Re: [PATCH] PCI: keystone: Fix build error while only CONFIG_PCI_KEYSTONE is set

From: Kishon Vijay Abraham I
Date: Mon Apr 22 2019 - 03:51:27 EST

Hi Bjorn,

On 19/04/19 11:17 PM, Bjorn Helgaas wrote:
> On Fri, Apr 19, 2019 at 09:46:39PM +0800, YueHaibing wrote:
>> On 2019/4/19 21:31, Bjorn Helgaas wrote:
>>> On Fri, Apr 19, 2019 at 10:58:55AM +0800, Yue Haibing wrote:
>>>> From: YueHaibing <yuehaibing@xxxxxxxxxx>
>>>> During randconfig builds, I occasionally run into an invalid configuration
>>>> drivers/pci/controller/dwc/pci-keystone.o: In function `ks_pcie_link_up':
>>>> pci-keystone.c:(.text+0x90): undefined reference to `__dw_pcie_read_dbi'
>>>> pci-keystone.c:(.text+0x90): relocation truncated to fit: R_AARCH64_CALL26 against undefined symbol `__dw_pcie_read_dbi'
>>>> drivers/pci/controller/dwc/pci-keystone.o: In function `ks_pcie_v3_65_scan_bus':
>>>> pci-keystone.c:(.text+0x4f0): undefined reference to `__dw_pcie_write_dbi'
>>>> pci-keystone.c:(.text+0x4f0): relocation truncated to fit: R_AARCH64_CALL26 against undefined symbol `__dw_pcie_write_dbi'
>>>> while CONFIG_PCI_KEYSTONE is selected but CONFIG_PCIE_DW
>>>> is not set, the building failed like this. This patch
>>>> selects PCIE_DW to fix it.
>>>> Reported-by: Hulk Robot <hulkci@xxxxxxxxxx>
>>>> Fixes: 5709114f0a97 ("PCI: keystone: Add support for PCIe EP in AM654x Platforms")
>>>> Signed-off-by: YueHaibing <yuehaibing@xxxxxxxxxx>
>>>> ---
>>>> drivers/pci/controller/dwc/Kconfig | 1 +
>>>> 1 file changed, 1 insertion(+)
>>>> diff --git a/drivers/pci/controller/dwc/Kconfig b/drivers/pci/controller/dwc/Kconfig
>>>> index b450ad2..641fa0f 100644
>>>> --- a/drivers/pci/controller/dwc/Kconfig
>>>> +++ b/drivers/pci/controller/dwc/Kconfig
>>>> @@ -105,6 +105,7 @@ config PCIE_SPEAR13XX
>>>> config PCI_KEYSTONE
>>>> bool "TI Keystone PCIe controller"
>>>> depends on ARCH_KEYSTONE || ARCH_K3 || ((ARM || ARM64) && COMPILE_TEST)
>>>> + select PCIE_DW
>>> What version are you building? In v5.0, PCI_KEYSTONE already selects
>>> $ git grep -A4 PCI_KEYSTONE v5.0:drivers/pci/controller/dwc/Kconfig | cat
>>> v5.0:drivers/pci/controller/dwc/Kconfig:config PCI_KEYSTONE
>>> v5.0:drivers/pci/controller/dwc/Kconfig- bool "TI Keystone PCIe controller"
>>> v5.0:drivers/pci/controller/dwc/Kconfig- depends on ARCH_KEYSTONE || (ARM && COMPILE_TEST)
>>> v5.0:drivers/pci/controller/dwc/Kconfig- depends on PCI_MSI_IRQ_DOMAIN
>>> v5.0:drivers/pci/controller/dwc/Kconfig- select PCIE_DW_HOST
>>> 08:30:42 ~/history (master)$
>> This patch based on linux-next:
> Oh, yep, looks like this was broken by ad5ca801bd7e ("PCI: keystone:
> Add support for PCIe EP in AM654x Platforms"), which added
> PCIE_DW_EP, respectively, which in turn select PCIE_DW.
> But I guess it's possible to enable PCI_KEYSTONE by itself without
> either PCI_KEYSTONE_HOST or PCI_KEYSTONE_EP. I'm not sure what that
> would mean, but we would still try to build pci-keystone.o
> I'm hoping we can amend that commit before the merge window.

Sometime back Niklas had fixed this for DRA7xx in

commit b052835c63857e13d9ada3ebc57a8f9e1d124f3a
Author: Niklas Cassel <niklas.cassel@xxxxxxxx>
Date: Wed Dec 20 00:29:28 2017 +0100

PCI: dwc: dra7xx: Refactor Kconfig and Makefile handling for host/ep mode

Refactor the Kconfig and Makefile handling for host/ep mode, since
the previous handling was a bit unorthodox and would have been a bit
bloated once more DWC based controllers added support for ep mode.

Signed-off-by: Niklas Cassel <niklas.cassel@xxxxxxxx>
Signed-off-by: Lorenzo Pieralisi <lorenzo.pieralisi@xxxxxxx>
Acked-by: Kishon Vijay Abraham I <kishon@xxxxxx>

Something similar is applicable for Keystone too as you mentioned without
selecting PCI_KEYSTONE_HOST or PCI_KEYSTONE_EP, compiling pci-keystone.c
doesn't make sense.

Please see if below looks okay to you. This can be merged with "PCI:
keystone:Add support for PCIe EP in AM654x Platforms"). Let me know if I have
to send it separately with Fixes tag.