On 07/11/2023 13:43, Mauricio Tavares wrote:
On Tue, Nov 7, 2023 at 7:15 AM lejeczek via users
<users@lists.fedoraproject.org> wrote:
Hi guys.

cannot get a client to get anything from 'in.tftp' server?

Connection seems to get established, _tcpdump_ shows:
...
13:13:08.331277 IP 10.3.1.100.35176 > whale.mine.priv.tftp: TFTP, length 19, RRQ "zabbix.zip" octet
13:13:13.331324 IP 10.3.1.100.35176 > whale.mine.priv.tftp: TFTP, length 19, RRQ "zabbix.zip" octet
...
journal log for tftp.service, so on the servers, shows:
...
tftpd: read(ack): No route to host
...

any & all thoughts on how troubleshoot & fix this are much appreciated.
many thanks, L.
___________________
      Can you connect from client host to tftp host using other services?
_______________________________________________
Yes, everything works as... well, as expected/normal.
Meanwhile...

I've read through my old notes and found:

.. When you use in.tftpd on interface like bridges, this probably won't work, it at least did not for Dell PowerConnect when you wanted to upload the image, it worked on regualar eth

And it today, a couple of years later and different hardware, seems that still with bridge ifaces, for both server & client, the issue remains.

2nd meanwhile - I'm trying _dnsmasq_ which I'm new to thus errors are possible but... it seems that the same issue remains.

with _dnsmasq_. tftp client - as with in.tftp as the server - "times out" but _dnsmasq_ server thinks and says that:

...

sent /var/lib/tftpboot/zabbix.zip to 10.3.1.100
...

This might go as far down as kernel's tweaking or perhaps bridge's iface - if possible at all - any kernel/net experts are most welcome to comment.

many thanks, L.