[PATCH] PCI: export pci_find_host_bridge() to fix MFD build error

[PATCH] PCI: export pci_find_host_bridge() to fix MFD build error

[Randy Dunlap]

Fix a build error in drivers/mfd/ioc.o by exporting
pci_find_host_bridge().

ERROR: modpost: "pci_find_host_bridge" [drivers/mfd/ioc3.ko] undefined!

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Cc: Thomas Bogendoerfer <tsbogend@alpha.franken.de>
Cc: Paul Burton <paulburton@kernel.org>
Cc: linux-mips@vger.kernel.org
Cc: Bjorn Helgaas <bhelgaas@google.com>
Cc: linux-pci@vger.kernel.org
---
 drivers/pci/host-bridge.c |    2 ++
 1 file changed, 2 insertions(+)

--- linux-next-20201016.orig/drivers/pci/host-bridge.c
+++ linux-next-20201016/drivers/pci/host-bridge.c
@@ -4,6 +4,7 @@
  */
 
 #include <linux/kernel.h>
+#include <linux/export.h>
 #include <linux/pci.h>
 #include <linux/module.h>
 
@@ -23,6 +24,7 @@ struct pci_host_bridge *pci_find_host_br
 
 	return to_pci_host_bridge(root_bus->bridge);
 }
+EXPORT_SYMBOL(pci_find_host_bridge);
 
 struct device *pci_get_host_bridge_device(struct pci_dev *dev)
 {

Re: [PATCH] PCI: export pci_find_host_bridge() to fix MFD build error

[Christoph Hellwig]

On Sun, Oct 18, 2020 at 11:14:53PM -0700, Randy Dunlap wrote:
> Fix a build error in drivers/mfd/ioc.o by exporting
> pci_find_host_bridge().
> 
> ERROR: modpost: "pci_find_host_bridge" [drivers/mfd/ioc3.ko] undefined!

I think the mfd code should be fixed to not depend on such an internal
symbol instead.

Re: [PATCH] PCI: export pci_find_host_bridge() to fix MFD build error

[Randy Dunlap]

On 10/20/20 1:02 AM, Christoph Hellwig wrote:
> On Sun, Oct 18, 2020 at 11:14:53PM -0700, Randy Dunlap wrote:
>> Fix a build error in drivers/mfd/ioc.o by exporting
>> pci_find_host_bridge().
>>
>> ERROR: modpost: "pci_find_host_bridge" [drivers/mfd/ioc3.ko] undefined!
> 
> I think the mfd code should be fixed to not depend on such an internal
> symbol instead.
> 

Thanks for commenting. and I am not surprised.

Adding Cc: for Thomas and Lee.

-- 
~Randy

Re: [PATCH] PCI: export pci_find_host_bridge() to fix MFD build error

[Randy Dunlap]

On 10/20/20 7:06 AM, Randy Dunlap wrote:
> On 10/20/20 1:02 AM, Christoph Hellwig wrote:
>> On Sun, Oct 18, 2020 at 11:14:53PM -0700, Randy Dunlap wrote:
>>> Fix a build error in drivers/mfd/ioc.o by exporting
>>> pci_find_host_bridge().
>>>
>>> ERROR: modpost: "pci_find_host_bridge" [drivers/mfd/ioc3.ko] undefined!
>>
>> I think the mfd code should be fixed to not depend on such an internal
>> symbol instead.
>>
> 
> Thanks for commenting. and I am not surprised.
> 
> Adding Cc: for Thomas and Lee.

BTW, if someone wants to use the MFD ioc3 driver now,
without any source code patches, they can just build it in
(SGI_MFD_IOC3=y) instead of building it as a loadable module.

It builds successfully and can use pci_find_host_bridge().
pci_find_host_bridge() just isn't exported for loadable modules.

So this could be useful (PCI=y):

config SGI_MFD_IOC3
	tristate "SGI IOC3 core driver"
	depends on PCI=y && MIPS && 64BIT
	select MFD_CORE

-- 
~Randy

Re: [PATCH] PCI: export pci_find_host_bridge() to fix MFD build error

[Randy Dunlap]

On 10/20/20 8:51 AM, Randy Dunlap wrote:
> On 10/20/20 7:06 AM, Randy Dunlap wrote:
>> On 10/20/20 1:02 AM, Christoph Hellwig wrote:
>>> On Sun, Oct 18, 2020 at 11:14:53PM -0700, Randy Dunlap wrote:
>>>> Fix a build error in drivers/mfd/ioc.o by exporting
>>>> pci_find_host_bridge().
>>>>
>>>> ERROR: modpost: "pci_find_host_bridge" [drivers/mfd/ioc3.ko] undefined!
>>>
>>> I think the mfd code should be fixed to not depend on such an internal
>>> symbol instead.
>>>
>>
>> Thanks for commenting. and I am not surprised.
>>
>> Adding Cc: for Thomas and Lee.
> 
> BTW, if someone wants to use the MFD ioc3 driver now,
> without any source code patches, they can just build it in
> (SGI_MFD_IOC3=y) instead of building it as a loadable module.
> 
> It builds successfully and can use pci_find_host_bridge().
> pci_find_host_bridge() just isn't exported for loadable modules.
> 
> So this could be useful (PCI=y):
> 
> config SGI_MFD_IOC3
> 	tristate "SGI IOC3 core driver"
> 	depends on PCI=y && MIPS && 64BIT
> 	select MFD_CORE

Wow. That's not correct at all. Need more coffee.

Just set SGI_MFD_IOC3=y and it will be fine.

-- 
~Randy

Re: [PATCH] PCI: export pci_find_host_bridge() to fix MFD build error

[Thomas Bogendoerfer]

On Tue, Oct 20, 2020 at 09:02:19AM +0100, Christoph Hellwig wrote:
> On Sun, Oct 18, 2020 at 11:14:53PM -0700, Randy Dunlap wrote:
> > Fix a build error in drivers/mfd/ioc.o by exporting
> > pci_find_host_bridge().
> > 
> > ERROR: modpost: "pci_find_host_bridge" [drivers/mfd/ioc3.ko] undefined!
> 
> I think the mfd code should be fixed to not depend on such an internal
> symbol instead.

are you talking about the EXPORT_SYMBOL or the usage of
pci_find_host_bridge() outside of drivers/pci ? If the latter the
function is used in arm/arm64 code and pci controller code... so
I doesn't look like a pure internal symbol to me.

Thomas.

-- 
Crap can work. Given enough thrust pigs will fly, but it's not necessarily a
good idea.                                                [ RFC1925, 2.3 ]

Re: [PATCH] PCI: export pci_find_host_bridge() to fix MFD build error

[Christoph Hellwig]

On Wed, Oct 21, 2020 at 12:08:45PM +0200, Thomas Bogendoerfer wrote:
> On Tue, Oct 20, 2020 at 09:02:19AM +0100, Christoph Hellwig wrote:
> > On Sun, Oct 18, 2020 at 11:14:53PM -0700, Randy Dunlap wrote:
> > > Fix a build error in drivers/mfd/ioc.o by exporting
> > > pci_find_host_bridge().
> > > 
> > > ERROR: modpost: "pci_find_host_bridge" [drivers/mfd/ioc3.ko] undefined!
> > 
> > I think the mfd code should be fixed to not depend on such an internal
> > symbol instead.
> 
> are you talking about the EXPORT_SYMBOL or the usage of
> pci_find_host_bridge() outside of drivers/pci ? If the latter the
> function is used in arm/arm64 code and pci controller code... so
> I doesn't look like a pure internal symbol to me.

All of those are built-in and related bits aren't exported at all, or
at best EXPORT_SYMBOL_GPL.  I think just not allowing the ioc3 mfd to
be built modular might be the easiest fix here.  Especially as for IP27
and co kernel it is everything but optional anyway.