From mboxrd@z Thu Jan 1 00:00:00 1970 From: Matthew Garrett Subject: Re: [PATCH v2] Add Qualcomm Gobi 2000 driver. Date: Fri, 8 Oct 2010 19:03:10 +0100 Message-ID: <20101008180310.GA27126@srcf.ucam.org> References: <20101006151208.GB1571@google.com> <20101008.102803.104075511.davem@davemloft.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: ellyjones@google.com, netdev@vger.kernel.org, jglasgow@google.com, msb@google.com, olofj@google.com To: David Miller Return-path: Received: from cavan.codon.org.uk ([93.93.128.6]:48056 "EHLO cavan.codon.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757346Ab0JHSDV (ORCPT ); Fri, 8 Oct 2010 14:03:21 -0400 Content-Disposition: inline In-Reply-To: <20101008.102803.104075511.davem@davemloft.net> Sender: netdev-owner@vger.kernel.org List-ID: On Fri, Oct 08, 2010 at 10:28:03AM -0700, David Miller wrote: > I've looked at the gobi_loader code and it's simpler than many > of the gigabit ethernet driver firmware loader sequences in > the tree already. > > Requiring udev et al. magic only makes this networking device > that much harder to use from an initrd. > > I understand how this might be a bit clumsy since we'd need to make a > dependency on the serial device since that is the mechanism by which > the firmware is uploaded, but really I'd like you to consider it > seriously. The device is capable of functioning in either UMTS or CDMA modes, and requires different firmware for each. Further, there's separate firmware images per carrier due to regulatory differences. To make things even more complicated, there's no consistency in the naming of the firmware (and we can't fix that, because the firmware is undistributable). The only way to do this in-kernel would seem to be to have a module parameter to pass a firmware name, which really doesn't seem elegant. -- Matthew Garrett | mjg59@srcf.ucam.org