Comment 13 for bug 1313196

I see, thanks Steffen for your explanation.

I did some better tests. With acpi_osi='!Windows 2012' and the default 3.11.0-12-generic kernel, when I deactivate wifi using the switch button I get the following output:

0: phy0: Wireless LAN
        Soft blocked: no
        Hard blocked: no
1: hp-wifi: Wireless LAN
        Soft blocked: no
        Hard blocked: yes
2: hp-bluetooth: Bluetooth
        Soft blocked: no
        Hard blocked: yes

That is: the phy0 entry is not blocked (sorry for the wrong information i gave previously), but the hp-bluetooh entry appears as hard blocked, while the hci0 has gone completely.

This said, I tested the following mainline kernel versions (without using acpi_osi='!Windows 2012') as suggested by Christopher. I did these tests directly on my Linux Mint 16 KDE installation, because installing a mainline kernel version on the live USB stick of Ubuntu 14.04 did not work. Here are the results:
- with 3.12-rc7-saucy (the latest one built specifically for Saucy, which Mint 16 is based on) the problem is still present (the wireless witch button does not work)
- with 3.15-rc2-trusty (the latest one built specifically for Trusty) the problem is fixed (the wireless switch button does work)

However, with 3.15-rc2 without the acpi_osi='!Windows 2012' boot parameter, the output of "rfkill list" is:
0: phy0: Wireless LAN
        Soft blocked: no
        Hard blocked: no
1: hci0: Bluetooth
        Soft blocked: no
        Hard blocked: no
and when the wifi is disabled it is:
0: phy0: Wireless LAN
        Soft blocked: yes
        Hard blocked: no
1: hci0: Bluetooth
        Soft blocked: yes
        Hard blocked: no

That is: the hp-wifi and hp-bluetooth entries are missing and the switch causes a soft-block rather than a hard block. I would say the most correct behaviour is the one I have with acpi_osi='!Windows 2012' (i.e.: the system sees the hp-* entries and the button performs a hard block), but I'm not an expert. After all, the final result from a user point of view is the same (the button works and the led on it reflects the current enabling state of the wifi card).

I still don't know what acpi_osi='!Windows 2012' is really meant to be and whether there may be any other side effects when using it.