb))
                return NULL;
 
-       n = cfq_rb_first(&cfqd->service_tree);
-       cfqq = rb_entry(n, struct cfq_queue, rb_node);
-
-       if (cfq_class_idle(cfqq)) {
-               /*
-                * if we have idle queues and no rt or be queues had
-                * pending requests, either allow immediate service if
-                * the grace period has passed or arm the idle grace
-                * timer
-                */
-               if (start_idle_class_timer(cfqd))
-                       cfqq = NULL;
-       }
-
-       return cfqq;
+       return cfq_rb_first(&cfqd->service_tree);
 }
 
 /*
@@ -895,7 +869,7 @@ static void cfq_arm_slice_timer(struct cfq_data *cfqd)
         * task has exited, don't wait
         */
        cic = cfqd->active_cic;
-       if (!cic || !cic->ioc->task)
+       if (!cic || !atomic_read(&cic->ioc->nr_tasks))
                return;
 
        /*
@@ -939,7 +913,7 @@ static void cfq_dispatch_insert(struct request_queue *q, struct request *rq)
 /*
  * return expired entry, or NULL to just start from scratch in rbtree
  */
-static inline struct request *cfq_check_fifo(struct cfq_queue *cfqq)
+static struct request *cfq_check_fifo(struct cfq_queue *cfqq)
 {
        struct cfq_data *cfqd = cfqq->cfqd;
        struct request *rq;
@@ -1034,7 +1008,8 @@ __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
                /*
                 * follow expired path, else get first next available
                 */
-               if ((rq = cfq_check_fifo(cfqq)) == NULL)
+               rq = cfq_check_fifo(cfqq);
+               if (rq == NULL)
                        rq = cfqq->next_rq;
 
                /*
@@ -1068,7 +1043,7 @@ __cfq_dispatch_requests(struct cfq_data *cfqd, struct cfq_queue *cfqq,
        return dispatched;
 }
 
-static inline int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
+static int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
 {
        int dispatched = 0;
 
@@ -1087,14 +1062,11 @@ static inline int __cfq_forced_dispatch_cfqq(struct cfq_queue *cfqq)
  */
 static int cfq_forced_dispatch(struct cfq_data *cfqd)
 {
+       struct cfq_queue *cfqq;
        int dispatched = 0;
-       struct rb_node *n;
-
-       while ((n = cfq_rb_first(&cfqd->service_tree)) != NULL) {
-               struct cfq_queue *cfqq = rb_entry(n, struct cfq_queue, rb_node);
 
+       while ((cfqq = cfq_rb_first(&cfqd->service_tree)) != NULL)
                dispatched += __cfq_forced_dispatch_cfqq(cfqq);
-       }
 
        cfq_slice_expired(cfqd, 0);
 
@@ -1170,20 +1142,51 @@ static void cfq_put_queue(struct cfq_queue *cfqq)
        kmem_cache_free(cfq_pool, cfqq);
 }
 
-static void cfq_free_io_context(struct io_context *ioc)
+/*
+ * Call func for each cic attached to this ioc. Returns number of cic's seen.
+ */
+static unsigned int
+call_for_each_cic(struct io_context *ioc,
+                 void (*func)(struct io_context *, struct cfq_io_context *))
 {
-       struct cfq_io_context *__cic;
-       struct rb_node *n;
-       int freed = 0;
-
-       ioc->ioc_data = NULL;
+       struct cfq_io_context *cic;
+       struct hlist_node *n;
+       int called = 0;
 
-       while ((n = rb_first(&ioc->cic_root)) != NULL) {
-               __cic = rb_entry(n, struct cfq_io_context, rb_node);
-               rb_erase(&__cic->rb_node, &ioc->cic_root);
-               kmem_cache_free(cfq_ioc_pool, __cic);
-               freed++;
+       rcu_read_lock();
+       hlist_for_each_entry_rcu(cic, n, &ioc->cic_list, cic_list) {
+               func(ioc, cic);
+               called++;
        }
+       rcu_read_unlock();
+
+       return called;
+}
+
+static void cic_free_func(struct io_context *ioc, struct cfq_io_context *cic)
+{
+       unsigned long flags;
+
+       BUG_ON(!cic->dead_key);
+
+       spin_lock_irqsave(&ioc->lock, flags);
+       radix_tree_delete(&ioc->radix_root, cic->dead_key);
+       hlist_del_rcu(&cic->cic_list);
+       spin_unlock_irqrestore(&ioc->lock, flags);
+
+       kmem_cache_free(cfq_ioc_pool, cic);
+}
+
+static void cfq_free_io_context(struct io_context *ioc)
+{
+       int freed;
+
+       /*
+        * ioc->refcount is zero here, so no more cic's are allowed to be
+        * linked into this ioc. So it should be ok to iterate over the known
+        * list, we will see all cic's since no new ones are added.
+        */
+       freed = call_for_each_cic(ioc, cic_free_func);
 
        elv_ioc_count_mod(ioc_count, -freed);
 
@@ -1205,7 +1208,12 @@ static void __cfq_exit_single_io_context(struct cfq_data *cfqd,
                                         struct cfq_io_context *cic)
 {
        list_del_init(&cic->queue_list);
+
+       /*
+        * Make sure key == NULL is seen for dead queues
+        */
        smp_wmb();
+       cic->dead_key = (unsigned long) cic->key;
        cic->key = NULL;
 
        if (cic->cfqq[ASYNC]) {
@@ -1219,16 +1227,18 @@ static void __cfq_exit_single_io_context(struct cfq_data *cfqd,
        }
 }
 
-static void cfq_exit_single_io_context(struct cfq_io_context *cic)
+static void cfq_exit_single_io_context(struct io_context *ioc,
+                                      struct cfq_io_context *cic)
 {
        struct cfq_data *cfqd = cic->key;
 
        if (cfqd) {
                struct request_queue *q = cfqd->queue;
+               unsigned long flags;
 
-               spin_lock_irq(q->queue_lock);
+               spin_lock_irqsave(q->queue_lock, flags);
                __cfq_exit_single_io_context(cfqd, cic);
-               spin_unlock_irq(q->queue_lock);
+               spin_unlock_irqrestore(q->queue_lock, flags);
        }
 }
 
@@ -1238,21 +1248,8 @@ static void cfq_exit_single_io_context(struct cfq_io_context *cic)
  */
 static void cfq_exit_io_context(struct io_context *ioc)
 {
-       struct cfq_io_context *__cic;
-       struct rb_node *n;
-
-       ioc->ioc_data = NULL;
-
-       /*
-        * put the reference this task is holding to the various queues
-        */
-       n = rb_first(&ioc->cic_root);
-       while (n != NULL) {
-               __cic = rb_entry(n, struct cfq_io_context, rb_node);
-
-               cfq_exit_single_io_context(__cic);
-               n = rb_next(n);
-       }
+       rcu_assign_pointer(ioc->ioc_data, NULL);
+       call_for_each_cic(ioc, cfq_exit_single_io_context);
 }
 
 static struct cfq_io_context *
@@ -1265,6 +1262,7 @@ cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
        if (cic) {
                cic->last_end_request = jiffies;
                INIT_LIST_HEAD(&cic->queue_list);
+               INIT_HLIST_NODE(&cic->cic_list);
                cic->dtor = cfq_free_io_context;
                cic->exit = cfq_exit_io_context;
                elv_ioc_count_inc(ioc_count);
@@ -1273,7 +1271,7 @@ cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
        return cic;
 }
 
-static void cfq_init_prio_data(struct cfq_queue *cfqq)
+static void cfq_init_prio_data(struct cfq_queue *cfqq, struct io_context *ioc)
 {
        struct task_struct *tsk = current;
        int ioprio_class;
@@ -1281,30 +1279,30 @@ static void cfq_init_prio_data(struct cfq_queue *cfqq)
        if (!cfq_cfqq_prio_changed(cfqq))
                return;
 
-       ioprio_class = IOPRIO_PRIO_CLASS(tsk->ioprio);
+       ioprio_class = IOPRIO_PRIO_CLASS(ioc->ioprio);
        switch (ioprio_class) {
-               default:
-                       printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
-               case IOPRIO_CLASS_NONE:
-                       /*
-                        * no prio set, place us in the middle of the BE classes
-                        */
-                       cfqq->ioprio = task_nice_ioprio(tsk);
-                       cfqq->ioprio_class = IOPRIO_CLASS_BE;
-                       break;
-               case IOPRIO_CLASS_RT:
-                       cfqq->ioprio = task_ioprio(tsk);
-                       cfqq->ioprio_class = IOPRIO_CLASS_RT;
-                       break;
-               case IOPRIO_CLASS_BE:
-                       cfqq->ioprio = task_ioprio(tsk);
-                       cfqq->ioprio_class = IOPRIO_CLASS_BE;
-                       break;
-               case IOPRIO_CLASS_IDLE:
-                       cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
-                       cfqq->ioprio = 7;
-                       cfq_clear_cfqq_idle_window(cfqq);
-                       break;
+       default:
+               printk(KERN_ERR "cfq: bad prio %x\n", ioprio_class);
+       case IOPRIO_CLASS_NONE:
+               /*
+                * no prio set, place us in the middle of the BE classes
+                */
+               cfqq->ioprio = task_nice_ioprio(tsk);
+               cfqq->ioprio_class = IOPRIO_CLASS_BE;
+               break;
+       case IOPRIO_CLASS_RT:
+               cfqq->ioprio = task_ioprio(ioc);
+               cfqq->ioprio_class = IOPRIO_CLASS_RT;
+               break;
+       case IOPRIO_CLASS_BE:
+               cfqq->ioprio = task_ioprio(ioc);
+               cfqq->ioprio_class = IOPRIO_CLASS_BE;
+               break;
+       case IOPRIO_CLASS_IDLE:
+               cfqq->ioprio_class = IOPRIO_CLASS_IDLE;
+               cfqq->ioprio = 7;
+               cfq_clear_cfqq_idle_window(cfqq);
+               break;
        }
 
        /*
@@ -1316,7 +1314,7 @@ static void cfq_init_prio_data(struct cfq_queue *cfqq)
        cfq_clear_cfqq_prio_changed(cfqq);
 }
 
-static inline void changed_ioprio(struct cfq_io_context *cic)
+static void changed_ioprio(struct io_context *ioc, struct cfq_io_context *cic)
 {
        struct cfq_data *cfqd = cic->key;
        struct cfq_queue *cfqq;
@@ -1330,8 +1328,7 @@ static inline void changed_ioprio(struct cfq_io_context *cic)
        cfqq = cic->cfqq[ASYNC];
        if (cfqq) {
                struct cfq_queue *new_cfqq;
-               new_cfqq = cfq_get_queue(cfqd, ASYNC, cic->ioc->task,
-                                        GFP_ATOMIC);
+               new_cfqq = cfq_get_queue(cfqd, ASYNC, cic->ioc, GFP_ATOMIC);
                if (new_cfqq) {
                        cic->cfqq[ASYNC] = new_cfqq;
                        cfq_put_queue(cfqq);
@@ -1347,29 +1344,19 @@ static inline void changed_ioprio(struct cfq_io_context *cic)
 
 static void cfq_ioc_set_ioprio(struct io_context *ioc)
 {
-       struct cfq_io_context *cic;
-       struct rb_node *n;
-
+       call_for_each_cic(ioc, changed_ioprio);
        ioc->ioprio_changed = 0;
-
-       n = rb_first(&ioc->cic_root);
-       while (n != NULL) {
-               cic = rb_entry(n, struct cfq_io_context, rb_node);
-
-               changed_ioprio(cic);
-               n = rb_next(n);
-       }
 }
 
 static struct cfq_queue *
 cfq_find_alloc_queue(struct cfq_data *cfqd, int is_sync,
-                    struct task_struct *tsk, gfp_t gfp_mask)
+                    struct io_context *ioc, gfp_t gfp_mask)
 {
        struct cfq_queue *cfqq, *new_cfqq = NULL;
        struct cfq_io_context *cic;
 
 retry:
-       cic = cfq_cic_rb_lookup(cfqd, tsk->io_context);
+       cic = cfq_cic_lookup(cfqd, ioc);
        /* cic always exists here */
        cfqq = cic_to_cfqq(cic, is_sync);
 
@@ -1404,15 +1391,16 @@ retry:
                atomic_set(&cfqq->ref, 0);
                cfqq->cfqd = cfqd;
 
-               if (is_sync) {
-                       cfq_mark_cfqq_idle_window(cfqq);
-                       cfq_mark_cfqq_sync(cfqq);
-               }
-
                cfq_mark_cfqq_prio_changed(cfqq);
                cfq_mark_cfqq_queue_new(cfqq);
 
-               cfq_init_prio_data(cfqq);
+               cfq_init_prio_data(cfqq, ioc);
+
+               if (is_sync) {
+                       if (!cfq_class_idle(cfqq))
+                               cfq_mark_cfqq_idle_window(cfqq);
+                       cfq_mark_cfqq_sync(cfqq);
+               }
        }
 
        if (new_cfqq)
@@ -1426,7 +1414,7 @@ out:
 static struct cfq_queue **
 cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio)
 {
-       switch(ioprio_class) {
+       switch (ioprio_class) {
        case IOPRIO_CLASS_RT:
                return &cfqd->async_cfqq[0][ioprio];
        case IOPRIO_CLASS_BE:
@@ -1439,11 +1427,11 @@ cfq_async_queue_prio(struct cfq_data *cfqd, int ioprio_class, int ioprio)
 }
 
 static struct cfq_queue *
-cfq_get_queue(struct cfq_data *cfqd, int is_sync, struct task_struct *tsk,
+cfq_get_queue(struct cfq_data *cfqd, int is_sync, struct io_context *ioc,
              gfp_t gfp_mask)
 {
-       const int ioprio = task_ioprio(tsk);
-       const int ioprio_class = task_ioprio_class(tsk);
+       const int ioprio = task_ioprio(ioc);
+       const int ioprio_class = task_ioprio_class(ioc);
        struct cfq_queue **async_cfqq = NULL;
        struct cfq_queue *cfqq = NULL;
 
@@ -1453,7 +1441,7 @@ cfq_get_queue(struct cfq_data *cfqd, int is_sync, struct task_struct *tsk,
        }
 
        if (!cfqq) {
-               cfqq = cfq_find_alloc_queue(cfqd, is_sync, tsk, gfp_mask);
+               cfqq = cfq_find_alloc_queue(cfqd, is_sync, ioc, gfp_mask);
                if (!cfqq)
                        return NULL;
        }
@@ -1470,28 +1458,43 @@ cfq_get_queue(struct cfq_data *cfqd, int is_sync, struct task_struct *tsk,
        return cfqq;
 }
 
+static void cfq_cic_free(struct cfq_io_context *cic)
+{
+       kmem_cache_free(cfq_ioc_pool, cic);
+       elv_ioc_count_dec(ioc_count);
+
+       if (ioc_gone && !elv_ioc_count_read(ioc_count))
+               complete(ioc_gone);
+}
+
 /*
  * We drop cfq io contexts lazily, so we may find a dead one.
  */
 static void
-cfq_drop_dead_cic(struct io_context *ioc, struct cfq_io_context *cic)
+cfq_drop_dead_cic(struct cfq_data *cfqd, struct io_context *ioc,
+                 struct cfq_io_context *cic)
 {
+       unsigned long flags;
+
        WARN_ON(!list_empty(&cic->queue_list));
 
+       spin_lock_irqsave(&ioc->lock, flags);
+
        if (ioc->ioc_data == cic)
-               ioc->ioc_data = NULL;
+               rcu_assign_pointer(ioc->ioc_data, NULL);
 
-       rb_erase(&cic->rb_node, &ioc->cic_root);
-       kmem_cache_free(cfq_ioc_pool, cic);
-       elv_ioc_count_dec(ioc_count);
+       radix_tree_delete(&ioc->radix_root, (unsigned long) cfqd);
+       hlist_del_rcu(&cic->cic_list);
+       spin_unlock_irqrestore(&ioc->lock, flags);
+
+       cfq_cic_free(cic);
 }
 
 static struct cfq_io_context *
-cfq_cic_rb_lookup(struct cfq_data *cfqd, struct io_context *ioc)
+cfq_cic_lookup(struct cfq_data *cfqd, struct io_context *ioc)
 {
-       struct rb_node *n;
        struct cfq_io_context *cic;
-       void *k, *key = cfqd;
+       void *k;
 
        if (unlikely(!ioc))
                return NULL;
@@ -1499,74 +1502,66 @@ cfq_cic_rb_lookup(struct cfq_data *cfqd, struct io_context *ioc)
        /*
         * we maintain a last-hit cache, to avoid browsing over the tree
         */
-       cic = ioc->ioc_data;
+       cic = rcu_dereference(ioc->ioc_data);
        if (cic && cic->key == cfqd)
                return cic;
 
-restart:
-       n = ioc->cic_root.rb_node;
-       while (n) {
-               cic = rb_entry(n, struct cfq_io_context, rb_node);
+       do {
+               rcu_read_lock();
+               cic = radix_tree_lookup(&ioc->radix_root, (unsigned long) cfqd);
+               rcu_read_unlock();
+               if (!cic)
+                       break;
                /* ->key must be copied to avoid race with cfq_exit_queue() */
                k = cic->key;
                if (unlikely(!k)) {
-                       cfq_drop_dead_cic(ioc, cic);
-                       goto restart;
+                       cfq_drop_dead_cic(cfqd, ioc, cic);
+                       continue;
                }
 
-               if (key < k)
-                       n = n->rb_left;
-               else if (key > k)
-                       n = n->rb_right;
-               else {
-                       ioc->ioc_data = cic;
-                       return cic;
-               }
-       }
+               rcu_assign_pointer(ioc->ioc_data, cic);
+               break;
+       } while (1);
 
-       return NULL;
+       return cic;
 }
 
-static inline void
-cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc,
-            struct cfq_io_context *cic)
+/*
+ * Add cic into ioc, using cfqd as the search key. This enables us to lookup
+ * the process specific cfq io context when entered from the block layer.
+ * Also adds the cic to a per-cfqd list, used when this queue is removed.
+ */
+static int cfq_cic_link(struct cfq_data *cfqd, struct io_context *ioc,
+                       struct cfq_io_context *cic, gfp_t gfp_mask)
 {
-       struct rb_node **p;
-       struct rb_node *parent;
-       struct cfq_io_context *__cic;
        unsigned long flags;
-       void *k;
-
-       cic->ioc = ioc;
-       cic->key = cfqd;
-
-restart:
-       parent = NULL;
-       p = &ioc->cic_root.rb_node;
-       while (*p) {
-               parent = *p;
-               __cic = rb_entry(parent, struct cfq_io_context, rb_node);
-               /* ->key must be copied to avoid race with cfq_exit_queue() */
-               k = __cic->key;
-               if (unlikely(!k)) {
-                       cfq_drop_dead_cic(ioc, __cic);
-                       goto restart;
+       int ret;
+
+       ret = radix_tree_preload(gfp_mask);
+       if (!ret) {
+               cic->ioc = ioc;
+               cic->key = cfqd;
+
+               spin_lock_irqsave(&ioc->lock, flags);
+               ret = radix_tree_insert(&ioc->radix_root,
+                                               (unsigned long) cfqd, cic);
+               if (!ret)
+                       hlist_add_head_rcu(&cic->cic_list, &ioc->cic_list);
+               spin_unlock_irqrestore(&ioc->lock, flags);
+
+               radix_tree_preload_end();
+
+               if (!ret) {
+                       spin_lock_irqsave(cfqd->queue->queue_lock, flags);
+                       list_add(&cic->queue_list, &cfqd->cic_list);
+                       spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
                }
-
-               if (cic->key < k)
-                       p = &(*p)->rb_left;
-               else if (cic->key > k)
-                       p = &(*p)->rb_right;
-               else
-                       BUG();
        }
 
-       rb_link_node(&cic->rb_node, parent, p);
-       rb_insert_color(&cic->rb_node, &ioc->cic_root);
+       if (ret)
+               printk(KERN_ERR "cfq: cic link failed!\n");
 
-       spin_lock_irqsave(cfqd->queue->queue_lock, flags);
-       list_add(&cic->queue_list, &cfqd->cic_list);
-       spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
+       return ret;
 }
 
 /*
@@ -1586,7 +1581,7 @@ cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
        if (!ioc)
                return NULL;
 
-       cic = cfq_cic_rb_lookup(cfqd, ioc);
+       cic = cfq_cic_lookup(cfqd, ioc);
        if (cic)
                goto out;
 
@@ -1594,13 +1589,17 @@ cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
        if (cic == NULL)
                goto err;
 
-       cfq_cic_link(cfqd, ioc, cic);
+       if (cfq_cic_link(cfqd, ioc, cic, gfp_mask))
+               goto err_free;
+
 out:
        smp_read_barrier_depends();
        if (unlikely(ioc->ioprio_changed))
                cfq_ioc_set_ioprio(ioc);
 
        return cic;
+err_free:
+       cfq_cic_free(cic);
 err:
        put_io_context(ioc);
        return NULL;
@@ -1655,12 +1654,15 @@ cfq_update_idle_window(struct cfq_data *cfqd, struct cfq_queue *cfqq,
 {
        int enable_idle;
 
-       if (!cfq_cfqq_sync(cfqq))
+       /*
+        * Don't idle for async or idle io prio class
+        */
+       if (!cfq_cfqq_sync(cfqq) || cfq_class_idle(cfqq))
                return;
 
        enable_idle = cfq_cfqq_idle_window(cfqq);
 
-       if (!cic->ioc->task || !cfqd->cfq_slice_idle ||
+       if (!atomic_read(&cic->ioc->nr_tasks) || !cfqd->cfq_slice_idle ||
            (cfqd->hw_tag && CIC_SEEKY(cic)))
                enable_idle = 0;
        else if (sample_valid(cic->ttime_samples)) {
@@ -1793,7 +1795,7 @@ static void cfq_insert_request(struct request_queue *q, struct request *rq)
        struct cfq_data *cfqd = q->elevator->elevator_data;
        struct cfq_queue *cfqq = RQ_CFQQ(rq);
 
-       cfq_init_prio_data(cfqq);
+       cfq_init_prio_data(cfqq, RQ_CIC(rq)->ioc);
 
        cfq_add_rq_rb(rq);
 
@@ -1834,7 +1836,7 @@ static void cfq_completed_request(struct request_queue *q, struct request *rq)
                        cfq_set_prio_slice(cfqd, cfqq);
                        cfq_clear_cfqq_slice_new(cfqq);
                }
-               if (cfq_slice_used(cfqq))
+               if (cfq_slice_used(cfqq) || cfq_class_idle(cfqq))
                        cfq_slice_expired(cfqd, 1);
                else if (sync && RB_EMPTY_ROOT(&cfqq->sort_list))
                        cfq_arm_slice_timer(cfqd);
@@ -1894,13 +1896,13 @@ static int cfq_may_queue(struct request_queue *q, int rw)
         * so just lookup a possibly existing queue, or return 'may queue'
         * if that fails
         */
-       cic = cfq_cic_rb_lookup(cfqd, tsk->io_context);
+       cic = cfq_cic_lookup(cfqd, tsk->io_context);
        if (!cic)
                return ELV_MQUEUE_MAY;
 
        cfqq = cic_to_cfqq(cic, rw & REQ_RW_SYNC);
        if (cfqq) {
-               cfq_init_prio_data(cfqq);
+               cfq_init_prio_data(cfqq, cic->ioc);
                cfq_prio_boost(cfqq);
 
                return __cfq_may_queue(cfqq);
@@ -1938,7 +1940,6 @@ static int
 cfq_set_request(struct request_queue *q, struct request *rq, gfp_t gfp_mask)
 {
        struct cfq_data *cfqd = q->elevator->elevator_data;
-       struct task_struct *tsk = current;
        struct cfq_io_context *cic;
        const int rw = rq_data_dir(rq);
        const int is_sync = rq_is_sync(rq);
@@ -1956,7 +1957,7 @@ cfq_set_request(struct request_queue *q, struct request *rq, gfp_t gfp_mask)
 
        cfqq = cic_to_cfqq(cic, is_sync);
        if (!cfqq) {
-               cfqq = cfq_get_queue(cfqd, is_sync, tsk, gfp_mask);
+               cfqq = cfq_get_queue(cfqd, is_sync, cic->ioc, gfp_mask);
 
                if (!cfqq)
                        goto queue_fail;
@@ -2007,7 +2008,8 @@ static void cfq_idle_slice_timer(unsigned long data)
 
        spin_lock_irqsave(cfqd->queue->queue_lock, flags);
 
-       if ((cfqq = cfqd->active_queue) != NULL) {
+       cfqq = cfqd->active_queue;
+       if (cfqq) {
                timed_out = 0;
 
                /*
@@ -2039,29 +2041,9 @@ 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 (!start_idle_class_timer(cfqd))
-               cfq_schedule_dispatch(cfqd);
-
-       spin_unlock_irqrestore(cfqd->queue->queue_lock, flags);
-}
-
 static void cfq_shutdown_timer_wq(struct cfq_data *cfqd)
 {
        del_timer_sync(&cfqd->idle_slice_timer);
-       del_timer_sync(&cfqd->idle_class_timer);
        kblockd_flush_work(&cfqd->unplug_work);
 }
 
@@ -2126,10 +2108,6 @@ static void *cfq_init_queue(struct request_queue *q)
        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);
 
        cfqd->last_end_request = jiffies;
@@ -2160,7 +2138,7 @@ static int __init cfq_slab_setup(void)
        if (!cfq_pool)
                goto fail;
 
-       cfq_ioc_pool = KMEM_CACHE(cfq_io_context, 0);
+       cfq_ioc_pool = KMEM_CACHE(cfq_io_context, SLAB_DESTROY_BY_RCU);
        if (!cfq_ioc_pool)
                goto fail;
 
@@ -2225,14 +2203,18 @@ static ssize_t __FUNC(elevator_t *e, const char *page, size_t count)    \
        return ret;                                                     \
 }
 STORE_FUNCTION(cfq_quantum_store, &cfqd->cfq_quantum, 1, UINT_MAX, 0);
-STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1, UINT_MAX, 1);
-STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1, UINT_MAX, 1);
+STORE_FUNCTION(cfq_fifo_expire_sync_store, &cfqd->cfq_fifo_expire[1], 1,
+               UINT_MAX, 1);
+STORE_FUNCTION(cfq_fifo_expire_async_store, &cfqd->cfq_fifo_expire[0], 1,
+               UINT_MAX, 1);
 STORE_FUNCTION(cfq_back_seek_max_store, &cfqd->cfq_back_max, 0, UINT_MAX, 0);
-STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1, UINT_MAX, 0);
+STORE_FUNCTION(cfq_back_seek_penalty_store, &cfqd->cfq_back_penalty, 1,
+               UINT_MAX, 0);
 STORE_FUNCTION(cfq_slice_idle_store, &cfqd->cfq_slice_idle, 0, UINT_MAX, 1);
 STORE_FUNCTION(cfq_slice_sync_store, &cfqd->cfq_slice[1], 1, UINT_MAX, 1);
 STORE_FUNCTION(cfq_slice_async_store, &cfqd->cfq_slice[0], 1, UINT_MAX, 1);
-STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1, UINT_MAX, 0);
+STORE_FUNCTION(cfq_slice_async_rq_store, &cfqd->cfq_slice_async_rq, 1,
+               UINT_MAX, 0);
 #undef STORE_FUNCTION
 
 #define CFQ_ATTR(name) \