[linux-2.6-block.git] / io_uring / msg_ring.c

// SPDX-License-Identifier: GPL-2.0
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/file.h>
#include <linux/slab.h>
#include <linux/nospec.h>
#include <linux/io_uring.h>

#include <uapi/linux/io_uring.h>

#include "io_uring.h"
#include "rsrc.h"
#include "filetable.h"
#include "msg_ring.h"

struct io_msg {
	struct file			*file;
	struct file			*src_file;
	u64 user_data;
	u32 len;
	u32 cmd;
	u32 src_fd;
	u32 dst_fd;
	u32 flags;
};

void io_msg_ring_cleanup(struct io_kiocb *req)
{
	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);

	if (WARN_ON_ONCE(!msg->src_file))
		return;

	fput(msg->src_file);
	msg->src_file = NULL;
}

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);

	if (msg->src_fd || msg->dst_fd || msg->flags)
		return -EINVAL;

	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
		return 0;

	return -EOVERFLOW;
}

static void io_double_unlock_ctx(struct io_ring_ctx *octx,
				 unsigned int issue_flags)
{
	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;
}

static struct file *io_msg_grab_file(struct io_kiocb *req, unsigned int issue_flags)
{
	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	struct io_ring_ctx *ctx = req->ctx;
	struct file *file = NULL;
	unsigned long file_ptr;
	int idx = msg->src_fd;

	io_ring_submit_lock(ctx, issue_flags);
	if (likely(idx < ctx->nr_user_files)) {
		idx = array_index_nospec(idx, ctx->nr_user_files);
		file_ptr = io_fixed_file_slot(&ctx->file_table, idx)->file_ptr;
		file = (struct file *) (file_ptr & FFS_MASK);
		if (file)
			get_file(file);
	}
	io_ring_submit_unlock(ctx, issue_flags);
	return file;
}

static int io_msg_send_fd(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);
	struct io_ring_ctx *ctx = req->ctx;
	struct file *src_file = msg->src_file;
	int ret;

	if (target_ctx == ctx)
		return -EINVAL;
	if (!src_file) {
		src_file = io_msg_grab_file(req, issue_flags);
		if (!src_file)
			return -EBADF;
		msg->src_file = src_file;
		req->flags |= REQ_F_NEED_CLEANUP;
	}

	if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
		return -EAGAIN;

	ret = __io_fixed_fd_install(target_ctx, src_file, msg->dst_fd);
	if (ret < 0)
		goto out_unlock;
	msg->src_file = NULL;
	req->flags &= ~REQ_F_NEED_CLEANUP;

	if (msg->flags & IORING_MSG_RING_CQE_SKIP)
		goto out_unlock;

	/*
	 * If this fails, the target still received the file descriptor but
	 * wasn't notified of the fact. This means that if this request
	 * completes with -EOVERFLOW, then the sender must ensure that a
	 * later IORING_OP_MSG_RING delivers the message.
	 */
	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);
	return ret;
}

int io_msg_ring_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
{
	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);

	if (unlikely(sqe->buf_index || sqe->personality))
		return -EINVAL;

	msg->src_file = NULL;
	msg->user_data = READ_ONCE(sqe->off);
	msg->len = READ_ONCE(sqe->len);
	msg->cmd = READ_ONCE(sqe->addr);
	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)
		return -EINVAL;

	return 0;
}

int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags)
{
	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	int ret;

	ret = -EBADFD;
	if (!io_is_uring_fops(req->file))
		goto done;

	switch (msg->cmd) {
	case IORING_MSG_DATA:
		ret = io_msg_ring_data(req);
		break;
	case IORING_MSG_SEND_FD:
		ret = io_msg_send_fd(req, issue_flags);
		break;
	default:
		ret = -EINVAL;
		break;
	}

done:
	if (ret == -EAGAIN)
		return -EAGAIN;
	if (ret < 0)
		req_set_fail(req);
	io_req_set_res(req, ret, 0);
	return IOU_OK;
}
Commit	Line	Data
36404b09 JA	1	// SPDX-License-Identifier: GPL-2.0
	2	#include <linux/kernel.h>
	3	#include <linux/errno.h>
	4	#include <linux/file.h>
	5	#include <linux/slab.h>
e6130eba	6	#include <linux/nospec.h>
36404b09 JA	7	#include <linux/io_uring.h>
	8
	9	#include <uapi/linux/io_uring.h>
	10
36404b09	11	#include "io_uring.h"
e6130eba JA	12	#include "rsrc.h"
e6130eba JA	13	#include "filetable.h"
36404b09 JA	14	#include "msg_ring.h"
	15
	16	struct io_msg {
	17	struct file *file;
11373026	18	struct file *src_file;
36404b09 JA	19	u64 user_data;
36404b09 JA	20	u32 len;
e6130eba JA	21	u32 cmd;
	22	u32 src_fd;
	23	u32 dst_fd;
	24	u32 flags;
36404b09 JA	25	};
36404b09 JA	26
11373026 PB	27	void io_msg_ring_cleanup(struct io_kiocb *req)
	28	{
	29	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	30
	31	if (WARN_ON_ONCE(!msg->src_file))
	32	return;
	33
	34	fput(msg->src_file);
	35	msg->src_file = NULL;
	36	}
	37
e6130eba JA	38	static int io_msg_ring_data(struct io_kiocb *req)
	39	{
	40	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	41	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
e6130eba JA	42
	43	if (msg->src_fd \|\| msg->dst_fd \|\| msg->flags)
	44	return -EINVAL;
	45
b529c96a	46	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
e6130eba JA	47	return 0;
	48
	49	return -EOVERFLOW;
	50	}
	51
11373026	52	static void io_double_unlock_ctx(struct io_ring_ctx *octx,
e6130eba JA	53	unsigned int issue_flags)
e6130eba JA	54	{
e6130eba JA	55	mutex_unlock(&octx->uring_lock);
	56	}
	57
11373026	58	static int io_double_lock_ctx(struct io_ring_ctx *octx,
e6130eba JA	59	unsigned int issue_flags)
	60	{
	61	/*
	62	* To ensure proper ordering between the two ctxs, we can only
	63	* attempt a trylock on the target. If that fails and we already have
	64	* the source ctx lock, punt to io-wq.
	65	*/
	66	if (!(issue_flags & IO_URING_F_UNLOCKED)) {
	67	if (!mutex_trylock(&octx->uring_lock))
	68	return -EAGAIN;
	69	return 0;
	70	}
11373026 PB	71	mutex_lock(&octx->uring_lock);
	72	return 0;
	73	}
e6130eba	74
11373026 PB	75	static struct file io_msg_grab_file(struct io_kiocb req, unsigned int issue_flags)
	76	{
	77	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	78	struct io_ring_ctx *ctx = req->ctx;
	79	struct file *file = NULL;
	80	unsigned long file_ptr;
	81	int idx = msg->src_fd;
	82
	83	io_ring_submit_lock(ctx, issue_flags);
	84	if (likely(idx < ctx->nr_user_files)) {
	85	idx = array_index_nospec(idx, ctx->nr_user_files);
	86	file_ptr = io_fixed_file_slot(&ctx->file_table, idx)->file_ptr;
	87	file = (struct file *) (file_ptr & FFS_MASK);
	88	if (file)
	89	get_file(file);
e6130eba	90	}
11373026 PB	91	io_ring_submit_unlock(ctx, issue_flags);
11373026 PB	92	return file;
e6130eba JA	93	}
	94
	95	static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags)
	96	{
	97	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	98	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
e6130eba	99	struct io_ring_ctx *ctx = req->ctx;
11373026	100	struct file *src_file = msg->src_file;
e6130eba JA	101	int ret;
	102
	103	if (target_ctx == ctx)
	104	return -EINVAL;
11373026 PB	105	if (!src_file) {
	106	src_file = io_msg_grab_file(req, issue_flags);
	107	if (!src_file)
	108	return -EBADF;
	109	msg->src_file = src_file;
	110	req->flags \|= REQ_F_NEED_CLEANUP;
	111	}
e6130eba	112
11373026 PB	113	if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
11373026 PB	114	return -EAGAIN;
e6130eba JA	115
e6130eba JA	116	ret = __io_fixed_fd_install(target_ctx, src_file, msg->dst_fd);
11373026	117	if (ret < 0)
e6130eba	118	goto out_unlock;
11373026 PB	119	msg->src_file = NULL;
11373026 PB	120	req->flags &= ~REQ_F_NEED_CLEANUP;
e6130eba JA	121
	122	if (msg->flags & IORING_MSG_RING_CQE_SKIP)
	123	goto out_unlock;
	124
	125	/*
	126	* If this fails, the target still received the file descriptor but
	127	* wasn't notified of the fact. This means that if this request
	128	* completes with -EOVERFLOW, then the sender must ensure that a
	129	* later IORING_OP_MSG_RING delivers the message.
	130	*/
b529c96a	131	if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
e6130eba JA	132	ret = -EOVERFLOW;
e6130eba JA	133	out_unlock:
11373026	134	io_double_unlock_ctx(target_ctx, issue_flags);
e6130eba JA	135	return ret;
	136	}
	137
36404b09 JA	138	int io_msg_ring_prep(struct io_kiocb req, const struct io_uring_sqe sqe)
36404b09 JA	139	{
f2ccb5ae	140	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09	141
e6130eba	142	if (unlikely(sqe->buf_index \|\| sqe->personality))
36404b09 JA	143	return -EINVAL;
36404b09 JA	144
11373026	145	msg->src_file = NULL;
36404b09 JA	146	msg->user_data = READ_ONCE(sqe->off);
36404b09 JA	147	msg->len = READ_ONCE(sqe->len);
e6130eba JA	148	msg->cmd = READ_ONCE(sqe->addr);
	149	msg->src_fd = READ_ONCE(sqe->addr3);
	150	msg->dst_fd = READ_ONCE(sqe->file_index);
	151	msg->flags = READ_ONCE(sqe->msg_ring_flags);
	152	if (msg->flags & ~IORING_MSG_RING_CQE_SKIP)
	153	return -EINVAL;
	154
36404b09 JA	155	return 0;
	156	}
	157
	158	int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags)
	159	{
f2ccb5ae	160	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09 JA	161	int ret;
	162
	163	ret = -EBADFD;
	164	if (!io_is_uring_fops(req->file))
	165	goto done;
	166
e6130eba JA	167	switch (msg->cmd) {
	168	case IORING_MSG_DATA:
	169	ret = io_msg_ring_data(req);
	170	break;
	171	case IORING_MSG_SEND_FD:
	172	ret = io_msg_send_fd(req, issue_flags);
	173	break;
	174	default:
	175	ret = -EINVAL;
	176	break;
	177	}
36404b09 JA	178
36404b09 JA	179	done:
4c979eae PB	180	if (ret == -EAGAIN)
4c979eae PB	181	return -EAGAIN;
36404b09 JA	182	if (ret < 0)
	183	req_set_fail(req);
	184	io_req_set_res(req, ret, 0);
36404b09 JA	185	return IOU_OK;
36404b09 JA	186	}