[syslinux] tftp open timeout but with no server side errors

[H. Peter Anvin]

Christopher Deneen wrote:
> Background,
> 
> Client - realtek rtl8111c
> tftpd version is 5.0
> options on use -l -v
> 
> Client:
> 
> PXE-EX32 TFTP Open Timeout
> 
> Server:
> 
> Jun 12 10:48:38 damar in.tftpd[30132]: RRQ from 192.168.1.107 filename
> gpxelinux.0
> Jun 12 10:48:48 damar in.tftpd[30133]: RRQ from 192.168.1.107 filename
> gpxelinux.0
> Jun 12 10:49:24 damar in.tftpd[30134]: RRQ from 192.168.1.107 filename
> gpxelinux.0
> Jun 12 10:50:36 damar in.tftpd[30135]: RRQ from 192.168.1.107 filename
> gpxelinux.0
> etc.
> 
> The client has a valid ip and the server can communicate , tftpd-hpa
> tries to send it the file multiple times
> but it appears that it does not receive it. In order to get more
> information I tried atftpd which has more detailed logging
> and it reports continuous timeouts.
> 
> I was wondering if anyone has experienced this problem before or has
> any suggestions. I'd like to note
> this machine is identical to 40 other machines , to which about half boot.
> 

Getting a packet trace via tcpdump or wireshark would be useful, but
this sounds like a case of something I call "dead receiver syndrome",
which means the IP stack sends, but does not receive, packets.  In your
case it seems to happen for the initial TFTP download, which is fairly
unusual; it is more common for it to happen after the initial TFTP
download as PXELINUX starts up.

Look for the specific PXE stack versions on your machine and see if
there is an updated one.  Since this is happening while still in ROM,
it's unlikely that anything other updating the BIOS to a different
version of the vendor stack or to another stack like gPXE is going to
help.  There is obviously nothing the software can do -- it's not
running yet.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.