[PATCH] of: handle NULL node in of_get_next_available_child

[PATCH] of: handle NULL node in of_get_next_available_child

[Florian Fainelli]

Add an early check for the node argument in
of_get_next_available_child() to avoid dereferencing a NULL node pointer
a few lines after.

CC: Daniel Mack <zonque-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
Signed-off-by: Florian Fainelli <f.fainelli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
---
 drivers/of/base.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drivers/of/base.c b/drivers/of/base.c
index e67b308819c9..9a71bfac78ae 100644
--- a/drivers/of/base.c
+++ b/drivers/of/base.c
@@ -746,6 +746,9 @@ struct device_node *of_get_next_available_child(const struct device_node *node,
 	struct device_node *next;
 	unsigned long flags;
 
+	if (!node)
+		return NULL;
+
 	raw_spin_lock_irqsave(&devtree_lock, flags);
 	next = prev ? prev->sibling : node->child;
 	for (; next; next = next->sibling) {
-- 
1.9.1

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] of: handle NULL node in of_get_next_available_child

[Grant Likely]

On Fri, 23 May 2014 12:43:11 -0700, Florian Fainelli <f.fainelli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> Add an early check for the node argument in
> of_get_next_available_child() to avoid dereferencing a NULL node pointer
> a few lines after.
> 
> CC: Daniel Mack <zonque-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> Signed-off-by: Florian Fainelli <f.fainelli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>

Is there a bug that exposed this path? I'd like some more justification
before merging this patch, particularly considering of_get_next_child()
doesn't check for it either, but you're not modifying that function.

g.

> ---
>  drivers/of/base.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/drivers/of/base.c b/drivers/of/base.c
> index e67b308819c9..9a71bfac78ae 100644
> --- a/drivers/of/base.c
> +++ b/drivers/of/base.c
> @@ -746,6 +746,9 @@ struct device_node *of_get_next_available_child(const struct device_node *node,
>  	struct device_node *next;
>  	unsigned long flags;
>  
> +	if (!node)
> +		return NULL;
> +
>  	raw_spin_lock_irqsave(&devtree_lock, flags);
>  	next = prev ? prev->sibling : node->child;
>  	for (; next; next = next->sibling) {
> -- 
> 1.9.1
> 

--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] of: handle NULL node in of_get_next_available_child

[Florian Fainelli]

Hi Grant,

2014-05-27 4:19 GMT-07:00 Grant Likely <grant.likely-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>:
> On Fri, 23 May 2014 12:43:11 -0700, Florian Fainelli <f.fainelli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
>> Add an early check for the node argument in
>> of_get_next_available_child() to avoid dereferencing a NULL node pointer
>> a few lines after.
>>
>> CC: Daniel Mack <zonque-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>> Signed-off-by: Florian Fainelli <f.fainelli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
>
> Is there a bug that exposed this path?

Daniel sent a patch recently [1] which makes us call
for_each_available_child_of_node() on a potentially NULL node
argument.

The reason for that is that the loop is supposed to iterate over child
nodes (Ethernet PHY nodes) of a given MDIO bus node. There might be
non-DT probed MDIO buses registered in the system (e.g: the fixed MDIO
bus), which do not have a valid of_node pointer. Since then, Daniel
updated his patch to also check for the MDIO bus device_node pointer
to be non-NULL, but I feel like we should also but defensive here.
Presumably this situation could happen with e.g: i2c busses, spi
busses in a system that mixes both DT-probed and non-DT probed
devices.


> I'd like some more justification
> before merging this patch, particularly considering of_get_next_child()
> doesn't check for it either, but you're not modifying that function.

Good catch, it should also be updated.

[1]: http://patchwork.ozlabs.org/patch/347712/

>
> g.
>
>> ---
>>  drivers/of/base.c | 3 +++
>>  1 file changed, 3 insertions(+)
>>
>> diff --git a/drivers/of/base.c b/drivers/of/base.c
>> index e67b308819c9..9a71bfac78ae 100644
>> --- a/drivers/of/base.c
>> +++ b/drivers/of/base.c
>> @@ -746,6 +746,9 @@ struct device_node *of_get_next_available_child(const struct device_node *node,
>>       struct device_node *next;
>>       unsigned long flags;
>>
>> +     if (!node)
>> +             return NULL;
>> +
>>       raw_spin_lock_irqsave(&devtree_lock, flags);
>>       next = prev ? prev->sibling : node->child;
>>       for (; next; next = next->sibling) {
>> --
>> 1.9.1
>>
>



-- 
Florian
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH] of: handle NULL node in of_get_next_available_child

[Grant Likely]

On Tue, 27 May 2014 10:36:10 -0700, Florian Fainelli <f.fainelli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> Hi Grant,
> 
> 2014-05-27 4:19 GMT-07:00 Grant Likely <grant.likely-QSEj5FYQhm4dnm+yROfE0A@public.gmane.org>:
> > On Fri, 23 May 2014 12:43:11 -0700, Florian Fainelli <f.fainelli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org> wrote:
> >> Add an early check for the node argument in
> >> of_get_next_available_child() to avoid dereferencing a NULL node pointer
> >> a few lines after.
> >>
> >> CC: Daniel Mack <zonque-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> >> Signed-off-by: Florian Fainelli <f.fainelli-Re5JQEeQqe8AvxtiuMwx3w@public.gmane.org>
> >
> > Is there a bug that exposed this path?
> 
> Daniel sent a patch recently [1] which makes us call
> for_each_available_child_of_node() on a potentially NULL node
> argument.

And yet in your reply to that patch you argued that the NULL check
should be in the caller!

Anyway, this isn't a big deal. Post a patch fixing both and I'll apply
it.

g.
--
To unsubscribe from this list: send the line "unsubscribe devicetree" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html