linux-kernel - Re: [PATCH] [RFC] CFQ: simplify radix tree lookup in cfq_cic

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]

Message-ID: <20110803211413.GE32385@redhat.com>
Date:	Wed, 3 Aug 2011 17:14:13 -0400
From:	Vivek Goyal <vgoyal@...hat.com>
To:	Paul Bolle <vivek.goyal2008@...il.com>
Cc:	Jens Axboe <axboe@...nel.dk>, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] [RFC] CFQ: simplify radix tree lookup in cfq_cic_lookup()

On Wed, Aug 03, 2011 at 10:14:32PM +0200, Paul Bolle wrote:
> 0) Not tested and not signed-off.
> 
> 1) This is to see whether I understand the cfqd->cic_index usage (at
> least, part of it).
> 
> 2) If the lookup of a cic in the radix tree turned up a "dead" cic, then
> that cic will be dropped. There's no reason to again try to lookup that
> cic: that lookup should return NULL. (If it doesn't return NULL, we seem
> to be in trouble.) So there's no need for a do {[...]} while (1) loop
> and this code can be simplified a little.
> 
> 3) Does this make sense?


When a request queue exits, cfq_exit_queue(), it will free up the
associated cic_index (ida_remove(&cic_index_ida, cfqd->cic_index)). All
the cic which are on the request queue will be marked as dead. Now this
cic_index is up for grab and can be re-allocated to a different request
queue.

Now if the same process does IO to this new queue we same cic_index as
old request queue, then it should find the dead key and drop it and
then allocated a new cic.

So it does sound that there can not be more than one dead key associated
with a cic_index at a time in ioc tree.

But keeping current code does not harm.

Thanks
Vivek

> 
> 
> Paul Bolle
> ---
>  block/cfq-iosched.c |   28 +++++++++++++---------------
>  1 files changed, 13 insertions(+), 15 deletions(-)
> 
> diff --git a/block/cfq-iosched.c b/block/cfq-iosched.c
> index 1f96ad6..0d33d8c 100644
> --- a/block/cfq-iosched.c
> +++ b/block/cfq-iosched.c
> @@ -3120,22 +3120,20 @@ cfq_cic_lookup(struct cfq_data *cfqd, struct io_context *ioc)
>  		return cic;
>  	}
>  
> -	do {
> -		cic = radix_tree_lookup(&ioc->radix_root, cfqd->cic_index);
> -		rcu_read_unlock();
> -		if (!cic)
> -			break;
> -		if (unlikely(cic->key != cfqd)) {
> -			cfq_drop_dead_cic(cfqd, ioc, cic);
> -			rcu_read_lock();
> -			continue;
> -		}
> +	cic = radix_tree_lookup(&ioc->radix_root, cfqd->cic_index);
> +	rcu_read_unlock();
>  
> -		spin_lock_irqsave(&ioc->lock, flags);
> -		rcu_assign_pointer(ioc->ioc_data, cic);
> -		spin_unlock_irqrestore(&ioc->lock, flags);
> -		break;
> -	} while (1);
> +	if (!cic)
> +		return NULL;
> +
> +	if (unlikely(cic->key != cfqd)) {
> +		cfq_drop_dead_cic(cfqd, ioc, cic);
> +		return NULL;
> +	}
> +
> +	spin_lock_irqsave(&ioc->lock, flags);
> +	rcu_assign_pointer(ioc->ioc_data, cic);
> +	spin_unlock_irqrestore(&ioc->lock, flags);
>  
>  	return cic;
>  }
> -- 
> 1.7.6
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@...r.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/