From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnd Bergmann Subject: Re: [PATCH] serial: relocate remaining serial drivers from tty/ to tty/serial/ Date: Wed, 4 Jan 2012 23:27:33 +0000 Message-ID: <201201042327.33739.arnd@arndb.de> References: <1325707287-31038-1-git-send-email-paul.gortmaker@windriver.com> <4F04BFA8.9010100@windriver.com> <20120104220332.GA22985@suse.de> Mime-Version: 1.0 Content-Type: Text/Plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Return-path: Received: from moutng.kundenserver.de ([212.227.17.10]:53365 "EHLO moutng.kundenserver.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932225Ab2ADX2M (ORCPT ); Wed, 4 Jan 2012 18:28:12 -0500 In-Reply-To: <20120104220332.GA22985@suse.de> Sender: linux-serial-owner@vger.kernel.org List-Id: linux-serial@vger.kernel.org To: Greg KH Cc: Paul Gortmaker , Jiri Slaby , alan@linux.intel.com, linux-serial@vger.kernel.org, linux-kernel@vger.kernel.org, Jiri Slaby On Wednesday 04 January 2012, Greg KH wrote: > > > > My thinking was that having a drivers/tty/serial dir and then > > not having all the serial drivers in that dir violated the > > principle of least surprise. Is there a reason why the dir should > > be the exclusive domain of drivers with a dependency on SERIAL_CORE? > > Because that is what the directory is for? :) > > We have other "serial" like drivers all over the kernel, this was for > the SERIAL_CORE drivers only at the moment. My initial plan when moving some files to drivers/tty was to have a separate directory for the non-SERIAL_CORE serial drivers next to drivers/tty/serial. I would still prefer this solution, but I think we never agreed on a good name for that directory. IIRC, I had suggested drivers/tty/legacy believing that SERIAL_CORE was the modern way to implement a serial driver, but that turned out not to be true and at lease one of these (bfin_jtag) is not a legacy driver in practice. Maybe drivers/tty/hw? I think that one has been suggested before, too. I don't remember any argument against it and I think it would be nice to separate the core implementation from actual device drivers. Arnd