linux-kernel - Re: [PATCH] drm/imx: Fix of

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] [day] [month] [year] [list]

Message-ID: <1473062797.2456.26.camel@pengutronix.de>
Date:   Mon, 05 Sep 2016 10:06:37 +0200
From:   Philipp Zabel <p.zabel@...gutronix.de>
To:     Christophe JAILLET <christophe.jaillet@...adoo.fr>
Cc:     airlied@...ux.ie, dri-devel@...ts.freedesktop.org,
        linux-kernel@...r.kernel.org, kernel-janitors@...r.kernel.org
Subject: Re: [PATCH] drm/imx: Fix of_node ref counting

Am Montag, den 05.09.2016, 10:01 +0200 schrieb Philipp Zabel:
> Hi Christophe,
> 
> Am Sonntag, den 04.09.2016, 08:45 +0200 schrieb Christophe JAILLET:
> > This code is spurious.
> > It takes a ref on a node, then call 'of_node_put' on it and then store
> > this node somewhere.
> 
> The node pointer is not stored. Note that np is not dereferenced at all,
> we just compare the pointer value against dev->of_node.
> It doesn't matter whether we drop the reference before or after that.
> 
> > It is likely that taking the ref on the parent node and releasing the child
> > node was expected instead.
> 
> Initially, np is assigned to the void *data parameter. The caller holds
> the reference to that, and we are not allowed to decrement its refcount
> (as of_get_next_parent does). Otherwise the iterator calling this match
> function would drop references of all the device_nodes it compares
> against.
> 
> > So, use 'of_get_next_parent' instead. It does all this in just one
> > function call.
> > 
> > Signed-off-by: Christophe JAILLET <christophe.jaillet@...adoo.fr>
> > ---
> > Un-tested
> > ---
> >  drivers/gpu/drm/imx/imx-drm-core.c | 6 ++----
> >  1 file changed, 2 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/gpu/drm/imx/imx-drm-core.c b/drivers/gpu/drm/imx/imx-drm-core.c
> > index 438bac8fbc2b..60fb388c80f8 100644
> > --- a/drivers/gpu/drm/imx/imx-drm-core.c
> > +++ b/drivers/gpu/drm/imx/imx-drm-core.c
> > @@ -449,10 +449,8 @@ static int compare_of(struct device *dev, void *data)
> >  	}
> >  
> >  	/* Special case for LDB, one device for two channels */
> > -	if (of_node_cmp(np->name, "lvds-channel") == 0) {
> > -		np = of_get_parent(np);
> > -		of_node_put(np);
> > -	}
> 
> This could have been written as:
> 
> 	bool match;
> 
> 	/* Special case for LDB, one device for two channels */
> 	if (of_node_cmp(np->name, "lvds-channel") == 0) {
> 		struct device_node *parent = of_get_parent(np);
> 
> 		match = dev->of_node == parent;
> 		of_node_put(parent);
> 	} else {
> 		match = dev->of_node == np;
> 	}
> 
> 	return match;
> 
> which does exactly the same. Maybe the reuse of np and the pointer
> comparison after of_node_put warrants a comment.

Actually, maybe we should just do

-       if (of_node_cmp(np->name, "lvds-channel") == 0) {
-               np = of_get_parent(np);
-               of_node_put(np);
-       }
+       if (of_node_cmp(np->name, "lvds-channel") == 0)
+               np = np->parent;

regards
Philipp