From mboxrd@z Thu Jan 1 00:00:00 1970 From: Arnd Bergmann Subject: Re: [PATCH] PCI: tegra: Use physical range for I/O mapping Date: Wed, 26 Nov 2014 21:30:04 +0100 Message-ID: <2943259.8mh79BAhne@wuerfel> References: <1417018897-3965-1-git-send-email-thierry.reding@gmail.com> <3072281.UjpXkitFjL@wuerfel> <20141126201758.GA19624@mithrandir> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7Bit Return-path: In-Reply-To: <20141126201758.GA19624@mithrandir> Sender: linux-tegra-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Thierry Reding Cc: Bjorn Helgaas , Liviu Dudau , Marc Zyngier , linux-pci-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-tegra-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-Id: linux-tegra@vger.kernel.org On Wednesday 26 November 2014 21:18:03 Thierry Reding wrote: > On Wed, Nov 26, 2014 at 05:55:05PM +0100, Arnd Bergmann wrote: > > On Wednesday 26 November 2014 17:21:37 Thierry Reding wrote: > > > - > > > + pci_add_resource_offset(&sys->resources, &pcie->pio, sys->io_offset); > > > pci_add_resource_offset(&sys->resources, &pcie->mem, sys->mem_offset); > > > pci_add_resource_offset(&sys->resources, &pcie->prefetch, > > > sys->mem_offset); > > > > > > > You don't set sys->io_offset anywhere, which is a bug if you have multiple > > instances of the PCI host in one system. In my draft patch, I was setting > > both io_offset and mem_offset for consistency, and while mem_offset would > > in practice be always zero (as discussed on IRC), the io_offset in fact has > > a realistic chance of being nonzero and you should definitely set it. > > I was going to do that in a follow-up patch since it isn't needed to fix > this particular regression. Makes sense, but then I think you can do the request_resource() and pci_add_resource_offset() additions in the same follow-up patch as well, and only change the computation in the regression fix. Arnd