Re: [PATCH 0/3] New input patches

[Dmitry Torokhov]

On Tuesday 21 September 2004 07:10 am, Vojtech Pavlik wrote:
> On Thu, Sep 16, 2004 at 11:58:27PM -0500, Dmitry Torokhov wrote:
> >           I think that command processing is now race free - instead of using
> >           bit operations on flags the ps2_command and ps2_send_byte simply
> >           take serio->lock (via serio_pause/continue_rx). Since serio->lock
> >           is also taken by interrupt handler anyway it gives us desired
> >           serialization. As wakeup routines take a spinlock as well and
> >           spinlock is guaranteed to be a barrier we should not miss wake up
> >           events either.
> 
> I hope the wait_event* functions also use memory barriers properly, but
> they probably must, otherwise they won't be useful, because there we're
> accessing the flags variable without a lock.

prepare_to_wait does spin_lock/spin_unlock on when adding a task to a wait queue
and (I hope someone corrects me if I'm wrong) spin lock/unlock are guaranteed to
be memory barriers.

> -       input_event(&atkbd->dev, EV_MSC, MSC_RAW, code);
> +       if (atkbd->softraw)
> +               input_event(&atkbd->dev, EV_MSC, MSC_RAW, code);
> 
> ... we definitely want the RAW codes to be sent when we're not in
> softraw mode, because that's when they're passed through keyboard.c to
> the console.
> 
> So the condition needs to be inverted. However, it's not necessary at
> all, since the input layer will not pass the RAW events when the MSC_RAW
> bit is not set.

I see, my bad. I will drop that bit.

Thanks for the comments!

-- 
Dmitry
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/