On Thu, Mar 3, 2016 at 9:16 AM, Ard Biesheuvel
[off-list ref] wrote:

On 3 March 2016 at 18:13, Russell King - ARM Linux
[off-list ref] wrote:

quoted

On Thu, Mar 03, 2016 at 05:24:41PM +0100, Ard Biesheuvel wrote:

quoted

If there are no remaining objections, may I suggest that Dan takes
patch #1 and #2, and that I put the third patch into Russell's patch
system? The only strict ordering requirement is that #1 is merged
before #2 and #3 both hit mainline, and that is solved by taking #1
and #2 in order.

How does the dependency between 1 and 3 get satisfied?  Your comment
makes no sense to me.

Patch #3 introduces a function arch_memremap_wb() on the ARM side
which will never get called unless patch #2 is also merged. Once both
#2 and #3 are in, the memremap() call that #1 reverts will use
MT_MEMORY_RW attributes rather than MT_DEVICE_CACHED, so #1 needs to
go first, but #2 and #3 can go in in either order.

In other words, there are no build dependencies between the patches,
but to prevent pxa2xx-flash from using MT_MEMORY_RW attributes, it
needs to go in before #2

Can we add a new MEMREMAP_ type for this case so that we don't need to
keep carrying the confusing ioremap_cache() which means different
things to different archs and silently falls back to uncached on some
archs.