Re: WARNING: at kernel/rcutorture.c:1243 rcu_torture_printk

[Fengguang Wu <fengguang.wu@intel.com>]

Paul,

On Mon, Jun 10, 2013 at 07:51:10AM -0700, Paul E. McKenney wrote:
> On Mon, Jun 10, 2013 at 03:47:28PM +0800, Fengguang Wu wrote:
> > Greetings,
> > 
> > I got the below dmesg and the first bad commit is
> > 
> > commit 911af505ef407c2511106c224dd640f882f0f590
> > Author: Paul E. McKenney <paul.mckenney@linaro.org>
> > Date:   Mon Feb 11 10:23:27 2013 -0800
> > 
> >     rcu: Provide compile-time control for no-CBs CPUs
> >     
> >     Currently, the only way to specify no-CBs CPUs is via the rcu_nocbs
> >     kernel command-line parameter.  This is inconvenient in some cases,
> >     particularly for randconfig testing, so this commit adds a new set of
> >     kernel configuration parameters.  CONFIG_RCU_NOCB_CPU_NONE (the default)
> >     retains the old behavior, CONFIG_RCU_NOCB_CPU_ZERO offloads callback
> >     processing from CPU 0 (along with any other CPUs specified by the
> >     rcu_nocbs boot-time parameter), and CONFIG_RCU_NOCB_CPU_ALL offloads
> >     callback processing from all CPUs.
> >     
> >     Signed-off-by: Paul E. McKenney <paul.mckenney@linaro.org>
> >     Signed-off-by: Paul E. McKenney <paulmck@linux.vnet.ibm.com>
> > 
> > However I guess it's a wrong bisect, the commit should be unmasking an
> > old bug because it merely provides more kconfig options.
> > 
> > This warning happened only once in 74 boots of the kernel.
> > 
> > I also attached the 2nd dmesg which shows boot hang for the same
> > kernel.
> > 
> > Note that there are also 2 kernel hangs in the 74 boots.  It should
> > not be an RCU problem, however for completeness, I also attach a dmesg
> > for your reference when considering this RCU warning.
> 
> Either way, that is a nasty warning!  You get this on recent kernels as
> well, I take it?

Yes, the warning is still in recent kernels, indicated from the last
lines of the bisect log.

911af505ef407c2511106c224dd640f882f0f590 is bad and its parent
34ed62461ae4970695974afb9a60ac3df0086830 is good:
> > git bisect  bad 911af505ef407c2511106c224dd640f882f0f590  #    65  2013-06-07 16:14:41  rcu: Provide compile-time control for no-CBs CPUs
> > git bisect good 34ed62461ae4970695974afb9a60ac3df0086830  #   900  2013-06-08 04:57:34  rcu: Remove restrictions on no-CBs CPUs

After bisect, the script tries to confirm that
34ed62461ae4970695974afb9a60ac3df0086830 is really good by boot
testing it 2700 more times:

> > git bisect good 34ed62461ae4970695974afb9a60ac3df0086830  #  2700  2013-06-10 14:33:56  rcu: Remove restrictions on no-CBs CPUs

And continue to find out whether the linus/master and
linux-next/master are good/bad (here the results are both bad):

> > git bisect  bad f43e7a34255a387930bc2bb826ec17c052ce975e  # 14:34      0  Merge branch 'for-next'
> > git bisect  bad c04efed734409f5a44715b54a6ca1b54b0ccf215  # 14:41     11  Add linux-next specific files for 20130607

Thanks,
Fengguang