[PATCHv6+ 01/13] of: introduce of_property_for_earch_phandle_with_args()

[Grant Likely]

On Thu, 21 Nov 2013 11:57:00 -0700, Stephen Warren <swarren at wwwdotorg.org> wrote:
> On 11/21/2013 10:17 AM, Hiroshi Doyu wrote:
> > Iterating over a property containing a list of phandles with arguments
> > is a common operation for device drivers. This patch adds a new
> > of_property_for_each_phandle_with_args() macro to make the iteration
> > simpler.
> > 
> > Signed-off-by: Hiroshi Doyu <hdoyu at nvidia.com>
> > ---
> > v6+:
> > Use the description, which Grant Likely proposed, to be full enough
> > that a future reader can figure out why a patch was written.
> >   http://lists.linuxfoundation.org/pipermail/iommu/2013-November/007062.html
> 
> This new version only addresses one of the concerns that Grant had,
> namely the commit message.
> 
> > diff --git a/include/linux/of.h b/include/linux/of.h
> 
> > +#define of_property_for_each_phandle_with_args(np, list, cells, i, args) \
> > +	for (i = 0; !of_parse_phandle_with_args(np, list, cells, i, args); i++)
> > +
> 
> Grant also wanted the actual implementation fixed so that it wasn't so
> inefficient.
> 
> What this current patch does is basically:
> 
> for every entry in the property:
>     for every entry in the property before the current index:
>         parse the phandle+specifier
> 
> That's roughly O(n^2). (n is # entries in the property)
> 
> Instead, what should happen is:
> 
> for every entry in the property:
>     parse the phandle+specifier
>     yield the result
> 
> That's roughly O(n).
> 
> In other words, an implementation more along the lines of
> include/linux/of.h's:
> 
> #define of_property_for_each_u32(np, propname, prop, p, u)      \
>         for (prop = of_find_property(np, propname, NULL),       \
>                 p = of_prop_next_u32(prop, NULL, &u);           \
>                 p;                                              \
>                 p = of_prop_next_u32(prop, p, &u))
> 
> ... so you'd need functions like of_prop_first_specifier() and
> of_prop_next_specifier(), and perhaps some associated set of state
> variables, perhaps with all the state wrapped into a single struct for
> simplicity.

That's right, I forgot I said that. Yes please fix the implementation.

g.