RE: BUG: sleeping function called from invalid context at mm/slub.c:1595 --- "hcitool cc" or "l2ping"

[Marcel Holtmann <marcel@holtmann.org>]

Hi Bing,

> > > > > > > I'm getting this error whenever I use "hcitool cc" or "l2ping
> > > > xx:xx:xx:xx:xx:xx"
> > > > > > command.
> > > > > > >
> > > > > > > Kernel: version 2.6.30-rc3 from bluetooth-testing tree
> > > > > > > HCI Tool: version 3.20
> > > > > > > Laptop: Lenovo X61s
> > > > > >
> > > > > > can you rebase the bluetooth-testing.git tree from a few hours ago and
> > > > > > test again.
> > > > >
> > > > > "git log" shows the last commit was ad4751fcada51292bb2967748667b582281647d0.
> > > > > Bluetooth: Fix errors and warnings in L2CAP reported by checkpatch.pl
> > > > >
> > > > > Is it the latest one?
> > > >
> > > > yes, that one is the latest. What happens if you revert:
> > > >
> > > > commit d97ebf3798fb5745953230e049011f5b20b7c6d9
> > > > Author: Roger Quadros <ext-roger.quadros@nokia.com>
> > > > Date:   Thu Apr 23 14:50:54 2009 +0300
> > > >
> > > >     Bluetooth: Ensure that HCI sysfs add/del is preempt safe
> > > >
> > > > Does this one introduces your problem?
> > > >
> > > > What is your kernel config and hardware you are running this one? Seems
> > > > not everybody is seeing this one.
> > > >
> > > Reverting d97ebf3798fb5745953230e049011f5b20b7c6d9 doesn't help.
> > >
> > > Today I got a chance to test it with some old tags.
> > >
> > > Tags:
> > > v2.6.29:                no issue
> > > v2.6.30-rc1:    no issue
> > > v2.6.30-rc2:    BUG: sleeping function called from invalid context at mm/slub.c:1595
> > > v2.6.30-rc3:    BUG: sleeping function called from invalid context at mm/slub.c:1595
> > >
> > > It happens on my laptops, Lenovo X61s and Fujitsu Lifebook A6110, with both v2.6.30-rc2 and
> > v2.6.30-rc3.
> > >
> > >
> > > The followings are from my Lifebook A6110,
> > > with blutooth-test git tag: v2.6.30-rc2 (commit 0882e8dd).
> > 
> > so I really think this is a kernel issue and not a Bluetooth issue. Can
> > clone Linus' tree from today. All the important Bluetooth fixes for
> > 2.6.30 went in today.
> > 
> > I did double check the flush_workqueue() and as far as I can tell we can
> > sleep in that context. So for some reason we are in the wrong context. I
> > really don't know why.
> 
> OK. I'll clone Linus's tree and give it a try when I get a chance.

I re-based bluetooth-testing.git against 2.6.30-rc4 and that is as
recent as it gets right now.

Regards

Marcel