Merge pull request #2803 from rorysexton/dev_branch

cpufreq plugin: Read number of p-state transitions and time spent in each p-state

diff --git a/src/collectd.conf.pod b/src/collectd.conf.pod
index e11e514..6f5ee7f 100644 (file)
--- a/src/collectd.conf.pod
+++ b/src/collectd.conf.pod
@@ -1662,6 +1662,10 @@ installed) to get the current CPU frequency. If this file does not exist make
 sure B<cpufreqd> (L<http://cpufreqd.sourceforge.net/>) or a similar tool is
 installed and an "cpu governor" (that's a kernel module) is loaded.
 
+If the system has the I<cpufreq-stats> kernel module loaded, this plugin reports
+the rate of p-state (cpu frequency) transitions and the percentage of time spent
+in each p-state.
+
 =head2 Plugin C<cpusleep>
 
 This plugin doesn't have any options. It reads CLOCK_BOOTTIME and

diff --git a/src/cpufreq.c b/src/cpufreq.c
index 851aad4..3e3244c 100644 (file)
--- a/src/cpufreq.c
+++ b/src/cpufreq.c
@@ -25,19 +25,63 @@
 #include "common.h"
 #include "plugin.h"
 
+#define MAX_AVAIL_FREQS 20
+
 static int num_cpu;
 
+struct cpu_data_t {
+  value_to_rate_state_t time_state[MAX_AVAIL_FREQS];
+} * cpu_data;
+
+/* Flags denoting capability of reporting CPU frequency statistics. */
+static bool report_p_stats = false;
+
+static void cpufreq_stats_init(void) {
+  cpu_data = calloc(num_cpu, sizeof(struct cpu_data_t));
+  if (cpu_data == NULL)
+    return;
+
+  report_p_stats = true;
+
+  /* Check for stats module and disable if not present. */
+  for (int i = 0; i < num_cpu; i++) {
+    char filename[PATH_MAX];
+
+    snprintf(filename, sizeof(filename),
+             "/sys/devices/system/cpu/cpu%d/cpufreq/stats/time_in_state", i);
+    if (access(filename, R_OK)) {
+      NOTICE("cpufreq plugin: File %s not exists or no access. P-State "
+             "statistics will not be reported. Check if `cpufreq-stats' kernel "
+             "module is loaded.",
+             filename);
+      report_p_stats = false;
+      break;
+    }
+
+    snprintf(filename, sizeof(filename),
+             "/sys/devices/system/cpu/cpu%d/cpufreq/stats/total_trans", i);
+    if (access(filename, R_OK)) {
+      NOTICE("cpufreq plugin: File %s not exists or no access. P-State "
+             "statistics will not be reported. Check if `cpufreq-stats' kernel "
+             "module is loaded.",
+             filename);
+      report_p_stats = false;
+      break;
+    }
+  }
+  return;
+}
+
 static int cpufreq_init(void) {
-  int status;
-  char filename[256];
+  char filename[PATH_MAX];
 
   num_cpu = 0;
 
   while (1) {
-    status = snprintf(filename, sizeof(filename),
-                      "/sys/devices/system/cpu/cpu%d/cpufreq/"
-                      "scaling_cur_freq",
-                      num_cpu);
+    int status = snprintf(filename, sizeof(filename),
+                          "/sys/devices/system/cpu/cpu%d/cpufreq/"
+                          "scaling_cur_freq",
+                          num_cpu);
     if ((status < 1) || ((unsigned int)status >= sizeof(filename)))
       break;
 
@@ -48,6 +92,7 @@ static int cpufreq_init(void) {
   }
 
   INFO("cpufreq plugin: Found %d CPU%s", num_cpu, (num_cpu == 1) ? "" : "s");
+  cpufreq_stats_init();
 
   if (num_cpu == 0)
     plugin_unregister_read("cpufreq");
@@ -55,23 +100,91 @@ static int cpufreq_init(void) {
   return 0;
 } /* int cpufreq_init */
 
-static void cpufreq_submit(int cpu_num, value_t value) {
+static void cpufreq_submit(int cpu_num, const char *type,
+                           const char *type_instance, value_t *value) {
   value_list_t vl = VALUE_LIST_INIT;
 
-  vl.values = &value;
+  vl.values = value;
   vl.values_len = 1;
   sstrncpy(vl.plugin, "cpufreq", sizeof(vl.plugin));
-  sstrncpy(vl.type, "cpufreq", sizeof(vl.type));
-  snprintf(vl.type_instance, sizeof(vl.type_instance), "%i", cpu_num);
+  snprintf(vl.plugin_instance, sizeof(vl.plugin_instance), "%i", cpu_num);
+  if (type != NULL)
+    sstrncpy(vl.type, type, sizeof(vl.type));
+  if (type_instance != NULL)
+    sstrncpy(vl.type_instance, type_instance, sizeof(vl.type_instance));
 
   plugin_dispatch_values(&vl);
 }
 
+static void cpufreq_read_stats(int cpu) {
+  char filename[PATH_MAX];
+  /* Read total transitions for cpu frequency */
+  snprintf(filename, sizeof(filename),
+           "/sys/devices/system/cpu/cpu%d/cpufreq/stats/total_trans", cpu);
+
+  value_t v;
+  if (parse_value_file(filename, &v, DS_TYPE_DERIVE) != 0) {
+    ERROR("cpufreq plugin: Reading \"%s\" failed.", filename);
+    return;
+  }
+  cpufreq_submit(cpu, "transitions", NULL, &v);
+
+  /* Determine percentage time in each state for cpu during previous
+   * interval. */
+  snprintf(filename, sizeof(filename),
+           "/sys/devices/system/cpu/cpu%d/cpufreq/stats/time_in_state", cpu);
+
+  FILE *fh = fopen(filename, "r");
+  if (fh == NULL) {
+    ERROR("cpufreq plugin: Reading \"%s\" failed.", filename);
+    return;
+  }
+
+  int state_index = 0;
+  cdtime_t now = cdtime();
+  char buffer[DATA_MAX_NAME_LEN];
+
+  while (fgets(buffer, sizeof(buffer), fh) != NULL) {
+    unsigned int frequency;
+    unsigned long long time;
+
+    /*
+     * State time units is 10ms. To get rate of seconds per second
+     * we have to divide by 100. To get percents we have to multiply it
+     * by 100 back. So, just use parsed value directly.
+     */
+    if (!sscanf(buffer, "%u%llu", &frequency, &time)) {
+      ERROR("cpufreq plugin: Reading \"%s\" failed.", filename);
+      break;
+    }
+
+    char state[DATA_MAX_NAME_LEN];
+    snprintf(state, sizeof(state), "%u", frequency);
+
+    if (state_index >= MAX_AVAIL_FREQS) {
+      NOTICE("cpufreq plugin: Found too many frequency states (%d > %d). "
+             "Plugin needs to be recompiled. Please open a bug report for "
+             "this.",
+             (state_index + 1), MAX_AVAIL_FREQS);
+      break;
+    }
+
+    gauge_t g;
+    if (value_to_rate(&g, (value_t){.derive = time}, DS_TYPE_DERIVE, now,
+                      &(cpu_data[cpu].time_state[state_index])) == 0) {
+      cpufreq_submit(cpu, "percent", state, &(value_t){.gauge = g});
+    }
+    state_index++;
+  }
+  fclose(fh);
+}
+
 static int cpufreq_read(void) {
-  for (int i = 0; i < num_cpu; i++) {
+  for (int cpu = 0; cpu < num_cpu; cpu++) {
     char filename[PATH_MAX];
+    /* Read cpu frequency */
     snprintf(filename, sizeof(filename),
-             "/sys/devices/system/cpu/cpu%d/cpufreq/scaling_cur_freq", i);
+             "/sys/devices/system/cpu/cpu%d/cpufreq/scaling_cur_freq", cpu);
 
     value_t v;
     if (parse_value_file(filename, &v, DS_TYPE_GAUGE) != 0) {
@@ -82,9 +195,11 @@ static int cpufreq_read(void) {
     /* convert kHz to Hz */
     v.gauge *= 1000.0;
 
-    cpufreq_submit(i, v);
-  }
+    cpufreq_submit(cpu, "cpufreq", NULL, &v);
 
+    if (report_p_stats)
+      cpufreq_read_stats(cpu);
+  }
   return 0;
 } /* int cpufreq_read */
 

diff --git a/src/types.db b/src/types.db
index 0370b7f..e9de64f 100644 (file)
--- a/src/types.db
+++ b/src/types.db
@@ -260,6 +260,7 @@ total_threads           value:DERIVE:0:U
 total_time_in_ms        value:DERIVE:0:U
 total_values            value:DERIVE:0:U
 turbo_enabled           value:GAUGE:0:1
+transitions             value:DERIVE:0:U
 uptime                  value:GAUGE:0:4294967295
 uncore_ratio            value:GAUGE:0:U
 users                   value:GAUGE:0:65535