bridges & tftp

[lejeczek]

hi guys.

I have a bit weird (mis)behavior with TFTP server which 
would serve on an _bridge_ iface a okey but sometimes it 
"snaps" and TFPT stops responding, errors out:
...
tftpd: read(ack): No route to host
...

Bridge itself is also used by _libvirt_.
When bridge "snaps" it does it only for TFTP (or I'm not 
aware of anything else being affected).
Interestingly, an Ubuntu VM on another node/host using 
virtually identical bridge(which is also used by _libvirt_ 
and that Ubuntu VM) can put/get to that TFTP server which 
otherwise errors out for Centos..
whereas a Centos VM on the host as Ubuntu ends up with TFTP 
server erroring out as above.

Centos, both VM and metal, seem can be helped, when a tftp 
client, with creation of firewall policy which example - 
clientConntrack - from man pages I borrow verbatim.
but..
that "fix" is really nothing in grand scheme of things for 
it leaves out netbooting, all if it I think, certainly PXE 
clients calling TFTP.

I hope some experts can suggest, how to possibly tune/tweak 
net-bridge part of the whole thing so that "snap" will not 
happen nor have affect if it did happen.
* "snap" I think happens when there is multiple at once or 
in frequent successions of _tftp_ connections, but it's more 
of a feeling.

many thanks, L.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/networkmanager/attachments/20231226/98cea8e0/attachment.htm>