netdev - Re: [PATCH net-next 2/2] net: pse-pd: pd692x0: Add sysfs interface for configuration save/reset

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: <20250828132901.76e00334@kmaincent-XPS-13-7390>
Date: Thu, 28 Aug 2025 13:29:01 +0200
From: Kory Maincent <kory.maincent@...tlin.com>
To: Jakub Kicinski <kuba@...nel.org>
Cc: Andrew Lunn <andrew@...n.ch>, Oleksij Rempel <o.rempel@...gutronix.de>,
 Andrew Lunn <andrew+netdev@...n.ch>, "David S. Miller"
 <davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>, Paolo Abeni
 <pabeni@...hat.com>, kernel@...gutronix.de, Dent Project
 <dentproject@...uxfoundation.org>, Thomas Petazzoni
 <thomas.petazzoni@...tlin.com>, netdev@...r.kernel.org,
 linux-kernel@...r.kernel.org, Maxime Chevallier
 <maxime.chevallier@...tlin.com>, Kyle Swenson <kyle.swenson@....tech>
Subject: Re: [PATCH net-next 2/2] net: pse-pd: pd692x0: Add sysfs interface
 for configuration save/reset

Le Mon, 25 Aug 2025 15:14:22 -0700,
Jakub Kicinski <kuba@...nel.org> a écrit :

> On Mon, 25 Aug 2025 14:18:25 +0200 Andrew Lunn wrote:
> > On Mon, Aug 25, 2025 at 11:14:03AM +0200, Oleksij Rempel wrote:  
> > > On Mon, Aug 25, 2025 at 10:47:21AM +0200, Kory Maincent wrote:    
>  [...]  
>  [...]  
> > > 
> > > My personal preference would be to use devlink (netlink based)
> > > interface.    
> > 
> > Yes, devlink also crossed my mind, probably devlink params. Although
> > saving the current configuration to non-volatile memory is more a meta
> > parameter.  
> 
> This is a bit of a perennial topic. Most datacenter NIC vendors have 
> a way to save link settings and alike to flash. None bothered with
> adding upstream APIs tho. If the configs are fully within ethtool
> I think we should be able to add an ethtool header flag that says 
> "this config request is to be written to flash". And vice versa 
> (get should read from flash)?

In fact there is the managers power budget parameter taken from the devicetree
which is not in ethtool config. It could be reconfigured after each reboot or
conf reset but it is an example of non ethtool configuration and more could
appear in the future. Talking about perennial, ethtool is then maybe not a good
idea because we still will need a way to save these new global configurations
saved to the non-volatile mem.
I am not really familiar with devlink but indeed after a quick look devlink
seems more suitable for the PSE global configurations.
I don't really know if we should use devlink param and devlink reload or only
devlink param or a new devlink conf.

Or we could save the configuration on every change but it will bring 70ms (I2C
read/write + store waiting time) latency for every command.

> Resetting would work either via devlink reload, or ethtool --reset,
> don't think we even need any API addition there.

Is there no way for NIC to reset their configuration except through ethtool
reload?

PS: got a client mail issue so you might have receive two mails. Sorry for that.
-- 
Köry Maincent, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com