Re: [PATCH] PPC64 enable hp-timing

[Benjamin Herrenschmidt <benh at kernel dot crashing dot org>]

> Wouldn't it make a bit more sense to put timebase into sometime else in 
> /proc?  Seems silly to have that value inside of /proc/cpuinfo in the 
> first place.

This is legacy. I will put it in the vDSO (hopefully before 2.6.15).

> Course on ppc64 there is
> 
> /proc/device-tree/cpus/PowerPC,POWER5@0/timebase-frequency which can get 
> the same value without having to deal with cpuinfo and it's variable 
> size.   Course the problem with looking in device-tree is that of course 
> is a config option to the kernel and really valid only on OF hardware. 

That, and the format of the path can be slightly different, and some
machines may have bogus info there, etc etc... I wouldn't want glibc to
get any knowledge about the device-tree, that is really out of bounds.

> Still that leads me back to my original observation,  shouldn't/couldn't 
> timebase be put into it's own file in /proc?

Adding something to /proc, it wouldn't be upstream before 2.6.15 anyway,
so let's do it the right way and add a __kernel_get_tbfreq() or whatever
you want to call it to the vDSO. I'll do that along with the other
additions you asked me this week, as soon as I'm done with the 1T
segment work.

Ben.