[ipxe-devel] Ipxe cannot handle bigger TFTP blocksize?

Michael Brown mbrown at fensystems.co.uk
Sat Jul 16 00:57:12 UTC 2011


On Wednesday 08 Jun 2011 10:15:21 Tom.LAUWERS at ext.ec.europa.eu wrote:
> This works, but is very slow. So I replace line 7 with:
> 
> 	Bcdedit /store %BCD-File% /set {ramdiskoptions} ramdisktftpblocksize 8192
> 
> Now it still works on pc's using the internal NIC's PXE-boot code (and it's
>  much faster), but it doesn't work anymore with iPXE.
> 
> My guess is that iPXE cannot handle TFTP-responses that are spread over
> multiple ethernet frames. 1432 bytes works well because with overhead it's
> lower than the MTU (1500 bytes). But when the TFTP-server sends 2 or more
> UDP packets (to send a bigger block of data, requiring less
> acknowledgements), these somehow get lost.

IPv4 fragmentation reassembly should now be functional, as of commit

  http://git.ipxe.org/ipxe.git/commitdiff/13186b6

Could you try with the latest git tree?

Thanks,

Michael



More information about the ipxe-devel mailing list