linux-kernel - Re: [PATCH 23/28] locking/lockdep: Update irqsafe lock bitmaps

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 25 Apr 2019 21:55:15 +0200
From:   Peter Zijlstra <peterz@...radead.org>
To:     Yuyang Du <duyuyang@...il.com>
Cc:     will.deacon@....com, mingo@...nel.org, bvanassche@....org,
        ming.lei@...hat.com, frederic@...nel.org, tglx@...utronix.de,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH 23/28] locking/lockdep: Update irqsafe lock bitmaps

On Wed, Apr 24, 2019 at 06:19:29PM +0800, Yuyang Du wrote:
> The bitmaps keep track of which locks are irqsafe. Update the bitmaps
> when there is new irqsafe usage and when an irqsafe lock is zapped.
> 
> Signed-off-by: Yuyang Du <duyuyang@...il.com>
> ---
>  kernel/locking/lockdep.c | 39 ++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 38 insertions(+), 1 deletion(-)
> 
> diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> index 291cc9c..1b78216 100644
> --- a/kernel/locking/lockdep.c
> +++ b/kernel/locking/lockdep.c
> @@ -3107,6 +3107,7 @@ typedef int (*check_usage_f)(struct task_struct *, struct held_lock *,
>  	int excl_bit = exclusive_bit(new_bit);
>  	int read = new_bit & LOCK_USAGE_READ_MASK;
>  	int dir = new_bit & LOCK_USAGE_DIR_MASK;
> +	struct lock_class *lock = hlock_class(this);
>  
>  	/*
>  	 * mark USED_IN has to look forwards -- to ensure no dependency
> @@ -3119,6 +3120,25 @@ typedef int (*check_usage_f)(struct task_struct *, struct held_lock *,
>  		check_usage_backwards : check_usage_forwards;
>  
>  	/*
> +	 * The bit is already marked so that we update the bitmaps
> +	 * before validation.
> +	 */
> +	if (!dir) {
> +		unsigned long *bitmaps[4] = {
> +			lock_classes_hardirq_safe,
> +			lock_classes_hardirq_safe_read,
> +			lock_classes_softirq_safe,
> +			lock_classes_softirq_safe_read

That again should be something CPP magic using lockdep_states.h.

Also, that array can be static const, right? It's just an index into the
static bitmaps.

> +		};
> +		int index = (new_bit >> 2) << 1;
> +
> +		if (read)
> +		       index += 1;
> +
> +		__set_bit(lock - lock_classes, bitmaps[index]);
> +	}
> +
> +	/*
>  	 * Validate that this particular lock does not have conflicting
>  	 * usage states.
>  	 */
> @@ -3146,7 +3166,7 @@ typedef int (*check_usage_f)(struct task_struct *, struct held_lock *,
>  			return 0;
>  	}
>  
> -	if (state_verbose(new_bit, hlock_class(this)))
> +	if (state_verbose(new_bit, lock))
>  		return 2;
>  
>  	return 1;
> @@ -4650,6 +4670,22 @@ static void remove_class_from_lock_chains(struct pending_free *pf,
>  	}
>  }
>  
> +static inline void remove_irqsafe_lock_bitmap(struct lock_class *class)
> +{
> +#if defined(CONFIG_TRACE_IRQFLAGS) && defined(CONFIG_PROVE_LOCKING)
> +	unsigned long usage = class->usage_mask;
> +
> +	if (usage & LOCKF_USED_IN_HARDIRQ)
> +		__clear_bit(class - lock_classes, lock_classes_hardirq_safe);
> +	if (usage & LOCKF_USED_IN_HARDIRQ_READ)
> +		__clear_bit(class - lock_classes, lock_classes_hardirq_safe_read);
> +	if (usage & LOCKF_USED_IN_SOFTIRQ)
> +		__clear_bit(class - lock_classes, lock_classes_softirq_safe);
> +	if (usage & LOCKF_USED_IN_SOFTIRQ_READ)
> +		__clear_bit(class - lock_classes, lock_classes_softirq_safe_read);

More CPP foo required here. Also, do we really need to test, we could
just unconditionally clear the bits.

> +#endif
> +}