PXE Boot vmware 8.x fails after pxeboot.

[Zaphod Beeblebrox]

Replying to my own message again, I tried booting the same
configuration on the raw computer hardware (as opposed to vmware).  On
the raw computer hardware, pxeboot correctly loads a kernel and boots
it.  This seems to imply that pxeboot and vmware 8.x don't play well
together.

On Tue, Apr 3, 2012 at 1:27 AM, Zaphod Beeblebrox <zbeeble at gmail.com> wrote:
> Replying to my own message, I add more below...
>
> On Sun, Apr 1, 2012 at 8:44 PM, Zaphod Beeblebrox <zbeeble at gmail.com> wrote:
>> I have had diskless FreeBSD machines before.  I started this project
>> with an eye to booting iscsi disks, but there seems to be no way to
>> communicate the root disk path (and parameters) to FreeBSD ---
>> something that might be solvable, but I need practical at the moment.
>> So I fall back on NFS diskless with PXE boot (I may have used
>> etherboot in the past --- it's been awhile).
>>
>> Anyways... this attempt is made with FreeBSD-9.0-RELEASE binaries.
>>
>> In my network, 192.168.0.1 is the DHCP and TFTP server.  192.168.0.52
>> is my NFS server.  The new vmware guest is bridged and gets
>> 192.168.0.135.  It successfully gets 'pxeboot' onto the vmware guest
>> --- pxeboot prints it's banner.  Then the only network traffic I
>> observe is DHCP Discover (vmware, presumably the pxeboot binary)
>> followed by DHCP Offer (192.168.0.1 again) and this repeats.
>>
>> Now the dhcp offer gives a root path of
>> "192.168.0.52:/vr/diskless/hit" ... and I've tried it with and without
>> a trailing slash.
>>
>> Obviously this is something within the pxeboot's binary as no attempt
>> to make the nfs mount occurs.
>
> ... With a few more variations of this test, I came across a
> configuration where the pxeboot client loaded into the vmware system
> would continue to spam the "options next-server" host to mount "/" ...
> interestingly here, it seems to completely ignore "options root-path"
> ... both the ip address and the path portions alike.
>
> ... but said behavior only occurs with some set of random
> configuration changes on the returned DHCP packets and/or slightly
> different versions of pxeboot (which I've pulled from various hosts
> from 8.x through the 9.x that I'm trying to boot with.
>
> It strikes me that the pxeboot process is hanging somewhere ... or
> overwriting memory ... or somesuch ... on this box.
>
> Has anyone seen or investigated this type of behavior?