Hello all!
Christian provided me with a linux-mainline-6.14rc4-1-x86_64 package to test with today and I am still experiencing the issue, so I am assuming that revert did not happen yet.
To give some additional context for the performance regression, I have an AT&T 500/500 fiber line. As of this morning on kernel 6.13.1, I get 550 down / 470 up over Wifi 6 on speedtest.net. Testing immediately afterwards with Christian's mainline package, I got a result of 170 down / 75 up, which is similar to the results that I got for all the other bisection kernels Christian gave me to test.
Thanks! Ben
On Friday, February 28th, 2025 at 9:14 AM, Peter Seiderer ps.report@gmx.net wrote:
Hello Christian,
On Fri, 28 Feb 2025 11:19:52 +0100, Christian Heusel christian@heusel.eu wrote:
Hello everyone,
on the Arch Linux Bugtracker1 Benjamin (also added in CC) reported that his MT7925 wifi card has halved it's throughput when updating from the v6.13.1 to the v6.13.2 stable kernel. The problem is still present in the 6.13.5 stable kernel.
We have bisected this issue together and found the backporting of the following commit responsible for this issue:
4cf9f08632c0 ("wifi: mt76: mt7925: Update mt7925_mcu_uni_[tx,rx]_ba for MLO")
Seems there is already a suggested revert of the mentioned commit, see
[PATCH v4 1/6] Revert "wifi: mt76: mt7925: Update mt7925_mcu_uni_[tx,rx]_ba for MLO" https://lore.kernel.org/linux-wireless/20250226025647.102904-1-sean.wang@ker...
Regards, Peter
We unfortunately didn't have a chance to test the mainline releases as the reporter uses the (out of tree) nvidia modules that were not compatible with mainline release at the time of testing. We will soon test against Mainline aswell.
I have attached dmesg outputs of a good and a bad boot aswell as his other hardware specs and will be available to debug this further.
Cheers, Christian