Re: [PATCH V7 07/11] pci, acpi: Handle ACPI companion assignment.

[Tomasz Nowicki <tn@semihalf.com>]

On 17.05.2016 05:11, Dongdong Liu wrote:
> Hi Tomasz
>
> I used the patchset and added "PATCH V6 11/13 specic quirks", tested on
> HiSilicon D02 board but met the below problem.
>
> [    2.614115] [<ffffff80083b13bc>] hisi_pcie_init+0x6c/0x1ec
> [    2.619571] [<ffffff80083ab060>] pci_ecam_create+0x130/0x1ec
> [    2.625209] [<ffffff80083f3764>] pci_acpi_scan_root+0x160/0x218
> [    2.631096] [<ffffff80083d1f6c>] acpi_pci_root_add+0x36c/0x42c
> [    2.636897] [<ffffff80083ce36c>] acpi_bus_attach+0xe4/0x1a8
> [    2.642438] [<ffffff80083ce3d8>] acpi_bus_attach+0x150/0x1a8
> [    2.648066] [<ffffff80083ce3d8>] acpi_bus_attach+0x150/0x1a8
> [    2.653693] [<ffffff80083ce55c>] acpi_bus_scan+0x64/0x74
> [    2.658975] [<ffffff8008ae665c>] acpi_scan_init+0x5c/0x19c
> [    2.664429] [<ffffff8008ae6408>] acpi_init+0x280/0x2a4
> [    2.669538] [<ffffff80080829e8>] do_one_initcall+0x8c/0x19c
> [    2.675080] [<ffffff8008ac3af8>] kernel_init_freeable+0x14c/0x1ec
> [    2.681139] [<ffffff80087a8438>] kernel_init+0x10/0xfc
> [    2.686248] [<ffffff8008085e10>] ret_from_fork+0x10/0x40
>
> In hisi_pcie_init, I used "struct acpi_device *device =
> ACPI_COMPANION(dev);".
> I found the reason is V7 lack the below code. I added the below code, it
> worked ok.
>
> [PATCH V6 01/13] pci, acpi, x86, ia64: Move ACPI host bridge device
> companion assignment to core code.
> --- a/drivers/acpi/pci_root.c
> +++ b/drivers/acpi/pci_root.c
> @@ -564,6 +564,11 @@ static int acpi_pci_root_add(struct acpi_device
> *device,
>           }
>       }
>
> +    /*
> +     * pci_create_root_bus() needs to detect the parent device type,
> +     * so initialize its companion data accordingly.
> +     */
> +    ACPI_COMPANION_SET(&device->dev, device);
>
> This code will be upstreamed with the "PATCH V6 11/13 specic quirks" in
> next time after the patchset is accepted.
> Right ?

We had that patch in previous series to retrieve PCI domain nicely. But 
that has bad implication to userspace. See:
https://lkml.org/lkml/2016/5/9/918

I understand that:
[PATCH V6 01/13] pci, acpi, x86, ia64: Move ACPI host bridge device 
companion assignment to core code.
helps to get firmware specific info in hisi_pcie_init but we need to 
figure out something better for quirk handling too.

Tomasz