BCMSMAC freezes system on trivial oeprations

BCMSMAC freezes system on trivial oeprations

[Cedric Sodhi]

I use an HP Touchsmart tm2 laptop which has an onboard BRCM 4727- I used to compile my kernel with BCMSMAC module.

At first, I got regular kernel panics that, according to the printed-out stack originated somewhere from the BRCM driver and for which I could not exactly tell the reason (I assume it had something to do with roaming between APs).

Also, the laptop has a dedicated button to activating/deactivating the WLAN device, besides it has a BIOS wired funtion key, that does the same - at least that what it does in Windows and what HP says it should do.

When pressing either of the buttons, the kernel would hard-lock without any further notice and the computer could only be brought down through a hard reset.

This was awful enough, so I though I could try enabling the rfkill kernel compoenent as it seems related (layman at work, as you can tell). Recompiling that way (linux-next, by the way, *not* on wireless-testing yet) and also modifiying another few minor settings resulted in a desaster:

Since then - *even after I have reerted all changes and gone back to the kernel which was known to have worked last!!!* - as soon as the BCMSMAC module is either ocmpiled in or loaded, trivial commands such as ifconfig, iwconfig or just exit (!!!) freeze in a State D (awaiting IO) and can not be killed by even kill -9 or -1.

Shutting the system down is equally impossible without the help of SysReqs. Pushing the WLAN-Button still hardfreezes the system.

You can safely assume that practically *any* operation that is remotely related to the device, such as echo'ing to the rfkill /sys device, too, plus many commands that appear not to be related at all, such as "exit" will freeze in state D and cannot be killed by any means possible.

Pressing aforementioned "WLAN Button" will make these commands resume, just for a split second, before the system hard-locks.

Most puzzling about that, I find, is that reverting to the known-as-having-worked setup (exactly the same) has not got the system back to working.

I have two theories: Either there is some evil, hidden hardware switch on the device that has been triggered and not been reset since and that is causing all this right now, or some very crucial part of the system has been damaged by the regular hard-reset (which I deem unlikely since it would not explain why all sorts of programs freeze for no apparent reason).

So far, the system is practically not usuable - at least not with the driver activated - I'm in desperate need for your help. I can reformat and resetup everything, but I need to know a solution to this in the long term and I really don't know what else to try.

BCMSMAC freezes system on trivial oeprations

[Cedric Sodhi]

I need to reply to my own message. After I just returned home, it
appeared that the computer had no problem whatsoever connecting to my
local home network. As if nothing had ever happend. Well, the
"WLAN-Button" still hardlocks the machine but apart from that I have no
indication of problems.

I'll let you know as soon as I know more and wpulwd appreciate any kind
of comments in the meantime.

regards,
Cedric


On Mon, May 09, 2011 at 03:44:16PM +0200, Cedric Sodhi wrote:
> I use an HP Touchsmart tm2 laptop which has an onboard BRCM 4727- I used to compile my kernel with BCMSMAC module.
> 
> At first, I got regular kernel panics that, according to the printed-out stack originated somewhere from the BRCM driver and for which I could not exactly tell the reason (I assume it had something to do with roaming between APs).
> 
> Also, the laptop has a dedicated button to activating/deactivating the WLAN device, besides it has a BIOS wired funtion key, that does the same - at least that what it does in Windows and what HP says it should do.
> 
> When pressing either of the buttons, the kernel would hard-lock without any further notice and the computer could only be brought down through a hard reset.
> 
> This was awful enough, so I though I could try enabling the rfkill kernel compoenent as it seems related (layman at work, as you can tell). Recompiling that way (linux-next, by the way, *not* on wireless-testing yet) and also modifiying another few minor settings resulted in a desaster:
> 
> Since then - *even after I have reerted all changes and gone back to the kernel which was known to have worked last!!!* - as soon as the BCMSMAC module is either ocmpiled in or loaded, trivial commands such as ifconfig, iwconfig or just exit (!!!) freeze in a State D (awaiting IO) and can not be killed by even kill -9 or -1.
> 
> Shutting the system down is equally impossible without the help of SysReqs. Pushing the WLAN-Button still hardfreezes the system.
> 
> You can safely assume that practically *any* operation that is remotely related to the device, such as echo'ing to the rfkill /sys device, too, plus many commands that appear not to be related at all, such as "exit" will freeze in state D and cannot be killed by any means possible.
> 
> Pressing aforementioned "WLAN Button" will make these commands resume, just for a split second, before the system hard-locks.
> 
> Most puzzling about that, I find, is that reverting to the known-as-having-worked setup (exactly the same) has not got the system back to working.
> 
> I have two theories: Either there is some evil, hidden hardware switch on the device that has been triggered and not been reset since and that is causing all this right now, or some very crucial part of the system has been damaged by the regular hard-reset (which I deem unlikely since it would not explain why all sorts of programs freeze for no apparent reason).
> 
> So far, the system is practically not usuable - at least not with the driver activated - I'm in desperate need for your help. I can reformat and resetup everything, but I need to know a solution to this in the long term and I really don't know what else to try.
> 
> 
> _______________________________________________
> b43-dev mailing list
> b43-dev at lists.infradead.org
> http://lists.infradead.org/mailman/listinfo/b43-dev

BCMSMAC freezes system on trivial oeprations

[Cedric Sodhi]

1 - The distro is Gentoo
2 - Initially? I don't understand the question
3 - No, I did not find an easy way to obtain the trace on a panic at
    first. I've by now learned about options which allow to dump the
    strace to flash.
4 - Gitweb tarball. Hence, I can no longer provide a SHA1, since it was
    an archive. As stated in my last email the best I can provide you
    with is the following string:

    Linux 2.6.39-rc6-next-20110506
5 - I think I've noted that the rfkill sysnode  freezes as ifconfig,
    iwconfig, exit and others do. Now, under these new circumstances
    where all problems (apart from the button-triggered freeze) seem to
    have vanished, I can try both, writing to it again and running
    rfkill. I'll do so after I have dispatched this mail as I do reckon
    with new freezes and hardlocks. I will provide further information
    in another email just in a short while.

regards,
Cedric

On Tue, May 10, 2011 at 12:57:01PM +0200, Arend van Spriel wrote:
> On 05/09/2011 03:44 PM, Cedric Sodhi wrote:
> > I use an HP Touchsmart tm2 laptop which has an onboard BRCM 4727- I used to compile my kernel with BCMSMAC module.
> Hi Cedric,
> 
> This story is quite erratic so I want to get some basic information first.
> 
> 1) what linux distro are you using?
> 2) what kernel version are you running (initially)?
> > At first, I got regular kernel panics that, according to the printed-out stack originated somewhere from the BRCM driver and for which I could not exactly tell the reason (I assume it had something to do with roaming between APs).
> 3) do you have any trace information on the 'regular kernel panics'?
> > Also, the laptop has a dedicated button to activating/deactivating the WLAN device, besides it has a BIOS wired funtion key, that does the same - at least that what it does in Windows and what HP says it should do.
> >
> > When pressing either of the buttons, the kernel would hard-lock without any further notice and the computer could only be brought down through a hard reset.
> >
> > This was awful enough, so I though I could try enabling the rfkill kernel compoenent as it seems related (layman at work, as you can tell). Recompiling that way (linux-next, by the way, *not* on wireless-testing yet) and also modifiying another few minor settings resulted in a desaster:
> 4) how did you obtain linux-next? git? when using git can you provide sha1?
> > Since then - *even after I have reerted all changes and gone back to the kernel which was known to have worked last!!!* - as soon as the BCMSMAC module is either ocmpiled in or loaded, trivial commands such as ifconfig, iwconfig or just exit (!!!) freeze in a State D (awaiting IO) and can not be killed by even kill -9 or -1.
> 5) could you try the 'rfkill' command?
> 
> Gr. AvS
> 
> -- 
> Almost nobody dances sober, unless they happen to be insane.
> -- H.P. Lovecraft --
>