Bizarre network issue
Greetings.....
I am working on a driver for generic serial-based radios (e.g, Coyote Datacomm DR-915 and Microhard MHX-910, etc..), that basically allows the radio to be used as a network interface, much in the spirit of STRIP. Kernel is 2.4.8 (mandrake 8.1).
Given that the radios pose an "unknown", I have gone to a NULL-modem cable
until this issue is resolved. Here's what *is* working across my interface:
1. ARP ... tcpdump shows both the request AND reply.
2. PING ... Getting approx 120ms round trip with the MHX-910s (23ms null modem)
3. UDP ... Works perfectly with netcat in both directions.
Here's what *is not* working:
4. TCP .... tcpdump shows the SYN packet, but no SYN/ACK ever appears
5. ICMP 3/3 ... If I try a UDP session when there's no-body "listening" on that remote port, no "Port Unreachable" message is ever sent back to the
sending host.
The fact that items 1 though 3 work, indicate that 4 and 5 should work as well, but they don't. I have added a debug statement to my driver's "hard_start_xmit" routine to write to syslog when it's called. The kernel does not even appear to be calling the routine to respond to TCP SYN's or UDP packets headed for an unreachable port.
The worst part of this is that TCP was working fine across this interface about a month ago. When I went to pick up where I left off from is when this behavior started to exhibit itself. Any ideas?
Please CC me personally on responses as I am not subscribed to the list.
TIA,
Larry
Signature · 4 lines
Lawrence A. Wimble 414 NE 3rd Street; Suite B Chief Software Engineer Crystal River, FL 34429 Design On Demand, Inc. Phone 352-563-1225 x112 [email protected] Fax 352-563-2098
Quoted 5 lines · Lawrence A. Wimble
1. ARP ... tcpdump shows both the request AND reply.
2. PING ... Getting approx 120ms round trip with the MHX-910s (23ms null modem)
3. UDP ... Works perfectly with netcat in both directions.
Quoted 7 lines
Here's what *is not* working:
4. TCP .... tcpdump shows the SYN packet, but no SYN/ACK ever appears
5. ICMP 3/3 ... If I try a UDP session when there's no-body "listening" on that
remote port, no "Port Unreachable" message is ever sent back to the
sending host.
I'll take a quick guess. Your faked mac address has bit 0 of the first byte set.
Quoted 8 lines
Greetings.....
I am working on a driver for generic serial-based radios (e.g, Coyote Datacomm
DR-915 and Microhard MHX-910, etc..), that basically allows the radio to be used
as a network interface, much in the spirit of STRIP. Kernel is 2.4.8 (mandrake 8.1).
Are you willing share your code? I'd love to see what you have so far since I have a similar project. It is basically using a wireless link that supposed to look like a terminal. I'm hoping that all I'll need to do is make a simple driver to implement hard_start_xmit and let pppd do the rest. Probably like your radio, it has a 0-7 network id which I assume you are including in your MAC addr some how.
Regards,
Mazhar
Quoted 1 line · Mazhar Memon
Are you willing share your code? I'd love to see what you have so far since I have a similar project. It is basically using a wireless link that supposed to look like a terminal. I'm hoping that all I'll need to do is make a simple driver to implement
hard_start_xmitand let pppd do the rest. Probably like your radio, it has a 0-7 network id which I assume you are including in your MAC addr some how.
For network protocols you might also want to look at the daemon called "scarabd" that Robin O'Leary and I did years ago for a low speed radio modem that needed software to manage things like tx/rx, carrier handling and so forth
Quoted 6 lines · Mazhar Memon
Are you willing share your code? I'd love to see what you have so far since I have a similar project. It is basically using a wireless link that supposed to look like a terminal. I'm hoping that all I'll need to do is make a simple driver to implement
hard_start_xmitand let pppd do the rest. Probably like your radio, it has a 0-7 network id which I assume you are including in your MAC addr some how.Regards,
Mazhar
Yes, I am quite willing to share the code. In fact, I hope to have everything wrapped up by the end of this week, and after a little testing, will make it available on our website for download. I'll send an email off-list to anyone interested once it's ready.
Cheers!
Larry
Signature · 4 lines
Lawrence A. Wimble 414 NE 3rd Street; Suite B Chief Software Engineer Crystal River, FL 34429 Design On Demand, Inc. Phone 352-563-1225 x112 [email protected] Fax 352-563-2098