Re: Re: [PATCH] mips: add arch_trigger_all_cpu_backtrace() function

Re: Re: [PATCH] mips: add arch_trigger_all_cpu_backtrace() function

[Eunbong Song]

> Hi Eubong,

> one small question inline ...

>> +void arch_trigger_all_cpu_backtrace(bool); +#define
>> arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace

> What is the purpose of this define ? is this maybe a leftover from
> some regex/cleanups ?

Hi John.
Actually, I just follow the same function of sparc architecture.
You can find this in arch/sparc/include/asm/irq_64.h as below

void arch_trigger_all_cpu_backtrace(bool);
#define arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace

I guess this is used for conditional compile. 
See below.
include/linux/nmi.h
#ifdef arch_trigger_all_cpu_backtrace
static inline bool trigger_all_cpu_backtrace(void)
{
        arch_trigger_all_cpu_backtrace(true);

        return true;
}
static inline bool trigger_allbutself_cpu_backtrace(void)
{
        arch_trigger_all_cpu_backtrace(false);
        return true;
}
#else
static inline bool trigger_all_cpu_backtrace(void)
{
        return false;
}
static inline bool trigger_allbutself_cpu_backtrace(void)
{
        return false;
}
#endif

Thanks. 
> John

Re: [PATCH] mips: add arch_trigger_all_cpu_backtrace() function

[John Crispin]

On 22/10/2014 08:54, Eunbong Song wrote:
> 
>> Hi Eubong,
> 
>> one small question inline ...
> 
>>> +void arch_trigger_all_cpu_backtrace(bool); +#define
>>> arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace
> 
>> What is the purpose of this define ? is this maybe a leftover from
>> some regex/cleanups ?
> 
> Hi John.
> Actually, I just follow the same function of sparc architecture.
> You can find this in arch/sparc/include/asm/irq_64.h as below
> 
> void arch_trigger_all_cpu_backtrace(bool);
> #define arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace
> 
> I guess this is used for conditional compile. 
> See below.
> include/linux/nmi.h
> #ifdef arch_trigger_all_cpu_backtrace
> static inline bool trigger_all_cpu_backtrace(void)
> {
>         arch_trigger_all_cpu_backtrace(true);
> 
>         return true;
> }
> static inline bool trigger_allbutself_cpu_backtrace(void)
> {
>         arch_trigger_all_cpu_backtrace(false);
>         return true;
> }
> #else
> static inline bool trigger_all_cpu_backtrace(void)
> {
>         return false;
> }
> static inline bool trigger_allbutself_cpu_backtrace(void)
> {
>         return false;
> }
> #endif
> 
> Thanks. 
>> John
> 

i don't see how this is required for conditional compiles. the code
define a->a which is bogus i think.

	John

Re: [PATCH] mips: add arch_trigger_all_cpu_backtrace() function

[James Hogan]

Hi John,

On Wed, Oct 22, 2014 at 09:11:39AM +0200, John Crispin wrote:
> On 22/10/2014 08:54, Eunbong Song wrote:
> >>> +void arch_trigger_all_cpu_backtrace(bool); +#define
> >>> arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace
> > 
> >> What is the purpose of this define ? is this maybe a leftover from
> >> some regex/cleanups ?
> > 
> > Hi John.
> > Actually, I just follow the same function of sparc architecture.
> > You can find this in arch/sparc/include/asm/irq_64.h as below
> > 
> > void arch_trigger_all_cpu_backtrace(bool);
> > #define arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace
> > 
> > I guess this is used for conditional compile. 
> > See below.
> > include/linux/nmi.h
> > #ifdef arch_trigger_all_cpu_backtrace
> > static inline bool trigger_all_cpu_backtrace(void)
> > {
> >         arch_trigger_all_cpu_backtrace(true);
> > 
> >         return true;
> > }
> > static inline bool trigger_allbutself_cpu_backtrace(void)
> > {
> >         arch_trigger_all_cpu_backtrace(false);
> >         return true;
> > }
> > #else
> > static inline bool trigger_all_cpu_backtrace(void)
> > {
> >         return false;
> > }
> > static inline bool trigger_allbutself_cpu_backtrace(void)
> > {
> >         return false;
> > }
> > #endif
> > 
> > Thanks. 
> >> John
> > 
> 
> i don't see how this is required for conditional compiles. the code
> define a->a which is bogus i think.

It's a pretty common pattern in the asm headers, in order to allow
generic code to use the preprocessor to see whether it was defined or
not.

Cheers
James

Re: [PATCH] mips: add arch_trigger_all_cpu_backtrace() function

[James Hogan]

Hi John,

On Wed, Oct 22, 2014 at 09:11:39AM +0200, John Crispin wrote:
> On 22/10/2014 08:54, Eunbong Song wrote:
> >>> +void arch_trigger_all_cpu_backtrace(bool); +#define
> >>> arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace
> > 
> >> What is the purpose of this define ? is this maybe a leftover from
> >> some regex/cleanups ?
> > 
> > Hi John.
> > Actually, I just follow the same function of sparc architecture.
> > You can find this in arch/sparc/include/asm/irq_64.h as below
> > 
> > void arch_trigger_all_cpu_backtrace(bool);
> > #define arch_trigger_all_cpu_backtrace arch_trigger_all_cpu_backtrace
> > 
> > I guess this is used for conditional compile. 
> > See below.
> > include/linux/nmi.h
> > #ifdef arch_trigger_all_cpu_backtrace
> > static inline bool trigger_all_cpu_backtrace(void)
> > {
> >         arch_trigger_all_cpu_backtrace(true);
> > 
> >         return true;
> > }
> > static inline bool trigger_allbutself_cpu_backtrace(void)
> > {
> >         arch_trigger_all_cpu_backtrace(false);
> >         return true;
> > }
> > #else
> > static inline bool trigger_all_cpu_backtrace(void)
> > {
> >         return false;
> > }
> > static inline bool trigger_allbutself_cpu_backtrace(void)
> > {
> >         return false;
> > }
> > #endif
> > 
> > Thanks. 
> >> John
> > 
> 
> i don't see how this is required for conditional compiles. the code
> define a->a which is bogus i think.

It's a pretty common pattern in the asm headers, in order to allow
generic code to use the preprocessor to see whether it was defined or
not.

Cheers
James

Re: Re: [PATCH] mips: add arch_trigger_all_cpu_backtrace() function

[Eunbong Song]

>> This patch adds arch_trigger_all_cpu_backtrace() for mips architecture.

> Don't forget your Signed-off-by

I'm sorry fot this. 

>> +static void arch_dump_stack(void *info)
>> +{
>> + struct pt_regs *regs;  
>> + 
>> + regs = get_irq_regs();
>> +
>> + if(regs)
>> + show_regs(regs);
>> +
>> + dump_stack();
>> +}
>> +
>> +void arch_trigger_all_cpu_backtrace(bool include_self)
>> +{
>> + smp_call_function(arch_dump_stack, NULL, 1);

> should this call arch_dump_stack directly too if include_self?
Currently, in case of mips there is no case include_self is true, so this is not a problem. 
arch_trigger_all_cpu_backtrace can only be called from trigger_allbutself_cpu_backtrace() in kernel/watchdog.c.
But as you said, if the case will be added, we should consider that.

Thanks.

> Cheers
> James