brcm80211 hangs when disabling wireless

brcm80211 hangs when disabling wireless

[Dave Hansen]

I've been getting hard hangs when I toggle the wireless on a Lenovo
S10-3.  It happens both with the physical switch and the Fn-F5 key
combo.  It's quite repeatable.  The system is quite stable when I'm not
trying to disable the wireless, though.

I've been able to get some screenshots, but the console messages don't
make it out to the disk:

	http://picasaweb.google.com/hansendc/BrcmOops#

netconsole also isn't working on this hardware.  Any other suggestions
for getting an oops out of a serial-port-challenged netbook?

It isn't completely clear to me where this bug actually is.  The stack
dump looks almost completely confined to net/mac80211/ functions, but
there are a few error printk's from brcm80211.

The first printk's are from this code, but it's apparently trying to
transmit in the mac80211 stack anyway:
        
        static int wl_ops_tx(struct ieee80211_hw *hw, struct sk_buff *skb)
        {
                int status;
                wl_info_t *wl = hw->priv;
                WL_LOCK(wl);
                if (!wl->pub->up) {
                        WL_ERROR(("ops->tx called while down\n"));
                        status = -ENETDOWN;
                        goto done;
                }
                status = wl_start(skb, wl);
         done:
                WL_UNLOCK(wl);
                return status;
        }

I'm not completely sure that this is what is causing the hang that I'm
seeing, but it's certainly the last thing I see on the console.  

With the hardware driver in -staging, I don't expect anybody to jump out
of their seats to fix this, I just figured someone might want to see the
bug report.

-- Dave

Re: brcm80211 hangs when disabling wireless

[Greg KH]

On Thu, Dec 09, 2010 at 01:26:50PM -0800, Dave Hansen wrote:
> I've been getting hard hangs when I toggle the wireless on a Lenovo
> S10-3.  It happens both with the physical switch and the Fn-F5 key
> combo.  It's quite repeatable.  The system is quite stable when I'm not
> trying to disable the wireless, though.

What kernel version is this?

Linus's latest tree has a known bug when the wireless is suspended and
then resumed.  Broadcom is working on it but no patch is forthcoming
yet.  Perhaps this is the same issue.

thanks,

greg k-h

Re: brcm80211 hangs when disabling wireless

[Dave Hansen]

On Thu, 2010-12-09 at 13:43 -0800, Greg KH wrote:
> On Thu, Dec 09, 2010 at 01:26:50PM -0800, Dave Hansen wrote:
> > I've been getting hard hangs when I toggle the wireless on a Lenovo
> > S10-3.  It happens both with the physical switch and the Fn-F5 key
> > combo.  It's quite repeatable.  The system is quite stable when I'm not
> > trying to disable the wireless, though.
> 
> What kernel version is this?
> 
> Linus's latest tree has a known bug when the wireless is suspended and
> then resumed.  Broadcom is working on it but no patch is forthcoming
> yet.  Perhaps this is the same issue.

Yeah, it's 2.6.37-rc5 plus a pull from:

	http://kernel.ubuntu.com/git?p=ikepanhc/ideapad-laptop.git;a=summary

That could easily be it.  Thanks, Greg.

-- Dave

Re: brcm80211 hangs when disabling wireless

[Henry Ptasinski]

On Thu, Dec 09, 2010 at 01:54:38PM -0800, Dave Hansen wrote:
> On Thu, 2010-12-09 at 13:43 -0800, Greg KH wrote:
> > On Thu, Dec 09, 2010 at 01:26:50PM -0800, Dave Hansen wrote:
> > > I've been getting hard hangs when I toggle the wireless on a Lenovo
> > > S10-3.  It happens both with the physical switch and the Fn-F5 key
> > > combo.  It's quite repeatable.  The system is quite stable when I'm not
> > > trying to disable the wireless, though.
> > 
> > What kernel version is this?
> > 
> > Linus's latest tree has a known bug when the wireless is suspended and
> > then resumed.  Broadcom is working on it but no patch is forthcoming
> > yet.  Perhaps this is the same issue.
> 
> Yeah, it's 2.6.37-rc5 plus a pull from:
> 
> 	http://kernel.ubuntu.com/git?p=ikepanhc/ideapad-laptop.git;a=summary
> 
> That could easily be it.  Thanks, Greg.

Actually, I think it's the rfkill support (or lack thereof) that's causing you
trouble.  But we're looking at that as well.

The outstanding reported issues that I'm aware of are:
	- 43224 crashes
	- missing rfkill support
	- suspend/resume problems
	- 4313 infinite loop in some cases

In terms of process, should these all get filed in bugzilla.kernel.org (or
somewhere else), or just track them via email and our TODO list?

Thanks,
- Henry

Re: brcm80211 hangs when disabling wireless

[Greg KH]

On Mon, Dec 13, 2010 at 11:31:11AM -0800, Henry Ptasinski wrote:
> On Thu, Dec 09, 2010 at 01:54:38PM -0800, Dave Hansen wrote:
> > On Thu, 2010-12-09 at 13:43 -0800, Greg KH wrote:
> > > On Thu, Dec 09, 2010 at 01:26:50PM -0800, Dave Hansen wrote:
> > > > I've been getting hard hangs when I toggle the wireless on a Lenovo
> > > > S10-3.  It happens both with the physical switch and the Fn-F5 key
> > > > combo.  It's quite repeatable.  The system is quite stable when I'm not
> > > > trying to disable the wireless, though.
> > > 
> > > What kernel version is this?
> > > 
> > > Linus's latest tree has a known bug when the wireless is suspended and
> > > then resumed.  Broadcom is working on it but no patch is forthcoming
> > > yet.  Perhaps this is the same issue.
> > 
> > Yeah, it's 2.6.37-rc5 plus a pull from:
> > 
> > 	http://kernel.ubuntu.com/git?p=ikepanhc/ideapad-laptop.git;a=summary
> > 
> > That could easily be it.  Thanks, Greg.
> 
> Actually, I think it's the rfkill support (or lack thereof) that's causing you
> trouble.  But we're looking at that as well.
> 
> The outstanding reported issues that I'm aware of are:
> 	- 43224 crashes
> 	- missing rfkill support
> 	- suspend/resume problems
> 	- 4313 infinite loop in some cases
> 
> In terms of process, should these all get filed in bugzilla.kernel.org (or
> somewhere else), or just track them via email and our TODO list?

Where ever you feel comfortable putting them, some like email, some like
bugzilla.kernel.org.  The TODO list might be a bit tougher as it doesn't
always propagate to Linus fast enough (look at the differences between
it already in Linus's tree and linux-next, right?)

thanks,

greg k-h