Re: [PATCH] dcache: separate slab for directory dentries

[Dave Kleikamp <shaggy@austin.ibm.com>]

On Fri, 2005-10-07 at 13:54 +1000, Greg Banks wrote:
> On Thu, 2005-10-06 at 22:28, Dave Kleikamp wrote:
> > On Thu, 2005-10-06 at 16:27 +1000, David Chinner wrote:
> > > +static struct dentry * d_realloc_for_inode(struct dentry * dentry,
> > > +					   struct inode *inode)
> > > +{
> > > +	int flags = 0;
> > > +	struct dentry *new;
> > > [...]
> > > +	new = __d_alloc(parent, &dentry->d_name, dentry->d_flags | flags);
> > > +
> > > +	spin_lock(&dcache_lock);
> > > +
> > > +	BUG_ON(new == NULL);	/* TODO */
> > > +	if (new) {
> > > +//		new->d_op = dentry->d_op;
> > > +//		new->d_fsdata = dentry->d_fsdata;
> > > +	}
> > > +	
> > > +	return new;
> > > +}
> > 
> > Isn't this leaking the original dentry?  Shouldn't it be doing a dput or
> > at least a d_free here?
> 
> While it has been some months since I wrote this, and it was not
> intended to be a production patch, I don't believe it leaks dentries.
> 
> The function is called on the path real_lookup() to i_op->lookup() to
> d_splice_alias(), and the original dentry allocated in real_lookup()
> is d_put() there if a non-NULL new dentry pointer is passed back up
> via d_splice_alias() and i_op->lookup().

Yes, you are right.  I've always found the dcache code a little
confusing.  Your patch looks good to me.
-- 
David Kleikamp
IBM Linux Technology Center

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>