tc multiple piped actions results performance hit, seek better way - linux-kernel

I have been experimenting with using tc, tc-nat and tc-pedit to perform stateless nat operations. My goal is to mirror packets from one interface (eth0) to another (eth1), but I also want to change the destination IP, and destination MAC, so the packets can be delivered to another host.
Host A [Eth0 - > Eth1 (change IP dst, change MAC dst)] -------> HOST B
Because I have other applications listening on Eth0, I can't perform either tc-nat or tc-pedit on the ingress/egress on Eth0. So this is what I did
Create a dummy interface
Mirror packets from Eth0 to dummy interface
Run tc-nat (to change destination IP) on egress side of dummy, pipe into tc-pedit (to change destination MAC), and pipe to mirror to Eth1.
The above contraption works as expected, but the performance is terrible (particularly when compared to a libpcap based application such as tcpbridge).
I started this approach hoping to achieve better performance in kernel space vs libpcap based application in user space. I'm curious is there a better way.

Related

How to get the underlying network interface used by a VPN connection in Windows

The closest thing to a solution I've found is using Get-NetConnectionProfile to return all active interfaces, which works fine when there's only an active physical interface and the VPN itself. However, this would not work if the user's machine has 2 active physical interfaces (e.g Wi-Fi + Ethernet) along with the VPN.
Ideally, I'd like a solution that works similarly to "ifconfig -v" in MacOS, which tells you the effective interface for a virtual interface:
Unfortunately it seems there is no sure-fire way to get the underlying physical adapter for a VPN using a Windows API. Short of involving a packet sniffer such as Wireshark, the best solution I found involves parsing the output of two PowerShell commands: Get-NetAdapter and Get-NetRoute.
With the information from these commands, I can know which interfaces are virtual and which ones are physical, and I can rank the physical interfaces by 3 different criteria (in case of tie, we move on to the next criteria):
Sorting the physical interfaces by the interface metric + the route metric to the default gateway (0.0.0.0).
Wired connections over wireless ones (PhysicalMediaType=802.3).
Prioritizing faster adapters.
With this logic all the VPNs I tested appear to reflect the expected network interface, although some VPNs let you force traffic through a particular physical adapter in which case obviously this all goes out the window.
First,
You can install wireshark or some other traffic monitoring tool and capture the relevant packets (filtering using openvVPN protocol or port etc.)
Second,
As far as I know there is no hard linking between the virtual network interface and the regular (ethernet, WIFI etc.) interface, at least not in OpenVPN (there are diffrent VPN protocols). The openVPN packets will be routed to the remote server using you OS routing table.
This way if your ethernet interface is your primary default gateway, and it gets unplugged, your VPN service will be able to recover, since it will have a route to your remote VPN-server address using your WIFI interface.

Is it possible to detect which interface is connected to device with known IP without changing interface network?

My knowledge of networking is very basic so please bear with what I am sure are some very fundamental questions.
Is it possible to determine which interface/adapter (on a machine with multiple physical and virtual interfaces/adapters) is directly connected to a device with a known static IPv4 address without changing the network (i.e. the first 3 octets of its IPv4 address) of the interface/adapter beforehand?
I am asking this because I am trying to automate the setting of interface/adapter static IPv4 addresses on multiple machines in order for them to be able to communicate with the device and for it to work even if the machines do not use the same interface/adapter.
My original idea was to brute-force it by
Getting the device's IPv4 address as user input
Getting all physical interfaces/adapters connected to an Unidentified.network
Saving their previous network settings somewhere and loop through them while configuring them to use an appropriate static IPv4 address based on input in step 1
Attempting to ping device for a specified duration and if timeout occurs, revert the network settings of the interface/adapter to its original one otherwise, keep the new network settings and break out of the loop
My current worry is that the above process may inadvertently cause original network settings to be lost if some unpredictable crash were to occur during it.
So, it would be ideal if I could skip steps 2 and 3 and immediately detect which interface/adapter needs to have its IP address changed but based on what I read, this does not seem possible at all since if you could detect the device, then there would be no need to set a static IP address for the interface/adapter in the first place (cause you can already detect it).
I've also seen that using the ARP cache would work if the first-time connection on all machines are done manually and then setting the appropriate entry to persistent so that it would survive across reboots but I would ideally want even first-time connections to be automated if possible.
Can anyone provide any insight as to whether I am attempting the impossible?
Or is there a better way to achieve what I want?
I have seen some recommendations to use wireshark to sniff for incoming ARP packets but that would require installing it on all the machines I would be deploying my automated approach on which may not be feasible.
For context, I am automating this on Windows 10 and would prefer using powershell/bash/python.
Edit 1
Thanks to the link provided in the comment, I have gained a better understanding of IPv4 terminology.
So I guess I can reformulate my question.
Device
IPv4: 192.168.0.216/24
Network: 192.168.0.0
Machine
First interface/adapter
IPv4: 169.254.19.133/16
Network: 169.254.0.0
Second interface/adapter
IPv4: 169.254.27.245/16
Network: 169.254.0.0
If there is already a physical connection between the second interface/adapter and the device, is there a way to detect (not communicate with) the device without having to change the network of the second interface/adapter beforehand?
Or is brute-forcing it the only way to achieve my goal?

Ruby sending packet to a known MAC address

I like to send packet to a device with only MAC address associated to it and also connecting to my local LAN.
Is there anything in ruby / gem library that I can use to perform such task?
I want to do something like tcpreplay but in ruby instead. Thanks!
MACs are the addresses used on OSI layer 2, the package type you are using on this layer are e.g Ethernet frames. IP Sockets live on layer 3/4, thus if you are using an IP socket, you are sending IP packet.
So if you want to send something "to a MAC address", you need to send raw Ethernet frames which need to be understood on the other side. Given the really low level of this stuff, this is rarely what you really want to do.
Instead, you should try to get the proper IP addresses of the remote device in your local network (technically the broadcast domain) with either a broadcast/reply mechanism or with multicast (which is used e.g. by Bonjour). That way you can always use the much more supported IP layer. For that, your device needs to have an IP address of course. You can establish those with mechanisms like DHCP or address auto configuration for IPv6.
As for your question, I'm not aware of any library for Ruby to send raw Ethernet frames. But given you would need to invent a new protocol besides IP anyway (which is hard, really hard), sending the package is probably the least of your concerns.
I suppose there are a couple of approaches. You didn't say whether you wanted to send an IP packet of some kind, and if so you didn't say which IP protocol. (I can tell you want to send an ethernet frame but ... what encapsulated protocol?)
Anyway, you can open a SOCK_RAW with Ruby, something like:
Socket.open(Socket::PF_INET, Socket::SOCK_RAW, Socket::IPPROTO_RAW)
Another approach might be to just pick a link-local IP and then establish an IP/MAC mapping in your local ARP cache with something like:
system "arp -s host hw_addr"
Then you could communicate with it using more familiar socket ops using the link-local IP address.
You need to be root for all of these things.

Find IP address of directly connected device

Is there a way to find out the IP address of a device that is directly connected to a specific ethernet interface? I.e. given one host, one wired ethernet connection and one second host connected to this wired connection, which layer or protocol below IP could be used to find this out.
I would also be comfortable with a Windows-only solution using some Windows-API function or callback.
(I know that the real way to do this would probably via DHCP, but this is about discovering a legacy device.)
Mmh ... there are many ways.
I answer another network discovery question, and I write a little getting started.
Some tcpip stacks reply to icmp broadcasts.
So you can try a PING to your network broadcast address.
For example, you have ip 192.168.1.1 and subnet 255.255.255.0
ping 192.168.1.255
stop the ping after 5 seconds
watch the devices replies : arp -a
Note : on step 3. you get the lists of the MAC-to-IP cached entries, so there are also the hosts in your subnet you exchange data to in the last minutes, even if they don't reply to icmp_get.
Note (2) : now I am on linux. I am not sure, but it can be windows doesn't reply to icm_get via broadcast.
Is it the only one device attached to your pc ?
Is it a router or another simple pc ?
To use DHCP, you'd have to run a DHCP server on the primary and a client on the secondary; the primary could then query the server to find out what address it handed out. Probably overkill.
I can't help you with Windows directly. On Unix, the "arp" command will tell you what IP addresses are known to be attached to the local ethernet segment. Windows will have this same information (since it's a core part of the IP/Ethernet interface) but I don't know how you get at it.
Of course, the networking stack will only know about the other host if it has previously seen traffic from it. You may have to first send a broadcast packet on the interface to elicit some sort of response and thus populate the local ARP table.
Windows 7 has the arp command within it.
arp -a should show you the static and dynamic type interfaces connected to your system.
Your Best Approach is to install Wireshark, reboot the device wait for the TCP/UDP stream , broadcasts will announce the IP address for both Ethernet ports
This is especially useful when the device connected does not have DHCP Client enabled, then you can go from there.
You can also get information from directly connected networking devices, such as network switches with LDWin, a portable and free Windows program published on github:
http://www.sysadmit.com/2016/11/windows-como-saber-la-ip-del-switch-al-que-estoy-conectado.html
LDWin supports the following methods of link discovery: CDP (Cisco Discovery Protocol) and LLDP (Link Layer Discovery Protocol).
You can obtain the model, management IP, VLAN identifier, Port identifier, firmware version, etc.

Win32 sockets - Forcing ip packets to leave physical interfaces when sending to other local interfaces

Summary: I'm trying to create sockets to pass data between two physical interfaces that exist on the same machine, and Win32 sockets always forwards the traffic directly in the kernel instead of pushing through the physical interfaces. Is there any way to disable this behavior, perhaps through device settings, registry tweaks, routing table shenanigans, or socket options? We're using Windows XP SP3.
Some background. I'm attempting to build some completely automated IP tests to exercise our custom IPv4 equipment. We have a large lab of Windows XP machines, and individual physical ethernet interfaces for each device we're connecting to. Our devices are effectively ethernet routers each with their own IPs.
We need to send data out our lab machines, through our devices, then back into the same computer. We will be sending Unicast and Multicast UDP, TCP, and broadcast IP traffic through the devices.
We want (and likely need) the traffic to originate on the same machine it is destined to.
To do this, we configure two separate NICs each with their own IP on their own subnet, for instance NIC #1 with 10.0.0.1/24 and NIC #2 with 10.0.1.1/24. Our devices then act like simple passthrough routers, and have two interfaces, one on the 10.0.0.0/24 subnet, one on the 10.0.1.0/24 subnet, which they just forward packets back and forth from.
To generate our data, we'd like to be able to use Win32 sockets, since it is well-understood, well-supported, what our customers are using, and would probably be the most rapid approach. Packet injection is probably feasible for UDP and broadcast IP, but very likely not so for TCP. I'd entertain ideas that used packet injection, but would strongly prefer standard Win32 sockets.
As stated in the summary, the packets never leave the machine. I've googled like a madman and I've not found much. Any ideas?
Use Windows' command-line ROUTE utility. You can configure it so any IP packet sent to a specific IP address on a specific Subnet gets sent to another IP/device. For example:
route ADD <NIC_1_IP> MASK <NIC_1_SUBNET> <DEVICE_IP_CONNECTED_TO_NIC_2> METRIC 1
route ADD <NIC_2_IP> MASK <NIC_2_SUBNET> <DEVICE_IP_CONNECTED_TO_NIC_1> METRIC 1
Alternatively, if you know the index numbers of the NIC interfaces, you can specify them instead:
route ADD <NIC_1_IP> MASK <NIC_1_SUBNET> METRIC 1 IF <NIC_2_INTF>
route ADD <NIC_2_IP> MASK <NIC_2_SUBNET> METRIC 1 IF <NIC_1_INTF>
This way, whenever a packet is sent to NIC #1's IP, the packet goes to the device connected to NIC #2, which will then pass it on to NIC #1. And vice versa for packets sent to NIC #2's IP.
For instance, this is a useful technique for allowing WireShark to capture local IP traffic if the PC is connected to a network with a router. Packets from one local IP/Port to another local IP/Port can be bounced off the router back to the PC so they travel through physical interfaces that WireShark can monitor (WireShark will see duplicate copies of each local packet - one outbound and one inbound - but you can filter out the duplicates).
Winsock is always going to bring the packet data up into the kernel space and deal with it there. Thats the whole point to a generic API is that any device is dealt with at the same "layer". If you want to stick with Winsock, I don't believe you can (or would want to) work around this behavior.
You can remove some of the buffer copying with TransmitPackets or TransmitFile, but not between two device interfaces.
That being said, are you having a performance issue with the additional buffer coping that Winsock performs? Security concerns?
How about running the endpoints of your tester inside of distinct virtual machines? Then you need only a single piece of hardware, but you'll have separate TCP/IP stacks that don't know each other are local (and most VM solutions pass the packet straight through the host unchanged, I don't think the host is going to grab the packet and send it straight to another VM unless you configure bridging between VMs... but you'll bind each VM to a different physical network adapter).

Resources