Re: [PATCH 0/3] Dell Airplane Mode Switch driver

["Pali Rohár" <pali.rohar@gmail.com>]

[-- Attachment #1: Type: Text/Plain, Size: 6397 bytes --]

On Monday 29 December 2014 08:27:21 Alex Hung wrote:
> On Fri, Dec 26, 2014 at 5:55 AM, Gabriele Mazzotta
> 
> <gabriele.mzt@gmail.com> wrote:
> > On Thursday 25 December 2014 21:11:05 Pali Rohár wrote:
> >> I will try to recap all information which we have...
> >> 
> >> *) We should not send wireless key press to userspace when
> >> BIOS already handles wireless state (and enable/disable
> >> wifi):
> >> http://www.spinics.net/lists/platform-driver-x86/msg05922.
> >> html
> >> 
> >> *) some tested dell machines does not implement GRBT method
> >> (or report constant value) which could return state of
> >> wireless (enabled/disabled) -- e.g. Inspiron 7447
> >> 
> >> *) dell-wireless driver is doing nothing on devices which
> >> have wireless slider switch (except calling CRBT/ARBT
> >> methods)
> >> 
> >> *) all tested machines emit key with keycode 240 (scancode
> >> is probably 136 = 0x88) to userspace via i8042 bus/AT
> >> keyboard when wireless button/slider is pressed/switched
> >> 
> >> *) both drivers dell-wireless and dell-rbtn do not
> >> implement setting soft rfkill state (or change wifi state)
> >> 
> >> So in my opinion: if we decide to use driver for acpi
> >> DELLABCE device we should use dell-rbnt for devices with
> >> hw slider switch and dell-wireless for devices with fn
> >> button. I think it does not make sense to use
> >> dell-wireless for devices with hw slider because it do
> >> nothing and dell-rbtn for devices with fn key button as
> >> GRBT does not working properly.
> > 
> > Here the problem. We are determining which laptop has an
> > hardware slider and which doesn't calling CRBT. If the
> > returned value is 2 or 3, then the laptop has an hardware
> > slider. However, it cannot be assumed the opposite if the
> > returned value is 0 or 1.
> 
> The spec defines as below:
> 1. When CRBT returns 0 or 1, a system has a toggle, ex.
> hotkey, radio button. 2. When CRBT returns 2 or 3, a system
> has a slider switch
> 
> The only difference of 0 and 1 (or 2 and 3) is the presence of
> a wireless LED. However I believe this is defined according
> to Microsoft's hardware requirement.
> 
> > See the acpidumps Alex uploaded. The Inspiron 3543 returns 0
> > when CRBT is called and so does the Inspiron 7447. However,
> > the former doesn't have a working ARBT method and, if I'm
> > not wrong, its BIOS doesn't handle radio devices. The BIOS
> > of the latter can handle radio and has a working ARBT
> > method to make it stop from doing that.
> > 
> > Since we determine when the BIOS might not be able to
> > control radio devices through CRBT and we can't say it for
> > sure (in the example above, CRBT returned the same value),
> > we make sure that the BIOS doesn't handle radio devices by
> > calling ARBT. We can ensure this (e.g. Inspirong 7447), but
> > we can't ensure the opposite (e.g. Inspiron 3453).
> > 
> > If there was a way to determine which laptop really needs
> > dell-wireless, calling ARBT wouldn't have been necessary,
> > but that's not the case. Users can always blacklist the
> > module if they know their laptop can work as expected
> > without it, but we have to ensure that everything always
> > works.
> > 
> > This is what I understood by looking at the acpidumps, so I
> > could be wrong.
> 
> I think ARBT acts as a switch for changing BIOS behaviours:
> 
> When ARBT is not called or is called with ARBT(0), BIOS's
> default behaviour is to change wireless states by hardware
> pin. When ARBT(1) is called by driver or userspace
> application (ex. required in Windows 8), wireless state is
> controlled by OS. Similar functions are defined in ASUS ATK
> and HP WMI. Such implementation will provide maximum
> capability for different OS with / without dedicate drivers.
> 
> The reality is never such wonderful; especially many systems
> are only tested with Win8 + driver (some may tested with Win7
> - and acpi_osi=!Windows 2012 / 2013 will probably work). In
> this case, we probably need to be more compatible with
> Windows 8 / 8.1.
> 
> >> And second note: Do we need some driver for acpi DELLABCE
> >> device? Which problem is trying acpi DELLABCE device to
> >> solve? Is not everything working fine without driver for
> >> DELLABCE device?
> > 
> > As I wrote above, DELLABCE is for those systems whose BIOS
> > doesn't handle radio devices and expects the OS do
> > everything when Fn keys are pressed.
> > (Well, it actually seems that something is done by the
> > keyboard controller, but this is not certain yet)
> > 
> >> My dell-rbtn approach is trying to export rfkill interface
> >> from DELLABCE device and eliminate using i8042 hook
> >> function in smbios dell-laptop driver.
> >> 
> >> Alex, can you check if scancode of wireless change
> >> generated by BIOS is on all machines same: 136 (0x88)? And
> >> is send by keyboard controller (not acpi/wmi)?
> > 
> > I can say that on my XPS13 the scancode is 0x88 and it is
> > sent by the keyboard controller.
> > 
> > Gabriele
> 
> Just before I check the scancode, I looked up the scancode
> table
> (http://download.microsoft.com/download/1/6/1/161ba512-40e2-4
> cc9-843a-923143f3456c/translate.pdf) and found 0x88 is already
> used - it is the "break" ("release" in Linux term) code for
> "7" (the one above Y/U, not number pad). It can be verified
> by "showkey -s": 0x08 for press and 0x88 for release. It is
> not the best idea to map this scancode to anything else.
> 
> I also sent emails for this scancode last week but it is a
> holiday season but I don't expect to receive any feedback
> this year...

Just to note, that there is only press key event with scancode 
136 (0x88). Release event is not sent to AT keyboard which cause 
problems (if you map this scancode to some well known keycode).

Alex, but I wrote you about this problem (in private email which 
you forwarded) together with other bios/fw problems for latitude 
machines...

Anyway I sent email to Matthew Garrett before Christmas and linux 
kernel could generate proper release event by fixing commit 
61579ba83934d397a4fa2bb7372de9ae112587d5 (adding also 9 and 10 to 
chassis_type). I believe Matthew Garrett will kernel provide 
workaround until Dell fix their BIOSes...

-- 
Pali Rohár
pali.rohar@gmail.com

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 198 bytes --]