Bug 681569

Summary:	anaconda stage1 loader does not pass essential "layer2" network option
Product:	Red Hat Enterprise Linux 5	Reporter:	Philip Rowlands <phr>
Component:	anaconda	Assignee:	David Cantrell <dcantrell>
Status:	CLOSED DUPLICATE	QA Contact:	Release Test Team <release-test-team>
Severity:	high	Docs Contact:
Priority:	unspecified
Version:	5.6	CC:	brueckner
Target Milestone:	rc
Target Release:	---
Hardware:	s390x
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2012-03-22 13:00:16 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:

Description Philip Rowlands 2011-03-02 15:41:11 UTC

Steps to reproduce:
- Buy IBM s390x mainframe
- Configure VSWITCH in "ETHERNET" mode (layer2)
- Punch kernel, parm and initrd files to the reader and IPL
- Continue through linuxrc.s390 to provide basic network and hostname info, including LAYER2=1
- SSH when prompted, to launch stage 1 loader
- Proceed through language selection to "Installation Method".

At this point we choose the installation media; either network-based (NFS, HTTP, FTP) or let's call it block-device based (CDROM, Hard drive).

If we choose network-based, loader performs an extra config step:

14:22:02 INFO : need to set up networking
14:22:02 INFO : going to pick interface
14:22:02 INFO : only have one network device: eth0
14:22:02 INFO : going to do getNetConfig

If we choose "Hard drive", this "set up networking" stage is skipped.

- Continue through the installation, including eth0 network details, and reboot
- Notice that eth0 will not come up

Problem:

anaconda doesn't carry through the LAYER2 config setting when using Hard drive media. For network installs, /tmp/netinfo is rewritten in or around getNetConfig to change LAYER2=1 into OPTIONS="layer2=1", among other edits. It's possible that other network-related parameters from CMSCONFFILE are affected; I only checked LAYER2.

This /tmp/netinfo rewrite is essential to allow later stages of anaconda to write a correct ifcfg-eth0, but the step is skipped.

There is no way to have anaconda correctly configure the network device for layer2 when installing from Hard drive media.

Comment 1 Philip Rowlands 2011-03-15 15:34:11 UTC

Testing in an environment where layer2=0 is used, "ARP=no" is omitted from ifcfg-eth0 when installing from Hard drive. (This doesn't seem as critical to network operation, compared to the layer2=1 scenario above.)

Comment 2 Philip Rowlands 2011-08-11 09:26:11 UTC

Was this fixed in RHBA-2011:0984?

 * The hard drive installation method and layer2 VSWITCH caused non-functional networking on the IBM System z. Now, anaconda writes out the LAYER2 and PORTNO options into ifcfg files also on non-network install methods. (BZ#649301)

As with so many of the bugs linked from RHBA notices, I can't view bug 649301 (not authorised).

Comment 3 David Cantrell 2012-03-22 13:00:16 UTC

Yes, it was.  Sorry this bug was not picked up as a dupe.  While I cannot make the bug visible to you, you can look at the git commit that fixed the issue:

59992e95263dfe13d5284ab44deb9e22f0f98deb

Information on getting the source for anaconda via git is at http://fedoraproject.org/wiki/Anaconda

*** This bug has been marked as a duplicate of bug 649301 ***