Skip to contentSkip to conversation

PowerPC: dynamic tick support

PATCH In review Sergei Shtylyov 1 participant 2 messages Nov 8, 2006 – Mar 16, 2007 PowerPC

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 PowerPC: dynamic tick support

Add dynamic tick support for PowerPC.

---
This patch is reworked against 2.6.18-hrt-dyntick2 patchset.

Known issues:
1) with 2.6.18-hrt-dyntick2, dynamic tick mode causes bootup slowdown after the

    HRT mode is entered (it seems to return to normal shortly);

2) with 2.6.18-rt7 (plus an extra patch to make it actually working :-) in the kernel preemption mode other than RT, that slowdown seems permanent (this was not happening with 2.6.18-rt6).

I don't have time to investigate now, so anybody interested is welcome to... or I may even return to this if/when the time permits. :-)

 arch/powerpc/kernel/idle.c |    4 ++++
 1 files changed, 4 insertions(+)
arch/powerpc/kernel/idle.c +4 -0
Index: linux-2.6/arch/powerpc/kernel/idle.c
===================================================================
@@ -49,6 +49,8 @@ void cpu_idle(void)
 
 	set_thread_flag(TIF_POLLING_NRFLAG);
 	while (1) {
+		hrtimer_stop_sched_tick();
+
 		while (!need_resched() && !cpu_should_die()) {
 			ppc64_runlatch_off();
 
@@ -82,6 +84,8 @@ void cpu_idle(void)
 		ppc64_runlatch_on();
 		if (cpu_should_die())
 			cpu_die();
+
+		hrtimer_restart_sched_tick();
 		preempt_enable_no_resched();
 		schedule();
 		preempt_disable();
#1 lore raw
Quoted 1 line · Hello, I wrote:

Add dynamic tick support for PowerPC.

[...]

Quoted 8 lines

Known issues:
1) with 2.6.18-hrt-dyntick2, dynamic tick mode causes bootup slowdown after the

    HRT mode is entered (it seems to return to normal shortly);

2) with 2.6.18-rt7 (plus an extra patch to make it actually working :-) in the kernel preemption mode other than RT, that slowdown seems permanent (this was not happening with 2.6.18-rt6).

I don't have time to investigate now, so anybody interested is welcome to...

... but I should have done it. :-/

Quoted 1 line

or I may even return to this if/when the time permits. :-)

Looks like that was caused by that missing !need_resched_delayed() below.
So, should be fixed with the latest 2.6.21-rc3-rt0 patch...

Quoted 16 lines
 arch/powerpc/kernel/idle.c |    4 ++++
 1 files changed, 4 insertions(+)

Index: linux-2.6/arch/powerpc/kernel/idle.c
===================================================================

arch/powerpc/kernel/idle.c +2 -0
@@ -49,6 +49,8 @@ void cpu_idle(void)
 
 	set_thread_flag(TIF_POLLING_NRFLAG);
 	while (1) {
+		hrtimer_stop_sched_tick();
+
 		while (!need_resched() && !cpu_should_die()) {
 			ppc64_runlatch_off();