cfq-iosched: compact io_context radix_tree
Use small consequent indexes as radix tree keys instead of sparse cfqd address.
This change will reduce radix tree depth from 11 (6 for 32-bit hosts)
to 1 if host have <=64 disks under cfq control, or to 0 if there only one disk.
So, this patch save 10*560 bytes for each process (5*296 for 32-bit hosts)
For each cfqd allocate cic index from ida.
To unlink dead cic from tree without cfqd access store index into ->key.
(bit 0 -- dead mark, bits 1..30 -- index: ida produce id in range 0..2^31-1)
Signed-off-by: Konstantin Khlebnikov <khlebnikov@openvz.org>
Signed-off-by: Jens Axboe <jens.axboe@oracle.com>
diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
index 4076023..c72e5ac 100644
--- a/block/cfq-iosched.c
+++ b/block/cfq-iosched.c
@@ -64,6 +64,9 @@
static struct completion *ioc_gone;
static DEFINE_SPINLOCK(ioc_gone_lock);
+static DEFINE_SPINLOCK(cic_index_lock);
+static DEFINE_IDA(cic_index_ida);
+
#define CFQ_PRIO_LISTS IOPRIO_BE_NR
#define cfq_class_idle(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_IDLE)
#define cfq_class_rt(cfqq) ((cfqq)->ioprio_class == IOPRIO_CLASS_RT)
@@ -271,6 +274,7 @@
unsigned int cfq_latency;
unsigned int cfq_group_isolation;
+ unsigned int cic_index;
struct list_head cic_list;
/*
@@ -431,10 +435,11 @@
}
#define CIC_DEAD_KEY 1ul
+#define CIC_DEAD_INDEX_SHIFT 1
static inline void *cfqd_dead_key(struct cfq_data *cfqd)
{
- return (void *)((unsigned long) cfqd | CIC_DEAD_KEY);
+ return (void *)(cfqd->cic_index << CIC_DEAD_INDEX_SHIFT | CIC_DEAD_KEY);
}
static inline struct cfq_data *cic_to_cfqd(struct cfq_io_context *cic)
@@ -2532,7 +2537,7 @@
BUG_ON(!(dead_key & CIC_DEAD_KEY));
spin_lock_irqsave(&ioc->lock, flags);
- radix_tree_delete(&ioc->radix_root, dead_key & ~CIC_DEAD_KEY);
+ radix_tree_delete(&ioc->radix_root, dead_key >> CIC_DEAD_INDEX_SHIFT);
hlist_del_rcu(&cic->cic_list);
spin_unlock_irqrestore(&ioc->lock, flags);
@@ -2906,7 +2911,7 @@
BUG_ON(ioc->ioc_data == cic);
- radix_tree_delete(&ioc->radix_root, (unsigned long) cfqd);
+ radix_tree_delete(&ioc->radix_root, cfqd->cic_index);
hlist_del_rcu(&cic->cic_list);
spin_unlock_irqrestore(&ioc->lock, flags);
@@ -2934,7 +2939,7 @@
}
do {
- cic = radix_tree_lookup(&ioc->radix_root, (unsigned long) cfqd);
+ cic = radix_tree_lookup(&ioc->radix_root, cfqd->cic_index);
rcu_read_unlock();
if (!cic)
break;
@@ -2971,7 +2976,7 @@
spin_lock_irqsave(&ioc->lock, flags);
ret = radix_tree_insert(&ioc->radix_root,
- (unsigned long) cfqd, cic);
+ cfqd->cic_index, cic);
if (!ret)
hlist_add_head_rcu(&cic->cic_list, &ioc->cic_list);
spin_unlock_irqrestore(&ioc->lock, flags);
@@ -3723,10 +3728,32 @@
cfq_shutdown_timer_wq(cfqd);
+ spin_lock(&cic_index_lock);
+ ida_remove(&cic_index_ida, cfqd->cic_index);
+ spin_unlock(&cic_index_lock);
+
/* Wait for cfqg->blkg->key accessors to exit their grace periods. */
call_rcu(&cfqd->rcu, cfq_cfqd_free);
}
+static int cfq_alloc_cic_index(void)
+{
+ int index, error;
+
+ do {
+ if (!ida_pre_get(&cic_index_ida, GFP_KERNEL))
+ return -ENOMEM;
+
+ spin_lock(&cic_index_lock);
+ error = ida_get_new(&cic_index_ida, &index);
+ spin_unlock(&cic_index_lock);
+ if (error && error != -EAGAIN)
+ return error;
+ } while (error);
+
+ return index;
+}
+
static void *cfq_init_queue(struct request_queue *q)
{
struct cfq_data *cfqd;
@@ -3734,10 +3761,16 @@
struct cfq_group *cfqg;
struct cfq_rb_root *st;
+ i = cfq_alloc_cic_index();
+ if (i < 0)
+ return NULL;
+
cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL | __GFP_ZERO, q->node);
if (!cfqd)
return NULL;
+ cfqd->cic_index = i;
+
/* Init root service tree */
cfqd->grp_service_tree = CFQ_RB_ROOT;
@@ -3999,6 +4032,7 @@
*/
if (elv_ioc_count_read(cfq_ioc_count))
wait_for_completion(&all_gone);
+ ida_destroy(&cic_index_ida);
cfq_slab_kill();
}
