[PATCH] ppc64: SMU partitions & device-tree

[Nathan Lynch]

Benjamin Herrenschmidt wrote:

> In order to update the device-tree, I made changes to
> prom_add_property() so that it now updates the proc entries as well
> (and gets some locking too while we are at it). I also removed the
> code in proc_devtree.c that tried to enforce the order of the
> properties & nodes as it really didn't look good and shouldn't be
> necessary.
> -void
> +int
>  prom_add_property(struct device_node* np, struct property* prop)
>  {
> -	struct property **next = &np->properties;
> +	struct property **next;
>  
>  	prop->next = NULL;	
> -	while (*next)
> +	write_lock(&devtree_lock);
> +	next = &np->properties;
> +	while (*next) {
> +		if (strcmp(prop->name, (*next)->name) == 0) {
> +			/* duplicate ! don't insert it */
> +			write_unlock(&devtree_lock);
> +			return -1;
> +		}
>  		next = &(*next)->next;
> +	}
>  	*next = prop;
> +	write_unlock(&devtree_lock);
> +
> +	/* try to add to proc as well if it was initialized */
> +	if (np->pde)
> +		proc_device_tree_add_prop(np->pde, prop);
> +
> +	return 0;
>  }

I think this is abusing devtree_lock, which is meant to protect
traversal of the device tree and not the contents of nodes.  True,
this serializes prom_add_property itself, and will prevent corruption
of the node's property list.  But I think I'd rather the SMU code take
care of this itself, i.e. use its own mutex to serialize the addition
of properties to /smu.