Discussion:
rump oddity
(too old to reply)
Patrick Welche
2017-05-12 16:25:15 UTC
Permalink
I believe the attached script sets up the following trivial network in rump:

+---------------------------+
| gw |
int 192.168.1.2 ---- + 192.168.1.1 192.168.2.1 +---- ext 192.168.2.2
+---------------------------+

The puzzle is: while sitting at 192.168.1.2, why can I ping 192.168.2.2
but not 192.168.2.1?


Cheers,

Patrick
(-current/amd64)
Robert Elz
2017-05-12 23:45:56 UTC
Permalink
Date: Fri, 12 May 2017 17:25:15 +0100
From: Patrick Welche <***@cam.ac.uk>
Message-ID: <***@quartz>

| The puzzle is: while sitting at 192.168.1.2, why can I ping 192.168.2.2
| but not 192.168.2.1?

I suspect you may be missing (in the export RUMP_SERVER=$sock_gw block)
something equivalent to
rump.sysctl -w net.inet.ip.forwarding=1
(add -q if you like.) (I don't know if there is a rump.sysctl, but there
must be a mechanism to allow that to be done...)

And while the little diagram you drew might be a logical descr of what
you're configuring, I think a more accurate one would be:

#
# ===================================================================
# | | |
# | +---------------------------+ |
# int 192.168.1.2 + 192.168.1.1 192.168.2.1 + ext 192.168.2.2
# | gw |
# +---------------------------+

gw would need to be configured with shmif0 & shmif1 to accurately
model the original diagram. Not that this should really matter
to the test you tried.

kre


--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Robert Elz
2017-05-17 16:55:28 UTC
Permalink
Date: Wed, 17 May 2017 15:56:58 +0100
From: Patrick Welche <***@cam.ac.uk>
Message-ID: <***@quartz>


| Re the ip.forwarding - I was hit by the POLS: it seems to default to 1,
| at least the experiment attached prints 1, which was unexpected...

If the kernel is configured with "options GATEWAY" it defaults that way.
How rump kernels are configured I have no idea (haven't looked.)

But explicitly setting it would be good defensive practice, if you need
packet forwarding, just in case the default setup changes sometime.

kre


--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Robert Elz
2017-05-13 03:09:23 UTC
Permalink
Date: Fri, 12 May 2017 17:25:15 +0100
From: Patrick Welche <***@cam.ac.uk>
Message-ID: <***@quartz>

| The puzzle is: while sitting at 192.168.1.2, why can I ping 192.168.2.2
| but not 192.168.2.1?

I actually misread that before, and while the correction needed might
be the same (might) for that one, the difference between the diagram
you drew, and the network you actually configured, might be more relevant.

kre


--
Posted automagically by a mail2news gateway at muc.de e.V.
Please direct questions, flames, donations, etc. to news-***@muc.de
Patrick Welche
2017-05-17 14:56:58 UTC
Permalink
Post by Robert Elz
Date: Fri, 12 May 2017 17:25:15 +0100
| The puzzle is: while sitting at 192.168.1.2, why can I ping 192.168.2.2
| but not 192.168.2.1?
I suspect you may be missing (in the export RUMP_SERVER=$sock_gw block)
something equivalent to
rump.sysctl -w net.inet.ip.forwarding=1
(add -q if you like.) (I don't know if there is a rump.sysctl, but there
must be a mechanism to allow that to be done...)
And while the little diagram you drew might be a logical descr of what
#
# ===================================================================
# | | |
# | +---------------------------+ |
# int 192.168.1.2 + 192.168.1.1 192.168.2.1 + ext 192.168.2.2
# | gw |
# +---------------------------+
gw would need to be configured with shmif0 & shmif1 to accurately
model the original diagram. Not that this should really matter
to the test you tried.
Thank you - that was what I was missing! (Implementation not matching diagram)

Re the ip.forwarding - I was hit by the POLS: it seems to default to 1,
at least the experiment attached prints 1, which was unexpected...

Cheers,

Patrick
Loading...