Re: [PATCH 0/4] workqueue_tracepoint: Add worklet tracepoints for worklet lifecycle tracing

[Andrew Morton <akpm@linux-foundation.org>]

On Fri, 24 Apr 2009 22:51:03 -0400 (EDT) Steven Rostedt <rostedt@goodmis.org> wrote:

> In the old -rt patch series, we had trace points scattered all over the 
> kernel. This was the original "event tracer". It was low overhead and can 
> still give a good overview of the system when the function tracer was too 
> much data. Yes, we solved many issues in -rt because of the event tracer. 

Sure, tracers can be useful.  The ext3 tracer I did way back when
maintained a 32-element trace buffer inside each buffer_head and then would
emit that trace when an assertion failed against that buffer_head, so
you can see the last 32 things which happened to that bh.  It would
have been nigh impossible to fix many of the things which were fixed
without that facility.  (I doubt, incidentally, whether ftrace can do
this sort of data-centric tracing?).

But I never merged it into Linux.  Some of the tracepoints are in there
(grep TRACE fs/ext3/*.c) but the core was kept out-of-tree.

> BTW, you work for Google,

Hey, it's the other way round ;)

> doesn't google claim to have some magical 
> 20-some tracepoints that is all they need? Could you give us a hint to 
> what and where they are?

Wouldn't have a clue.  Jiaying Zhang should be able to find out.