[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Y4UslxhfRPVGXzS/@mars>
Date: Mon, 28 Nov 2022 22:48:07 +0100
From: Christoph Fritz <christoph.fritz@...dev.de>
To: Ryan Edwards <ryan.edwards@...il.com>,
Oliver Hartkopp <socketcan@...tkopp.net>,
Pavel Pisa <pisa@....felk.cvut.cz>,
Andreas Lauser <andreas.lauser@...tion.io>,
Richard Weinberger <richard@....at>,
Wolfgang Grandegger <wg@...ndegger.com>,
Marc Kleine-Budde <mkl@...gutronix.de>,
"David S . Miller" <davem@...emloft.net>,
Jakub Kicinski <kuba@...nel.org>,
Eric Dumazet <edumazet@...gle.com>,
Paolo Abeni <pabeni@...hat.com>
Cc: Jonathan Corbet <corbet@....net>, linux-can@...r.kernel.org,
netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [RFC][PATCH 0/2] LIN support for Linux
Thanks for all the participation and feedback. Here is my attempt at a
summary of what we have discussed so far:
- Common goal: A solid LIN UAPI and API
- LIN fits CAN well and could be embedded into Linux CAN infrastructure
- LIN support cannot be a tty-line-discipline driver for all devices
out there
- slLIN should become a user of this API
- slLIN itself needs some more options in the line-discipline
configuraion (to tweak UART FIFO settings and e.g. enable
LIN-break-detection for supported UARTs) _or_ tty-LIN support
becomes something more like RS485 and get integrated into
tty-drivers directly.
- LIN devices with off loading capabilities are a bit special.
- one approach is to have a kfifo for the slavetable (64 entries a 8
bytes + checksum and some extra flags for some LIN special cases)
while updating it from userland through CAN or a simple sysfs
interface
- when we agree that "dumb" UARTs have no need for an
in-kernel-off-load slavetable (because of RT-Linux), then there
are only devices left (e.g. some USB adapters) maintaining their
own table, so a simple e.g. sysfs update mechanism would be
enough (without the need for an in-kernel kfifo buffer).
- LIN slavetable might need another name
- LIN needs rx, tx, set bitrate, set checksum variant and maybe update
a slavetable (or whatever it is called then...)
What do you think? Any objections, corrections, enhancements?
My question is: Which approach of an API is favored: Patch 1 or 2 of
this RFC series or something completely different?
Thanks
-- Christoph
Powered by blists - more mailing lists