On Fri, Apr 24, 2020 at 3:11 AM Roman Penyaev <rpenyaev@suse.de> wrote:
>
> Hi Khazhismel,
>
> That seems to be correct. The patch you refer 339ddb53d373
> relies on callback path, which *should* wake up, not the path
> which harvests events (thus unnecessary wakeups).  When we add
> a new event to the ->ovflist nobody wakes up the waiters,
> thus missing wakeup. You are right.
>
> May I suggest a small change in order to avoid one new goto?
> We can add a new event in either ->ovflist or ->rdllist and
> then wakeup should happen. So simple 'else if' branch should
> do things right, something like the following:
>

Thanks for the review! I agree, I'll send a v2 without new goto