Re: tty_init_dev: 24 callbacks suppressed

[Jiri Slaby <jslaby@suse.cz>]

CCing Joe.

On 10/04/2012 03:11 PM, Markus Trippelsdorf wrote:
> On 2012.10.04 at 14:40 +0200, Borislav Petkov wrote:
>> On Thu, Oct 04, 2012 at 01:51:57PM +0200, Markus Trippelsdorf wrote:
>>> diff --git a/include/linux/ratelimit.h b/include/linux/ratelimit.h
>>> index e11ccb4..d8de255 100644
>>> --- a/include/linux/ratelimit.h
>>> +++ b/include/linux/ratelimit.h
>>> @@ -46,20 +46,17 @@ extern int ___ratelimit(struct ratelimit_state *rs, const char *func);
>>>  #define WARN_ON_RATELIMIT(condition, state)			\
>>>  		WARN_ON((condition) && __ratelimit(state))
>>>  
>>> -#define __WARN_RATELIMIT(condition, state, format...)		\
>>> -({								\
>>> -	int rtn = 0;						\
>>> -	if (unlikely(__ratelimit(state)))			\
>>> -		rtn = WARN(condition, format);			\
>>> -	rtn;							\
>>> -})
>>> -
>>> -#define WARN_RATELIMIT(condition, format...)			\
>>> +#define WARN_RATELIMIT(condition, fmt, ...)			\
>>>  ({								\
>>>  	static DEFINE_RATELIMIT_STATE(_rs,			\
>>>  				      DEFAULT_RATELIMIT_INTERVAL,	\
>>>  				      DEFAULT_RATELIMIT_BURST);	\
>>> -	__WARN_RATELIMIT(condition, &_rs, format);		\
>>> +	int rtn = !!(condition);				\
>>> +								\
>>> +	if (unlikely(rtn && __ratelimit(&_rs)))		\
>>> +		WARN(rtn, fmt, ##__VA_ARGS__);			\
>>> +								\
>>> +	rtn;							\
>>>  })
>>
>> Aha, I see it. We need to look at the condition before the __ratelimit,
>> otherwise we WARN unnecessarily, good catch.
>>
>>>  #else
>>> @@ -67,15 +64,9 @@ extern int ___ratelimit(struct ratelimit_state *rs, const char *func);
>>>  #define WARN_ON_RATELIMIT(condition, state)			\
>>>  	WARN_ON(condition)
>>>  
>>> -#define __WARN_RATELIMIT(condition, state, format...)		\
>>> -({								\
>>> -	int rtn = WARN(condition, format);			\
>>> -	rtn;							\
>>> -})
>>> -
>>> -#define WARN_RATELIMIT(condition, format...)			\
>>> +#define WARN_RATELIMIT(condition, fmt, ...)			\
>>
>> ... except this change is unrelated and unneeded - there's enough room
>> in 80 cols to leave it as "format" instead of shortening it.
>>
>> Other than that:
>>
>> Acked-and-tested-by: Borislav Petkov <borislav.petkov@amd.com>
> 
> I'll let Jiri handle this :). It's his patch anyway.

Actually this is Joe's version of the patch. Joe, people started hitting
the bug [1]. Could you resend your patch?

[1] https://patchwork.kernel.org/patch/1339221/

BTW what scares me that nobody noticed that bug until this is in the
Linus's tree. Do people use -next at all or am I the only one user? (I
didn't hit it as I have the patch in my local queue.)

thanks,
-- 
js
suse labs