Re: [PATCH] Fix i_mutex handling in nfsd readdir.

[David Woodhouse <dwmw2@infradead.org>]

On Sat, 2009-04-18 at 12:11 +0900, hooanon05@yahoo.co.jp wrote:
> David Woodhouse:
> > This patch fixes it by locking the directory's i_mutex again before
> > calling the filldir functions. The original deadlocks which commit
> > 14f7dd63 was designed to avoid are still avoided, because they were due
> > to fs-internal locking, not i_mutex.
> 	:::
> > --- a/fs/namei.c
> > +++ b/fs/namei.c
> > @@ -1248,6 +1248,8 @@ struct dentry *lookup_one_len(const char *name, struct dentry *base, int len)
> >  	int err;
> >  	struct qstr this;
> >  
> > +	WARN_ON_ONCE(!mutex_is_locked(&base->d_inode->i_mutex));
> > +
> >  	err = __lookup_one_len(name, &this, base, len);
> 
> I'd suggest this checking is done only when CONFIG_DEBUG_KERNEL (or
> something) is enabled. Because unconditional checking costs high for the
> well-reviewed lookup code.

It's supposed to be locked. It's likely to have been locked quite
recently, so it'll be in the cache. I don't think the mutex_is_locked()
check is going to be that expensive, is it?

-- 
David Woodhouse                            Open Source Technology Centre
David.Woodhouse@intel.com                              Intel Corporation