Ysurac / openmptcprouter

OpenMPTCProuter is an open source solution to aggregate multiple internet connections using Multipath TCP (MPTCP) on OpenWrt
https://www.openmptcprouter.com/
GNU General Public License v3.0
1.84k stars 265 forks source link

omr-bypass domains ignored for AAAA records #2714

Closed kevinh-csalabs closed 1 year ago

kevinh-csalabs commented 1 year ago

Expected Behavior

Adding a domain to omr-bypass should use the specific interface and bypass vps

Current Behavior

Currently this works if the domain uses A records but AAAA records still go through VPS

Possible Solution

Steps to Reproduce the Problem

  1. Add adafruit.com to omr-bypass and set it to use the default master route. enable this rule and save/apply
  2. run traceroute -4 to adafruit.com and note that it uses the master interface
  3. run traceroute -6 to adafruit.com and note that it goes through the vps

Context (Environment)

Specifications

Kalimeiro commented 1 year ago

i'm agree, but :

for mptcp-next, full support IPv4 and IPv6 on mptcp is planned with 6.3 kernel.

for mptcp v0 : i dont know.

Ysurac commented 1 year ago

I'm working on ignoring AAAA request via option in omr-bypass (all code is done, only need to test and commit) MPTCP work on IPv4 and IPv6 but in separate way: IPv4 WANs are aggregated and IPv6 WANs are aggregated, not a mix of both, not supported on MPTCP for now and not sure this will ever be supported.

Kalimeiro commented 1 year ago

I'm working on ignoring AAAA request via option in omr-bypass (all code is done, only need to test and commit) MPTCP work on IPv4 and IPv6 but in separate way: IPv4 WANs are aggregated and IPv6 WANs are aggregated, not a mix of both, not supported on MPTCP for now and not sure this will ever be supported.

Only with MPTCP-NEXT :

From matttbe on his WIP 6.3 kernel :

" Allow having a mix of v4/v6 subflows for the same socket

It is important to support MPTCP connections with some subflows in v6 and others in v4. But so far, MPTCP does not yet support mixed IPv6 and IPv4 endpoints for the same socket.

Some use-cases:

Using IPv4 and IPv6 only networks
Some paths behind different in v4 and v6
Having only one links with both v4 and v6 where paths to reach the end server are different or using one or the other as a failover mechanism.
etc.

"

kevinh-csalabs commented 1 year ago

Thanks - this will be very helpful as I’m finding the ipv4 handling on some cellular networks such as t-mobile is abysmal (much higher latency, packet loss) due to their tunneling.

Regards, Kevin Hart (424) 332-7507


From: Maxime R. @.> Sent: Monday, January 2, 2023 9:55:38 AM To: Ysurac/openmptcprouter @.> Cc: Kevin Hart @.>; Author @.> Subject: Re: [Ysurac/openmptcprouter] omr-bypass domains ignored for AAAA records (Issue #2714)

I'm working on ignoring AAAA request via option in omr-bypass (all code is done, only need to test and commit) MPTCP work on IPv4 and IPv6 but in separate way: IPv4 WANs are aggregated and IPv6 WANs are aggregated, not a mix of both, not supported on MPTCP for now and not sure this will ever be supported.

Only with MPTCP-NEXT :

From matttbehttps://github.com/matttbe on his WIP 6.3 kernel :

" Allow having a mix of v4/v6 subflows for the same socket #269https://github.com/Ysurac/openmptcprouter/issues/269

It is important to support MPTCP connections with some subflows in v6 and others in v4. But so far, MPTCP does not yet support mixed IPv6 and IPv4 endpoints for the same socket.

Some use-cases:

Using IPv4 and IPv6 only networks Some paths behind different in v4 and v6 Having only one links with both v4 and v6 where paths to reach the end server are different or using one or the other as a failover mechanism. etc.

"

— Reply to this email directly, view it on GitHubhttps://github.com/Ysurac/openmptcprouter/issues/2714#issuecomment-1369010150, or unsubscribehttps://github.com/notifications/unsubscribe-auth/ARHBLPMJDFNGO2QATXRMUT3WQLT6VANCNFSM6AAAAAATM722YI. You are receiving this because you authored the thread.Message ID: @.***>

github-actions[bot] commented 1 year ago

This issue is stale because it has been open 90 days with no activity. Remove stale label or comment or this will be closed in 5 days