[linux-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;
	u64 user_data;
	u32 len;
	u32 cmd;
	u32 src_fd;
	u32 dst_fd;
	u32 flags;
};

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 *ctx,
				 struct io_ring_ctx *octx,
				 unsigned int issue_flags)
{
	if (issue_flags & IO_URING_F_UNLOCKED)
		mutex_unlock(&ctx->uring_lock);
	mutex_unlock(&octx->uring_lock);
}

static int io_double_lock_ctx(struct io_ring_ctx *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;
	}

	/* Always grab smallest value ctx first. We know ctx != octx. */
	if (ctx < octx) {
		mutex_lock(&ctx->uring_lock);
		mutex_lock(&octx->uring_lock);
	} else {
		mutex_lock(&octx->uring_lock);
		mutex_lock(&ctx->uring_lock);
	}

	return 0;
}

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;
	unsigned long file_ptr;
	struct file *src_file;
	int ret;

	if (target_ctx == ctx)
		return -EINVAL;

	ret = io_double_lock_ctx(ctx, target_ctx, issue_flags);
	if (unlikely(ret))
		return ret;

	ret = -EBADF;
	if (unlikely(msg->src_fd >= ctx->nr_user_files))
		goto out_unlock;

	msg->src_fd = array_index_nospec(msg->src_fd, ctx->nr_user_files);
	file_ptr = io_fixed_file_slot(&ctx->file_table, msg->src_fd)->file_ptr;
	if (!file_ptr)
		goto out_unlock;

	src_file = (struct file *) (file_ptr & FFS_MASK);
	get_file(src_file);

	ret = __io_fixed_fd_install(target_ctx, src_file, msg->dst_fd);
	if (ret < 0) {
		fput(src_file);
		goto out_unlock;
	}

	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(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->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;
	18	u64 user_data;
	19	u32 len;
e6130eba JA	20	u32 cmd;
	21	u32 src_fd;
	22	u32 dst_fd;
	23	u32 flags;
36404b09 JA	24	};
36404b09 JA	25
e6130eba JA	26	static int io_msg_ring_data(struct io_kiocb *req)
	27	{
	28	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	29	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
e6130eba JA	30
	31	if (msg->src_fd \|\| msg->dst_fd \|\| msg->flags)
	32	return -EINVAL;
	33
b529c96a	34	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
e6130eba JA	35	return 0;
	36
	37	return -EOVERFLOW;
	38	}
	39
	40	static void io_double_unlock_ctx(struct io_ring_ctx *ctx,
	41	struct io_ring_ctx *octx,
	42	unsigned int issue_flags)
	43	{
	44	if (issue_flags & IO_URING_F_UNLOCKED)
	45	mutex_unlock(&ctx->uring_lock);
	46	mutex_unlock(&octx->uring_lock);
	47	}
	48
	49	static int io_double_lock_ctx(struct io_ring_ctx *ctx,
	50	struct io_ring_ctx *octx,
	51	unsigned int issue_flags)
	52	{
	53	/*
	54	* To ensure proper ordering between the two ctxs, we can only
	55	* attempt a trylock on the target. If that fails and we already have
	56	* the source ctx lock, punt to io-wq.
	57	*/
	58	if (!(issue_flags & IO_URING_F_UNLOCKED)) {
	59	if (!mutex_trylock(&octx->uring_lock))
	60	return -EAGAIN;
	61	return 0;
	62	}
	63
	64	/* Always grab smallest value ctx first. We know ctx != octx. */
	65	if (ctx < octx) {
	66	mutex_lock(&ctx->uring_lock);
	67	mutex_lock(&octx->uring_lock);
	68	} else {
	69	mutex_lock(&octx->uring_lock);
	70	mutex_lock(&ctx->uring_lock);
	71	}
	72
	73	return 0;
	74	}
	75
	76	static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags)
	77	{
	78	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	79	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
e6130eba JA	80	struct io_ring_ctx *ctx = req->ctx;
	81	unsigned long file_ptr;
	82	struct file *src_file;
	83	int ret;
	84
	85	if (target_ctx == ctx)
	86	return -EINVAL;
	87
	88	ret = io_double_lock_ctx(ctx, target_ctx, issue_flags);
	89	if (unlikely(ret))
	90	return ret;
	91
	92	ret = -EBADF;
	93	if (unlikely(msg->src_fd >= ctx->nr_user_files))
	94	goto out_unlock;
	95
	96	msg->src_fd = array_index_nospec(msg->src_fd, ctx->nr_user_files);
	97	file_ptr = io_fixed_file_slot(&ctx->file_table, msg->src_fd)->file_ptr;
16bbdfe5 HM	98	if (!file_ptr)
	99	goto out_unlock;
	100
e6130eba JA	101	src_file = (struct file *) (file_ptr & FFS_MASK);
	102	get_file(src_file);
	103
	104	ret = __io_fixed_fd_install(target_ctx, src_file, msg->dst_fd);
	105	if (ret < 0) {
	106	fput(src_file);
	107	goto out_unlock;
	108	}
	109
	110	if (msg->flags & IORING_MSG_RING_CQE_SKIP)
	111	goto out_unlock;
	112
	113	/*
	114	* If this fails, the target still received the file descriptor but
	115	* wasn't notified of the fact. This means that if this request
	116	* completes with -EOVERFLOW, then the sender must ensure that a
	117	* later IORING_OP_MSG_RING delivers the message.
	118	*/
b529c96a	119	if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
e6130eba JA	120	ret = -EOVERFLOW;
	121	out_unlock:
	122	io_double_unlock_ctx(ctx, target_ctx, issue_flags);
	123	return ret;
	124	}
	125
36404b09 JA	126	int io_msg_ring_prep(struct io_kiocb req, const struct io_uring_sqe sqe)
36404b09 JA	127	{
f2ccb5ae	128	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09	129
e6130eba	130	if (unlikely(sqe->buf_index \|\| sqe->personality))
36404b09 JA	131	return -EINVAL;
	132
	133	msg->user_data = READ_ONCE(sqe->off);
	134	msg->len = READ_ONCE(sqe->len);
e6130eba JA	135	msg->cmd = READ_ONCE(sqe->addr);
	136	msg->src_fd = READ_ONCE(sqe->addr3);
	137	msg->dst_fd = READ_ONCE(sqe->file_index);
	138	msg->flags = READ_ONCE(sqe->msg_ring_flags);
	139	if (msg->flags & ~IORING_MSG_RING_CQE_SKIP)
	140	return -EINVAL;
	141
36404b09 JA	142	return 0;
	143	}
	144
	145	int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags)
	146	{
f2ccb5ae	147	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09 JA	148	int ret;
	149
	150	ret = -EBADFD;
	151	if (!io_is_uring_fops(req->file))
	152	goto done;
	153
e6130eba JA	154	switch (msg->cmd) {
	155	case IORING_MSG_DATA:
	156	ret = io_msg_ring_data(req);
	157	break;
	158	case IORING_MSG_SEND_FD:
	159	ret = io_msg_send_fd(req, issue_flags);
	160	break;
	161	default:
	162	ret = -EINVAL;
	163	break;
	164	}
36404b09 JA	165
36404b09 JA	166	done:
4c979eae PB	167	if (ret == -EAGAIN)
4c979eae PB	168	return -EAGAIN;
36404b09 JA	169	if (ret < 0)
	170	req_set_fail(req);
	171	io_req_set_res(req, ret, 0);
36404b09 JA	172	return IOU_OK;
36404b09 JA	173	}