[ltt-dev] some question about ltt

loody miloody at gmail.com
Sat Feb 5 08:59:44 EST 2011


hi Mathieu:
thanks for ur kind reply :)

2011/2/1 Mathieu Desnoyers <compudj at krystal.dyndns.org>:
> Hi,
>
> * loody (miloody at gmail.com) wrote:
>> hi all:
>> I have some questions about ltt:
>> 1. what is the out format about ltt?
>> it seems kernel.softirq_entry mean it calls softirq_entry and belongs
>> to kernel group
>
> yes
>
>> but what does metadata.core_marker_format mean?
>
> group metadata (lttng trace metadta), which outputs the id/format of
> events. It describes the rest of the trace.
>
>> Calling core_marker_format? and what does metadata group mean?
>>
>> 2. ltt is used for monitoring kernel behavior, why there is a group
>> named "userspace"?
>
> It contains the events written through the debugfs file ltt/write_event.
> See "Userspace tracing" in the LTTng Manual on the lttng.org website.
> This is a "slow" tracing mode that takes a simple string as input. UST
> is a much faster tracer that hooks directly onto tracepoints declared in
> userspace code.
>
>> 3. can I monitor the system calls, "set_thread_area" and "mprotect"?
>
> Yes. But you will see them as "syscall_entry/syscall_exit" events, with
> the syscall number mapping to the syscall name (first event field). We
> don't gather the syscall arguments at this point: you'll need to add
> instrumentation for that.
>
> Thanks,
>
> Mathieu
>
>> I found these 2 functions in strace but cannot found it in ltt.
>> Should I open additional config ?
there is a document under doc/developer/format.html, and it seems a
doc which used to describe structure used in source code.
Is there any doc which describe the format of output log as below:
kernel.page_fault_entry: 1.962735214
(/media/D230-4695/trace2.11.01.31.2/kernel_0), 194, 0, , , 0, 0x0,
TRAP { ip = 0x2aae0ef0, address = 0x2aae0ef0, trap_id = 2,
write_access = 0 }
appreciate your help,
miloody


-- 
Regards,




More information about the lttng-dev mailing list