Re: [RFC][PATCH] rcu: Use typeof(p) instead of typeof(*p) *

On Tue, 5 Oct 2021 15:40:29 -0400
Steven Rostedt [off-list ref] wrote:

struct trace_pid_list {
	unsigned long		ignore;
};

Rename the above struct trace_pid_list to struct trace_pid_internal.

And internally have:

union trace_pid_data {
	struct trace_pid_list		external;
	struct trace_pid_internal	internal;
};

Then use the internal version within the C file that modifies it, and just
return a pointer to the external part.

So this has proved to be a PITA.

That should follow the "C standard".

Really, thinking about abstraction, I don't believe there's anything wrong
with returning a pointer of one type, and then typecasting it to a pointer
of another type. Is there? As long as whoever uses the returned type does
nothing with it.

That is, if I simply do:

In the header file:

struct trace_pid_list {
	void *ignore;
};

struct trace_pid_list *trace_pid_list_alloc(void);
void trace_pid_list_free(struct trace_pid_list *pid_list);


And then in the C file:

struct pid_list {
	[...]
};

struct trace_pid_list *trace_pid_list_alloc(void)
{
	struct pid_list *pid_list;

	pid_list = kmalloc(sizeof(*pid_list), GFP_KERNEL);
	[..]

	return (struct trace_pid_list *)pid_list;
}

void trace_pid_list_free(struct trace_pid_list *list)
{
	struct pid_list *pid_list = (struct pid_list *)list;

	[..]
	kfree(pid_list);
}

That should be perfectly fine for standard C. Right?

-- Steve