Re: [PATCH] tracing: fix a build error on alpha and m68k

[Frederic Weisbecker <fweisbec@gmail.com>]

Ingo Molnar wrote:
> * Alexey Dobriyan <adobriyan@gmail.com> wrote:
> 
>> On Thu, Oct 23, 2008 at 07:27:48PM +0200, Frederic Weisbecker wrote:
>>> When tracing is enabled, some arch have included <linux/irqflags.h>
>>> on their <asm/system.h> but others like alpha or m68k don't.
>>>
>>> Build error on alpha:
>>>
>>> kernel/trace/trace.c: In function 'tracing_generic_entry_update':
>>> kernel/trace/trace.c:658: error: implicit declaration of function 'irqs_disabled_flags'
>>> kernel/trace/trace.c: In function 'tracing_cpumask_write':
>>> kernel/trace/trace.c:2145: error: implicit declaration of function 'raw_local_irq_disable'
>>> kernel/trace/trace.c:2162: error: implicit declaration of function 'raw_local_irq_enable'
>>> kernel/trace/trace.c: In function 'trace_die_handler':
>>> kernel/trace/trace.c:3039: error: 'DIE_OOPS' undeclared (first use in this function)
>>>
>>> Build error on m68k:
>>> http://kisskb.ellerman.id.au/kisskb/buildresult/50641/
>>>
>>>
>>> Include it on kernel/trace/trace.c
>>>
>>> Reported-by: Alexey Dobriyan <adobriyan@gmail.com>
>>> Signed-off-by: Frederic Weisbecker <fweisbec@gmail.com>
>>> ---
>>> diff --git a/kernel/trace/trace.c b/kernel/trace/trace.c
>>> index 78d5661..deb9684 100644
>>> --- a/kernel/trace/trace.c
>>> +++ b/kernel/trace/trace.c
>>> @@ -34,6 +34,7 @@
>>>  
>>>  #include <linux/stacktrace.h>
>>>  #include <linux/ring_buffer.h>
>>> +#include <linux/irqflags.h>
>> Sure, except it doesn't fix anything.
> 
> hm, zapped the commit then.
> 
> the problem is most likely that none of these architectures is lockdep 
> enabled, hence they have no irqtrace wrappers, hence not all of the 
> tracers can be built on them?
> 
> 	Ingo
> 

I just testes my patch by building with a cross compiler on Alpha.
Before the patch:

kernel/trace/trace.c: In function 'tracing_generic_entry_update':
kernel/trace/trace.c:664: error: implicit declaration of function 'irqs_disabled_flags'
kernel/trace/trace.c: In function 'tracing_cpumask_write':
kernel/trace/trace.c:2150: error: implicit declaration of function 'raw_local_irq_disable'
kernel/trace/trace.c:2167: error: implicit declaration of function 'raw_local_irq_enable'
kernel/trace/trace.c: In function 'trace_die_handler':
kernel/trace/trace.c:3044: error: 'DIE_OOPS' undeclared (first use in this function)
kernel/trace/trace.c:3044: error: (Each undeclared identifier is reported only once
kernel/trace/trace.c:3044: error: for each function it appears in.)

After the patch:

kernel/trace/trace.c: In function 'tracing_generic_entry_update':
kernel/trace/trace.c:664: error: implicit declaration of function 'irqs_disabled_flags'
kernel/trace/trace.c: In function 'trace_die_handler':
kernel/trace/trace.c:3044: error: 'DIE_OOPS' undeclared (first use in this function)
kernel/trace/trace.c:3044: error: (Each undeclared identifier is reported only once
kernel/trace/trace.c:3044: error: for each function it appears in.)

So the raw_local_irq_* functions are defined in linux/irqflags.h, even for arch that doesn't have CONFIG_TRACE_IRQFLAGS_SUPPORT. They just do a local_irq_save/restore.
That's what we wanted. And we have to include it on trace.c for archs like alpha.

So this patch fixes it.

The DIE_OOPS should be fixed by a recent patch from Steven (ftrace: move nmi die handler to arch specific).

The last issue is irqs_disabled_flags. But it's one other problem which is in discussion.