Re: [Qemu-devel] [PATCH] monitor: intervally send down events to guest in hold time

[Luiz Capitulino]

On Mon, 22 Apr 2013 11:33:20 +0200
Paolo Bonzini <address@hidden> wrote:

> Il 22/04/2013 10:09, Amos Kong ha scritto:
> > On Mon, Apr 22, 2013 at 03:32:52PM +0800, Amos Kong wrote:
> >> On Sat, Apr 20, 2013 at 10:06:28AM -0600, Eric Blake wrote:
> >>> On 04/18/2013 10:44 PM, Amos Kong wrote:
> >>>> (qemu) sendkey a 1000
> >>>>
> >>>> Current design is that qemu only send one down event to guest,
> >>>> and delay sometime, then send one up event. In this case, only
> >>>> key can be identified by guest.
> >>>>
> >>>> This patch changed qemu to intervally send down events to guest
> >>>> in the hold time, the interval is 100ms.
> >>>
> >>> I don't like this.
> >>
> >>> When you hold a key for a long time on bare metal,
> >>> there is only one down and one up event;
> >>
> >> Really? I do check events by 'showkey', the output of showkey is not the
> >> events sent from keyboard?
> >>
> >> # showkey -s (show keys' scancode)
> >> I can always see many down scancodes, and one up scancode.
> >> It's same when I disable / enable auto-repeat mode in system.
> >>
> >> In the real host / vnc guest/ sdl guest, hold one key, many down
> >> events can be checked by showkey.
> >  
> > # watch cat /proc/interrupts
> >           CPU0       CPU1       CPU2       CPU3
> >  1:       1692      40309       1462       1795   IO-APIC-edge  i8042
> > 
> > hit a botton without long-time holding, interrupt count increased 2.
> > hit a botton with long-time holding, interrupt count increased a lot (more 
> > than 2)
> 
> You're right.  The typematic delay/rate is implemented within the i8042
> keyboard microcontroller (QEMU does not implement that register).
> 
> It is possible that software ignores interrupts for a key that is
> already down, and reimplements autorepeat in software, but your patch is
> correct.

But isn't this patch the equivalent of repeatedly pressing and releasing a
key? Shouldn't this be implemented at a lower-level layer like the input
subsystem?

Say, the input subsystem detects a key is being hold and asks the keyboard
emulation driver to keep sending interrupts for that key like Amos described?