[PATCH v3 1/3] ARM: Introduce atomic MMIO modify

[Catalin Marinas]

On Fri, Aug 30, 2013 at 10:08:07AM +0100, Will Deacon wrote:
> On Fri, Aug 23, 2013 at 12:48:05PM +0100, Catalin Marinas wrote:
> > On Fri, Aug 23, 2013 at 12:32:26PM +0100, Ezequiel Garcia wrote:
> > > On Fri, Aug 23, 2013 at 08:07:50AM -0300, Ezequiel Garcia wrote:
> > > > On Fri, Aug 23, 2013 at 01:38:02PM +0300, Baruch Siach wrote:
> > > > > On Fri, Aug 23, 2013 at 07:24:03AM -0300, Ezequiel Garcia wrote:
> > > > > > Some SoC have MMIO regions that are shared across orthogonal
> > > > > > subsystems. This commit implements a possible solution for the
> > > > > > thread-safe access of such regions through a spinlock-protected API.
> > > > > > 
> > > > > > Concurrent access is protected with a single spinlock for the
> > > > > > entire MMIO address space. While this protects shared-registers,
> > > > > > it also serializes access to unrelated/unshared registers.
> > > > > > 
> > > > > > We add relaxed and non-relaxed variants, by using writel_relaxed and writel,
> > > > > > respectively. The rationale for this is that some users may not require
> > > > > > register write completion but only thread-safe access to a register.
> > > > > > 
> > > > > > Signed-off-by: Ezequiel Garcia <ezequiel.garcia at free-electrons.com>
> > > > > 
> > > > > Is there a reason why this should be limited to ARM? I haven't found anything 
> > > > > ARM specific in the code.
> > > > 
> > > > I guess not.
> > > 
> > > ... or maybe yes. I'm not seeing {readl,writel}_relaxed as guaranteed
> > > to exist in every architecture. So, indeed, this seems to be ARM-dependent.
> > 
> > There was a discussion couple of years ago to make these part of the IO
> > specification since many architectures define them:
> > 
> > http://thread.gmane.org/gmane.linux.ports.arm.kernel/117626
> > 
> > (and some older threads on linux-arch which I haven't searched)
> > 
> > We could have some default implementation pointing to readl/writel while
> > letting the arch code to define more optimised variants.
> 
> The main thing I dislike about that is the back-to-back dsbs that you will
> get from the read-(modify)-write. It really makes the non-optimised version
> needlessly expensive.

Yes, it's pretty bad. But we don't have relaxed (write) accessors on
other architectures and I'm not sure about their semantics either. I
guess here it's a data dependency so you cannot write the value before
reading it, especially since sane architectures should speculate reads
or writes to device memory.

What about making it always use *_relaxed() accessors if the
architecture provides them? No need for atomic_io_modify_relaxed().

-- 
Catalin