On Mon, May 25, 2026 at 07:09:27PM +0200, Kirill A. Korinsky wrote:

> tech@,
> 
> on my debug about 40Mhz width for 11n I had spoted one more differences with
> ath12k which worth to be delivered and tested as small dedicated step: avoid
> hardcoded one RX and TX steram and use actual number of channels at that
> device that matches ath12k behaviour.
> 
> Tested on my honor under ~30 minute load by iperf, no regression.
> 
> Ok?

Makes sense to me.  Mirrors Linux ath12k.  No regression spotted with
the Samsung qwz.

ok mglocker@
 
> Index: sys/dev/ic/qwz.c
> ===================================================================
> RCS file: /home/cvs/src/sys/dev/ic/qwz.c,v
> diff -u -p -r1.35 qwz.c
> --- sys/dev/ic/qwz.c	25 May 2026 08:50:13 -0000	1.35
> +++ sys/dev/ic/qwz.c	25 May 2026 16:07:26 -0000
> @@ -21641,8 +21641,8 @@ qwz_mac_vdev_start_restart(struct qwz_so
>  	arg.channel.max_reg_power = 20; /* XXX */
>  	arg.channel.max_antenna_gain = 0; /* XXX */
>  
> -	arg.pref_tx_streams = 1;
> -	arg.pref_rx_streams = 1;
> +	arg.pref_tx_streams = sc->num_tx_chains;
> +	arg.pref_rx_streams = sc->num_rx_chains;
>  
>  	arg.mbssid_flags = 0;
>  	arg.mbssid_tx_vdev_id = 0;
> 
> 
> -- 
> wbr, Kirill
>