As the title says, some images like core-image-sato-sdk have 2 DHCP clients and this causes a conflict. The the case of core-image-sato-sdk, there is connman and busibox's udhcpc. Steps to reproduce: 1) Boot the image inside a network that has a fast IP address renewing DHCP server. In our case the IP was given on a per-MAC basis. 2) issue 'ps | grep dhcp" and notice: root@genericx86-64:~# ps | grep dhcp 650 root 4728 S udhcpc -R -b -p /var/run/udhcpc.eth0.pid -i eth0 3) Wait for the dhcp lease to expire. Notice in /var/log/messages: ... Jul 10 11:30:27 genericx86-64 daemon.info connmand[660]: eth0 {del} route 0.0.0.0 gw 192.168.99.1 scope 0 <UNIVERSE> Jul 10 11:30:27 genericx86-64 daemon.info connmand[660]: eth0 {add} route 0.0.0.0 gw 192.168.99.1 scope 0 <UNIVERSE> Jul 10 11:30:27 genericx86-64 daemon.info connmand[660]: eth0 {del} route 0.0.0.0 gw 192.168.99.1 scope 0 <UNIVERSE> Jul 10 11:30:27 genericx86-64 daemon.info connmand[660]: eth0 {add} route 0.0.0.0 gw 192.168.99.1 scope 0 <UNIVERSE> Jul 10 11:30:27 genericx86-64 daemon.info connmand[660]: eth0 {del} route 0.0.0.0 gw 192.168.99.1 scope 0 <UNIVERSE> Jul 10 11:30:27 genericx86-64 daemon.info connmand[660]: eth0 {add} route 0.0.0.0 gw 192.168.99.1 scope 0 <UNIVERSE> Jul 10 11:30:27 genericx86-64 daemon.info connmand[660]: eth0 {del} route 0.0.0.0 gw 192.168.99.1 scope 0 <UNIVERSE> Jul 10 11:30:27 genericx86-64 daemon.info connmand[660]: eth0 {add} route 0.0.0.0 gw 192.168.99.1 scope 0 <UNIVERSE> ... These messages go on forever and the result is intermittent loss connectivity. Operations such as using the smart package manager are affected and sometimes fail with 'Network is unreachable'. 4) issue 'ps | grep dhcp" and notice: root@genericx86-64:~# ps | grep dhcp 650 root 4728 S udhcpc -R -b -p /var/run/udhcpc.eth0.pid -i eth0 807 root 4728 S run-parts -a renew /etc/udhcpc.d 808 root 18840 R {50default} /bin/sh /etc/udhcpc.d/50default renew NOTES: - killing the "{50default} /bin/sh /etc/udhcpc.d/50default renew" process solves the issue for the moment. Rebooting the image will bring it back though. - Issuing 'smart remove busybox-udhcpc' will permanently solve the issue, even though 'ps | grep dhcp' still shows the same as with the package installed: root@genericx86-64:~# ps | grep dhcp 650 root 4728 S udhcpc -R -b -p /var/run/udhcpc.eth0.pid -i eth0 - The only images I tested so far for this are core-image-sato-sdk for genericx86-64 and beaglebone. - An email thread on the subject: https://lists.connman.net/pipermail/connman/2013-December/016272.html
Good news, I can reproduce this, so I will continue to bang on it
This patch makes busybox aware of EXTERNAL client and adds busybox into the known clients http://git.yoctoproject.org/cgit/cgit.cgi/poky/commit/?id=a40db96621fdbc36889a53279aa8840513444a76
Hello, Just checked on an core-image-sato-sdk on an emenlow, and now only one dhcp client (via connman) is available. Also did checks on /var/log/messages and no occurrences of the reported issue can be found. I will mark this bug as verified. Andra,