Re: [PATCH 0/4] [GIT PULL] tracing: recursion and compile fixes

[Frederic Weisbecker <fweisbec@gmail.com>]

On Mon, Apr 20, 2009 at 04:17:32PM -0400, Steven Rostedt wrote:
> > > [   16.033417] Running tests again, along with the function tracer
> > > [   16.039141] Running tests on trace events:
> > > [   16.043063] Testing event kfree_skb: OK
> > > [   16.052521] Testing event kmalloc: OK
> > > [   16.061510] Testing event kmem_cache_alloc: OK
> > > [   16.071902] Testing event kmalloc_node: OK
> > > [   16.081756] Testing event kmem_cache_alloc_node: OK
> > > [   16.091759] Testing event kfree: OK
> > > [   16.100517] Testing event kmem_cache_free: OK
> > > [   16.110523] Testing event irq_handler_exit: OK
> > > [   16.120524] Testing event irq_handler_entry: OK
> > > [   16.130521] Testing event softirq_entry: OK
> > > [   16.139833] Testing event softirq_exit: OK
> > > [   16.149597] Testing event lock_acquired: OK
> > > [   16.160062] Testing event lock_acquire: OK
> > > [   16.171719] Testing event lock_release: <0>BUG: spinlock lockup on CPU#0, swapper/0, 80cd0240
> > > [   16.176013] Pid: 0, comm: swapper Not tainted 2.6.30-rc2-tip-01538-ge8201b9-dirty #34096
> > > [   16.176013] Call Trace:
> > > [   16.176013]  [<80944340>] ? printk+0x15/0x1d
> > > [   16.176013]  [<803a410a>] _raw_spin_lock+0x19a/0x1b0
> > > [   16.176013]  [<80947bb7>] _spin_lock_irqsave+0x57/0x70
> > > [   16.176013]  [<8012b2ca>] __wake_up+0x1a/0x50
> > > [   16.176013]  [<8019392f>] trace_wake_up+0x2f/0x40
> > 
> > 
> > The function tracer, used in the selftest, uses trace_current_buffer_unlock_commit
> > which in turn calls trace_wake_up.
> > 
> > The function tracer shouldn't call trace_wake_up() (might hold
> > the rq lock already).
> > 
> > I'm preparing a fix.
> > 
> 
> Hmm, doesn't the trace wakeup test if the runqueue lock is locked or not?
> 
> -- Steve
> 


Hmm, yes it does but that's not the first time we meet this problem
(sched switch event tracing recursions by the past). So either the
test doesn't work well or this is about another lock that
wake_up_common takes...