Fixing __tmp network interface problems in Linux: Google Bait
Having spent most of today wrestling with an internet gateway box which suddenly decided it was not going to pass packets, and having trawled through the net only to run across multiple incompatible causes and solutions, I can say:
If your ifconfig -a output has a __tmpnnnnnnnnnn entry where there should be an ethN entry (eth0, eth1 etc), the following may fix it:
TEMPORARY FIX:
ifrename -i __tmpnnnnnnnnnn -n ethN
service network restart
PERMANENT FIX:
Edit the appropriate ifcfg-ethN file, which will usually be somewhere under /etc depending on your distro. Lock the interface down by setting the appropriate options (if you have multiple interfaces, look in the other ifcfg-ethN files for inspiration) - in particular, HWADDR might help. The data for that entry can be usually found using ifconfig -a and looking for the __tmp interface. Once the file is edited, run service network restart to reload, and give it a minute before checking ifconfig -a again. If you are after a ppp interface (ppp0 etc), this should come up at the same time, assuming it was running fine before one of the interfaces got stuck as a __tmp.
Note also that service network restart is RedHat-centric; your system may use something like /etc/init.d/networking restart instead. You can also restart just the one interface using ifdown interfacename and ifup interfacename (followed by bringing up ppp if you need it), if you don't want to also bounce an interface you're using to remote into the box. I've maintained remote terminal sessions through a full network bounce, but I wouldn't necessarily risk it on a server I couldn't physically get to easily, particularly if the root cause of the networking problem hadn't been nailed down and all the config files checked beforehand.
If your ifconfig -a output has a __tmpnnnnnnnnnn entry where there should be an ethN entry (eth0, eth1 etc), the following may fix it:
TEMPORARY FIX:
ifrename -i __tmpnnnnnnnnnn -n ethN
service network restart
PERMANENT FIX:
Edit the appropriate ifcfg-ethN file, which will usually be somewhere under /etc depending on your distro. Lock the interface down by setting the appropriate options (if you have multiple interfaces, look in the other ifcfg-ethN files for inspiration) - in particular, HWADDR might help. The data for that entry can be usually found using ifconfig -a and looking for the __tmp interface. Once the file is edited, run service network restart to reload, and give it a minute before checking ifconfig -a again. If you are after a ppp interface (ppp0 etc), this should come up at the same time, assuming it was running fine before one of the interfaces got stuck as a __tmp.
Note also that service network restart is RedHat-centric; your system may use something like /etc/init.d/networking restart instead. You can also restart just the one interface using ifdown interfacename and ifup interfacename (followed by bringing up ppp if you need it), if you don't want to also bounce an interface you're using to remote into the box. I've maintained remote terminal sessions through a full network bounce, but I wouldn't necessarily risk it on a server I couldn't physically get to easily, particularly if the root cause of the networking problem hadn't been nailed down and all the config files checked beforehand.