[PATCH RFC 5/6] arm64/xen: implement xen_remap on arm64

[Stefano Stabellini]

On Fri, 31 May 2013, Stefano Stabellini wrote:
> On Fri, 31 May 2013, Catalin Marinas wrote:
> > On Fri, May 31, 2013 at 02:21:16PM +0100, Stefano Stabellini wrote:
> > > On Fri, 31 May 2013, Catalin Marinas wrote:
> > > > On Thu, May 30, 2013 at 05:18:32PM +0100, Stefano Stabellini wrote:
> > > > > --- a/arch/arm/include/asm/xen/page.h
> > > > > +++ b/arch/arm/include/asm/xen/page.h
> > > > > @@ -90,6 +90,10 @@ static inline bool set_phys_to_machine(unsigned long pfn, unsigned long mfn)
> > > > >  	return __set_phys_to_machine(pfn, mfn);
> > > > >  }
> > > > >  
> > > > > +#ifdef CONFIG_ARM64
> > > > > +#define xen_remap(cookie, size) __ioremap((cookie), (size), __pgprot(PROT_NORMAL))
> > > > > +#else
> > > > >  #define xen_remap(cookie, size) __arm_ioremap((cookie), (size), MT_MEMORY);
> > > > > +#endif
> > > > 
> > > > Now I saw the ARM-specific part. Can you not use something like
> > > > ioremap_cached() which would give normal cacheable memory (at least on
> > > > ARMv7).
> > > 
> > > No, I cannot because ioremap_cached uses MT_DEVICE_CACHED, while this
> > > needs to be MT_MEMORY. It is used for normal memory pages, not device
> > > memory.
> > 
> > MT_DEVICE_CACHED is Normal memory for ARMv7.
> 
> I didn't realize that MT_DEVICE_CACHED and MT_MEMORY end up having the
> same AttrIndx encoding!
> In that case yes, I should be able to use it.
> I take that I should just implement ioremap_cached on arm64 as well?
> 

Should just use MT_MEMORY and call it PROT_NORMAL?

#define PROT_NORMAL     (PROT_DEFAULT | PTE_ATTRINDX(MT_NORMAL))
#define ioremap_cached(addr, size)     __ioremap((addr), (size), __pgprot(PROT_NORMAL))