[PATCH] of/pdt: allow DT device matching by fixing 'name' brokenness

[Grant Likely]

On Fri, Feb 18, 2011 at 07:06:59PM -0800, Andres Salomon wrote:
> On Fri, 18 Feb 2011 23:42:57 +0000
> Daniel Drake <dsd at laptop.org> wrote:
> 
> > On 16 February 2011 22:44, David Woodhouse <dwmw2 at infradead.org>
> > wrote:
> > > On Wed, 2011-02-16 at 22:28 +0000, Daniel Drake wrote:
> > >>
> > >> +static int __init add_common_platform_devices(void)
> > >> +{
> > >> +       struct platform_device *pdev;
> > >> +
> > >> +       pdev = platform_device_register_simple("olpc-battery", -1,
> > >> NULL, 0);
> > >> +       if (IS_ERR(pdev))
> > >> +               return PTR_ERR(pdev);
> > >> +
> > >> +       return 0;
> > >> +}
> > >> +
> > >
> > > Still kind of sucks that you have to do this, and can't bind to
> > > something in the device-tree.
> > 
> > OK, feel free to put this patch on hold for now. I started looking at
> > the device tree approach today. It looks doable but first we have to
> > fix a DT bug/inconsistency that is preventing us from correctly
> > binding to the tree's devices.
> > 
> > Daniel
> 
> 
> Mea culpa.  The patch below fixes a bug I introduced earlier.
> Cc'ing the sparc folks, as this probably affects them
> (although I would think that it fixes broken behavior for them..?)

Wait; why are you binding to a device based on name?  Binding by name
and/or device_type is strongly discouraged for new code.  Use compatible
instead.

As for this patch, comments below...

> From: Andres Salomon <dilinger at queued.net>
> 
> Commit e2f2a93b changed dp->name from using the 'name' property to
> using package-to-path.  This fixed /proc/device-tree creation by
> eliminating conflicts between names (the 'name' property provides
> names like 'battery', whereas package-to-path provides names like
> '/foo/bar/battery at 0', which we stripped to 'battery at 0').  However,
> it also breaks of_device_id table matching.
> 
> The fix that we _really_ wanted was to keep dp->name based upon
> the name property ('battery'), but based dp->full_name upon
> package-to-path ('battery at 0').  This patch does just that.
> 
> Signed-off-by: Andres Salomon <dilinger at queued.net>
> Reported-by: Daniel Drake <dsd at laptop.org>

>From what I can tell, this only affects OLPC, correct?  It looks like
SPARC's implementation of of_pdt_node_name() will sidestep most of this
name retrieval code.

> ---
>  drivers/of/pdt.c |   42 +++++++++++++++++++-----------------------
>  1 files changed, 19 insertions(+), 23 deletions(-)
> 
> diff --git a/drivers/of/pdt.c b/drivers/of/pdt.c
> index 28295d0..b39d584 100644
> --- a/drivers/of/pdt.c
> +++ b/drivers/of/pdt.c
> @@ -48,7 +48,7 @@ static inline void irq_trans_init(struct device_node *dp) { }
>  
>  static inline const char *of_pdt_node_name(struct device_node *dp)
>  {
> -	return dp->name;
> +	return NULL;
>  }

Rather than using this hook; perhaps the sparc .path_component_name
should be enabled for all architectures.  It is a useful data item to
keep a pointer to, and it would simplify the of_pdt code.

>  
>  #endif /* !CONFIG_SPARC */
> @@ -156,23 +156,6 @@ static char * __init of_pdt_try_pkg2path(phandle node)
>  	return res+1;
>  }
>  
> -/*
> - * When fetching the node's name, first try using package-to-path; if
> - * that fails (either because the arch hasn't supplied a PROM callback,
> - * or some other random failure), fall back to just looking at the node's
> - * 'name' property.
> - */
> -static char * __init of_pdt_build_name(phandle node)
> -{
> -	char *buf;
> -
> -	buf = of_pdt_try_pkg2path(node);
> -	if (!buf)
> -		buf = of_pdt_get_one_property(node, "name");
> -
> -	return buf;
> -}
> -
>  static struct device_node * __init of_pdt_create_node(phandle node,
>  						    struct device_node *parent)
>  {
> @@ -187,7 +170,7 @@ static struct device_node * __init of_pdt_create_node(phandle node,
>  
>  	kref_init(&dp->kref);
>  
> -	dp->name = of_pdt_build_name(node);
> +	dp->name = of_pdt_get_one_property(node, "name");
>  	dp->type = of_pdt_get_one_property(node, "device_type");
>  	dp->phandle = node;
>  
> @@ -198,13 +181,26 @@ static struct device_node * __init of_pdt_create_node(phandle node,
>  	return dp;
>  }
>  
> -static char * __init of_pdt_build_full_name(struct device_node *dp)
> +static char * __init of_pdt_build_full_name(struct device_node *dp,
> +		phandle node)
>  {
>  	int len, ourlen, plen;
> +	const char *name;
>  	char *n;
>  
> +	/*
> +	 * When fetching the full name, rather than what we see with the
> +	 * name property (ie, 'battery'), we want the name we see with
> +	 * package-to-path (ie, 'battery at 0').
> +	 */
> +	name = of_pdt_node_name(dp);
> +	if (!name)
> +		name = of_pdt_try_pkg2path(node);
> +	if (!name)
> +		name = dp->name;
> +

Rather than this thrashing about looking for a way to get the path
component, it should be fairly certain how to obtain the node's name
before getting into of_pdt_build_full_name().  I'd follow SPARC's
lead and create an implementation of build_path_component();

However, considering that this is creating the full name; wouldn't the
raw output of pkg2path() provide exactly the full name string you need
here instead of breaking it down into components and adding it back up
again?

>  	plen = strlen(dp->parent->full_name);
> -	ourlen = strlen(of_pdt_node_name(dp));
> +	ourlen = strlen(name);
>  	len = ourlen + plen + 2;
>  
>  	n = prom_early_alloc(len);
> @@ -213,7 +209,7 @@ static char * __init of_pdt_build_full_name(struct device_node *dp)
>  		strcpy(n + plen, "/");
>  		plen++;
>  	}
> -	strcpy(n + plen, of_pdt_node_name(dp));
> +	strcpy(n + plen, name);
>  
>  	return n;
>  }
> @@ -243,7 +239,7 @@ static struct device_node * __init of_pdt_build_tree(struct device_node *parent,
>  #if defined(CONFIG_SPARC)
>  		dp->path_component_name = build_path_component(dp);
>  #endif
> -		dp->full_name = of_pdt_build_full_name(dp);
> +		dp->full_name = of_pdt_build_full_name(dp, node);
>  
>  		dp->child = of_pdt_build_tree(dp,
>  				of_pdt_prom_ops->getchild(node), nextp);
> -- 
> 1.7.2.3
>