Peter Zijlstra wrote:
> On Fri, 2009-02-06 at 11:34 +0100, Matthias Luescher wrote:
>
>> Hi
>>
>> I finally found out which patch does not work well with my configuration:
>>
>> 2.6.24.7-rt12 seems to be ok, with 2.6.24.7-rt13 the command cyclictest -p99 -t10 -n -i250 crashes the kernel.
>>
>> 2.6.24.7-rt13 has the follwing new patches:
>> ftrace-hotplug-fix.patch
>> ftrace-wakeup-rawspinlock.patch
>> radix-tree-lockdep-plus1.patch
>> sched-cpupri-hotplug-support.patch
>> sched-cpupri-priocount.patch
>>
>> No I went again to 2.6.24.7-rt26 but removed some patches:
>> 2.6.24.7-rt26 with all patches -> CRASH
>> 2.6.24.7-rt26 without sched-cpupri-hotplug-support.patch and without sched-cpupri-priocount.patch ->OK
>> 2.6.24.7-rt26 without sched-cpupri-hotplug-support.patch ->CRASH
>>
>
> sched-cpupri-priocount.patch is buggy
>
> +#define CPUPRI_NR_PRI_WORDS (CPUPRI_NR_PRIORITIES + BITS_PER_LONG/2)/BITS_PER_LONG
>
Thomas fixed this somewhere along the way with the inlined patch.
Steven, this should be pulled into all actively maintained -rt branches
that do not already have it.
commit e539d8fcd11af811db70707d47ea436d5621d0da
Author: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
Date: Thu Jun 5 10:28:00 2008 +0200
sched: fix the cpuprio count really
Peter pointed out that the last version of the "fix" was still one off
under certain circumstances. Use BITS_TO_LONG instead to get an
accurate result.
Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
diff --git a/kernel/sched_cpupri.h b/kernel/sched_cpupri.h
index 6b38355..f25811b 100644
--- a/kernel/sched_cpupri.h
+++ b/kernel/sched_cpupri.h
@@ -3,8 +3,8 @@
#include <linux/sched.h>
-#define CPUPRI_NR_PRIORITIES 2+MAX_RT_PRIO
-#define CPUPRI_NR_PRI_WORDS (CPUPRI_NR_PRIORITIES +
BITS_PER_LONG/2)/BITS_PER_LONG
+#define CPUPRI_NR_PRIORITIES (MAX_RT_PRIO + 2)
+#define CPUPRI_NR_PRI_WORDS BITS_TO_LONGS(CPUPRI_NR_PRIORITIES)
#define CPUPRI_INVALID -1
#define CPUPRI_IDLE 0
Attachment:
signature.asc
Description: OpenPGP digital signature
