Bug 679188

Summary:	gpxe virtio-net driver incorrectly hardcodes frame length always as 1514
Product:	Red Hat Enterprise Linux 6	Reporter:	Matt Domsch <matt_domsch>
Component:	gpxe	Assignee:	john cooper <john.cooper>
Status:	CLOSED DUPLICATE	QA Contact:	Virtualization Bugs <virt-bugs>
Severity:	medium	Docs Contact:
Priority:	unspecified
Version:	6.0	CC:	cra, gcosta, jforbes, khong, lfarkas, matt_domsch, nobody, tburke
Target Milestone:	rc	Keywords:	Triaged
Target Release:	---
Hardware:	All
OS:	All
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:	678789	Environment:
Last Closed:	2011-03-09 16:31:01 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:	678789
Bug Blocks:

Description Matt Domsch 2011-02-21 20:29:27 UTC

+++ This bug was initially created as a clone of Bug #678789 +++

Created attachment 479710 [details]
virtio-net: set size of transmitted ethernet frame to header + data length

Description of problem:

All ethernet frames sent by gpxe's virtio-net driver have a length of 1514, regardless of the actual size needed to encapsulate the data.  This messes up at least DHCP on some networks, because the DHCP relay agent can't manipulate the "full" ethernet frame.  This was causing KVM PXE failures for me since the DHCP transaction wouldn't complete.

Version-Release number of selected component (if applicable):
gpxe-1.0.1-1

How reproducible:
always

Steps to Reproduce:
1. Use virt-manager to create a VM with network driver virtio
2. Try to PXE boot the VM
3. Monitor the traffic--all frames have an ethernet length of 1514
  
Actual results:
fail to boot via PXE

Expected results:
Should boot via PXE.  virtio-net should set the ethernet frame length to the actual size of the data to be sent + headers.

Additional info:
See attached patch for the fix.

--- Additional comment from cra on 2011-02-19 12:29:03 EST ---

Same fix needs to be applied to F13, F14, F15, rawhide/F16, and EL5 branches.  Thanks.

--- Additional comment from lfarkas on 2011-02-20 15:05:37 EST ---

what's up with el6?

--- Additional comment from cra on 2011-02-21 11:58:13 EST ---

Same bug exists in el6.  Could someone who has a RHEL subscription please open a ticket?  Thanks.