Re: please update -head if you're running it - testing changes to NULL frames and sequence number offload

From: Adrian Chadd <adrian_at_freebsd.org>
Date: Sun, 21 Sep 2025 01:18:53 UTC
On Sat, 20 Sept 2025 at 17:26, Bjoern A. Zeeb <bz@freebsd.org> wrote:

> On Sat, 20 Sep 2025, Adrian Chadd wrote:
>
> > On Sat, 20 Sept 2025 at 16:10, Nuno Teixeira <eduardo@freebsd.org>
> wrote:
> >
> >> 1. git rebase -i and drop revert
> >> 2. apply patch
> >>
> >> OK, iperf several times without any prob.
> >>
> >
> > Interesting. ok. I'll go line this up in my diff stack and get it
> reviewed.
> > I wonder what the difference is between AX201 and AX210 firmware!
>
> The AX201 is a 22000 chipset not an ax210.
> But beware that AX210 != AX210 in some cases too!
>
> man iwlwififw can give you some hints ;-)
>

Interesting, ok. I can't see where the iwlwifi mvm code is doing anything
differently here.
It looks like  iwl_mvm_set_tx_cmd() sets/clears TX_CMD_FLG_SEQ_CTL based on
some
logic and whether mac80211 has specified a sequence number to use.

So from my reading it actually does look like we need to at least set
sequence numbers
if required, but leave the SEQNO_ASSIGN set so we don't involve that global
lock I'm
trying to get rid of.

Also, the iwx TX path isn't setting all of the TX flags anyway that
iwlwifi/mvm is setting.
I'll worry about that later. Much later. :-)

Thanks,



-adrian



>
> --
> Bjoern A. Zeeb                                                     r15:7
>