From: Kuniyuki Iwashima <kuniyu@amazon.com>
To: <gnault@redhat.com>
Cc: <dsahern@gmail.com>, <edumazet@google.com>, <kuniyu@amazon.com>,
<mkubecek@suse.cz>, <netdev@vger.kernel.org>
Subject: [PATCH iproute2-next] ss: Add support for dumping TCP bound-inactive sockets.
Date: Mon, 11 Dec 2023 23:19:17 +0900 [thread overview]
Message-ID: <20231211141917.42613-1-kuniyu@amazon.com> (raw)
In-Reply-To: <87947b2975508804d4efc49b9380041288eaa0f6.1702301488.git.gnault@redhat.com>
From: Guillaume Nault <gnault@redhat.com>
Date: Mon, 11 Dec 2023 14:35:15 +0100
> Make ss aware of the new "bound-inactive" pseudo-state for TCP (see
> Linux commit 91051f003948 ("tcp: Dump bound-only sockets in inet_diag.")).
> These are TCP sockets that have been bound, but are neither listening nor
> connecting.
>
> With this patch, these sockets can now be dumped with:
>
> * the existing -a (--all) option, to dump all sockets, including
> bound-inactive ones,
>
> * the new -B (--bound-inactive) option, to dump them exclusively,
>
> * the new "bound-inactive" state, to be used in a STATE-FILTER.
>
> The SS_NEW_SYN_RECV pseudo-state is added in this patch only for code
> consistency, so that SS_BOUND_INACTIVE gets assigned the right value
> without manual assignment.
>
> Note that the SS_BOUND_INACTIVE state is a pseudo-state used for queries
> only. The kernel returns them as SS_CLOSE.
>
> Signed-off-by: Guillaume Nault <gnault@redhat.com>
> ---
> man/man8/ss.8 | 7 +++++++
> misc/ss.c | 13 ++++++++++++-
> 2 files changed, 19 insertions(+), 1 deletion(-)
>
> diff --git a/man/man8/ss.8 b/man/man8/ss.8
> index 073e9f03..4ece41fa 100644
> --- a/man/man8/ss.8
> +++ b/man/man8/ss.8
> @@ -40,6 +40,10 @@ established connections) sockets.
> .B \-l, \-\-listening
> Display only listening sockets (these are omitted by default).
> .TP
> +.B \-B, \-\-bound-inactive
> +Display only TCP bound but inactive (not listening, connecting, etc.) sockets
> +(these are omitted by default).
> +.TP
> .B \-o, \-\-options
> Show timer information. For TCP protocol, the output format is:
> .RS
> @@ -456,6 +460,9 @@ states except for
> - opposite to
> .B bucket
>
> +.B bound-inactive
> +- bound but otherwise inactive sockets (not listening, connecting, etc.)
> +
> .SH EXPRESSION
>
> .B EXPRESSION
> diff --git a/misc/ss.c b/misc/ss.c
> index 16ffb6c8..19adc1b7 100644
> --- a/misc/ss.c
> +++ b/misc/ss.c
> @@ -210,6 +210,8 @@ enum {
> SS_LAST_ACK,
> SS_LISTEN,
> SS_CLOSING,
> + SS_NEW_SYN_RECV,
I think this is bit confusing as TCP_NEW_SYN_RECV is usually
invisible from user.
TCP_NEW_SYN_RECV was originally split from TCP_SYN_RECV for a
non-{TFO,cross-SYN} request.
So, both get_openreq4() (/proc/net/tcp) and inet_req_diag_fill()
(inet_diag) maps TCP_NEW_SYN_RECV to TCP_SYN_RECV.
> + SS_BOUND_INACTIVE,
I prefer explicitly assigning a number to SS_BOUND_INACTIVE.
> SS_MAX
> };
>
> @@ -1382,6 +1384,8 @@ static void sock_state_print(struct sockstat *s)
> [SS_LAST_ACK] = "LAST-ACK",
> [SS_LISTEN] = "LISTEN",
> [SS_CLOSING] = "CLOSING",
> + [SS_NEW_SYN_RECV] = "NEW-SYN-RECV",
If we need to define SS_NEW_SYN_RECV, I prefer not setting
it or setting "" or "SYN-RECV".
> + [SS_BOUND_INACTIVE] = "BOUND-INACTIVE",
> };
>
> switch (s->local.family) {
> @@ -5339,6 +5343,7 @@ static void _usage(FILE *dest)
> " -r, --resolve resolve host names\n"
> " -a, --all display all sockets\n"
> " -l, --listening display listening sockets\n"
> +" -B, --bound-inactive display TCP bound but inactive sockets\n"
> " -o, --options show timer information\n"
> " -e, --extended show detailed socket information\n"
> " -m, --memory show socket memory usage\n"
> @@ -5421,6 +5426,8 @@ static int scan_state(const char *state)
> [SS_LAST_ACK] = "last-ack",
> [SS_LISTEN] = "listening",
> [SS_CLOSING] = "closing",
> + [SS_NEW_SYN_RECV] = "new-syn-recv",
Same here.
Thanks!
> + [SS_BOUND_INACTIVE] = "bound-inactive",
> };
> int i;
>
> @@ -5487,6 +5494,7 @@ static const struct option long_opts[] = {
> { "vsock", 0, 0, OPT_VSOCK },
> { "all", 0, 0, 'a' },
> { "listening", 0, 0, 'l' },
> + { "bound-inactive", 0, 0, 'B' },
> { "ipv4", 0, 0, '4' },
> { "ipv6", 0, 0, '6' },
> { "packet", 0, 0, '0' },
> @@ -5525,7 +5533,7 @@ int main(int argc, char *argv[])
> int state_filter = 0;
>
> while ((ch = getopt_long(argc, argv,
> - "dhaletuwxnro460spTbEf:mMiA:D:F:vVzZN:KHSO",
> + "dhalBetuwxnro460spTbEf:mMiA:D:F:vVzZN:KHSO",
> long_opts, NULL)) != EOF) {
> switch (ch) {
> case 'n':
> @@ -5590,6 +5598,9 @@ int main(int argc, char *argv[])
> case 'l':
> state_filter = (1 << SS_LISTEN) | (1 << SS_CLOSE);
> break;
> + case 'B':
> + state_filter = 1 << SS_BOUND_INACTIVE;
> + break;
> case '4':
> filter_af_set(¤t_filter, AF_INET);
> break;
> --
> 2.39.2
next prev parent reply other threads:[~2023-12-11 14:19 UTC|newest]
Thread overview: 4+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-12-11 13:35 [PATCH iproute2-next] ss: Add support for dumping TCP bound-inactive sockets Guillaume Nault
2023-12-11 14:19 ` Kuniyuki Iwashima [this message]
2023-12-11 17:58 ` Guillaume Nault
2023-12-13 17:16 ` David Ahern
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=20231211141917.42613-1-kuniyu@amazon.com \
--to=kuniyu@amazon.com \
--cc=dsahern@gmail.com \
--cc=edumazet@google.com \
--cc=gnault@redhat.com \
--cc=mkubecek@suse.cz \
--cc=netdev@vger.kernel.org \
/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;
as well as URLs for NNTP newsgroup(s).