Re: Preparing 2.6.16-7, Call for discussion on SECCOMP and HZ_250

[maximilian attems <maks@sternwelten.at>]

On Wed, 12 Apr 2006, Frederik Schueler wrote:

> Hello,
> 
> I would like to schedule the upload of 2.6.16-7, containing 2.6.16.3 and
> 2.6.16.4, for tomorrow Apr. 13th.
> 
> Additionally, I would like to call for comments on the following options:
> 
> - SECCOMP: already deactivated in trunk, should be deactivated in sid
>   too, for more info on the issue please see
>   http://marc.theaimsgroup.com/?l=linux-kernel&m=113679955125413&w=2
>   It seems to not change the ABI, so it could still be changed for -7.

done
   
> - HZ_250 instead of HZ_1000: If I remember correctly, we kept HZ_1000 
>   because it was the prior default. IMHO, we should follow upstream and 
>   change the default to HZ_250 too. 
> 
>   A short pro/cons is here: http://kerneltrap.org/node/5411
> 
>   Considering we plan to use the smp-alternatives patches for both amd64
>   and i386, at least here HZ_250 makes sense. For the other arches and
>   flavours, either HZ_100 for smp and HZ_1000 for up/desktop flavours
>   could make sense, or just HZ_250 too for the sake of simplicity.
> 
>   There will for sure be users requesting to revert this change, similar
>   to PREEMPT, so I would like to have a concensus in the team regarding
>   this option.

i would stick with HZ_250 everywhere.
the upstream change seems not contested since.

also you may get subtile driver breaks on those that don't get HZ right.
much safer calculation is upstream value.
i'm no fanboy of dyn_hz.

-- 
maks