[ipxe-devel] Windows having problems parsing iBFT from recent iPXE versions?

[Floris Bos]

On 10/30/2014 02:02 PM, Michael Brown wrote:
>
>> - iPXE communcation seems to end with an iSCSI read response, iPXE ACKs
>> nicely
>> - then there is this long wait on Windows startup (waiting for disks?),
>> and during that there are some TCP retransmissions of an iSCSI NOP
>> command trying to keep the connection warm from SAN to virtualbox.
>> - straight after that Windows takes over, there is some DHCP/ARP traffic
>> (not shown below), a LLMNR request for wpad, and a new iSCSI login.
>>
>> <snip>
>>
>> - Seems the iSCSI read response is retransmitted lacking the last ACK.
>> Those packets may arrive when Windows is about to take over.
>> - Windows does not seem to do any iSCSI communication
>
> Fantastic debugging work.  Could you possibly copies post the two 
> captures (either as attached raw .pcap files, or uploaded to 
> https://appliance.cloudshark.org/upload/ ?
>

https://www.cloudshark.org/captures/a781ff622adc
https://www.cloudshark.org/captures/0968b06118f5


Note that some communication is missing.
E.g. traffic between the virtualbox VM where Windows is being installed 
(192.168.178.4), and the virtualbox VM that plays for DHCP/installation 
server (192.168.178.100), doesn't touch the host system and therefore 
isn't in this capture.


>>> The iBFT is not created until you attempt to boot from the SAN target.
>
> My mistake; the iBFT is actually created at the moment of the 
> "sanhook" command, which would give time for any warning message to be 
> displayed.  The architectural issue is then that iPXE has no mechanism 
> in place for displaying non-fatal error messages.  We could make 
> int13_describe() return an error, but this would then prevent the SAN 
> device from being hooked.  We could hack a printf() into int13.c, but 
> that kind of crossover between mechanism and policy is the kind of 
> ugliness that I've worked hard for over a decade to eliminate from 
> this codebase.
>
> Since it's a fairly pathological system in which the memory above 
> 512kB is all used by the time iPXE starts up, and since it's 
> definitely not the cause of the problem you're experiencing, I don't 
> propose to do anything about it now.

Did had that issue on other systems, and still think not getting any 
warning and it mysterically failing is very annoying.
Affects the Intel NUC series, and doubt I'm the only one that tried to 
use those in diskless setups.


Yours sincerely,

Floris Bos