netdev - Re: Rx issues with Linux Bridge and thunderbolt-net

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Fri, 8 Dec 2023 10:34:25 -0700
From: Arjun Mehta <arjunmeht@...il.com>
To: Stephen Hemminger <stephen@...workplumber.org>
Cc: netdev@...r.kernel.org
Subject: Re: Rx issues with Linux Bridge and thunderbolt-net

Also I want to clarify that the issue. You state "Most likely this is a hardware issue on the thunderbolt interface where it will not allow sending with a different source MAC address."

Actually the issue is that the interface will not _receive_. Transmission to another device from the interface works at full speed. Receiving is not working as expected.


> On Dec 8, 2023, at 10:30 AM, Arjun Mehta <arjunmeht@...il.com> wrote:
> 
> Hi Stephen, thank you for the reply.
> 
> Proxmox does use a kernel derived from Ubuntu I believe (eg. kernel for Proxmox 8.1 which is what I'm using is 6.5.11-6-pve derived from Ubuntu 23.10). Not sure if there have been any modifications to the Linux Bridge in it.
> https://pve.proxmox.com/wiki/Proxmox_VE_Kernel#Proxmox_VE_8.x
> 
> Long shot, but do you happen to know of any workarounds with the Wifi interface issue you mentioned that would mitigate this issue? Maybe they would apply here.
> 
> I will also post to the Proxmox forums about this issue to flag for them.
> 
> Arjun
> 
> 
>> On Dec 7, 2023, at 3:37 PM, Stephen Hemminger <stephen@...workplumber.org> wrote:
>> 
>> On Thu, 7 Dec 2023 12:57:08 -0700
>> Arjun Mehta <arjunmeht@...il.com> wrote:
>> 
>>> Hi there, I’d like to report what I believe to be a bug with either Linux Bridge (maybe and/or thunderbolt-net as well).
>>> 
>>> Problem: Rx on bridged Thunderbolt interface are blocked
>>> 
>>> Reported Behavior:
>>> Tested on Proxmox host via iperf3, between B550 Vision D-P and MacBook Pro (2019 intel). On a direct interface, thunderbolt bridge Tx and Rx speeds are equal and full speed (in my case 9GB/s each). However, when a thunderbolt bridge is passed through via Linux Bridge to a VM or container (in my case a Proxmox LXC container or VM) the bridge achieves full Tx speeds, but Rx speeds are reporting limited to ~30kb/s
>>> 
>>> Expected:
>>> The VM/CT should have the same general performance for Tx AND Rx as the host
>>> 
>>> Reproducing:
>>> - Setup for the bridge was done by following this guide: https://gist.github.com/scyto/67fdc9a517faefa68f730f82d7fa3570
>>> - Both devices on Thunderbolt interfaces have static IPs
>>> - VM is given the same IP, but unique MAC address
>>> - BIOS has Thunderbolt security mode set to “No security”
>>> 
>>> Further reading:
>>> The problem is outlined more with screenshots and further details in this Reddit post: https://www.reddit.com/r/Proxmox/comments/17kq5st/slow_rx_speed_from_thunderbolt_3_port_to_vm_over/.
>>> 
>>> Please let me know if there is any further action I can do to help investigate or where else I can direct the bug/concern
>> 
>> Most likely this is a hardware issue on the thunderbolt interface where it will not
>> allow sending with a different source MAC address.  Some Wifi interfaces have this
>> problem.
>> 
>> Is Promox using a kernel from upstream Linux repository directly.
>> Netdev developers are unwilling to assist if there are any non-upstream kernel modules in use.
> 
>