rpms/kernel/F-7 kernel-2.6.spec, 1.3251, 1.3252 linux-2.6-ondemand-timer.patch, 1.5, 1.6

[Dave Jones (davej)]

Author: davej

Update of /cvs/pkgs/rpms/kernel/F-7
In directory cvs-int.fedora.redhat.com:/tmp/cvs-serv2970

Modified Files:
	kernel-2.6.spec linux-2.6-ondemand-timer.patch 
Log Message:
* Thu Jul 05 2007 Dave Jones <davej at redhat.com>
- Further tickless improvements to ondemand cpufreq governor.



Index: kernel-2.6.spec
===================================================================
RCS file: /cvs/pkgs/rpms/kernel/F-7/kernel-2.6.spec,v
retrieving revision 1.3251
retrieving revision 1.3252
diff -u -r1.3251 -r1.3252
--- kernel-2.6.spec	5 Jul 2007 20:36:55 -0000	1.3251
+++ kernel-2.6.spec	5 Jul 2007 20:43:43 -0000	1.3252
@@ -2465,6 +2465,9 @@
 %endif
 
 %changelog
+* Thu Jul 05 2007 Dave Jones <davej at redhat.com>
+- Further tickless improvements to ondemand cpufreq governor.
+
 * Thu Jul 05 2007 Chuck Ebbert <cebbert at redhat.com>
 - ACPI: fix possible hang on suspend
 - pata_it821x: partially fix DMA in RAID mode

linux-2.6-ondemand-timer.patch:

Index: linux-2.6-ondemand-timer.patch
===================================================================
RCS file: /cvs/pkgs/rpms/kernel/F-7/linux-2.6-ondemand-timer.patch,v
retrieving revision 1.5
retrieving revision 1.6
diff -u -r1.5 -r1.6
--- linux-2.6-ondemand-timer.patch	29 Mar 2007 00:00:56 -0000	1.5
+++ linux-2.6-ondemand-timer.patch	5 Jul 2007 20:43:43 -0000	1.6
@@ -328,3 +328,172 @@
  	do {							\
  		INIT_WORK_NAR(&(_work)->work, (_func));		\
 
+
+Userspace governor registers a frequency change notifier at init time, even
+when no CPU is set to userspace governor. Make it register only when
+atleast one CPU is using userspace.
+
+Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi at intel.com>
+
+Index: linux-2.6.22-rc-mm/drivers/cpufreq/cpufreq_userspace.c
+===================================================================
+--- linux-2.6.22-rc-mm.orig/drivers/cpufreq/cpufreq_userspace.c
++++ linux-2.6.22-rc-mm/drivers/cpufreq/cpufreq_userspace.c
+@@ -37,6 +37,7 @@ static unsigned int	cpu_set_freq[NR_CPUS
+ static unsigned int	cpu_is_managed[NR_CPUS];
+ 
+ static DEFINE_MUTEX	(userspace_mutex);
++static int cpus_using_userspace_governor;
+ 
+ #define dprintk(msg...) cpufreq_debug_printk(CPUFREQ_DEBUG_GOVERNOR, "userspace", msg)
+ 
+@@ -47,7 +48,11 @@ userspace_cpufreq_notifier(struct notifi
+ {
+         struct cpufreq_freqs *freq = data;
+ 
+-	dprintk("saving cpu_cur_freq of cpu %u to be %u kHz\n", freq->cpu, freq->new);
++	if (!cpu_is_managed[freq->cpu])
++		return 0;
++
++	dprintk("saving cpu_cur_freq of cpu %u to be %u kHz\n",
++			freq->cpu, freq->new);
+ 	cpu_cur_freq[freq->cpu] = freq->new;
+ 
+         return 0;
+@@ -142,6 +147,13 @@ static int cpufreq_governor_userspace(st
+ 		if (rc)
+ 			goto start_out;
+ 
++		if (cpus_using_userspace_governor == 0) {
++			cpufreq_register_notifier(
++					&userspace_cpufreq_notifier_block,
++					CPUFREQ_TRANSITION_NOTIFIER);
++		}
++		cpus_using_userspace_governor++;
++
+ 		cpu_is_managed[cpu] = 1;
+ 		cpu_min_freq[cpu] = policy->min;
+ 		cpu_max_freq[cpu] = policy->max;
+@@ -153,6 +165,13 @@ start_out:
+ 		break;
+ 	case CPUFREQ_GOV_STOP:
+ 		mutex_lock(&userspace_mutex);
++		cpus_using_userspace_governor--;
++		if (cpus_using_userspace_governor == 0) {
++        		cpufreq_unregister_notifier(
++					&userspace_cpufreq_notifier_block,
++					CPUFREQ_TRANSITION_NOTIFIER);
++		}
++
+ 		cpu_is_managed[cpu] = 0;
+ 		cpu_min_freq[cpu] = 0;
+ 		cpu_max_freq[cpu] = 0;
+@@ -198,7 +217,6 @@ EXPORT_SYMBOL(cpufreq_gov_userspace);
+ 
+ static int __init cpufreq_gov_userspace_init(void)
+ {
+-	cpufreq_register_notifier(&userspace_cpufreq_notifier_block, CPUFREQ_TRANSITION_NOTIFIER);
+ 	return cpufreq_register_governor(&cpufreq_gov_userspace);
+ }
+ 
+@@ -206,7 +224,6 @@ static int __init cpufreq_gov_userspace_
+ static void __exit cpufreq_gov_userspace_exit(void)
+ {
+ 	cpufreq_unregister_governor(&cpufreq_gov_userspace);
+-        cpufreq_unregister_notifier(&userspace_cpufreq_notifier_block, CPUFREQ_TRANSITION_NOTIFIER);
+ }
+ 
+ 
+
+Due to rounding and inexact jiffy accounting, idle_ticks can sometimes
+be higher than total_ticks. Make sure those cases are handled as
+zero load case.
+
+Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi at intel.com>
+
+Index: linux-2.6.22-rc-mm/drivers/cpufreq/cpufreq_ondemand.c
+===================================================================
+--- linux-2.6.22-rc-mm.orig/drivers/cpufreq/cpufreq_ondemand.c
++++ linux-2.6.22-rc-mm/drivers/cpufreq/cpufreq_ondemand.c
+@@ -325,7 +325,7 @@ static struct attribute_group dbs_attr_g
+ static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info)
+ {
+ 	unsigned int idle_ticks, total_ticks;
+-	unsigned int load;
++	unsigned int load = 0;
+ 	cputime64_t cur_jiffies;
+ 
+ 	struct cpufreq_policy *policy;
+@@ -370,7 +370,8 @@ static void dbs_check_cpu(struct cpu_dbs
+ 		if (tmp_idle_ticks < idle_ticks)
+ 			idle_ticks = tmp_idle_ticks;
+ 	}
+-	load = (100 * (total_ticks - idle_ticks)) / total_ticks;
++	if (likely(total_ticks > idle_ticks))
++		load = (100 * (total_ticks - idle_ticks)) / total_ticks;
+ 
+ 	/* Check for frequency increase */
+ 	if (load > dbs_tuners_ins.up_threshold) {
+
+With tickless kernel and software coordination os P-states, ondemand
+can look at wrong idle statistics. This can happen when ondemand sampling
+is happening on CPU 0 and due to software coordination sampling also looks at
+utilization of CPU 1. If CPU 1 is in tickless state at that moment, its idle
+statistics will not be uptodate and CPU 0 thinks CPU 1 is idle for less
+amount of time than it actually is.
+
+This can be resolved by looking at all the busy times of CPUs, which is
+accurate, even with tickless, and use that to determine idle time in a
+round about way (total time - busy time).
+
+Thanks to Arjan for originally reporting this ondemand issue on
+Lenovo T61.
+
+Signed-off-by: Venkatesh Pallipadi <venkatesh.pallipadi at intel.com>
+
+Index: linux-2.6.22-rc-mm/drivers/cpufreq/cpufreq_ondemand.c
+===================================================================
+--- linux-2.6.22-rc-mm.orig/drivers/cpufreq/cpufreq_ondemand.c
++++ linux-2.6.22-rc-mm/drivers/cpufreq/cpufreq_ondemand.c
+@@ -96,15 +96,25 @@ static struct dbs_tuners {
+ 
+ static inline cputime64_t get_cpu_idle_time(unsigned int cpu)
+ {
+-	cputime64_t retval;
++	cputime64_t idle_time;
++	cputime64_t cur_jiffies;
++	cputime64_t busy_time;
+ 
+-	retval = cputime64_add(kstat_cpu(cpu).cpustat.idle,
+-			kstat_cpu(cpu).cpustat.iowait);
++	cur_jiffies = jiffies64_to_cputime64(get_jiffies_64());
++	busy_time = cputime64_add(kstat_cpu(cpu).cpustat.user,
++			kstat_cpu(cpu).cpustat.system);
+ 
+-	if (dbs_tuners_ins.ignore_nice)
+-		retval = cputime64_add(retval, kstat_cpu(cpu).cpustat.nice);
++	busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.irq);
++	busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.softirq);
++	busy_time = cputime64_add(busy_time, kstat_cpu(cpu).cpustat.steal);
+ 
+-	return retval;
++	if (!dbs_tuners_ins.ignore_nice) {
++		busy_time = cputime64_add(busy_time,
++				kstat_cpu(cpu).cpustat.nice);
++	}
++
++	idle_time = cputime64_sub(cur_jiffies, busy_time);
++	return idle_time;
+ }
+ 
+ /*
+@@ -339,7 +349,8 @@ static void dbs_check_cpu(struct cpu_dbs
+ 	cur_jiffies = jiffies64_to_cputime64(get_jiffies_64());
+ 	total_ticks = (unsigned int) cputime64_sub(cur_jiffies,
+ 			this_dbs_info->prev_cpu_wall);
+-	this_dbs_info->prev_cpu_wall = cur_jiffies;
++	this_dbs_info->prev_cpu_wall = get_jiffies_64();
++
+ 	if (!total_ticks)
+ 		return;
+ 	/*