Re: Breakage in arch/mips/kernel/traps.c for 64bit

["Maciej W. Rozycki" <macro@xxxxxxxxxxxxxx>]

On Thu, 1 May 2008, Thomas Bogendoerfer wrote:

> it would be nice, if people started thinking before supplying such
> crappy^Winteresting code:
> 
> arch/mips/kernel/traps.c:
> 
> #define IS_KVA01(a) ((((unsigned int)a) & 0xc0000000) == 0x80000000)
> 
> Kills every 64bit kernel build...

 Not everybody tests 64-bit stuff as some people limit themselves to
32-bit systems only.  It looks like a step backwards, but there you go.

> Why is this needed at all ?

 It looks like an attempt to avoid TLB exceptions for the stack dump -- if
that is the case, then obviously a piece of code like one in
arch/mips/lib/uncached.c should be used to check for CKSEG0/1 and XKPHYS.  
If there are two uses of this code, then it should be wrapped in an inline
function and put in a header; <asm/addrspace.h>, perhaps.

  Maciej