From: ebiederm@xmission.com (Eric W. Biederman) Subject: Re: [PATCH v2 2/2] Intel-IOMMU, intr-remap: source-id checking Date: Thu, 21 May 2009 03:04:11 -0700 Message-ID: References: <1242757912-6041-1-git-send-email-weidong.han@intel.com> <1242757912-6041-3-git-send-email-weidong.han@intel.com> <20090519115055.GB14305@elte.hu> <20090520082114.GC6736@elte.hu> <715D42877B251141A38726ABF5CABF2C0545690223@pdsmsx503.ccr.corp.intel.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: "'Ingo Molnar'" , "'Yinghai Lu'" , "'Joerg Roedel'" , "'dwmw2\@infradead.org'" , "Siddha\, Suresh B" , "'linux-kernel\@vger.kernel.org'" , "'iommu\@lists.linux-foundation.org'" , "'kvm\@vger.kernel.org'" To: "Han\, Weidong" Return-path: Received: from out02.mta.xmission.com ([166.70.13.232]:54915 "EHLO out02.mta.xmission.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752277AbZEUKET (ORCPT ); Thu, 21 May 2009 06:04:19 -0400 In-Reply-To: <715D42877B251141A38726ABF5CABF2C0545690223@pdsmsx503.ccr.corp.intel.com> (Weidong Han's message of "Thu\, 21 May 2009 17\:00\:34 +0800") Sender: kvm-owner@vger.kernel.org List-ID: "Han, Weidong" writes: > Ingo Molnar wrote: >> * Eric W. Biederman wrote: >> >>> Not finding an upstream pcie_bridge and then concluding we are a >>> pcie device seems bogus. >>> > > If device is pcie, pci_find_upstream_pcie_bridge() will return > NULL. For root complex integrated device, it won't find upstream > bridge, and also return NULL. What's more, pcie device and root > complex integrated device will be handled as the same way to set > sid, so I mix them here. But it also returns NULL for busted > hardware. I think no parent bus can be considered Root Complex > integrated device, right? If so, I think can handle it as follows: I have problems with pci_find_upstream_pcie_bridge. The name is actively misleading about what that function does. Returning a pci_to_pci bridge is strongly counter intuitive give that name. Can we change it to pci_find_upstream_pcie_to_pci_bridge. Returning NULL in all cases when there is not an upstream pcie_to_pci bridge. For the set_sid case that is ideal. For the other cases in intel-iommu.c it may be a problem. Is it even possible to have a genuine pci device not behind a pcie to pci bridge on an intel chipset with this iommu? > ... > if (dev->is_pcie || !dev->bus->parent) { > set_irte_sid(...); > return 0; > } > > bridge = pci_find_upstream_pcie_bridge(dev); > if (bridge) { > if (bridge->is_pcie) /* PCIE-to-PCI/PCIx bridge */ > set_irte_sid(...); > else /* legacy PCI bridge */ > set_irte_sid(..); > } > > return 0; > >>> Why if we do have an upstream pcie bridge do we only want to do a >>> bus range verification instead of checking just for the bus >>>> devfn? >>> >>> The legacy PCI case seems even stranger. > > Why? If a PCI device isn't connected to a PCIe bridge, it should be a legacy bridge. I am not deep in the IOV specification at the moment. I am mostly wondering why we pick the parts we pick to verify. I recall bus and devfn being on the pcie packets so that makes sense. Why would we ever want to do something different? Does a pcie to pci bridge do something different in it's translation? Eric