On Sun, Sep 12, 2021 at 5:56 AM Linus Torvalds
<torvalds@xxxxxxxxxxxxxxxxxxxx> wrote:
>
> On Thu, Sep 9, 2021 at 6:56 AM Michael S. Tsirkin <mst@xxxxxxxxxx> wrote:
> >
> > NB: when merging this with
> > b542e383d8c0 ("eventfd: Make signal recursion protection a task bit")
> > from Linus' tree, replace eventfd_signal_count with
> > eventfd_signal_allowed, and drop the export of eventfd_wake_count from
> > ("eventfd: Export eventfd_wake_count to modules").
>
> What? No. That can't be right.
>
> Do you mean "replace eventfd_signal_count with !eventfd_signal_allowed()"?
>
> Because if I read the logic correctly, the issue is that
> 'vduse_vq_kick()' will call eventfd_signal().
>
> Which it must not do it eventfd_signal_allowed() returns false.
>
> So if eventfd_signal_allowed() is _not_ set, the code needs to defer
> it to the workqueue.
>
> No?
>
Yes, that's my fault. I just check how the eventfd_signal_allowed()
works in fs/aio.c (also needs a fix) when I see the fix for the merge
conflicts. Sorry for that.
Thanks,
Yongji
