Al Viro wrote:
> On Mon, Sep 04, 2017 at 09:12:38PM +0900, Tetsuo Handa wrote:
> > Hello.
> >
> > I hit kmemcheck splat on commit 49d31c2f389acfe8 ("dentry name snapshots") using linux-next-20170901.
> > Changing to strncpy() fixes this problem, but using strncpy() only if CONFIG_KMEMCHECK=y is better
> > for performance?
>
> *shrug*
>
> If anything, use dentry->d_name.len + 1 instead of DNAME_INLINE_LEN there, but
> that's really a false positive.
A false positive? Isn't it because we do
dentry = kmem_cache_alloc(dentry_cache, GFP_KERNEL);
(...snipped...)
dname = dentry->d_iname;
(...snipped...)
memcpy(dname, name->name, name->len);
dname[name->len] = 0;
at __d_alloc() which leaves a room for some of d_iname[] bytes uninitialized?
So, I think either pad explicitly at __f_alloc() or use dentry->d_name.len + 1 is needed.
