[PATCH -mm] kobject_add error handling when set name failed

[PATCH -mm] kobject_add error handling when set name failed

[Dave Young]

In kobject_add, when kobject_set_name failed we should put the kobj and return.

Signed-off-by: Dave Young <hidave.darkstar@gmail.com> 

---
lib/kobject.c |   11 ++++++++---
1 file changed, 8 insertions(+), 3 deletions(-)

diff -upr linux/lib/kobject.c linux.new/lib/kobject.c
--- linux/lib/kobject.c	2007-12-14 17:14:03.000000000 +0800
+++ linux.new/lib/kobject.c	2007-12-14 17:17:48.000000000 +0800
@@ -228,8 +228,13 @@ int kobject_add(struct kobject * kobj)
 
 	if (!(kobj = kobject_get(kobj)))
 		return -ENOENT;
-	if (!kobj->k_name)
-		kobject_set_name(kobj, "NO_NAME");
+	if (!kobj->k_name) {
+		error = kobject_set_name(kobj, "NO_NAME"));
+		if (error) {
+			kobject_put(kobj);
+			return error;
+		}
+	}
 	if (!*kobj->k_name) {
 		pr_debug("kobject (%p) attempted to be registered with no "
 			 "name!\n", kobj);
@@ -371,7 +376,7 @@ void kobject_init_ng(struct kobject *kob
 		goto error;
 	}
 	if (!ktype) {
-		err_str = "must have a ktype to be initialized properly!\n";
+		err_str = "must have a ktype to be initialized properly!";
 		goto error;
 	}
 	if (atomic_read(&kobj->kref.refcount)) {

Re: [PATCH -mm] kobject_add error handling when set name failed

[Greg KH]

On Fri, Dec 14, 2007 at 05:24:05PM +0800, Dave Young wrote:
> In kobject_add, when kobject_set_name failed we should put the kobj and return.
> 
> Signed-off-by: Dave Young <hidave.darkstar@gmail.com> 
> 
> ---
> lib/kobject.c |   11 ++++++++---
> 1 file changed, 8 insertions(+), 3 deletions(-)
> 
> diff -upr linux/lib/kobject.c linux.new/lib/kobject.c
> --- linux/lib/kobject.c	2007-12-14 17:14:03.000000000 +0800
> +++ linux.new/lib/kobject.c	2007-12-14 17:17:48.000000000 +0800
> @@ -228,8 +228,13 @@ int kobject_add(struct kobject * kobj)
>  
>  	if (!(kobj = kobject_get(kobj)))
>  		return -ENOENT;
> -	if (!kobj->k_name)
> -		kobject_set_name(kobj, "NO_NAME");
> +	if (!kobj->k_name) {
> +		error = kobject_set_name(kobj, "NO_NAME"));
> +		if (error) {
> +			kobject_put(kobj);
> +			return error;
> +		}
> +	}

Thanks, but I have a pending patch from Kay in my inbox that messes up
this patch, as it drops most of this nonsense :)

>  	if (!*kobj->k_name) {
>  		pr_debug("kobject (%p) attempted to be registered with no "
>  			 "name!\n", kobj);
> @@ -371,7 +376,7 @@ void kobject_init_ng(struct kobject *kob
>  		goto error;
>  	}
>  	if (!ktype) {
> -		err_str = "must have a ktype to be initialized properly!\n";
> +		err_str = "must have a ktype to be initialized properly!";
>  		goto error;

Oops, good catch, I'll just go add this change to my tree.

thanks,

greg k-h