Re: [rfc][patch] store-free path walking

[Jens Axboe <jens.axboe@oracle.com>]

On Fri, Oct 09 2009, Nick Piggin wrote:
> On Fri, Oct 09, 2009 at 11:51:19AM +0200, Jens Axboe wrote:
> > Nick,
> > 
> > One more thing... I see you converted part of btrfs, but there's still
> > this one sitting in btrfs_invalidate_inodes()
> > 
> >         if (atomic_read(&inode->i_count) > 1)
> >                 d_prune_aliases(inode);
> > 
> > Not sure how best to solve that, with a __d_prune_aliases() that assumed
> > the lock was held it would be easy. But perhaps you have better ideas,
> > this email is more of a heads-up since perhaps you missed this spot
> > (CC'ing Chris).
> 
> It's OK, you can load inode->i_count integer atomically -- in this
> sequence d_prune_aliases can't have assumed anything about i_count
> anyway because regardless of its type it might have changed in
> between.

Right, it was already racy wrt i_count. So the below should be OK.

diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index d3dadb6..7e02b5d 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -3419,8 +3419,10 @@ again:
 		inode = igrab(&entry->vfs_inode);
 		if (inode) {
 			spin_unlock(&root->inode_lock);
-			if (atomic_read(&inode->i_count) > 1)
+
+			if (inode->i_count > 1)
 				d_prune_aliases(inode);
+
 			/*
 			 * btrfs_drop_inode will remove it from
 			 * the inode cache when its usage count

-- 
Jens Axboe