On Sat, May 27, 2017 at 4:01 AM, Pali Rohár <pali.rohar@xxxxxxxxx> wrote:
> On Saturday 27 May 2017 07:16:19 Darren Hart wrote:
>> From: Andy Lutomirski <luto@xxxxxxxxxx>
>>
>> According to Mario at Dell, the DELLABC6 device should not be used on
>> a Linux system. It also conflicts with Intel-HID and its
>> interactions with Network Manager. Document that we are aware of the
>> device, but that we are intentionally ignoring it.
>>
>> Signed-off-by: Andy Lutomirski <luto@xxxxxxxxxx>
>> [dvhart: New commit message and minor comment wording fixes]
>> Cc: Mario Limonciello <mario_limonciello@xxxxxxxx>
>> Cc: "Pali Rohár" <pali.rohar@xxxxxxxxx>
>> Signed-off-by: Darren Hart (VMware) <dvhart@xxxxxxxxxxxxx>
>> ---
>> drivers/platform/x86/dell-rbtn.c | 26 +++++++++++++++++++-------
>> 1 file changed, 19 insertions(+), 7 deletions(-)
>>
>> diff --git a/drivers/platform/x86/dell-rbtn.c
>> b/drivers/platform/x86/dell-rbtn.c index dcd9f40..2eeef03 100644
>> --- a/drivers/platform/x86/dell-rbtn.c
>> +++ b/drivers/platform/x86/dell-rbtn.c
>> @@ -223,14 +223,26 @@ static const struct acpi_device_id rbtn_ids[] =
>> { * This driver can also handle the "DELLABC6" device that
>> * appears on the XPS 13 9350, but that device is disabled
>> * by the DSDT unless booted with acpi_osi="!Windows 2012"
>> - * acpi_osi="!Windows 2013". Even if we boot that and bind
>> - * the driver, we seem to have inconsistent behavior in
>> - * which NetworkManager can get out of sync with the rfkill
>> - * state.
>> + * acpi_osi="!Windows 2013".
>> *
>> - * On the XPS 13 9350 and similar laptops, we're not supposed to
>> - * use DELLABC6 at all. Instead, we handle the rfkill button
>> - * via the intel-hid driver.
>> + * According to Mario at Dell:
>> + *
>> + * DELLABC6 is a custom interface that was created solely to
>> + * have airplane mode support for Windows 7. For Windows 10
>> + * the proper interface is to use that which is handled by
>> + * intel-hid. A OEM airplane mode driver is not used.
>> + *
>> + * Since the kernel doesn't identify as Windows 7 it would be
>> + * incorrect to do attempt to use that interface.
>> + *
>> + * Even if we override _OSI and bind to DELLABC6, we end up
>> + * with inconsistent behavior in which NetworkManager can get
>> + * out of sync with the rfkill state. This happens because
>> + * NetworkManager receives events from intel-hid and fights with
>> + * dell-rbtn for control.
>> + *
>> + * The upshot is that it's better to just ignore DELLABC6
>> + * devices.
>> */
>>
>> { "", 0 },
>
> Just one note: Kernel code should not depend on one particular software
> which implements networking (in userspace). Either behaviour is
> independent of used software and therefore comment does not apply only
> to Network Manager OR behaviour is strictly bounded to Network Manager
> which is IMHO not a kernel bug, but rather userspace software
> application bug. If there is a bug in userspace, then userspace should
> be fixed instead of adding hacks/workarounds in kernel.
Fair enough. NetworkManager is just an example here. The general
kernel behavior is that, if the kernel sends KEY_RFKILL or similar,
that means "the button was pressed and it's up to userspace to handle
it". Sending KEY_RFKILL *and* handling it in the kernel is not going
to go well. This should be true with any other reasonably modern
userspace (connmgr or whatever it's called, perhaps?).
