How can I figure out which process is opening the certain tcp port?

Question

I usually use fuser command to check pid opening the certain tcp port like the following

fuser 22/tcp //To get pid opening the 22 tcp port

I've got a reference board which running a embedded linux. It have been already opening 22 tcp port for ssh connection. But fuser doesn't display output anything about 22 port. So I tried another ssh daemon to open 322 port then tried to check pid using fuser, it worked fine.

root@imx6qsabreauto:~# netstat -nlt | grep 22
tcp        0      0 0.0.0.0:4224            0.0.0.0:*               LISTEN
tcp        0      0 0.0.0.0:322             0.0.0.0:*               LISTEN
tcp        0      0 :::322                  :::*                    LISTEN
tcp        0      0 :::22                   :::*                    LISTEN

root@imx6qsabreauto:~# fuser 322/tcp
351

root@imx6qsabreauto:~# ps -ef | grep 351
root       351     1  0 01:46 ?        00:00:00 /usr/sbin/dropbear -r /etc/dropbear/dropbear_rsa_host_key -p 322 -B

root       379   315  0 02:11 ttymxc3  00:00:00 grep 351


root@imx6qsabreauto:~# fuser 22/tcp
==> This output nothing !!

How can I figure out which process is opening tcp 22 port. (In the board, lsof command is not available and.. netstat doesn't have -p option.)

Answer 1

I you have /proc mounted and bash and readlink both installed, You can write a small bash script that parses /proc/net/tcp, and scan /proc/*/fd/ to find the corresponding socket.

I'm not so familiar with embedded linux, but if you cannot find readlink, it may be included in busybox.

/proc/net/tcp is something like

sl  local_address rem_address   st tx_queue rx_queue tr tm->when retrnsmt   uid  timeout inode
0: 00000000:4E7A 00000000:0000 0A 00000000:00000000 00:00000000 00000000     0        0 13128 1 ffff8800cf960740 99 0 0 10 0

The local_address is hex string of HOST:PORT, so the script searches for :0016 when you want to search tcp 22 port.

Once it founds the row which contains :0016 in local_address, the inode is the corresponding socket number.

Then it searchs for /proc/*/fd/* which has the socket number using readlink command.

#!/bin/bash
PORT="$1"
HEX_PORT=$(printf %04X $PORT)
INODE=""
if ! [ "$PORT" ];then
  echo "usage $0 [PORT]"
  exit
fi
while read num host_port _ _ _ _ _ _ _ inode _; do
  if [[ $host_port =~ :"$HEX_PORT"$ ]];then
    INODE=$inode
  fi
done < /proc/net/tcp
if ! [ "$INODE" ];then
  echo "no process using $PORT"
  exit
fi
for fn in /proc/[1-9]*/fd/*; do
  if [ "$(readlink $fn)" = "socket:[$INODE]" ];then
    tmp=${fn%/fd*}
    echo ${tmp#/proc/}
  fi
done

Answer 2

Here's a version of ymonad's script that runs on a Tomato/BusyBox router:

#!/bin/sh

# This works with BusyBox on a Tomato router

PORT="$1"
HEX_PORT=`printf %04X $PORT`
INODE=""
if ! [ "$PORT" ]; then
    echo "Find the process that is listening on an open TCP network port."
    echo "usage $0 [PORT]"
    exit 2
fi
# sl localip:port remip:port st tx_q:rx_q tr:when retrns uid timeout inode ...
while read num host_port _ _ _ _ _ _ _ inode _; do
    port=`echo "$host_port" | awk -F: '{print $2}'`
    if [ "$port" = "$HEX_PORT" ]; then
        INODE=$inode
    fi
done < /proc/net/tcp
if ! [ "$INODE" ]; then
    echo "no process using $PORT"
    exit 1
fi
echo "found inode $INODE"
f=`ls -l /proc/[1-9]*/fd/* 2>/dev/null | fgrep "socket:[$INODE]" | awk '{print $9}'`
if ! [ "$f" ] ; then
    echo "no process found using inode $INODE"
    exit 1
fi
pid=`echo "$f" | awk -F/ '{print $3}'`
echo "Process matching PID=$pid:"
ps w | awk "\$1==$pid {print}"

Answer 3

If you have or can get ss on your device it can show you the PID:

ss -ltp # for TCP
ss -lup # for UDP

Answer 4

Thanks @ymonad !! :) As your mention, I've been able to get pid corresponding to the port like the following.

root@imx6qsabreauto:~# cat /proc/net/tcp
  sl  local_address rem_address   st tx_queue rx_queue tr tm->when retrnsmt   uid  timeout inode

   0: 00000000:1080 00000000:0000 0A 00000000:00000000 00:00000000 00000000     0        0 1018 1 d8d90a00 100 0 0 10 0

   1: 00000000:0DA2 00000000:0000 0A 00000000:00000000 00:00000000 00000000     0        0 842 1 d8d90000 100 0 0 10 0

   2: 00000000:006F 00000000:0000 0A 00000000:00000000 00:00000000 00000000     0        0 2515 1 d8dc8000 100 0 0 10 0

   3: 0100007F:0035 00000000:0000 0A 00000000:00000000 00:00000000 00000000     0        0 877 1 d8d90500 100 0 0 10 0

root@imx6qsabreauto:~# cat /proc/net/tcp6
  sl  local_address                         remote_address                        st tx_queue rx_queue tr tm->when retrnsmt   uid  timeo
ut inode
   0: 00000000000000000000000000000000:006F 00000000000000000000000000000000:0000 0A 00000000:00000000 00:00000000 00000000     0
 0 2518 1 d8dd0000 100 0 0 10 -1
   1: 00000000000000000000000001000000:0035 00000000000000000000000000000000:0000 0A 00000000:00000000 00:00000000 00000000     0
 0 881 1 d8de0000 100 0 0 10 -1
   2: 00000000000000000000000000000000:0016 00000000000000000000000000000000:0000 0A 00000000:00000000 00:00000000 00000000     0
 0 4933 1 d8da0000 100 0 0 10 -1

Your shell script is working fine, it can get pid properly as like below.

root@imx6qsabreauto:~# /tmp/find.sh 22
1

Odd thing is the result pid is 1. It's init process ;;

UID        PID  PPID  C STIME TTY          TIME CMD
root         1     0  0 04:21 ?        00:00:04 /sbin/init

I think I need to figure out more how init process opens 22 tcp port. Really thank you. :D I've learned a lot. Thanks again !!