Re: 2.4.22pre7aa1 ;-) Mit O(1) aber leider kein pre-emption

[Manuel Krause <manuelkrause@netscape.net>]

[-- Attachment #1: Type: text/plain, Size: 1600 bytes --]

On 07/21/03 17:55, Manuel Krause wrote:
> On 07/21/03 17:38, Manuel Krause wrote:
> 
>> On 07/21/03 16:39, Dieter Nützel wrote:
>>
>>> Data-logging in the works?
>>>
>>> Regards,
>>>   Dieter
>>
>> Mmmh. ATM of .22-pre7 only "old" rml-preempt for 2.4.21-final does not 
>> apply well on vanilla .22-pre7.  But that only in 
>> linux/arch/sh/kernel/entry.S and linux/arch/ppc/mm/tlb.c compared to 
>> the adjusted version for -pre5. Do you really need those files for now?!
>>
>> Recent data-logging without quota does apply without errors on here 
>> tough.
> 
> Heaven, sorry! I definitely don't have the O(1) in here... and no aaX...
> So, this info was only for 2.4.22-pre7 plus preempt plus data-logging 
> plus search_reada... Sorry for the inconvenience...
> 
>> But I haven't got the compile up and running yet.
>>
>> BTW, the HZ=1000 setting makes my system time get some minutes 
>> backwards  when having a longer uptime. It's solved on every bootup. I 
>> this typical or have I missed something?!
> 
> Bye,
> 
>  Manuel

Yess, I missed something as I don't read LKML usually. They have too 
much more SPAM compared to ReiserFSlist nowadays... (No kidding, but I 
don't need LKML in its whole fashion. But great thanks to the 
Hero-Anti-Spam-Worker of this list, Alexander Lyamin !!!, BTW, as spam 
rate drops to ZERO.)

The attached patch was posted to LKML and is supposed to solve what I 
asked for in the last paragraph of my previous mail that stayed 
unanswered so far. I finally found it myself last hour.

Thanks and bye,

  Manuel



[-- Attachment #2: patch.mkYY.timer_compensation_1000HZ.diff --]
[-- Type: text/plain, Size: 631 bytes --]

diff -Naurd old/kernel/timer.c linux/kernel/timer.c
--- old/kernel/timer.c  2003-07-18 17:27:01.000000000 -0400
+++ linux/kernel/timer.c    2003-07-18 17:32:19.000000000 -0400
@@ -606,6 +606,15 @@
     else
    time_adj += (time_adj >> 2) + (time_adj >> 5);
 #endif
+#if HZ == 1000
+    /* Compensate for (HZ==1000) != (1 << SHIFT_HZ).
+     * Add 1.5625% and 0.78125% to get 1023.4375; => only 0.05% error (p. 14)
+     */
+    if (time_adj < 0)
+   time_adj -= (-time_adj >> 6) + (-time_adj >> 7);
+    else
+   time_adj += (time_adj >> 6) + (time_adj >> 7);
+#endif
 }
 
 /* in the NTP reference this is called "hardclock()" */