Re: [patch, lock validator] rcu_torture_lock deadlock fix

["Paul E. McKenney" <paulmck@us.ibm.com>]

On Wed, Jan 25, 2006 at 07:50:12PM +0100, Ingo Molnar wrote:
> rcu_torture_lock is used in a softirq-unsafe manner, but it is also
> taken by rcu_torture_cb(), which may execute in softirq-context,
> resulting in potential deadlocks.

Good catch!!!  :-/

							Thanx, Paul

Acked-by: <paulmck@us.ibm.com>

> this bug was found by the lock-validator:
> 
>   ============================
>   [ BUG: illegal lock usage! ]
>   ----------------------------
>   illegal {enabled-softirqs} -> {used-in-softirq} usage.
>   swapper/0 [HC0[0]:SC1[2]:HE1:SE0] takes {rcu_torture_lock [u:25]}, at:
>    [<c014dab3>] rcu_torture_cb+0x49/0x7e
>   {enabled-softirqs} state was registered at:
>    [<c01269a0>] copy_process+0x25d/0xed5
>   hardirqs last enabled at: [<c0137a7a>] __rcu_process_callbacks+0x5f/0x14a
>   softirqs last enabled at: [<c012ce90>] irq_exit+0x36/0x38
>   
>   other info that might help in debugging this:
>   ------------------------------
>   | showing all locks held by: |  (swapper/0 [f7c06750, 140]): <none>
>   ------------------------------
>   
>    [<c0103e82>] show_trace+0xd/0xf
>    [<c0103e99>] dump_stack+0x15/0x17
>    [<c013e581>] print_usage_bug+0x176/0x181
>    [<c013ea58>] mark_lock+0x9b/0x22b
>    [<c013efa2>] debug_lock_chain+0x3ba/0xd0c
>    [<c013f925>] debug_lock_chain_spin+0x31/0x48
>    [<c041076e>] _raw_spin_lock+0x34/0x7f
>    [<c0d29435>] _spin_lock+0x8/0xa
>    [<c014dab3>] rcu_torture_cb+0x49/0x7e
>    [<c0137b1f>] __rcu_process_callbacks+0x104/0x14a
>    [<c0137e33>] rcu_process_callbacks+0x26/0x3f
>    [<c012cfc4>] tasklet_action+0x64/0xc8
>    [<c012d174>] __do_softirq+0x84/0xff
>    [<c0104d72>] do_softirq+0x52/0xbb
>    =======================
>    [<c012ce90>] irq_exit+0x36/0x38
>    [<c0118809>] smp_apic_timer_interrupt+0x4e/0x51
>    [<c010376b>] apic_timer_interrupt+0x27/0x2c
>    [<c0101b4f>] cpu_idle+0x9a/0xb3
>    [<c0117d1e>] start_secondary+0x3c3/0x3cb
>    [<00000000>] 0x0
>    [<f7c07fb4>] 0xf7c07fb4
> 
> the fix is to acquire rcu_torture_lock in a softirq-safe manner.
> With this fix applied, the rcu-torture code passes validation.
> 
> Signed-off-by: Ingo Molnar <mingo@elte.hu>
> 
> ----
> 
>  kernel/rcutorture.c |   10 +++++-----
>  1 files changed, 5 insertions(+), 5 deletions(-)
> 
> Index: linux/kernel/rcutorture.c
> ===================================================================
> --- linux.orig/kernel/rcutorture.c
> +++ linux/kernel/rcutorture.c
> @@ -114,16 +114,16 @@ rcu_torture_alloc(void)
>  {
>  	struct list_head *p;
>  
> -	spin_lock(&rcu_torture_lock);
> +	spin_lock_bh(&rcu_torture_lock);
>  	if (list_empty(&rcu_torture_freelist)) {
>  		atomic_inc(&n_rcu_torture_alloc_fail);
> -		spin_unlock(&rcu_torture_lock);
> +		spin_unlock_bh(&rcu_torture_lock);
>  		return NULL;
>  	}
>  	atomic_inc(&n_rcu_torture_alloc);
>  	p = rcu_torture_freelist.next;
>  	list_del_init(p);
> -	spin_unlock(&rcu_torture_lock);
> +	spin_unlock_bh(&rcu_torture_lock);
>  	return container_of(p, struct rcu_torture, rtort_free);
>  }
>  
> @@ -134,9 +134,9 @@ static void
>  rcu_torture_free(struct rcu_torture *p)
>  {
>  	atomic_inc(&n_rcu_torture_free);
> -	spin_lock(&rcu_torture_lock);
> +	spin_lock_bh(&rcu_torture_lock);
>  	list_add_tail(&p->rtort_free, &rcu_torture_freelist);
> -	spin_unlock(&rcu_torture_lock);
> +	spin_unlock_bh(&rcu_torture_lock);
>  }
>  
>  static void
>