/**
- * collectd - src/xmms.c
- * Copyright (C) 2013
+ * collectd - src/mic.c
+ * Copyright (C) 2013 Battelle Memorial Institute
*
* This program is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License as published by the
* 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
*
* Authors:
- * Evan Felix <evan.felix@pnnl.gov
+ * Evan Felix <evan.felix at pnnl.gov>
**/
#include "collectd.h"
-#include "plugin.h"
+
#include "common.h"
+#include "plugin.h"
+#include "utils_ignorelist.h"
-#include <MicAccessTypes.h>
-#include <MicAccessErrorTypes.h>
#include <MicAccessApi.h>
+#include <MicAccessErrorTypes.h>
+#include <MicAccessTypes.h>
+#include <MicPowerManagerAPI.h>
#include <MicThermalAPI.h>
#define MAX_MICS 32
#define MAX_CORES 256
static MicDeviceOnSystem mics[MAX_MICS];
-static U32 numMics = MAX_MICS;
-static HANDLE micHandle=NULL;
-#define NUM_THERMS 7
-static const int therms[NUM_THERMS] = {eMicThermalDie,eMicThermalDevMem,eMicThermalFin,eMicThermalFout,eMicThermalVccp,eMicThermalVddg,eMicThermalVddq};
-static const char *thermNames[NUM_THERMS] = {"die","devmem","fin","fout","vccp","vddg","vddq"};
+static U32 num_mics = 0;
+static HANDLE mic_handle = NULL;
+
+static int const therm_ids[] = {
+ eMicThermalDie, eMicThermalDevMem, eMicThermalFin, eMicThermalFout,
+ eMicThermalVccp, eMicThermalVddg, eMicThermalVddq};
+static char const *const therm_names[] = {"die", "devmem", "fin", "fout",
+ "vccp", "vddg", "vddq"};
+
+static const char *config_keys[] = {
+ "ShowCPU", "ShowCPUCores", "ShowMemory",
+ "ShowTemperatures", "Temperature", "IgnoreSelectedTemperature",
+ "ShowPower", "Power", "IgnoreSelectedPower"};
+static int config_keys_num = STATIC_ARRAY_SIZE(config_keys);
+
+static _Bool show_cpu = 1;
+static _Bool show_cpu_cores = 1;
+static _Bool show_memory = 1;
+static _Bool show_temps = 1;
+static ignorelist_t *temp_ignore = NULL;
+static _Bool show_power = 1;
+static ignorelist_t *power_ignore = NULL;
+
+static int mic_init(void) {
+ U32 ret;
+ U32 mic_count;
+
+ if (mic_handle)
+ return (0);
+
+ mic_count = (U32)STATIC_ARRAY_SIZE(mics);
+ ret = MicInitAPI(&mic_handle, eTARGET_SCIF_DRIVER, mics, &mic_count);
+ if (ret != MIC_ACCESS_API_SUCCESS) {
+ ERROR("mic plugin: Problem initializing MicAccessAPI: %s",
+ MicGetErrorString(ret));
+ }
+ DEBUG("mic plugin: found: %" PRIu32 " MIC(s)", mic_count);
+
+ if (mic_count < 0 || mic_count >= MAX_MICS) {
+ ERROR("mic plugin: No Intel MICs in system");
+ return (1);
+ } else {
+ num_mics = mic_count;
+ return (0);
+ }
+}
+
+static int mic_config(const char *key, const char *value) {
+ if (temp_ignore == NULL)
+ temp_ignore = ignorelist_create(1);
+ if (power_ignore == NULL)
+ power_ignore = ignorelist_create(1);
+ if (temp_ignore == NULL || power_ignore == NULL)
+ return (1);
+
+ if (strcasecmp("ShowCPU", key) == 0) {
+ show_cpu = IS_TRUE(value);
+ } else if (strcasecmp("ShowCPUCores", key) == 0) {
+ show_cpu_cores = IS_TRUE(value);
+ } else if (strcasecmp("ShowTemperatures", key) == 0) {
+ show_temps = IS_TRUE(value);
+ } else if (strcasecmp("ShowMemory", key) == 0) {
+ show_memory = IS_TRUE(value);
+ } else if (strcasecmp("ShowPower", key) == 0) {
+ show_power = IS_TRUE(value);
+ } else if (strcasecmp("Temperature", key) == 0) {
+ ignorelist_add(temp_ignore, value);
+ } else if (strcasecmp("IgnoreSelectedTemperature", key) == 0) {
+ int invert = 1;
+ if (IS_TRUE(value))
+ invert = 0;
+ ignorelist_set_invert(temp_ignore, invert);
+ } else if (strcasecmp("Power", key) == 0) {
+ ignorelist_add(power_ignore, value);
+ } else if (strcasecmp("IgnoreSelectedPower", key) == 0) {
+ int invert = 1;
+ if (IS_TRUE(value))
+ invert = 0;
+ ignorelist_set_invert(power_ignore, invert);
+ } else {
+ return (-1);
+ }
+ return (0);
+}
+
+static void mic_submit_memory_use(int micnumber, const char *type_instance,
+ U32 value) {
+ value_list_t vl = VALUE_LIST_INIT;
+
+ /* MicAccessAPI reports KB's of memory, adjust for this */
+ DEBUG("mic plugin: Memory Value Report; %u %lf", value,
+ ((gauge_t)value) * 1024.0);
+
+ vl.values = &(value_t){.gauge = ((gauge_t)value) * 1024.0};
+ vl.values_len = 1;
+
+ strncpy(vl.plugin, "mic", sizeof(vl.plugin));
+ ssnprintf(vl.plugin_instance, sizeof(vl.plugin_instance), "%i", micnumber);
+ strncpy(vl.type, "memory", sizeof(vl.type));
+ strncpy(vl.type_instance, type_instance, sizeof(vl.type_instance));
+ plugin_dispatch_values(&vl);
+}
-static int mic_init (void)
-{
+/* Gather memory Utilization */
+static int mic_read_memory(int mic) {
U32 ret;
+ U32 mem_total, mem_free, mem_bufs;
- ret = MicInitAPI(&micHandle, eTARGET_SCIF_DRIVER, mics, &numMics);
+ ret = MicGetMemoryUtilization(mic_handle, &mem_total, &mem_free, &mem_bufs);
if (ret != MIC_ACCESS_API_SUCCESS) {
- ERROR("Problem initializing MicAccessAPI: %s",MicGetErrorString(ret));
+ ERROR("mic plugin: Problem getting Memory Utilization: %s",
+ MicGetErrorString(ret));
+ return (1);
}
- INFO("MICs found: %d",numMics);
- if (numMics<0 || numMics>=MAX_MICS)
- return (1);
- else
- return (0);
+ mic_submit_memory_use(mic, "free", mem_free);
+ mic_submit_memory_use(mic, "used", mem_total - mem_free - mem_bufs);
+ mic_submit_memory_use(mic, "buffered", mem_bufs);
+ DEBUG("mic plugin: Memory Read: %u %u %u", mem_total, mem_free, mem_bufs);
+ return (0);
}
-static void mic_submit_memory_use(int micnumber, const char *type, gauge_t val)
-{
- value_t values[1];
+static void mic_submit_temp(int micnumber, const char *type, gauge_t value) {
value_list_t vl = VALUE_LIST_INIT;
- values[0].gauge = val;
+ vl.values = &(value_t){.gauge = value};
+ vl.values_len = 1;
- vl.values=values;
- vl.values_len=1;
+ strncpy(vl.host, hostname_g, sizeof(vl.host));
+ strncpy(vl.plugin, "mic", sizeof(vl.plugin));
+ ssnprintf(vl.plugin_instance, sizeof(vl.plugin_instance), "%i", micnumber);
+ strncpy(vl.type, "temperature", sizeof(vl.type));
+ strncpy(vl.type_instance, type, sizeof(vl.type_instance));
- strncpy (vl.host, hostname_g, sizeof (vl.host));
- strncpy (vl.plugin, "mic", sizeof (vl.plugin));
- ssnprintf (vl.plugin_instance, sizeof (vl.plugin_instance), "%i", micnumber);
- strncpy (vl.type, "memory", sizeof (vl.type));
- strncpy (vl.type_instance, type, sizeof (vl.type_instance));
+ plugin_dispatch_values(&vl);
+}
- plugin_dispatch_values (&vl);
-}
+/* Gather Temperature Information */
+static int mic_read_temps(int mic) {
+ size_t num_therms = STATIC_ARRAY_SIZE(therm_ids);
+
+ for (size_t j = 0; j < num_therms; j++) {
+ U32 status;
+ U32 temp_buffer;
+ U32 buffer_size = (U32)sizeof(temp_buffer);
+ char const *name = therm_names[j];
+
+ if (ignorelist_match(temp_ignore, name) != 0)
+ continue;
+
+ status =
+ MicGetTemperature(mic_handle, therm_ids[j], &temp_buffer, &buffer_size);
+ if (status != MIC_ACCESS_API_SUCCESS) {
+ ERROR("mic plugin: Error reading temperature \"%s\": "
+ "%s",
+ name, MicGetErrorString(status));
+ return (1);
+ }
+ mic_submit_temp(mic, name, temp_buffer);
+ }
+ return (0);
+}
-static void mic_submit_temp(int micnumber, const char *type, gauge_t val)
-{
- value_t values[1];
+static void mic_submit_cpu(int micnumber, const char *type_instance, int core,
+ derive_t value) {
value_list_t vl = VALUE_LIST_INIT;
- values[0].gauge = val;
+ vl.values = &(value_t){.derive = value};
+ vl.values_len = 1;
- vl.values=values;
- vl.values_len=1;
+ strncpy(vl.host, hostname_g, sizeof(vl.host));
+ strncpy(vl.plugin, "mic", sizeof(vl.plugin));
+ if (core < 0) /* global aggregation */
+ ssnprintf(vl.plugin_instance, sizeof(vl.plugin_instance), "%i", micnumber);
+ else /* per-core statistics */
+ ssnprintf(vl.plugin_instance, sizeof(vl.plugin_instance), "%i-cpu-%i",
+ micnumber, core);
+ strncpy(vl.type, "cpu", sizeof(vl.type));
+ strncpy(vl.type_instance, type_instance, sizeof(vl.type_instance));
- strncpy (vl.host, hostname_g, sizeof (vl.host));
- strncpy (vl.plugin, "mic", sizeof (vl.plugin));
- ssnprintf (vl.plugin_instance, sizeof (vl.plugin_instance), "%i", micnumber);
- strncpy (vl.type, "temperature", sizeof (vl.type));
- strncpy (vl.type_instance, type, sizeof (vl.type_instance));
+ plugin_dispatch_values(&vl);
+}
+
+/*Gather CPU Utilization Information */
+static int mic_read_cpu(int mic) {
+ MicCoreUtil core_util;
+ MicCoreJiff core_jiffs[MAX_CORES];
+ U32 core_jiffs_size;
+ U32 status;
+
+ core_jiffs_size = MAX_CORES * sizeof(MicCoreJiff);
+ status = MicGetCoreUtilization(mic_handle, &core_util, core_jiffs,
+ &core_jiffs_size);
+ if (status != MIC_ACCESS_API_SUCCESS) {
+ ERROR("mic plugin: Problem getting CPU utilization: %s",
+ MicGetErrorString(status));
+ return (-1);
+ }
- plugin_dispatch_values (&vl);
-}
+ if (show_cpu) {
+ mic_submit_cpu(mic, "user", -1, core_util.sum.user);
+ mic_submit_cpu(mic, "sys", -1, core_util.sum.sys);
+ mic_submit_cpu(mic, "nice", -1, core_util.sum.nice);
+ mic_submit_cpu(mic, "idle", -1, core_util.sum.idle);
+ }
-static void mic_submit_cpu(int micnumber, const char *type, int core, derive_t val)
-{
- value_t values[1];
+ if (show_cpu_cores) {
+ for (int j = 0; j < core_util.core; j++) {
+ mic_submit_cpu(mic, "user", j, core_jiffs[j].user);
+ mic_submit_cpu(mic, "sys", j, core_jiffs[j].sys);
+ mic_submit_cpu(mic, "nice", j, core_jiffs[j].nice);
+ mic_submit_cpu(mic, "idle", j, core_jiffs[j].idle);
+ }
+ }
+ return (0);
+}
+
+static void mic_submit_power(int micnumber, const char *type,
+ const char *type_instance, gauge_t value) {
value_list_t vl = VALUE_LIST_INIT;
- values[0].derive = val;
+ vl.values = &(value_t){.gauge = value};
+ vl.values_len = 1;
- vl.values=values;
- vl.values_len=1;
+ strncpy(vl.host, hostname_g, sizeof(vl.host));
+ strncpy(vl.plugin, "mic", sizeof(vl.plugin));
+ ssnprintf(vl.plugin_instance, sizeof(vl.plugin_instance), "%i", micnumber);
+ strncpy(vl.type, type, sizeof(vl.type));
+ strncpy(vl.type_instance, type_instance, sizeof(vl.type_instance));
- strncpy (vl.host, hostname_g, sizeof (vl.host));
- strncpy (vl.plugin, "mic", sizeof (vl.plugin));
- ssnprintf (vl.plugin_instance, sizeof (vl.plugin_instance), "%i", micnumber);
- strncpy (vl.type, "cpu", sizeof (vl.type));
- if (core < 0)
- strncpy (vl.type_instance, type, sizeof (vl.type_instance));
- else
- ssnprintf (vl.type_instance, sizeof (vl.type_instance), "%i-%s", core, type);
+ plugin_dispatch_values(&vl);
+}
- plugin_dispatch_values (&vl);
-}
+/* Gather Power Information */
+static int mic_read_power(int mic) {
+ U32 ret;
+ MicPwrUsage power_use;
+ ret = MicGetPowerUsage(mic_handle, &power_use);
+ if (ret != MIC_ACCESS_API_SUCCESS) {
+ ERROR("mic plugin: Problem getting Power Usage: %s",
+ MicGetErrorString(ret));
+ return (1);
+ }
+/* power is in uWatts, current in mA, voltage in uVolts.. convert to
+ * base unit */
+#define SUB_POWER(name) \
+ do { \
+ if (ignorelist_match(power_ignore, #name) == 0) \
+ mic_submit_power(mic, "power", #name, \
+ (gauge_t)power_use.name.prr * 0.000001); \
+ } while (0)
+#define SUB_VOLTS(name) \
+ do { \
+ if (ignorelist_match(power_ignore, #name) == 0) { \
+ mic_submit_power(mic, "power", #name, \
+ (gauge_t)(power_use.name.pwr * 0.000001)); \
+ mic_submit_power(mic, "current", #name, \
+ (gauge_t)(power_use.name.cur * 0.001)); \
+ mic_submit_power(mic, "voltage", #name, \
+ (gauge_t)(power_use.name.volt * 0.000001)); \
+ } \
+ } while (0)
+
+ SUB_POWER(total0);
+ SUB_POWER(total1);
+ SUB_POWER(inst);
+ SUB_POWER(imax);
+ SUB_POWER(pcie);
+ SUB_POWER(c2x3);
+ SUB_POWER(c2x4);
+ SUB_VOLTS(vccp);
+ SUB_VOLTS(vddg);
+ SUB_VOLTS(vddq);
-static int mic_read (void)
-{
- int i,j;
- U32 ret,bufferSize;
- U32 *tempBuffer;
+ return (0);
+}
+
+static int mic_read(void) {
+ U32 ret;
int error;
- U32 mem_total,mem_used,mem_bufs;
- MicCoreUtil coreUtil;
- MicCoreJiff coreJiffs[MAX_CORES];
-
- error=0;
- for (i=0;i<numMics;i++) {
- ret = MicInitAdapter(&micHandle,&mics[i]);
- if (ret != MIC_ACCESS_API_SUCCESS) {
- ERROR("Problem initializing MicAdapter: %s",MicGetErrorString(ret));
- error=1;
- break;
- }
-
- /* Gather memory Utilization */
- ret = MicGetMemoryUtilization(micHandle,&mem_total,&mem_used,&mem_bufs);
- if (ret != MIC_ACCESS_API_SUCCESS) {
- ERROR("Problem getting Memory Utilization: %s",MicGetErrorString(ret));
- error=3;
- break;
- }
- /* API reprots KB's of memory, adjust for this */
- mic_submit_memory_use(i,"total",mem_total*1024);
- mic_submit_memory_use(i,"used",mem_used*1024);
- mic_submit_memory_use(i,"bufs",mem_bufs*1024);
- /*INFO("Memory Read: %u %u %u",mem_total,mem_used,mem_bufs);*/
-
- /* Gather Temperature Information */
- bufferSize = sizeof(U32);
- tempBuffer = malloc(bufferSize);
- for (j=0;j<NUM_THERMS;j++) {
- ret = MicGetTemperature(micHandle,therms[j],tempBuffer,&bufferSize);
- if (ret != MIC_ACCESS_API_SUCCESS) {
- ERROR("Problem getting Temperature(%d) %s",j,MicGetErrorString(ret));
- error=4;
- break;
- }
- /*INFO("Temp Read: %u: %u %s",j,tempBuffer[0],thermNames[j]);*/
- mic_submit_temp(i,thermNames[j],tempBuffer[0]);
- }
- if (error)
- break;
-
- /*Gather CPU Utilization Information */
- bufferSize=MAX_CORES*sizeof(MicCoreJiff);
- ret = MicGetCoreUtilization(micHandle,&coreUtil,coreJiffs,&bufferSize);
- if (ret != MIC_ACCESS_API_SUCCESS) {
- ERROR("Problem getting CPU utilization: %s",MicGetErrorString(ret));
- error=5;
- break;
- }
- mic_submit_cpu(i,"user",-1,coreUtil.sum.user);
- mic_submit_cpu(i,"sys",-1,coreUtil.sum.sys);
- mic_submit_cpu(i,"nice",-1,coreUtil.sum.nice);
- mic_submit_cpu(i,"idle",-1,coreUtil.sum.idle);
- for (j=0;j<coreUtil.core;j++) {
- mic_submit_cpu(i,"user",j,coreJiffs[j].user);
- mic_submit_cpu(i,"sys",j,coreJiffs[j].sys);
- mic_submit_cpu(i,"nice",j,coreJiffs[j].nice);
- mic_submit_cpu(i,"idle",j,coreJiffs[j].idle);
- }
-
- ret = MicCloseAdapter(micHandle);
- if (ret != MIC_ACCESS_API_SUCCESS) {
- ERROR("Problem initializing MicAdapter: %s",MicGetErrorString(ret));
- error=2;
- break;
- }
+
+ error = 0;
+ for (int i = 0; i < num_mics; i++) {
+ ret = MicInitAdapter(&mic_handle, &mics[i]);
+ if (ret != MIC_ACCESS_API_SUCCESS) {
+ ERROR("mic plugin: Problem initializing MicAdapter: %s",
+ MicGetErrorString(ret));
+ error = 1;
+ }
+
+ if (error == 0 && show_memory)
+ error = mic_read_memory(i);
+
+ if (error == 0 && show_temps)
+ error = mic_read_temps(i);
+
+ if (error == 0 && (show_cpu || show_cpu_cores))
+ error = mic_read_cpu(i);
+
+ if (error == 0 && (show_power))
+ error = mic_read_power(i);
+
+ ret = MicCloseAdapter(mic_handle);
+ if (ret != MIC_ACCESS_API_SUCCESS) {
+ ERROR("mic plugin: Problem closing MicAdapter: %s",
+ MicGetErrorString(ret));
+ error = 2;
+ break;
+ }
}
+ if (num_mics == 0)
+ error = 3;
return error;
}
+static int mic_shutdown(void) {
+ if (mic_handle)
+ MicCloseAPI(&mic_handle);
+ mic_handle = NULL;
-static int mic_shutdown (void)
-{
- if (micHandle)
- MicCloseAPI(&micHandle);
return (0);
}
-void module_register (void)
-{
- plugin_register_init ("mic", mic_init);
- plugin_register_shutdown ("mic", mic_shutdown);
- plugin_register_read ("mic", mic_read);
+void module_register(void) {
+ plugin_register_init("mic", mic_init);
+ plugin_register_shutdown("mic", mic_shutdown);
+ plugin_register_read("mic", mic_read);
+ plugin_register_config("mic", mic_config, config_keys, config_keys_num);
} /* void module_register */
/*
- * vim: shiftwidth=2:softtabstop=2:textwidth=78
+ * vim: set shiftwidth=8 softtabstop=8 noet textwidth=78 :
*/