Re: [RFC][PATCH 2.6.11-rc3-mm2] Relay Fork Module

[Greg KH <greg@kroah.com>]

On Fri, Feb 11, 2005 at 04:08:40PM +0100, Guillaume Thouvenin wrote:
> +void kobject_fork(struct kobject *kobj, pid_t parent, pid_t child)
> +{
> +#ifdef CONFIG_KOBJECT_UEVENT

No, provide two different functions.  In a header file make it a static
inline function that does nothing if this option is not selected, so as
to make the code just go away, and the call is never made.


> +	char *kobj_path = NULL;
> +	char *action_string = NULL;
> +	char **envp = NULL;
> +	char ppid_string[FORK_BUFFER_SIZE];
> +	char cpid_string[FORK_BUFFER_SIZE];
> +
> +	if (!uevent_sock)
> +		return;
> +	
> +	action_string = action_to_string(KOBJ_FORK);
> +	if (!action_string)
> +		return;
> +	
> +	kobj_path = kobject_get_path(kobj, GFP_KERNEL);
> +	if (!kobj_path)
> +		return;

How is there a path for a kobject that is never registered with sysfs?

I agree with Andrew, why are you using a kobject for this?  Have you
looked at the "connector" code that is in the -mm tree?  That might be a
better solution for this, and it will be going into the kernel tree
after 2.6.11 is released.

> +EXPORT_SYMBOL(kobject_fork);

EXPORT_SYMBOL_GPL() for something like this please.

thanks,

greg k-h