From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from gate.crashing.org (gate.crashing.org [63.228.1.57]) (using TLSv1 with cipher DHE-RSA-AES256-SHA (256/256 bits)) (Client did not present a certificate) by ozlabs.org (Postfix) with ESMTP id 65871DDEC4 for ; Tue, 2 Jan 2007 08:36:13 +1100 (EST) Subject: Re: [PATCH] Fix Maple PATA IRQ assignment. From: Benjamin Herrenschmidt To: Segher Boessenkool In-Reply-To: <32106685eb2ecb2fb4adcf6e51f71482@kernel.crashing.org> References: <1167679875.22068.350.camel@pmac.infradead.org> <1167682254.23340.131.camel@localhost.localdomain> <32106685eb2ecb2fb4adcf6e51f71482@kernel.crashing.org> Content-Type: text/plain Date: Tue, 02 Jan 2007 08:33:50 +1100 Message-Id: <1167687230.23340.147.camel@localhost.localdomain> Mime-Version: 1.0 Cc: linuxppc-dev@ozlabs.org, David Woodhouse , alan@lxorguk.ukuu.org.uk, jgarzik@pobox.com List-Id: Linux on PowerPC Developers Mail List List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , On Mon, 2007-01-01 at 21:44 +0100, Segher Boessenkool wrote: > > I'm not sure you can switch it to native mode in sw... worth double > > checking though. All boards based on 8111 on powerpc have this problem, > > I think. What of the js20 and js21 with IBM firmware ? > > At least with SLOF, we use native mode. > > > Or does it have > > the ide strapped to native mode ? > > It's not a strapping, it's a purely software thing -- see the PCIIDE > specification. I know the PCI IDE spec, but there are a number of chips out there that cannot be changed without a strap. In this case, it's easy to test though we'll need fixups in the device-tree to indicate the change in routing, or a quirk. The thing is, there -is- a cascaded 8259 in there, and that's where the IRQ14 and IRQ15 end up in legacy mode (edge sensitive even) iirc, unless they are aslo routed to the APIC directly, but in anyway it's hidden by the firmware. So we can have a quirk in the Maple code to reconfigure that all in native mode. That will mean a bit more IRQ sharing since both IDE channels will suddenly share PIRQA though... along with whatever else uses that pin. Maybe not the best solution... Or keep the hook and use it in libata. It makes sense I think in general. I reckon other platforms might have good use of being able to "remap" those IRQs 14 and 15 from IDE controllers in legacy mode. I've seen plenty of cases where they are kept in that mode just to keep the IRQs non-shared. > Lots of controllers get the interrupt thing wrong though -- but > AMD8111 isn't one of those. Ben.