--- /dev/null
+/*
+Copyright 2018 Evgeny Naumov
+
+Permission is hereby granted, free of charge, to any person obtaining a copy of
+this software and associated documentation files (the "Software"), to deal in
+the Software without restriction, including without limitation the rights to
+use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies
+of the Software, and to permit persons to whom the Software is furnished to do
+so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.
+*/
+
+#include "daemon/collectd.h"
+#include "daemon/common.h"
+#include "daemon/plugin.h"
+
+#include <nvml.h>
+#include <stdint.h>
+#include <stdio.h>
+
+#define MAX_DEVNAME_LEN 256
+#define PLUGIN_NAME "gpu_nvidia"
+
+static nvmlReturn_t nv_status = NVML_SUCCESS;
+static char *nv_errline = "";
+
+#define TRY_CATCH(f, catch) \
+ if ((nv_status = f) != NVML_SUCCESS) { \
+ nv_errline = #f; \
+ goto catch; \
+ }
+
+#define TRY_CATCH_OPTIONAL(f, catch) \
+ if ((nv_status = f) != NVML_SUCCESS && \
+ nv_status != NVML_ERROR_NOT_SUPPORTED) { \
+ nv_errline = #f; \
+ goto catch; \
+ }
+
+#define TRY(f) TRY_CATCH(f, catch)
+#define TRYOPT(f) TRY_CATCH_OPTIONAL(f, catch)
+
+#define KEY_GPUINDEX "GPUIndex"
+#define KEY_IGNORESELECTED "IgnoreSelected"
+
+static const char *config_keys[] = {
+ KEY_GPUINDEX, KEY_IGNORESELECTED,
+};
+static const unsigned int n_config_keys = STATIC_ARRAY_SIZE(config_keys);
+
+// This is a bitflag, necessitating the (extremely conservative) assumption
+// that there are no more than 64 GPUs on this system.
+static uint64_t conf_match_mask = 0;
+static bool conf_mask_is_exclude = 0;
+
+static int nvml_config(const char *key, const char *value) {
+
+ if (strcasecmp(key, KEY_GPUINDEX) == 0) {
+ char *eptr;
+ unsigned long device_ix = strtoul(value, &eptr, 10);
+ if (eptr == value) {
+ ERROR(PLUGIN_NAME ": Failed to parse GPUIndex value \"%s\"", value);
+ return -1;
+ }
+ if (device_ix >= 64) {
+ ERROR(PLUGIN_NAME
+ ": At most 64 GPUs (0 <= GPUIndex < 64) are supported!");
+ return -2;
+ }
+ conf_match_mask |= (1 << device_ix);
+ } else if (strcasecmp(key, KEY_IGNORESELECTED)) {
+ conf_mask_is_exclude = IS_TRUE(value);
+ } else {
+ ERROR(PLUGIN_NAME ": Unrecognized config option %s", key);
+ return -10;
+ }
+
+ return 0;
+}
+
+static int nvml_init(void) {
+ TRY(nvmlInit());
+ return 0;
+
+ catch : ERROR(PLUGIN_NAME ": NVML init failed with %d", nv_status);
+ return -1;
+}
+
+static int nvml_shutdown(void) {
+ TRY(nvmlShutdown())
+ return 0;
+
+ catch : ERROR(PLUGIN_NAME ": NVML shutdown failed with %d", nv_status);
+ return -1;
+}
+
+static void nvml_submit_gauge(const char *plugin_instance, const char *type,
+ const char *type_instance, gauge_t nvml) {
+
+ value_list_t vl = VALUE_LIST_INIT;
+
+ vl.values = &(value_t){.gauge = nvml};
+ vl.values_len = 1;
+
+ sstrncpy(vl.plugin, PLUGIN_NAME, sizeof(vl.plugin));
+ sstrncpy(vl.plugin_instance, plugin_instance, sizeof(vl.plugin_instance));
+
+ 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 int nvml_read(void) {
+
+ unsigned int device_count;
+ TRY_CATCH(nvmlDeviceGetCount(&device_count), catch_nocount);
+
+ if (device_count > 64) {
+ device_count = 64;
+ }
+
+ for (unsigned int ix = 0; ix < device_count; ix++) {
+
+ unsigned int is_match =
+ ((1 << ix) & conf_match_mask) || (conf_match_mask == 0);
+ if (conf_mask_is_exclude == !!is_match) {
+ continue;
+ }
+
+ nvmlDevice_t dev;
+ TRY(nvmlDeviceGetHandleByIndex(ix, &dev));
+
+ char dev_name[MAX_DEVNAME_LEN + 1] = {0};
+ TRY(nvmlDeviceGetName(dev, dev_name, sizeof(dev_name) - 1));
+
+ // Try to be as lenient as possible with the variety of devices that are
+ // out there, ignoring any NOT_SUPPORTED errors gently.
+ nvmlMemory_t meminfo;
+ TRYOPT(nvmlDeviceGetMemoryInfo(dev, &meminfo))
+ if (nv_status == NVML_SUCCESS) {
+ nvml_submit_gauge(dev_name, "memory", "used", meminfo.used);
+ nvml_submit_gauge(dev_name, "memory", "free", meminfo.free);
+ }
+
+ nvmlUtilization_t utilization;
+ TRYOPT(nvmlDeviceGetUtilizationRates(dev, &utilization))
+ if (nv_status == NVML_SUCCESS)
+ nvml_submit_gauge(dev_name, "percent", "gpu_used", utilization.gpu);
+
+ unsigned int fan_speed;
+ TRYOPT(nvmlDeviceGetFanSpeed(dev, &fan_speed))
+ if (nv_status == NVML_SUCCESS)
+ nvml_submit_gauge(dev_name, "fanspeed", NULL, fan_speed);
+
+ unsigned int core_temp;
+ TRYOPT(nvmlDeviceGetTemperature(dev, NVML_TEMPERATURE_GPU, &core_temp))
+ if (nv_status == NVML_SUCCESS)
+ nvml_submit_gauge(dev_name, "temperature", "core", core_temp);
+
+ unsigned int sm_clk_mhz;
+ TRYOPT(nvmlDeviceGetClockInfo(dev, NVML_CLOCK_SM, &sm_clk_mhz))
+ if (nv_status == NVML_SUCCESS)
+ nvml_submit_gauge(dev_name, "frequency", "multiprocessor",
+ 1e6 * sm_clk_mhz);
+
+ unsigned int mem_clk_mhz;
+ TRYOPT(nvmlDeviceGetClockInfo(dev, NVML_CLOCK_MEM, &mem_clk_mhz))
+ if (nv_status == NVML_SUCCESS)
+ nvml_submit_gauge(dev_name, "frequency", "memory", 1e6 * mem_clk_mhz);
+
+ unsigned int power_mW;
+ TRYOPT(nvmlDeviceGetPowerUsage(dev, &power_mW))
+ if (nv_status == NVML_SUCCESS)
+ nvml_submit_gauge(dev_name, "power", NULL, 1e-3 * power_mW);
+
+ continue;
+
+ // Failures here indicate transient errors or removal of GPU. In either
+ // case it will either be resolved or the GPU will no longer be enumerated
+ // the next time round.
+ catch : WARNING(PLUGIN_NAME
+ ": NVML call \"%s\" failed (%d) on dev at index %d!",
+ nv_errline, nv_status, ix);
+ continue;
+ }
+
+ return 0;
+
+// Failures here indicate serious misconfiguration; we bail out totally.
+catch_nocount:
+ ERROR(PLUGIN_NAME ": Failed to enumerate NVIDIA GPUs (\"%s\" returned %d)",
+ nv_errline, nv_status);
+ return -1;
+}
+
+void module_register(void) {
+ plugin_register_init(PLUGIN_NAME, nvml_init);
+ plugin_register_config(PLUGIN_NAME, nvml_config, config_keys, n_config_keys);
+ plugin_register_read(PLUGIN_NAME, nvml_read);
+ plugin_register_shutdown(PLUGIN_NAME, nvml_shutdown);
+}
projects / collectd.git / blobdiff