From: lizetao <[email protected]>
To: David Wei <[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: Sat, 22 Feb 2025 08:54:59 +0000 [thread overview]
Message-ID: <[email protected]> (raw)
In-Reply-To: <[email protected]>
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?
> + 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
next prev parent reply other threads:[~2025-02-22 8:55 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 [this message]
2025-02-24 0:17 ` David Wei
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