> On May 5, 2023, at 10:06 AM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
>
> On Thu, 4 May 2023 12:19:38 -0400
> Joel Fernandes <joel@xxxxxxxxxxxxxxxxx> wrote:
>
>>> void f(void)
>>> {
>>> struct test *t1;
>>> struct test **t2 = &t1;
>>>
>>> list_for_each_entry_rcu((*t2), &testlist, node) { /* works */
>>> //...
>>> }
>>> list_for_each_entry_rcu(*t2, &testlist, node) { /* broken */
>>> //...
>>> }
>>
>> Yeah it is not clear why anyone would ever want to use it like this.
>> Why don't they just pass t1 to list_for_each_entry_rcu() ? I would
>> rather it break them and they re-think their code ;).
>
> Remember interfaces should not be enforcing policy unless it's key to the
> way the interface works.
Oh yeah, 100% agree. I am not particularly against this particular patch but I also dont see it as solving any problem. Feel free to Ack the patch if you feel strongly about wanting it.
- Joel
>
> -- Steve
