Message ID | 1382205252-14988-1-git-send-email-hagen@jauu.net |
---|---|
State | Changes Requested, archived |
Delegated to: | stephen hemminger |
Headers | show |
On Sat, 2013-10-19 at 19:54 +0200, Hagen Paul Pfeifer wrote: > Currently "ss -emoi" output is nearly perfect: key value(s) tuples are > delimited by colon. Except three groups: socket options, congestion > control algorithm and send data. Especially the first two groups prevent > automated parsing: what if a future congestion control algorithm is > named "faster" - is this a TCP flag or a congestion control algorithm? > The current syntax allow no parsing. > > This patch harmonize the syntax for the last three remaining groups. > > Signed-off-by: Hagen Paul Pfeifer <hagen@jauu.net> > --- Seems fine to me, maybe we want to support json for better parsing capabilities ? -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
* Eric Dumazet | 2013-10-19 14:56:16 [-0700]: >Seems fine to me, maybe we want to support json for better parsing >capabilities ? Currently the parsing is straight forward with Python: split for whitespaces and subsequent split for colon to get a key/value list. Supporting an additional format may be overhead - not sure?! But yeah, JSON formated output is nice too. The current format has the advantage that it is human readable and "good" machine parseable. Except the first line: tcp ESTAB 0 0 192.168.1.29:58951 208.68.163.218:xmpp-cli The line has a strict order, so it is not that hard to write a correct parser. Btw: for people not following google plus circus: I wrote an parser for ss(8) for plotting TCP stack variables: http://research.protocollabs.com/captcp/doc-socket-statistic-module.html Cheers, Hagen
On Sun, 2013-10-20 at 15:24 +0200, Hagen Paul Pfeifer wrote: > * Eric Dumazet | 2013-10-19 14:56:16 [-0700]: > > >Seems fine to me, maybe we want to support json for better parsing > >capabilities ? > > Currently the parsing is straight forward with Python: split for > whitespaces and subsequent split for colon to get a key/value list. Supporting > an additional format may be overhead - not sure?! But yeah, JSON formated > output is nice too. The current format has the advantage that it is human > readable and "good" machine parseable. Except the first line: > Yes, I was only suggesting adding the -j flag, as done lately on other iproute2 commands. Default would stay as is ;) > tcp ESTAB 0 0 192.168.1.29:58951 208.68.163.218:xmpp-cli > > The line has a strict order, so it is not that hard to write a correct parser. > > > Btw: for people not following google plus circus: I wrote an parser for ss(8) > for plotting TCP stack variables: > > http://research.protocollabs.com/captcp/doc-socket-statistic-module.html Yeah, I saw that on your Google+ page, very nice tool ! Thanks -- To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/misc/ss.c b/misc/ss.c index c0369f1..beca2eb 100644 --- a/misc/ss.c +++ b/misc/ss.c @@ -1386,20 +1386,22 @@ static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r, info = RTA_DATA(tb[INET_DIAG_INFO]); if (show_options) { + int is_not_first_element = 0; + printf(" options:"); if (info->tcpi_options & TCPI_OPT_TIMESTAMPS) - printf(" ts"); + printf("%cts", is_not_first_element++ ? ',' : '\0'); if (info->tcpi_options & TCPI_OPT_SACK) - printf(" sack"); + printf("%csack", is_not_first_element++ ? ',' : '\0'); if (info->tcpi_options & TCPI_OPT_ECN) - printf(" ecn"); + printf("%cecn", is_not_first_element++ ? ',' : '\0'); if (info->tcpi_options & TCPI_OPT_ECN_SEEN) - printf(" ecnseen"); + printf("%cecnseen", is_not_first_element++ ? ',' : '\0'); if (info->tcpi_options & TCPI_OPT_SYN_DATA) - printf(" fastopen"); + printf("%cfastopen", is_not_first_element++ ? ',' : '\0'); } if (tb[INET_DIAG_CONG]) - printf(" %s", rta_getattr_str(tb[INET_DIAG_CONG])); + printf(" cc:%s", rta_getattr_str(tb[INET_DIAG_CONG])); if (info->tcpi_options & TCPI_OPT_WSCALE) printf(" wscale:%d,%d", info->tcpi_snd_wscale, @@ -1429,7 +1431,7 @@ static void tcp_show_info(const struct nlmsghdr *nlh, struct inet_diag_msg *r, } if (rtt > 0 && info->tcpi_snd_mss && info->tcpi_snd_cwnd) { - printf(" send %sbps", + printf(" send:%sbps", sprint_bw(b1, (double) info->tcpi_snd_cwnd * (double) info->tcpi_snd_mss * 8000000. / rtt));
Currently "ss -emoi" output is nearly perfect: key value(s) tuples are delimited by colon. Except three groups: socket options, congestion control algorithm and send data. Especially the first two groups prevent automated parsing: what if a future congestion control algorithm is named "faster" - is this a TCP flag or a congestion control algorithm? The current syntax allow no parsing. This patch harmonize the syntax for the last three remaining groups. Signed-off-by: Hagen Paul Pfeifer <hagen@jauu.net> --- misc/ss.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-)