From owner-freebsd-current Tue Jan 5 10:08:56 1999 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id KAA14349 for freebsd-current-outgoing; Tue, 5 Jan 1999 10:08:56 -0800 (PST) (envelope-from owner-freebsd-current@FreeBSD.ORG) Received: from nlsystems.com (nlsys.demon.co.uk [158.152.125.33]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id KAA14308 for ; Tue, 5 Jan 1999 10:08:44 -0800 (PST) (envelope-from dfr@nlsystems.com) Received: from localhost (dfr@localhost) by nlsystems.com (8.9.1/8.8.5) with SMTP id SAA12610; Tue, 5 Jan 1999 18:08:19 GMT Date: Tue, 5 Jan 1999 18:08:19 +0000 (GMT) From: Doug Rabson To: Matthew Jacob cc: Garrett Wollman , Warner Losh , Bill Paul , current@FreeBSD.ORG Subject: Re: bus_space_foo and bus_dmamap_foo In-Reply-To: Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-current@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG On Tue, 5 Jan 1999, Matthew Jacob wrote: > > > > How do you view this integration? I'm confused... What needs to be > > > > integrated? > > > > > > When you activate a memory or I/O resource, the appropriate magic > > > should happen such that included in the stuff you get back is > > > something which can be used to access that resource quickly. Stuff > > > like bus_space_alloc should be handled by the bus driver in > > > coordination with the resource manager. (It should be implemented as > > > a method call on the device's parent, so that it has all the necessary > > > context pre-computed.) > > > > I think that this is the right approach. The interface for PCI busses > > should have methods to manage the various device resources and these will > > be able to provide bus_space handles as necessary. The resources should > > not need to be mapped directly by the pci driver. I'm currently > > prototyping the new PCI interface (or I will be as soon as I can get that > > wretched sysinstall program working properly on the alpha) and I'll add > > something like this. > > > > Sounds like the Solaris DKI/DDI. Will you be making the bus_space handles > hierarchical and bidirectional? That is, you can do bus_space_XXX for > a CPU's access to i/o or memory space on a PCI device, but can you also > see about a PCI device's view of memory (or another PCI device) such that > dma handles are managed similarily? I'm not sure that this fits with the existing bus_dma apis which we have taken from NetBSD. I wasn't planning to change that api much if at all. It does make sense for the two to be at least similar though. -- Doug Rabson Mail: dfr@nlsystems.com Nonlinear Systems Ltd. Phone: +44 181 442 9037 To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message