From e944475e69849273ca8f1fe04a3ce81b5901d165 Mon Sep 17 00:00:00 2001
From: Jens Axboe <axboe@kernel.dk>
Date: Tue, 26 Nov 2019 15:02:04 -0700
Subject: [PATCH] io_uring: make poll->wait dynamically allocated

In the quest to bring io_kiocb down to 3 cachelines, this one does
the trick. Make the wait_queue_entry for the poll command come out
of kmalloc instead of embedding it in struct io_poll_iocb, as the
latter is the largest member of io_kiocb. Once we trim this down a
bit, we're back at a healthy 192 bytes for struct io_kiocb.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 fs/io_uring.c | 29 +++++++++++++++++------------
 1 file changed, 17 insertions(+), 12 deletions(-)

diff --git a/fs/io_uring.c b/fs/io_uring.c
index e44b0e01f1b5..2c2e8c25da01 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -291,7 +291,7 @@ struct io_poll_iocb {
 	__poll_t			events;
 	bool				done;
 	bool				canceled;
-	struct wait_queue_entry		wait;
+	struct wait_queue_entry		*wait;
 };
 
 struct io_timeout_data {
@@ -2030,8 +2030,8 @@ static void io_poll_remove_one(struct io_kiocb *req)
 
 	spin_lock(&poll->head->lock);
 	WRITE_ONCE(poll->canceled, true);
-	if (!list_empty(&poll->wait.entry)) {
-		list_del_init(&poll->wait.entry);
+	if (!list_empty(&poll->wait->entry)) {
+		list_del_init(&poll->wait->entry);
 		io_queue_async_work(req);
 	}
 	spin_unlock(&poll->head->lock);
@@ -2104,6 +2104,7 @@ static void io_poll_complete(struct io_kiocb *req, __poll_t mask, int error)
 	struct io_ring_ctx *ctx = req->ctx;
 
 	req->poll.done = true;
+	kfree(req->poll.wait);
 	if (error)
 		io_cqring_fill_event(req, error);
 	else
@@ -2141,7 +2142,7 @@ static void io_poll_complete_work(struct io_wq_work **workptr)
 	 */
 	spin_lock_irq(&ctx->completion_lock);
 	if (!mask && ret != -ECANCELED) {
-		add_wait_queue(poll->head, &poll->wait);
+		add_wait_queue(poll->head, poll->wait);
 		spin_unlock_irq(&ctx->completion_lock);
 		return;
 	}
@@ -2161,8 +2162,7 @@ static void io_poll_complete_work(struct io_wq_work **workptr)
 static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
 			void *key)
 {
-	struct io_poll_iocb *poll = container_of(wait, struct io_poll_iocb,
-							wait);
+	struct io_poll_iocb *poll = wait->private;
 	struct io_kiocb *req = container_of(poll, struct io_kiocb, poll);
 	struct io_ring_ctx *ctx = req->ctx;
 	__poll_t mask = key_to_poll(key);
@@ -2172,7 +2172,7 @@ static int io_poll_wake(struct wait_queue_entry *wait, unsigned mode, int sync,
 	if (mask && !(mask & poll->events))
 		return 0;
 
-	list_del_init(&poll->wait.entry);
+	list_del_init(&poll->wait->entry);
 
 	/*
 	 * Run completion inline if we can. We're using trylock here because
@@ -2213,7 +2213,7 @@ static void io_poll_queue_proc(struct file *file, struct wait_queue_head *head,
 
 	pt->error = 0;
 	pt->req->poll.head = head;
-	add_wait_queue(head, &pt->req->poll.wait);
+	add_wait_queue(head, pt->req->poll.wait);
 }
 
 static void io_poll_req_insert(struct io_kiocb *req)
@@ -2252,6 +2252,10 @@ static int io_poll_add(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	if (!poll->file)
 		return -EBADF;
 
+	poll->wait = kmalloc(sizeof(*poll->wait), GFP_KERNEL);
+	if (!poll->wait)
+		return -ENOMEM;
+
 	req->sqe = NULL;
 	INIT_IO_WORK(&req->work, io_poll_complete_work);
 	events = READ_ONCE(sqe->poll_events);
@@ -2268,8 +2272,9 @@ static int io_poll_add(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	ipt.error = -EINVAL; /* same as no support for IOCB_CMD_POLL */
 
 	/* initialized the list so that we can do list_empty checks */
-	INIT_LIST_HEAD(&poll->wait.entry);
-	init_waitqueue_func_entry(&poll->wait, io_poll_wake);
+	INIT_LIST_HEAD(&poll->wait->entry);
+	init_waitqueue_func_entry(poll->wait, io_poll_wake);
+	poll->wait->private = poll;
 
 	INIT_LIST_HEAD(&req->list);
 
@@ -2278,14 +2283,14 @@ static int io_poll_add(struct io_kiocb *req, const struct io_uring_sqe *sqe,
 	spin_lock_irq(&ctx->completion_lock);
 	if (likely(poll->head)) {
 		spin_lock(&poll->head->lock);
-		if (unlikely(list_empty(&poll->wait.entry))) {
+		if (unlikely(list_empty(&poll->wait->entry))) {
 			if (ipt.error)
 				cancel = true;
 			ipt.error = 0;
 			mask = 0;
 		}
 		if (mask || ipt.error)
-			list_del_init(&poll->wait.entry);
+			list_del_init(&poll->wait->entry);
 		else if (cancel)
 			WRITE_ONCE(poll->canceled, true);
 		else if (!poll->done) /* actually waiting for an event */