From mboxrd@z Thu Jan 1 00:00:00 1970 From: "David S. Miller" Subject: Re: Raylink/WebGear testing - ray_cs.c iomem bug? Date: Wed, 6 Oct 2004 10:54:53 -0700 Sender: linux-kernel-owner@vger.kernel.org Message-ID: <20041006105453.5f7d1888.davem@davemloft.net> References: Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: jgarzik@pobox.com, acme@conectiva.com.br, corey@world.std.com, netdev@oss.sgi.com, linux-kernel@vger.kernel.org Return-path: To: Linus Torvalds In-Reply-To: List-Id: netdev.vger.kernel.org On Wed, 6 Oct 2004 10:43:50 -0700 (PDT) Linus Torvalds wrote: > That cleanup in turn seems to show that the driver was fundamentally buggy > in a way that really surprises me: it adds "CCS_BASE" to the PCI window > base in order to get to both the "struct ccs" pointer _and_ to the "struct > rcs" pointer. In the spot where this occurs, it adds both CCS_BASE and 'rcsindex' to the sram base, and only when rcsindex >= NUMBER_OF_CCS. NUMBER_OF_CCS is 64, and the difference between CCS_BASE and RCS_BASE is 0x400 so this really doesn't account for anything. I can't see how you've changed the behavior, so it should work as well as it did before your changes. Sorry, I don't have a ray_cs handy :)