From: David Gibson <david@gibson.dropbear.id.au>
To: Stefano Brivio <sbrivio@redhat.com>
Cc: passt-dev@passt.top
Subject: Re: [PATCH 15/32] tcp: Unify part of spliced and non-spliced conn_from_sock path
Date: Thu, 17 Nov 2022 12:37:04 +1100 [thread overview]
Message-ID: <Y3WQQL3wXk4Go1F/@yekko> (raw)
In-Reply-To: <20221117005358.324ca3a0@elisabeth>
[-- Attachment #1: Type: text/plain, Size: 8427 bytes --]
On Thu, Nov 17, 2022 at 12:53:58AM +0100, Stefano Brivio wrote:
> On Wed, 16 Nov 2022 15:41:55 +1100
> David Gibson <david@gibson.dropbear.id.au> wrote:
>
> > In tcp_sock_handler() we split off to handle spliced sockets before
> > checking anything else. However the first steps of the "new connection"
> > path for each case are the same: allocate a connection entry and accept()
> > the connection.
> >
> > Remove this duplication by making tcp_conn_from_sock() handle both spliced
> > and non-spliced cases, with help from more specific tcp_tap_conn_from_sock
> > and tcp_splice_conn_from_sock functions for the later stages which differ.
> >
> > Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> > ---
> > tcp.c | 68 ++++++++++++++++++++++++++++++++++------------------
> > tcp_splice.c | 58 +++++++++++++++++++++++---------------------
> > tcp_splice.h | 4 ++++
> > 3 files changed, 80 insertions(+), 50 deletions(-)
> >
> > diff --git a/tcp.c b/tcp.c
> > index 72d3b49..e66a82a 100644
> > --- a/tcp.c
> > +++ b/tcp.c
> > @@ -2753,28 +2753,19 @@ static void tcp_connect_finish(struct ctx *c, struct tcp_tap_conn *conn)
> > }
> >
> > /**
> > - * tcp_conn_from_sock() - Handle new connection request from listening socket
> > + * tcp_tap_conn_from_sock() - Initialize state for non-spliced connection
> > * @c: Execution context
> > * @ref: epoll reference of listening socket
> > + * @conn: connection structure to initialize
> > + * @s: Accepted socket
> > + * @sa: Peer socket address (from accept())
> > * @now: Current timestamp
> > */
> > -static void tcp_conn_from_sock(struct ctx *c, union epoll_ref ref,
> > - const struct timespec *now)
> > +static void tcp_tap_conn_from_sock(struct ctx *c, union epoll_ref ref,
> > + struct tcp_tap_conn *conn, int s,
> > + struct sockaddr *sa,
> > + const struct timespec *now)
> > {
> > - struct sockaddr_storage sa;
> > - struct tcp_tap_conn *conn;
> > - socklen_t sl;
> > - int s;
> > -
> > - if (c->tcp.conn_count >= TCP_MAX_CONNS)
> > - return;
> > -
> > - sl = sizeof(sa);
> > - s = accept4(ref.r.s, (struct sockaddr *)&sa, &sl, SOCK_NONBLOCK);
> > - if (s < 0)
> > - return;
> > -
> > - conn = CONN(c->tcp.conn_count++);
> > conn->c.spliced = false;
> > conn->sock = s;
> > conn->timer = -1;
> > @@ -2784,7 +2775,7 @@ static void tcp_conn_from_sock(struct ctx *c, union epoll_ref ref,
> > if (ref.r.p.tcp.tcp.v6) {
> > struct sockaddr_in6 sa6;
> >
> > - memcpy(&sa6, &sa, sizeof(sa6));
> > + memcpy(&sa6, sa, sizeof(sa6));
> >
> > if (IN6_IS_ADDR_LOOPBACK(&sa6.sin6_addr) ||
> > IN6_ARE_ADDR_EQUAL(&sa6.sin6_addr, &c->ip6.addr_seen) ||
> > @@ -2813,7 +2804,7 @@ static void tcp_conn_from_sock(struct ctx *c, union epoll_ref ref,
> > } else {
> > struct sockaddr_in sa4;
> >
> > - memcpy(&sa4, &sa, sizeof(sa4));
> > + memcpy(&sa4, sa, sizeof(sa4));
> >
> > memset(&conn->a.a4.zero, 0, sizeof(conn->a.a4.zero));
> > memset(&conn->a.a4.one, 0xff, sizeof(conn->a.a4.one));
> > @@ -2846,6 +2837,37 @@ static void tcp_conn_from_sock(struct ctx *c, union epoll_ref ref,
> > tcp_get_sndbuf(conn);
> > }
> >
> > +/**
> > + * tcp_conn_from_sock() - Handle new connection request from listening socket
> > + * @c: Execution context
> > + * @ref: epoll reference of listening socket
> > + * @now: Current timestamp
> > + */
> > +static void tcp_conn_from_sock(struct ctx *c, union epoll_ref ref,
> > + const struct timespec *now)
> > +{
> > + struct sockaddr_storage sa;
> > + union tcp_conn *conn;
> > + socklen_t sl;
> > + int s;
> > +
> > + if (c->tcp.conn_count >= TCP_MAX_CONNS)
> > + return;
> > +
> > + sl = sizeof(sa);
> > + s = accept4(ref.r.s, (struct sockaddr *)&sa, &sl, SOCK_NONBLOCK);
>
> Combined with 16/32 I'm not sure this is simplifying much -- it looks a
> bit unnatural there to get the peer address not "directly" from
> accept4(). On the other hand you drop a few lines -- I'm fine with
> it either way.
Um.. I'm not really sure what you're getting at here.
> > + if (s < 0)
> > + return;
> > +
> > + conn = tc + c->tcp.conn_count++;
> > +
> > + if (ref.r.p.tcp.tcp.splice)
> > + tcp_splice_conn_from_sock(c, ref, &conn->splice, s);
> > + else
> > + tcp_tap_conn_from_sock(c, ref, &conn->tap, s,
> > + (struct sockaddr *)&sa, now);
> > +}
> > +
> > /**
> > * tcp_timer_handler() - timerfd events: close, send ACK, retransmit, or reset
> > * @c: Execution context
> > @@ -2925,13 +2947,13 @@ void tcp_sock_handler(struct ctx *c, union epoll_ref ref, uint32_t events,
> > return;
> > }
> >
> > - if (ref.r.p.tcp.tcp.splice) {
> > - tcp_sock_handler_splice(c, ref, events);
> > + if (ref.r.p.tcp.tcp.listen) {
> > + tcp_conn_from_sock(c, ref, now);
> > return;
> > }
> >
> > - if (ref.r.p.tcp.tcp.listen) {
> > - tcp_conn_from_sock(c, ref, now);
> > + if (ref.r.p.tcp.tcp.splice) {
> > + tcp_sock_handler_splice(c, ref, events);
> > return;
> > }
> >
> > diff --git a/tcp_splice.c b/tcp_splice.c
> > index 7a06252..7007501 100644
> > --- a/tcp_splice.c
> > +++ b/tcp_splice.c
> > @@ -501,6 +501,36 @@ static void tcp_splice_dir(struct tcp_splice_conn *conn, int ref_sock,
> > *pipes = *from == conn->a ? conn->pipe_a_b : conn->pipe_b_a;
> > }
> >
> > +/**
> > + * tcp_splice_conn_from_sock() - Initialize state for spliced connection
> > + * @c: Execution context
> > + * @ref: epoll reference of listening socket
> > + * @conn: connection structure to initialize
> > + * @s: Accepted socket
> > + *
> > + * #syscalls:pasta setsockopt
> > + */
> > +void tcp_splice_conn_from_sock(struct ctx *c, union epoll_ref ref,
> > + struct tcp_splice_conn *conn, int s)
> > +{
> > + assert(c->mode == MODE_PASTA);
> > +
> > + if (setsockopt(s, SOL_TCP, TCP_QUICKACK, &((int){ 1 }),
> > + sizeof(int))) {
> > + trace("TCP (spliced): failed to set TCP_QUICKACK on %i",
> > + s);
>
> This could be indented sanely, now.
Fixed.
> > + }
> > +
> > + conn->c.spliced = true;
> > + c->tcp.splice_conn_count++;
> > + conn->a = s;
> > + conn->flags = ref.r.p.tcp.tcp.v6 ? SPLICE_V6 : 0;
> > +
> > + if (tcp_splice_new(c, conn, ref.r.p.tcp.tcp.index,
> > + ref.r.p.tcp.tcp.outbound))
> > + conn_flag(c, conn, CLOSING);
> > +}
> > +
> > /**
> > * tcp_sock_handler_splice() - Handler for socket mapped to spliced connection
> > * @c: Execution context
> > @@ -517,33 +547,7 @@ void tcp_sock_handler_splice(struct ctx *c, union epoll_ref ref,
> > uint32_t *seq_read, *seq_write;
> > struct tcp_splice_conn *conn;
> >
> > - if (ref.r.p.tcp.tcp.listen) {
> > - int s;
> > -
> > - if (c->tcp.conn_count >= TCP_MAX_CONNS)
> > - return;
> > -
> > - if ((s = accept4(ref.r.s, NULL, NULL, SOCK_NONBLOCK)) < 0)
> > - return;
> > -
> > - if (setsockopt(s, SOL_TCP, TCP_QUICKACK, &((int){ 1 }),
> > - sizeof(int))) {
> > - trace("TCP (spliced): failed to set TCP_QUICKACK on %i",
> > - s);
> > - }
> > -
> > - conn = CONN(c->tcp.conn_count++);
> > - conn->c.spliced = true;
> > - c->tcp.splice_conn_count++;
> > - conn->a = s;
> > - conn->flags = ref.r.p.tcp.tcp.v6 ? SPLICE_V6 : 0;
> > -
> > - if (tcp_splice_new(c, conn, ref.r.p.tcp.tcp.index,
> > - ref.r.p.tcp.tcp.outbound))
> > - conn_flag(c, conn, CLOSING);
> > -
> > - return;
> > - }
> > + assert(!ref.r.p.tcp.tcp.listen);
> >
> > conn = CONN(ref.r.p.tcp.tcp.index);
> >
> > diff --git a/tcp_splice.h b/tcp_splice.h
> > index 22024d6..f9462ae 100644
> > --- a/tcp_splice.h
> > +++ b/tcp_splice.h
> > @@ -6,8 +6,12 @@
> > #ifndef TCP_SPLICE_H
> > #define TCP_SPLICE_H
> >
> > +struct tcp_splice_conn;
> > +
> > void tcp_sock_handler_splice(struct ctx *c, union epoll_ref ref,
> > uint32_t events);
> > +void tcp_splice_conn_from_sock(struct ctx *c, union epoll_ref ref,
> > + struct tcp_splice_conn *conn, int s);
> > void tcp_splice_init(struct ctx *c);
> >
> > #endif /* TCP_SPLICE_H */
>
--
David Gibson | I'll have my music baroque, and my code
david AT gibson.dropbear.id.au | minimalist, thank you. NOT _the_ _other_
| _way_ _around_!
http://www.ozlabs.org/~dgibson
[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 833 bytes --]
next prev parent reply other threads:[~2022-11-17 2:08 UTC|newest]
Thread overview: 57+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-16 4:41 [PATCH 00/32] Use dual stack sockets to listen for inbound TCP connections David Gibson
2022-11-16 4:41 ` [PATCH 01/32] clang-tidy: Suppress warning about assignments in if statements David Gibson
2022-11-16 23:10 ` Stefano Brivio
2022-11-17 1:20 ` David Gibson
2022-11-16 4:41 ` [PATCH 02/32] style: Minor corrections to function comments David Gibson
2022-11-16 23:11 ` Stefano Brivio
2022-11-17 1:21 ` David Gibson
2022-11-16 4:41 ` [PATCH 03/32] tcp_splice: #include tcp_splice.h in tcp_splice.c David Gibson
2022-11-16 4:41 ` [PATCH 04/32] tcp: Remove unused TCP_MAX_SOCKS constant David Gibson
2022-11-16 4:41 ` [PATCH 05/32] tcp: Better helpers for converting between connection pointer and index David Gibson
2022-11-16 23:11 ` Stefano Brivio
2022-11-17 1:24 ` David Gibson
2022-11-16 4:41 ` [PATCH 06/32] tcp_splice: Helpers for converting from index to/from tcp_splice_conn David Gibson
2022-11-16 4:41 ` [PATCH 07/32] tcp: Move connection state structures into a shared header David Gibson
2022-11-16 4:41 ` [PATCH 08/32] tcp: Add connection union type David Gibson
2022-11-16 4:41 ` [PATCH 09/32] tcp: Improved helpers to update connections after moving David Gibson
2022-11-16 4:41 ` [PATCH 10/32] tcp: Unify spliced and non-spliced connection tables David Gibson
2022-11-16 4:41 ` [PATCH 11/32] tcp: Unify tcp_defer_handler and tcp_splice_defer_handler() David Gibson
2022-11-16 4:41 ` [PATCH 12/32] tcp: Partially unify tcp_timer() and tcp_splice_timer() David Gibson
2022-11-16 4:41 ` [PATCH 13/32] tcp: Unify the IN_EPOLL flag David Gibson
2022-11-16 4:41 ` [PATCH 14/32] tcp: Separate helpers to create ns listening sockets David Gibson
2022-11-16 23:51 ` Stefano Brivio
2022-11-17 1:32 ` David Gibson
2022-11-16 4:41 ` [PATCH 15/32] tcp: Unify part of spliced and non-spliced conn_from_sock path David Gibson
2022-11-16 23:53 ` Stefano Brivio
2022-11-17 1:37 ` David Gibson [this message]
2022-11-17 7:30 ` Stefano Brivio
2022-11-17 8:58 ` David Gibson
2022-11-16 4:41 ` [PATCH 16/32] tcp: Use the same sockets to listen for spliced and non-spliced connections David Gibson
2022-11-16 23:54 ` Stefano Brivio
2022-11-17 1:43 ` David Gibson
2022-11-16 4:41 ` [PATCH 17/32] tcp: Remove splice from tcp_epoll_ref David Gibson
2022-11-16 4:41 ` [PATCH 18/32] tcp: Don't store hash bucket in connection structures David Gibson
2022-11-16 4:41 ` [PATCH 19/32] inany: Helper functions for handling addresses which could be IPv4 or IPv6 David Gibson
2022-11-16 23:54 ` Stefano Brivio
2022-11-17 1:48 ` David Gibson
2022-11-16 4:42 ` [PATCH 20/32] tcp: Hash IPv4 and IPv4-mapped-IPv6 addresses the same David Gibson
2022-11-16 4:42 ` [PATCH 21/32] tcp: Take tcp_hash_insert() address from struct tcp_conn David Gibson
2022-11-16 4:42 ` [PATCH 22/32] tcp: Simplify tcp_hash_match() to take an inany_addr David Gibson
2022-11-16 4:42 ` [PATCH 23/32] tcp: Unify initial sequence number calculation for IPv4 and IPv6 David Gibson
2022-11-16 4:42 ` [PATCH 24/32] tcp: Have tcp_seq_init() take its parameters from struct tcp_conn David Gibson
2022-11-16 4:42 ` [PATCH 25/32] tcp: Fix small errors in tcp_seq_init() time handling David Gibson
2022-11-16 4:42 ` [PATCH 26/32] tcp: Remove v6 flag from tcp_epoll_ref David Gibson
2022-11-17 0:15 ` Stefano Brivio
2022-11-17 1:50 ` David Gibson
2022-11-16 4:42 ` [PATCH 27/32] tcp: NAT IPv4-mapped IPv6 addresses like IPv4 addresses David Gibson
2022-11-17 0:15 ` Stefano Brivio
2022-11-17 2:00 ` David Gibson
2022-11-16 4:42 ` [PATCH 28/32] tcp_splice: Allow splicing of connections from IPv4-mapped loopback David Gibson
2022-11-17 0:15 ` Stefano Brivio
2022-11-17 2:05 ` David Gibson
2022-11-16 4:42 ` [PATCH 29/32] tcp: Consolidate tcp_sock_init[46] David Gibson
2022-11-16 4:42 ` [PATCH 30/32] util: Allow sock_l4() to open dual stack sockets David Gibson
2022-11-16 4:42 ` [PATCH 31/32] util: Always return -1 on error in sock_l4() David Gibson
2022-11-16 4:42 ` [PATCH 32/32] tcp: Use dual stack sockets for port forwarding when possible David Gibson
2022-11-17 0:15 ` Stefano Brivio
2022-11-17 2:08 ` David Gibson
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 \
--in-reply-to=Y3WQQL3wXk4Go1F/@yekko \
--to=david@gibson.dropbear.id.au \
--cc=passt-dev@passt.top \
--cc=sbrivio@redhat.com \
/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.
Code repositories for project(s) associated with this public inbox
https://passt.top/passt
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for IMAP folder(s).