Khalid Aziz [off-list ref] writes:

On 08/10/2017 07:20 AM, Michael Ellerman wrote:

quoted

Khalid Aziz [off-list ref] writes:
=20

quoted

A protection flag may not be valid across entire address space and
hence arch_validate_prot() might need the address a protection bit is
being set on to ensure it is a valid protection flag. For example, sparc
processors support memory corruption detection (as part of ADI feature)
flag on memory addresses mapped on to physical RAM but not on PFN mapped
pages or addresses mapped on to devices. This patch adds address to the
parameters being passed to arch_validate_prot() so protection bits can
be validated in the relevant context.

Signed-off-by: Khalid Aziz <redacted>
Cc: Khalid Aziz <khalid@gonehiking.org>
---
v7:
	- new patch

  arch/powerpc/include/asm/mman.h | 2 +-
  arch/powerpc/kernel/syscalls.c  | 2 +-
  include/linux/mman.h            | 2 +-
  mm/mprotect.c                   | 2 +-
  4 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/powerpc/include/asm/mman.h b/arch/powerpc/include/asm=

/mman.h

quoted

quoted

index 30922f699341..bc74074304a2 100644

--- a/arch/powerpc/include/asm/mman.h
+++ b/arch/powerpc/include/asm/mman.h

@@ -40,7 +40,7 @@ static inline bool arch_validate_prot(unsigned long p=

rot)

quoted

quoted

  		return false;
  	return true;
  }
-#define arch_validate_prot(prot) arch_validate_prot(prot)
+#define arch_validate_prot(prot, addr) arch_validate_prot(prot)

=20
This can be simpler, as just:
=20
#define arch_validate_prot arch_validate_prot
=20

Hi Michael,

Thanks for reviewing!

My patch expands parameter list for arch_validate_prot() from one to two=

=20

parameters. Existing powerpc version of arch_validate_prot() is written=20
with one parameter. If I use the above #define, compilation fails with:

mm/mprotect.c: In function =E2=80=98do_mprotect_pkey=E2=80=99:
mm/mprotect.c:399: error: too many arguments to function=20
=E2=80=98arch_validate_prot=E2=80=99

Another way to solve it would be to add the new addr parameter to=20
powerpc version of arch_validate_prot() but I chose the less disruptive=20
solution of tackling it through #define and expanded the existing=20
#define to include the new parameter. Make sense?

Yes, it makes sense. But it's a bit gross.

At first glance it looks like our arch_validate_prot() has an incorrect
signature.

I'd prefer you just updated it to have the correct signature, I think
you'll have to change one more line in do_mmap2(). So it's not very
intrusive.

cheers