Syslinux - Jul 2014 - lpxelinux hangs under Intel Boot Agent 1.3.81 (2.1 build 089) on Dell Optiplex 990 BIOS A16

On Thu, Jul 10, 2014 at 7:24 PM, H. Peter Anvin <hpa at zytor.com>
wrote:
> On 07/10/2014 04:07 PM, Alexander Perlis wrote:
>> Trying to PXE boot a Dell Optiplex 990 with BIOS A16, which has Intel
>> Boot Agent 1.3.81 (2.1 build 089). It gets network and launches
>> lpxelinux.0, I see the PXELINUX banner line, but there it hangs.
>>
>> I tried both 6.03-pre18 and 5.10, same hanging behavior. If I instead
>> use pxelinux.0 or gpxelinux.0, either 6.03-pre18 or 5.10 or 4.07, there
>> are no problems. The problem seems isolated to lpxelinux.0, and
isolated
>> to this specific hardware/BIOS. We're otherwise PXE booting to
>> lpxelinux.0 6.03-pre18 on a bunch of different machines without
trouble.
>>
>> I can update the BIOS to A18 and try again with possibly newer Intel
>> Boot Agent code, if that report would help, but I only have this one
>> machine of this type. Thus, before I upgrade to A18, in case any
>> developers on this list want me to try anything with the current broken
>> setup, if that will help isolate a problem in lpxelinux.0 (or a way for
>> lpxelinux.0 to work around a problem in the Intel Boot Agent or Dell
>> BIOS), I'll be glad to try some things and report back.
>>
>
> Could you get a packet trace from the server, to see if you see any
> traffic at all before it dies?
If the traffic appears like any responses to the client after
lpxelinux.0 starts running are probably ignored by the client PC, try
4.10-pre22 and report the value of flags and the first 3 octets of the
MAC address.  This may require the same workaround that some other
models needed (specifically a 760).

-- 
-Gene

On 07/10/2014 06:39 PM, Gene Cumm wrote:
> On Thu, Jul 10, 2014 at 7:24 PM, H. Peter Anvin <hpa at zytor.com>
wrote:
>> On 07/10/2014 04:07 PM, Alexander Perlis wrote:
>>> Trying to PXE boot a Dell Optiplex 990 with BIOS A16, which has
Intel
>>> Boot Agent 1.3.81 (2.1 build 089). It gets network and launches
>>> lpxelinux.0, I see the PXELINUX banner line, but there it hangs.
>>
>> Could you get a packet trace from the server, to see if you see any
>> traffic at all before it dies?
>
> If the traffic appears like any responses to the client after
> lpxelinux.0 starts running are probably ignored by the client PC, try
> 4.10-pre22 and report the value of flags and the first 3 octets of the
> MAC address.  This may require the same workaround that some other
> models needed (specifically a 760).
>
Trace shows that lpxelinux.0 6.03-pre18 is transferred, and after that 
there is only ARP traffic: a "gratuitous" ARP announcement by the 
client, and then a string of ARP requests and replies in which the 
client is asking for the MAC of the PXE server, about every 3 seconds, 
for about 2 minutes, after which I see a message on the client that the 
boot failed and to press a key to try again. (So I previously 
misreported: it's not a total hang...)

Switching to 4.10-pre22, there is no lpxelinux.0, but instead using 
pxelinux.0 4.10-pre22, flags is 0xdc1b, and the first three octets of 
the client MAC are 18:03:73 (which tshark reports as "Dell_"). Note
that
looking at the packet capture in this case, after transferring 
pxelinux.0, there is only that initial "gratuitous" ARP announcement, 
then the wire goes silent, and this time it seems like a true hang, no 
ctrl-alt-del, had to press the power button to reboot.

Alex

On 07/11/2014 10:17 AM, Alexander Perlis wrote:
> 
> Trace shows that lpxelinux.0 6.03-pre18 is transferred, and after that
> there is only ARP traffic: a "gratuitous" ARP announcement by the
> client, and then a string of ARP requests and replies in which the
> client is asking for the MAC of the PXE server, about every 3 seconds,
> for about 2 minutes, after which I see a message on the client that the
> boot failed and to press a key to try again. (So I previously
> misreported: it's not a total hang...)
> 
So it is sending packets but not getting the response.

	-hpa