[ltt-dev] [PATCH v5 0/5] Debugfs-based ltt trace controller
zhaolei at cn.fujitsu.com
Sun Nov 23 19:57:00 EST 2008
* From: "Mathieu Desnoyers" <compudj at krystal.dyndns.org>
> Hi Zhaolei,
> I've just released your work in LTTng 0.58. Thanks again !
Glad to hear that.
I will continue to make user-mode lttctl support this interface.
> * Zhaolei (zhaolei at cn.fujitsu.com) wrote:
>> Hello, Mathieu
>> This is v5 of debugfs-based ltt trace controller.
>> >> > But IMHO, mode is not necessary now, because we can control each channel's
>> >> > overwrite.
>> >> > So maybe I'd like to remove conception of "mode" in ltt's kernel source.
>> >> > It will delete some struct member, some function, and modify some function,
>> >> > so we can do it with another patch.
>> >> > User-mode tools can have conception of "mode". They can set each channel's
>> >> > overwrite attribute by mode.
>> >> > (OR let user-mode tools get rid of conception of mode, and control each
>> >> > channel's overwrite attribute directly)
>> >> >
>> > Yes, the "mode" concept can go away form the kernel, it's not required
>> > anymore. Having one "overwrite" per channel is a superset of "mode"
>> > (flight, normal, hybrid) which is basically just fixing overwrite modes
>> > to specific channels.
>> > If you plan to leave a "mode" file for transition (I don't think it is
>> > necessary), it should not go into
>> > debugfs/control/trace_name/channel/cpu/mode
>> > but rather in
>> > debugfs/control/trace_name/mode
>> > because it fixes the mode for the whole trace, not only for a particular
>> > channel.
>> Concept of "mode" is removed from ltt-kernel and ltt-trace-control.
>> Other small fix(as ctrate -> create in comment) are also included.
>> We may begin to work on ltt-user-space tools after it got merged.
>> ltt-dev mailing list
>> ltt-dev at lists.casi.polymtl.ca
> Mathieu Desnoyers
> OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
More information about the lttng-dev