linux-kernel - Re: [PATCH v2 1/1] mux: mux-core: Add NULL check for dev->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] [thread-next>] [day] [month] [year] [list]

Message-ID: <27cdacb6-b775-f678-9ff0-6a3f2ad11e17@gmail.com>
Date:   Sun, 9 Jul 2017 00:35:45 -0700
From:   "Kuppuswamy, Sathyanarayanan" <sathyaosid@...il.com>
To:     Peter Rosin <peda@...ntia.se>,
        sathyanarayanan.kuppuswamy@...ux.intel.com
Cc:     linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 1/1] mux: mux-core: Add NULL check for dev->of_node

Hi,


On 7/9/2017 12:07 AM, Peter Rosin wrote:
> On 2017-07-09 01:12, Kuppuswamy, Sathyanarayanan wrote:
>> Hi Peter,
>>
>>
>> On 7/8/2017 2:00 PM, Peter Rosin wrote:
>>> On 2017-07-07 23:46, sathyanarayanan.kuppuswamy@...ux.intel.com wrote:
>>>> From: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@...ux.intel.com>
>>>>
>>>> If dev->of_node is NULL, then calling mux_control_get()
>>>> function can lead to NULL pointer exception. So adding
>>>> a NULL check for dev->of_node.
>>>>
>>>> Signed-off-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@...ux.intel.com>
>>> Do you have a driver that might call mux_control_get and not have any
>>> of_node?
>> For non-device tree drivers, this case is valid. I hit this issue when I
>> was working on Intel USB MUX driver.
>>>    If not, I don't see the point of this check.
>> Since this is an API for other consumers, I think its better to have
>> some sanity checks.
>>
>> If a non device tree driver call this API , I think its better to fail
>> with some error no instead of creating null pointer exception.
> Is it? When authoring a new driver, and you make some error like this, why
> is a "nice" error better than a big fat fail? If you get a null deref,
> you will presumably also get a call stack etc, which will help you find
> where you made the error, w/o adding a bunch of traces to find out exactly
> what you did wrong.
In this case, I think the error can happen even if there is any 
configuration mismatch in device tree blob. So its not only
about API usage in driver. If you think that you need to fail the system 
if the API is used without proper dt node configuration,
  then we should use something like BUG or WARN_ON to explicitly mention 
this dependency.  But I think its better to
leave this decision to the MUX consumers because there use cases where 
MUX control can be optional
>
> So, I'm skeptic...
>
> Cheers,
> peda
>
>>> Cheers,
>>> peda
>>>
>>>> ---
>>>>    drivers/mux/mux-core.c | 3 +++
>>>>    1 file changed, 3 insertions(+)
>>>>
>>>> Changes since v1:
>>>>    * Removed dummy new line.
>>>>
>>>> diff --git a/drivers/mux/mux-core.c b/drivers/mux/mux-core.c
>>>> index 90b8995..924c983 100644
>>>> --- a/drivers/mux/mux-core.c
>>>> +++ b/drivers/mux/mux-core.c
>>>> @@ -438,6 +438,9 @@ struct mux_control *mux_control_get(struct device *dev, const char *mux_name)
>>>>    	int index = 0;
>>>>    	int ret;
>>>>    
>>>> +	if (!np)
>>>> +		return ERR_PTR(-ENODEV);
>>>> +
>>>>    	if (mux_name) {
>>>>    		index = of_property_match_string(np, "mux-control-names",
>>>>    						 mux_name);
>>>>