Re: [PATCH 0/13] Time: Generic Timeofday Subsystem (v B10)

[Andi Kleen <ak@suse.de>]

On Sunday 13 November 2005 08:32, Ingo Molnar wrote:

> there are 3 "generic" components needed right now to clean up all time 
> related stuff: GTOD, ktimers and clockevents. [you know the first two, 
> and clockevents is new code from Thomas Gleixner that generalizes timer 
> interrupts and introduces one compact notion for 'clock chips'.]

Both noidletick and the per cpu gettimeofday change significantly
how timer interrupts work. I hope your generalizations will be still
compatible to that. It's a bit dangerous to generalize
before things have their final shape. 

Also vsyscalls make it all more difficult, because they don't map
very well to any kind of "timer drivers". 
 
> what is the point? Ontop of these, a previously difficult feature, High 
> Resolution Timers became _massively_ simpler. All of these patches exist 
> together in the -rt tree, so it's not handwaving. The same will be the 
> case for idle ticks / dynamic ticks [we started with HRT because it is 
> so much harder than idle ticks]. So i do agree with you that GTOD needs 
> more work, but it also makes time related features all that much easier.
> 
> right now it's GTOD that needs the most work before it can be merged 
> upstream, so you picked the right one to criticise :-)

My point was basically that there is a lot of feature work going on
on x86-64 in this area, and that has priority over  any "cleanups" like this 
from my side. If it has settled again later maybe it can be generalized,
or maybe not. I will only do it if it truly makes the code cleaner in the end,
just lots of indirect pointers by itself isn't necessarily something
that does this.

-Andi