Re: [PATCH 0/3] tracepoints: delay argument evaluation

[Li Zefan <lizf@cn.fujitsu.com>]

Jiaying Zhang wrote:
> Is it possible to convert blktrace to use event tracer? Then in this case we

Yes, I'm doing this, see:
	http://marc.info/?l=linux-kernel&m=124228198011297&w=2

> can pass 'md' as the parameter to trace_block_bio_complete and dereference
> md->queue during assignment.
> 

But the problem discussed here exists whether you use plain tracepoints
or TRACE_EVENT.

Though we can add a new tracepoint named trace_md_bio_complete, this is
not the way to solve it.

> Jiaying
> 
> On Wed, May 20, 2009 at 8:42 AM, Jason Baron <jbaron@redhat.com> wrote:
>> On Wed, May 20, 2009 at 09:33:48AM +0200, Ingo Molnar wrote:
>>> hm, this is really a compiler bug in essence - the compiler should
>>> delay the construction of arguments into unlikely branches - if the
>>> arguments are only used there.
>>>
>>> We'd basically open-code a clear-cut:
>>>
>>>    trace_block_bio_complete(md->queue, bio);
>>>
>>> into this form:
>>>
>>>    trace(block_bio_complete, md->queue, bio);
>>>
>>> .. and this latter form could become moot (and a nuisance) if the
>>> compiler is fixed.
>>>
>>> Have you tried very latest GCC, does it still have this optimization
>>> problem?
>>>
>>> Note that the compiler getting this right would help a _lot_ of
>>> other inline functions in the kernel as well. Arguments only used
>>> within unlikely() branches are quite common.
>>>
>>>       Ingo
>> hi,
>>
>> I e-mailed the gcc list, where they suggested using a macro, as I've
>> done. They also suggested filing an enhancement request for this, which
>> I've done: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=40207 It seems
>> like they agree with the suggestion.
>>
>> It still might make sense to make this requirement explicit (by adding
>> the extra macro), as the tracepoint off case should really be as optimized as
>> possible.
>>
>> thanks,
>>
>> -Jason
>>