Re: [PATCH] workaround minor lockdep bug triggered by mm_take_all_locks

[Peter Zijlstra <a.p.zijlstra@chello.nl>]

On Mon, 2008-08-04 at 19:57 +0200, Andrea Arcangeli wrote:
> From: Andrea Arcangeli <andrea@qumranet.com>
> 
> Lockdep can't recognize if spinlocks are at a different address. So
> using trylock in a loop is one way to avoid lockdep to generate false
> positives. After lockdep will be fixed this change can and should be
> reverted.
> 
> Signed-off-by: Andrea Arcangeli <andrea@qumranet.com>

NAK, come-on, you didn't even bother to look at the available
annotations..

> ---
> 
> Resubmit because I didn't update the subject and so I improved the
> comments too. This is mostly to show it's a bug in lockdep and it can
> be trivially worked around without having to fix lockdep for real, any
> superior solution to this hack is more than welcome and recommended.
> 
> diff -r 3469dce61df1 mm/mmap.c
> --- a/mm/mmap.c	Tue Jul 29 20:01:28 2008 +0200
> +++ b/mm/mmap.c	Mon Aug 04 19:54:27 2008 +0200
> @@ -2279,8 +2279,12 @@ static void vm_lock_anon_vma(struct anon
>  		/*
>  		 * The LSB of head.next can't change from under us
>  		 * because we hold the mm_all_locks_mutex.
> +		 *
> +		 * spin_lock would confuse lockdep who doesn't notice
> +		 * the 'anon_vma' always changing address.
>  		 */
> -		spin_lock(&anon_vma->lock);
> +		while (!spin_trylock(&anon_vma->lock))
> +			cpu_relax();
>  		/*
>  		 * We can safely modify head.next after taking the
>  		 * anon_vma->lock. If some other vma in this mm shares
> @@ -2310,7 +2314,12 @@ static void vm_lock_mapping(struct addre
>  		 */
>  		if (test_and_set_bit(AS_MM_ALL_LOCKS, &mapping->flags))
>  			BUG();
> -		spin_lock(&mapping->i_mmap_lock);
> +		/*
> +		 * spin_lock would confuse lockdep who doesn't notice
> +		 * the 'mapping' always changing address.
> +		 */
> +		while (!spin_trylock(&mapping->i_mmap_lock))
> +			cpu_relax();
>  	}
>  }
>