Red Hat Bugzilla – Bug 195250
anaconda hangs when given eth device that gets no dhcp response
Last modified: 2007-11-30 17:11:35 EST
Description of problem:
This was seen on an HP Integrity ia64 server.
During a network based install I selected an eth device that is disconnected.
Anaconda tries to get an ip address via dhcp and hangs instead of timing out and
allowing the user to select the correct eth device.
Version-Release number of selected component (if applicable):
100% on this system
Steps to Reproduce:
1. start an nfs install
2. when prompted for the ethernet device choose one that is not connected
3. select DHCP
should time out and allow selecting a different device
I have been experimenting with this on other systems. I don't hit it 100% of
the time but I do see it elsewhere.
I can reporoduce it every time after a few tries of:
1. start the install
2. when prompted select an eth device that is disconnected
3. select dhcp
4. if the dhcp times out as it should just re-try dhcp
It will usually hang on the first try. I have not seen it take more than 2
tries to reproduce the hang.
Looks like the IPv6 DHCP request is what fails to timeout. The IPv4 request
times out after 10 seconds, then it moves to IPv6 and you can see it just loops
over and over (check log on tty3). Looking at this in libdhcp6client.
For now you can boot with the 'noipv6' boot parameter and that should disable
IPv6 throughout the installation.
I've patched libdhcp6client, which fixes the timeout problem. Next build in
rawhide will have this fix integrated.
*** Bug 195935 has been marked as a duplicate of this bug. ***
I still see the hang in rawhide-20060623. I am going to re-open but put it into
the state "NEEDINFO_REPORTER" so it is on me to re-verify this next week (in
case the fix did not make it in to the 0623 build yet).
I just verified that this is fixed in rawhide-20060626. The timeout however
does seem to take a very long time (which might be why I thought it was still
broken last week). It certainly takes longer than I remember back in RHEL4. I
will do some comparisons and open a different BZ if it does seem to be a real issue.
The timeout is 45 seconds each for IPv6, per the RFC requirement.