Bug 860767

Summary:	cxgb3 on vlan-enabled network config does not pass TCP traffic
Product:	Red Hat Enterprise Linux 5	Reporter:	Jay Fenlason <fenlason>
Component:	kernel	Assignee:	Doug Ledford <dledford>
Status:	CLOSED NOTABUG	QA Contact:	Red Hat Kernel QE team <kernel-qe>
Severity:	unspecified	Docs Contact:
Priority:	unspecified
Version:	5.8	CC:	dledford, jfeeney
Target Milestone:	rc
Target Release:	---
Hardware:	Unspecified
OS:	Unspecified
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2013-02-20 15:36:02 UTC	Type:	Bug
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:

Description Jay Fenlason 2012-09-26 16:40:51 UTC

Description of problem:
If you configure a cxgb3 adapter to use a specific vlan, tcp traffic on that interface is dropped/blocked/othewise not sent to userspace.

Version-Release number of selected component (if applicable):
kernel-2.6.18-308.12.1.el5

How reproducible:
Always

Steps to Reproduce:
1.Configure two machines with cxgb3 interfaces to talk over a vlan
2.ping one machine from the other one, confirm that icmp packets are not blocked
3.attempt to ssh from one machine to the other, observe that no connection happens
  
Actual results:
Packets lost

Expected results:
packets passed

Additional info:
We've seen this on multiple network configurations.

Comment 1 Jay Fenlason 2012-10-02 15:42:34 UTC

This only appears to happen when the base interface (the one with no vlan configuration) is not given an IP address and brought up.  Simply doing an "ifconfig cxgb3 0.0.0.0 up" causes the interface to start correctly passing traffic.  I'm investigating what piece of init code isn't being run.

Comment 2 Doug Ledford 2013-02-20 15:36:02 UTC

The base interface must be up before the vlan interfaces are expected to work.  You've confirmed that it works once the interface comes up, so this really is a configuration issue, not a bug.