On 8/6/2012 7:14 AM, Russell King - ARM Linux wrote:

On Tue, Jul 31, 2012 at 07:04:39PM -0400, Cyril Chemparathy wrote:

quoted

This patch fixes up the types used when converting back and forth between
physical and virtual addresses.

Signed-off-by: Vitaly Andrianov <redacted>
Signed-off-by: Cyril Chemparathy <redacted>
---
  arch/arm/include/asm/memory.h |   17 +++++++++++------
  1 file changed, 11 insertions(+), 6 deletions(-)

diff --git a/arch/arm/include/asm/memory.h b/arch/arm/include/asm/memory.h
index 01c710d..4a0108f 100644
--- a/arch/arm/include/asm/memory.h
+++ b/arch/arm/include/asm/memory.h

@@ -157,22 +157,27 @@ extern unsigned long __pv_phys_offset;

  extern unsigned long __pv_offset;

-static inline unsigned long __virt_to_phys(unsigned long x)
+static inline phys_addr_t __virt_to_phys(unsigned long x)
  {
  	unsigned long t;
  	early_patch_imm8(x, t, "add", __pv_offset);
  	return t;
  }

-static inline unsigned long __phys_to_virt(unsigned long x)
+static inline unsigned long __phys_to_virt(phys_addr_t x)
  {
  	unsigned long t;
  	early_patch_imm8(x, t, "sub", __pv_offset);
  	return t;
  }
  #else
-#define __virt_to_phys(x)	((x) - PAGE_OFFSET + PHYS_OFFSET)
-#define __phys_to_virt(x)	((x) - PHYS_OFFSET + PAGE_OFFSET)
+
+#define __virt_to_phys(x)		\
+	((phys_addr_t)(x) - PAGE_OFFSET + PHYS_OFFSET)
+
+#define __phys_to_virt(x)		\
+	((unsigned long)((phys_addr_t)(x) - PHYS_OFFSET + PAGE_OFFSET))
+
  #endif
  #endif

@@ -207,14 +212,14 @@ static inline phys_addr_t virt_to_phys(const volatile void *x)

  static inline void *phys_to_virt(phys_addr_t x)
  {
-	return (void *)(__phys_to_virt((unsigned long)(x)));
+	return (void *)__phys_to_virt(x);
  }

  /*
   * Drivers should NOT use these either.
   */
  #define __pa(x)			__virt_to_phys((unsigned long)(x))
-#define __va(x)			((void *)__phys_to_virt((unsigned long)(x)))
+#define __va(x)			((void *)__phys_to_virt((phys_addr_t)(x)))
  #define pfn_to_kaddr(pfn)	__va((pfn) << PAGE_SHIFT)

This as a whole does not fill me with a great amount of enthusiasm,
because it breaks some of the typechecking that we have here.

The whole point of __phys_to_virt() and __virt_to_phys() is that they work
on integer types, and warn if you dare to use them with pointers.  Adding
a cast into them breaks that.

Understood.  Thanks.  The casts were needed to upgrade to 64-bit before 
arithmetic.  We should convert the non-patch __phys_to_virt and 
__virt_to_phys to inlines to keep the typechecking intact.

The whole point is that the typecasting is explicitly inside phys_to_virt()
and virt_to_phys() and not their macro counterparts.

Secondly, are you sure that this patch is correct on its own?  You're
passing a u64 into assembly only expecting a 32-bit register.  Have you
checked it does the right thing with a 64-bit phys_addr_t on both LE
and BE?

We should explicitly pass in the lower order bits here, at least until 
the next patch in the series fixes things up for 64-bit.  Thanks.

We've tested with 64-bit and 32-bit phys_addr_t, but only on LE.  Thanks 
for pointing this out, we'll figure out a way to run BE as well.

-- 
Thanks
- Cyril