[Regression] - 2.6.32-rc5 hangs using rfkill on an eeePC

[Regression] - 2.6.32-rc5 hangs using rfkill on an eeePC

[Fabio Comolli]

Hi all.
Pressing rfkill button on my eeePC 900 to disable wireless hangs the system.

In console I can see (warning: what follows is copied by hand):

acpid: client 148[6:100] has disconnected
ACPI group/action not found: hotkey / ATKD
unregister netdevice: waiting for wlan0 to become free; usage count=1

(last line repeats forever). Hard reset is needed.

With any previous kernel version (I'm running -rc4 right now) everything works:

ACPI group/action undefined: hotkey / ATKD
ath5k 0000:01:00.0: PCI INT A disabled

(there's no acpid line)


Regards,
Fabio

Re: [Regression] - 2.6.32-rc5 hangs using rfkill on an eeePC

[Fabio Comolli]

Actually it was pretty easy to find the culprit, as thereare only two
eeepc-related commit between -rc4 and -rc5 and one is about the
camera.

Reverting:

[PATCH] eeepc-laptop: Prevent a panic when disabling RT2860 wireless
when associated

fixes the problem.

Commit-id is b56ab33d68638e6aafdbfc694025e8354a628f49

Regards,
Fabio


On Fri, Oct 16, 2009 at 9:53 PM, Fabio Comolli <fabio.comolli@gmail.com> wrote:
> Hi all.
> Pressing rfkill button on my eeePC 900 to disable wireless hangs the system.
>
> In console I can see (warning: what follows is copied by hand):
>
> acpid: client 148[6:100] has disconnected
> ACPI group/action not found: hotkey / ATKD
> unregister netdevice: waiting for wlan0 to become free; usage count=1
>
> (last line repeats forever). Hard reset is needed.
>
> With any previous kernel version (I'm running -rc4 right now) everything works:
>
> ACPI group/action undefined: hotkey / ATKD
> ath5k 0000:01:00.0: PCI INT A disabled
>
> (there's no acpid line)
>
>
> Regards,
> Fabio
>

Re: [Regression] - 2.6.32-rc5 hangs using rfkill on an eeePC

[Frans Pop]

Fabio Comolli wrote:
> Pressing rfkill button on my eeePC 900 to disable wireless hangs the
> system.
> 
> In console I can see (warning: what follows is copied by hand):
> 
> acpid: client 148[6:100] has disconnected
> ACPI group/action not found: hotkey / ATKD
> unregister netdevice: waiting for wlan0 to become free; usage count=1
> (last line repeats forever). Hard reset is needed.
>
> With any previous kernel version (I'm running -rc4 right now) everything
> works:
>
> ACPI group/action undefined: hotkey / ATKD
> ath5k 0000:01:00.0: PCI INT A disabled

AFAICT there are only two commits since -rc4 that are likely candidates:

commit b56ab33d68638e6aafdbfc694025e8354a628f49
Author: Darren Salt <linux@youmustbejoking.demon.co.uk>
Date:   Tue Oct 13 00:13:33 2009 +0200
    eeepc-laptop: Prevent a panic when disabling RT2860 wireless when
		  associated

commit 0af49167b1e5ba154e90d2c454bf4624ee47df80
Author: Darren Salt <linux@youmustbejoking.demon.co.uk>
Date:   Wed Oct 14 02:19:22 2009 +0100
    Staging: rt2860sta: prevent a panic when disabling when associated

They look related and which one is relevant for you probably depends on 
which driver you're using.

Could you try reverting them both to see if that solves the problem?

Cheers,
FJP

Re: [Regression] - 2.6.32-rc5 hangs using rfkill on an eeePC

[Fabio Comolli]

Hi.

On Fri, Oct 16, 2009 at 10:46 PM, Frans Pop <elendil@planet.nl> wrote:
> Fabio Comolli wrote:
>> Pressing rfkill button on my eeePC 900 to disable wireless hangs the
>> system.
>>
>> In console I can see (warning: what follows is copied by hand):
>>
>> acpid: client 148[6:100] has disconnected
>> ACPI group/action not found: hotkey / ATKD
>> unregister netdevice: waiting for wlan0 to become free; usage count=1
>> (last line repeats forever). Hard reset is needed.
>>
>> With any previous kernel version (I'm running -rc4 right now) everything
>> works:
>>
>> ACPI group/action undefined: hotkey / ATKD
>> ath5k 0000:01:00.0: PCI INT A disabled
>
> AFAICT there are only two commits since -rc4 that are likely candidates:
>
> commit b56ab33d68638e6aafdbfc694025e8354a628f49
> Author: Darren Salt <linux@youmustbejoking.demon.co.uk>
> Date:   Tue Oct 13 00:13:33 2009 +0200
>    eeepc-laptop: Prevent a panic when disabling RT2860 wireless when
>                  associated
>
> commit 0af49167b1e5ba154e90d2c454bf4624ee47df80
> Author: Darren Salt <linux@youmustbejoking.demon.co.uk>
> Date:   Wed Oct 14 02:19:22 2009 +0100
>    Staging: rt2860sta: prevent a panic when disabling when associated
>
> They look related and which one is relevant for you probably depends on
> which driver you're using.
>
> Could you try reverting them both to see if that solves the problem?

I already answered my own post in the thread: the culprit is:

eeepc-laptop: Prevent a panic when disabling RT2860 wireless when associated

Funnily enough, my eeePC has an Atheros card.

>
> Cheers,
> FJP
>

Regards,
Fabio

Re: [Regression] - 2.6.32-rc5 hangs using rfkill on an eeePC

[Darren Salt]

I demand that Fabio Comolli may or may not have written...

[snip]
> I already answered my own post in the thread: the culprit is:

> eeepc-laptop: Prevent a panic when disabling RT2860 wireless when
> associated

> Funnily enough, my eeePC has an Atheros card.

We (myself and Corentin Chary) tried to stop that patch once I'd noticed the
problem – I updated the patch, tested (all was fine with the rt2860 card),
upgraded from -rc3 to -rc4, installed an Intel 3945 card, tested, sent the
patch, restored the rt2860 card to its proper place, happened to toggle
wireless, got a panic due to having failed to remember to actually apply the
patch to -rc4... at this point, I rechecked with the Intel card (this time
with a patched kernel), saw a failure (stuck process – hmm, should rfkill be
re-entrant?), reported it, went digging into rt2860sta and produced, tested
and sent the replacement patch, then tried to stop the broken patch from
going upstream.

A reversion patch has been sent by Corentin (shortly after you reported this,
though the commit message says that he's reacted to its being in rc5) and,
earlier today, I directly requested that it be reverted; so hopefully, one
way or another, -rc6 will be fine for Atheros users and anybody else affected
by this, much as -rc5 is for rt2860 users.

Oh well. So much for Linus's "no new regressions". ;-}

-- 
| Darren Salt            | linux at youmustbejoking | nr. Ashington, | Doon
| using Debian GNU/Linux | or ds    ,demon,co,uk    | Northumberland | Army
| + Travel less. Share transport more.           PRODUCE LESS CARBON DIOXIDE.

Look! Over there!... Ha. Made you look.