Comment 8 for bug 1913259

Tested in a AN 100Mbps wireless network:

* Killer1650i [8086:43f0] subsystem [1a56:1652]
  * G15-TGL-DVT2-C2
    * 5.10.0-2017.18+lp1913259.1.new.card.so.qu: download 70Mbps, upload 97Mbps
* AX201 [8086:43f0] subsystem [8086:0074]
  * G15-TGL-DVT2-C2
    * 5.10.0-2017.18+lp1913259.1.new.card.so.qu: download 70Mbps, upload 97Mbps
    * 5.10.0-1017: download 80, upload 90
* 9260 [8086:2526] subsystem [8086:0010]
  * G15-TGL-DVT2-C2
    * 5.10.0-2017.18+lp1913259.1.new.card.so.qu: download 80Mbps, upload 1.2Mbps
    * 5.10.0-1017: same poor upload rate
    * 5.8.0-45-generic: same poor upload rate
    * 5.6.0-1050-oem: same poor upload rate
    * 5.4.0-67-generic: download 18Mbps, upload 49.56Mbps
  * KRM3-DVT2-C1
    * 5.10.0-2017.18+lp1913259.1.new.card.so.qu: download 80Mbps, upload 83Mbps
    * 5.10.0-1017: download 89Mbps, upload 1Mbps
* 8265 [8086:24fd] subsystem [8086:0010]
  * KRM3-DVT2-C1
    * 5.10.0-2017.18+lp1913259.1.new.card.so.qu: ok

So I think the most correct interpretation for these results is that 9260 has been broken at upload rate for a while, and it takes a few conditions in both hw and sw to fix. This backport should be good and I'm going to SRU it.