[RFC] HZ entry in /proc/sys/kernel

[RFC] HZ entry in /proc/sys/kernel

[Grzegorz Jaskiewicz]

I have seen few scripts allready that are assuming HZ==100.
Afaik this value is different in 2.5/2.4 for the same arch.
So is it worth to add on both simple /proc/sys/kernel/HZ ?
   
Simple patch for 2.4.21-rc3 , if it is allright i will do same thing for
2.5

----------------------------------------------------------

diff -ur 2/include/linux/sysctl.h 1/include/linux/sysctl.h
--- 2/include/linux/sysctl.h    2003-05-24 12:10:35.000000000 +0100
+++ 1/include/linux/sysctl.h    2003-05-26 12:44:16.000000000 +0100
@@ -125,6 +125,7 @@
        KERN_TAINTED=53,        /* int: various kernel tainted flags */
        KERN_CADPID=54,         /* int: PID of the process to notify on
CAD */
        KERN_CORE_PATTERN=56,   /* string: pattern for core-files */
+       KERN_HZ=57,             /* int: HZ value */
 };


diff -ur 2/kernel/sysctl.c 1/kernel/sysctl.c
--- 2/kernel/sysctl.c   2003-05-26 12:33:37.000000000 +0100
+++ 1/kernel/sysctl.c   2003-05-26 12:45:31.000000000 +0100
@@ -32,6 +32,7 @@
 #include <linux/highuid.h>

 #include <asm/uaccess.h>
+#include <asm/param.h>

 #ifdef CONFIG_ROOT_NFS
 #include <linux/nfs_fs.h>
@@ -123,6 +124,8 @@

 #ifdef CONFIG_PROC_FS

+static unsigned int hz=HZ;
+
 static ssize_t proc_readsys(struct file *, char *, size_t, loff_t *);
 static ssize_t proc_writesys(struct file *, const char *, size_t,
loff_t *);
 static int proc_sys_permission(struct inode *, int);
@@ -244,6 +247,8 @@
         0600, NULL, &proc_dointvec},
        {KERN_MAX_THREADS, "threads-max", &max_threads, sizeof(int),
         0644, NULL, &proc_dointvec},
+       {KERN_HZ, "HZ", &hz, sizeof(unsigned int),
+        0444, NULL, &proc_dointvec},
        {KERN_RANDOM, "random", NULL, 0, 0555, random_table},
        {KERN_OVERFLOWUID, "overflowuid", &overflowuid, sizeof(int),
0644, NULL,
         &proc_dointvec_minmax, &sysctl_intvec, NULL,

  
-- 
Grzegorz Jaskiewicz <gj@pointblue.com.pl>
K4 labs

Re: [RFC] HZ entry in /proc/sys/kernel

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 324 bytes --]

On Mon, 2003-05-26 at 13:53, Grzegorz Jaskiewicz wrote:
> I have seen few scripts allready that are assuming HZ==100.
> Afaik this value is different in 2.5/2.4 for the same arch.

No it's not actually.
The userspace interface is constant/stable and in units of HZ=100 even
though the kernel HZ might be different.


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [RFC] HZ entry in /proc/sys/kernel

[Daniele]

On Mon, May 26, 2003 at 02:15:57PM +0200, Arjan van de Ven wrote:
> On Mon, 2003-05-26 at 13:53, Grzegorz Jaskiewicz wrote:
> > I have seen few scripts allready that are assuming HZ==100.
> > Afaik this value is different in 2.5/2.4 for the same arch.
> 
> No it's not actually.
> The userspace interface is constant/stable and in units of HZ=100 even
> though the kernel HZ might be different.
> 

What's this HZ value related to?
cheers,
	Daniele

Re: [RFC] HZ entry in /proc/sys/kernel

[Arjan van de Ven]

On Tue, May 27, 2003 at 09:39:57PM +0200, Daniele wrote:
> On Mon, May 26, 2003 at 02:15:57PM +0200, Arjan van de Ven wrote:
> > On Mon, 2003-05-26 at 13:53, Grzegorz Jaskiewicz wrote:
> > > I have seen few scripts allready that are assuming HZ==100.
> > > Afaik this value is different in 2.5/2.4 for the same arch.
> > 
> > No it's not actually.
> > The userspace interface is constant/stable and in units of HZ=100 even
> > though the kernel HZ might be different.
> > 
> 
> What's this HZ value related to?

the userspace interface is not in units of kernel HZ, but in
"theoretical time units that happen to match a defined
default" which happens to match 10ms on x86.