Re: [RFC PATCH 2/5] can: slcan: remove legacy infrastructure - Marc Kleine-Budde

From: Marc Kleine-Budde <mkl@pengutronix.de>
To: Max Staudt <max@enpas.org>
Cc: Oliver Hartkopp <socketcan@hartkopp.net>,
	Dario Binacchi <dario.binacchi@amarulasolutions.com>,
	linux-kernel@vger.kernel.org,
	Jeroen Hofstee <jhofstee@victronenergy.com>,
	michael@amarulasolutions.com,
	Amarula patchwork <linux-amarula@amarulasolutions.com>,
	"David S. Miller" <davem@davemloft.net>,
	Eric Dumazet <edumazet@google.com>,
	Greg Kroah-Hartman <gregkh@linuxfoundation.org>,
	Jakub Kicinski <kuba@kernel.org>,
	Jiri Slaby <jirislaby@kernel.org>,
	Paolo Abeni <pabeni@redhat.com>,
	Vincent Mailhol <mailhol.vincent@wanadoo.fr>,
	Wolfgang Grandegger <wg@grandegger.com>,
	linux-can@vger.kernel.org, netdev@vger.kernel.org
Subject: Re: [RFC PATCH 2/5] can: slcan: remove legacy infrastructure
Date: Tue, 19 Jul 2022 21:59:27 +0200	[thread overview]
Message-ID: <87cze0swkm.fsf@hardanger.blackshift.org> (raw)
In-Reply-To: <20220719030326.49f204f6.max@enpas.org>

[-- Attachment #1: Type: text/plain, Size: 1985 bytes --]

On 19.07.2022 03:03:26, Max Staudt wrote:
> On Mon, 18 Jul 2022 12:23:05 +0200
> Oliver Hartkopp <socketcan@hartkopp.net> wrote:
> 
> > IMO it does not break user space when slcan gets the common naming 
> > schema for CAN interface names.
> 
> For what it's worth, slcan provides a specific ioctl to resolve from
> (tty fd) to (netdev name). As far as I can see, any sane program should
> use this API to find whatever device it has created, irrespective of
> the netdev name.
> 
> 
> @ Marc: slcand and slcan_attach in can-utils do not depend on the
> slcanX name, as far as I can see. They are the main interfaces to
> slcan, and apart from that I can only think of scripts doing netlink
> stuff and scanning for slcanX. Does this change your opinion, or is
> breaking such scripts already a step too far?
> 
> I'm thinking that in other circumstances, I've had cases where devices
> were numbered differently just because they enumerated faster or slower
> at different boots, or with other kernel versions, or because of media
> inserted, etc. - though renumbering is of course one step less than
> changing the prefix.

Convinced! Please change the patch to use standard canX interface
naming. Can you add a pointer to the ioctl() to get the network
interface name to the tty fd in the patch description.

> > We had the same thing with 'eth0' which is now named enblablabla or 
> > 'wlan0' now named wlp2s0.
> 
> Actually, the kernel still calls them ethX and wlanX. It's systemd+udev
> that renames them to the "unique" names afterwards. It's an extra step
> that happens in userspace.
> 
> udev is immensely useful, but also great stuff for race conditions :)

Marc

-- 
Pengutronix e.K.                 | Marc Kleine-Budde           |
Embedded Linux                   | https://www.pengutronix.de  |
Vertretung West/Dortmund         | Phone: +49-231-2826-924     |
Amtsgericht Hildesheim, HRA 2686 | Fax:   +49-5121-206917-5555 |

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 488 bytes --]