[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"


/* All valid masks for MSG_RING */
#define IORING_MSG_RING_MASK		(IORING_MSG_RING_CQE_SKIP | \
					IORING_MSG_RING_FLAGS_PASS)

struct io_msg {
	struct file			*file;
	struct file			*src_file;
	struct callback_head		tw;
	u64 user_data;
	u32 len;
	u32 cmd;
	u32 src_fd;
	union {
		u32 dst_fd;
		u32 cqe_flags;
	};
	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 {
		u32 flags = 0;

		if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
			flags = msg->cqe_flags;

		/*
		 * 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, flags))
			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);
	u32 flags = 0;
	int ret;

	if (msg->src_fd || msg->flags & ~IORING_MSG_RING_FLAGS_PASS)
		return -EINVAL;
	if (!(msg->flags & IORING_MSG_RING_FLAGS_PASS) && msg->dst_fd)
		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);

	if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
		flags = msg->cqe_flags;

	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, flags))
			ret = 0;
		io_double_unlock_ctx(target_ctx);
	} else {
		if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
			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;
	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 = io_file_from_index(&ctx->file_table, idx);
		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, ret, 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 (msg->len)
		return -EINVAL;
	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_MASK)
		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"
36404b09 JA	15
cbeb47a7 BL	16
	17	/* All valid masks for MSG_RING */
	18	#define IORING_MSG_RING_MASK (IORING_MSG_RING_CQE_SKIP \| \
	19	IORING_MSG_RING_FLAGS_PASS)
	20
36404b09 JA	21	struct io_msg {
36404b09 JA	22	struct file *file;
11373026	23	struct file *src_file;
6d043ee1	24	struct callback_head tw;
36404b09 JA	25	u64 user_data;
36404b09 JA	26	u32 len;
e6130eba JA	27	u32 cmd;
e6130eba JA	28	u32 src_fd;
cbeb47a7 BL	29	union {
	30	u32 dst_fd;
	31	u32 cqe_flags;
	32	};
e6130eba	33	u32 flags;
36404b09 JA	34	};
36404b09 JA	35
423d5081 JA	36	static void io_double_unlock_ctx(struct io_ring_ctx *octx)
	37	{
	38	mutex_unlock(&octx->uring_lock);
	39	}
	40
	41	static int io_double_lock_ctx(struct io_ring_ctx *octx,
	42	unsigned int issue_flags)
	43	{
	44	/*
	45	* To ensure proper ordering between the two ctxs, we can only
	46	* attempt a trylock on the target. If that fails and we already have
	47	* the source ctx lock, punt to io-wq.
	48	*/
	49	if (!(issue_flags & IO_URING_F_UNLOCKED)) {
	50	if (!mutex_trylock(&octx->uring_lock))
	51	return -EAGAIN;
	52	return 0;
	53	}
	54	mutex_lock(&octx->uring_lock);
	55	return 0;
	56	}
	57
11373026 PB	58	void io_msg_ring_cleanup(struct io_kiocb *req)
	59	{
	60	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	61
	62	if (WARN_ON_ONCE(!msg->src_file))
	63	return;
	64
	65	fput(msg->src_file);
	66	msg->src_file = NULL;
	67	}
	68
56d8e318 PB	69	static inline bool io_msg_need_remote(struct io_ring_ctx *target_ctx)
	70	{
	71	if (!target_ctx->task_complete)
	72	return false;
	73	return current != target_ctx->submitter_task;
	74	}
	75
	76	static int io_msg_exec_remote(struct io_kiocb *req, task_work_func_t func)
	77	{
	78	struct io_ring_ctx *ctx = req->file->private_data;
	79	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
8579538c PB	80	struct task_struct *task = READ_ONCE(ctx->submitter_task);
	81
	82	if (unlikely(!task))
	83	return -EOWNERDEAD;
56d8e318 PB	84
	85	init_task_work(&msg->tw, func);
	86	if (task_work_add(ctx->submitter_task, &msg->tw, TWA_SIGNAL))
	87	return -EOWNERDEAD;
	88
	89	return IOU_ISSUE_SKIP_COMPLETE;
	90	}
	91
6d043ee1 PB	92	static void io_msg_tw_complete(struct callback_head *head)
	93	{
	94	struct io_msg *msg = container_of(head, struct io_msg, tw);
	95	struct io_kiocb *req = cmd_to_io_kiocb(msg);
	96	struct io_ring_ctx *target_ctx = req->file->private_data;
	97	int ret = 0;
	98
e12d7a46	99	if (current->flags & PF_EXITING) {
6d043ee1	100	ret = -EOWNERDEAD;
e12d7a46	101	} else {
8572df94 JA	102	u32 flags = 0;
	103
	104	if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
	105	flags = msg->cqe_flags;
	106
e12d7a46 JA	107	/*
	108	* If the target ring is using IOPOLL mode, then we need to be
	109	* holding the uring_lock for posting completions. Other ring
	110	* types rely on the regular completion locking, which is
	111	* handled while posting.
	112	*/
	113	if (target_ctx->flags & IORING_SETUP_IOPOLL)
	114	mutex_lock(&target_ctx->uring_lock);
8572df94	115	if (!io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
e12d7a46 JA	116	ret = -EOVERFLOW;
	117	if (target_ctx->flags & IORING_SETUP_IOPOLL)
	118	mutex_unlock(&target_ctx->uring_lock);
	119	}
6d043ee1 PB	120
	121	if (ret < 0)
	122	req_set_fail(req);
	123	io_req_queue_tw_complete(req, ret);
	124	}
	125
e12d7a46	126	static int io_msg_ring_data(struct io_kiocb *req, unsigned int issue_flags)
e6130eba JA	127	{
e6130eba JA	128	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	129	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
cbeb47a7	130	u32 flags = 0;
e12d7a46	131	int ret;
e6130eba	132
cbeb47a7 BL	133	if (msg->src_fd \|\| msg->flags & ~IORING_MSG_RING_FLAGS_PASS)
	134	return -EINVAL;
	135	if (!(msg->flags & IORING_MSG_RING_FLAGS_PASS) && msg->dst_fd)
e6130eba	136	return -EINVAL;
8579538c PB	137	if (target_ctx->flags & IORING_SETUP_R_DISABLED)
8579538c PB	138	return -EBADFD;
e6130eba	139
56d8e318 PB	140	if (io_msg_need_remote(target_ctx))
56d8e318 PB	141	return io_msg_exec_remote(req, io_msg_tw_complete);
6d043ee1	142
cbeb47a7 BL	143	if (msg->flags & IORING_MSG_RING_FLAGS_PASS)
	144	flags = msg->cqe_flags;
	145
e12d7a46 JA	146	ret = -EOVERFLOW;
	147	if (target_ctx->flags & IORING_SETUP_IOPOLL) {
	148	if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
	149	return -EAGAIN;
cbeb47a7	150	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
e12d7a46 JA	151	ret = 0;
	152	io_double_unlock_ctx(target_ctx);
	153	} else {
cbeb47a7	154	if (io_post_aux_cqe(target_ctx, msg->user_data, msg->len, flags))
e12d7a46 JA	155	ret = 0;
	156	}
	157	return ret;
e6130eba JA	158	}
e6130eba JA	159
11373026 PB	160	static struct file io_msg_grab_file(struct io_kiocb req, unsigned int issue_flags)
	161	{
	162	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	163	struct io_ring_ctx *ctx = req->ctx;
	164	struct file *file = NULL;
11373026 PB	165	int idx = msg->src_fd;
	166
	167	io_ring_submit_lock(ctx, issue_flags);
	168	if (likely(idx < ctx->nr_user_files)) {
	169	idx = array_index_nospec(idx, ctx->nr_user_files);
f432c8c8	170	file = io_file_from_index(&ctx->file_table, idx);
11373026 PB	171	if (file)
11373026 PB	172	get_file(file);
e6130eba	173	}
11373026 PB	174	io_ring_submit_unlock(ctx, issue_flags);
11373026 PB	175	return file;
e6130eba JA	176	}
e6130eba JA	177
17211310	178	static int io_msg_install_complete(struct io_kiocb *req, unsigned int issue_flags)
e6130eba JA	179	{
e6130eba JA	180	struct io_ring_ctx *target_ctx = req->file->private_data;
f2ccb5ae	181	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
11373026	182	struct file *src_file = msg->src_file;
e6130eba JA	183	int ret;
e6130eba JA	184
11373026 PB	185	if (unlikely(io_double_lock_ctx(target_ctx, issue_flags)))
11373026 PB	186	return -EAGAIN;
e6130eba JA	187
e6130eba JA	188	ret = __io_fixed_fd_install(target_ctx, src_file, msg->dst_fd);
11373026	189	if (ret < 0)
e6130eba	190	goto out_unlock;
17211310	191
11373026 PB	192	msg->src_file = NULL;
11373026 PB	193	req->flags &= ~REQ_F_NEED_CLEANUP;
e6130eba JA	194
	195	if (msg->flags & IORING_MSG_RING_CQE_SKIP)
	196	goto out_unlock;
e6130eba JA	197	/*
	198	* If this fails, the target still received the file descriptor but
	199	* wasn't notified of the fact. This means that if this request
	200	* completes with -EOVERFLOW, then the sender must ensure that a
	201	* later IORING_OP_MSG_RING delivers the message.
	202	*/
5da28edd	203	if (!io_post_aux_cqe(target_ctx, msg->user_data, ret, 0))
e6130eba JA	204	ret = -EOVERFLOW;
e6130eba JA	205	out_unlock:
423d5081	206	io_double_unlock_ctx(target_ctx);
e6130eba JA	207	return ret;
	208	}
	209
6d043ee1 PB	210	static void io_msg_tw_fd_complete(struct callback_head *head)
	211	{
	212	struct io_msg *msg = container_of(head, struct io_msg, tw);
	213	struct io_kiocb *req = cmd_to_io_kiocb(msg);
	214	int ret = -EOWNERDEAD;
	215
	216	if (!(current->flags & PF_EXITING))
	217	ret = io_msg_install_complete(req, IO_URING_F_UNLOCKED);
	218	if (ret < 0)
	219	req_set_fail(req);
	220	io_req_queue_tw_complete(req, ret);
	221	}
	222
17211310 PB	223	static int io_msg_send_fd(struct io_kiocb *req, unsigned int issue_flags)
	224	{
	225	struct io_ring_ctx *target_ctx = req->file->private_data;
	226	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
	227	struct io_ring_ctx *ctx = req->ctx;
	228	struct file *src_file = msg->src_file;
	229
5da28edd PB	230	if (msg->len)
5da28edd PB	231	return -EINVAL;
17211310 PB	232	if (target_ctx == ctx)
17211310 PB	233	return -EINVAL;
8579538c PB	234	if (target_ctx->flags & IORING_SETUP_R_DISABLED)
8579538c PB	235	return -EBADFD;
17211310 PB	236	if (!src_file) {
	237	src_file = io_msg_grab_file(req, issue_flags);
	238	if (!src_file)
	239	return -EBADF;
	240	msg->src_file = src_file;
	241	req->flags \|= REQ_F_NEED_CLEANUP;
	242	}
6d043ee1	243
56d8e318 PB	244	if (io_msg_need_remote(target_ctx))
56d8e318 PB	245	return io_msg_exec_remote(req, io_msg_tw_fd_complete);
17211310 PB	246	return io_msg_install_complete(req, issue_flags);
	247	}
	248
36404b09 JA	249	int io_msg_ring_prep(struct io_kiocb req, const struct io_uring_sqe sqe)
36404b09 JA	250	{
f2ccb5ae	251	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09	252
e6130eba	253	if (unlikely(sqe->buf_index \|\| sqe->personality))
36404b09 JA	254	return -EINVAL;
36404b09 JA	255
11373026	256	msg->src_file = NULL;
36404b09 JA	257	msg->user_data = READ_ONCE(sqe->off);
36404b09 JA	258	msg->len = READ_ONCE(sqe->len);
e6130eba JA	259	msg->cmd = READ_ONCE(sqe->addr);
	260	msg->src_fd = READ_ONCE(sqe->addr3);
	261	msg->dst_fd = READ_ONCE(sqe->file_index);
	262	msg->flags = READ_ONCE(sqe->msg_ring_flags);
cbeb47a7	263	if (msg->flags & ~IORING_MSG_RING_MASK)
e6130eba JA	264	return -EINVAL;
e6130eba JA	265
36404b09 JA	266	return 0;
	267	}
	268
	269	int io_msg_ring(struct io_kiocb *req, unsigned int issue_flags)
	270	{
f2ccb5ae	271	struct io_msg *msg = io_kiocb_to_cmd(req, struct io_msg);
36404b09 JA	272	int ret;
	273
	274	ret = -EBADFD;
	275	if (!io_is_uring_fops(req->file))
	276	goto done;
	277
e6130eba JA	278	switch (msg->cmd) {
e6130eba JA	279	case IORING_MSG_DATA:
e12d7a46	280	ret = io_msg_ring_data(req, issue_flags);
e6130eba JA	281	break;
	282	case IORING_MSG_SEND_FD:
	283	ret = io_msg_send_fd(req, issue_flags);
	284	break;
	285	default:
	286	ret = -EINVAL;
	287	break;
	288	}
36404b09 JA	289
36404b09 JA	290	done:
6d043ee1 PB	291	if (ret < 0) {
	292	if (ret == -EAGAIN \|\| ret == IOU_ISSUE_SKIP_COMPLETE)
	293	return ret;
36404b09	294	req_set_fail(req);
6d043ee1	295	}
36404b09	296	io_req_set_res(req, ret, 0);
36404b09 JA	297	return IOU_OK;
36404b09 JA	298	}