[ Removed CC to "Venkatesh Pallipadi <venkatesh.pallipadi@xxxxxxxxx>"
(email-adsress seems not to exist) ]
Indeed, the patch fixes the cpufreq issue here - Thanks.
Feel free to add:
Tested-by: Sedat Dilek <sedat.dilek@xxxxxxxxx>
- Sedat -
P.S.:
I have attached the patch from cpufreq ML, so it is available from
<patchwork.kernel.org>.
On Mon, May 31, 2010 at 8:43 AM, Sedat Dilek <sedat.dilek@xxxxxxxxxxxxxx> wrote:
> Damn!
> Sorry, I have found the following fix at cpufreq ML from Dmitry Monakhov [1].
>
> I will try it and report.
>
> - Sedat -
>
> [1] "[PATCH] cpufreq: call nr_iowait_cpu with disabled preemption"
> http://www.spinics.net/lists/cpufreq/msg01631.html
>
> On Mon, May 31, 2010 at 8:27 AM, Sedat Dilek <sedat.dilek@xxxxxxxxxxxxxx> wrote:
>> Hi,
>>
>> with 2.6.35-rc1 I get a cpufreq-related Call trace on startup on a
>> 32-bit Debian/sid system.
>>
>> Below I added (see investigations at the end):
>>
>> 1. the version of cpufrequtils I am using
>> 2. an output of cpufreq-related loaded kernel-modules
>> 3. an output of cpufreq-related setting in kernel-config
>>
>> Furthermore, I have attched the full dmesg output and kernel-config file.
>>
>> Here is the Call trace:
>>
>> [ 31.647049] BUG: using smp_processor_id() in preemptible [00000000]
>> code: cpufreq-set/2344
>> [ 31.647299] caller is nr_iowait_cpu+0xb/0x1a
>> [ 31.647303] Pid: 2344, comm: cpufreq-set Not tainted
>> 2.6.35-rc1-iniza-686-kms #1
>> [ 31.647306] Call Trace:
>> [ 31.647314] [<c111fdd9>] ? debug_smp_processor_id+0xa5/0xb8
>> [ 31.647319] [<c1022f93>] ? nr_iowait_cpu+0xb/0x1a
>> [ 31.647325] [<c104a93c>] ? update_ts_time_stats+0x36/0x92
>> [ 31.647329] [<c1045fcb>] ? ktime_get+0x5b/0xd9
>> [ 31.647333] [<c104aae9>] ? get_cpu_idle_time_us+0x36/0x62
>> [ 31.647340] [<f94e433d>] ? get_cpu_idle_time+0x12/0x127 [cpufreq_ondemand]
>> [ 31.647345] [<c10dea11>] ? sysfs_add_file_mode+0x48/0x66
>> [ 31.647350] [<c10e0827>] ? internal_create_group+0xcf/0x134
>> [ 31.647355] [<f94e4a3b>] ? cpufreq_governor_dbs+0xaa/0x2a3
>> [cpufreq_ondemand]
>> [ 31.647361] [<c119d94a>] ? __cpufreq_governor+0x2c/0x63
>> [ 31.647365] [<c119e80a>] ? __cpufreq_set_policy+0xea/0x11d
>> [ 31.647370] [<c119ea74>] ? store_scaling_governor+0x58/0x7c
>> [ 31.647375] [<c119ee09>] ? handle_update+0x0/0x8
>> [ 31.647381] [<c1225d70>] ? _raw_spin_unlock_irqrestore+0x22/0x2c
>> [ 31.647385] [<c119ea1c>] ? store_scaling_governor+0x0/0x7c
>> [ 31.647389] [<c119e708>] ? store+0x3c/0x54
>> [ 31.647393] [<c10de76d>] ? sysfs_write_file+0xb8/0xe5
>> [ 31.647397] [<c10de6b5>] ? sysfs_write_file+0x0/0xe5
>> [ 31.647403] [<c109f7c6>] ? vfs_write+0x7f/0xda
>> [ 31.647408] [<c101dad8>] ? do_page_fault+0x1fb/0x25f
>> [ 31.647412] [<c109f8b4>] ? sys_write+0x3c/0x60
>> [ 31.647417] [<c1002e18>] ? sysenter_do_call+0x12/0x28
>>
>> Hope this helps fixing the problem.
>>
>> Kind Regards,
>> - Sedat -
>>
>> P.S.:
>> Recently I watched a video with a speeh of Greg K-H [1] and extracted
>> the maintainers/contributors via:
>>
>> $ ./scripts/get_maintainer.pl -f drivers/cpuidle/governors/menu.c <---
>> nr_iowait_cpu()
>> $ ./scripts/get_maintainer.pl -f drivers/cpufreq/cpufreq_ondemand.c
>> <--- [cpufreq_ondemand]
>>
>> [1] Write and Submit your first Linux kernel Patch:
>> http://fosdem.org/2010/schedule/events/linuxkernelpatch
>>
>> ------ INVESTIGATIONS -----
>>
>> $ env COLUMNS=120 dpkg -l | grep -i cpufreq
>> ii cpufrequtils 007-1 utilities to
>> deal with the cpufreq Linux kernel feature
>> ii libcpufreq0 007-1 shared library
>> to deal with the cpufreq Linux kernel feature
>>
>> $ sudo lsmod | grep freq | sort
>> acpi_cpufreq 4343 1
>> cpufreq_conservative 6260 0
>> cpufreq_ondemand 5810 2
>> cpufreq_performance 590 0
>> cpufreq_powersave 582 0
>> cpufreq_stats 1969 0
>> freq_table 1647 3 acpi_cpufreq,cpufreq_ondemand,cpufreq_stats
>> mperf 899 1 acpi_cpufreq
>> processor 22099 3 acpi_cpufreq
>>
>> $ grep -i freq /boot/config-2.6.35-rc1-iniza-686-kms | grep -i cpu
>> # CPU Frequency scaling
>> CONFIG_CPU_FREQ=y
>> CONFIG_CPU_FREQ_TABLE=m
>> # CONFIG_CPU_FREQ_DEBUG is not set
>> CONFIG_CPU_FREQ_STAT=m
>> # CONFIG_CPU_FREQ_STAT_DETAILS is not set
>> # CONFIG_CPU_FREQ_DEFAULT_GOV_PERFORMANCE is not set
>> # CONFIG_CPU_FREQ_DEFAULT_GOV_POWERSAVE is not set
>> CONFIG_CPU_FREQ_DEFAULT_GOV_USERSPACE=y
>> # CONFIG_CPU_FREQ_DEFAULT_GOV_ONDEMAND is not set
>> # CONFIG_CPU_FREQ_DEFAULT_GOV_CONSERVATIVE is not set
>> CONFIG_CPU_FREQ_GOV_PERFORMANCE=m
>> CONFIG_CPU_FREQ_GOV_POWERSAVE=m
>> CONFIG_CPU_FREQ_GOV_USERSPACE=y
>> CONFIG_CPU_FREQ_GOV_ONDEMAND=m
>> CONFIG_CPU_FREQ_GOV_CONSERVATIVE=m
>> # CPUFreq processor drivers
>> # CONFIG_X86_PCC_CPUFREQ is not set
>> CONFIG_X86_ACPI_CPUFREQ=m
>> CONFIG_X86_CPUFREQ_NFORCE2=m
>>
>
This patch fix complain from preemptive debug
BUG: using smp_processor_id() in preemptible [00000000] code: S06cpuspeed/655
caller is nr_iowait_cpu+0x1a/0x30
Pid: 655, comm: S06cpuspeed Not tainted 2.6.34-quota-cleanups #46
Call Trace:
[<ffffffff812de2ae>] debug_smp_processor_id+0xc2/0xdc
[<ffffffff8102baea>] nr_iowait_cpu+0x1a/0x30
[<ffffffff810671b3>] update_ts_time_stats+0x3c/0x77
[<ffffffff81062894>] ? ktime_get+0x5e/0x98
[<ffffffff8106737c>] get_cpu_idle_time_us+0x49/0x68
[<ffffffffa004d4d1>] get_cpu_idle_time+0x21/0xa9 [cpufreq_ondemand]
[<ffffffffa004dc7f>] cpufreq_governor_dbs+0xcd/0x316 [cpufreq_ondemand]
[<ffffffff813d6b46>] __cpufreq_governor+0x89/0xc7
[<ffffffff813d7936>] __cpufreq_set_policy+0x1ad/0x227
[<ffffffff813d7dee>] store_scaling_governor+0x1a0/0x1e8
[<ffffffff813d8058>] ? handle_update+0x0/0x38
Signed-off-by: Dmitry Monakhov <dmonakhov@xxxxxxxxxx>
---
drivers/cpufreq/cpufreq_ondemand.c | 7 ++++++-
1 files changed, 6 insertions(+), 1 deletions(-)
diff --git a/drivers/cpufreq/cpufreq_ondemand.c b/drivers/cpufreq/cpufreq_ondemand.c
index e131421..c1b1a96 100644
--- a/drivers/cpufreq/cpufreq_ondemand.c
+++ b/drivers/cpufreq/cpufreq_ondemand.c
@@ -363,6 +363,7 @@ static ssize_t store_ignore_nice_load(struct kobject *a, struct attribute *b,
dbs_tuners_ins.ignore_nice = input;
/* we need to re-evaluate prev_cpu_idle */
+ get_cpu();
for_each_online_cpu(j) {
struct cpu_dbs_info_s *dbs_info;
dbs_info = &per_cpu(od_cpu_dbs_info, j);
@@ -372,6 +373,7 @@ static ssize_t store_ignore_nice_load(struct kobject *a, struct attribute *b,
dbs_info->prev_cpu_nice = kstat_cpu(j).cpustat.nice;
}
+ put_cpu();
mutex_unlock(&dbs_mutex);
return count;
@@ -493,9 +495,10 @@ static void dbs_check_cpu(struct cpu_dbs_info_s *this_dbs_info)
j_dbs_info = &per_cpu(od_cpu_dbs_info, j);
+ get_cpu();
cur_idle_time = get_cpu_idle_time(j, &cur_wall_time);
cur_iowait_time = get_cpu_iowait_time(j, &cur_wall_time);
-
+ put_cpu();
wall_time = (unsigned int) cputime64_sub(cur_wall_time,
j_dbs_info->prev_cpu_wall);
j_dbs_info->prev_cpu_wall = cur_wall_time;
@@ -695,6 +698,7 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
}
dbs_enable++;
+ get_cpu();
for_each_cpu(j, policy->cpus) {
struct cpu_dbs_info_s *j_dbs_info;
j_dbs_info = &per_cpu(od_cpu_dbs_info, j);
@@ -708,6 +712,7 @@ static int cpufreq_governor_dbs(struct cpufreq_policy *policy,
}
}
this_dbs_info->cpu = cpu;
+ put_cpu();
ondemand_powersave_bias_init_cpu(cpu);
/*
* Start the timerschedule work, when this governor
--
1.6.6.1
