Re: [PATCH -tip] Tracing/fastboot: Only trace non-module initcalls

[Alan Jenkins <alan-jenkins@tuffmail.co.uk>]

Frédéric Weisbecker wrote:
> 2008/10/3 Alan Jenkins <aj504@student.cs.york.ac.uk>:
>   
>> Sorry to butt in.  I think tracing module initcalls was a deliberate addition by Arjan (a06e1c2b448b317bc141934879cbbbed8319b4d4).
>>
>> Maybe you use an initrd and that affects the trace?  It would be nice to see the "before" svg as well as the "after", to explain the change.
>>
>> Thanks
>> Alan
>>     
>
>
> Hi Alan,
>
> The script bootgraph.pl in -tip stops the analyzing when it sees
> "Freeing unused kernel memory". That's
> exactly when the built-in initcalls are finished.
> And as a result, the graph produced with dmesg always finishes at this point.
>
> If you try to generate a graph whithout this limitation, you will not
> be able to see something interesting because
> the script will not find a lot of initcalls with a long enough time
> proportion against the total boot time.
> Perhaps it can be improved in that way but unfortunately I don't know
> any word in Perl....
>   

If you want to disable non-module initcalls, then what do you have left
that you want to trace after "Freeing unused kernel memory"?

On my desktop dmesg I can see that the initramfs is loaded and started
some time before the initcalls finish.  The initramfs includes udev
which loads many modules; I see network, usb, sata being initialised
before the "Freeing unused kernel memory" message.  Um...  your changes
will _not_ stop the initcalls from these modules being shown.

So I think your patch description is a bit inaccurate, and I don't
understand what you are trying to achieve.

Thanks
Alan