[parisc-linux] Re: [parisc-linux-cvs] linux-2.5 varenet

[Thibaut VARENE]

Le jeudi, 24 oct 2002, =E0 09:18 Europe/Paris, Richard Hirst a =E9crit :

> On Wed, Oct 23, 2002 at 10:46:11AM -0600, Thibaut Varene wrote:
>> CVSROOT:	/var/cvs
>> Module name:	linux-2.5
>> Changes by:	varenet	02/10/23 10:46:11
>>
>> Modified files:
>> 	include/asm-parisc: param.h
>>
>> Log message:
>> Switched to HZ =3D=3D 1000 for CONFIG_PA20.
>> Patch submitted by Laurent Canet.
>
> Don't know if this was simply a #define change, but if so...
> Robert Love has been arguing lately that you can't just do that as
> various userspace stuff is compiled with a knowledge of the kernel HZ
> value for your specific arch and uses that in interpreting various
> entries in /proc.  He had a patch that kept the userspace view
> consistent.  That was 2.4 though; maybe 2.5 is better in this respect.

Hm, I'm not sure we really broke something, we added the lines marked=20
'->':

#ifdef __KERNEL__
-># ifdef CONFIG_PA20
->#  define HZ            1000            /* Faster machines */
-># else
#  define HZ            100             /* Internal kernel timer=20
frequency */
-># endif
# define USER_HZ        100             /* .. some user interfaces are=20=

in "ticks
" */
# define CLOCKS_PER_SEC (USER_HZ)       /* like times() */
#endif

#ifndef HZ
#define HZ 100
#endif

As you can see, these defines are protected by a #ifdef __KERNEL__,=20
otherwise HZ is 100 in all cases. Since userspace stuff isn't supposed=20=

to define __KERNEL__ at compile time, HZ from a userspace point of view=20=

hasn't changed.

If you look at other archs (like i386, ppc64) they also have different=20=

HZ vs USER_HZ and *__KERNEL__* HZ.

In 2.4, there wasn't such #ifdef protection, so such change could have=20=

broken things.
In 2.5, I think this is quite safe IMHO.

HTH,


Thibaut VARENE
The PA/Linux ESIEE Team
http://pateam.esiee.fr/=