Re: [LARTC] filter ingress policy rates -> slow!!

[Ed Wildgoose <lists@wildgooses.com>]

>
>
>I think it will change because: 
>I'm not going to use the same distro, but change from Fedora to Suse.
>What's more, I've managed to compile this same kernel source time ago, so 
>something wrong must be happening :-). 
>  
>

One workaround is to look at which module is erroring (in your case an 
IDE bit of code).  And then see if you live without that code (disable 
it). 

Also if you are using a redhat .config preprovided file then it tends to 
build everything as a module - in my experience this becomes very 
fragile.  I prefer to have few or no modules (and this tends to fix a 
lot of compile probs as well)

>As I said, I'll get Suse 9.1 Prof soon (the 10th most probably). I'm curious 
>about the PSCHED_CLOCK_SOURCE value. Is it wrong or deprecated the default 
>Fedora's value (PSCHED_JIFFIES instead of PSCHED_CPU)? Or is it just that you 
>have to compile your own kernel in order to get ingress rate policies working 
>fine ?
>  
>

I'm a BIG fan of gentoo.  You never need to upgrade again...!  (Debian 
sounds nice as well)

I think the psched_clock_source code just changes how the scheduler 
works out how often to fire.  Basically its hard to get accurate high 
resolution timers on most operating systems.  I haven't studied the 
details, but assume the CPU method gives more accurate results, perhaps 
at the expense of more CPU required..?

Ed W
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/