On Wed, Mar 02, 2016 at 01:35:04PM +0100, Michał Kępień wrote:
> > On Mon, Feb 29, 2016 at 09:56:03PM +0100, Pali Rohár wrote:
> > > On Monday 29 February 2016 21:49:27 you wrote:
> > > > > On Monday 29 February 2016 21:31:23 Michał Kępień wrote:
I understand this better now, thank you for the additional explanation.
> > > > > > > > diff --git a/drivers/platform/x86/dell-wmi.c
> > > > > > > > b/drivers/platform/x86/dell-wmi.c index 65edd93..ffc957b5
> > > > > > > > 100644 --- a/drivers/platform/x86/dell-wmi.c
> > > > > > > > +++ b/drivers/platform/x86/dell-wmi.c
> > > > > > > > @@ -111,7 +111,7 @@ static const struct key_entry
> > > > > > > > dell_wmi_legacy_keymap[] __initconst = {
> > > > > > > >
> > > > > > > > { KE_IGNORE, 0xe020, { KEY_MUTE } },
> > > > > > > >
> > > > > > > > /* Shortcut and audio panel keys */
> > > > > > > >
> > > > > > > > - { KE_IGNORE, 0xe025, { KEY_RESERVED } },
> > > > > > > > + { KE_KEY, 0xe025, { KEY_PROG4 } },
> > > > > > > >
Your description below helped explain why the KE_KEY change was necessary, the
commit message didn't do that for me. Just explicitly stating "when there is no
i8042 interrupt, the WMI even must generate a valid KE_KEY" or something along
those lines would help.
> > > > > > > > { KE_IGNORE, 0xe026, { KEY_RESERVED } },
> > > > > > > >
> > > > > > > > { KE_IGNORE, 0xe02e, { KEY_VOLUMEDOWN } },
> > > > > > > >
> > > > > > > > @@ -235,6 +235,9 @@ static void dell_wmi_process_key(int
> > > > > > > > reported_key)
> > > > > > > >
> > > > > > > > acpi_video_handles_brightness_key_presses())
> > > > > > > >
> > > > > > > > return;
> > > > > > > >
> > > > > > > > + if (key->keycode == KEY_PROG4 &&
> > > > > > > > !wmi_requires_smbios_request) + return;
> > > > > > > > +
> > > > > > >
> > > > > > > Here I would rather test against reported_key, not keycode. If
> > > > > > > somebody in future adds KEY_PROG4 for something else we will
> > > > > > > have problem...
And ultimately, that is under our control. So let's just not do that :-)
A comment by the definition of KEY_PROG4 that notes it's meaning in this driver
should prevent any future attempts at overloading it and breaking this.
> > > > > >
> > > > > > As 0xe025 is currently the only event we know about that should
> > > > > > be ignored on some machines and processed on others, this makes
> > > > > > sense,
> > > > > >
> > > > > > at least for now. If I change the first condition to:
> > > > > > reported_key == 0xe025
> > > > >
> > > > > There will be need also change for 5/5 patch...
> > > >
> > > > Why? Are you aware of any model which sends a 0xe029 WMI event _and_
> > > > generates an i8042 interrupt? If not, WMI event 0xe029 should always
> > > > be turned into a key event, as per the keymap.
> > >
> > > No, but your current patch 4/5 and 5/5 do that (because it checks
> > > KEY_PROG4). But if it is not needed, I'm happy because of one hook less.
> >
> > From my reading, patch 5/5 adds 0xe029 to the reported keys that need to be
> > ignored, so the test would need to include both if it isn't using the common
> > keycode KEY_PROG4. I believe that is what Pali is saying as well. Is this not
> > the correct reading of 5/5?
>
> It's the other way round :) Perhaps explaining the issue once again
> will help.
Got it! Thanks! A couple of comments and I think this is a reasonable solution.
>
> Until this patch series, dell-wmi was only "aware" of laptops which
> generate _both_ an i8042 interrupt and a WMI event when Dell Instant
> Launch is pressed. Thus, as the i8042 interrupt already caused a key
> event to be generated, there was no point in generating another one for
> the WMI event, hence the KE_IGNORE entry for 0xe025 in the keymap.
>
> Enter Vostro V131, which does _not_ generate an i8042 interrupt when
> Dell Instant Launch is pressed, yet still generates a WMI event. In
> other words, there was no way of generating a key event for Dell Instant
> Launch on that model without changing the relevant keymap entry into a
> KE_KEY one. However, I still needed to ensure that for most machines
> that WMI event would _not_ be turned into a key event. That's why patch
> 4/5 changes the keymap entry to a KE_KEY one, while also adding a
> conditional return to the key processing function, thus making sure that
> the previous behavior (ignoring WMI event 0xe025) is preserved on most
> machines.
>
> I used KEY_PROG4 in the conditional expression because a mapped keycode
> is also used in a comparison just above the code inserted by patch 4/5.
> For event 0xe029, however, the conditional expression added by patch 4/5
> should always evaluate to false. Note that if we use KEY_PROG4 in the
> comparison, the second condition will be false; if we use 0xe025 in the
> comparison, the first condition will be false. The latter variant
> (suggested by Pali) will work fine until we hear of a model which
> generates _both_ WMI event 0xe029 and an i8042 interrupt upon pressing
> some hotkey.
>
> I was hoping the commit message for patch 4/5 would be clear enough to
> convey my intent, yet it seems I managed to confuse you. Maybe that's a
> sign that the commit message and/or code should be rephrased, but I
> can't be a judge on this one myself. Let me know what you think. Pali,
> if you think this is good enough as it is, please add your Reviewed-by.
> If you have doubts, I'm all ears.
>
> --
> Best regards,
> Michał Kępień
>
--
Darren Hart
Intel Open Source Technology Center
--
To unsubscribe from this list: send the line "unsubscribe platform-driver-x86" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
