linux-kernel - Re: [PATCH 2/2] new bd_mutex lockdep annotation

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, 14 Sep 2006 17:11:21 +1000
From:	Neil Brown <neilb@...e.de>
To:	Peter Zijlstra <a.p.zijlstra@...llo.nl>,
	Al Viro <viro@...iv.linux.org.uk>
Cc:	linux-kernel@...r.kernel.org, Ingo Molnar <mingo@...e.hu>,
	Arjan van de Ven <arjan@...ux.intel.com>,
	Andrew Morton <akpm@...l.org>, Jason Baron <jbaron@...hat.com>
Subject: Re: [PATCH 2/2] new bd_mutex lockdep annotation

On Wednesday September 13, a.p.zijlstra@...llo.nl wrote:
> Use the gendisk partition number to set a lock class.

Yes, this does look a lot nicer, thanks.

Two observations.
1/ I was confused that you added a call to mutex_init.   One would
 normally expect to only have one of these for any given mutex, so
 adding one was a surprise.
 I now realise that the purpose of this call is not exactly to init
 the mutex, but to init the lockdep class in case this inode was
 previously used for a partition but is now being used for a whole
 device.  This makes sense, but renders the mutex_init in 
 init_once pointless.  Maybe that should be removed?

2/ You are introducing a new call to get_gendisk.
   This bothers me for two reasons.  Both relate to a comparison
   with the call to get_gendisk in block_dev.c:do_open.
   a/ That call is protected by lock_kernel.  Your call is not.
   b/ That call is followed by a test for '!disk' implying that it
       can return NULL.  Yours is not - at least not obviously
       (put_disk does have the check).

   I'm not sure if these are actually problems, but the do bother me.

   Thinking through the possibly reasons for the lock_kernel, I wonder
   it the current device number mapping scheme actually allows you
   to determine if something is partitioned or not in a static sense.
   Maybe that is only guaranteed to be stable while the device is
   open...
   I wonder if Al Viro could put my mind at rest .... Al - do you have
   a moment to look at this?  Thanks.

NeilBrown
    
> 
> Signed-off-by: Peter Zijlstra <a.p.zijlstra@...llo.nl>
> Cc: Neil Brown <neilb@....unsw.edu.au>
> Cc: Ingo Molnar <mingo@...e.hu>
> Cc: Arjan van de Ven <arjan@...ux.intel.com>
> Cc: Andrew Morton <akpm@...l.org>
> Cc: Jason Baron <jbaron@...hat.com>
> ---
>  fs/block_dev.c |    9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> Index: linux-2.6-mm/fs/block_dev.c
> ===================================================================
> --- linux-2.6-mm.orig/fs/block_dev.c
> +++ linux-2.6-mm/fs/block_dev.c
> @@ -357,10 +357,14 @@ static int bdev_set(struct inode *inode,
>  
>  static LIST_HEAD(all_bdevs);
>  
> +static struct lock_class_key bdev_part_lock_key;
> +
>  struct block_device *bdget(dev_t dev)
>  {
>  	struct block_device *bdev;
>  	struct inode *inode;
> +	struct gendisk *disk;
> +	int part = 0;
>  
>  	inode = iget5_locked(bd_mnt->mnt_sb, hash(dev),
>  			bdev_test, bdev_set, &dev);
> @@ -386,6 +390,11 @@ struct block_device *bdget(dev_t dev)
>  		list_add(&bdev->bd_list, &all_bdevs);
>  		spin_unlock(&bdev_lock);
>  		unlock_new_inode(inode);
> +		mutex_init(&bdev->bd_mutex);
> +		disk = get_gendisk(dev, &part);
> +		if (part)
> +			lockdep_set_class(&bdev->bd_mutex, &bdev_part_lock_key);
> +		put_disk(disk);
>  	}
>  	return bdev;
>  }
> 
> --
-
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/