public inbox for [email protected]
 help / color / mirror / Atom feed
From: David Wei <[email protected]>
To: lizetao <[email protected]>
Cc: Jens Axboe <[email protected]>,
	Pavel Begunkov <[email protected]>,
	"[email protected]" <[email protected]>
Subject: Re: [PATCH v2 1/2] io_uring/zcrx: add single shot recvzc
Date: Sun, 23 Feb 2025 16:17:12 -0800	[thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>

On 2025-02-22 00:54, lizetao wrote:
> Hi,
> 
>> -----Original Message-----
>> From: David Wei <[email protected]>
>> Sent: Saturday, February 22, 2025 4:52 AM
>> To: [email protected]
>> Cc: Jens Axboe <[email protected]>; Pavel Begunkov <[email protected]>
>> Subject: [PATCH v2 1/2] io_uring/zcrx: add single shot recvzc
>>
>> Currently only multishot recvzc requests are supported, but sometimes there is
>> a need to do a single recv e.g. peeking at some data in the socket. Add single
>> shot recvzc requests where IORING_RECV_MULTISHOT is _not_ set and the sqe-
>>> len field is set to the number of bytes to read N.
>>
>> There could be multiple completions containing data, like the multishot case,
>> since N bytes could be split across multiple frags. This is followed by a final
>> completion with res and cflags both set to 0 that indicate the completion of the
>> request, or a -res that indicate an error.
>>
>> Signed-off-by: David Wei <[email protected]>
>> ---
>>  io_uring/net.c  | 19 +++++++++++++++++--  io_uring/zcrx.c | 17 ++++++++++++--
>> ---  io_uring/zcrx.h |  2 +-
>>  3 files changed, 30 insertions(+), 8 deletions(-)
>>
>> diff --git a/io_uring/net.c b/io_uring/net.c index 000dc70d08d0..cae34a24266c
>> 100644
>> --- a/io_uring/net.c
>> +++ b/io_uring/net.c
>> @@ -94,6 +94,7 @@ struct io_recvzc {
>>  	struct file			*file;
>>  	unsigned			msg_flags;
>>  	u16				flags;
>> +	u32				len;
>>  	struct io_zcrx_ifq		*ifq;
>>  };
>>
>> @@ -1241,7 +1242,7 @@ int io_recvzc_prep(struct io_kiocb *req, const struct
>> io_uring_sqe *sqe)
>>  	unsigned ifq_idx;
>>
>>  	if (unlikely(sqe->file_index || sqe->addr2 || sqe->addr ||
>> -		     sqe->len || sqe->addr3))
>> +		     sqe->addr3))
>>  		return -EINVAL;
>>
>>  	ifq_idx = READ_ONCE(sqe->zcrx_ifq_idx); @@ -1250,6 +1251,12 @@
>> int io_recvzc_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
>>  	zc->ifq = req->ctx->ifq;
>>  	if (!zc->ifq)
>>  		return -EINVAL;
>> +	zc->len = READ_ONCE(sqe->len);
>> +	if (zc->len == UINT_MAX)
>> +		return -EINVAL;
>> +	/* UINT_MAX means no limit on readlen */
> 
> What is the difference between unlimited read length and IO_URING_F_MULTISHOT?
> In what specific scenarios would unlimited read length be used?

I will clean up the terminology in the next revision. There isn't a
singleshot vs multishot as such. Rather, it's an unlimited multishot vs
a limited multishot. The unlimited multishot request would be a typical
multishot recv that will keep reading data from the socket. The limited
multishot request will read up to a limit N before completing the
request.

> 
>> +	if (!zc->len)
>> +		zc->len = UINT_MAX;
>>
>>  	zc->flags = READ_ONCE(sqe->ioprio);
>>  	zc->msg_flags = READ_ONCE(sqe->msg_flags); @@ -1269,6 +1276,7
>> @@ int io_recvzc_prep(struct io_kiocb *req, const struct io_uring_sqe *sqe)
>> int io_recvzc(struct io_kiocb *req, unsigned int issue_flags)  {
>>  	struct io_recvzc *zc = io_kiocb_to_cmd(req, struct io_recvzc);
>> +	bool limit = zc->len != UINT_MAX;
> 
> It is not used.
> 
>>  	struct socket *sock;
>>  	int ret;
>>
>> @@ -1281,7 +1289,7 @@ int io_recvzc(struct io_kiocb *req, unsigned int
>> issue_flags)
>>  		return -ENOTSOCK;
>>
>>  	ret = io_zcrx_recv(req, zc->ifq, sock, zc->msg_flags | MSG_DONTWAIT,
>> -			   issue_flags);
>> +			   issue_flags, &zc->len);
>>  	if (unlikely(ret <= 0) && ret != -EAGAIN) {
>>  		if (ret == -ERESTARTSYS)
>>  			ret = -EINTR;
>> @@ -1296,6 +1304,13 @@ int io_recvzc(struct io_kiocb *req, unsigned int
>> issue_flags)
>>  		return IOU_OK;
>>  	}
>>
>> +	if (zc->len == 0) {
>> +		io_req_set_res(req, 0, 0);
>> +
>> +		if (issue_flags & IO_URING_F_MULTISHOT)
>> +			return IOU_STOP_MULTISHOT;
>> +		return IOU_OK;
>> +	}
>>  	if (issue_flags & IO_URING_F_MULTISHOT)
>>  		return IOU_ISSUE_SKIP_COMPLETE;
>>  	return -EAGAIN;
>> diff --git a/io_uring/zcrx.c b/io_uring/zcrx.c index
>> f2d326e18e67..74bca4e471bc 100644
>> --- a/io_uring/zcrx.c
>> +++ b/io_uring/zcrx.c
>> @@ -817,6 +817,7 @@ io_zcrx_recv_skb(read_descriptor_t *desc, struct
>> sk_buff *skb,
>>  	int i, copy, end, off;
>>  	int ret = 0;
>>
>> +	len = min_t(size_t, len, desc->count);
>>  	if (unlikely(args->nr_skbs++ > IO_SKBS_PER_CALL_LIMIT))
>>  		return -EAGAIN;
>>
>> @@ -894,26 +895,32 @@ io_zcrx_recv_skb(read_descriptor_t *desc, struct
>> sk_buff *skb,
>>  out:
>>  	if (offset == start_off)
>>  		return ret;
>> +	if (desc->count != UINT_MAX)
>> +		desc->count -= (offset - start_off);
>>  	return offset - start_off;
>>  }
>>
>>  static int io_zcrx_tcp_recvmsg(struct io_kiocb *req, struct io_zcrx_ifq *ifq,
>>  				struct sock *sk, int flags,
>> -				unsigned issue_flags)
>> +				unsigned issue_flags, unsigned int *outlen)
>>  {
>> +	unsigned int len = *outlen;
>> +	bool limit = len != UINT_MAX;
>>  	struct io_zcrx_args args = {
>>  		.req = req,
>>  		.ifq = ifq,
>>  		.sock = sk->sk_socket,
>>  	};
>>  	read_descriptor_t rd_desc = {
>> -		.count = 1,
>> +		.count = len,
>>  		.arg.data = &args,
>>  	};
>>  	int ret;
>>
>>  	lock_sock(sk);
>>  	ret = tcp_read_sock(sk, &rd_desc, io_zcrx_recv_skb);
>> +	if (limit && ret)
>> +		*outlen = len - ret;
>>  	if (ret <= 0) {
>>  		if (ret < 0 || sock_flag(sk, SOCK_DONE))
>>  			goto out;
>> @@ -930,7 +937,7 @@ static int io_zcrx_tcp_recvmsg(struct io_kiocb *req,
>> struct io_zcrx_ifq *ifq,
>>  		ret = IOU_REQUEUE;
>>  	} else if (sock_flag(sk, SOCK_DONE)) {
>>  		/* Make it to retry until it finally gets 0. */
>> -		if (issue_flags & IO_URING_F_MULTISHOT)
>> +		if (!limit && (issue_flags & IO_URING_F_MULTISHOT))
>>  			ret = IOU_REQUEUE;
>>  		else
>>  			ret = -EAGAIN;
>> @@ -942,7 +949,7 @@ static int io_zcrx_tcp_recvmsg(struct io_kiocb *req,
>> struct io_zcrx_ifq *ifq,
>>
>>  int io_zcrx_recv(struct io_kiocb *req, struct io_zcrx_ifq *ifq,
>>  		 struct socket *sock, unsigned int flags,
>> -		 unsigned issue_flags)
>> +		 unsigned issue_flags, unsigned int *len)
>>  {
>>  	struct sock *sk = sock->sk;
>>  	const struct proto *prot = READ_ONCE(sk->sk_prot); @@ -951,5
>> +958,5 @@ int io_zcrx_recv(struct io_kiocb *req, struct io_zcrx_ifq *ifq,
>>  		return -EPROTONOSUPPORT;
>>
>>  	sock_rps_record_flow(sk);
>> -	return io_zcrx_tcp_recvmsg(req, ifq, sk, flags, issue_flags);
>> +	return io_zcrx_tcp_recvmsg(req, ifq, sk, flags, issue_flags, len);
>>  }
>> diff --git a/io_uring/zcrx.h b/io_uring/zcrx.h index
>> a16bdd921f03..1b4042dc48e2 100644
>> --- a/io_uring/zcrx.h
>> +++ b/io_uring/zcrx.h
>> @@ -46,7 +46,7 @@ void io_unregister_zcrx_ifqs(struct io_ring_ctx *ctx);  void
>> io_shutdown_zcrx_ifqs(struct io_ring_ctx *ctx);  int io_zcrx_recv(struct io_kiocb
>> *req, struct io_zcrx_ifq *ifq,
>>  		 struct socket *sock, unsigned int flags,
>> -		 unsigned issue_flags);
>> +		 unsigned issue_flags, unsigned int *len);
>>  #else
>>  static inline int io_register_zcrx_ifq(struct io_ring_ctx *ctx,
>>  					struct io_uring_zcrx_ifq_reg __user
>> *arg)
>> --
>> 2.43.5
>>
>>
> 
> ---
> Li Zetao

  reply	other threads:[~2025-02-24  0:17 UTC|newest]

Thread overview: 20+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2025-02-21 20:51 [PATCH v2 0/2] io_uring zc rx fixed len recvzc David Wei
2025-02-21 20:51 ` [PATCH v2 1/2] io_uring/zcrx: add single shot recvzc David Wei
2025-02-22  0:08   ` Jens Axboe
2025-02-22  1:01     ` Pavel Begunkov
2025-02-22  1:07       ` Jens Axboe
2025-02-23 22:35         ` David Wei
2025-02-24 12:49           ` Pavel Begunkov
2025-02-23 22:39     ` David Wei
2025-02-22  0:40   ` Pavel Begunkov
2025-02-22  0:52     ` Jens Axboe
2025-02-22  1:06       ` Pavel Begunkov
2025-02-22  1:09         ` Jens Axboe
2025-02-22  1:15           ` Pavel Begunkov
2025-02-22  1:09         ` Pavel Begunkov
2025-02-23 22:43     ` David Wei
2025-02-22  0:56   ` Pavel Begunkov
2025-02-23 22:44     ` David Wei
2025-02-22  8:54   ` lizetao
2025-02-24  0:17     ` David Wei [this message]
2025-02-21 20:51 ` [PATCH v2 2/2] io_uring/zcrx: add selftest case for " David Wei

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    [email protected] \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox