[PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[Jens Axboe]

This behaves like IORING_OP_READ, except:

1) It only supports pollable files (eg pipes, sockets, etc). Note that
   for sockets, you probably want to use recv/recvmsg with multishot
   instead.

2) It supports multishot mode, meaning it will repeatedly trigger a
   read and fill a buffer when data is available. This allows similar
   use to recv/recvmsg but on non-sockets, where a single request will
   repeatedly post a CQE whenever data is read from it.

3) Because of #2, it must be used with provided buffers. This is
   uniformly true across any request type that supports multishot and
   transfers data, with the reason being that it's obviously not
   possible to pass in a single buffer for the data, as multiple reads
   may very well trigger before an application has a chance to process
   previous CQEs and the data passed from them.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/uapi/linux/io_uring.h |  1 +
 io_uring/opdef.c              | 13 +++++++
 io_uring/rw.c                 | 66 +++++++++++++++++++++++++++++++++++
 io_uring/rw.h                 |  2 ++
 4 files changed, 82 insertions(+)

diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index daa363d1a502..c35438af679a 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -246,6 +246,7 @@ enum io_uring_op {
 	IORING_OP_FUTEX_WAIT,
 	IORING_OP_FUTEX_WAKE,
 	IORING_OP_FUTEX_WAITV,
+	IORING_OP_READ_MULTISHOT,
 
 	/* this goes last, obviously */
 	IORING_OP_LAST,
diff --git a/io_uring/opdef.c b/io_uring/opdef.c
index bfb7c53389c0..03e1a6f26fa5 100644
--- a/io_uring/opdef.c
+++ b/io_uring/opdef.c
@@ -460,6 +460,16 @@ const struct io_issue_def io_issue_defs[] = {
 		.prep			= io_eopnotsupp_prep,
 #endif
 	},
+	[IORING_OP_READ_MULTISHOT] = {
+		.needs_file		= 1,
+		.unbound_nonreg_file	= 1,
+		.pollin			= 1,
+		.buffer_select		= 1,
+		.audit_skip		= 1,
+		.ioprio			= 1,
+		.prep			= io_read_mshot_prep,
+		.issue			= io_read_mshot,
+	},
 };
 
 const struct io_cold_def io_cold_defs[] = {
@@ -692,6 +702,9 @@ const struct io_cold_def io_cold_defs[] = {
 	[IORING_OP_FUTEX_WAITV] = {
 		.name			= "FUTEX_WAITV",
 	},
+	[IORING_OP_READ_MULTISHOT] = {
+		.name			= "READ_MULTISHOT",
+	},
 };
 
 const char *io_uring_get_opcode(u8 opcode)
diff --git a/io_uring/rw.c b/io_uring/rw.c
index c3bf38419230..7305792fbbbf 100644
--- a/io_uring/rw.c
+++ b/io_uring/rw.c
@@ -123,6 +123,22 @@ int io_prep_rw(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	return 0;
 }
 
+/*
+ * Multishot read is prepared just like a normal read/write request, only
+ * difference is that we set the MULTISHOT flag.
+ */
+int io_read_mshot_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+{
+	int ret;
+
+	ret = io_prep_rw(req, sqe);
+	if (unlikely(ret))
+		return ret;
+
+	req->flags |= REQ_F_APOLL_MULTISHOT;
+	return 0;
+}
+
 void io_readv_writev_cleanup(struct io_kiocb *req)
 {
 	struct io_async_rw *io = req->async_data;
@@ -869,6 +885,56 @@ int io_read(struct io_kiocb *req, unsigned int issue_flags)
 	return kiocb_done(req, ret, issue_flags);
 }
 
+int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
+{
+	unsigned int cflags = 0;
+	int ret;
+
+	/*
+	 * Multishot MUST be used on a pollable file
+	 */
+	if (!file_can_poll(req->file))
+		return -EBADFD;
+
+	ret = __io_read(req, issue_flags);
+
+	/*
+	 * If we get -EAGAIN, recycle our buffer and just let normal poll
+	 * handling arm it.
+	 */
+	if (ret == -EAGAIN) {
+		io_kbuf_recycle(req, issue_flags);
+		return -EAGAIN;
+	}
+
+	/*
+	 * Any error will terminate a multishot request
+	 */
+	if (ret <= 0) {
+finish:
+		io_req_set_res(req, ret, cflags);
+		if (issue_flags & IO_URING_F_MULTISHOT)
+			return IOU_STOP_MULTISHOT;
+		return IOU_OK;
+	}
+
+	/*
+	 * Put our buffer and post a CQE. If we fail to post a CQE, then
+	 * jump to the termination path. This request is then done.
+	 */
+	cflags = io_put_kbuf(req, issue_flags);
+
+	if (io_fill_cqe_req_aux(req, issue_flags & IO_URING_F_COMPLETE_DEFER,
+				ret, cflags | IORING_CQE_F_MORE)) {
+		if (issue_flags & IO_URING_F_MULTISHOT)
+			return IOU_ISSUE_SKIP_COMPLETE;
+		else
+			return -EAGAIN;
+	}
+
+	goto finish;
+}
+
 int io_write(struct io_kiocb *req, unsigned int issue_flags)
 {
 	struct io_rw *rw = io_kiocb_to_cmd(req, struct io_rw);
diff --git a/io_uring/rw.h b/io_uring/rw.h
index 4b89f9659366..c5aed03d42a4 100644
--- a/io_uring/rw.h
+++ b/io_uring/rw.h
@@ -23,3 +23,5 @@ int io_writev_prep_async(struct io_kiocb *req);
 void io_readv_writev_cleanup(struct io_kiocb *req);
 void io_rw_fail(struct io_kiocb *req);
 void io_req_rw_complete(struct io_kiocb *req, struct io_tw_state *ts);
+int io_read_mshot_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
+int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags);
-- 
2.40.1

Re: [PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[Gabriel Krisman Bertazi]

Jens Axboe <axboe@kernel.dk> writes:

> This behaves like IORING_OP_READ, except:
>
> 1) It only supports pollable files (eg pipes, sockets, etc). Note that
>    for sockets, you probably want to use recv/recvmsg with multishot
>    instead.
>
> 2) It supports multishot mode, meaning it will repeatedly trigger a
>    read and fill a buffer when data is available. This allows similar
>    use to recv/recvmsg but on non-sockets, where a single request will
>    repeatedly post a CQE whenever data is read from it.
>
> 3) Because of #2, it must be used with provided buffers. This is
>    uniformly true across any request type that supports multishot and
>    transfers data, with the reason being that it's obviously not
>    possible to pass in a single buffer for the data, as multiple reads
>    may very well trigger before an application has a chance to process
>    previous CQEs and the data passed from them.
>
> Signed-off-by: Jens Axboe <axboe@kernel.dk>

This is a really cool feature.  Just two comments inline.


> +/*
> + * Multishot read is prepared just like a normal read/write request, only
> + * difference is that we set the MULTISHOT flag.
> + */
> +int io_read_mshot_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
> +{
> +	int ret;
> +
> +	ret = io_prep_rw(req, sqe);
> +	if (unlikely(ret))
> +		return ret;
> +
> +	req->flags |= REQ_F_APOLL_MULTISHOT;
> +	return 0;
> +}
> +
>  void io_readv_writev_cleanup(struct io_kiocb *req)
>  {
>  	struct io_async_rw *io = req->async_data;
> @@ -869,6 +885,56 @@ int io_read(struct io_kiocb *req, unsigned int issue_flags)
>  	return kiocb_done(req, ret, issue_flags);
>  }
>  
> +int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
> +{
> +	unsigned int cflags = 0;
> +	int ret;
> +
> +	/*
> +	 * Multishot MUST be used on a pollable file
> +	 */
> +	if (!file_can_poll(req->file))
> +		return -EBADFD;

io_uring is pollable, so I think you want to also reject when
req->file->f_ops == io_uring_fops to avoid the loop where a ring
monitoring itself will cause a recursive completion? Maybe this can't
happen here for some reason I miss?

> +
> +	ret = __io_read(req, issue_flags);
> +
> +	/*
> +	 * If we get -EAGAIN, recycle our buffer and just let normal poll
> +	 * handling arm it.
> +	 */
> +	if (ret == -EAGAIN) {
> +		io_kbuf_recycle(req, issue_flags);
> +		return -EAGAIN;
> +	}
> +
> +	/*
> +	 * Any error will terminate a multishot request
> +	 */
> +	if (ret <= 0) {
> +finish:
> +		io_req_set_res(req, ret, cflags);
> +		if (issue_flags & IO_URING_F_MULTISHOT)
> +			return IOU_STOP_MULTISHOT;
> +		return IOU_OK;

Just a style detail, but I'd prefer to unfold this on the end of the function
instead of jumping backwards here..

> +	}
> +
> +	/*
> +	 * Put our buffer and post a CQE. If we fail to post a CQE, then
> +	 * jump to the termination path. This request is then done.
> +	 */
> +	cflags = io_put_kbuf(req, issue_flags);
> +
> +	if (io_fill_cqe_req_aux(req, issue_flags & IO_URING_F_COMPLETE_DEFER,
> +				ret, cflags | IORING_CQE_F_MORE)) {
> +		if (issue_flags & IO_URING_F_MULTISHOT)
> +			return IOU_ISSUE_SKIP_COMPLETE;
> +		else
> +			return -EAGAIN;
> +	}
> +
> +	goto finish;
> +}
> +
>  int io_write(struct io_kiocb *req, unsigned int issue_flags)
>  {
>  	struct io_rw *rw = io_kiocb_to_cmd(req, struct io_rw);
> diff --git a/io_uring/rw.h b/io_uring/rw.h
> index 4b89f9659366..c5aed03d42a4 100644
> --- a/io_uring/rw.h
> +++ b/io_uring/rw.h
> @@ -23,3 +23,5 @@ int io_writev_prep_async(struct io_kiocb *req);
>  void io_readv_writev_cleanup(struct io_kiocb *req);
>  void io_rw_fail(struct io_kiocb *req);
>  void io_req_rw_complete(struct io_kiocb *req, struct io_tw_state *ts);
> +int io_read_mshot_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
> +int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags);

-- 
Gabriel Krisman Bertazi

Re: [PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[Gabriel Krisman Bertazi]

Jens Axboe <axboe@kernel.dk> writes:

> This behaves like IORING_OP_READ, except:
>
> 1) It only supports pollable files (eg pipes, sockets, etc). Note that
>    for sockets, you probably want to use recv/recvmsg with multishot
>    instead.
>
> 2) It supports multishot mode, meaning it will repeatedly trigger a
>    read and fill a buffer when data is available. This allows similar
>    use to recv/recvmsg but on non-sockets, where a single request will
>    repeatedly post a CQE whenever data is read from it.
>
> 3) Because of #2, it must be used with provided buffers. This is
>    uniformly true across any request type that supports multishot and
>    transfers data, with the reason being that it's obviously not
>    possible to pass in a single buffer for the data, as multiple reads
>    may very well trigger before an application has a chance to process
>    previous CQEs and the data passed from them.
>
> Signed-off-by: Jens Axboe <axboe@kernel.dk>
> ---
>  include/uapi/linux/io_uring.h |  1 +
>  io_uring/opdef.c              | 13 +++++++
>  io_uring/rw.c                 | 66 +++++++++++++++++++++++++++++++++++
>  io_uring/rw.h                 |  2 ++
>  4 files changed, 82 insertions(+)
>
> diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
> index daa363d1a502..c35438af679a 100644
> --- a/include/uapi/linux/io_uring.h
> +++ b/include/uapi/linux/io_uring.h
> @@ -246,6 +246,7 @@ enum io_uring_op {
>  	IORING_OP_FUTEX_WAIT,
>  	IORING_OP_FUTEX_WAKE,
>  	IORING_OP_FUTEX_WAITV,
> +	IORING_OP_READ_MULTISHOT,
>  
>  	/* this goes last, obviously */
>  	IORING_OP_LAST,
> diff --git a/io_uring/opdef.c b/io_uring/opdef.c
> index bfb7c53389c0..03e1a6f26fa5 100644
> --- a/io_uring/opdef.c
> +++ b/io_uring/opdef.c
> @@ -460,6 +460,16 @@ const struct io_issue_def io_issue_defs[] = {
>  		.prep			= io_eopnotsupp_prep,
>  #endif
>  	},
> +	[IORING_OP_READ_MULTISHOT] = {
> +		.needs_file		= 1,
> +		.unbound_nonreg_file	= 1,
> +		.pollin			= 1,
> +		.buffer_select		= 1,
> +		.audit_skip		= 1,
> +		.ioprio			= 1,
> +		.prep			= io_read_mshot_prep,
> +		.issue			= io_read_mshot,
> +	},
>  };
>  
>  const struct io_cold_def io_cold_defs[] = {
> @@ -692,6 +702,9 @@ const struct io_cold_def io_cold_defs[] = {
>  	[IORING_OP_FUTEX_WAITV] = {
>  		.name			= "FUTEX_WAITV",
>  	},
> +	[IORING_OP_READ_MULTISHOT] = {
> +		.name			= "READ_MULTISHOT",
> +	},
>  };
>  
>  const char *io_uring_get_opcode(u8 opcode)
> diff --git a/io_uring/rw.c b/io_uring/rw.c
> index c3bf38419230..7305792fbbbf 100644
> --- a/io_uring/rw.c
> +++ b/io_uring/rw.c
> @@ -123,6 +123,22 @@ int io_prep_rw(struct io_kiocb *req, const struct io_uring_sqe *sqe)
>  	return 0;
>  }
>  
> +/*
> + * Multishot read is prepared just like a normal read/write request, only
> + * difference is that we set the MULTISHOT flag.
> + */
> +int io_read_mshot_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
> +{
> +	int ret;
> +
> +	ret = io_prep_rw(req, sqe);
> +	if (unlikely(ret))
> +		return ret;
> +
> +	req->flags |= REQ_F_APOLL_MULTISHOT;
> +	return 0;
> +}
> +
>  void io_readv_writev_cleanup(struct io_kiocb *req)
>  {
>  	struct io_async_rw *io = req->async_data;
> @@ -869,6 +885,56 @@ int io_read(struct io_kiocb *req, unsigned int issue_flags)
>  	return kiocb_done(req, ret, issue_flags);
>  }
>  
> +int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
> +{
> +	unsigned int cflags = 0;
> +	int ret;
> +
> +	/*
> +	 * Multishot MUST be used on a pollable file
> +	 */
> +	if (!file_can_poll(req->file))
> +		return -EBADFD;
> +

Please disregard my previous comment about checking for
io_uring_fops. It is not necessary because this kind of file can't be
read in the first place, so it would never get here.

(Also, things seems to be misbehaving on my MUA archive and Lore didn't
get my own message yet, so I'm not replying directly to it)

-- 
Gabriel Krisman Bertazi

Re: [PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[Jens Axboe]

On 9/11/23 5:57 PM, Gabriel Krisman Bertazi wrote:
>> +int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
>> +{
>> +	unsigned int cflags = 0;
>> +	int ret;
>> +
>> +	/*
>> +	 * Multishot MUST be used on a pollable file
>> +	 */
>> +	if (!file_can_poll(req->file))
>> +		return -EBADFD;
> 
> io_uring is pollable, so I think you want to also reject when
> req->file->f_ops == io_uring_fops to avoid the loop where a ring
> monitoring itself will cause a recursive completion? Maybe this can't
> happen here for some reason I miss?

I saw your followup, but we do actually handle that case - if this fd is
an io_uring context, then we track the inflight state of it so we can
appropriately cancel to break that loop.

But yeah, doesn't matter for this case, as you cannot read or write to
an io_uring fd in the first place.

>> +	ret = __io_read(req, issue_flags);
>> +
>> +	/*
>> +	 * If we get -EAGAIN, recycle our buffer and just let normal poll
>> +	 * handling arm it.
>> +	 */
>> +	if (ret == -EAGAIN) {
>> +		io_kbuf_recycle(req, issue_flags);
>> +		return -EAGAIN;
>> +	}
>> +
>> +	/*
>> +	 * Any error will terminate a multishot request
>> +	 */
>> +	if (ret <= 0) {
>> +finish:
>> +		io_req_set_res(req, ret, cflags);
>> +		if (issue_flags & IO_URING_F_MULTISHOT)
>> +			return IOU_STOP_MULTISHOT;
>> +		return IOU_OK;
> 
> Just a style detail, but I'd prefer to unfold this on the end of the function
> instead of jumping backwards here..

Sure, that might look better. I'll make the edit.

-- 
Jens Axboe

Re: [PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[Jens Axboe]

On 9/11/23 6:38 PM, Gabriel Krisman Bertazi wrote:
> (Also, things seems to be misbehaving on my MUA archive and Lore didn't
> get my own message yet, so I'm not replying directly to it)

I'm having all sorts of delays too, both last week and this week...
Pretty unfortunate for probably the most core of kernel development
infrastructure.

-- 
Jens Axboe

Re: [PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[Jens Axboe]

On 9/11/23 6:46 PM, Jens Axboe wrote:
>>> +	ret = __io_read(req, issue_flags);
>>> +
>>> +	/*
>>> +	 * If we get -EAGAIN, recycle our buffer and just let normal poll
>>> +	 * handling arm it.
>>> +	 */
>>> +	if (ret == -EAGAIN) {
>>> +		io_kbuf_recycle(req, issue_flags);
>>> +		return -EAGAIN;
>>> +	}
>>> +
>>> +	/*
>>> +	 * Any error will terminate a multishot request
>>> +	 */
>>> +	if (ret <= 0) {
>>> +finish:
>>> +		io_req_set_res(req, ret, cflags);
>>> +		if (issue_flags & IO_URING_F_MULTISHOT)
>>> +			return IOU_STOP_MULTISHOT;
>>> +		return IOU_OK;
>>
>> Just a style detail, but I'd prefer to unfold this on the end of the
>> function instead of jumping backwards here..
> 
> Sure, that might look better. I'll make the edit.

Actually we can just indent the next case and get rid of the goto
completely:

int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
{
	unsigned int cflags = 0;
	int ret;

	/*
	 * Multishot MUST be used on a pollable file
	 */
	if (!file_can_poll(req->file))
		return -EBADFD;

	ret = __io_read(req, issue_flags);

	/*
	 * If we get -EAGAIN, recycle our buffer and just let normal poll
	 * handling arm it.
	 */
	if (ret == -EAGAIN) {
		io_kbuf_recycle(req, issue_flags);
		return -EAGAIN;
	}

	/*
	 * Any successful return value will keep the multishot read armed.
	 */
	if (ret > 0) {
		/*
		 * Put our buffer and post a CQE. If we fail to post a CQE, then
		 * jump to the termination path. This request is then done.
		 */
		cflags = io_put_kbuf(req, issue_flags);

		if (io_fill_cqe_req_aux(req,
					issue_flags & IO_URING_F_COMPLETE_DEFER,
					ret, cflags | IORING_CQE_F_MORE)) {
			if (issue_flags & IO_URING_F_MULTISHOT)
				return IOU_ISSUE_SKIP_COMPLETE;
			return -EAGAIN;
		}
	}

	/*
	 * Either an error, or we've hit overflow posting the CQE. For any
	 * multishot request, hitting overflow will terminate it.
	 */
	io_req_set_res(req, ret, cflags);
	if (issue_flags & IO_URING_F_MULTISHOT)
		return IOU_STOP_MULTISHOT;
	return IOU_OK;
}

-- 
Jens Axboe

[PATCHSET v2 0/3] Add support for multishot reads

[Jens Axboe]

Hi,

We support multishot for other request types, generally in the shape of
a flag for the request. Doing a flag based approach with reads isn't
straightforward, as the read/write flags are in the RWF_ space. Instead,
add a separate opcode for this, IORING_OP_READ_MULTISHOT.

This can only be used provided buffers, like other multishot request
types that read/receive data.

It can also only be used for pollable file types, like a tun device or
pipes, for example. File types that are always readable (or seekable),
like regular files, cannot be used with multishot reads.

This is based on the io_uring-futex branch (which, in turn, is based on
the io_uring-waitid branch). No dependencies as such between them,
except the opcode numbering.

Can also be found here:

https://git.kernel.dk/cgit/linux/log/?h=io_uring-mshot-read

and there's a liburing branch with some basic support and some test
cases here:

https://git.kernel.dk/cgit/liburing/log/?h=read-mshot

 include/uapi/linux/io_uring.h |  1 +
 io_uring/opdef.c              | 14 ++++++
 io_uring/opdef.h              |  2 +
 io_uring/rw.c                 | 92 ++++++++++++++++++++++++++++++++---
 io_uring/rw.h                 |  2 +
 5 files changed, 105 insertions(+), 6 deletions(-)

Changes since v1:
- Code reformatting
- Add/expand a few comments

-- 
Jens Axboe

[PATCH 1/3] io_uring/rw: split io_read() into a helper

[Jens Axboe]

Add __io_read() which does the grunt of the work, leaving the completion
side to the new io_read(). No functional changes in this patch.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 io_uring/rw.c | 13 ++++++++++++-
 1 file changed, 12 insertions(+), 1 deletion(-)

diff --git a/io_uring/rw.c b/io_uring/rw.c
index c8c822fa7980..402e8bf002d6 100644
--- a/io_uring/rw.c
+++ b/io_uring/rw.c
@@ -708,7 +708,7 @@ static int io_rw_init_file(struct io_kiocb *req, fmode_t mode)
 	return 0;
 }
 
-int io_read(struct io_kiocb *req, unsigned int issue_flags)
+static int __io_read(struct io_kiocb *req, unsigned int issue_flags)
 {
 	struct io_rw *rw = io_kiocb_to_cmd(req, struct io_rw);
 	struct io_rw_state __s, *s = &__s;
@@ -853,6 +853,17 @@ int io_read(struct io_kiocb *req, unsigned int issue_flags)
 	/* it's faster to check here then delegate to kfree */
 	if (iovec)
 		kfree(iovec);
+	return ret;
+}
+
+int io_read(struct io_kiocb *req, unsigned int issue_flags)
+{
+	int ret;
+
+	ret = __io_read(req, issue_flags);
+	if (unlikely(ret < 0))
+		return ret;
+
 	return kiocb_done(req, ret, issue_flags);
 }
 
-- 
2.40.1

[PATCH 2/3] io_uring/rw: mark readv/writev as vectored in the opcode definition

[Jens Axboe]

This is cleaner than gating on the opcode type, particularly as more
read/write type opcodes may be added.

Then we can use that for the data import, and for __io_read() on
whether or not we need to copy state.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 io_uring/opdef.c |  2 ++
 io_uring/opdef.h |  2 ++
 io_uring/rw.c    | 10 ++++++----
 3 files changed, 10 insertions(+), 4 deletions(-)

diff --git a/io_uring/opdef.c b/io_uring/opdef.c
index 1d26ef10fc16..bfb7c53389c0 100644
--- a/io_uring/opdef.c
+++ b/io_uring/opdef.c
@@ -65,6 +65,7 @@ const struct io_issue_def io_issue_defs[] = {
 		.ioprio			= 1,
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
+		.vectored		= 1,
 		.prep			= io_prep_rw,
 		.issue			= io_read,
 	},
@@ -78,6 +79,7 @@ const struct io_issue_def io_issue_defs[] = {
 		.ioprio			= 1,
 		.iopoll			= 1,
 		.iopoll_queue		= 1,
+		.vectored		= 1,
 		.prep			= io_prep_rw,
 		.issue			= io_write,
 	},
diff --git a/io_uring/opdef.h b/io_uring/opdef.h
index c22c8696e749..9e5435ec27d0 100644
--- a/io_uring/opdef.h
+++ b/io_uring/opdef.h
@@ -29,6 +29,8 @@ struct io_issue_def {
 	unsigned		iopoll_queue : 1;
 	/* opcode specific path will handle ->async_data allocation if needed */
 	unsigned		manual_alloc : 1;
+	/* vectored opcode, set if 1) vectored, and 2) handler needs to know */
+	unsigned		vectored : 1;
 
 	int (*issue)(struct io_kiocb *, unsigned int);
 	int (*prep)(struct io_kiocb *, const struct io_uring_sqe *);
diff --git a/io_uring/rw.c b/io_uring/rw.c
index 402e8bf002d6..c3bf38419230 100644
--- a/io_uring/rw.c
+++ b/io_uring/rw.c
@@ -388,8 +388,7 @@ static struct iovec *__io_import_iovec(int ddir, struct io_kiocb *req,
 	buf = u64_to_user_ptr(rw->addr);
 	sqe_len = rw->len;
 
-	if (opcode == IORING_OP_READ || opcode == IORING_OP_WRITE ||
-	    (req->flags & REQ_F_BUFFER_SELECT)) {
+	if (!io_issue_defs[opcode].vectored || req->flags & REQ_F_BUFFER_SELECT) {
 		if (io_do_buffer_select(req)) {
 			buf = io_buffer_select(req, &sqe_len, issue_flags);
 			if (!buf)
@@ -776,8 +775,11 @@ static int __io_read(struct io_kiocb *req, unsigned int issue_flags)
 
 	if (ret == -EAGAIN || (req->flags & REQ_F_REISSUE)) {
 		req->flags &= ~REQ_F_REISSUE;
-		/* if we can poll, just do that */
-		if (req->opcode == IORING_OP_READ && file_can_poll(req->file))
+		/*
+		 * If we can poll, just do that. For a vectored read, we'll
+		 * need to copy state first.
+		 */
+		if (file_can_poll(req->file) && !io_issue_defs[req->opcode].vectored)
 			return -EAGAIN;
 		/* IOPOLL retry should happen for io-wq threads */
 		if (!force_nonblock && !(req->ctx->flags & IORING_SETUP_IOPOLL))
-- 
2.40.1

[PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[Jens Axboe]

This behaves like IORING_OP_READ, except:

1) It only supports pollable files (eg pipes, sockets, etc). Note that
   for sockets, you probably want to use recv/recvmsg with multishot
   instead.

2) It supports multishot mode, meaning it will repeatedly trigger a
   read and fill a buffer when data is available. This allows similar
   use to recv/recvmsg but on non-sockets, where a single request will
   repeatedly post a CQE whenever data is read from it.

3) Because of #2, it must be used with provided buffers. This is
   uniformly true across any request type that supports multishot and
   transfers data, with the reason being that it's obviously not
   possible to pass in a single buffer for the data, as multiple reads
   may very well trigger before an application has a chance to process
   previous CQEs and the data passed from them.

Signed-off-by: Jens Axboe <axboe@kernel.dk>
---
 include/uapi/linux/io_uring.h |  1 +
 io_uring/opdef.c              | 12 ++++++
 io_uring/rw.c                 | 73 +++++++++++++++++++++++++++++++++--
 io_uring/rw.h                 |  2 +
 4 files changed, 85 insertions(+), 3 deletions(-)

diff --git a/include/uapi/linux/io_uring.h b/include/uapi/linux/io_uring.h
index daa363d1a502..c35438af679a 100644
--- a/include/uapi/linux/io_uring.h
+++ b/include/uapi/linux/io_uring.h
@@ -246,6 +246,7 @@ enum io_uring_op {
 	IORING_OP_FUTEX_WAIT,
 	IORING_OP_FUTEX_WAKE,
 	IORING_OP_FUTEX_WAITV,
+	IORING_OP_READ_MULTISHOT,
 
 	/* this goes last, obviously */
 	IORING_OP_LAST,
diff --git a/io_uring/opdef.c b/io_uring/opdef.c
index bfb7c53389c0..74b3c1ef22b0 100644
--- a/io_uring/opdef.c
+++ b/io_uring/opdef.c
@@ -460,6 +460,15 @@ const struct io_issue_def io_issue_defs[] = {
 		.prep			= io_eopnotsupp_prep,
 #endif
 	},
+	[IORING_OP_READ_MULTISHOT] = {
+		.needs_file		= 1,
+		.unbound_nonreg_file	= 1,
+		.pollin			= 1,
+		.buffer_select		= 1,
+		.audit_skip		= 1,
+		.prep			= io_read_mshot_prep,
+		.issue			= io_read_mshot,
+	},
 };
 
 const struct io_cold_def io_cold_defs[] = {
@@ -692,6 +701,9 @@ const struct io_cold_def io_cold_defs[] = {
 	[IORING_OP_FUTEX_WAITV] = {
 		.name			= "FUTEX_WAITV",
 	},
+	[IORING_OP_READ_MULTISHOT] = {
+		.name			= "READ_MULTISHOT",
+	},
 };
 
 const char *io_uring_get_opcode(u8 opcode)
diff --git a/io_uring/rw.c b/io_uring/rw.c
index c3bf38419230..ec0cc38ea682 100644
--- a/io_uring/rw.c
+++ b/io_uring/rw.c
@@ -123,6 +123,22 @@ int io_prep_rw(struct io_kiocb *req, const struct io_uring_sqe *sqe)
 	return 0;
 }
 
+/*
+ * Multishot read is prepared just like a normal read/write request, only
+ * difference is that we set the MULTISHOT flag.
+ */
+int io_read_mshot_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
+{
+	int ret;
+
+	ret = io_prep_rw(req, sqe);
+	if (unlikely(ret))
+		return ret;
+
+	req->flags |= REQ_F_APOLL_MULTISHOT;
+	return 0;
+}
+
 void io_readv_writev_cleanup(struct io_kiocb *req)
 {
 	struct io_async_rw *io = req->async_data;
@@ -863,10 +879,61 @@ int io_read(struct io_kiocb *req, unsigned int issue_flags)
 	int ret;
 
 	ret = __io_read(req, issue_flags);
-	if (unlikely(ret < 0))
-		return ret;
+	if (ret >= 0)
+		return kiocb_done(req, ret, issue_flags);
+
+	return ret;
+}
+
+int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags)
+{
+	unsigned int cflags = 0;
+	int ret;
+
+	/*
+	 * Multishot MUST be used on a pollable file
+	 */
+	if (!file_can_poll(req->file))
+		return -EBADFD;
+
+	ret = __io_read(req, issue_flags);
+
+	/*
+	 * If we get -EAGAIN, recycle our buffer and just let normal poll
+	 * handling arm it.
+	 */
+	if (ret == -EAGAIN) {
+		io_kbuf_recycle(req, issue_flags);
+		return -EAGAIN;
+	}
+
+	/*
+	 * Any successful return value will keep the multishot read armed.
+	 */
+	if (ret > 0) {
+		/*
+		 * Put our buffer and post a CQE. If we fail to post a CQE, then
+		 * jump to the termination path. This request is then done.
+		 */
+		cflags = io_put_kbuf(req, issue_flags);
+
+		if (io_fill_cqe_req_aux(req,
+					issue_flags & IO_URING_F_COMPLETE_DEFER,
+					ret, cflags | IORING_CQE_F_MORE)) {
+			if (issue_flags & IO_URING_F_MULTISHOT)
+				return IOU_ISSUE_SKIP_COMPLETE;
+			return -EAGAIN;
+		}
+	}
 
-	return kiocb_done(req, ret, issue_flags);
+	/*
+	 * Either an error, or we've hit overflow posting the CQE. For any
+	 * multishot request, hitting overflow will terminate it.
+	 */
+	io_req_set_res(req, ret, cflags);
+	if (issue_flags & IO_URING_F_MULTISHOT)
+		return IOU_STOP_MULTISHOT;
+	return IOU_OK;
 }
 
 int io_write(struct io_kiocb *req, unsigned int issue_flags)
diff --git a/io_uring/rw.h b/io_uring/rw.h
index 4b89f9659366..c5aed03d42a4 100644
--- a/io_uring/rw.h
+++ b/io_uring/rw.h
@@ -23,3 +23,5 @@ int io_writev_prep_async(struct io_kiocb *req);
 void io_readv_writev_cleanup(struct io_kiocb *req);
 void io_rw_fail(struct io_kiocb *req);
 void io_req_rw_complete(struct io_kiocb *req, struct io_tw_state *ts);
+int io_read_mshot_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe);
+int io_read_mshot(struct io_kiocb *req, unsigned int issue_flags);
-- 
2.40.1

Re: [PATCH 3/3] io_uring/rw: add support for IORING_OP_READ_MULTISHOT

[Jens Axboe]

On 9/12/23 11:24 AM, Jens Axboe wrote:
> @@ -863,10 +879,61 @@ int io_read(struct io_kiocb *req, unsigned int issue_flags)
>  	int ret;
>  
>  	ret = __io_read(req, issue_flags);
> -	if (unlikely(ret < 0))
> -		return ret;
> +	if (ret >= 0)
> +		return kiocb_done(req, ret, issue_flags);
> +
> +	return ret;
> +}

Looks like this one got folded into the wrong patch, this belongs with
patch 1 of course. The current git tree is correct, patch 1 does it this
way and patch 3 doesn't touch it.

-- 
Jens Axboe

Re: [PATCHSET v2 0/3] Add support for multishot reads

[Gabriel Krisman Bertazi]

Jens Axboe <axboe@kernel.dk> writes:

> Hi,
>
> We support multishot for other request types, generally in the shape of
> a flag for the request. Doing a flag based approach with reads isn't
> straightforward, as the read/write flags are in the RWF_ space. Instead,
> add a separate opcode for this, IORING_OP_READ_MULTISHOT.
>
> This can only be used provided buffers, like other multishot request
> types that read/receive data.
>
> It can also only be used for pollable file types, like a tun device or
> pipes, for example. File types that are always readable (or seekable),
> like regular files, cannot be used with multishot reads.
>
> This is based on the io_uring-futex branch (which, in turn, is based on
> the io_uring-waitid branch). No dependencies as such between them,
> except the opcode numbering.
>
> Can also be found here:
>
> https://git.kernel.dk/cgit/linux/log/?h=io_uring-mshot-read
>
> and there's a liburing branch with some basic support and some test
> cases here:
>
> https://git.kernel.dk/cgit/liburing/log/?h=read-mshot

Hey Jens,

For the entire series, feel free to take:

Reviewed-by: Gabriel Krisman Bertazi <krisman@suse.de>


-- 
Gabriel Krisman Bertazi

Re: [PATCHSET v2 0/3] Add support for multishot reads

[Jens Axboe]

On 9/12/23 12:39 PM, Gabriel Krisman Bertazi wrote:
> Jens Axboe <axboe@kernel.dk> writes:
> 
>> Hi,
>>
>> We support multishot for other request types, generally in the shape of
>> a flag for the request. Doing a flag based approach with reads isn't
>> straightforward, as the read/write flags are in the RWF_ space. Instead,
>> add a separate opcode for this, IORING_OP_READ_MULTISHOT.
>>
>> This can only be used provided buffers, like other multishot request
>> types that read/receive data.
>>
>> It can also only be used for pollable file types, like a tun device or
>> pipes, for example. File types that are always readable (or seekable),
>> like regular files, cannot be used with multishot reads.
>>
>> This is based on the io_uring-futex branch (which, in turn, is based on
>> the io_uring-waitid branch). No dependencies as such between them,
>> except the opcode numbering.
>>
>> Can also be found here:
>>
>> https://git.kernel.dk/cgit/linux/log/?h=io_uring-mshot-read
>>
>> and there's a liburing branch with some basic support and some test
>> cases here:
>>
>> https://git.kernel.dk/cgit/liburing/log/?h=read-mshot
> 
> Hey Jens,
> 
> For the entire series, feel free to take:
> 
> Reviewed-by: Gabriel Krisman Bertazi <krisman@suse.de>

Thanks, added!

-- 
Jens Axboe