netdev - Re: [PATCH net-next v6 1/7] net: ethtool: pass ethtool_rxfh to get/set

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

Message-ID: <20240603161752.70eee7a4@kernel.org>
Date: Mon, 3 Jun 2024 16:17:52 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Edward Cree <ecree.xilinx@...il.com>
Cc: Ahmed Zaki <ahmed.zaki@...el.com>, netdev@...r.kernel.org
Subject: Re: [PATCH net-next v6 1/7] net: ethtool: pass ethtool_rxfh to
 get/set_rxfh ethtool ops

On Mon, 3 Jun 2024 19:54:49 +0100 Edward Cree wrote:
> On 27/11/2023 18:04, Jakub Kicinski wrote:
> > On Mon, 27 Nov 2023 17:10:37 +0000 Edward Cree wrote:  
> >> Yep, I had noticed.  Was wondering how the removal of the old
> >>  [sg]et_rxfh_context functions would interact with my new API,
> >>  which has three ops (create/modify/delete) and thus can't
> >>  really be wedged into the [sg]et_rxfh() like that.  
> > 
> > Set side looks fairly straightforward. Get is indeed more tricky.  
> Looking at this now, and wondering if the create/modify/delete
>  ops should use Ahmed's struct ethtool_rxfh_param, or keep the
>  separated out fields (indir, key, hfunc, and now xfrm) as in
>  my previous versions.
> Arguments for keeping the separate arguments:
> * Ensures that if the API is further extended (e.g. another
>   parameter added, like input_xfrm was) then old drivers will
>   refuse to build until fixed, rather than potentially silently
>   ignoring new members of struct ethtool_rxfh_param.

We add "supported" fields to the ethtool_ops (e.g.
supported_coalesce_params) and reject settings in the core
if the driver didn't opt in.

BTW I have no attachment to the existing cap_rss_* bits, feel free 
to rejig.

> * Avoids potential confusion by driver developers seeing the
>   @indir_size, @key_size, and @rss_delete members of struct
>   ethtool_rxfh_param, which are all superseded by or at least
>   duplicative with parts of the new API

Can we avoid the confusion by careful wording of the related kdoc?
"context" is the current state, while "params" describe the intended
configuration. If we move the "no_change" bits over to "params", 
I hope it wouldn't be all that confusing.

>   (struct ethtool_rxfh_context has @indir_size and @key_size members;
>   the separate 'delete' op in the new API obviates the need
>   for a 'delete' flag).
> However, you presumably had reasons for wanting the arguments
>  wrapped in a struct in the first place, and they may still
>  apply to the new API.  Guidance & comments would be welcome.