Skip site navigation (1)Skip section navigation (2)
Date:      Mon, 06 Feb 2012 17:32:43 +0100
From:      =?ISO-8859-1?Q?Erik_N=F8rgaard?= <>
To:        Rick Miller <>
Subject:   Re: and vlan tagging
Message-ID:  <>
In-Reply-To: <>
References:  <> <> <>

Next in thread | Previous in thread | Raw E-Mail | Index | Archive | Help
On 06/02/2012 17:11, Rick Miller wrote:
> See my responses inline...
> On Mon, Feb 6, 2012 at 10:54 AM, Erik Nørgaard<>  wrote:
>> On 06/02/2012 16:35, Rick Miller wrote:
>>> We are PXE booting into FreeBSD 8.2-RELEASE to perform system builds.
>>> The file was recompiled with TFTP support.  8.2-RELEASE
>>> builds were working fine until we attempted a build on bare metal in
>>> an environment that utilizes vlan tagging.  When the system loaded the
>>> file it prompted that a disk containing the mfsroot.gz be
>>> inserted.
>> Hi, a few questions:
>> 1) do you use tftp or nfs? if you built pxeboot with tftp, did you also
>> build the kernel with tftp?
> was compiled with TFTP enabled.  We made 3 modifications to
> the kernel not related to tftp.  One change was to the NIC source code
> to improve small packet performance, we compiled a custom kernel that
> disabled FLOWTABLE and enabled ROUTETABLES.
> TFTP is the intent for PXE, we did not want to rely on NFS.  For the
> remainder of the install, we've written code into sysinstall to
> support pure HTTP installs.
>> 2) do you fetch the kernel successfully?
>> When using tftp, The kernel and kernel modules are fetched before the memory
>> file system, so do pxeboot fetch the kernel but not the mfsroot?
>> The reason for these questions is that your problem may be with the kernel
>> and kernel modules and not pxeboot. Just to be sure.
> I see what you are saying.  We will have to look at the packet
> captures to make that determination.

There used to be some kernel options in older versions but I think the 
necessity for these were removed with 7.0. Can't find them right now.

Also, there is a trick when you want to use tftp, by default the kernel 
will try nfs if it has been built with nfs code, so check for references 
to nfs. The generic kernel is built with with the option

options 	NFSCL
options 	NFSD
options 	NFSLOCKD
options 	NFS_ROOT

these should be disabled.

> Thanks for this information, it has given me some other things to
> think about and could potentially lead to a proper solution.

My approach to this is to create a closed network with one server 
providing all the necessary services (dhcp, tftp and ftp/http) as well 
as a local mirror with all the required files for installation.

If you have a lot of servers, it makes no sense that each sould fetch 
packages from the public mirrors, hence for installation there should be 
no need for access to external networks.

If you do this you can configure your switches accordingly without any 
need for vlan tagging.

BR, Erik
M: +34 666 334 818
T: +34 915 211 157

Want to link to this message? Use this URL: <>