SMBIOS problem with PXE booting

[Evan Bigall <evan.bigall@scalent.com>]

I am using Xen 3.03 on IBM T60s running FC5 and RHEL4 with VT enabled.

We have a standard linux bootable iso that we use to run diagnostics 
(2.6.11 kernel).  When I boot that image in an HVM guest, the 
diagnostics run fine, and SMBIOS looks normal.

When I change nothing in the config but the image to boot, and then boot 
eb-5.4.2-rtl8139.iso (downloaded from *rom*-*o*-*matic*.net), the iso 
boots, correctly, PXEs, tftps our kernel and ramdisk, but then if I run 
dmidecode on the ramdisk, it looks like SMBIOS has been corrupted somehow.

A correct dmidecode starts out like this:

SMBIOS 2.4 present.
10 structures occupying 283 bytes
Table at 0x0009f01
Handle 0x0000
    DMI type 0, 24 bytes
    BIOS information
       Vendor: Xen
       Version 3.0.3-0
       [etc...]      

The corrupted one looks like this:

SMBIOS 2.4 present.
10 structures occupying 283 bytes
Table at 0x0009f01f
Handle 0x009e
    DMI type 0, 0 bytes.
    BIOS Information
Handle 0x09e0
    DMI type 158, 0 bytes.

I've tried integrating the rom image directly into 
/usr/lib/xen/boot/hvmloader, but get the same results.

Any ideas?

I've also noted that in the HVM guest, if I try to boot an SMP kernel, 
it hangs, only the non-SMP kernel successfully boots.

Evan