Re: hp-wmi and hp dv9000

Re: hp-wmi and hp dv9000

[Alan Jenkins]

Peter Chubb wrote:
> Hi Alan.
> 	I noticed you were last working on hp-wmi.  I'm finding that
> if I load hp-wmi then the wireless network does not work --- dmesg
> output always says `iwlagn 0000:02:00: Radio disabled by HW RF Kill switch' 
> regardless of the position of the switch.
>
>
> This machine has no bluetooth, just the IWL-AGN mini-pci card.
> I've been testing with versions from 2.6.32 to 2.6.33-rc7, none work
> with hp-wmi loaded.
>
> I'm happy to provide traces, logs, etc., if that'll help --- just ask
> for what you want.
>   

Thanks Peter.  I do have some worries about my patch that went into 
2.6.33-rc1.  What I can't see how this happens on 2.6.32, unless your 
laptop completely inverts the hp ACPI interface.

I'd like to look at the output of "acpidump", but that may not help.  
(Other acpidumps I've seen just pass the calls through to the BIOS via SMI).


Also try running rfkill 
(http://wireless.kernel.org/en/users/Documentation/rfkill) on 2.6.33-rc1 
or later.  What do you see with -

"rfkill list" when your wireless is working, before you load hp-wmi.

"rfkill event" - this will stay running and log rfkill changes.  What 
does it show when you load hp-wmi?

"rfkill unblock wlan" after loading hp-wmi.  Does it get the wireless 
working?  What does "rfkill list" show afterwords?


If acpidump doesn't tell me anything, the rfkill details might help with 
crafting a work-around.


Regards
Alan

Re: hp-wmi and hp dv9000

[Peter Chubb]

>>>>> "Alan" == Alan Jenkins <alan-jenkins@tuffmail.co.uk> writes:

Alan> Peter Chubb wrote:
>> Hi Alan.  I noticed you were last working on hp-wmi.  I'm finding
>> that if I load hp-wmi then the wireless network does not work ---
>> dmesg output always says `iwlagn 0000:02:00: Radio disabled by HW
>> RF Kill switch' regardless of the position of the switch.

I've discovered that if I load the rfkill module *before* hp-wmi then
everything works properly.  Maybe there's a missing dependency?  I
hadn't even been building the rfkill input module before --- it
doesn't get automatically selected.

Peter C


--
Dr Peter Chubb  http://www.gelato.unsw.edu.au  peterc AT gelato.unsw.edu.au
http://www.ertos.nicta.com.au           ERTOS within National ICT Australia

Re: hp-wmi and hp dv9000

[Alan Jenkins]

On 2/10/10, Peter Chubb <peter.chubb@nicta.com.au> wrote:
>>>>>> "Alan" == Alan Jenkins <alan-jenkins@tuffmail.co.uk> writes:
>
> Alan> Peter Chubb wrote:
>>> Hi Alan.  I noticed you were last working on hp-wmi.  I'm finding
>>> that if I load hp-wmi then the wireless network does not work ---
>>> dmesg output always says `iwlagn 0000:02:00: Radio disabled by HW
>>> RF Kill switch' regardless of the position of the switch.
>
> I've discovered that if I load the rfkill module *before* hp-wmi then
> everything works properly.

Can you show me what "udevadm monitor -ke" reports in both cases?  Use
one of the 2.6.33-rc kernels, and show me your kernel config.

> Maybe there's a missing dependency?

If you've built the rfkill module, "modprobe hp-wmi" should load it
automatically.  If it didn't, then hp-wmi would fail to load.  So it
sounds more like a race condition.

> I
> hadn't even been building the rfkill input module before --- it
> doesn't get automatically selected.

Um.  rfkill-input is no longer a separate module.  You can disable it
individually, *if* you have CONFIG_EMBEDDED enabled.  If that causes
problems, we need to fix that as well (because rfkill-input is
theoretically going away).

Or perhaps you have rfkill as a whole disabled?  In that case, we'd
expect that hp-wmi wouldn't have any effect on the wireless.  If it
does, that's also a bug (albeit a more obscure one with an easy
workaround).

Regards
Alan

Re: hp-wmi and hp dv9000

[Peter Chubb]

Hi Alan.	
	I dunno what's happened, but I can no longer get the problem
to appear.  I've spent half the day today recompiling different
kernels with different configs -- some that I *know* didn't work
before -- and now they work.

	The only thing I can think of (and I can't see in the code
where it could have happened) is that somehow getting the kill switch
to work reset something in the NVRAM, and now it works all the time,
regardless of whether rfkill and hp-wmi are in or out of the kernel in
any combination.
But that's far fetched, and I'm not sure I believe it.


Sorry about this -- I really hate bugs that disappear after I start to
investigate them, especially if I don't understand why.
--
Dr Peter Chubb		www.nicta.com.au    peter DOT chubb AT nicta.com.au
http://www.ertos.nicta.com.au           ERTOS within National ICT Australia