-Wincompatible-pointer-types in arch/powerpc/platforms/embedded6xx/mvme5100.c

-Wincompatible-pointer-types in arch/powerpc/platforms/embedded6xx/mvme5100.c

[Nathan Chancellor]

Hi all,

0day reported a build error in arch/powerpc/platforms/embedded6xx/mvme5100.c
when building with clang [1]. This is not a clang specific issue since
it also happens with gcc:

$ curl -LSs https://lore.kernel.org/lkml/202004131704.6MH1jcq3%25lkp@intel.com/2-a.bin | gzip -d > .config
$ make -j$(nproc) -s ARCH=powerpc CROSS_COMPILE=powerpc-linux- olddefconfig arch/powerpc/platforms/embedded6xx/mvme5100.o
arch/powerpc/platforms/embedded6xx/mvme5100.c: In function 'mvme5100_add_bridge':
arch/powerpc/platforms/embedded6xx/mvme5100.c:135:58: error: passing argument 5 of 'early_read_config_dword' from incompatible pointer type [-Werror=incompatible-pointer-types]
  135 |  early_read_config_dword(hose, 0, 0, PCI_BASE_ADDRESS_1, &pci_membase);
      |                                                          ^~~~~~~~~~~~
      |                                                          |
      |                                                          phys_addr_t * {aka long long unsigned int *}
In file included from arch/powerpc/platforms/embedded6xx/mvme5100.c:18:
./arch/powerpc/include/asm/pci-bridge.h:139:32: note: expected 'u32 *' {aka 'unsigned int *'} but argument is of type 'phys_addr_t *' {aka 'long long unsigned int *'}
  139 |    int dev_fn, int where, u32 *val);
      |                           ~~~~~^~~
In file included from ./include/linux/printk.h:7,
                 from ./include/linux/kernel.h:15,
                 from ./include/linux/list.h:9,
                 from ./include/linux/rculist.h:10,
                 from ./include/linux/pid.h:5,
                 from ./include/linux/sched.h:14,
                 from ./include/linux/ratelimit.h:6,
                 from ./include/linux/dev_printk.h:16,
                 from ./include/linux/device.h:15,
                 from ./include/linux/of_platform.h:9,
                 from arch/powerpc/platforms/embedded6xx/mvme5100.c:15:
./include/linux/kern_levels.h:5:18: error: format '%x' expects argument of type 'unsigned int', but argument 2 has type 'phys_addr_t' {aka 'long long unsigned int'} [-Werror=format=]
    5 | #define KERN_SOH "\001"  /* ASCII Start Of Header */
      |                  ^~~~~~
./include/linux/kern_levels.h:14:19: note: in expansion of macro 'KERN_SOH'
   14 | #define KERN_INFO KERN_SOH "6" /* informational */
      |                   ^~~~~~~~
./include/linux/printk.h:305:9: note: in expansion of macro 'KERN_INFO'
  305 |  printk(KERN_INFO pr_fmt(fmt), ##__VA_ARGS__)
      |         ^~~~~~~~~
arch/powerpc/platforms/embedded6xx/mvme5100.c:142:2: note: in expansion of macro 'pr_info'
  142 |  pr_info("mvme5100_pic_init: pci_membase: %x\n", pci_membase);
      |  ^~~~~~~
arch/powerpc/platforms/embedded6xx/mvme5100.c:142:44: note: format string is defined here
  142 |  pr_info("mvme5100_pic_init: pci_membase: %x\n", pci_membase);
      |                                           ~^
      |                                            |
      |                                            unsigned int
      |                                           %llx
cc1: all warnings being treated as errors
make[4]: *** [scripts/Makefile.build:267: arch/powerpc/platforms/embedded6xx/mvme5100.o] Error 1
make[3]: *** [scripts/Makefile.build:488: arch/powerpc/platforms/embedded6xx] Error 2
make[2]: *** [scripts/Makefile.build:488: arch/powerpc/platforms] Error 2
make[1]: *** [Makefile:1722: arch/powerpc] Error 2
make: *** [Makefile:328: __build_one_by_one] Error 2

I am not sure how exactly this should be fixed. Should this driver just
not be selectable when CONFIG_PHYS_ADDR_T_64BIT is selected or is there
something else that I am missing?

[1]: https://lore.kernel.org/lkml/202004131704.6MH1jcq3%25lkp@intel.com/

Cheers,
Nathan

Re: -Wincompatible-pointer-types in arch/powerpc/platforms/embedded6xx/mvme5100.c

[Michael Ellerman]

Hi Nathan,

Thanks for the report.

Nathan Chancellor <natechancellor@gmail.com> writes:
> Hi all,
>
> 0day reported a build error in arch/powerpc/platforms/embedded6xx/mvme5100.c
> when building with clang [1]. This is not a clang specific issue since
> it also happens with gcc:
>
> $ curl -LSs https://lore.kernel.org/lkml/202004131704.6MH1jcq3%25lkp@intel.com/2-a.bin | gzip -d > .config
> $ make -j$(nproc) -s ARCH=powerpc CROSS_COMPILE=powerpc-linux- olddefconfig arch/powerpc/platforms/embedded6xx/mvme5100.o
> arch/powerpc/platforms/embedded6xx/mvme5100.c: In function 'mvme5100_add_bridge':
> arch/powerpc/platforms/embedded6xx/mvme5100.c:135:58: error: passing argument 5 of 'early_read_config_dword' from incompatible pointer type [-Werror=incompatible-pointer-types]
>   135 |  early_read_config_dword(hose, 0, 0, PCI_BASE_ADDRESS_1, &pci_membase);
>       |                                                          ^~~~~~~~~~~~
>       |                                                          |
>       |                                                          phys_addr_t * {aka long long unsigned int *}


Yuck.

> ...
> I am not sure how exactly this should be fixed. Should this driver just
> not be selectable when CONFIG_PHYS_ADDR_T_64BIT is selected or is there
> something else that I am missing?

I'm not sure TBH. This is all ancient history as far as I can tell, none
of it's been touched for ~7 years.

Your config has:

CONFIG_EMBEDDED6xx=y
CONFIG_PPC_BOOK3S_32=y
CONFIG_PPC_BOOK3S_6xx=y
CONFIG_PPC_MPC52xx=y
CONFIG_PPC_86xx=y


Which I'm not sure really makes sense at all, ie. it's trying to build a
kernel for multiple platforms at once (EMBEDDED6xx, MPC52xx, 86xx), but
the Kconfig doesn't exclude that so I guess we have to live with it for
now.

Then Kconfig has:

config PHYS_64BIT
	bool 'Large physical address support' if E500 || PPC_86xx
	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx
	select PHYS_ADDR_T_64BIT


So it's PPC_86xx that allows 64-bit phys_addr_t.

That was added in:

  4ee7084eb11e ("POWERPC: Allow 32-bit hashed pgtable code to support 36-bit physical")

Which did:

config PHYS_64BIT
-	bool 'Large physical address support' if E500
-	depends on 44x || E500
+	bool 'Large physical address support' if E500 || PPC_86xx
+	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx


ie. it wanted to add support for PPC_86xx but had to deliberately
exclude some of the other BOOK3S_32 based platforms.

So I'm going to guess it should have also excluded embedded6xx, and this
seems to fix it:

diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype
index 0c3c1902135c..134fc383daf7 100644
--- a/arch/powerpc/platforms/Kconfig.cputype
+++ b/arch/powerpc/platforms/Kconfig.cputype
@@ -278,7 +278,7 @@ config PTE_64BIT
 
 config PHYS_64BIT
 	bool 'Large physical address support' if E500 || PPC_86xx
-	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx
+	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx && !EMBEDDED6xx
 	select PHYS_ADDR_T_64BIT
 	---help---
 	  This option enables kernel support for larger than 32-bit physical


So unless anyone can tell me otherwise I'm inclined to commit that ^

cheers

Re: -Wincompatible-pointer-types in arch/powerpc/platforms/embedded6xx/mvme5100.c

[Nathan Chancellor]

Hi Michael,

On Tue, Apr 14, 2020 at 05:33:45PM +1000, Michael Ellerman wrote:
> Hi Nathan,
> 
> Thanks for the report.
> 
> Nathan Chancellor <natechancellor@gmail.com> writes:
> > Hi all,
> >
> > 0day reported a build error in arch/powerpc/platforms/embedded6xx/mvme5100.c
> > when building with clang [1]. This is not a clang specific issue since
> > it also happens with gcc:
> >
> > $ curl -LSs https://lore.kernel.org/lkml/202004131704.6MH1jcq3%25lkp@intel.com/2-a.bin | gzip -d > .config
> > $ make -j$(nproc) -s ARCH=powerpc CROSS_COMPILE=powerpc-linux- olddefconfig arch/powerpc/platforms/embedded6xx/mvme5100.o
> > arch/powerpc/platforms/embedded6xx/mvme5100.c: In function 'mvme5100_add_bridge':
> > arch/powerpc/platforms/embedded6xx/mvme5100.c:135:58: error: passing argument 5 of 'early_read_config_dword' from incompatible pointer type [-Werror=incompatible-pointer-types]
> >   135 |  early_read_config_dword(hose, 0, 0, PCI_BASE_ADDRESS_1, &pci_membase);
> >       |                                                          ^~~~~~~~~~~~
> >       |                                                          |
> >       |                                                          phys_addr_t * {aka long long unsigned int *}
> 
> 
> Yuck.
> 
> > ...
> > I am not sure how exactly this should be fixed. Should this driver just
> > not be selectable when CONFIG_PHYS_ADDR_T_64BIT is selected or is there
> > something else that I am missing?
> 
> I'm not sure TBH. This is all ancient history as far as I can tell, none
> of it's been touched for ~7 years.
> 
> Your config has:
> 
> CONFIG_EMBEDDED6xx=y
> CONFIG_PPC_BOOK3S_32=y
> CONFIG_PPC_BOOK3S_6xx=y
> CONFIG_PPC_MPC52xx=y
> CONFIG_PPC_86xx=y
> 
> 
> Which I'm not sure really makes sense at all, ie. it's trying to build a
> kernel for multiple platforms at once (EMBEDDED6xx, MPC52xx, 86xx), but
> the Kconfig doesn't exclude that so I guess we have to live with it for
> now.

c'est la randconfig :)

> Then Kconfig has:
> 
> config PHYS_64BIT
> 	bool 'Large physical address support' if E500 || PPC_86xx
> 	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx
> 	select PHYS_ADDR_T_64BIT
> 
> 
> So it's PPC_86xx that allows 64-bit phys_addr_t.
> 
> That was added in:
> 
>   4ee7084eb11e ("POWERPC: Allow 32-bit hashed pgtable code to support 36-bit physical")
> 
> Which did:
> 
> config PHYS_64BIT
> -	bool 'Large physical address support' if E500
> -	depends on 44x || E500
> +	bool 'Large physical address support' if E500 || PPC_86xx
> +	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx
> 
> 
> ie. it wanted to add support for PPC_86xx but had to deliberately
> exclude some of the other BOOK3S_32 based platforms.
> 
> So I'm going to guess it should have also excluded embedded6xx, and this
> seems to fix it:

This is what I was thinking as well; I agree with your analysis. Feel
free to slap the following tags on:

Reviewed-by: Nathan Chancellor <natechancellor@gmail.com>
Tested-by: Nathan Chancellor <natechancellor@gmail.com> # build

> diff --git a/arch/powerpc/platforms/Kconfig.cputype b/arch/powerpc/platforms/Kconfig.cputype
> index 0c3c1902135c..134fc383daf7 100644
> --- a/arch/powerpc/platforms/Kconfig.cputype
> +++ b/arch/powerpc/platforms/Kconfig.cputype
> @@ -278,7 +278,7 @@ config PTE_64BIT
>  
>  config PHYS_64BIT
>  	bool 'Large physical address support' if E500 || PPC_86xx
> -	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx
> +	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx && !EMBEDDED6xx
>  	select PHYS_ADDR_T_64BIT
>  	---help---
>  	  This option enables kernel support for larger than 32-bit physical
> 
> 
> So unless anyone can tell me otherwise I'm inclined to commit that ^
> 
> cheers

Cheers,
Nathan

Re: -Wincompatible-pointer-types in arch/powerpc/platforms/embedded6xx/mvme5100.c

[Scott Wood]

On Tue, 2020-04-14 at 17:33 +1000, Michael Ellerman wrote:
> I'm not sure TBH. This is all ancient history as far as I can tell, none
> of it's been touched for ~7 years.
> 
> Your config has:
> 
> CONFIG_EMBEDDED6xx=y
> CONFIG_PPC_BOOK3S_32=y
> CONFIG_PPC_BOOK3S_6xx=y
> CONFIG_PPC_MPC52xx=y
> CONFIG_PPC_86xx=y
> 
> 
> Which I'm not sure really makes sense at all, ie. it's trying to build a
> kernel for multiple platforms at once (EMBEDDED6xx, MPC52xx, 86xx), but
> the Kconfig doesn't exclude that so I guess we have to live with it for
> now.

I thought supporting multiple platforms in a kernel was something we tried to
support when practical?

> So I'm going to guess it should have also excluded embedded6xx, and this
> seems to fix it:
> 
> diff --git a/arch/powerpc/platforms/Kconfig.cputype
> b/arch/powerpc/platforms/Kconfig.cputype
> index 0c3c1902135c..134fc383daf7 100644
> --- a/arch/powerpc/platforms/Kconfig.cputype
> +++ b/arch/powerpc/platforms/Kconfig.cputype
> @@ -278,7 +278,7 @@ config PTE_64BIT
>  
>  config PHYS_64BIT
>  	bool 'Large physical address support' if E500 || PPC_86xx
> -	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx
> +	depends on (44x || E500 || PPC_86xx) && !PPC_83xx && !PPC_82xx &&
> !EMBEDDED6xx
>  	select PHYS_ADDR_T_64BIT
>  	---help---
>  	  This option enables kernel support for larger than 32-bit physical
> 
> 
> So unless anyone can tell me otherwise I'm inclined to commit that ^

This could silently break someone's config who's depending on PHYS_64BIT (e.g.
an 86xx kernel that happens to include an embedded6xx target as well, even if
just by accident).  It'd be better to have the MVME500 depend on
!CONFIG_PHYS_ADDR_T_64BIT as Nathan suggested (if there's nobody around to
test a fix to the actual bug), which shouldn't break anyone since it already
didn't build.

-Scott