Re: [v2 0/9] Early boot time stamps for x86

[Pasha Tatashin]

Hi Thomas,

The second versions was actually meant as a reply to your e-mail: the 
code differences were minimal: the main differences were in the cover 
letter. You mentioned that it is not necessary to have early boot time 
stamps, and  I wanted to show examples how this data is useful to track 
scalability bugs and avoid future regressions.

Anyway, you asked for some time to think about this problem, I won't 
send any replies to this thread for the next two weeks. So, please 
consider this solution. The feature is well abstracted, does not harm 
the performance of the fast path, and if necessary it can also be made 
optional with something like:
CONFIG_HAVE_UNSTABLE_EARLY_CLOCK

Thank you,
Pasha

On 03/25/2017 06:25 AM, Thomas Gleixner wrote:
> On Fri, 24 Mar 2017, Pavel Tatashin wrote:
>
> > changelog
> > ---------
> > v1 - v2
> > 	In patch "x86/tsc: tsc early":
> > 	- added tsc_adjusted_early()
> > 	- fixed 32-bit compile error use do_div()
>
> Did you actually read my last reply on V1 of this?
>
> I made it entirely clear that the way this is done, i.e. hacking it into
> the earliest boost stage is not going to happen.
>
> Further I asked you to hold off until I found some time to look into this
> in detail.
>
> So what's the point of ignoring what I said and resending the whole lot
> with some more hackery applied?
>
> I don't care about you wasting your time, but I very much care about my
> time.
>
> Thanks,
>
> 	tglx