This is the mail archive of the systemtap@sourceware.org mailing list for the systemtap project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [Ltt-dev] LTTng 0.10.0-pre23 (fix)


* Mike Mason (mmlnx@us.ibm.com) wrote:
> Mathieu Desnoyers wrote:
>> * Mathieu Desnoyers (compudj@krystal.dyndns.org) wrote:
>>> * Mathieu Desnoyers (compudj@krystal.dyndns.org) wrote:
>>>> Hi,
>>>>
>>>> I just released lttng 0.10.0-pre22 for Linux 2.6.24-rc2, which adds
>>>> support for multiple callbacks connected to a single marker. I make
>>>> extensive use the kernel RCU mechanism so batch connexion of probes to
>>>> different markers can be done as quickly as possible, requiring to wait
>>>> for quiescent state only when necessary. It proves to be important when
>>>> connecting probes on busy systems, where waiting for ~50
>>>> synchronize_sched() could take about 1-2 minutes.
>>>>
>>> Actually this release is pre21, not pre22.
>>>
>> LTTng 0.10.0-pre23 fixes a bug in an inaccurate assumption that a
>> function pointer is always aligned on (at least) 2 bytes boundaries.
>> It's not on some architectures.
>
> I tested your changes on x86_64 using multiple instances of the 
> probe-example code.  Everything seems to work fine.  The code itself looks 
> fine, too, though I'm no RCU expert.  The SystemTap support for markers 
> will have to change again since the register and unregister interfaces have 
> changed.  I also noticed the arm and disarm functions are gone.  I never 
> quite understood why those were necessary anyway.
>
> When do you anticipate pushing multiple marker support upstream?  Also, are 
> there any upcoming changes that may change the interfaces again?  It'll be 
> nice when the interface is locked in.

I just sent an RFC on LKML about this patch this afternoon. If
everything goes smoothly, we can expect to post it in a 1-2 weeks
window, I guess. It really depends on the comments we will receive
(which could range from : "It's godawful" to "Wow, we like that!")

Since we can argue that this API change should be done before 2.6.24
final, this could give an incentive to Andrew to merge it soon enough.
Until we are close to 2.6.24 final, the API could change. Even after..
this is kernel code and there is no guaranteed stable API in the kernel.
But The multiple probes support is the only API change I foresee.

Thanks for the review and testing, I'm glad you like it.

Mathieu

>
> Mike
>
>> Mathieu
>>>> Mathieu
>>>>
>>>> -- 
>>>> Mathieu Desnoyers
>>>> Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal
>>>> OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 
>>>> 9A68
>>>> _______________________________________________
>>>> Ltt-dev mailing list
>>>> Ltt-dev@listserv.shafik.org
>>>> http://listserv.shafik.org/mailman/listinfo/ltt-dev
>>>>
>>> -- 
>>> Mathieu Desnoyers
>>> Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal
>>> OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 
>>> 9A68
>>> _______________________________________________
>>> Ltt-dev mailing list
>>> Ltt-dev@listserv.shafik.org
>>> http://listserv.shafik.org/mailman/listinfo/ltt-dev
>>>
>

-- 
Mathieu Desnoyers
Computer Engineering Ph.D. Student, Ecole Polytechnique de Montreal
OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F  BA06 3F25 A8FE 3BAE 9A68


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]