[ipxe-devel] Problem chainloading iPXE from some legacy Intel PXE

[Jedrzej Kalinowski]

 On Thu, 5 May 2011 19:10:58 +0100, Michael Brown wrote:
> On Thursday 05 May 2011 18:03:35 Jedrzej Kalinowski wrote:
>>
>>  First, I've altered the patch a bit, because the pxeprefix_dump 
>> section
>>  was added after the return statement so it wasn't accessed. I've 
>> moved
>>  it 3 lines up.
>>
>>  Finally, I got this dump:
>>  http://www.iem.pw.edu.pl/~kalinoj1/public/pxeprefix_dump.jpg
>
> Thanks.  Sorry about the patch brokennesss.
>
> This is properly strange.  The structure in .prefix appears to be
> intact, but
> the copy made in .data16 is empty by the time undibus_probe() 
> executes.  I
> have no idea what could be causing this, especially since your binary 
> is
> known-good on several other machines.
>
> I would suggest that you try adding
>
>   #include <undipreload.h>
>
>   DBG_HD ( &preloaded_undi, sizeof ( preloaded_undi ) );
>
> at various points in the code (starting with the top of main()) to 
> find out
> when this structure gets zeroed out.
>

 I've placed this in the top of main(), before initialise() function 
 call, and still got zeroes, starting at address 322e0e5c.


 But I've got a hint ;)
 I decided to try old gPXE 1.0.0 undionly.kpxe image and it worked fine!

-- 
 JK