io_uring: add support for IORING_OP_PIPE

Commit Message

Jens Axboe April 8, 2025, 4:45 p.m. UTC

This works just like pipe2(2), except it also supports fixed file
descriptors. Used in a similar fashion as for other fd instantiating
opcodes (like accept, socket, open, etc), where sqe->file_slot is set
appropriately if two direct descriptors are desired rather than a set
of normal file descriptors.

sqe->addr must be set to a pointer to an array of 2 integers, which
is where the fixed/normal file descriptors are copied to.

sqe->pipe_flags contains flags, same as what is allowed for pipe2(2).

Future expansion of per-op private flags can go in sqe->ioprio,
like we do for other opcodes that take both a "syscall" flag set and
an io_uring opcode specific flag set.

Signed-off-by: Jens Axboe <[e-mail]>

---

Patch

diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index ed2beb4def3f..431c156f9fe8 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -73,6 +73,7 @@  struct io_uring_sqe {
 		__u32		futex_flags;
 		__u32		install_fd_flags;
 		__u32		nop_flags;
+		__u32		pipe_flags;
 	};
 	__u64	user_data;	/* data to be passed back at completion time */
 	/* pack this to avoid bogus arm OABI complaints */
@@ -283,6 +284,7 @@  enum io_uring_op {
 	IORING_OP_EPOLL_WAIT,
 	IORING_OP_READV_FIXED,
 	IORING_OP_WRITEV_FIXED,
+	IORING_OP_PIPE,
 
 	/* this goes last, obviously */
 	IORING_OP_LAST,
diff --git a/io_uring/opdef.c b/io_uring/opdef.c
index 489384c0438b..db36433c2294 100644
--- a/io_uring/opdef.c
+++ b/io_uring/opdef.c
@@ -569,6 +569,10 @@  const struct io_issue_def io_issue_defs[] = {
 		.prep			= io_prep_writev_fixed,
 		.issue			= io_write,
 	},
+	[IORING_OP_PIPE] = {
+		.prep			= io_pipe_prep,
+		.issue			= io_pipe,
+	},
 };
 
 const struct io_cold_def io_cold_defs[] = {
@@ -815,6 +819,9 @@  const struct io_cold_def io_cold_defs[] = {
 		.cleanup		= io_readv_writev_cleanup,
 		.fail			= io_rw_fail,
 	},
+	[IORING_OP_PIPE] = {
+		.name			= "PIPE",
+	},
 };
 
 const char *io_uring_get_opcode(u8 opcode)
diff --git a/io_uring/openclose.c b/io_uring/openclose.c
index e3357dfa14ca..4dd461163457 100644
--- a/io_uring/openclose.c
+++ b/io_uring/openclose.c
@@ -6,6 +6,8 @@ 
 #include <linux/fdtable.h>
 #include <linux/fsnotify.h>
 #include <linux/namei.h>
+#include <linux/pipe_fs_i.h>
+#include <linux/watch_queue.h>
 #include <linux/io_uring.h>
 
 #include <uapi/linux/io_uring.h>
@@ -302,3 +304,134 @@  int io_install_fixed_fd(struct io_kiocb *req, unsigned int issue_flags)
 	io_req_set_res(req, ret, 0);
 	return IOU_OK;
 }
+
+struct io_pipe {
+	struct file *file;
+	int __user *fds;
+	int flags;
+	int file_slot;
+	unsigned long nofile;
+};
+
+int io_pipe_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+{
+	struct io_pipe *p = io_kiocb_to_cmd(req, struct io_pipe);
+
+	if (sqe->fd || sqe->off || sqe->addr3)
+		return -EINVAL;
+
+	p->fds = u64_to_user_ptr(READ_ONCE(sqe->addr));
+	p->flags = READ_ONCE(sqe->pipe_flags);
+	if (p->flags & ~(O_CLOEXEC | O_NONBLOCK | O_DIRECT | O_NOTIFICATION_PIPE))
+		return -EINVAL;
+
+	p->file_slot = READ_ONCE(sqe->file_index);
+	p->nofile = rlimit(RLIMIT_NOFILE);
+	return 0;
+}
+
+static int io_pipe_fixed(struct io_kiocb *req, struct file **files,
+			 unsigned int issue_flags)
+{
+	struct io_pipe *p = io_kiocb_to_cmd(req, struct io_pipe);
+	struct io_ring_ctx *ctx = req->ctx;
+	int ret, fds[2] = { -1, -1 };
+	int slot = p->file_slot;
+
+	if (p->flags & O_CLOEXEC)
+		return -EINVAL;
+
+	io_ring_submit_lock(ctx, issue_flags);
+
+	ret = __io_fixed_fd_install(ctx, files[0], slot);
+	if (ret < 0)
+		goto err;
+	fds[0] = ret;
+	files[0] = NULL;
+
+	/*
+	 * If a specific slot is given, next one will be used for
+	 * the write side.
+	 */
+	if (slot != IORING_FILE_INDEX_ALLOC)
+		slot++;
+
+	ret = __io_fixed_fd_install(ctx, files[1], slot);
+	if (ret < 0)
+		goto err;
+	fds[1] = ret;
+	files[1] = NULL;
+
+	io_ring_submit_unlock(ctx, issue_flags);
+
+	if (!copy_to_user(p->fds, fds, sizeof(fds)))
+		return 0;
+
+	ret = -EFAULT;
+	io_ring_submit_lock(ctx, issue_flags);
+err:
+	if (fds[0] != -1)
+		io_fixed_fd_remove(ctx, fds[0]);
+	if (fds[1] != -1)
+		io_fixed_fd_remove(ctx, fds[1]);
+	io_ring_submit_unlock(ctx, issue_flags);
+	return ret;
+}
+
+static int io_pipe_fd(struct io_kiocb *req, struct file **files)
+{
+	struct io_pipe *p = io_kiocb_to_cmd(req, struct io_pipe);
+	int ret, fds[2] = { -1, -1 };
+
+	ret = __get_unused_fd_flags(p->flags, p->nofile);
+	if (ret < 0)
+		goto err;
+	fds[0] = ret;
+
+	ret = __get_unused_fd_flags(p->flags, p->nofile);
+	if (ret < 0)
+		goto err;
+	fds[1] = ret;
+
+	if (!copy_to_user(p->fds, fds, sizeof(fds))) {
+		fd_install(fds[0], files[0]);
+		fd_install(fds[1], files[1]);
+		return 0;
+	}
+	ret = -EFAULT;
+err:
+	if (fds[0] != -1)
+		put_unused_fd(fds[0]);
+	if (fds[1] != -1)
+		put_unused_fd(fds[1]);
+	return ret;
+}
+
+int io_pipe(struct io_kiocb *req, unsigned int issue_flags)
+{
+	struct io_pipe *p = io_kiocb_to_cmd(req, struct io_pipe);
+	struct file *files[2];
+	int ret;
+
+	ret = create_pipe_files(files, p->flags);
+	if (ret)
+		return ret;
+	files[0]->f_mode |= FMODE_NOWAIT;
+	files[1]->f_mode |= FMODE_NOWAIT;
+
+	if (!!p->file_slot)
+		ret = io_pipe_fixed(req, files, issue_flags);
+	else
+		ret = io_pipe_fd(req, files);
+
+	io_req_set_res(req, ret, 0);
+	if (!ret)
+		return IOU_OK;
+
+	req_set_fail(req);
+	if (files[0])
+		fput(files[0]);
+	if (files[1])
+		fput(files[1]);
+	return ret;
+}
diff --git a/io_uring/openclose.h b/io_uring/openclose.h
index 8a93c98ad0ad..4ca2a9935abc 100644
--- a/io_uring/openclose.h
+++ b/io_uring/openclose.h
@@ -13,5 +13,8 @@  int io_openat2(struct io_kiocb *req, unsigned int issue_flags);
 int io_close_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_close(struct io_kiocb *req, unsigned int issue_flags);
 
+int io_pipe_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
+int io_pipe(struct io_kiocb *req, unsigned int issue_flags);
+
 int io_install_fixed_fd_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
 int io_install_fixed_fd(struct io_kiocb *req, unsigned int issue_flags);