Re: [PATCH v6 2/2] io_uring: avoid unnecessary io_wq_work copy for fast poll feature

[Xiaoguang Wang <xiaoguang.wang@xxxxxxxxxxxxxxxxx>]

hi,

I also use below debug patch to run test cases in liburing:
diff --git a/fs/io_uring.c b/fs/io_uring.c
index 3bec6057c189..119764d18a61 100644
--- a/fs/io_uring.c
+++ b/fs/io_uring.c
@@ -5819,6 +5819,13 @@ static int io_init_req(struct io_ring_ctx *ctx, struct io_kiocb *req,
        refcount_set(&req->refs, 2);
        req->task = NULL;
        req->result = 0;
+       req->work.list.next = 0x1;
+       req->work.files = 0x2;
+       req->work.mm = 0x3;
+       req->work.creds = 0x4;
+       req->work.fs = 0x5;
+       req->work.flags = 0x6;
+       req->work.task_pid = 0x7;

All test cases pass.

Regards,
Xiaoguang Wang

> Basically IORING_OP_POLL_ADD command and async armed poll handlers
> for regular commands don't touch io_wq_work, so only REQ_F_WORK_INITIALIZED
> is set, can we do io_wq_work copy and restore.
>
> Signed-off-by: Xiaoguang Wang <xiaoguang.wang@xxxxxxxxxxxxxxxxx>
>
> ---
> V3:
>    drop the REQ_F_WORK_NEED_RESTORE flag introduced in V2 patch, just
>    use REQ_F_WORK_INITIALIZED to control whether to do io_wq_work copy
>    and restore.
>
> V6:
>    rebase to io_uring-5.8.
> ---
>   fs/io_uring.c | 13 +++++++++----
>   1 file changed, 9 insertions(+), 4 deletions(-)
>
> diff --git a/fs/io_uring.c b/fs/io_uring.c
> index bde8b17a7275..3bec6057c189 100644
> --- a/fs/io_uring.c
> +++ b/fs/io_uring.c
> @@ -4261,7 +4261,8 @@ static void io_async_task_func(struct callback_head *cb)
>   	spin_unlock_irq(&ctx->completion_lock);
>   
>   	/* restore ->work in case we need to retry again */
> -	memcpy(&req->work, &apoll->work, sizeof(req->work));
> +	if (req->flags & REQ_F_WORK_INITIALIZED)
> +		memcpy(&req->work, &apoll->work, sizeof(req->work));
>   	kfree(apoll);
>   
>   	if (!canceled) {
> @@ -4358,7 +4359,8 @@ static bool io_arm_poll_handler(struct io_kiocb *req)
>   		return false;
>   
>   	req->flags |= REQ_F_POLLED;
> -	memcpy(&apoll->work, &req->work, sizeof(req->work));
> +	if (req->flags & REQ_F_WORK_INITIALIZED)
> +		memcpy(&apoll->work, &req->work, sizeof(req->work));
>   	had_io = req->io != NULL;
>   
>   	get_task_struct(current);
> @@ -4383,7 +4385,8 @@ static bool io_arm_poll_handler(struct io_kiocb *req)
>   		if (!had_io)
>   			io_poll_remove_double(req);
>   		spin_unlock_irq(&ctx->completion_lock);
> -		memcpy(&req->work, &apoll->work, sizeof(req->work));
> +		if (req->flags & REQ_F_WORK_INITIALIZED)
> +			memcpy(&req->work, &apoll->work, sizeof(req->work));
>   		kfree(apoll);
>   		return false;
>   	}
> @@ -4428,7 +4431,9 @@ static bool io_poll_remove_one(struct io_kiocb *req)
>   			 * io_req_work_drop_env below when dropping the
>   			 * final reference.
>   			 */
> -			memcpy(&req->work, &apoll->work, sizeof(req->work));
> +			if (req->flags & REQ_F_WORK_INITIALIZED)
> +				memcpy(&req->work, &apoll->work,
> +				       sizeof(req->work));
>   			kfree(apoll);
>   		}
>   	}