netdev - Re: [PATCH v1 wl-next 1/3] wifi: wext: Move wext

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <a4e27207087bc94ce120a1e1c54742f97c9ad046.camel@sipsolutions.net>
Date: Thu, 17 Oct 2024 10:06:06 +0200
From: Johannes Berg <johannes@...solutions.net>
To: Kuniyuki Iwashima <kuniyu@...zon.com>
Cc: alexandre.ferrieux@...il.com, kuni1840@...il.com, 
	linux-wireless@...r.kernel.org, netdev@...r.kernel.org
Subject: Re: [PATCH v1 wl-next 1/3] wifi: wext: Move wext_nlevents to
 net->gen[].

On Wed, 2024-10-16 at 16:58 -0700, Kuniyuki Iwashima wrote:
> 
> Btw, why WEXT cannot be module ?

TBH, I don't remember well. I feel like I may have tried ~20 years ago,
but hit issues, and just made the built-in parts minimal. Might've been
we didn't have net->gen yet (did we? I don't recall), but I wouldn't be
surprised if there are other issues with it as well with ioctl linkage
and /proc and whatever else it does.

Not sure it's worth trying, WEXT really ought to be on the way out now,
and with WiFi7 (and higher) devices it's completely disabled.

Btw, if you really wanted to, I suspect you _could_ use net->gen[], make
the .size only a pointer size and then allocate the real data only if a
wireless capable device shows up in the namespace? Then that'd actually
be a win (vs. the other discussion we just had above) since wireless
devices are probably almost never in a netns. Not sure you'd be able to
easily free it when the last wifi capable devices leaves a netns, but
that probably also doesn't matter.

I don't know though how much the size of the netns matters for the
scalability issue you have in mind, seems the O(N) time behaviour here
is more problematic than a handful of bytes.

johannes