Re: [PATCH] fb.h: ARM uses __raw_{read/write}

[Catalin Marinas]

On Mon, Jun 13, 2011 at 01:00:04PM +0900, Paul Mundt wrote:
> On Fri, Jun 10, 2011 at 05:31:08PM -0700, H Hartley Sweeten wrote:
> > ARM provides __raw_{read/write}* functions for memory access. These
> > should be used instead of the default '(*(volatile' stuff to make sure the
> > memory accesses are typesafe (void __iomem *).
> > 
> > This also fixes a number of sparse warning like:
> > 
> >   warning: cast removes address space of expression
> > 
> > Signed-off-by: H Hartley Sweeten <hsweeten@xxxxxxxxxxxxxxxxxxx>
> > Cc: Paul Mundt <lethal@xxxxxxxxxxxx>
> > Cc: Russell King <linux@xxxxxxxxxxxxxxxx>
> 
> I'm not sure what semantics are desirable for ARM here, so I'll wait for
> Russell to reply.
> 
> This wrapping will basically mean that the fb_read/write ops are using
> __raw_xxx variants while the memset and memcpy wrappers will be using the
> regular read/write[bwl] routines which contain __iormb() calls. Given
> that ioread/write and friends all wrap in to the normal versions with the
> barriers, I would suppose that this is the default behaviour that is
> desired, as opposed to wrapping in to the __raw_xxx variants directly.

The intention for the __iormb/__iowmb calls is in relation to DMA
transfers where a buffer in normal RAM is filled in with data and the
transfer started by a writel() to a device. We need to make sure that
the normal RAM writing completes before the writel().

The change proposed by Hartley wouldn't make much difference from the
current volatile accesses (__raw_* accessors are implemented as volatile
on ARM).

I think the memcpy_(from|to)io could be optimised on ARM to only add a
barrier before or after he copying loop.

-- 
Catalin
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/