netdev - Re: [PATCH net-next 12/13] genetlink: allow families to use split ops directly

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: <20221021080113.6cee1270@kernel.org>
Date:   Fri, 21 Oct 2022 08:01:13 -0700
From:   Jakub Kicinski <kuba@...nel.org>
To:     Johannes Berg <johannes@...solutions.net>
Cc:     davem@...emloft.net, netdev@...r.kernel.org, edumazet@...gle.com,
        pabeni@...hat.com, jiri@...nulli.us, razor@...ckwall.org,
        nicolas.dichtel@...nd.com, gnault@...hat.com,
        jacob.e.keller@...el.com, fw@...len.de
Subject: Re: [PATCH net-next 12/13] genetlink: allow families to use split
 ops directly

On Fri, 21 Oct 2022 13:02:31 +0200 Johannes Berg wrote:
> > Perhaps we could hang it of the .resv_start_op as well?  
> 
> Yes, hopefully? Maybe?
> 
> > Any op past that would treat policy == NULL as reject all?  
> 
> Right. The only danger is that someone already added new stuff somewhere
> and bad/broken userspace already used it with garbage attrs.
> 
> But the chances of that are probably low.
> 
> So I'd say go for it, and worst case we bump up the resv_start_op for
> anything that breaks? Wouldn't be a huge loss either.

resv_start_op are only present in -rc kernels, so I think we can break
things risking only common anger not uAPI wrath :)

> > We'd need to add GENL_DONT_VALIDATE_DO for families which 
> > want to parse inside the callbacks. I wonder if people would
> > get annoyed.  
> 
> Why would anyone really want to _parse_ in the callbacks?

Until recently that was the only way to do per-op policies, I don't
know if anyone actually used per-op policies outside of ethtool tho.