Received-SPF: pass (google.com: domain of linux-wireless-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <096c975e7287cbe22fd355bcfd4b3889c890bdc7.camel@sipsolutions.net>
Subject: Re: [PATCH 1/9] wifi: mac80211: allow enabling chanctx until hw
 registration
From:   Johannes Berg <johannes@sipsolutions.net>
To:     Sean Wang <sean.wang@kernel.org>
Cc:     Sean Wang =?UTF-8?Q?=28=E7=8E=8B=E5=BF=97=E4=BA=98=29?= 
        <sean.wang@mediatek.com>, Felix Fietkau <nbd@nbd.name>,
        lorenzo.bianconi@redhat.com, Soul.Huang@mediatek.com,
        YN.Chen@mediatek.com, Leon.Yen@mediatek.com,
        Eric-SY.Chang@mediatek.com, Deren Wu <Deren.Wu@mediatek.com>,
        km.lin@mediatek.com, jenhao.yang@mediatek.com,
        robin.chiu@mediatek.com, Eddie.Chen@mediatek.com,
        ch.yeh@mediatek.com, posh.sun@mediatek.com, ted.huang@mediatek.com,
        Stella.Chang@mediatek.com, Tom.Chou@mediatek.com,
        steve.lee@mediatek.com, jsiuda@google.com, frankgor@google.com,
        kuabhs@google.com, druth@google.com, abhishekpandit@google.com,
        shawnku@google.com, linux-wireless@vger.kernel.org,
        "moderated list:ARM/Mediatek SoC support" 
        <linux-mediatek@lists.infradead.org>
Date:   Fri, 19 Aug 2022 19:16:58 +0200
In-Reply-To: <CAGp9LzqmFKs8wp-g7WsJ=uUj_OqucQskdMDwM6yf2dZyuP0ABA@mail.gmail.com>
References: <cover.1660606893.git.objelf@gmail.com>
         <473fc7b169f288b7815a7736cf33ac0ec1599a09.1660606893.git.objelf@gmail.com>
         <edd45a1e732a0eef1894b9f6eda2d4e3c692359b.camel@sipsolutions.net>
         <CAGp9LzrqYsXAPsbZvTy-_3Tnu23oc11KHdxNs9WvYfe0JHQaDg@mail.gmail.com>
         <5b24421363048bff1a9f03174cb0223b3e226bf9.camel@sipsolutions.net>
         <CAGp9Lzrag1b3G+QfW0RCQPGLnR_WzGSg4k+Zdx4zi4b_oeunRQ@mail.gmail.com>
         <dc45daf23fc8084596f1da8cb41afb422b24160d.camel@sipsolutions.net>
         <CAGp9LzqmFKs8wp-g7WsJ=uUj_OqucQskdMDwM6yf2dZyuP0ABA@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
User-Agent: Evolution 3.44.4 (3.44.4-1.fc36) 
MIME-Version: 1.0
Precedence: bulk

Hi Sean,

> > If we really, really need to allow changing the ops, then we should
> > probably make a much larger change to not even pass the ops until
> > register, though I'm not really sure it's worth it just to have mt7921
> > avoid loading the firmware from disk before allocation?

> Thanks for your input. I thought I'd try to write a patch to follow up
> on the idea you mentioned here.
>=20

I think you will introduce a bug into mt7921 when you do this, and I'm
curious if you will find it ;-)

Seriously though, this approach also seems fragile, and I don't even
know if other bugs would be introduced. And splitting into three
functions (alloc -> set_ops -> register) also feels a bit awkward.


Is there really no chance you could add bits to the firmware _file_
format so you can query the capabilities before you actually _run_ the
firmware? I guess you could even validate it at runtime again (and just
fail is somebody messed up the file), but it would make things a lot
simpler, I'm sure.

johannes