-
zip
I have a silly networking question
-
zip
I have a machine that is, at least at present, currently sitting on two vlans: 1 and 50
-
rtprio
ask away, perhaps you'll get a silly answer
-
zip
if a machine on vlan 50 happens to try to talk to the interface on br0.1, or a machine on vlan 1 happens to talk to something on br0.50, the return packet comes out the other interface
-
zip
basically it notices that it's got a direct way to talk to the other network segment and just pops the packet out on that path
-
rtprio
by "other interface" you mean "out the properly tagged vlan"?
-
zip
so, my question is: is the problem that I'm doing something daft, or is there a way to tell freebsd to use the appropriate interface
-
zip
well, I hope it's properly tagged
-
rtprio
but br0.50 should ignore vlan 1
-
zip
you'd hope
-
rtprio
but it sounds like that's not the case
-
zip
I mean, maybe it's stupid to have two vlans and hope that traffic between them definitely goes to the router
-
zip
and I think it does the right thing if I'm, eg, talking to a vnet jail
-
rtprio
... isn't it on your switch not sending packets on the right vlan?
-
zip
okay so lessee
-
zip
`ifconfig_br0="vlanfilter defuntagged 50 addm igc0 untagged 1 tagged 50"`
-
zip
so we've got `igc0` is the real interface, `br0.1` and `br0.50` being the interfaces for the machine itself on the same bridge
-
zip
our players here being bluebell (my desktop) and raven (the server)
-
zip
if I use "curl raven50:9100" on bluebell, what raven sees is:
-
zip
br0.50 gets the syn packet, and then br0.1 and igc0 see the syn/ack packet, and so on
-
zip
so we're going around in a little loop
-
zip
bluebell > [vlan1] router [vlan50] > [vlan50/igc0] raven [vlan1/igc0] > bluebell
-
rtprio
someone else is probably more qualified to comment
-
zip
basically I think raven is quite happy to _route_ a packet correctly, but if it _originates_ a packet it notices that it can just dump a packet for the subnet on vlan1 directly into vlan1
-
zip
and yeah, perhaps the answer to "why does this stupid thing happen" is "because you're doing a stupid thing"
-
zip
For now I have solved by not having the machine bridging interfaces
-
zip
which is probably the sensible thing to do
-
sotov1ch
Ahoi
-
antranigv
if I have a port named foo, which relies on files from port bar, how would I mention that in the Makefile?
-
vkarlsen
antranigv: Is it a build, lib or run dependency?
-
antranigv
vkarlsen build, but its not an .so.
-
antranigv
nor is it an executable
-
vkarlsen
antranigv: bar>0:category/bar
-
antranigv
Yeah I keep getting voc-pipes-0.1.0.m.20260124 depends on executable: voc-strutils - not found
-
antranigv
oh wait, I didn't notice the >0
-
antranigv
that helped
-
antranigv
now I need to copy the files of bar into WRKSRC/build/
-
antranigv
manual CP I guess?
-
vkarlsen
antranigv: ${CP} would be more correct(tm), but is it necessary for them to be in there?
-
antranigv
vkarlsen well, yeah, the compiler is looking for the deps.
-
vkarlsen
antranigv: Doesn't it look, or can't you tell it to look for them on the system?
-
antranigv
vkarlsen I probably can tell it to look into the system but then I have to rewrite the Makefile thats shipped with it.
-
vkarlsen
antranigv: Can I have a look?
-
antranigv
-
antranigv
will commit my whole branch of ports now, because that would give you a better picture
-
vkarlsen
Looks like it will download stuff during the build phase too, which poudriere will not allow
-
antranigv
vkarlsen yup, so I ported those too
-
vkarlsen
Projects like these can be a bit tricky. I have to go to bed now, but would love to see more tomorrow
-
antranigv
vkarlsen thank you dear.
-
antranigv