[syslinux] A few questions for the braintrust

[H. Peter Anvin]

On 08/18/2015 01:13 PM, Jeff via Syslinux wrote:
> I have a few questions to throw out to the collective braintrust.
> 
> 1. in core/fs/pxe/tftp.c, when communicating using udp, in connecting, 
> transmitting and receiving functions, errors are not communicated to the 
> user. In fact the comment after transmit specifies "/* If the WRITE call 
> fails, we let the timeout take care of it... */"
> 
> The question is: Should we notify the user that there has been a problem? 
> In our scenario, there was just a very long pause (all of the retries) and 
> nothing hitting the wire. Since we have a watchdog in our system, it 
> reboots prior to completing all of the timeouts due to prolonged 
> inactivity. Is there a better, timelier or more user friendly way to 
> handle the errors? 
> 
> 2. Just curious as to why the decision was made to create new udp 
> connection instead of using the existing pxebc? udp is already set up and 
> configured. The reason I ask is we were having problems in our customer 
> environment with the routetable. We modified the code to use pxebc 
> protocol and it works perfectly. This environment consists of three 
> separate subnets, client on one, dhcp server on a second and proxy and 
> tftp servers on the third. This scenario was causing problems with routing 
> so the udp connection failed and the transmit never hit the wire. Of 
> course, the receive also failed.
> 
> If anyone would like to see the changes we made to use pxebc, I would be 
> happy to post but thought there is probably a very good reason in other 
> environments to recreate the udp connections and didn't want to muddy the 
> waters. I also think it may not be a complete solution.
> 

The problem with the pxebc is that it only supports one concurrent TFTP
connection, and Syslinux expects to be able to keep multiple files open
at the same time.

	-hpa