From: Stefano Brivio <sbrivio@redhat.com>
To: David Gibson <david@gibson.dropbear.id.au>
Cc: passt-dev@passt.top
Subject: Re: [PATCH v2 09/22] tcp_splice: Simplify clean up logic
Date: Sun, 18 Feb 2024 21:59:37 +0100 [thread overview]
Message-ID: <20240218215937.126b2839@elisabeth> (raw)
In-Reply-To: <20240206011734.884138-10-david@gibson.dropbear.id.au>
On Tue, 6 Feb 2024 12:17:21 +1100
David Gibson <david@gibson.dropbear.id.au> wrote:
> Currently tcp_splice_flow_defer() contains specific logic to determine if
> we're far enough initialised that we need to close pipes and/or sockets.
> This is potentially fragile if we change something about the order in which
> we do things. We can simplify this by initialising the pipe and socket
> fields to -1 very early, then close()ing them if and only if they're non
> negative.
>
> This lets us remove a special case cleanup if our connect() fails. This
> will already trigger a CLOSING event, and the socket fd in question is
> populated in the connection structure. Thus we can let the new cleanup
> logic handle it rather than requiring an explicit close().
>
> Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> ---
> tcp_splice.c | 27 +++++++++++----------------
> 1 file changed, 11 insertions(+), 16 deletions(-)
>
> diff --git a/tcp_splice.c b/tcp_splice.c
> index 40ecb5d4..f0343eb5 100644
> --- a/tcp_splice.c
> +++ b/tcp_splice.c
> @@ -246,16 +246,14 @@ bool tcp_splice_flow_defer(union flow *flow)
> return false;
>
> for (side = 0; side < SIDES; side++) {
> - if (conn->events & SPLICE_ESTABLISHED) {
> - /* Flushing might need to block: don't recycle them. */
> - if (conn->pipe[side][0] != -1) {
> - close(conn->pipe[side][0]);
> - close(conn->pipe[side][1]);
> - conn->pipe[side][0] = conn->pipe[side][1] = -1;
> - }
> + /* Flushing might need to block: don't recycle them. */
> + if (conn->pipe[side][0] >= 0) {
> + close(conn->pipe[side][0]);
> + close(conn->pipe[side][1]);
> + conn->pipe[side][0] = conn->pipe[side][1] = -1;
> }
>
> - if (side == 0 || conn->events & SPLICE_CONNECT) {
> + if (conn->s[side] >= 0) {
> close(conn->s[side]);
> conn->s[side] = -1;
> }
> @@ -284,8 +282,6 @@ static int tcp_splice_connect_finish(const struct ctx *c,
> int i = 0;
>
> for (side = 0; side < SIDES; side++) {
> - conn->pipe[side][0] = conn->pipe[side][1] = -1;
> -
> for (; i < TCP_SPLICE_PIPE_POOL_SIZE; i++) {
> if (splice_pipe_pool[i][0] >= 0) {
> SWAP(conn->pipe[side][0],
> @@ -361,12 +357,8 @@ static int tcp_splice_connect(const struct ctx *c, struct tcp_splice_conn *conn,
> }
>
> if (connect(conn->s[1], sa, sl)) {
> - if (errno != EINPROGRESS) {
> - int ret = -errno;
> -
> - close(sock_conn);
> - return ret;
> - }
> + if (errno != EINPROGRESS)
> + return -errno;
Nit: a newline here would be nice.
> conn_event(c, conn, SPLICE_CONNECT);
> } else {
> conn_event(c, conn, SPLICE_ESTABLISHED);
> @@ -459,6 +451,9 @@ bool tcp_splice_conn_from_sock(const struct ctx *c,
> conn->f.type = FLOW_TCP_SPLICE;
> conn->flags = inany_v4(&aany) ? 0 : SPLICE_V6;
> conn->s[0] = s;
> + conn->s[1] = -1;
> + conn->pipe[0][0] = conn->pipe[0][1] = -1;
> + conn->pipe[1][0] = conn->pipe[1][1] = -1;
>
> if (tcp_splice_new(c, conn, ref.port, ref.pif))
> conn_flag(c, conn, CLOSING);
next prev parent reply other threads:[~2024-02-18 21:00 UTC|newest]
Thread overview: 38+ messages / expand[flat|nested] mbox.gz Atom feed top
2024-02-06 1:17 [PATCH v2 00/22] More flow table preliminaries: address handling improvements David Gibson
2024-02-06 1:17 ` [PATCH v2 01/22] treewide: Use sa_family_t for address family variables David Gibson
2024-02-06 1:17 ` [PATCH v2 02/22] inany: Helper to test for various address types David Gibson
2024-02-18 20:58 ` Stefano Brivio
2024-02-19 1:48 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 03/22] inany: Add inany_ntop() helper David Gibson
2024-02-06 1:17 ` [PATCH v2 04/22] inany: Provide more conveniently typed constants for special addresses David Gibson
2024-02-06 1:17 ` [PATCH v2 05/22] inany: Introduce union sockaddr_inany David Gibson
2024-02-06 1:17 ` [PATCH v2 06/22] util: Allow IN4_IS_* macros to operate on untyped addresses David Gibson
2024-02-06 1:17 ` [PATCH v2 07/22] tcp, udp: Don't precompute port remappings in epoll references David Gibson
2024-02-06 1:17 ` [PATCH v2 08/22] flow: Add helper to determine a flow's protocol David Gibson
2024-02-06 1:17 ` [PATCH v2 09/22] tcp_splice: Simplify clean up logic David Gibson
2024-02-18 20:59 ` Stefano Brivio [this message]
2024-02-19 1:50 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 10/22] tcp_splice: Don't use flow_trace() before setting flow type David Gibson
2024-02-06 1:17 ` [PATCH v2 11/22] flow: Clarify flow entry life cycle, introduce uniform logging David Gibson
2024-02-18 21:00 ` Stefano Brivio
2024-02-19 1:58 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 12/22] tcp, tcp_splice: Helpers for getting sockets from the pools David Gibson
2024-02-18 21:00 ` Stefano Brivio
2024-02-19 1:51 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 13/22] tcp_splice: More specific variable names in new splice path David Gibson
2024-02-18 21:00 ` Stefano Brivio
2024-02-19 1:53 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 14/22] tcp_splice: Merge tcp_splice_new() into its caller David Gibson
2024-02-06 1:17 ` [PATCH v2 15/22] tcp_splice: Make tcp_splice_connect() create its own sockets David Gibson
2024-02-06 1:17 ` [PATCH v2 16/22] tcp_splice: Improve error reporting on connect path David Gibson
2024-02-18 21:01 ` Stefano Brivio
2024-02-19 3:23 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 17/22] tcp_splice: Improve logic deciding when to splice David Gibson
2024-02-06 1:17 ` [PATCH v2 18/22] tcp, tcp_splice: Parse listening socket epoll ref in tcp_listen_handler() David Gibson
2024-02-06 1:17 ` [PATCH v2 19/22] tcp: Validate TCP endpoint addresses David Gibson
2024-02-22 12:45 ` Stefano Brivio
2024-02-23 3:56 ` David Gibson
2024-02-06 1:17 ` [PATCH v2 20/22] tap: Disallow loopback addresses on tap interface David Gibson
2024-02-06 1:17 ` [PATCH v2 21/22] port_fwd: Fix copypasta error in port_fwd_scan_udp() comments David Gibson
2024-02-06 1:17 ` [PATCH v2 22/22] fwd: Rename port_fwd.[ch] and their contents David Gibson
2024-02-27 14:22 ` [PATCH v2 00/22] More flow table preliminaries: address handling improvements Stefano Brivio
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=20240218215937.126b2839@elisabeth \
--to=sbrivio@redhat.com \
--cc=david@gibson.dropbear.id.au \
--cc=passt-dev@passt.top \
/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).