From: Vadim Kochan <vadim4j@gmail.com>
To: Miha Marolt <miham@beyondsemi.com>
Cc: Vadim Kochan <vadim4j@gmail.com>,
Nikolay Aleksandrov <nikolay@cumulusnetworks.com>,
netdev@vger.kernel.org
Subject: Re: [BUG] "$ ss -a" incorrectly displays raw sockets as udp sockets
Date: Fri, 10 Jul 2015 10:16:19 +0300 [thread overview]
Message-ID: <20150710071619.GA13710@angus-think.lan> (raw)
In-Reply-To: <559F6FBA.9080807@beyondsemi.com>
On Fri, Jul 10, 2015 at 09:09:46AM +0200, Miha Marolt wrote:
>
>
> On 07/09/2015 05:15 PM, Vadim Kochan wrote:
> >On Thu, Jul 09, 2015 at 05:09:27PM +0200, Miha Marolt wrote:
> >>
> >>On 07/09/2015 04:57 PM, Nikolay Aleksandrov wrote:
> >>>On 07/09/2015 04:55 PM, Vadim Kochan wrote:
> >>>>On Thu, Jul 09, 2015 at 04:50:06PM +0200, Nikolay Aleksandrov wrote:
> >>>>>On 07/09/2015 04:13 PM, Miha Marolt wrote:
> >>>>>>Hi!
> >>>>>>
> >>>>>>I hope this is the right place to reports bugs. I apologize if it isn't.
> >>>>>>
> >>>>>>I have written a C program (see below for source code) that opens a raw socket on CentOS 7.1 Linux and binds it to some address (it doesn't use the port that I supplied, but that is not the point here). The "netstat" program correctly recognizes the socket as "raw", while "ss" program says it is "udp". Here are the relevant lines from the ss and netstat commands:
> >>>>>>
> >>>>>>$ netstat -an
> >>>>>>raw 0 0 127.0.0.1:6 0.0.0.0:* 7
> >>>>>>
> >>>>>>$ ./ss -an
> >>>>>>udp UNCONN 21569 0 127.0.0.1:6 *:*
> >>>>>>
> >>>>>>Here is the version information
> >>>>>>
> >>>>>>$ netstat --version # From CentOS 7.1.
> >>>>>>net-tools 2.10-alpha
> >>>>>>
> >>>>>>$ ./ss --version # Built from git.
> >>>>>>ss utility, iproute2-ss150626
> >>>>>>
> >>>>>>
> >>>>>>C source follows. If you store it in "main.c", then compile it with "$ gcc main.c -o main" and then run it by executing "$ sudo ./main".
> >>>>>>
> >>>>>>#include <arpa/inet.h>
> >>>>>>#include <assert.h>
> >>>>>>#include <sys/socket.h>
> >>>>>>#include <stdio.h>
> >>>>>>#include <unistd.h>
> >>>>>>
> >>>>>>
> >>>>>>int main(void)
> >>>>>>{
> >>>>>> // Create a raw socket.
> >>>>>> int sock = socket(AF_INET, SOCK_RAW, IPPROTO_TCP);
> >>>>>> if (sock == -1) { perror(NULL); goto exc_cleanup; }
> >>>>>>
> >>>>>> // Bind socket to an address.
> >>>>>> struct sockaddr_in addr;
> >>>>>> addr.sin_family = AF_INET;
> >>>>>> inet_pton(AF_INET, "127.0.0.1", &addr.sin_addr);
> >>>>>> addr.sin_port = htons(27183);
> >>>>>>
> >>>>>> int rc = bind(sock, (struct sockaddr*)&addr, sizeof(addr));
> >>>>>> if (rc != 0) { perror(NULL); goto exc_cleanup; }
> >>>>>>
> >>>>>> // Wait until user presses <ENTER>.
> >>>>>> printf("\nPress <ENTER> to quit the program.\n");
> >>>>>> getchar();
> >>>>>>
> >>>>>>exc_cleanup:
> >>>>>> assert(!close(sock));
> >>>>>>}
> >>>>>>
> >>>>>>
> >>>>>>Best regards,
> >>>>>>Miha
> >>>>>Hi,
> >>>>>I think this was changed by commit:
> >>>>>8250bc9ff4e5 ("ss: Unify inet sockets output")
> >>>>>Because dgram_show_line() is used for both UDP and RAW sockets
> >>>>>IPPROTO_UDP is used for both now and proto_name() returns "udp".
> >>>>>CCed the patch author and attached a possible solution.
> >>>>>
> >>>>>Cheers,
> >>>>> Nik
> >>>>>diff --git a/misc/ss.c b/misc/ss.c
> >>>>>index 870cad185341..4de77e92c319 100644
> >>>>>--- a/misc/ss.c
> >>>>>+++ b/misc/ss.c
> >>>>>@@ -1554,6 +1554,8 @@ out:
> >>>>> static char *proto_name(int protocol)
> >>>>> {
> >>>>> switch (protocol) {
> >>>>>+ case IPPROTO_RAW:
> >>>>>+ return "raw";
> >>>>> case IPPROTO_UDP:
> >>>>> return "udp";
> >>>>> case IPPROTO_TCP:
> >>>>>@@ -2398,7 +2400,7 @@ static int dgram_show_line(char *line, const struct filter *f, int family)
> >>>>> if (n < 9)
> >>>>> opt[0] = 0;
> >>>>>- inet_stats_print(&s, IPPROTO_UDP);
> >>>>>+ inet_stats_print(&s, dg_proto == UDP_PROTO ? IPPROTO_UDP : IPPROTO_RAW);
> >>>>> if (show_details && opt[0])
> >>>>> printf(" opt:\"%s\"", opt);
> >>>>Yeah, it fixed the issue, just tested the fix.
> >>>>
> >>>Great, I'll submit it in a minute.
> >>>Thanks for testing!
> >>Thank you very much for the fast fix!
> >>
> >>I've found another bug: command "$ ss" doesn't display only tcp sockets as
> >>the man page ss(8) says (file man/man8/ss.8 in iproute2 git repo). It also
> >>displays e.g. unix sockets. The man page says "When no option is used ss
> >>displays a list of open non-listening TCP sockets that have established
> >>connection.".
> >>
> >>
> >>Regards,
> >>Miha
> >That seems OK, I see that unix stream sockets are displayed which are
> >connection oriented, and UDP can also have established state.
>
> According to the man page, **only** TCP sockets should be listed by "$ ss".
> In reality, also Unix and UDP sockets are listed, so there is a discrepancy
> between what man page says and what actually happens.
Yes, you are right man page says so, but seems such behaviour is for a
long time and may be it is better to change man page rather than change
the default behaviour.
next prev parent reply other threads:[~2015-07-10 7:18 UTC|newest]
Thread overview: 15+ messages / expand[flat|nested] mbox.gz Atom feed top
2015-07-09 14:13 [BUG] "$ ss -a" incorrectly displays raw sockets as udp sockets Miha Marolt
2015-07-09 14:50 ` Nikolay Aleksandrov
2015-07-09 14:55 ` Vadim Kochan
2015-07-09 14:57 ` Nikolay Aleksandrov
2015-07-09 15:09 ` Miha Marolt
2015-07-09 15:15 ` Vadim Kochan
2015-07-10 7:09 ` Miha Marolt
2015-07-10 7:16 ` Vadim Kochan [this message]
2015-07-15 12:10 ` Miha Marolt
2015-07-15 12:50 ` [PATCH iproute2] ss: fix display of raw sockets Nikolay Aleksandrov
2015-07-20 21:58 ` Stephen Hemminger
2015-07-09 15:14 ` [BUG] "$ ss -a" incorrectly displays raw sockets as udp sockets Eric Dumazet
2015-07-09 15:21 ` Eric Dumazet
2015-07-09 15:28 ` Vadim Kochan
2015-07-09 15:30 ` Nikolay Aleksandrov
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=20150710071619.GA13710@angus-think.lan \
--to=vadim4j@gmail.com \
--cc=miham@beyondsemi.com \
--cc=netdev@vger.kernel.org \
--cc=nikolay@cumulusnetworks.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.
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).