[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;
	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 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);
	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->flags & IORING_SETUP_R_DISABLED)
		return -EBADFD;

	if (io_msg_need_remote(target_ctx))
		return io_msg_exec_remote(req, io_msg_tw_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 (target_ctx->flags & IORING_SETUP_R_DISABLED)
		return -EBADFD;
	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 (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);
}

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
56d8e318 PB	61	static inline bool io_msg_need_remote(struct io_ring_ctx *target_ctx)
	62	{
	63	if (!target_ctx->task_complete)
	64	return false;
	65	return current != target_ctx->submitter_task;
	66	}
	67
	68	static int io_msg_exec_remote(struct io_kiocb *req, task_work_func_t func)
	69	{
	70	struct io_ring_ctx *ctx = req->file->private_data;
	71	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
8579538c PB	72	struct task_struct *task = READ_ONCE(ctx->submitter_task);
	73
	74	if (unlikely(!task))
	75	return -EOWNERDEAD;
56d8e318 PB	76
	77	init_task_work(&msg->tw, func);
	78	if (task_work_add(ctx->submitter_task, &msg->tw, TWA_SIGNAL))
	79	return -EOWNERDEAD;
	80
	81	return IOU_ISSUE_SKIP_COMPLETE;
	82	}
	83
6d043ee1 PB	84	static void io_msg_tw_complete(struct callback_head *head)
	85	{
	86	struct io_msg *msg = container_of(head, struct io_msg, tw);
	87	struct io_kiocb *req = cmd_to_io_kiocb(msg);
	88	struct io_ring_ctx *target_ctx = req->file->private_data;
	89	int ret = 0;
	90
e12d7a46	91	if (current->flags & PF_EXITING) {
6d043ee1	92	ret = -EOWNERDEAD;
e12d7a46 JA	93	} else {
	94	/*
	95	* If the target ring is using IOPOLL mode, then we need to be
	96	* holding the uring_lock for posting completions. Other ring
	97	* types rely on the regular completion locking, which is
	98	* handled while posting.
	99	*/
	100	if (target_ctx->flags & IORING_SETUP_IOPOLL)
	101	mutex_lock(&target_ctx->uring_lock);
	102	if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
	103	ret = -EOVERFLOW;
	104	if (target_ctx->flags & IORING_SETUP_IOPOLL)
	105	mutex_unlock(&target_ctx->uring_lock);
	106	}
6d043ee1 PB	107
	108	if (ret < 0)
	109	req_set_fail(req);
	110	io_req_queue_tw_complete(req, ret);
	111	}
	112
e12d7a46	113	static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags)
e6130eba JA	114	{
e6130eba JA	115	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	116	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
e12d7a46	117	int ret;
e6130eba JA	118
	119	if (msg->src_fd \|\| msg->dst_fd \|\| msg->flags)
	120	return -EINVAL;
8579538c PB	121	if (target_ctx->flags & IORING_SETUP_R_DISABLED)
8579538c PB	122	return -EBADFD;
e6130eba	123
56d8e318 PB	124	if (io_msg_need_remote(target_ctx))
56d8e318 PB	125	return io_msg_exec_remote(req, io_msg_tw_complete);
6d043ee1	126
e12d7a46 JA	127	ret = -EOVERFLOW;
	128	if (target_ctx->flags & IORING_SETUP_IOPOLL) {
	129	if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
	130	return -EAGAIN;
	131	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
	132	ret = 0;
	133	io_double_unlock_ctx(target_ctx);
	134	} else {
	135	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
	136	ret = 0;
	137	}
	138	return ret;
e6130eba JA	139	}
e6130eba JA	140
11373026 PB	141	static struct file io_msg_grab_file(struct io_kiocb req, unsigned int issue_flags)
	142	{
	143	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	144	struct io_ring_ctx *ctx = req->ctx;
	145	struct file *file = NULL;
	146	unsigned long file_ptr;
	147	int idx = msg->src_fd;
	148
	149	io_ring_submit_lock(ctx, issue_flags);
	150	if (likely(idx < ctx->nr_user_files)) {
	151	idx = array_index_nospec(idx, ctx->nr_user_files);
	152	file_ptr = io_fixed_file_slot(&ctx->file_table, idx)->file_ptr;
	153	file = (struct file *) (file_ptr & FFS_MASK);
	154	if (file)
	155	get_file(file);
e6130eba	156	}
11373026 PB	157	io_ring_submit_unlock(ctx, issue_flags);
11373026 PB	158	return file;
e6130eba JA	159	}
e6130eba JA	160
17211310	161	static int io_msg_install_complete(struct io_kiocb *req, unsigned int issue_flags)
e6130eba JA	162	{
e6130eba JA	163	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	164	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
11373026	165	struct file *src_file = msg->src_file;
e6130eba JA	166	int ret;
e6130eba JA	167
11373026 PB	168	if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
11373026 PB	169	return -EAGAIN;
e6130eba JA	170
e6130eba JA	171	ret = __io_fixed_fd_install(target_ctx, src_file, msg->dst_fd);
11373026	172	if (ret < 0)
e6130eba	173	goto out_unlock;
17211310	174
11373026 PB	175	msg->src_file = NULL;
11373026 PB	176	req->flags &= ~REQ_F_NEED_CLEANUP;
e6130eba JA	177
	178	if (msg->flags & IORING_MSG_RING_CQE_SKIP)
	179	goto out_unlock;
e6130eba JA	180	/*
	181	* If this fails, the target still received the file descriptor but
	182	* wasn't notified of the fact. This means that if this request
	183	* completes with -EOVERFLOW, then the sender must ensure that a
	184	* later IORING_OP_MSG_RING delivers the message.
	185	*/
b529c96a	186	if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, 0))
e6130eba JA	187	ret = -EOVERFLOW;
e6130eba JA	188	out_unlock:
423d5081	189	io_double_unlock_ctx(target_ctx);
e6130eba JA	190	return ret;
	191	}
	192
6d043ee1 PB	193	static void io_msg_tw_fd_complete(struct callback_head *head)
	194	{
	195	struct io_msg *msg = container_of(head, struct io_msg, tw);
	196	struct io_kiocb *req = cmd_to_io_kiocb(msg);
	197	int ret = -EOWNERDEAD;
	198
	199	if (!(current->flags & PF_EXITING))
	200	ret = io_msg_install_complete(req, IO_URING_F_UNLOCKED);
	201	if (ret < 0)
	202	req_set_fail(req);
	203	io_req_queue_tw_complete(req, ret);
	204	}
	205
17211310 PB	206	static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags)
	207	{
	208	struct io_ring_ctx *target_ctx = req->file->private_data;
	209	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	210	struct io_ring_ctx *ctx = req->ctx;
	211	struct file *src_file = msg->src_file;
	212
	213	if (target_ctx == ctx)
	214	return -EINVAL;
8579538c PB	215	if (target_ctx->flags & IORING_SETUP_R_DISABLED)
8579538c PB	216	return -EBADFD;
17211310 PB	217	if (!src_file) {
	218	src_file = io_msg_grab_file(req, issue_flags);
	219	if (!src_file)
	220	return -EBADF;
	221	msg->src_file = src_file;
	222	req->flags \|= REQ_F_NEED_CLEANUP;
	223	}
6d043ee1	224
56d8e318 PB	225	if (io_msg_need_remote(target_ctx))
56d8e318 PB	226	return io_msg_exec_remote(req, io_msg_tw_fd_complete);
17211310 PB	227	return io_msg_install_complete(req, issue_flags);
	228	}
	229
36404b09 JA	230	int io_msg_ring_prep(struct io_kiocb req, const struct io_uring_sqe sqe)
36404b09 JA	231	{
f2ccb5ae	232	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09	233
e6130eba	234	if (unlikely(sqe->buf_index \|\| sqe->personality))
36404b09 JA	235	return -EINVAL;
36404b09 JA	236
11373026	237	msg->src_file = NULL;
36404b09 JA	238	msg->user_data = READ_ONCE(sqe->off);
36404b09 JA	239	msg->len = READ_ONCE(sqe->len);
e6130eba JA	240	msg->cmd = READ_ONCE(sqe->addr);
	241	msg->src_fd = READ_ONCE(sqe->addr3);
	242	msg->dst_fd = READ_ONCE(sqe->file_index);
	243	msg->flags = READ_ONCE(sqe->msg_ring_flags);
	244	if (msg->flags & ~IORING_MSG_RING_CQE_SKIP)
	245	return -EINVAL;
	246
36404b09 JA	247	return 0;
	248	}
	249
	250	int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags)
	251	{
f2ccb5ae	252	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09 JA	253	int ret;
	254
	255	ret = -EBADFD;
	256	if (!io_is_uring_fops(req->file))
	257	goto done;
	258
e6130eba JA	259	switch (msg->cmd) {
e6130eba JA	260	case IORING_MSG_DATA:
e12d7a46	261	ret = io_msg_ring_data(req, issue_flags);
e6130eba JA	262	break;
	263	case IORING_MSG_SEND_FD:
	264	ret = io_msg_send_fd(req, issue_flags);
	265	break;
	266	default:
	267	ret = -EINVAL;
	268	break;
	269	}
36404b09 JA	270
36404b09 JA	271	done:
6d043ee1 PB	272	if (ret < 0) {
	273	if (ret == -EAGAIN \|\| ret == IOU_ISSUE_SKIP_COMPLETE)
	274	return ret;
36404b09	275	req_set_fail(req);
6d043ee1	276	}
36404b09	277	io_req_set_res(req, ret, 0);
36404b09 JA	278	return IOU_OK;
36404b09 JA	279	}