[PATCH 3/4][Trivial] lib/kobject.c : Check ktype in kobject_add_internal

[PATCH 3/4][Trivial] lib/kobject.c : Check ktype in kobject_add_internal

[yan]

If someone calls kobject_init with null kobj_type, kobject_init can 
detect it. But it does not have a return value, it just dumps stack.
In this case, a kobject_add after that will result in:
   -->kobject_add_varg
    -->kobject_add_internal
     -->create_dir
      -->sysfs_create_dir
       -->sysfs_read_ns_type
        -->kobj_child_ns_ops
In the end, it will call ->ktype->child_ns_type directly and we have
an oops.

Check ktype before really add a kobject.

Signed-off-by: Yan Hong <clouds.yan@gmail.com>
---
 lib/kobject.c |    3 +++
 1 files changed, 3 insertions(+), 0 deletions(-)

diff --git a/lib/kobject.c b/lib/kobject.c
index 85d5e9b..3ba6216 100644
--- a/lib/kobject.c
+++ b/lib/kobject.c
@@ -164,6 +164,9 @@ static int kobject_add_internal(struct kobject *kobj)
 	if (!kobj)
 		return -ENOENT;
 
+	if (!kobj->ktype)
+		return -EINVAL;
+
 	if (!kobj->name || !kobj->name[0]) {
 		WARN(1, "kobject: (%p): attempted to be registered with empty "
 			 "name!\n", kobj);
-- 
1.7.5.1

Re: [PATCH 3/4][Trivial] lib/kobject.c : Check ktype in kobject_add_internal

[greg]

On Sat, Apr 21, 2012 at 05:58:31PM +0800, yan wrote:
> If someone calls kobject_init with null kobj_type, kobject_init can 
> detect it. But it does not have a return value, it just dumps stack.
> In this case, a kobject_add after that will result in:
>    -->kobject_add_varg
>     -->kobject_add_internal
>      -->create_dir
>       -->sysfs_create_dir
>        -->sysfs_read_ns_type
>         -->kobj_child_ns_ops
> In the end, it will call ->ktype->child_ns_type directly and we have
> an oops.
> 
> Check ktype before really add a kobject.

But how can we add a kobject without a valid ktype?

At the most, I'd prefer to warn about this, much like we warn about not
having a name, and not just returning an error without a hint as to what
went wrong.

greg k-h