[BUG] brcmsmac: Invalid killswitch handling

[BUG] brcmsmac: Invalid killswitch handling

[Sebastian Poehn]

I had a discussion with Matthew Garrett about handling of a wifi
killswitch on a Dell laptop.

Some (rfkill) code was removed from platform/dell-laptop. On my machine
a bug now occurs which is caused by brcmsmac in the view of Matthew.

So is this a bug, or do you have no killsitch handling, so far?

Bug description:
> physical switch is OFF
> rfkill state is ON
> => Laptop hangs on boot (maybe because a physical ascent device is
> used by network subsystem)
> 
> physical switch is OFF
> rfkill state is OFF
> => Activate interface in Gnome Connection Manager => System hang
> 
> test setup:
> Dell Vostro 3300
> Broadcom 4313 Wifi
> Kernel 3.5.[0-2] & 3.6
> Fedora 17


What Matthew wrote to me:
> > With the old dell-laptop code included everything was fine with
> > brcmsmac. So either generic platform code does not handle the switch
> > correct or there is no handling at all?
> 
> brcmsmac is broken. I suspect it was happily claiming to be
> unblocked, 
> but was ignored because dell-laptop claimed to be blocked.

Full discussion thread:
https://lkml.org/lkml/2012/8/23/234

Re: [BUG] brcmsmac: Invalid killswitch handling

[Arend van Spriel]

On 08/24/2012 10:51 AM, Sebastian Poehn wrote:
> I had a discussion with Matthew Garrett about handling of a wifi
> killswitch on a Dell laptop.
>
> Some (rfkill) code was removed from platform/dell-laptop. On my machine
> a bug now occurs which is caused by brcmsmac in the view of Matthew.
>
> So is this a bug, or do you have no killsitch handling, so far?

brcmsmac handled the rfkill switch properly so this would be a 
regression. I also have Dell laptops here with a physical rfkill switch. 
What rfkill switch do you have?

Anyway, I will test over here on my hardware as this regression has been 
flagged internally here as well. Would you be willing to test a fix on 
your system when I have one?

Gr. AvS

Re: [BUG] brcmsmac: Invalid killswitch handling

[Sebastian Poehn]

On Fri, 2012-08-24 at 12:59 +0200, Arend van Spriel wrote:
> brcmsmac handled the rfkill switch properly so this would be a 
> regression. I also have Dell laptops here with a physical rfkill switch. 
> What rfkill switch do you have?

I have no idea how to detect the type of the switch (looks like it's
neither connected via usb nor pci). That's what rfkill tells me:

[sebastian@localhost ~]$ rfkill list
0: dell-wifi: Wireless LAN
	Soft blocked: yes
	Hard blocked: yes
1: phy0: Wireless LAN
	Soft blocked: no
	Hard blocked: no

It is a Vostro 3300 with a physical slide switch on the front being
either black and enabled or red and disabled.

> 
> Anyway, I will test over here on my hardware as this regression has been 
> flagged internally here as well. Would you be willing to test a fix on 
> your system when I have one?

Would be a joy for me