Re: drivers/pci/controller/cadence/pci-j721e.c:648:undefined reference to `cdns_pcie_host_disable'

[Siddharth Vadapalli <s-vadapalli@ti.com>]

On Tue, 2025-11-11 at 16:24 +0530, Manivannan Sadhasivam wrote:

Hello Mani,

> + Siddharth

Thank you for notifying me of this.

> 
> On Tue, Nov 11, 2025 at 05:28:54PM +0800, kernel test robot wrote:
> > tree:   https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git master
> > head:   4427259cc7f7571a157fbc9b5011e1ef6fe0a4a8
> > commit: 1c72774df028429836eec3394212f2921bb830fc PCI: sg2042: Add Sophgo SG2042 PCIe driver
> > date:   8 weeks ago
> > config: loongarch-randconfig-r113-20251110 (https://download.01.org/0day-ci/archive/20251111/202511111705.MZ7ls8Hm-lkp@intel.com/config)
> > compiler: loongarch64-linux-gcc (GCC) 15.1.0
> > reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20251111/202511111705.MZ7ls8Hm-lkp@intel.com/reproduce)
> > 
> > If you fix the issue in a separate patch/commit (i.e. not just a new version of
> > the same patch/commit), kindly add following tags
> > > Reported-by: kernel test robot <lkp@intel.com>
> > > Closes: https://lore.kernel.org/oe-kbuild-all/202511111705.MZ7ls8Hm-lkp@intel.com/
> > 
> > All errors (new ones prefixed by >>):
> > 
> >    loongarch64-linux-ld: drivers/pci/controller/cadence/pci-j721e.o: in function `j721e_pcie_remove':
> > > > drivers/pci/controller/cadence/pci-j721e.c:648:(.text+0x83c): undefined reference to `cdns_pcie_host_disable'
> > 
> 
> From .config:
> 
> CONFIG_PCIE_CADENCE=y
> CONFIG_PCIE_CADENCE_HOST=m
> CONFIG_PCIE_CADENCE_EP=y
> CONFIG_PCIE_CADENCE_PLAT=y
> # CONFIG_PCIE_CADENCE_PLAT_HOST is not set
> CONFIG_PCIE_CADENCE_PLAT_EP=y
> CONFIG_PCIE_SG2042_HOST=m
> CONFIG_PCI_J721E=y
> # CONFIG_PCI_J721E_HOST is not set
> CONFIG_PCI_J721E_EP=y
> 
> PCI_J721E selects PCIE_CADENCE_HOST only if PCI_J721E_HOST is selected,
> otherwise, it will not select it. This will take care of the dependency between
> PCI_J721E and PCIE_CADENCE_{HOST/EP}.
> 
> But if PCIE_CADENCE_HOST is selected as a module by other drivers like,
> CONFIG_PCIE_SG2042_HOST=m, then if PCI_J721E is selected as a built-in using
> CONFIG_PCI_J721E_EP=y, it results in this build error as the built-in driver
> becomes dependent on a symbol from a loadable module.

While I don't deny the build error associated with the above config, it is
an invalid config in the sense that the Glue drivers for two different
devices are being enabled. This seems to be a generic issue wherein
multiple drivers tend to depend on a library/common driver. How is it
handled in such cases?
Is there a notion of reordering configs to ensure that such build errors
are avoided?

If PCI_J721E_EP was selected as 'y' before 'PCI_SG2042_HOST' being selected
as 'm', it would have resulted in 'PCIE_CADENCE_EP' being selected as 'y'
and this won't cause a build error even with 'PCI_SG2042_HOST' selected as
'm'.

> 
> I guess, we should force PCIE_CADENCE_{HOST/EP} to be 'bool' as it is getting
> selected by multiple drivers.

This will defeat the purpose of the series that enabled loadable module
support for the pci-j721e.c driver and the pcie-cadence-host/ep.c drivers.

Is there a way to address the issue by updating Kconfig? Specifically, is
there a way to re-order the 'select' scheme to fix the build error?


Regards,
Siddharth.