Re: [PATCH v2 09/19] ARM: at91: make sdram/ddr register base soc independent

[Nicolas Ferre <nicolas.ferre@atmel.com>]

On 02/23/2012 11:51 AM, Ryan Mallon :
> On 23/02/12 20:58, Nicolas Ferre wrote:
> 
>> On 02/23/2012 09:56 AM, Nicolas Ferre :
>>> On 02/22/2012 11:33 PM, Ryan Mallon :
>>>> On 22/02/12 20:39, Nicolas Ferre wrote:
>>>>
>>>>> From: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
>>>>>
>>>>> Signed-off-by: Jean-Christophe PLAGNIOL-VILLARD <plagnioj@jcrosoft.com>
>>>>> Acked-by: Nicolas Ferre <nicolas.ferre@atmel.com>
>>>>> ---
>>>>
>>>> <snip>
>>>>
>>>>> +void __init at91_ioremap_ramc(int id, u32 addr, u32 size)
>>>>> +{
>>>>> +	if (id > 1) {
>>>>> +		pr_warn("%s: id > 2\n", __func__);
>>>>> +		return;
>>>>> +	}
>>>>> +	at91_ramc_base[id] = ioremap(addr, size);
>>>>> +	if (!at91_ramc_base[id])
>>>>> +		pr_warn("Impossible to ioremap ramc.%d 0x%x\n", id, addr);
>>>>> +}
>>>>
>>>>
>>>> If this fails then you will oops if you call either at91_ramc_read or
>>>> at91_ramc_write since they don't check if at91_ramc_base[id] is a valid
>>>> pointer. Either this function should panic, like the other at91_ioremap
>>>> functions, or the at91_ramc_read/write functions should check for a
>>>> valid pointer.
>>>
>>> Yes, as you pointed out, it is done in a not-related following patch.
>>> I will bring the code here.
>>>
>>>
>>>> Nitpick: The id check should probably also be BUG() or WARN() since it
>>>> indicates a bug in the core AT91 code. pr_warn is likely to missed and
>>>> not reported by users. Since the value is int, the check should be:
>>>>
>>>>   if (id < 0 || id > 1)
>>>>
>>>> Obviously the chance of this error happening are slim, but if you are
>>>> going to check and warn for it, it should be done properly :-).
>>>
>>> Yes, I agree and modify it at the very moment.
>>
>>
>> What do you think about that:
>> If id is not setup properly, I try to find a way out by using id = 0...
>> Then, a panic is added if the iremap() fails.
>>
>>
>> --- a/arch/arm/mach-at91/pm.c
>> +++ b/arch/arm/mach-at91/pm.c
>> @@ -200,13 +200,14 @@ void __iomem *at91_ramc_base[2];
>>  
>>  void __init at91_ioremap_ramc(int id, u32 addr, u32 size)
>>  {
>> -       if (id > 1) {
>> -               pr_warn("%s: id > 2\n", __func__);
>> -               return;
>> +       if (id < 0 || id > 1) {
>> +               WARN(1, "%s: Wrong RAM controller id (%d), set it to 0\n",
>> +                                                               __func__, id);
>> +               id = 0;
> 
> 
> I don't think you should try to fix the id, just issue the warning and
> return. 

Well, I was trying hard to find a way out: but you may be right: a plain
and good-old BUG() should be a safer solution!

Bye,
-- 
Nicolas Ferre