From: Zijun Hu <quic_zijuhu@quicinc.com>

__of_add_property() compares property name by strcmp(), and that is
improper for SPARC which wants strcasecmp().

Fix by using dedicated property name comparison macro of_prop_cmp().

Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>
---
 drivers/of/base.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/of/base.c b/drivers/of/base.c
index 001ff6ce4abf85c07f13649d5a9f691f549a8ccc..c810014957e81171675b63f25eaabe391cc903e4 100644
--- a/drivers/of/base.c
+++ b/drivers/of/base.c
@@ -1651,7 +1651,7 @@ int __of_add_property(struct device_node *np, struct property *prop)
 	prop->next = NULL;
 	next = &np->properties;
 	while (*next) {
-		if (strcmp(prop->name, (*next)->name) == 0) {
+		if (of_prop_cmp(prop->name, (*next)->name) == 0) {
 			/* duplicate ! don't insert it */
 			rc = -EEXIST;
 			goto out_unlock;

-- 
2.34.1

On Mon, Feb 24, 2025 at 10:27:59PM +0800, Zijun Hu wrote:
> From: Zijun Hu <quic_zijuhu@quicinc.com>
> 
> __of_add_property() compares property name by strcmp(), and that is
> improper for SPARC which wants strcasecmp().

Except that 'name' is the nodename (usually, with a few rare 
exceptions). Sparc node names are case sensitive, so strcmp was correct. 

My hope is to get rid of case insensitive comparisions, so if nothing 
cares that we're doing a case sensitive comparision, I want to keep 
that. 

I also want to get rid of storing both 'name' as a property and 
device_node.name. The name property is an ABI issue though if no one is 
looking, then it's not an ABI issue. Also, we should be able to generate 
device_node.name from device_node.full_name. There's still a bunch of 
direct users of device_node.name which have to be fixed. Mostly in clock 
drivers from what I remember.

> Fix by using dedicated property name comparison macro of_prop_cmp().
> 
> Signed-off-by: Zijun Hu <quic_zijuhu@quicinc.com>
> ---
>  drivers/of/base.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/of/base.c b/drivers/of/base.c
> index 001ff6ce4abf85c07f13649d5a9f691f549a8ccc..c810014957e81171675b63f25eaabe391cc903e4 100644
> --- a/drivers/of/base.c
> +++ b/drivers/of/base.c
> @@ -1651,7 +1651,7 @@ int __of_add_property(struct device_node *np, struct property *prop)
>  	prop->next = NULL;
>  	next = &np->properties;
>  	while (*next) {
> -		if (strcmp(prop->name, (*next)->name) == 0) {
> +		if (of_prop_cmp(prop->name, (*next)->name) == 0) {
>  			/* duplicate ! don't insert it */
>  			rc = -EEXIST;
>  			goto out_unlock;
> 
> -- 
> 2.34.1
>

On 2025/2/25 22:38, Rob Herring wrote:
> On Mon, Feb 24, 2025 at 10:27:59PM +0800, Zijun Hu wrote:
>> From: Zijun Hu <quic_zijuhu@quicinc.com>
>>
>> __of_add_property() compares property name by strcmp(), and that is
>> improper for SPARC which wants strcasecmp().
> Except that 'name' is the nodename (usually, with a few rare 
> exceptions). Sparc node names are case sensitive, so strcmp was correct. 

Here, it is property's name and NOT node's name.

arch/sparc/include/asm/prom.h:

#define of_compat_cmp(s1, s2, l) strncmp((s1), (s2), (l))
#define of_prop_cmp(s1, s2)	strcasecmp((s1), (s2)) //SPARC HERE
#define of_node_cmp(s1, s2)	strcmp((s1), (s2))


include/linux/of.h:
/* Default string compare functions, Allow arch asm/prom.h to override */
#if !defined(of_compat_cmp)
#define of_compat_cmp(s1, s2, l) strcasecmp((s1), (s2))
#define of_prop_cmp(s1, s2)	strcmp((s1), (s2))      // others HERE
#define of_node_cmp(s1, s2)	strcasecmp((s1), (s2))
#endif

actually, later __of_update_property() and __of_find_property() use
of_prop_cmp() instead of strcmp() to compare property name.

perhaps, compatible and node name also have similar issues in current
OF codes.

On Tue, Feb 25, 2025 at 11:04:55PM +0800, Zijun Hu wrote:
> On 2025/2/25 22:38, Rob Herring wrote:
> > On Mon, Feb 24, 2025 at 10:27:59PM +0800, Zijun Hu wrote:
> >> From: Zijun Hu <quic_zijuhu@quicinc.com>
> >>
> >> __of_add_property() compares property name by strcmp(), and that is
> >> improper for SPARC which wants strcasecmp().
> > Except that 'name' is the nodename (usually, with a few rare 
> > exceptions). Sparc node names are case sensitive, so strcmp was correct. 
> 
> Here, it is property's name and NOT node's name.

Sigh, indeed... Applied.

Rob