On Thu, Feb 24, 2022 at 2:33 AM Greg Kroah-Hartman
<gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
>
> So, to follow the proposed solution in this thread, the following change
> is the "correct" one to make, right?
That would indeed be my preference.
If we ever are able to change
list_for_each_entry(tmp, &ep->queue, queue)...
to simply declare 'tmp' as the right type itself, then the e need to do
struct gr_request *tmp;
outside the loop goes away, and this kind of "set a variable that is
declared in the outside context" becomes the only way to do things.
Linus
