[PATCH v2] io_uring: Avoid polling configuration errors

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



If user doesn't configured poll queue but do the polled IO, it will get
a low performeance than regular poll, but 100% CPU uage. And there's no
prompts or warnings.

This patch aims to help users more easily verify their configurations
correctly, avoiding time and performance losses.

--
changes from v1:
- without disrupting the original I/O process.
- move judgement from block to io_uring.

Signed-off-by: hexue <xue01.he@xxxxxxxxxxx>
---
 include/linux/io_uring_types.h |  1 +
 io_uring/io_uring.c            |  4 +++-
 io_uring/rw.c                  | 18 ++++++++++++++++++
 3 files changed, 22 insertions(+), 1 deletion(-)

diff --git a/include/linux/io_uring_types.h b/include/linux/io_uring_types.h
index 91224bbcfa73..270c3edbbf21 100644
--- a/include/linux/io_uring_types.h
+++ b/include/linux/io_uring_types.h
@@ -428,6 +428,7 @@ struct io_ring_ctx {
 	unsigned short			n_sqe_pages;
 	struct page			**ring_pages;
 	struct page			**sqe_pages;
+	bool				check_poll_queue;
 };
 
 struct io_tw_state {
diff --git a/io_uring/io_uring.c b/io_uring/io_uring.c
index 816e93e7f949..1b45b4c52ae0 100644
--- a/io_uring/io_uring.c
+++ b/io_uring/io_uring.c
@@ -3463,8 +3463,10 @@ static __cold int io_uring_create(unsigned entries, struct io_uring_params *p,
 	    !(ctx->flags & IORING_SETUP_SQPOLL))
 		ctx->task_complete = true;
 
-	if (ctx->task_complete || (ctx->flags & IORING_SETUP_IOPOLL))
+	if (ctx->task_complete || (ctx->flags & IORING_SETUP_IOPOLL)) {
 		ctx->lockless_cq = true;
+		ctx->check_poll_queue = false;
+	}
 
 	/*
 	 * lazy poll_wq activation relies on ->task_complete for synchronisation
diff --git a/io_uring/rw.c b/io_uring/rw.c
index 1a2128459cb4..20f417152a17 100644
--- a/io_uring/rw.c
+++ b/io_uring/rw.c
@@ -772,6 +772,23 @@ static bool need_complete_io(struct io_kiocb *req)
 		S_ISBLK(file_inode(req->file)->i_mode);
 }
 
+static void check_poll_queue_state(struct io_ring_ctx *ctx, struct io_kiocb *req)
+{
+	if (!ctx->check_poll_queue) {
+		struct block_device *bdev;
+		struct request_queue *q;
+		struct inode *inode = req->file->f_inode;
+
+		if (inode->i_rdev) {
+			bdev = blkdev_get_no_open(inode->i_rdev);
+			q = bdev->bd_queue;
+			if (!test_bit(QUEUE_FLAG_POLL, &q->queue_flags))
+				pr_warn("the device does't configured with poll queues\n");
+		}
+		ctx->check_poll_queue = true;
+	}
+}
+
 static int io_rw_init_file(struct io_kiocb *req, fmode_t mode)
 {
 	struct io_rw *rw = io_kiocb_to_cmd(req, struct io_rw);
@@ -804,6 +821,7 @@ static int io_rw_init_file(struct io_kiocb *req, fmode_t mode)
 	if (ctx->flags & IORING_SETUP_IOPOLL) {
 		if (!(kiocb->ki_flags & IOCB_DIRECT) || !file->f_op->iopoll)
 			return -EOPNOTSUPP;
+		check_poll_queue_state(ctx, req);
 
 		kiocb->private = NULL;
 		kiocb->ki_flags |= IOCB_HIPRI;
-- 
2.40.1





[Index of Archives]     [Linux Samsung SoC]     [Linux Rockchip SoC]     [Linux Actions SoC]     [Linux for Synopsys ARC Processors]     [Linux NFS]     [Linux NILFS]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]


  Powered by Linux