commit 46a310b ([CPUFREQ] Don't set stat->last_index to -1 if the pol->cur has incorrect value.) tries to handle case where policy->cur does not match any entry in freq_table.
As indicated in the above commit, the exact match search of freq_table_get index will return a -1 which is stored in stat->last_index. However, as a result of the above commit, cpufreq_stat_notifier_trans which updates the statistics, fails to update any *further* valid transitions that take place as stat->last_index is -1 as the condition occurs at boot and never solved.
To fix this issue, lets create another entry for time_in_state and trans_table that will tell the user that CPU was running on unknown frequency for some time.
This is how the output looks like on my thinkpad (Removed some entries to keep it simple):
$ cat /sys/devices/system/cpu/cpu0/cpufreq/stats/time_in_state 2800000 46 2600000 138 1200000 65 1000000 152 800000 34803 unknown 0
$ cat /sys/devices/system/cpu/cpu0/cpufreq/stats/trans_table From : To : 2801000 2800000 2600000 2400000 2200000 2000000 unknown 2801000: 0 15 20 9 13 17 0 2800000: 13 0 4 1 0 1 0 2600000: 26 1 0 5 1 1 0 2400000: 11 0 6 0 1 1 0 2200000: 8 1 5 3 0 0 0 2000000: 11 1 2 1 2 0 0 unknown: 0 0 0 0 0 0 0
Reported-by: Carlos Hernandez ceh@ti.com Reported-and-tested-by: Nishanth Menon nm@ti.com Signed-off-by: Viresh Kumar viresh.kumar@linaro.org --- drivers/cpufreq/cpufreq_stats.c | 45 ++++++++++++++++++++++++++++------------- 1 file changed, 31 insertions(+), 14 deletions(-)
diff --git a/drivers/cpufreq/cpufreq_stats.c b/drivers/cpufreq/cpufreq_stats.c index 4cf0d28..ebb21cd 100644 --- a/drivers/cpufreq/cpufreq_stats.c +++ b/drivers/cpufreq/cpufreq_stats.c @@ -72,9 +72,13 @@ static ssize_t show_time_in_state(struct cpufreq_policy *policy, char *buf) return 0; cpufreq_stats_update(stat->cpu); for (i = 0; i < stat->state_num; i++) { - len += sprintf(buf + len, "%u %llu\n", stat->freq_table[i], - (unsigned long long) - jiffies_64_to_clock_t(stat->time_in_state[i])); + if (stat->freq_table[i] == -1) + return sprintf(buf + len, "unknown"); + else + return sprintf(buf + len, "%u", stat->freq_table[i]); + + len += sprintf(buf + len, " %llu\n", (unsigned long long) + jiffies_64_to_clock_t(stat->time_in_state[i])); } return len; } @@ -94,8 +98,12 @@ static ssize_t show_trans_table(struct cpufreq_policy *policy, char *buf) for (i = 0; i < stat->state_num; i++) { if (len >= PAGE_SIZE) break; - len += snprintf(buf + len, PAGE_SIZE - len, "%9u ", - stat->freq_table[i]); + if (stat->freq_table[i] == -1) + len += snprintf(buf + len, PAGE_SIZE - len, "%9s ", + "unknown"); + else + len += snprintf(buf + len, PAGE_SIZE - len, "%9u ", + stat->freq_table[i]); } if (len >= PAGE_SIZE) return PAGE_SIZE; @@ -106,8 +114,12 @@ static ssize_t show_trans_table(struct cpufreq_policy *policy, char *buf) if (len >= PAGE_SIZE) break;
- len += snprintf(buf + len, PAGE_SIZE - len, "%9u: ", - stat->freq_table[i]); + if (stat->freq_table[i] == -1) + len += snprintf(buf + len, PAGE_SIZE - len, "%9s: ", + "unknown"); + else + len += snprintf(buf + len, PAGE_SIZE - len, "%9u: ", + stat->freq_table[i]);
for (j = 0; j < stat->state_num; j++) { if (len >= PAGE_SIZE) @@ -145,10 +157,12 @@ static struct attribute_group stats_attr_group = { static int freq_table_get_index(struct cpufreq_stats *stat, unsigned int freq) { int index; - for (index = 0; index < stat->max_state; index++) + for (index = 0; index < stat->max_state - 1; index++) if (stat->freq_table[index] == freq) return index; - return -1; + + /* Last state is INVALID, to mark out of table frequency */ + return stat->max_state - 1; }
/* should be called late in the CPU removal sequence so that the stats @@ -222,6 +236,9 @@ static int cpufreq_stats_create_table(struct cpufreq_policy *policy, count++; }
+ /* An extra entry for Invalid frequencies */ + count++; + alloc_size = count * sizeof(int) + count * sizeof(u64);
#ifdef CONFIG_CPU_FREQ_STAT_DETAILS @@ -243,9 +260,13 @@ static int cpufreq_stats_create_table(struct cpufreq_policy *policy, unsigned int freq = table[i].frequency; if (freq == CPUFREQ_ENTRY_INVALID) continue; - if (freq_table_get_index(stat, freq) == -1) + if (freq_table_get_index(stat, freq) == stat->max_state - 1) stat->freq_table[j++] = freq; } + + /* Mark Invalid freq as max value to indicate Invalid freq */ + stat->freq_table[j++] = -1; + stat->state_num = j; spin_lock(&cpufreq_stats_lock); stat->last_time = get_jiffies_64(); @@ -315,10 +336,6 @@ static int cpufreq_stat_notifier_trans(struct notifier_block *nb, old_index = stat->last_index; new_index = freq_table_get_index(stat, freq->new);
- /* We can't do stat->time_in_state[-1]= .. */ - if (old_index == -1 || new_index == -1) - return 0; - cpufreq_stats_update(freq->cpu);
if (old_index == new_index)
On Tue, 2013-11-19 at 11:10 +0530, Viresh Kumar wrote:
commit 46a310b ([CPUFREQ] Don't set stat->last_index to -1 if the pol->cur has incorrect value.) tries to handle case where policy->cur does not match any entry in freq_table.
As indicated in the above commit, the exact match search of freq_table_get index will return a -1 which is stored in stat->last_index. However, as a result of the above commit, cpufreq_stat_notifier_trans which updates the statistics, fails to update any *further* valid transitions that take place as stat->last_index is -1 as the condition occurs at boot and never solved.
To fix this issue, lets create another entry for time_in_state and trans_table that will tell the user that CPU was running on unknown frequency for some time.
This is how the output looks like on my thinkpad (Removed some entries to keep it simple):
$ cat /sys/devices/system/cpu/cpu0/cpufreq/stats/time_in_state 2800000 46 2600000 138 1200000 65 1000000 152 800000 34803 unknown 0
I'm wondering if this would break any user-side code parsing this output and does that matter? (E.g. powertop will end up doing strtoull on "unknown" and get zero as the frequency.) Would omitting 'unknown' if the time is zero be safer in this repect? Possibly not, inconsistency is probably worse than change.
On 19 November 2013 14:12, Jon Medhurst (Tixy) tixy@linaro.org wrote:
I'm wondering if this would break any user-side code parsing this output and does that matter? (E.g. powertop will end up doing strtoull on "unknown" and get zero as the frequency.) Would omitting 'unknown' if the time is zero be safer in this repect? Possibly not, inconsistency is probably worse than change.
We wouldn't know about this on bootup all the time. I thought of doing this only when an invalid frequency is configured at boot. Because after that cpufreq will never configure any freq out of freq-table.
But in case of suspend/resume control again goes to bootloaders and they might end up configuring unknown freq again. And we aren't allocating these data structures on suspend resume but only during system boot.
Even if powertop reads it zero then also it isn't that bad.. because the other option we had was to mark it zero or -1 (uint max).. But unknown looks better for readability..
-- viresh
On Tuesday 19 November 2013 11:10 AM, Viresh Kumar wrote:
diff --git a/drivers/cpufreq/cpufreq_stats.c b/drivers/cpufreq/cpufreq_stats.c index 4cf0d28..ebb21cd 100644 --- a/drivers/cpufreq/cpufreq_stats.c +++ b/drivers/cpufreq/cpufreq_stats.c @@ -72,9 +72,13 @@ static ssize_t show_time_in_state(struct cpufreq_policy *policy, char *buf) return 0; cpufreq_stats_update(stat->cpu); for (i = 0; i < stat->state_num; i++) {
len += sprintf(buf + len, "%u %llu\n", stat->freq_table[i],
(unsigned long long)
jiffies_64_to_clock_t(stat->time_in_state[i]));
if (stat->freq_table[i] == -1)
return sprintf(buf + len, "unknown");
else
return sprintf(buf + len, "%u", stat->freq_table[i]);
len += sprintf(buf + len, " %llu\n", (unsigned long long)
} return len;jiffies_64_to_clock_t(stat->time_in_state[i]));
Somehow I forgot to add below fix with this patch, which I have tested at that time.
diff --git a/drivers/cpufreq/cpufreq_stats.c b/drivers/cpufreq/cpufreq_stats.c index ebb21cd..89012d1 100644 --- a/drivers/cpufreq/cpufreq_stats.c +++ b/drivers/cpufreq/cpufreq_stats.c @@ -73,9 +73,9 @@ static ssize_t show_time_in_state(struct cpufreq_policy *policy, char *buf) cpufreq_stats_update(stat->cpu); for (i = 0; i < stat->state_num; i++) { if (stat->freq_table[i] == -1) - return sprintf(buf + len, "unknown"); + len+= sprintf(buf + len, "unknown"); else - return sprintf(buf + len, "%u", stat->freq_table[i]); + len+= sprintf(buf + len, "%u", stat->freq_table[i]);
len += sprintf(buf + len, " %llu\n", (unsigned long long) jiffies_64_to_clock_t(stat->time_in_state[i]));
linaro-kernel@lists.linaro.org