IO-Uring Archive on lore.kernel.org
 help / color / Atom feed
* [PATCH] io-wq: shrink io_wq_work a bit
@ 2019-11-26 19:02 Jens Axboe
  0 siblings, 0 replies; only message in thread
From: Jens Axboe @ 2019-11-26 19:02 UTC (permalink / raw)
  To: io-uring

Currently we're using 40 bytes for the io_wq_work structure, and 16 of
those is the doubly link list node. We don't need doubly linked lists,
we always add to tail to keep things ordered, and any other use case
is list traversal with deletion. For the deletion case, we can easily
support any node deletion by keeping track of the previous entry.

This shrinks io_wq_work to 32 bytes, and subsequently io_kiock from
io_uring to 216 to 208 bytes.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

---

diff --git a/fs/io-wq.c b/fs/io-wq.c
index d3e8907cc182..91b85df0861e 100644
--- a/fs/io-wq.c
+++ b/fs/io-wq.c
@@ -84,7 +84,7 @@ enum {
 struct io_wqe {
 	struct {
 		spinlock_t lock;
-		struct list_head work_list;
+		struct io_wq_work_list work_list;
 		unsigned long hash_map;
 		unsigned flags;
 	} ____cacheline_aligned_in_smp;
@@ -236,7 +236,8 @@ static void io_worker_exit(struct io_worker *worker)
 static inline bool io_wqe_run_queue(struct io_wqe *wqe)
 	__must_hold(wqe->lock)
 {
-	if (!list_empty(&wqe->work_list) && !(wqe->flags & IO_WQE_FLAG_STALLED))
+	if (!wq_list_empty(&wqe->work_list) &&
+	    !(wqe->flags & IO_WQE_FLAG_STALLED))
 		return true;
 	return false;
 }
@@ -375,12 +376,15 @@ static bool __io_worker_idle(struct io_wqe *wqe, struct io_worker *worker)
 static struct io_wq_work *io_get_next_work(struct io_wqe *wqe, unsigned *hash)
 	__must_hold(wqe->lock)
 {
+	struct io_wq_work_node *node, *prev;
 	struct io_wq_work *work;
 
-	list_for_each_entry(work, &wqe->work_list, list) {
+	wq_list_for_each(node, prev, &wqe->work_list) {
+		work = container_of(node, struct io_wq_work, list);
+
 		/* not hashed, can run anytime */
 		if (!(work->flags & IO_WQ_WORK_HASHED)) {
-			list_del(&work->list);
+			wq_node_del(&wqe->work_list, node, prev);
 			return work;
 		}
 
@@ -388,7 +392,7 @@ static struct io_wq_work *io_get_next_work(struct io_wqe *wqe, unsigned *hash)
 		*hash = work->flags >> IO_WQ_HASH_SHIFT;
 		if (!(wqe->hash_map & BIT_ULL(*hash))) {
 			wqe->hash_map |= BIT_ULL(*hash);
-			list_del(&work->list);
+			wq_node_del(&wqe->work_list, node, prev);
 			return work;
 		}
 	}
@@ -416,7 +420,7 @@ static void io_worker_handle_work(struct io_worker *worker)
 		work = io_get_next_work(wqe, &hash);
 		if (work)
 			__io_worker_busy(wqe, worker, work);
-		else if (!list_empty(&wqe->work_list))
+		else if (!wq_list_empty(&wqe->work_list))
 			wqe->flags |= IO_WQE_FLAG_STALLED;
 
 		spin_unlock_irq(&wqe->lock);
@@ -526,7 +530,7 @@ static int io_wqe_worker(void *data)
 
 	if (test_bit(IO_WQ_BIT_EXIT, &wq->state)) {
 		spin_lock_irq(&wqe->lock);
-		if (!list_empty(&wqe->work_list))
+		if (!wq_list_empty(&wqe->work_list))
 			io_worker_handle_work(worker);
 		else
 			spin_unlock_irq(&wqe->lock);
@@ -714,7 +718,7 @@ static void io_wqe_enqueue(struct io_wqe *wqe, struct io_wq_work *work)
 	}
 
 	spin_lock_irqsave(&wqe->lock, flags);
-	list_add_tail(&work->list, &wqe->work_list);
+	wq_list_add_tail(&work->list, &wqe->work_list);
 	wqe->flags &= ~IO_WQE_FLAG_STALLED;
 	spin_unlock_irqrestore(&wqe->lock, flags);
 
@@ -829,14 +833,17 @@ static enum io_wq_cancel io_wqe_cancel_cb_work(struct io_wqe *wqe,
 		.cancel = cancel,
 		.caller_data = cancel_data,
 	};
+	struct io_wq_work_node *node, *prev;
 	struct io_wq_work *work;
 	unsigned long flags;
 	bool found = false;
 
 	spin_lock_irqsave(&wqe->lock, flags);
-	list_for_each_entry(work, &wqe->work_list, list) {
+	wq_list_for_each(node, prev, &wqe->work_list) {
+		work = container_of(node, struct io_wq_work, list);
+
 		if (cancel(work, cancel_data)) {
-			list_del(&work->list);
+			wq_node_del(&wqe->work_list, node, prev);
 			found = true;
 			break;
 		}
@@ -894,6 +901,7 @@ static bool io_wq_worker_cancel(struct io_worker *worker, void *data)
 static enum io_wq_cancel io_wqe_cancel_work(struct io_wqe *wqe,
 					    struct io_wq_work *cwork)
 {
+	struct io_wq_work_node *node, *prev;
 	struct io_wq_work *work;
 	unsigned long flags;
 	bool found = false;
@@ -906,9 +914,11 @@ static enum io_wq_cancel io_wqe_cancel_work(struct io_wqe *wqe,
 	 * no completion will be posted for it.
 	 */
 	spin_lock_irqsave(&wqe->lock, flags);
-	list_for_each_entry(work, &wqe->work_list, list) {
+	wq_list_for_each(node, prev, &wqe->work_list) {
+		work = container_of(node, struct io_wq_work, list);
+
 		if (work == cwork) {
-			list_del(&work->list);
+			wq_node_del(&wqe->work_list, node, prev);
 			found = true;
 			break;
 		}
@@ -1023,7 +1033,7 @@ struct io_wq *io_wq_create(unsigned bounded, struct io_wq_data *data)
 		wqe->node = node;
 		wqe->wq = wq;
 		spin_lock_init(&wqe->lock);
-		INIT_LIST_HEAD(&wqe->work_list);
+		INIT_WQ_LIST(&wqe->work_list);
 		INIT_HLIST_NULLS_HEAD(&wqe->free_list, 0);
 		INIT_HLIST_NULLS_HEAD(&wqe->busy_list, 1);
 		INIT_LIST_HEAD(&wqe->all_list);
diff --git a/fs/io-wq.h b/fs/io-wq.h
index 5cd8c7697e88..600e0158cba7 100644
--- a/fs/io-wq.h
+++ b/fs/io-wq.h
@@ -22,18 +22,60 @@ enum io_wq_cancel {
 	IO_WQ_CANCEL_NOTFOUND,	/* work not found */
 };
 
+struct io_wq_work_node {
+	struct io_wq_work_node *next;
+};
+
+struct io_wq_work_list {
+	struct io_wq_work_node *first;
+	struct io_wq_work_node *last;
+};
+
+static inline void wq_list_add_tail(struct io_wq_work_node *node,
+				    struct io_wq_work_list *list)
+{
+	if (!list->first) {
+		list->first = list->last = node;
+	} else {
+		list->last->next = node;
+		list->last = node;
+	}
+}
+
+static inline void wq_node_del(struct io_wq_work_list *list,
+			       struct io_wq_work_node *node,
+			       struct io_wq_work_node *prev)
+{
+	if (node == list->first)
+		list->first = node->next;
+	if (node == list->last)
+		list->last = prev;
+	if (prev)
+		prev->next = node->next;
+}
+
+#define wq_list_for_each(pos, prv, head)			\
+	for (pos = (head)->first, prv = NULL; pos; prv = pos, pos = (pos)->next)
+
+#define wq_list_empty(list)	((list)->first == NULL)
+#define INIT_WQ_LIST(list)	do {				\
+	(list)->first = NULL;					\
+	(list)->last = NULL;					\
+} while (0)
+
 struct io_wq_work {
 	union {
-		struct list_head list;
+		struct io_wq_work_node list;
 		void *data;
 	};
 	void (*func)(struct io_wq_work **);
-	unsigned flags;
 	struct files_struct *files;
+	unsigned flags;
 };
 
 #define INIT_IO_WORK(work, _func)			\
 	do {						\
+		(work)->list.next = NULL;		\
 		(work)->func = _func;			\
 		(work)->flags = 0;			\
 		(work)->files = NULL;			\

-- 
Jens Axboe


^ permalink raw reply	[flat|nested] only message in thread

only message in thread, back to index

Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2019-11-26 19:02 [PATCH] io-wq: shrink io_wq_work a bit Jens Axboe

IO-Uring Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/io-uring/0 io-uring/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 io-uring io-uring/ https://lore.kernel.org/io-uring \
		io-uring@vger.kernel.org
	public-inbox-index io-uring

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.io-uring


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git