[U-Boot] [RFC PATCH v2] ARM: Avoid compiler optimization for usages of readb, writeb and friends.

[John Rigby]

Earlier in this thread Alexander said:
> I haven't add the definitions which are using a memory barrier because I haven't found
> a place in the kernel where they were actually enabled
> (CONFIG_ARM_DMA_MEM_BUFFERABLE).

I think this is the problem because it is indeed defined for all v6
and v7 arm platforms.  Here is the config snippet from
arch/arm/mm/Kconfig:

config ARM_DMA_MEM_BUFFERABLE
	bool "Use non-cacheable memory for DMA" if CPU_V6 && !CPU_V7
	depends on !(MACH_REALVIEW_PB1176 || REALVIEW_EB_ARM11MP || \
		     MACH_REALVIEW_PB11MP)
	default y if CPU_V6 || CPU_V7
	help
	  Historically, the kernel has used strongly ordered mappings to
	  provide DMA coherent memory.  With the advent of ARMv7, mapping
	  memory with differing types results in unpredictable behaviour,
	  so on these CPUs, this option is forced on.

	  Multiple mappings with differing attributes is also unpredictable
	  on ARMv6 CPUs, but since they do not have aggressive speculative
	  prefetch, no harm appears to occur.

	  However, drivers may be missing the necessary barriers for ARMv6,
	  and therefore turning this on may result in unpredictable driver
	  behaviour.  Therefore, we offer this as an option.

	  You are recommended say 'Y' here and debug any affected drivers.