Re: [PATCH 1/7] of: base: Add of_count_phandle_with_fixed_args()

[Richard Fitzgerald]

On 16/10/2020 14:31, Rob Herring wrote:
> On Thu, Oct 15, 2020 at 11:52 AM Robin Murphy <robin.murphy@xxxxxxx> wrote:
> > On 2020-10-14 19:39, Rob Herring wrote:
> > > On Wed, Oct 14, 2020 at 9:54 AM Richard Fitzgerald
> > > <rf@xxxxxxxxxxxxxxxxxxxxx> wrote:
> > > > Add an equivalent of of_count_phandle_with_args() for fixed argument
> > > > sets, to pair with of_parse_phandle_with_fixed_args().
> > > >
> > > > Signed-off-by: Richard Fitzgerald <rf@xxxxxxxxxxxxxxxxxxxxx>
> > > > ---
> > > >    drivers/of/base.c  | 42 ++++++++++++++++++++++++++++++++++++++++++
> > > >    include/linux/of.h |  9 +++++++++
> > > >    2 files changed, 51 insertions(+)
> > > >
> > > > diff --git a/drivers/of/base.c b/drivers/of/base.c
> > > > index ea44fea99813..45d8b0e65345 100644
> > > > --- a/drivers/of/base.c
> > > > +++ b/drivers/of/base.c
> > > > @@ -1772,6 +1772,48 @@ int of_count_phandle_with_args(const struct device_node *np, const char *list_na
> > > >    }
> > > >    EXPORT_SYMBOL(of_count_phandle_with_args);
> > > >
> > > > +/**
> > > > + * of_count_phandle_with_fixed_args() - Find the number of phandles references in a property
> > > > + * @np:                pointer to a device tree node containing a list
> > > > + * @list_name: property name that contains a list
> > > > + * @cell_count: number of argument cells following the phandle
> > > > + *
> > > > + * Returns the number of phandle + argument tuples within a property. It
> > > > + * is a typical pattern to encode a list of phandle and variable
> > > > + * arguments into a single property.
> > > > + */
> > > > +int of_count_phandle_with_fixed_args(const struct device_node *np,
> > > > +                                    const char *list_name,
> > > > +                                    int cells_count)
> > > > +{
> > >
> > > Looks to me like you can refactor of_count_phandle_with_args to handle
> > > both case and then make this and of_count_phandle_with_args simple
> > > wrapper functions.
> >
> > Although for just counting the number of phandles each with n arguments
> > that a property contains, isn't that simply a case of dividing the
> > property length by n + 1? The phandles themselves will be validated by
> > any subsequent of_parse_phandle*() call anyway, so there doesn't seem
> > much point in doing more work then necessary here.
> >
> > > > +       struct of_phandle_iterator it;
> > > > +       int rc, cur_index = 0;
> > > > +
> > > > +       if (!cells_count) {
> > > > +               const __be32 *list;
> > > > +               int size;
> > > > +
> > > > +               list = of_get_property(np, list_name, &size);
> > > > +               if (!list)
> > > > +                       return -ENOENT;
> > > > +
> > > > +               return size / sizeof(*list);
> >
> > Case in point - if it's OK to do exactly that for n == 0, then clearly
> > we're *aren't* fussed about validating anything, so the n > 0 code below
> > is nothing more than a massively expensive way to check for a nonzero
> > remainder :/
>
> Indeed. We should just generalize this. It can still be refactored to
> shared code.
>
> It's probably worthwhile to check for a remainder here IMO.

Ok, I looked at the implementation of of_phandle_iterator_next() and
it is in fact simply incrementing by 'count' 32-bit words. So as Robin
said the count_phandle_with_x_args()functions could simply divide the
length by count+1.

However, may I suggest that should be done in a separate patch after my
patch to add count_phandle_with_fixed_args()? That way, if replacing the
iteration with the simple length divide causes any unforeseen problems
the patch can just be reverted.

> Rob