As you can see the problem is not with drivers or kernel. Both interfaces are 2,5Gb. Luci probably shows current status of the link.
1 Like
I'm following this thread now over a month, ordered a an AX6000 beginning of Dec and have it running smoothly without a single issue since a week now.
just wanted to say many thanks for all your efforts you put into this 
5 Likes
There is one big problem. When downloading large files (over 100GB) via SCP or SFTP, after a few minutes the connection is terminated with the error: Incorrect MAC (Message Authentication Code) received on packet. Today I restored the stock firmware and that error does not appear. It is possible that it is a minor configuration error, but the problem is too big for me because I download such files very often.
4 Likes
luci does show current link speed instead of the maximum speed. Ports eth1 (wan) and lan5 are 2.5GbE compatible ports.
Both WAN (eth1) and last port (lan5) are 2.5GbE compatible. luci shows currently negotiated connection speed.
Regarding that eth1 thing, I found this (unrelated to this router or this kernel version):
So it's still possible that DSA is used. And as long as you are using 2.5GbE WAN port for internet, you will get 2.5Gbps in total routed for both up/down speeds (for example symetric 1Gbps internet).
i'd like to test this on my box (even i think i never downloaded a file of that size via scp/sftp), do you have a source that i could retry?
I download backups of my servers to my NAS every day. Unfortunately, I cannot send you that source.
you can try reproduce it with command like below:
execute from local machine to remote router
cat /dev/urandom | ssh router_user@192.168.10.1 "cat > /dev/null"
@whitedd which ssh server have you used?
EDIT: after 200GB transfer is still stable. I use openSSH server on router.
To double-confirm: do I understand correctly that you download large files over SCP/SFTP from some server on the Internet to a NAS, which is in your home network and which is not the same as your router? In other words - I want to confirm that the problem concerns transit traffic, not necessarily traffic terminated by your router.
The problem relates exclusively to traffic that goes through the router. I tried to download to the NAS as well as to the computer, the result is the same. The NAS and the computer are located behind the router. When I return the router to the stock firmware, there are no problems. The same problem occurs when I use transfer via rsync, not only over SCP/SFTP.
1 Like
I found something else. If the transfer is performed via VPN (surfShark), then there is no problem.
1 Like
Are They both ( NAS and server) connected to the LAN ports? Please supply more information about configuration. I'll try to reproduce it.
Good finding!
Can you also check whether the corruption happens with flow offloading off/software/hardware?
Its heppend in any situation...i try all
Another idea that came... if you are using PPPoE, then one known scenario in which data corruption happens without the TCP stack noticing is when Van Jacobson header compression is enabled (which is the case by default in OpenWrt, but all properly configured PPPoE access concentrators should reject it) and packets are reordered. The mitigation is to add the "novj" line into /etc/ppp/options. Can you try that?
Of course, ignore this comment if you are not using PPP.
nope...iam not using PPPoE
How can I reproduce the problem? Connect to wifi, wired LAN ? I cant reproduce it ....
i use wired lan
NAS is connected over Lan and server is remotly
Can you explain how the WAN and the VPN are configured? Ideally, show the relevant parts of your configuration files minus passwords.
2 Likes