[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;
	struct callback_head		tw;
	u64 user_data;
	u32 len;
	u32 cmd;
	u32 src_fd;
	u32 dst_fd;
	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);

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

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

static void io_msg_tw_complete(struct callback_head *head)
{
	struct io_msg *msg = container_of(head, struct io_msg, tw);
	struct io_kiocb *req = cmd_to_io_kiocb(msg);
	struct io_ring_ctx *target_ctx = req->file->private_data;
	int ret = 0;

	if (current->flags & PF_EXITING) {
		ret = -EOWNERDEAD;
	} else {
		/*
		 * 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, 0))
			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, 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);
	int ret;

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

	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;
	}

	ret = -EOVERFLOW;
	if (target_ctx->flags & IORING_SETUP_IOPOLL) {
		if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
			return -EAGAIN;
		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
			ret = 0;
		io_double_unlock_ctx(target_ctx);
	} else {
		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
			ret = 0;
	}
	return ret;
}

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_install_complete(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 file *src_file = msg->src_file;
	int ret;

	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);
	return ret;
}

static void io_msg_tw_fd_complete(struct callback_head *head)
{
	struct io_msg *msg = container_of(head, struct io_msg, tw);
	struct io_kiocb *req = cmd_to_io_kiocb(msg);
	int ret = -EOWNERDEAD;

	if (!(current->flags & PF_EXITING))
		ret = io_msg_install_complete(req, IO_URING_F_UNLOCKED);
	if (ret < 0)
		req_set_fail(req);
	io_req_queue_tw_complete(req, ret);
}

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;

	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 (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;
	}
	return io_msg_install_complete(req, issue_flags);
}

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, issue_flags);
		break;
	case IORING_MSG_SEND_FD:
		ret = io_msg_send_fd(req, issue_flags);
		break;
	default:
		ret = -EINVAL;
		break;
	}

done:
	if (ret < 0) {
		if (ret == -EAGAIN || ret == IOU_ISSUE_SKIP_COMPLETE)
			return ret;
		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;
6d043ee1	19	struct callback_head tw;
36404b09 JA	20	u64 user_data;
36404b09 JA	21	u32 len;
e6130eba JA	22	u32 cmd;
	23	u32 src_fd;
	24	u32 dst_fd;
	25	u32 flags;
36404b09 JA	26	};
36404b09 JA	27
423d5081 JA	28	static void io_double_unlock_ctx(struct io_ring_ctx *octx)
	29	{
	30	mutex_unlock(&octx->uring_lock);
	31	}
	32
	33	static int io_double_lock_ctx(struct io_ring_ctx *octx,
	34	unsigned int issue_flags)
	35	{
	36	/*
	37	* To ensure proper ordering between the two ctxs, we can only
	38	* attempt a trylock on the target. If that fails and we already have
	39	* the source ctx lock, punt to io-wq.
	40	*/
	41	if (!(issue_flags & IO_URING_F_UNLOCKED)) {
	42	if (!mutex_trylock(&octx->uring_lock))
	43	return -EAGAIN;
	44	return 0;
	45	}
	46	mutex_lock(&octx->uring_lock);
	47	return 0;
	48	}
	49
11373026 PB	50	void io_msg_ring_cleanup(struct io_kiocb *req)
	51	{
	52	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	53
	54	if (WARN_ON_ONCE(!msg->src_file))
	55	return;
	56
	57	fput(msg->src_file);
	58	msg->src_file = NULL;
	59	}
	60
6d043ee1 PB	61	static void io_msg_tw_complete(struct callback_head *head)
	62	{
	63	struct io_msg *msg = container_of(head, struct io_msg, tw);
	64	struct io_kiocb *req = cmd_to_io_kiocb(msg);
	65	struct io_ring_ctx *target_ctx = req->file->private_data;
	66	int ret = 0;
	67
e12d7a46	68	if (current->flags & PF_EXITING) {
6d043ee1	69	ret = -EOWNERDEAD;
e12d7a46 JA	70	} else {
	71	/*
	72	* If the target ring is using IOPOLL mode, then we need to be
	73	* holding the uring_lock for posting completions. Other ring
	74	* types rely on the regular completion locking, which is
	75	* handled while posting.
	76	*/
	77	if (target_ctx->flags & IORING_SETUP_IOPOLL)
	78	mutex_lock(&target_ctx->uring_lock);
	79	if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
	80	ret = -EOVERFLOW;
	81	if (target_ctx->flags & IORING_SETUP_IOPOLL)
	82	mutex_unlock(&target_ctx->uring_lock);
	83	}
6d043ee1 PB	84
	85	if (ret < 0)
	86	req_set_fail(req);
	87	io_req_queue_tw_complete(req, ret);
	88	}
	89
e12d7a46	90	static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags)
e6130eba JA	91	{
e6130eba JA	92	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	93	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
e12d7a46	94	int ret;
e6130eba JA	95
	96	if (msg->src_fd \|\| msg->dst_fd \|\| msg->flags)
	97	return -EINVAL;
	98
6d043ee1 PB	99	if (target_ctx->task_complete && current != target_ctx->submitter_task) {
	100	init_task_work(&msg->tw, io_msg_tw_complete);
	101	if (task_work_add(target_ctx->submitter_task, &msg->tw,
	102	TWA_SIGNAL_NO_IPI))
	103	return -EOWNERDEAD;
	104
761c61c1	105	atomic_or(IORING_SQ_TASKRUN, &target_ctx->rings->sq_flags);
6d043ee1 PB	106	return IOU_ISSUE_SKIP_COMPLETE;
	107	}
	108
e12d7a46 JA	109	ret = -EOVERFLOW;
	110	if (target_ctx->flags & IORING_SETUP_IOPOLL) {
	111	if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
	112	return -EAGAIN;
	113	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
	114	ret = 0;
	115	io_double_unlock_ctx(target_ctx);
	116	} else {
	117	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
	118	ret = 0;
	119	}
	120	return ret;
e6130eba JA	121	}
e6130eba JA	122
11373026 PB	123	static struct file io_msg_grab_file(struct io_kiocb req, unsigned int issue_flags)
	124	{
	125	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	126	struct io_ring_ctx *ctx = req->ctx;
	127	struct file *file = NULL;
	128	unsigned long file_ptr;
	129	int idx = msg->src_fd;
	130
	131	io_ring_submit_lock(ctx, issue_flags);
	132	if (likely(idx < ctx->nr_user_files)) {
	133	idx = array_index_nospec(idx, ctx->nr_user_files);
	134	file_ptr = io_fixed_file_slot(&ctx->file_table, idx)->file_ptr;
	135	file = (struct file *) (file_ptr & FFS_MASK);
	136	if (file)
	137	get_file(file);
e6130eba	138	}
11373026 PB	139	io_ring_submit_unlock(ctx, issue_flags);
11373026 PB	140	return file;
e6130eba JA	141	}
e6130eba JA	142
17211310	143	static int io_msg_install_complete(struct io_kiocb *req, unsigned int issue_flags)
e6130eba JA	144	{
e6130eba JA	145	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	146	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
11373026	147	struct file *src_file = msg->src_file;
e6130eba JA	148	int ret;
e6130eba JA	149
11373026 PB	150	if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
11373026 PB	151	return -EAGAIN;
e6130eba JA	152
e6130eba JA	153	ret = __io_fixed_fd_install(target_ctx, src_file, msg->dst_fd);
11373026	154	if (ret < 0)
e6130eba	155	goto out_unlock;
17211310	156
11373026 PB	157	msg->src_file = NULL;
11373026 PB	158	req->flags &= ~REQ_F_NEED_CLEANUP;
e6130eba JA	159
	160	if (msg->flags & IORING_MSG_RING_CQE_SKIP)
	161	goto out_unlock;
e6130eba JA	162	/*
	163	* If this fails, the target still received the file descriptor but
	164	* wasn't notified of the fact. This means that if this request
	165	* completes with -EOVERFLOW, then the sender must ensure that a
	166	* later IORING_OP_MSG_RING delivers the message.
	167	*/
b529c96a	168	if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
e6130eba JA	169	ret = -EOVERFLOW;
e6130eba JA	170	out_unlock:
423d5081	171	io_double_unlock_ctx(target_ctx);
e6130eba JA	172	return ret;
	173	}
	174
6d043ee1 PB	175	static void io_msg_tw_fd_complete(struct callback_head *head)
	176	{
	177	struct io_msg *msg = container_of(head, struct io_msg, tw);
	178	struct io_kiocb *req = cmd_to_io_kiocb(msg);
	179	int ret = -EOWNERDEAD;
	180
	181	if (!(current->flags & PF_EXITING))
	182	ret = io_msg_install_complete(req, IO_URING_F_UNLOCKED);
	183	if (ret < 0)
	184	req_set_fail(req);
	185	io_req_queue_tw_complete(req, ret);
	186	}
	187
17211310 PB	188	static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags)
	189	{
	190	struct io_ring_ctx *target_ctx = req->file->private_data;
	191	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	192	struct io_ring_ctx *ctx = req->ctx;
	193	struct file *src_file = msg->src_file;
	194
	195	if (target_ctx == ctx)
	196	return -EINVAL;
	197	if (!src_file) {
	198	src_file = io_msg_grab_file(req, issue_flags);
	199	if (!src_file)
	200	return -EBADF;
	201	msg->src_file = src_file;
	202	req->flags \|= REQ_F_NEED_CLEANUP;
	203	}
6d043ee1 PB	204
	205	if (target_ctx->task_complete && current != target_ctx->submitter_task) {
	206	init_task_work(&msg->tw, io_msg_tw_fd_complete);
	207	if (task_work_add(target_ctx->submitter_task, &msg->tw,
	208	TWA_SIGNAL))
	209	return -EOWNERDEAD;
	210
	211	return IOU_ISSUE_SKIP_COMPLETE;
	212	}
17211310 PB	213	return io_msg_install_complete(req, issue_flags);
	214	}
	215
36404b09 JA	216	int io_msg_ring_prep(struct io_kiocb req, const struct io_uring_sqe sqe)
36404b09 JA	217	{
f2ccb5ae	218	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09	219
e6130eba	220	if (unlikely(sqe->buf_index \|\| sqe->personality))
36404b09 JA	221	return -EINVAL;
36404b09 JA	222
11373026	223	msg->src_file = NULL;
36404b09 JA	224	msg->user_data = READ_ONCE(sqe->off);
36404b09 JA	225	msg->len = READ_ONCE(sqe->len);
e6130eba JA	226	msg->cmd = READ_ONCE(sqe->addr);
	227	msg->src_fd = READ_ONCE(sqe->addr3);
	228	msg->dst_fd = READ_ONCE(sqe->file_index);
	229	msg->flags = READ_ONCE(sqe->msg_ring_flags);
	230	if (msg->flags & ~IORING_MSG_RING_CQE_SKIP)
	231	return -EINVAL;
	232
36404b09 JA	233	return 0;
	234	}
	235
	236	int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags)
	237	{
f2ccb5ae	238	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09 JA	239	int ret;
	240
	241	ret = -EBADFD;
	242	if (!io_is_uring_fops(req->file))
	243	goto done;
	244
e6130eba JA	245	switch (msg->cmd) {
e6130eba JA	246	case IORING_MSG_DATA:
e12d7a46	247	ret = io_msg_ring_data(req, issue_flags);
e6130eba JA	248	break;
	249	case IORING_MSG_SEND_FD:
	250	ret = io_msg_send_fd(req, issue_flags);
	251	break;
	252	default:
	253	ret = -EINVAL;
	254	break;
	255	}
36404b09 JA	256
36404b09 JA	257	done:
6d043ee1 PB	258	if (ret < 0) {
	259	if (ret == -EAGAIN \|\| ret == IOU_ISSUE_SKIP_COMPLETE)
	260	return ret;
36404b09	261	req_set_fail(req);
6d043ee1	262	}
36404b09	263	io_req_set_res(req, ret, 0);
36404b09 JA	264	return IOU_OK;
36404b09 JA	265	}