1 files changed, 103 insertions, 50 deletions

diff --git a/io_uring/msg_ring.c b/io_uring/msg_ring.c
index 2d3cd945a531..8803c0979e2a 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;
@@ -21,10 +26,35 @@ struct io_msg {
 	u32 len;
 	u32 cmd;
 	u32 src_fd;
-	u32 dst_fd;
+	union {
+		u32 dst_fd;
+		u32 cqe_flags;
+	};
 	u32 flags;
 };
 
+static void io_double_unlock_ctx(struct io_ring_ctx *octx)
+{
+	mutex_unlock(&octx->uring_lock);
+}
+
+static int io_double_lock_ctx(struct io_ring_ctx *octx,
+			      unsigned int issue_flags)
+{
+	/*
+	 * To ensure proper ordering between the two ctxs, we can only
+	 * attempt a trylock on the target. If that fails and we already have
+	 * the source ctx lock, punt to io-wq.
+	 */
+	if (!(issue_flags & IO_URING_F_UNLOCKED)) {
+		if (!mutex_trylock(&octx->uring_lock))
+			return -EAGAIN;
+		return 0;
+	}
+	mutex_lock(&octx->uring_lock);
+	return 0;
+}
+
 void io_msg_ring_cleanup(struct io_kiocb *req)
 {
 	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
@@ -36,6 +66,29 @@ void io_msg_ring_cleanup(struct io_kiocb *req)
 	msg->src_file = NULL;
 }
 
+static inline bool io_msg_need_remote(struct io_ring_ctx *target_ctx)
+{
+	if (!target_ctx->task_complete)
+		return false;
+	return current != target_ctx->submitter_task;
+}
+
+static int io_msg_exec_remote(struct io_kiocb *req, task_work_func_t func)
+{
+	struct io_ring_ctx *ctx = req->file->private_data;
+	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
+	struct task_struct *task = READ_ONCE(ctx->submitter_task);
+
+	if (unlikely(!task))
+		return -EOWNERDEAD;
+
+	init_task_work(&msg->tw, func);
+	if (task_work_add(ctx->submitter_task, &msg->tw, TWA_SIGNAL))
+		return -EOWNERDEAD;
+
+	return IOU_ISSUE_SKIP_COMPLETE;
+}
+
 static void io_msg_tw_complete(struct callback_head *head)
 {
 	struct io_msg *msg = container_of(head, struct io_msg, tw);
@@ -43,61 +96,65 @@ static void io_msg_tw_complete(struct callback_head *head)
 	struct io_ring_ctx *target_ctx = req->file->private_data;
 	int ret = 0;
 
-	if (current->flags & PF_EXITING)
+	if (current->flags & PF_EXITING) {
 		ret = -EOWNERDEAD;
-	else if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
-		ret = -EOVERFLOW;
+	} else {
+		u32 flags = 0;
+
+		if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
+			flags = msg->cqe_flags;
+
+		/*
+		 * If the target ring is using IOPOLL mode, then we need to be
+		 * holding the uring_lock for posting completions. Other ring
+		 * types rely on the regular completion locking, which is
+		 * handled while posting.
+		 */
+		if (target_ctx->flags & IORING_SETUP_IOPOLL)
+			mutex_lock(&target_ctx->uring_lock);
+		if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
+			ret = -EOVERFLOW;
+		if (target_ctx->flags & IORING_SETUP_IOPOLL)
+			mutex_unlock(&target_ctx->uring_lock);
+	}
 
 	if (ret < 0)
 		req_set_fail(req);
 	io_req_queue_tw_complete(req, ret);
 }
 
-static int io_msg_ring_data(struct io_kiocb *req)
+static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags)
 {
 	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;
+	int ret;
 
-	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 (target_ctx->flags & IORING_SETUP_R_DISABLED)
+		return -EBADFD;
 
-	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,
-				  TWA_SIGNAL_NO_IPI))
-			return -EOWNERDEAD;
-
-		atomic_or(IORING_SQ_TASKRUN, &target_ctx->rings->sq_flags);
-		return IOU_ISSUE_SKIP_COMPLETE;
-	}
-
-	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
-		return 0;
-
-	return -EOVERFLOW;
-}
+	if (io_msg_need_remote(target_ctx))
+		return io_msg_exec_remote(req, io_msg_tw_complete);
 
-static void io_double_unlock_ctx(struct io_ring_ctx *octx,
-				 unsigned int issue_flags)
-{
-	mutex_unlock(&octx->uring_lock);
-}
+	if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
+		flags = msg->cqe_flags;
 
-static int io_double_lock_ctx(struct io_ring_ctx *octx,
-			      unsigned int issue_flags)
-{
-	/*
-	 * To ensure proper ordering between the two ctxs, we can only
-	 * attempt a trylock on the target. If that fails and we already have
-	 * the source ctx lock, punt to io-wq.
-	 */
-	if (!(issue_flags & IO_URING_F_UNLOCKED)) {
-		if (!mutex_trylock(&octx->uring_lock))
+	ret = -EOVERFLOW;
+	if (target_ctx->flags & IORING_SETUP_IOPOLL) {
+		if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
 			return -EAGAIN;
-		return 0;
+		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
+			ret = 0;
+		io_double_unlock_ctx(target_ctx);
+	} else {
+		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
+			ret = 0;
 	}
-	mutex_lock(&octx->uring_lock);
-	return 0;
+	return ret;
 }
 
 static struct file *io_msg_grab_file(struct io_kiocb *req, unsigned int issue_flags)
@@ -148,7 +205,7 @@ static int io_msg_install_complete(struct io_kiocb *req, unsigned int issue_flag
 	if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
 		ret = -EOVERFLOW;
 out_unlock:
-	io_double_unlock_ctx(target_ctx, issue_flags);
+	io_double_unlock_ctx(target_ctx);
 	return ret;
 }
 
@@ -174,6 +231,8 @@ static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags)
 
 	if (target_ctx == ctx)
 		return -EINVAL;
+	if (target_ctx->flags & IORING_SETUP_R_DISABLED)
+		return -EBADFD;
 	if (!src_file) {
 		src_file = io_msg_grab_file(req, issue_flags);
 		if (!src_file)
@@ -182,14 +241,8 @@ static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags)
 		req->flags |= REQ_F_NEED_CLEANUP;
 	}
 
-	if (target_ctx->task_complete && current != target_ctx->submitter_task) {
-		init_task_work(&msg->tw, io_msg_tw_fd_complete);
-		if (task_work_add(target_ctx->submitter_task, &msg->tw,
-				  TWA_SIGNAL))
-			return -EOWNERDEAD;
-
-		return IOU_ISSUE_SKIP_COMPLETE;
-	}
+	if (io_msg_need_remote(target_ctx))
+		return io_msg_exec_remote(req, io_msg_tw_fd_complete);
 	return io_msg_install_complete(req, issue_flags);
 }
 
@@ -207,7 +260,7 @@ 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;
@@ -224,7 +277,7 @@ int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags)
 
 	switch (msg->cmd) {
 	case IORING_MSG_DATA:
-		ret = io_msg_ring_data(req);
+		ret = io_msg_ring_data(req, issue_flags);
 		break;
 	case IORING_MSG_SEND_FD:
 		ret = io_msg_send_fd(req, issue_flags);