Re: [PATCH tip/core/rcu 0/2] more lockdep-RCU and RCU_FAST_NO_HZfixes

From: Paul E. McKenney
Date: Sun Feb 28 2010 - 11:32:26 EST

On Sun, Feb 28, 2010 at 09:55:05AM +0100, Ingo Molnar wrote:
>
> * Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx> wrote:
>
> > Hello!
> >
> > This series (1) makes RCU_FAST_NO_HZ actually do the full number of
> > iterations specified instead of just one and (2) fixes an RCU lockdep
> > splat.
>
> Thanks, applied.
>
> This one is pretty stubborn, it still triggers:
>
> [ 0.000000] soft-irq read-recursion/321: ok |
> [ 0.000000] -------------------------------------------------------
> [ 0.000000] Good, all 218 testcases passed! |
> [ 0.000000] ---------------------------------
> [ 0.000000] allocated 10485760 bytes of page_cgroup
> [ 0.000000] please try 'cgroup_disable=memory' option if you don't want memory cgroups
> [ 0.000000]
> [ 0.000000] ===================================================
> [ 0.000000] [ INFO: suspicious rcu_dereference_check() usage. ]
> [ 0.000000] ---------------------------------------------------
> [ 0.000000] kernel/sched_fair.c:3479 invoked rcu_dereference_check() without protection!
> [ 0.000000]
> [ 0.000000] other info that might help us debug this:
> [ 0.000000]
> [ 0.000000] no locks held by swapper/0.
> [ 0.000000]
> [ 0.000000] stack backtrace:
> [ 0.000000] Pid: 0, comm: swapper Not tainted 2.6.33-tip-01015-g97e7da3-dirty #10753
> [ 0.000000] Call Trace:
> [ 0.000000] <IRQ> [<ffffffff8107f5d1>] lockdep_rcu_dereference+0xa1/0xb0
> [ 0.000000] [<ffffffff8104afd1>] scheduler_tick+0x221/0x2c0
> [ 0.000000] [<ffffffff8105f6d5>] update_process_times+0x55/0x70
> [ 0.000000] [<ffffffff8107ac27>] tick_periodic+0x27/0x70
> [ 0.000000] [<ffffffff8107ac8f>] tick_handle_periodic+0x1f/0x90
> [ 0.000000] [<ffffffff8100e29a>] timer_interrupt+0x1a/0x30
> [ 0.000000] [<ffffffff810a28aa>] handle_IRQ_event+0x3a/0xe0
> [ 0.000000] [<ffffffff810a53ab>] handle_level_irq+0x8b/0x130
> [ 0.000000] [<ffffffff8100da94>] handle_irq+0x44/0x70
> [ 0.000000] [<ffffffff8100cfde>] do_IRQ+0x6e/0xf0
> [ 0.000000] [<ffffffff817eea13>] ret_from_intr+0x0/0x16
> [ 0.000000] <EOI> [<ffffffff8102dfe6>] ? native_restore_fl+0x6/0x10
> [ 0.000000] [<ffffffff817ee3df>] _raw_spin_unlock_irqrestore+0x6f/0x80
> [ 0.000000] [<ffffffff810a37f3>] __setup_irq+0x1b3/0x3b0
> [ 0.000000] [<ffffffff810a3a18>] setup_irq+0x28/0x40
> [ 0.000000] [<ffffffff81ef2914>] setup_default_timer_irq+0x12/0x14
> [ 0.000000] [<ffffffff81ef292d>] hpet_time_init+0x17/0x19
> [ 0.000000] [<ffffffff81ef28fb>] x86_late_time_init+0xa/0x11
> [ 0.000000] [<ffffffff81eefcf4>] start_kernel+0x25f/0x2e6
> [ 0.000000] [<ffffffff81eef30b>] x86_64_start_reservations+0xf6/0xfa
> [ 0.000000] [<ffffffff81eef409>] x86_64_start_kernel+0xfa/0x109
> [ 0.000000] Fast TSC calibration using PIT
> [ 0.000000] Detected 2002.507 MHz processor.
> [ 0.004004] Calibrating delay loop (skipped), value calculated using timer frequency.. 4005.01 BogoMIPS (lpj=2002507)

Ah -- this is a new arrival from my viewpoint. Patch below!

This won't apply to tip/core/rcu, but does against current tip/master.

Thanx, Paul

------------------------------------------------------------------------

sched: fix rcu_dereference() for RCU-lockde

Make rcu_dereference() of runqueue data structures be
rcu_dereference_sched().

Located-by: Ingo Molnar <mingo@xxxxxxx>
Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxxxxxxxxxx>
---

sched_fair.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/kernel/sched_fair.c b/kernel/sched_fair.c
index 3e1fd96..5a5ea2c 100644
--- a/kernel/sched_fair.c
+++ b/kernel/sched_fair.c
@@ -3476,7 +3476,7 @@ static void run_rebalance_domains(struct softirq_action *h)

static inline int on_null_domain(int cpu)
{
- return !rcu_dereference(cpu_rq(cpu)->sd);
+ return !rcu_dereference_sched(cpu_rq(cpu)->sd);
}

/*
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/