Open fbi-ranger opened 7 years ago
Florian, it would help tremendously if you did not overwrite what you already wrote. Use the empty box at the bottom of the page to add to the incident.
All I can say from above is that if you want to talk to the VM stack, a TUN interface works. Why not try it?
Sorry, I had corrected only a typing error in the message.
OK, lets do it the other way: What I would like to have is a way to integrate the VM stack in my LAN without any subneting and routing and other voodoo.
AFAIK the CTC connection is not doing that as it is a PTP connection. If I remember correctly according to the description, there must be a different network between the host and Hercules. In order to reach VM and the guests on it form other than the host system, that network has to be NATed. And that is point were IPtables and all the other stuff comes into the game isn't it?
The LCS thing worked well, even I would prefer of course to have an OSA3 and under VM with a VSWITCH. But it was second best.
Now after changing the OS version, it seems that there are some bugs in the code.
Now I changed the operating system and LCS is NOT working again.
If the VM system is in the same subnet as the hosting UNIX, all you need to do is enable proxy ARP and IP forwarding for TUN to work.
I tried, never worked...
If you have a CTCI interface with VM/TCPIP (FAL) which is a point to point interface, then there is no way to do this through ARP proxying. It HAS to go through routing.
Which means all other hosts have to be made aware of how to route....
LCS allows the host to be on the same subnet and to use ARP proxying/publication
--Ivan
On 1/7/2017 7:16 PM, John P. Hartmann wrote:
If the VM system is in the same subnet as the hosting UNIX, all you need to do is enable proxy ARP and IP forwarding for TUN to work.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/hercules-390/hyperion/issues/182#issuecomment-271100197, or mute the thread https://github.com/notifications/unsubscribe-auth/ABjMW5lpmqXCajD7U-xwKON6K_2ta8kJks5rP9Z0gaJpZM4LddUy.
That's why I use LCS. It works great if the init-problems would be solved. I don't know what an interrupted system call means. I didn't see that error before. The other error with the fork is also strange. Maybe a consequence?
I guess I have to set fire to my flat in Copenhagen when I return, then, so that I can destroy the evidence. Anyhow, this is likely to be different between a UNIX host and Windows host, so there may be more to it than meets the eye.
Within a bridged network all you need is proxy arp (and throw in IP forwarding--it cannot hurt), if your VM system and the other end of the tunnel are in the subnet.
Surely it would be better to publish rather than destroy the evidence? Like Ivan, I have also tried to get CTCI working without routing and have always failed. John, please publish your methodology, so that those of us less capable than yourself might discover the wrinkle that has prevented our success.
Perhaps the key is that I use a preassigned IP address for the tunnel, thus obviating the need for hercifc. Otherwise, you'll have to be patient. No doubt there is a difference between a UNIX host and a Windows host here. And most of the time I use the host as my terminal, which also changes the picture somewhat.
Unfortunately there are two other problems with LCS:
New start of Hercules brings sometimes the error:
HHC00138E Error setting TUN/TAP: Interrupted system call.
This event is not stable: Exiting Hercules and restart triggers same error OR NOT.
If it works correctly message is displayed: HHC00901I 0:F00 LCS: Interface tap0, type TAP opened.
IPL zVM and starting TCP/IP machine leads to sometimes:
HHC00136E Error in function fork(): Cannot allocate memory.
Hercules continues to open LCS devices F00 and F01. Messages: HHC00942I CTC: lcs device tap0 using mac 7A:AA:0D:DA:59:D8 are shown. But the IP address of z/VM is not reachable.
It appears that the second problem (HHC00136E) is forced when Hercules has more than 4G memory. Reducing it to 4G seems to start the stack correctly.