Re: [PATCH 2/2] ASoC: Add BUG() assertion if max98095_get_bq_channelreturns -EINVAL

[Ryan Mallon]

On 29/09/11 11:59, Dave Young wrote:

> On Thu, Sep 29, 2011 at 9:52 AM, Ryan Mallon <rmallon@xxxxxxxxx> wrote:
>> On 29/09/11 11:35, Dave Young wrote:
>>
>>> On Thu, Sep 29, 2011 at 7:19 AM, Ryan Mallon <rmallon@xxxxxxxxx> wrote:
>>>> On 29/09/11 00:02, Axel Lin wrote:
>>>>> The callers use the return value of max98095_get_bq_channel as array index to
>>>>> access max98095->dai[] array.
>>>>> Add BUG() assertion for out of bound access of max98095->dai[] array.
>>>>
>>>> Same here, fix the problem in the callers.
>>>>
>>>> ----
>>>> Check the return value of max98095_get_bq_channel in the callers and
>>>> propagate any errors up. Remove the BUG_ON(channel > 1) since
>>>> max98095_get_bq_channel never returns a value larger than 1.
>>>>
>>>> Signed-off-by: Ryan Mallon <rmallon@xxxxxxxxx>
>>>> ---
>>>>
>>>> diff --git a/sound/soc/codecs/max98095.c b/sound/soc/codecs/max98095.c
>>>> index 668434d..55eccea 100644
>>>> --- a/sound/soc/codecs/max98095.c
>>>> +++ b/sound/soc/codecs/max98095.c
>>>> @@ -2014,7 +2014,8 @@ static int max98095_put_bq_enum(struct snd_kcontrol *kcontrol,
>>>>        int fs, best, best_val, i;
>>>>        int regmask, regsave;
>>>>
>>>> -       BUG_ON(channel > 1);
>>>> +       if (channel < 0)
>>>> +               return channel;
>>>
>>> If use BUG() happens in  max98095_get_bq_channel, it will not return here?
>>
>>
>> Not quite sure what you mean?
> 
> I means if Axel Lin's patch applied, and CONFIG_BUG is on, it will
> panic firstly the if condition will be never entered.

My patch is a replacement for Axel's patch, not on top of it. For Axel's
patch it would panic if channel was less than zero if CONFIG_BUG was
enabled, but would still have undefined behaviour if CONFIG_BUG was not
enabled.

~Ryan

>>
>> If CONFIG_BUG was not enabled for the original version, then it would
>> not return at the BUG_ON and would either crash or cause odd behaviour
>> if it tried to index channel as -1.
>>
>> My patch is removing the BUG_ON and replacing it with a proper check and
>> return. It doesn't need to check > 1 since max98095_get_bq_channel never
>> returns that.
>>
>> My understanding is that device drivers, in general, should not call
>> BUG. BUG is for unrecoverable errors which leave the kernel in some
>> unstable state. Here we can just return an error code.
> 
> Agree
> 
>>
>> ~Ryan
>>
>>>
>>>>
>>>>        if (!pdata || !max98095->bq_textcnt)
>>>>                return 0;
>>>> @@ -2069,6 +2070,9 @@ static int max98095_get_bq_enum(struct snd_kcontrol *kcontrol,
>>>>        int channel = max98095_get_bq_channel(kcontrol->id.name);
>>>>        struct max98095_cdata *cdata;
>>>>
>>>> +       if (channel < 0)
>>>> +               return channel;
>>>> +
>>>>        cdata = &max98095->dai[channel];
>>>>        ucontrol->value.enumerated.item[0] = cdata->bq_sel;
>>>>
>>>>
>>>> --
>>>> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
>>>> the body of a message to majordomo@xxxxxxxxxxxxxxx
>>>> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>>>> Please read the FAQ at  http://www.tux.org/lkml/
>>>>
>>>
>>>
>>>
>>
>>
>>
> 
> 
> 


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/