Syslinux - Jul 2016 - PXELINUX: lpxelinux.0 configuration

Hi Gene,

Am 05.07.2016 um 12:42 schrieb Gene Cumm:
> Version 6.03.  Is this a distribution build or fresh from the
> binary/source archive from kernel.org without any running of
'make'?
it's actually a freshly downloaded lpxelinux.0 from kernel.org. I get:
$ md5sum lpxelinux.0
d77a175ea1a0a8c05b315d179992e1bd  lpxelinux.0
>> nothing more. Sniffing revealed that there are no further attempts to
load
>> anything.
>
> Where/how are you sniffing?  I'm guessing from the system with the
> tftpd?  pxelinux.0 and lpxelinux.0 should both load ldlinux.c32 next
> then try for configuration file.  Which file was last attempted?  What
> file was last successful?  Do you see pxelinux.0 loading ldlinux.c32?
I am sniffing old school with a hub right next to the thin client.

With pxelinux.0 everything works fine: DHCP request and ACK, ARP request 
and reply, pxelinux.0 gets loaded, ldlinux.c32 and the pxelinux.cfg/* 
files as well, and finally vmlinuz and initrd get downloaded.

It's strange that after the same start (DHCP request and ACK, ARP 
request and reply) lpxelinux.0 gets loaded and then follows an 
(unnecessary) gratuitous ARP for the client IP (172.16.12.1) plus an 
unanswered (!) second ARP request for the server's IP (172.16.0.96). No 
further network traffic is send out by the client.
>
>> I expected lpxelinux.0 to load the same configuration file as
pxelinux.0,
>> either by TFTP or HTTP.
>> This is from my default config file:
>> KERNEL http://172.16.0.96/boot/vmlinuz
>> APPEND initrd=http://172.16.0.96/boot/initrd
>
> I'm guessing this is but one LABEL within your config that you're
> hoping to load.
Yes, this is the full file:

TIMEOUT 0
DEFAULT default
DISPLAY product.txt
LABEL default
         KERNEL http://172.16.0.96/boot/vmlinuz
         APPEND initrd=http://172.16.0.96/boot/initrd [?]
#       SAY Hallo
>> What am I missing? Do I have to set any further DHCP options?
>
> Nope.  There's a chance you have a deaf client or one that doesn't
> properly report interrupt status.
As said, with pxelinux.0 this client is booting up flawlessly. Strange.

Greetings
Christof

On Tue, Jul 5, 2016 at 11:49 AM, Christof B?ckler <cb at wvsgym.de>
wrote:
> Hi Gene,
>
> Am 05.07.2016 um 12:42 schrieb Gene Cumm:
>>
>> Version 6.03.  Is this a distribution build or fresh from the
>> binary/source archive from kernel.org without any running of
'make'?
>
>
> it's actually a freshly downloaded lpxelinux.0 from kernel.org. I get:
> $ md5sum lpxelinux.0
> d77a175ea1a0a8c05b315d179992e1bd  lpxelinux.0
>
>>> nothing more. Sniffing revealed that there are no further attempts
to
>>> load
>>> anything.
>>
>>
>> Where/how are you sniffing?  I'm guessing from the system with the
>> tftpd?  pxelinux.0 and lpxelinux.0 should both load ldlinux.c32 next
>> then try for configuration file.  Which file was last attempted?  What
>> file was last successful?  Do you see pxelinux.0 loading ldlinux.c32?
>
>
> I am sniffing old school with a hub right next to the thin client.
>
> With pxelinux.0 everything works fine: DHCP request and ACK, ARP request
and
> reply, pxelinux.0 gets loaded, ldlinux.c32 and the pxelinux.cfg/* files as
> well, and finally vmlinuz and initrd get downloaded.
>
> It's strange that after the same start (DHCP request and ACK, ARP
request
> and reply) lpxelinux.0 gets loaded and then follows an (unnecessary)
> gratuitous ARP for the client IP (172.16.12.1) plus an unanswered (!)
second
Unnecessary but not hurting anything.
> ARP request for the server's IP (172.16.0.96). No further network
traffic is
This is a major problem.  What's your actual subnet?  172.16.0.0/20?
> send out by the client.
>>
>>
>>> I expected lpxelinux.0 to load the same configuration file as
pxelinux.0,
>>> either by TFTP or HTTP.
>>> This is from my default config file:
>>> KERNEL http://172.16.0.96/boot/vmlinuz
>>> APPEND initrd=http://172.16.0.96/boot/initrd
>>
>>
>> I'm guessing this is but one LABEL within your config that
you're
>> hoping to load.
>
>
> Yes, this is the full file:
>
> TIMEOUT 0
> DEFAULT default
> DISPLAY product.txt
> LABEL default
>         KERNEL http://172.16.0.96/boot/vmlinuz
>         APPEND initrd=http://172.16.0.96/boot/initrd [?]
> #       SAY Hallo
>
>>> What am I missing? Do I have to set any further DHCP options?
>>
>>
>> Nope.  There's a chance you have a deaf client or one that
doesn't
>> properly report interrupt status.
>
>
> As said, with pxelinux.0 this client is booting up flawlessly. Strange.
>
> Greetings
> Christof


-- 
-Gene

On Wed, Jul 6, 2016 at 7:34 AM, Christof B?ckler <cb at wvsgym.de>
wrote:
> Am 06.07.2016 um 12:38 schrieb Gene Cumm:
>>>
>>> It's strange that after the same start (DHCP request and ACK,
ARP request
>>> and reply) lpxelinux.0 gets loaded and then follows an
(unnecessary)
>>> gratuitous ARP for the client IP (172.16.12.1) plus an unanswered
(!)
>>> second
>>
>>
>> Unnecessary but not hurting anything.
>>
>>> ARP request for the server's IP (172.16.0.96). No further
network traffic
>>> is
>>
>>
>> This is a major problem.  What's your actual subnet? 
172.16.0.0/20?
>
>
> It's 172.16.0.0/16.
>
> Greetings
> Christof
Good.  Nothing silly like ARP for out of subnet address due to missing
default route. Sound like an intra-segment filter/firewall throttling
ARP, including possibly an OS firewall unless the ARP is malformed or
the frame was otherwise dropped (ie saturated buffer for link).

Any chance of seeing a capture off the hub and off the server?



-- 
-Gene