Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S933698AbcJXG50 (ORCPT ); Mon, 24 Oct 2016 02:57:26 -0400 Received: from szxga02-in.huawei.com ([119.145.14.65]:26403 "EHLO szxga02-in.huawei.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S933376AbcJXG5W (ORCPT ); Mon, 24 Oct 2016 02:57:22 -0400 Subject: Re: [PATCH V3 2/2] PCI/ACPI: hisi: Add ACPI support for HiSilicon SoCs Host Controllers To: Lorenzo Pieralisi References: <1476933034-46044-1-git-send-email-liudongdong3@huawei.com> <1476933034-46044-3-git-send-email-liudongdong3@huawei.com> <5809B1DC.2070605@huawei.com> <20161021160838.GA29335@red-moon> CC: "Rafael J. Wysocki" , Bjorn Helgaas , Arnd Bergmann , Tomasz Nowicki , , , Linux PCI , ACPI Devel Maling List , Linux Kernel Mailing List , Jon Masters , , , Hanjun Guo , From: Dongdong Liu Message-ID: <580DB0A5.20406@huawei.com> Date: Mon, 24 Oct 2016 14:56:37 +0800 User-Agent: Mozilla/5.0 (Windows NT 6.1; rv:38.0) Gecko/20100101 Thunderbird/38.5.0 MIME-Version: 1.0 In-Reply-To: <20161021160838.GA29335@red-moon> Content-Type: text/plain; charset="UTF-8"; format=flowed Content-Transfer-Encoding: 8bit X-Originating-IP: [10.61.21.156] X-CFilter-Loop: Reflected Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 4091 Lines: 129 Hi Lorenzo Many thanks for your review. 在 2016/10/22 0:08, Lorenzo Pieralisi 写道: > On Fri, Oct 21, 2016 at 02:12:44PM +0800, Dongdong Liu wrote: > > [...] > >>>> +static int hisi_pcie_init(struct pci_config_window *cfg) >>>> +{ >>>> + int ret; >>>> + struct acpi_device *adev = to_acpi_device(cfg->parent); >>> >>> Why is this expected to be struct acpi_device? >> >> I use this acpi device(Device (PCI2)) to get it's child acpi device(Device (RES2)), then >> obtain rc base addresses from PNP0C02 as subdevice of PNP0A03. >> >> The procedure to get the value of cfg->parent is as the below. >> arch/arm64/kernel/pci.c >> pci_acpi_scan_root(struct acpi_pci_root *root) >> --->pci_acpi_setup_ecam_mapping(root, ri); >> --->pci_ecam_create(&root->device->dev, &cfgres, bus_res, ecam_ops); >> --->cfg->parent = dev >> ; >> PCIe DSDT table defines as below. >> Device (PCI2) >> { >> Name (_HID, "PNP0A08") // PCI Express Root Bridge >> Name (_CID, "PNP0A03") // Compatible PCI Root Bridge >> ...... >> Device (RES2) >> { >> Name (_HID, "HISI0081") // HiSi PCIe RC config base address >> Name (_CID, "PNP0C02") // Motherboard reserved resource >> Name (_CRS, ResourceTemplate (){ >> Memory32Fixed (ReadWrite, 0xa00a0000 , 0x10000) >> }) >> ...... >> } >>> >>> Shouldn't it be a "physical" device whose companion is an ACPI device object? >> >> Sorry, I don't understand. What do you mean ? > > I think Rafael meant the physical node (that in this case is the pci > bridge device) that is associated with the acpi device (the struct > pci_config_window.parent pointer points at the PNP0A03/PNP0A08 acpi > device object though, not the RES2 acpi device that's what you need). > > Have a look at: > > Documentation/acpi/enumeration.txt > Documentation/acpi/namespace.txt Thanks for explaining this. > >> It should be a "physical" device. but I have not saw about >> ACPI_COMPANION_SET() of this device. > > The PNP0A08 acpi_device (that is what is pointed at by > struct acpi_device = to_acpi_device(pci_config_window.parent) > is the respective pci bridge device companion. > > arch/arm64/kernel/pci.c (pcibios_root_bridge_prepare()). > > Now for something completely different :), your RES2 pseudo-device. > > IIUC platform device (physical node) is created by the ACPI enumeration > code for your RES2 pseudo-device and its resources are already > initialized (by parsing its _CRS) in acpi_create_platform_device(). This is not right. About RES2 pseudo-device,it's _CID is PNP0C02, this will attach acpi_pnp_attach() (drivers/acpi/acpi_pnp.c) . acpi_pnp_attach() return 1;ret = acpi_scan_attach_handler() will return 1. So this will not call acpi_default_enumeration(), this also will not call acpi_create_platform_device(). drivers/acpi/scan.c static void acpi_bus_attach(struct acpi_device *device) { ...... ret = acpi_scan_attach_handler(device); if (ret < 0) return; device->flags.match_driver = true; if (!ret) { ret = device_attach(&device->dev); if (ret < 0) return; if (!ret && device->pnp.type.platform_id) acpi_default_enumeration(device); } ...... } Dongdong Thanks > > This means that by the time you match the PNP0A03/PNP0A08 child with an > acpi_device with _HID == "HISI0081", you can get its associated > physical node and get its resources IOMEM through the physical > node (ie platform device and related resources) instead of re-parsing > the _CRS if I am not mistaken (and that's an IF because I did not > test it). As above said this does not re-parse the _CRS for RES2 pseudo-device. Dongdong Thanks > > Regardless, I am not entirely sure there are kernel drivers/control > paths already using this mechanism to handle child devices (keeping in > mind that RES2 is not a real device at all, it is there to represent > PNP0A03 sub-address space for PCI quirks), so it would be good to get > Rafael's agreement on this approach to prevent abusing the current ACPI > platform devices glue code assumptions. > > It would also be good to agree on the whole approach soundness. > > Thanks, > Lorenzo > > . >