[PATCH] Make sure all block/io scheduler setups are node aware

Some were kmalloc_node(), some were still kmalloc(). Change them all to
kmalloc_node().

Signed-off-by: Jens Axboe <axboe@suse.de>

block/as-iosched.c

@@ -210,9 +210,9 @@ static struct as_io_context *alloc_as_io_context(void)
  * If the current task has no AS IO context then create one and initialise it.
  * Then take a ref on the task's io context and return it.
  */
-static struct io_context *as_get_io_context(void)
+static struct io_context *as_get_io_context(int node)
 {
-	struct io_context *ioc = get_io_context(GFP_ATOMIC);
+	struct io_context *ioc = get_io_context(GFP_ATOMIC, node);
 	if (ioc && !ioc->aic) {
 		ioc->aic = alloc_as_io_context();
 		if (!ioc->aic) {
@@ -1148,7 +1148,7 @@ static void as_add_request(request_queue_t *q, struct request *rq)
 
 	data_dir = rq_is_sync(rq);
 
-	rq->elevator_private = as_get_io_context();
+	rq->elevator_private = as_get_io_context(q->node);
 
 	if (RQ_IOC(rq)) {
 		as_update_iohist(ad, RQ_IOC(rq)->aic, rq);
@@ -1292,7 +1292,7 @@ static int as_may_queue(request_queue_t *q, int rw)
 	struct io_context *ioc;
 	if (ad->antic_status == ANTIC_WAIT_REQ ||
 			ad->antic_status == ANTIC_WAIT_NEXT) {
-		ioc = as_get_io_context();
+		ioc = as_get_io_context(q->node);
 		if (ad->io_context == ioc)
 			ret = ELV_MQUEUE_MUST;
 		put_io_context(ioc);

block/cfq-iosched.c

@@ -1148,8 +1148,9 @@ static void cfq_exit_io_context(struct io_context *ioc)
 static struct cfq_io_context *
 cfq_alloc_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
 {
-	struct cfq_io_context *cic = kmem_cache_alloc(cfq_ioc_pool, gfp_mask);
+	struct cfq_io_context *cic;
 
+	cic = kmem_cache_alloc_node(cfq_ioc_pool, gfp_mask, cfqd->queue->node);
 	if (cic) {
 		memset(cic, 0, sizeof(*cic));
 		cic->last_end_request = jiffies;
@@ -1277,11 +1278,11 @@ retry:
 			 * free memory.
 			 */
 			spin_unlock_irq(cfqd->queue->queue_lock);
-			new_cfqq = kmem_cache_alloc(cfq_pool, gfp_mask|__GFP_NOFAIL);
+			new_cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask|__GFP_NOFAIL, cfqd->queue->node);
 			spin_lock_irq(cfqd->queue->queue_lock);
 			goto retry;
 		} else {
-			cfqq = kmem_cache_alloc(cfq_pool, gfp_mask);
+			cfqq = kmem_cache_alloc_node(cfq_pool, gfp_mask, cfqd->queue->node);
 			if (!cfqq)
 				goto out;
 		}
@@ -1407,7 +1408,7 @@ cfq_get_io_context(struct cfq_data *cfqd, gfp_t gfp_mask)
 
 	might_sleep_if(gfp_mask & __GFP_WAIT);
 
-	ioc = get_io_context(gfp_mask);
+	ioc = get_io_context(gfp_mask, cfqd->queue->node);
 	if (!ioc)
 		return NULL;
 
@@ -1955,7 +1956,7 @@ static void *cfq_init_queue(request_queue_t *q, elevator_t *e)
 	struct cfq_data *cfqd;
 	int i;
 
-	cfqd = kmalloc(sizeof(*cfqd), GFP_KERNEL);
+	cfqd = kmalloc_node(sizeof(*cfqd), GFP_KERNEL, q->node);
 	if (!cfqd)
 		return NULL;
 
@@ -1970,7 +1971,7 @@ static void *cfq_init_queue(request_queue_t *q, elevator_t *e)
 	INIT_LIST_HEAD(&cfqd->empty_list);
 	INIT_LIST_HEAD(&cfqd->cic_list);
 
-	cfqd->cfq_hash = kmalloc(sizeof(struct hlist_head) * CFQ_QHASH_ENTRIES, GFP_KERNEL);
+	cfqd->cfq_hash = kmalloc_node(sizeof(struct hlist_head) * CFQ_QHASH_ENTRIES, GFP_KERNEL, q->node);
 	if (!cfqd->cfq_hash)
 		goto out_free;
 

block/elevator.c

@@ -161,12 +161,12 @@ __setup("elevator=", elevator_setup);
 
 static struct kobj_type elv_ktype;
 
-static elevator_t *elevator_alloc(struct elevator_type *e)
+static elevator_t *elevator_alloc(request_queue_t *q, struct elevator_type *e)
 {
 	elevator_t *eq;
 	int i;
 
-	eq = kmalloc(sizeof(elevator_t), GFP_KERNEL);
+	eq = kmalloc_node(sizeof(elevator_t), GFP_KERNEL, q->node);
 	if (unlikely(!eq))
 		goto err;
 
@@ -178,7 +178,8 @@ static elevator_t *elevator_alloc(struct elevator_type *e)
 	eq->kobj.ktype = &elv_ktype;
 	mutex_init(&eq->sysfs_lock);
 
-	eq->hash = kmalloc(sizeof(struct hlist_head) * ELV_HASH_ENTRIES, GFP_KERNEL);
+	eq->hash = kmalloc_node(sizeof(struct hlist_head) * ELV_HASH_ENTRIES,
+					GFP_KERNEL, q->node);
 	if (!eq->hash)
 		goto err;
 
@@ -224,7 +225,7 @@ int elevator_init(request_queue_t *q, char *name)
 		e = elevator_get("noop");
 	}
 
-	eq = elevator_alloc(e);
+	eq = elevator_alloc(q, e);
 	if (!eq)
 		return -ENOMEM;
 
@@ -987,7 +988,7 @@ static int elevator_switch(request_queue_t *q, struct elevator_type *new_e)
 	/*
 	 * Allocate new elevator
 	 */
-	e = elevator_alloc(new_e);
+	e = elevator_alloc(q, new_e);
 	if (!e)
 		return 0;
 

block/ll_rw_blk.c

@@ -39,6 +39,7 @@ static void blk_unplug_timeout(unsigned long data);
 static void drive_stat_acct(struct request *rq, int nr_sectors, int new_io);
 static void init_request_from_bio(struct request *req, struct bio *bio);
 static int __make_request(request_queue_t *q, struct bio *bio);
+static struct io_context *current_io_context(gfp_t gfp_flags, int node);
 
 /*
  * For the allocated request tables
@@ -2114,7 +2115,7 @@ static struct request *get_request(request_queue_t *q, int rw, struct bio *bio,
 
 	if (rl->count[rw]+1 >= queue_congestion_on_threshold(q)) {
 		if (rl->count[rw]+1 >= q->nr_requests) {
-			ioc = current_io_context(GFP_ATOMIC);
+			ioc = current_io_context(GFP_ATOMIC, q->node);
 			/*
 			 * The queue will fill after this allocation, so set
 			 * it as full, and mark this process as "batching".
@@ -2234,7 +2235,7 @@ static struct request *get_request_wait(request_queue_t *q, int rw,
 			 * up to a big batch of them for a small period time.
 			 * See ioc_batching, ioc_set_batching
 			 */
-			ioc = current_io_context(GFP_NOIO);
+			ioc = current_io_context(GFP_NOIO, q->node);
 			ioc_set_batching(q, ioc);
 
 			spin_lock_irq(q->queue_lock);
@@ -3641,7 +3642,7 @@ void exit_io_context(void)
  * but since the current task itself holds a reference, the context can be
  * used in general code, so long as it stays within `current` context.
  */
-struct io_context *current_io_context(gfp_t gfp_flags)
+static struct io_context *current_io_context(gfp_t gfp_flags, int node)
 {
 	struct task_struct *tsk = current;
 	struct io_context *ret;
@@ -3650,7 +3651,7 @@ struct io_context *current_io_context(gfp_t gfp_flags)
 	if (likely(ret))
 		return ret;
 
-	ret = kmem_cache_alloc(iocontext_cachep, gfp_flags);
+	ret = kmem_cache_alloc_node(iocontext_cachep, gfp_flags, node);
 	if (ret) {
 		atomic_set(&ret->refcount, 1);
 		ret->task = current;
@@ -3674,10 +3675,10 @@ EXPORT_SYMBOL(current_io_context);
  *
  * This is always called in the context of the task which submitted the I/O.
  */
-struct io_context *get_io_context(gfp_t gfp_flags)
+struct io_context *get_io_context(gfp_t gfp_flags, int node)
 {
 	struct io_context *ret;
-	ret = current_io_context(gfp_flags);
+	ret = current_io_context(gfp_flags, node);
 	if (likely(ret))
 		atomic_inc(&ret->refcount);
 	return ret;

block/noop-iosched.c

@@ -69,7 +69,7 @@ static void *noop_init_queue(request_queue_t *q, elevator_t *e)
 {
 	struct noop_data *nd;
 
-	nd = kmalloc(sizeof(*nd), GFP_KERNEL);
+	nd = kmalloc_node(sizeof(*nd), GFP_KERNEL, q->node);
 	if (!nd)
 		return NULL;
 	INIT_LIST_HEAD(&nd->queue);

include/linux/blkdev.h

@@ -104,8 +104,7 @@ struct io_context {
 
 void put_io_context(struct io_context *ioc);
 void exit_io_context(void);
-struct io_context *current_io_context(gfp_t gfp_flags);
-struct io_context *get_io_context(gfp_t gfp_flags);
+struct io_context *get_io_context(gfp_t gfp_flags, int node);
 void copy_io_context(struct io_context **pdst, struct io_context **psrc);
 void swap_io_context(struct io_context **ioc1, struct io_context **ioc2);