Re: linux-next: Tree for May 24 (ratelimits, CONFIG_PRINTK not enabled)

Re: linux-next: Tree for May 24 (ratelimits, CONFIG_PRINTK not enabled)

[Randy Dunlap]

On Tue, 24 May 2011 14:02:12 +1000 Stephen Rothwell wrote:

> Hi all,
> 
> Changes since 20110523:

when CONFIG_PRINTK is not enabled:

net/core/filter.c:353: warning: type defaults to 'int' in declaration of 'DEFINE_RATELIMIT_STATE'
net/core/filter.c:353: warning: parameter names (without types) in function declaration
net/core/filter.c:353: error: invalid storage class for function 'DEFINE_RATELIMIT_STATE'
net/core/filter.c:353: error: implicit declaration of function '__ratelimit'
net/core/filter.c:353: error: '_rs' undeclared (first use in this function)


Adding <linux/ratelimit.h> to <asm-generic/bug.h> causes other problems:

In file included from linux-next-20110524/include/linux/thread_info.h:53,
                 from linux-next-20110524/include/linux/preempt.h:9,
                 from linux-next-20110524/include/linux/spinlock.h:50,
                 from linux-next-20110524/include/linux/ratelimit.h:5,
                 from linux-next-20110524/include/asm-generic/bug.h:5,
                 from linux-next-20110524/arch/x86/include/asm/bug.h:38,
                 from linux-next-20110524/include/linux/kernel.h:23,
                 from linux-next-20110524/arch/x86/include/asm/percpu.h:44,
                 from linux-next-20110524/arch/x86/include/asm/current.h:5,
                 from linux-next-20110524/arch/x86/include/asm/processor.h:15,
                 from linux-next-20110524/arch/x86/include/asm/atomic.h:6,
                 from linux-next-20110524/include/linux/crypto.h:20,
                 from linux-next-20110524/arch/x86/kernel/asm-offsets.c:8:
linux-next-20110524/arch/x86/include/asm/thread_info.h:34: error: expected specifier-qualifier-list before 'mm_segment_t'
In file included from linux-next-20110524/include/linux/thread_info.h:53,
                 from linux-next-20110524/include/linux/preempt.h:9,
                 from linux-next-20110524/include/linux/spinlock.h:50,
                 from linux-next-20110524/include/linux/ratelimit.h:5,
                 from linux-next-20110524/include/asm-generic/bug.h:5,
                 from linux-next-20110524/arch/x86/include/asm/bug.h:38,
                 from linux-next-20110524/include/linux/kernel.h:23,
                 from linux-next-20110524/arch/x86/include/asm/percpu.h:44,
                 from linux-next-20110524/arch/x86/include/asm/current.h:5,
                 from linux-next-20110524/arch/x86/include/asm/processor.h:15,
                 from linux-next-20110524/arch/x86/include/asm/atomic.h:6,
                 from linux-next-20110524/include/linux/crypto.h:20,
                 from linux-next-20110524/arch/x86/kernel/asm-offsets.c:8:
linux-next-20110524/arch/x86/include/asm/thread_info.h:217: error: expected declaration specifiers or '...' before 'kernel_stack'
linux-next-20110524/arch/x86/include/asm/thread_info.h:217: warning: data definition has no type or storage class
linux-next-20110524/arch/x86/include/asm/thread_info.h:217: warning: type defaults to 'int' in declaration of 'DECLARE_PER_CPU'
linux-next-20110524/arch/x86/include/asm/thread_info.h: In function 'current_thread_info':
linux-next-20110524/arch/x86/include/asm/thread_info.h:222: error: implicit declaration of function 'percpu_read_stable'
linux-next-20110524/arch/x86/include/asm/thread_info.h:222: error: 'kernel_stack' undeclared (first use in this function)


---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

Re: linux-next: Tree for May 24 (ratelimits, CONFIG_PRINTK not enabled)

[Joe Perches]

On Tue, 2011-05-24 at 11:33 -0700, Randy Dunlap wrote:
> On Tue, 24 May 2011 14:02:12 +1000 Stephen Rothwell wrote:
> > Hi all,
> > Changes since 20110523:
> when CONFIG_PRINTK is not enabled:
> net/core/filter.c:353: warning: type defaults to 'int' in declaration of 'DEFINE_RATELIMIT_STATE'
> net/core/filter.c:353: warning: parameter names (without types) in function declaration
> net/core/filter.c:353: error: invalid storage class for function 'DEFINE_RATELIMIT_STATE'
> net/core/filter.c:353: error: implicit declaration of function '__ratelimit'
> net/core/filter.c:353: error: '_rs' undeclared (first use in this function)

Thanks for the report Randy.

I wonder if it's better to new create sections for
!CONFIG_PRINTK in bug.h.  That might make the image
a bit smaller when !CONFIG_PRINTK too.

> Adding <linux/ratelimit.h> to <asm-generic/bug.h> causes other problems:

Yup, that's not a good solution.

Re: linux-next: Tree for May 24 (ratelimits, CONFIG_PRINTK not enabled)

[David Miller]

From: Joe Perches <joe@perches.com>
Date: Tue, 24 May 2011 11:49:48 -0700

> On Tue, 2011-05-24 at 11:33 -0700, Randy Dunlap wrote:
>> Adding <linux/ratelimit.h> to <asm-generic/bug.h> causes other problems:
> 
> Yup, that's not a good solution.

Please come up with a fix or I'll have to revert, thanks Joe.

Re: linux-next: Tree for May 24 (ratelimits, CONFIG_PRINTK not enabled)

[Joe Perches]

On Tue, 2011-05-24 at 15:32 -0400, David Miller wrote:
> From: Joe Perches <joe@perches.com>
> Date: Tue, 24 May 2011 11:49:48 -0700
> > On Tue, 2011-05-24 at 11:33 -0700, Randy Dunlap wrote:
> >> Adding <linux/ratelimit.h> to <asm-generic/bug.h> causes other problems:
> > Yup, that's not a good solution.
> Please come up with a fix or I'll have to revert, thanks Joe.

I have some errands to run for awhile.

Probably a revert today and another patch set tomorrow
might be best.

This seems to work and doesn't cause any extra state to be
instantiated when !CONFIG_PRINTK.  What do you think?

#ifdef CONFIG_PRINTK

#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...)			\
({								\
	static DEFINE_RATELIMIT_STATE(_rs,			\
				      DEFAULT_RATELIMIT_INTERVAL,	\
				      DEFAULT_RATELIMIT_BURST);	\
	__WARN_RATELIMIT(condition, &_rs, format);		\
})

#else

#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...)			\
({								\
	int rtn = WARN(condition, format);			\
	rtn;							\
})

#endif

Re: linux-next: Tree for May 24 (ratelimits, CONFIG_PRINTK not enabled)

[David Miller]

From: Joe Perches <joe@perches.com>
Date: Tue, 24 May 2011 12:49:40 -0700

> On Tue, 2011-05-24 at 15:32 -0400, David Miller wrote:
>> From: Joe Perches <joe@perches.com>
>> Date: Tue, 24 May 2011 11:49:48 -0700
>> > On Tue, 2011-05-24 at 11:33 -0700, Randy Dunlap wrote:
>> >> Adding <linux/ratelimit.h> to <asm-generic/bug.h> causes other problems:
>> > Yup, that's not a good solution.
>> Please come up with a fix or I'll have to revert, thanks Joe.
> 
> I have some errands to run for awhile.
> 
> Probably a revert today and another patch set tomorrow
> might be best.
> 
> This seems to work and doesn't cause any extra state to be
> instantiated when !CONFIG_PRINTK.  What do you think?

Seems fine, here it is as a patch.

Randy please test this, thanks:

diff --git a/include/asm-generic/bug.h b/include/asm-generic/bug.h
index 12b250c..9178484 100644
--- a/include/asm-generic/bug.h
+++ b/include/asm-generic/bug.h
@@ -162,6 +162,8 @@ extern void warn_slowpath_null(const char *file, const int line);
 	unlikely(__ret_warn_once);				\
 })
 
+#ifdef CONFIG_PRINTK
+
 #define WARN_ON_RATELIMIT(condition, state)			\
 		WARN_ON((condition) && __ratelimit(state))
 
@@ -181,6 +183,25 @@ extern void warn_slowpath_null(const char *file, const int line);
 	__WARN_RATELIMIT(condition, &_rs, format);		\
 })
 
+#else
+
+#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...)			\
+({								\
+	int rtn = WARN(condition, format);			\
+	rtn;							\
+})
+
+#endif
+
 /*
  * WARN_ON_SMP() is for cases that the warning is either
  * meaningless for !SMP or may even cause failures.

Re: linux-next: Tree for May 24 (ratelimits, CONFIG_PRINTK not enabled)

[Randy Dunlap]

On 05/24/11 12:53, David Miller wrote:
> From: Joe Perches <joe@perches.com>
> Date: Tue, 24 May 2011 12:49:40 -0700
> 
>> On Tue, 2011-05-24 at 15:32 -0400, David Miller wrote:
>>> From: Joe Perches <joe@perches.com>
>>> Date: Tue, 24 May 2011 11:49:48 -0700
>>>> On Tue, 2011-05-24 at 11:33 -0700, Randy Dunlap wrote:
>>>>> Adding <linux/ratelimit.h> to <asm-generic/bug.h> causes other problems:
>>>> Yup, that's not a good solution.
>>> Please come up with a fix or I'll have to revert, thanks Joe.
>>
>> I have some errands to run for awhile.
>>
>> Probably a revert today and another patch set tomorrow
>> might be best.
>>
>> This seems to work and doesn't cause any extra state to be
>> instantiated when !CONFIG_PRINTK.  What do you think?
> 
> Seems fine, here it is as a patch.
> 
> Randy please test this, thanks:

Works here.

Acked-by: Randy Dunlap <randy.dunlap@oracle.com>

Thanks.

> diff --git a/include/asm-generic/bug.h b/include/asm-generic/bug.h
> index 12b250c..9178484 100644
> --- a/include/asm-generic/bug.h
> +++ b/include/asm-generic/bug.h
> @@ -162,6 +162,8 @@ extern void warn_slowpath_null(const char *file, const int line);
>  	unlikely(__ret_warn_once);				\
>  })
>  
> +#ifdef CONFIG_PRINTK
> +
>  #define WARN_ON_RATELIMIT(condition, state)			\
>  		WARN_ON((condition) && __ratelimit(state))
>  
> @@ -181,6 +183,25 @@ extern void warn_slowpath_null(const char *file, const int line);
>  	__WARN_RATELIMIT(condition, &_rs, format);		\
>  })
>  
> +#else
> +
> +#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...)			\
> +({								\
> +	int rtn = WARN(condition, format);			\
> +	rtn;							\
> +})
> +
> +#endif
> +
>  /*
>   * WARN_ON_SMP() is for cases that the warning is either
>   * meaningless for !SMP or may even cause failures.


-- 
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***

Re: linux-next: Tree for May 24 (ratelimits, CONFIG_PRINTK not enabled)

[David Miller]

From: Randy Dunlap <randy.dunlap@oracle.com>
Date: Tue, 24 May 2011 13:11:51 -0700

> On 05/24/11 12:53, David Miller wrote:
>> From: Joe Perches <joe@perches.com>
>> Date: Tue, 24 May 2011 12:49:40 -0700
>> 
>>> On Tue, 2011-05-24 at 15:32 -0400, David Miller wrote:
>>>> From: Joe Perches <joe@perches.com>
>>>> Date: Tue, 24 May 2011 11:49:48 -0700
>>>>> On Tue, 2011-05-24 at 11:33 -0700, Randy Dunlap wrote:
>>>>>> Adding <linux/ratelimit.h> to <asm-generic/bug.h> causes other problems:
>>>>> Yup, that's not a good solution.
>>>> Please come up with a fix or I'll have to revert, thanks Joe.
>>>
>>> I have some errands to run for awhile.
>>>
>>> Probably a revert today and another patch set tomorrow
>>> might be best.
>>>
>>> This seems to work and doesn't cause any extra state to be
>>> instantiated when !CONFIG_PRINTK.  What do you think?
>> 
>> Seems fine, here it is as a patch.
>> 
>> Randy please test this, thanks:
> 
> Works here.
> 
> Acked-by: Randy Dunlap <randy.dunlap@oracle.com>

Thanks Randy, once my allmodconfig+!CONFIG_PRINTK build finishes
I'll push this out.