Change Details

Linux sets[1] this to '1', incorrect '0' was damaging rtw89 driver, when it was trying to determine SS number by substracting '1' from rx_nss and passing it to the hardware. This fixes rtw89 driver, still more diffs are needed to make it robust with iperf3 tests and hw crypto, but we're not ready to share those yet. [1] - https://github.com/torvalds/linux/blob/master/net/mac80211/vht.c#L633 Sponsored by: Future Crew LLC

linux80211: fix default deflink.rx_nss Native Linux implementation sets this as a maximum between 1 and ht/vht/eht rx SS'es, FreeBSD does the same, but uses 0 as a minimum, which leads setting it to 0 if we're not in ht/vht case. This 0 was damaging rtw89 driver, when it was trying to determine SS number by subtracting 1 from rx_nss and passing the value to the hardware. After that patch rtw89 association and simple ping work reliably, but more work is needed to make the driver robust with heavy traffic (iperf3) and being long idle. Sponsored by: Future Crew LLC

Linux sets[1] this to '1', incorrect '0' was damaging rtw89 driver, when it was trying to determine SS number by substracting '1' from rx_nss and passing it to the hardware.linux80211: fix default deflink.rx_nss This fixes rtw89 driverNative Linux implementation sets this as a maximum between 1 and ht/vht/eht rx SS'es, still more diffs are needed to make it robust with iperf3 tests and hw cryptoFreeBSD does the same, but we're not ready to share those yetuses 0 as a minimum, which leads setting it to 0 if we're not in ht/vht case. [1] - https://github.com/torvalds/linux/blob/master/net/mac80211/vht.c#L633This 0 was damaging rtw89 driver, when it was trying to determine SS number by subtracting 1 from rx_nss and passing the value to the hardware. After that patch rtw89 association and simple ping work reliably, but more work is needed to make the driver robust with heavy traffic (iperf3) and being long idle. Sponsored by: Future Crew LLC