Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
From:   leitao@debian.org
To:     dylany@meta.com, axboe@kernel.dk, asml.silence@gmail.com,
        io-uring@vger.kernel.org
Cc:     leit@meta.com, linux-kernel@vger.kernel.org,
        Breno Leitao <leitao@debian.org>
Subject: [PATCH] io_uring/msg_ring: Pass custom flags to the cqe
Date:   Tue,  3 Jan 2023 08:05:07 -0800
Message-Id: <20230103160507.617416-1-leitao@debian.org>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

From: Breno Leitao <leitao@debian.org>

This patch adds a new flag (IORING_MSG_RING_FLAGS_PASS) in the message
ring operations (IORING_OP_MSG_RING). This new flag enables the sender
to specify custom flags, which will be copied over to cqe->flags in the
receiving ring.  These custom flags should be specified using the
sqe->file_index field.

This mechanism provides additional flexibility when sending messages
between rings.

The changes on liburing side could be seen at
https://github.com/axboe/liburing/pull/765

Signed-off-by: Breno Leitao <leitao@debian.org>
---
 include/uapi/linux/io_uring.h |  2 ++
 io_uring/msg_ring.c           | 23 +++++++++++++++++++----
 2 files changed, 21 insertions(+), 4 deletions(-)

diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index 2780bce62faf..636a4c2c1294 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -347,6 +347,8 @@ enum {
  *				applicable for IORING_MSG_DATA, obviously.
  */
 #define IORING_MSG_RING_CQE_SKIP	(1U << 0)
+/* Pass through the flags from sqe->file_index to cqe->flags */
+#define IORING_MSG_RING_FLAGS_PASS	(1U << 1)
 
 /*
  * IO completion data structure (Completion Queue Entry)
diff --git a/io_uring/msg_ring.c b/io_uring/msg_ring.c
index 2d3cd945a531..dc233f72a541 100644
--- a/io_uring/msg_ring.c
+++ b/io_uring/msg_ring.c
@@ -13,6 +13,11 @@
 #include "filetable.h"
 #include "msg_ring.h"
 
+
+/* All valid masks for MSG_RING */
+#define IORING_MSG_RING_MASK		(IORING_MSG_RING_CQE_SKIP | \
+					IORING_MSG_RING_FLAGS_PASS)
+
 struct io_msg {
 	struct file			*file;
 	struct file			*src_file;
@@ -21,7 +26,10 @@ struct io_msg {
 	u32 len;
 	u32 cmd;
 	u32 src_fd;
-	u32 dst_fd;
+	union {
+		u32 dst_fd;
+		u32 cqe_flags;
+	};
 	u32 flags;
 };
 
@@ -57,10 +65,17 @@ static int io_msg_ring_data(struct io_kiocb *req)
 {
 	struct io_ring_ctx *target_ctx = req->file->private_data;
 	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
+	u32 flags = 0;
 
-	if (msg->src_fd || msg->dst_fd || msg->flags)
+	if (msg->src_fd || msg->flags & ~IORING_MSG_RING_FLAGS_PASS)
 		return -EINVAL;
 
+	if (!(msg->flags & IORING_MSG_RING_FLAGS_PASS) && msg->dst_fd)
+		return -EINVAL;
+
+	if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
+		flags = msg->cqe_flags;
+
 	if (target_ctx->task_complete && current != target_ctx->submitter_task) {
 		init_task_work(&msg->tw, io_msg_tw_complete);
 		if (task_work_add(target_ctx->submitter_task, &msg->tw,
@@ -71,7 +86,7 @@ static int io_msg_ring_data(struct io_kiocb *req)
 		return IOU_ISSUE_SKIP_COMPLETE;
 	}
 
-	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
+	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
 		return 0;
 
 	return -EOVERFLOW;
@@ -207,7 +222,7 @@ int io_msg_ring_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	msg->src_fd = READ_ONCE(sqe->addr3);
 	msg->dst_fd = READ_ONCE(sqe->file_index);
 	msg->flags = READ_ONCE(sqe->msg_ring_flags);
-	if (msg->flags & ~IORING_MSG_RING_CQE_SKIP)
+	if (msg->flags & ~IORING_MSG_RING_MASK)
 		return -EINVAL;
 
 	return 0;
-- 
2.30.2