Re: [Patch -tip 0/4] Creation of the initcall tracer

[Ingo Molnar]

* Frédéric Weisbecker <fweisbec@xxxxxxxxx> wrote:

> This patchset is about the initcall tracer requested by Ingo Molnar. 
> It is supposed to replace the printk involved during fastboot if the 
> initcall_debug parameter has been selected.

that was fast! :)

i've created a new -tip topic for your new tracer: tip/tracing/fastboot. 
That topic is based on the tip/tracing/ftrace + tip/fastboot trees.

Note that i've done a couple of small changes to your patches:

 - i've done the following renames:

      :1,$s/INITCALL_TRACER/BOOT_TRACER/g
      :1,$s/trace_initcall/trace_boot/g
      :1,$s/initcall_trace/boot_trace/g

   the reasoning: we eventually want to extend this new tracer to all 
   things "bootup delay", not just initcalls - hence calling it an
   'initcall tracer' would be too limiting.

   The goal is to help the Arjan's fastboot project to achieve 5 seconds 
   to-GUI bootup times like this:

      http://www.youtube.com/watch?v=s7NxCM8ryF8

   [ Even if you know what a 5 seconds bootup is, you've got to check 
     this "From Naught to Sixty in 5 seconds" video ;-) ]

 - there were a few overlong lines in the commit logs and in the Kconfig 
   desciption - i fixed them up. You should check your email client and 
   your editor for line length settings - currently you seem to have no
   line breaks at all.

 - I also corrected a few capitalization things in commit subject lines 
   and corrected a few minor coding style errors - you can avoid these 
   by running scripts/checkpatch.pl over your patches. Nothing overly
   serious - but you should diff against tip/master to see what the 
   changes are.

> Currently this tracer is a test. Its output is not yet adapted for 
> scripts/bootgraph.pl which uses the old initcalls printk to produce a 
> graph. But since this tracer will output sched_switch and stack_trace, 
> we should first talk about the future output format of this tracer in 
> sync with bootgraph.pl
> 
> It currently not uses the sched_switch and stack_trace tracers since 
> related functions of insertion into the ring buffer perhaps need first 
> to be moved on their own tracer file. But it have to be discussed.
> 
> Note that you need the initcall_debug to enable this tracer. When 
> printk from initcall_debug will disappear, we could surely enable this 
> tracer by default without this parameter.
> 
> Old printk have been kept for backward compatibility.

yes, it's necessary to keep that backwards compatibility.

i've integrated tip/tracing/fastboot into tip/master and have started 
testing it. It passed the basic tests already so i've just pushed out 
the new tip/master. Please double-check that i have not messed up the 
rename or the integration somewhere.

all in one, nice stuff!

	Ingo
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/