Fixed a latent memory allocation issue with NVML.

Makefile

@@ -1,6 +1,6 @@
 .PHONY : clean debug
 
-OBJS = evga-icx.o evga-card.o icx3.o
+OBJS = evga-icx.o evga-card.o icx3.o board-sensors.o
 LDLIBS = -li2c
 CFLAGS = -MD
 
@@ -18,7 +18,7 @@ endif
 
 evga-icx : $(OBJS)
 
-debug : CFLAGS += -g -O0
+debug : CFLAGS += -g -Og
 debug : evga-icx
 
 clean :

README.md

@@ -14,6 +14,8 @@ A supported EVGA 30-series card with iCX3.  This includes:
 
 The number of fans supported depends, of course, on your particular model.
 
+You must have the `i2c-dev` kernel module loaded with `modprobe i2c-dev`
+
 Access to the `/dev/i2c` device files, which means either:
 * Run as root, or
 * Install udev rules to allow user access. If you have the OpenRGB udev rules installed to control the LEDs you already have this set up.
@@ -39,7 +41,7 @@ Note that when controlling fans directly through iCX3 they will fall offline fro
 
 ```text
 Available options:
---i2c N      : Only probe I2C bus N instead of all (may help with stuttering with --watch)
+--i2c N      : Only probe I2C bus N instead of all (may help with stuttering or freezing when probing I2C devices)
 --gpu N      : Control only GPU N instead of all supported cards
 --fan SPEED  : Set all fans at once to SPEED (see below)
 --fanN SPEED : Set fan N (0-3) to SPEED
@@ -53,6 +55,7 @@ Available options:
 --watch N    : Keep printing output every N seconds
 --overwrite  : Overwrite previously displayed info with --watch and --compact instead of continuously logging new lines
 --color      : Print headers in color in --compact mode for better readability
+--no-reasons : Do not query NVML for clock reasons (can cause stuttering)
 ```
 
 ### Examples:

board-sensors.c

@@ -0,0 +1,122 @@
+#include <dirent.h>
+#include <string.h>
+#include <stdlib.h>
+
+#include "board-sensors.h"
+
+int find_board_sensors(struct hwmon_avail_sensor *board_sensors, int max_sensors)
+{
+	const char *hwmon_path = "/sys/class/hwmon/";
+	char device_path[NAME_MAX];
+	char sensor_path[NAME_MAX];
+	
+	char driver_name[256];
+	
+	FILE *file;
+	DIR *dir;
+	struct dirent *ent;
+	
+	int num_sensors = 0;
+	
+	/* Start looking for hwmon devices in /sys/class/hwmon/ */
+	dir = opendir(hwmon_path);
+	
+	/* make sure we can open the device directory */
+	if(dir == NULL)
+		return 0;
+	
+	/* loop over all hwmon devices */
+	while((ent = readdir(dir)) != NULL)
+	{
+		/* Ignore any non-hwmon dirs */
+		if(strncmp(ent->d_name, "hwmon", 5) != 0) 
+			continue;
+		
+		strcpy(device_path, hwmon_path);
+		strcat(device_path, ent->d_name);
+		
+		/* Read in the name of the device */
+		strcpy(sensor_path, device_path);
+		strcat(sensor_path, "/name");
+		
+		file = fopen(sensor_path, "r");
+		if (file == NULL)
+			continue;
+		
+		if (fgets(driver_name, sizeof(driver_name), file) == NULL) {
+			fclose(file);
+			continue;
+		}
+		
+		fclose(file);
+
+		/* Driver names have a linebreak at the end so let's remove that for comparison*/
+		driver_name[strlen(driver_name) - 1] = '\0';
+		
+		/* Loop through all supported sensors and see if any are present in this device */
+		for (int i = 0; i < (sizeof(hwmon_sensor_info) / sizeof(struct hwmon_sensor)); i++) {
+			if (strcmp(driver_name, hwmon_sensor_info[i].driver_name) == 0) {
+				/* We matched the driver name, try to open the files */
+				strcpy(sensor_path, device_path);
+				strcat(sensor_path, "/");
+				strcat(sensor_path, hwmon_sensor_info[i].sensor_file_name);
+				strcat(sensor_path, "_input");
+				
+				file = fopen(sensor_path, "r");
+				if (file != NULL) {
+					fclose(file);
+					/* Good open of the sensor file */
+					board_sensors[num_sensors].file = calloc(NAME_MAX, sizeof(char));
+					strcpy(board_sensors[num_sensors].file, sensor_path);
+					board_sensors[num_sensors].sort_index = i;
+					board_sensors[num_sensors].sensor_info = &hwmon_sensor_info[i];
+					
+					/* Read in the sensor name */
+					board_sensors[num_sensors].sensor_name = calloc(MAX_SENSOR_NAME_LENGTH, sizeof(char));
+					strcpy(sensor_path, device_path);
+					strcat(sensor_path, "/");
+					strcat(sensor_path, hwmon_sensor_info[i].sensor_file_name);
+					strcat(sensor_path, "_label");
+					
+					file = fopen(sensor_path, "r");
+					if (file != NULL)
+						fgets(board_sensors[num_sensors].sensor_name, MAX_SENSOR_NAME_LENGTH, file);
+
+					/* Sensor name seems to always have a trailing newline we don't want */
+					size_t len_without_newline = strcspn(board_sensors[num_sensors].sensor_name, "\n");
+					board_sensors[num_sensors].sensor_name[len_without_newline] = '\0';
+					
+					if (num_sensors == max_sensors)
+						return num_sensors;
+					num_sensors++;
+				}
+			}
+		}
+		
+	}
+	
+	return num_sensors;
+}
+
+/* Returns 0 on a bad read or missing sensor, 1 on OK */
+int get_sensor_reading(struct hwmon_avail_sensor *sensor, float *reading) {
+	char buf[256] = {0};
+	long int raw;
+	FILE *file;
+	
+	file = fopen(sensor->file, "r");
+	if (file == NULL)
+		return 0;
+
+	fgets(buf, 256, file);
+	raw = strtol(buf, NULL, 10);
+	
+	fclose(file);
+	
+	*reading = (float)raw / sensor->sensor_info->divisor;
+
+	if (*reading == sensor->sensor_info->bad_value)
+		return 0;
+	
+	return 1;
+}

board-sensors.h

@@ -0,0 +1,45 @@
+#include <stdio.h>
+
+#define MAX_SENSOR_NAME_LENGTH 256
+
+struct hwmon_sensor {
+	char *driver_name;				/* Contents of /sys/class/hwmon/hwmonX/name */
+	char *sensor_file_name;			/* Sysfs file to read */
+	char *name_prefix;				/* Prefix to attach to temp*_label for clarity */
+	char *short_name;				/* 'Category' name when using compact mode */
+	char *units;					/* Units string */
+	float divisor;					/* Divisor to convert temp* to units */
+	float bad_value;	/* Raw value that indicates a bad (missing) sensor TODO: verfiy most of these*/
+};
+
+/* Note the order here matters, it's the order these will be printed in */
+static struct hwmon_sensor hwmon_sensor_info[] = 
+{
+	{"zenpower",	"temp1",	"CPU ",			"CPU",	"°C",	1000.0,	-40.0},	/* Tdie */
+	{"asusec",		"temp2",	"Motherboard ",	"CPU",	"°C",	1000.0,	-40.0},	/* CPU */
+	{"zenpower",	"temp3",	"CPU ",			"CCD",	"°C",	1000.0,	-40.0},	/* Tccd1 */
+	{"zenpower",	"temp4",	"CPU ",			"CCD",	"°C",	1000.0,	-40.0},	/* Tccd2 */
+	{"zenpower",	"temp5",	"CPU ",			"CCD",	"°C",	1000.0,	-40.0},	/* Tccd3 */
+	{"zenpower",	"temp6",	"CPU ",			"CCD",	"°C",	1000.0,	-40.0},	/* Tccd4 */
+	{"zenpower",	"temp7",	"CPU ",			"CCD",	"°C",	1000.0,	-40.0},	/* Tccd5 */
+	{"zenpower",	"temp8",	"CPU ",			"CCD",	"°C",	1000.0,	-40.0},	/* Tccd6 */
+	{"zenpower",	"temp9",	"CPU ",			"CCD",	"°C",	1000.0,	-40.0},	/* Tccd7 */
+	{"zenpower",	"temp10",	"CPU ",			"CCD",	"°C",	1000.0,	-40.0},	/* Tccd8 */
+	{"asusec",		"temp1",	"Motherboard ",	"CHIP",	"°C",	1000.0,	-40.0},	/* Chipset */
+	{"asusec",		"temp5",	"Motherboard ",	"VRM",	"°C",	1000.0,	-40.0},	/* VRM */
+	{"asusec",		"temp3",	"",				"MOBO",	"°C",	1000.0,	-40.0},	/* Motherboard */
+	{"asusec",		"temp4",	"Motherboard ",	"SENS",	"°C",	1000.0,	-40.0},	/* T_Sensor */
+	{"asusec",		"temp6",	"Motherboard ",	"H2O",	"°C",	1000.0,	-40.0},	/* Water_In */
+	{"asusec",		"temp7",	"Motherboard ",	"H2O",	"°C",	1000.0,	-40.0},	/* Water_Out */
+	{"nvme",		"temp1",	"NVMe ",		"NVME",	"°C",	1000.0,	-40.0},	/* NVME Composite */
+};
+
+struct hwmon_avail_sensor {
+	char *sensor_name;					/* Sensor name as read from the sysfs file */
+	int sort_index;						/* Sort index for order to display in */
+	char *file;							/* File to read from */
+	struct hwmon_sensor *sensor_info;	/* Associated sensor info struct */
+};
+
+int find_board_sensors(struct hwmon_avail_sensor *board_sensors, int max_sensors);
+int get_sensor_reading(struct hwmon_avail_sensor *sensor, float *reading);

evga-card.h

@@ -91,6 +91,9 @@ struct card_info {
 	int i2c_fd;							/* File descriptor for the i2c device file, for re-use */
 	int product_id;						/* EVGA internal product ID, as reported by the iCX3 controller */
 	unsigned int bar0;					/* Address of the card's PCI base address register */
+	void *nvml_device;					/* Pointer to nvmlDevice_t for use in NVML calls */
+	void *vram_addr;					/* Memory mapping for GDDR6 temps */
+	void *hotspot_addr;					/* Memory mapping for hotspot temperature */
 };
 
 struct gpu_pci_info {

evga-icx.c

@@ -15,8 +15,10 @@
 
 #include "icx3.h"
 #include "evga-card.h"
+#include "board-sensors.h"
 
 #define MAX_GPUS 16
+#define MAX_BOARD_SENSORS 256
 #define HEADER_COLOR_START "\x1b[36m"
 #define HEADER_COLOR_END "\x1b[39m"
 
@@ -24,7 +26,7 @@ char *header_start = "";
 char *header_end = "";
 
 static const char helpstring[] = "Available options:\n"
-								 "--i2c N      : Only probe I2C bus N instead of all (may help with stuttering with --watch)\n"
+								 "--i2c N      : Only probe I2C bus N instead of all (may help with stuttering or freezing when probing I2C devices)\n"
 								 "--gpu N      : Control only GPU N instead of all supported cards\n"
 								 "--fan SPEED  : Set all fans at once to SPEED (see below)\n"
 								 "--fanN SPEED : Set fan N (0-3) to SPEED\n"
@@ -37,20 +39,26 @@ static const char helpstring[] = "Available options:\n"
 								 "--compact    : Print sensor reading in a compact one-line per card format\n"
 								 "--watch N    : Keep printing output every N seconds\n"
 								 "--overwrite  : Overwrite previously displayed info with --watch and --compact instead of continuously logging\n"
-								 "--color      : Print headers in color in --compact mode for better readability\n";
+								 "--color      : Print headers in color in --compact mode for better readability\n"
+								 "--no-reasons : Do not query NVML for clock reasons (can cause stuttering)\n"
+								 "--board      : Also print temperatures from the CPU, motherboard, and other sensors";
 								 
-void print_gpu_info(int gpu_num, struct card_info gpus[], int compact);
+void print_gpu_info(int gpu_num, struct card_info gpus[], int compact, int no_reasons);
+void print_board_info(struct hwmon_avail_sensor *board_sensors, int num_sensors, int compact);
 
 int main (int argc, char **argv)
 {
 	struct card_info gpus[MAX_GPUS];
-	int gpu_count;
+	struct hwmon_avail_sensor board_sensors[MAX_BOARD_SENSORS];
+	int gpu_count, board_sensor_count;
 	int print_info = 0;
 	int compact = 0;			
 	int gpu_num = -1;				/* Card to control */
-	int i2c_bus = -1;
-	int overwrite = 0;
-	unsigned int watch = 0;
+	int i2c_bus = -1;				/* Specific i2c bus to probe instead of all */
+	int overwrite = 0;				/* Overwrite printed console info in compact mode */
+	int no_reasons = 0;				/* Don't probe or display NVML clock reasons */
+	unsigned int watch = 0;			/* Refresh display every this many seconds */
+	int print_board_sensors = 0;	/* Print CPU/motherbord/other sensors as well */
 	char *fan_speed[ICX3_MAX_FANS] = {NULL}; 
 	
 	/* Input parsing */
@@ -110,6 +118,10 @@ int main (int argc, char **argv)
 		} else if (strcmp(argv[i], "--color") == 0) {
 			header_start = HEADER_COLOR_START;
 			header_end = HEADER_COLOR_END;
+		} else if (strcmp(argv[i], "--no-reasons") == 0) {
+			no_reasons = 1;
+		} else if (strcmp(argv[i], "--board") == 0) {
+			print_board_sensors = 1;
 		} else {
 			printf(helpstring);
 			return 0;
@@ -129,8 +141,10 @@ int main (int argc, char **argv)
 	if (overwrite && !compact)
 		overwrite = 0;
 	
+	/* Scan for supported GPUs */
 	gpu_count = find_evga_gpu_i2cs(gpus, MAX_GPUS, i2c_bus);
 
+	/* Check for no GPUs found or other errors */
 	if (gpu_count == -1) {
 		printf("Error scanning I2C devices\n");
 		return -1;
@@ -138,12 +152,15 @@ int main (int argc, char **argv)
 		printf("No supported GPUs found.\nAre you root or do you have udev access to i2c devices?\nDo you need to run `modprobe i2c-dev`?\n");
 		return -1;
 	}
-	
+	/* Check for invalid GPUs */
 	if (gpu_num > gpu_count - 1) {
 		printf("Invalid GPU number specified (%d, max %d)\n", gpu_num, gpu_count - 1);
 		return -1;
 	}
 		
+	/* Scan for motherboard/CPU/other sensors */
+	if (print_board_sensors)
+		board_sensor_count = find_board_sensors(board_sensors, MAX_BOARD_SENSORS);
 	
 	/* execute fan commands */
 	if (gpu_num == -1) {
@@ -163,6 +180,10 @@ int main (int argc, char **argv)
 	/* NVML init */
 #ifdef USE_NVML
     init_nvml();
+	for (int i = 0; i < gpu_count; i++) {
+		gpus[i].nvml_device = malloc(sizeof(nvmlDevice_t));
+		get_nvml_handle(&gpus[i]);
+	}
 #endif
 
 	/* PCI init for VRAM/hotspot temps */
@@ -177,15 +198,18 @@ int main (int argc, char **argv)
 			if (overwrite)
 				printf("\x1b[K");	/* Clear current console line */
 			
+			if (print_board_sensors)
+				print_board_info(board_sensors, board_sensor_count, compact);
+
 			if (gpu_num == -1) {
 				/* No GPU specified on command line, loop over all supported GPUs */
 				for (int i = 0; i < gpu_count; i++){
 					if (i > 0)
 						printf("\n");
-					print_gpu_info(i, &gpus[i], compact);
+					print_gpu_info(i, &gpus[i], compact, no_reasons);
 				}		
 			} else if (gpu_num <= gpu_count - 1) {
-				print_gpu_info(gpu_num, &gpus[gpu_num], compact);
+				print_gpu_info(gpu_num, &gpus[gpu_num], compact, no_reasons);
 			}
 
 			if (!overwrite)
@@ -193,8 +217,8 @@ int main (int argc, char **argv)
 			
 			if (overwrite && compact) {
 				printf("\x1b[1G");						/* Move cursor back to column 1 */
-				if (gpu_count > 1)
-					printf("\x1b[%dA", gpu_count-1);	/* Move cursor back up to the top of gpu list */
+				if (gpu_count > 1 || print_board_sensors)
+					printf("\x1b[%dA", gpu_count-1+print_board_sensors);	/* Move cursor back up to the top of gpu list */
 			}
 			
 			fflush(stdout);
@@ -207,7 +231,64 @@ int main (int argc, char **argv)
 #endif
 }
 
-void print_gpu_info(int gpu_num, struct card_info *gpu, int compact) {
+void print_board_info(struct hwmon_avail_sensor *board_sensors, int num_sensors, int compact)
+{
+	int printed_sensors = 0;
+	int current_sort_index = 0;
+	float current_reading = 0.0;
+	int good_reading = 0;
+	/* These allow us to 'summarize' units and categories by only printing them when they change */
+	char *last_short_name = NULL;
+	char *last_units = NULL;
+	
+	while (printed_sensors < num_sensors) {
+		for (int i=0; i < num_sensors; i++){
+			/* Loop over all sensors, but only output those with the current sort index so they come out sort of sorted
+			Duplicates (e.g.) multiple NVMe will come out in whatever sort of order the directory listing happened to */
+			if (board_sensors[i].sort_index == current_sort_index) {
+				printed_sensors++;
+				good_reading = get_sensor_reading(&board_sensors[i], &current_reading);
+				
+				if (!good_reading)
+					continue;
+				
+				if (compact) {
+					/* Print units if needed */
+					if (last_units != NULL && strcmp(board_sensors[i].sensor_info->units, last_units))
+						printf("%s", last_units);
+					
+					/* Print new section header if needed */
+					if (last_short_name == NULL || strcmp(board_sensors[i].sensor_info->short_name, last_short_name)) {
+						if (last_short_name != NULL) /* Spacer for all headings not the first one */
+							printf(" ");
+						printf("%s%s%s", header_start, board_sensors[i].sensor_info->short_name, header_end);
+					}
+					printf(" %3.0f", current_reading);
+					
+					last_short_name = board_sensors[i].sensor_info->short_name;
+					last_units = board_sensors[i].sensor_info->units;
+					
+				} else {
+					printf("%s%s: %+.1f%s\n", 
+						board_sensors[i].sensor_info->name_prefix,
+						board_sensors[i].sensor_name,
+						current_reading,
+						board_sensors[i].sensor_info->units);
+				}
+				
+			}
+		}
+		current_sort_index++;
+	}
+	
+	if (compact && last_units != NULL)
+		printf("%s", last_units);
+	
+	printf("\n");
+}
+
+void print_gpu_info(int gpu_num, struct card_info *gpu, int compact, int no_reasons)
+{
 	if (compact) {
 		/* One line per GPU */
 		printf("%s#%d FAN%s", header_start, gpu_num, header_end);
@@ -239,8 +320,10 @@ void print_gpu_info(int gpu_num, struct card_info *gpu, int compact) {
 #ifdef USE_NVML
 		printf("%s MEM %s", header_start, header_end);
 		printf("%3d%%", get_nvml_mem_util(gpu));
+		if (!no_reasons) {
 			printf("%s CLK %s", header_start, header_end);
 			print_nvml_clock_reason(1, gpu);
+		}
 #endif
 
 
@@ -270,8 +353,10 @@ void print_gpu_info(int gpu_num, struct card_info *gpu, int compact) {
 
 #ifdef USE_NVML
 		printf("Mem util: %d%%\n", get_nvml_mem_util(gpu));
+		if (!no_reasons) {
 			printf("Clock reasons: ");
 			print_nvml_clock_reason(0, gpu);
+		}
 		printf("\n");
 #endif
 	}

gddr6.c

@@ -36,70 +36,69 @@ void init_gddr6(struct card_info *card)
 	card->bar0 = (pci_dev->base_addr[0] & 0xFFFFFFFF);
 
 	pci_cleanup(pacc);
+	
+	/* Open our memory mappings */
+	card->vram_addr = NULL;
+	card->hotspot_addr = NULL;
+	
+	int fd;
+	if ((fd = open("/dev/mem", O_RDWR | O_SYNC)) == -1)
+    {
+        printf("Can't read memory for VRAM and Hotspot temperatures. If you are root, enable kernel parameter iomem=relaxed\n");
+		return;
+    }
+	
+	unsigned int phys_addr, base_offset;
+	void *map_base;
+	
+	for (int i = 0; i < sizeof(device_offset_info) / sizeof(struct device_offset); i++) {
+		if (card->pci_device_id == device_offset_info[i].device_id){
+			/* Map for VRAM */
+			phys_addr = (card->bar0 + device_offset_info[i].vram_offset);
+			base_offset = phys_addr & ~(PG_SZ-1);
+			map_base = mmap(0, PG_SZ, PROT_READ, MAP_SHARED, fd, base_offset);
+			if(map_base == (void *) -1)
+				printf("Can't map memory for VRAM temperature. If you are root, enable kernel parameter iomem=relaxed\n");
+			else
+				card->vram_addr = (void *) map_base + (phys_addr - base_offset);
+			
+			
+			/* Map for hotspot */
+			phys_addr = (card->bar0 + device_offset_info[i].hotspot_offset);
+			base_offset = phys_addr & ~(PG_SZ-1);
+			map_base = mmap(0, PG_SZ, PROT_READ, MAP_SHARED, fd, base_offset);
+			if(map_base == (void *) -1)
+				printf("Can't map memory for Hotspot temperature. If you are root, enable kernel parameter iomem=relaxed\n");
+			else
+				card->hotspot_addr = (void *) map_base + (phys_addr - base_offset);
+		}
+	}
+	
+	close(fd);
 }
 
 float get_vram_temp(struct card_info *card)
 {
-	int fd;
 	float temp = 0.0;
 	
-	if ((fd = open("/dev/mem", O_RDWR | O_SYNC)) == -1)
-    {
-        printf("Can't read memory. If you are root, enable kernel parameter iomem=relaxed\n");
-		return 0;
-    }
+	if(card->vram_addr == NULL)
+		return 0.0;
 
-	for (int i = 0; i < sizeof(device_offset_info) / sizeof(struct device_offset); i++) {
-		if (card->pci_device_id == device_offset_info[i].device_id){
-			unsigned int phys_addr = (card->bar0 + device_offset_info[i].vram_offset);
-			unsigned int base_offset = phys_addr & ~(PG_SZ-1);
-			void *map_base = mmap(0, PG_SZ, PROT_READ, MAP_SHARED, fd, base_offset);
-			if(map_base == (void *) -1)
-			{
-				if (fd != -1)
-					close(fd);
-				printf("Can't read memory for VRAM temperature. If you are root, enable kernel parameter iomem=relaxed\n");
-			}
-			void *virt_addr = (char *) map_base + (phys_addr - base_offset);
-			int read_result = *((unsigned int *) virt_addr);
+	int read_result = *((unsigned int *) card->vram_addr);
 	temp = ((read_result & 0x00000fff) / 0x20);
-			munmap(map_base, PG_SZ);
-		}
-	}
 
-	close(fd);
 	return temp;
 }
 
 float get_hotspot_temp(struct card_info *card)
 {
-	int fd;
 	float temp = 0.0;
 	
-	if ((fd = open("/dev/mem", O_RDWR | O_SYNC)) == -1)
-    {
-        printf("Can't read memory. If you are root, enable kernel parameter iomem=relaxed\n");
-		return 0;
-    }
+	if(card->hotspot_addr == NULL)
+		return 0.0;
 
-	for (int i = 0; i < sizeof(device_offset_info) / sizeof(struct device_offset); i++) {
-		if (card->pci_device_id == device_offset_info[i].device_id){
-			unsigned int phys_addr = (card->bar0 + device_offset_info[i].hotspot_offset);
-			unsigned int base_offset = phys_addr & ~(PG_SZ-1);
-			void *map_base = mmap(0, PG_SZ, PROT_READ, MAP_SHARED, fd, base_offset);
-			if(map_base == (void *) -1)
-			{
-				if (fd != -1)
-					close(fd);
-				printf("Can't read memory for hotspot. If you are root, enable kernel parameter iomem=relaxed\n");
-			}
-			void *virt_addr = (char *) map_base + (phys_addr - base_offset);
-			int read_result = *((unsigned int *) virt_addr);
+	int read_result = *((unsigned int *) card->hotspot_addr);
 	temp = (read_result >> 8) & 0xff;
-			munmap(map_base, PG_SZ);
-		}
-	}
 
-	close(fd);
 	return temp;
 }

icx3.c

@@ -237,6 +237,6 @@ void set_fan(int fan, char *setting, struct card_info *card)
 	if (fan_readback.fanmode    != fan_control.fanmode ||
 		fan_readback.rpm_offset != fan_control.rpm_offset ||
 		fan_readback.duty       != fan_control.duty)
-		printf("Error setting fan %d on %s\n", fan, card->i2c_fd);
+		printf("Error setting fan %d on %s\n", fan, card->i2c_dev_path);
 }
 

nvidia-sensors.c

@@ -10,15 +10,14 @@ void init_nvml()
 		printf("Could not init NVML: %s\n", nvmlErrorString(result));
 }
 
-int get_nvml_handle(nvmlDevice_t *device, struct card_info *card)
+void get_nvml_handle(struct card_info *card)
 {
 	nvmlReturn_t result;
-	result = nvmlDeviceGetHandleByPciBusId_v2(card->pci_id, device);
+	result = nvmlDeviceGetHandleByPciBusId_v2(card->pci_id, card->nvml_device);
 	if (result != NVML_SUCCESS) {
-		printf("Failed to get device handle for card at %s: %s\n", card->pci_id, nvmlErrorString(result));
-		return 0;
+		printf("Failed to get NVML device handle for card at %s: %s\n", card->pci_id, nvmlErrorString(result));
+		card->nvml_device = NULL;
 	}
-	return 1;
 }
 
 void print_nvml_clock_reason(int compact, struct card_info *card)
@@ -52,12 +51,11 @@ void print_nvml_clock_reason(int compact, struct card_info *card)
 
 unsigned int get_nvml_temp(struct card_info *card)
 {
-	nvmlDevice_t nvml_device;
-	if (!get_nvml_handle(&nvml_device, card))
+	if (card->nvml_device == NULL)
 		return 0;
 	
 	unsigned int temp;
-	nvmlReturn_t result = nvmlDeviceGetTemperature(nvml_device, NVML_TEMPERATURE_GPU, &temp); 
+	nvmlReturn_t result = nvmlDeviceGetTemperature(*(nvmlDevice_t*)(card->nvml_device), NVML_TEMPERATURE_GPU, &temp); 
 	if (result != NVML_SUCCESS) {
 		printf("Failed to get temperature for card at %s: %s\n", card->pci_id, nvmlErrorString(result));
 		return 0;
@@ -67,12 +65,11 @@ unsigned int get_nvml_temp(struct card_info *card)
 
 unsigned long long get_nvml_clock_reasons(struct card_info *card)
 {
-	nvmlDevice_t nvml_device;
-	if (!get_nvml_handle(&nvml_device, card))
+	if (card->nvml_device == NULL)
 		return 0;
 	
 	unsigned long long reasons;
-	nvmlReturn_t result = nvmlDeviceGetCurrentClocksEventReasons(nvml_device, &reasons) ; 
+	nvmlReturn_t result = nvmlDeviceGetCurrentClocksEventReasons(*(nvmlDevice_t*)(card->nvml_device), &reasons) ; 
 	if (result != NVML_SUCCESS) {
 		printf("Failed to get clock reasons for card at %s: %s\n", card->pci_id, nvmlErrorString(result));
 		return 0;
@@ -83,12 +80,11 @@ unsigned long long get_nvml_clock_reasons(struct card_info *card)
 
 unsigned int get_nvml_mem_util(struct card_info *card)
 {
-	nvmlDevice_t nvml_device;
-	if (!get_nvml_handle(&nvml_device, card))
+	if (card->nvml_device == NULL)
 		return 0;
 	
 	nvmlUtilization_t util;
-	nvmlReturn_t result = nvmlDeviceGetUtilizationRates(nvml_device, &util);
+	nvmlReturn_t result = nvmlDeviceGetUtilizationRates(*(nvmlDevice_t*)(card->nvml_device), &util);
 	if (result != NVML_SUCCESS) {
 		printf("Failed to get clock reasons for card at %s: %s\n", card->pci_id, nvmlErrorString(result));
 		return 0;

nvidia-sensors.h

@@ -22,7 +22,7 @@ static struct clock_reason clock_reason_names[] =
 };
 
 void init_nvml();
-int get_nvml_handle(nvmlDevice_t *device, struct card_info *card);
+void get_nvml_handle(struct card_info *card);
 void print_nvml_clock_reason(int compact, struct card_info *card);
 unsigned int get_nvml_temp(struct card_info *card);
 unsigned long long get_nvml_clock_reasons(struct card_info *card);