// SPDX-License-Identifier: GPL-2.0/* * Memory bandwidth monitoring and allocation library * * Copyright (C) 2018 Intel Corporation * * Authors: *    Sai Praneeth Prakhya <sai.praneeth.prakhya@intel.com>, *    Fenghua Yu <fenghua.yu@intel.com> */#include "resctrl.h" #define UNCORE_IMC		"uncore_imc"#define READ_FILE_NAME		"events/cas_count_read"#define WRITE_FILE_NAME		"events/cas_count_write"#define DYN_PMU_PATH		"/sys/bus/event_source/devices"#define SCALE			0.00006103515625#define MAX_IMCS		20#define MAX_TOKENS		5#define READ			0#define WRITE			1 #define CON_MBM_LOCAL_BYTES_PATH		\	"%s/%s/mon_data/mon_L3_%02d/mbm_local_bytes" struct membw_read_format {	__u64 value;         /* The value of the event */	__u64 time_enabled;  /* if PERF_FORMAT_TOTAL_TIME_ENABLED */	__u64 time_running;  /* if PERF_FORMAT_TOTAL_TIME_RUNNING */	__u64 id;            /* if PERF_FORMAT_ID */}; struct imc_counter_config {	__u32 type;	__u64 event;	__u64 umask;	struct perf_event_attr pe;	struct membw_read_format return_value;	int fd;}; static char mbm_total_path[1024];static int imcs;static struct imc_counter_config imc_counters_config[MAX_IMCS][2];static const struct resctrl_test *current_test; void membw_initialize_perf_event_attr(int i, int j){	memset(&imc_counters_config[i][j].pe, 0,	       sizeof(struct perf_event_attr));	imc_counters_config[i][j].pe.type = imc_counters_config[i][j].type;	imc_counters_config[i][j].pe.size = sizeof(struct perf_event_attr);	imc_counters_config[i][j].pe.disabled = 1;	imc_counters_config[i][j].pe.inherit = 1;	imc_counters_config[i][j].pe.exclude_guest = 0;	imc_counters_config[i][j].pe.config =		imc_counters_config[i][j].umask << 8 |		imc_counters_config[i][j].event;	imc_counters_config[i][j].pe.sample_type = PERF_SAMPLE_IDENTIFIER;	imc_counters_config[i][j].pe.read_format =		PERF_FORMAT_TOTAL_TIME_ENABLED | PERF_FORMAT_TOTAL_TIME_RUNNING;} void membw_ioctl_perf_event_ioc_reset_enable(int i, int j){	ioctl(imc_counters_config[i][j].fd, PERF_EVENT_IOC_RESET, 0);	ioctl(imc_counters_config[i][j].fd, PERF_EVENT_IOC_ENABLE, 0);} void membw_ioctl_perf_event_ioc_disable(int i, int j){	ioctl(imc_counters_config[i][j].fd, PERF_EVENT_IOC_DISABLE, 0);} /* * get_event_and_umask:	Parse config into event and umask * @cas_count_cfg:	Config * @count:		iMC number * @op:			Operation (read/write) */void get_event_and_umask(char *cas_count_cfg, int count, bool op){	char *token[MAX_TOKENS];	int i = 0; 	strcat(cas_count_cfg, ",");	token[0] = strtok(cas_count_cfg, "=,"); 	for (i = 1; i < MAX_TOKENS; i++)		token[i] = strtok(NULL, "=,"); 	for (i = 0; i < MAX_TOKENS; i++) {		if (!token[i])			break;		if (strcmp(token[i], "event") == 0) {			if (op == READ)				imc_counters_config[count][READ].event =				strtol(token[i + 1], NULL, 16);			else				imc_counters_config[count][WRITE].event =				strtol(token[i + 1], NULL, 16);		}		if (strcmp(token[i], "umask") == 0) {			if (op == READ)				imc_counters_config[count][READ].umask =				strtol(token[i + 1], NULL, 16);			else				imc_counters_config[count][WRITE].umask =				strtol(token[i + 1], NULL, 16);		}	}} static int open_perf_event(int i, int cpu_no, int j){	imc_counters_config[i][j].fd =		perf_event_open(&imc_counters_config[i][j].pe, -1, cpu_no, -1,				PERF_FLAG_FD_CLOEXEC); 	if (imc_counters_config[i][j].fd == -1) {		fprintf(stderr, "Error opening leader %llx\n",			imc_counters_config[i][j].pe.config); 		return -1;	} 	return 0;} /* Get type and config (read and write) of an iMC counter */static int read_from_imc_dir(char *imc_dir, int count){	char cas_count_cfg[1024], imc_counter_cfg[1024], imc_counter_type[1024];	FILE *fp; 	/* Get type of iMC counter */	sprintf(imc_counter_type, "%s%s", imc_dir, "type");	fp = fopen(imc_counter_type, "r");	if (!fp) {		ksft_perror("Failed to open iMC counter type file"); 		return -1;	}	if (fscanf(fp, "%u", &imc_counters_config[count][READ].type) <= 0) {		ksft_perror("Could not get iMC type");		fclose(fp); 		return -1;	}	fclose(fp); 	imc_counters_config[count][WRITE].type =				imc_counters_config[count][READ].type; 	/* Get read config */	sprintf(imc_counter_cfg, "%s%s", imc_dir, READ_FILE_NAME);	fp = fopen(imc_counter_cfg, "r");	if (!fp) {		ksft_perror("Failed to open iMC config file"); 		return -1;	}	if (fscanf(fp, "%s", cas_count_cfg) <= 0) {		ksft_perror("Could not get iMC cas count read");		fclose(fp); 		return -1;	}	fclose(fp); 	get_event_and_umask(cas_count_cfg, count, READ); 	/* Get write config */	sprintf(imc_counter_cfg, "%s%s", imc_dir, WRITE_FILE_NAME);	fp = fopen(imc_counter_cfg, "r");	if (!fp) {		ksft_perror("Failed to open iMC config file"); 		return -1;	}	if  (fscanf(fp, "%s", cas_count_cfg) <= 0) {		ksft_perror("Could not get iMC cas count write");		fclose(fp); 		return -1;	}	fclose(fp); 	get_event_and_umask(cas_count_cfg, count, WRITE); 	return 0;} /* * A system can have 'n' number of iMC (Integrated Memory Controller) * counters, get that 'n'. For each iMC counter get it's type and config. * Also, each counter has two configs, one for read and the other for write. * A config again has two parts, event and umask. * Enumerate all these details into an array of structures. * * Return: >= 0 on success. < 0 on failure. */static int num_of_imcs(void){	char imc_dir[512], *temp;	unsigned int count = 0;	struct dirent *ep;	int ret;	DIR *dp; 	dp = opendir(DYN_PMU_PATH);	if (dp) {		while ((ep = readdir(dp))) {			temp = strstr(ep->d_name, UNCORE_IMC);			if (!temp)				continue; 			/*			 * imc counters are named as "uncore_imc_<n>", hence			 * increment the pointer to point to <n>. Note that			 * sizeof(UNCORE_IMC) would count for null character as			 * well and hence the last underscore character in			 * uncore_imc'_' need not be counted.			 */			temp = temp + sizeof(UNCORE_IMC); 			/*			 * Some directories under "DYN_PMU_PATH" could have			 * names like "uncore_imc_free_running", hence, check if			 * first character is a numerical digit or not.			 */			if (temp[0] >= '0' && temp[0] <= '9') {				sprintf(imc_dir, "%s/%s/", DYN_PMU_PATH,					ep->d_name);				ret = read_from_imc_dir(imc_dir, count);				if (ret) {					closedir(dp); 					return ret;				}				count++;			}		}		closedir(dp);		if (count == 0) {			ksft_print_msg("Unable to find iMC counters\n"); 			return -1;		}	} else {		ksft_perror("Unable to open PMU directory"); 		return -1;	} 	return count;} int initialize_mem_bw_imc(void){	int imc, j; 	imcs = num_of_imcs();	if (imcs <= 0)		return imcs; 	/* Initialize perf_event_attr structures for all iMC's */	for (imc = 0; imc < imcs; imc++) {		for (j = 0; j < 2; j++)			membw_initialize_perf_event_attr(imc, j);	} 	return 0;} static void perf_close_imc_mem_bw(void){	int mc; 	for (mc = 0; mc < imcs; mc++) {		if (imc_counters_config[mc][READ].fd != -1)			close(imc_counters_config[mc][READ].fd);		if (imc_counters_config[mc][WRITE].fd != -1)			close(imc_counters_config[mc][WRITE].fd);	}} /* * perf_open_imc_mem_bw - Open perf fds for IMCs * @cpu_no: CPU number that the benchmark PID is bound to * * Return: = 0 on success. < 0 on failure. */static int perf_open_imc_mem_bw(int cpu_no){	int imc, ret; 	for (imc = 0; imc < imcs; imc++) {		imc_counters_config[imc][READ].fd = -1;		imc_counters_config[imc][WRITE].fd = -1;	} 	for (imc = 0; imc < imcs; imc++) {		ret = open_perf_event(imc, cpu_no, READ);		if (ret)			goto close_fds;		ret = open_perf_event(imc, cpu_no, WRITE);		if (ret)			goto close_fds;	} 	return 0; close_fds:	perf_close_imc_mem_bw();	return -1;} /* * do_mem_bw_test - Perform memory bandwidth test * * Runs memory bandwidth test over one second period. Also, handles starting * and stopping of the IMC perf counters around the test. */static void do_imc_mem_bw_test(void){	int imc; 	for (imc = 0; imc < imcs; imc++) {		membw_ioctl_perf_event_ioc_reset_enable(imc, READ);		membw_ioctl_perf_event_ioc_reset_enable(imc, WRITE);	} 	sleep(1); 	/* Stop counters after a second to get results (both read and write) */	for (imc = 0; imc < imcs; imc++) {		membw_ioctl_perf_event_ioc_disable(imc, READ);		membw_ioctl_perf_event_ioc_disable(imc, WRITE);	}} /* * get_mem_bw_imc - Memory bandwidth as reported by iMC counters * @bw_report: Bandwidth report type (reads, writes) * * Memory bandwidth utilized by a process on a socket can be calculated * using iMC counters. Perf events are used to read these counters. * * Return: = 0 on success. < 0 on failure. */static int get_mem_bw_imc(const char *bw_report, float *bw_imc){	float reads, writes, of_mul_read, of_mul_write;	int imc; 	/* Start all iMC counters to log values (both read and write) */	reads = 0, writes = 0, of_mul_read = 1, of_mul_write = 1; 	/*	 * Get results which are stored in struct type imc_counter_config	 * Take overflow into consideration before calculating total bandwidth.	 */	for (imc = 0; imc < imcs; imc++) {		struct imc_counter_config *r =			&imc_counters_config[imc][READ];		struct imc_counter_config *w =			&imc_counters_config[imc][WRITE]; 		if (read(r->fd, &r->return_value,			 sizeof(struct membw_read_format)) == -1) {			ksft_perror("Couldn't get read bandwidth through iMC");			return -1;		} 		if (read(w->fd, &w->return_value,			 sizeof(struct membw_read_format)) == -1) {			ksft_perror("Couldn't get write bandwidth through iMC");			return -1;		} 		__u64 r_time_enabled = r->return_value.time_enabled;		__u64 r_time_running = r->return_value.time_running; 		if (r_time_enabled != r_time_running)			of_mul_read = (float)r_time_enabled /					(float)r_time_running; 		__u64 w_time_enabled = w->return_value.time_enabled;		__u64 w_time_running = w->return_value.time_running; 		if (w_time_enabled != w_time_running)			of_mul_write = (float)w_time_enabled /					(float)w_time_running;		reads += r->return_value.value * of_mul_read * SCALE;		writes += w->return_value.value * of_mul_write * SCALE;	} 	if (strcmp(bw_report, "reads") == 0) {		*bw_imc = reads;		return 0;	} 	if (strcmp(bw_report, "writes") == 0) {		*bw_imc = writes;		return 0;	} 	*bw_imc = reads + writes;	return 0;} /* * initialize_mem_bw_resctrl:	Appropriately populate "mbm_total_path" * @param:	Parameters passed to resctrl_val() * @domain_id:	Domain ID (cache ID; for MB, L3 cache ID) */void initialize_mem_bw_resctrl(const struct resctrl_val_param *param,			       int domain_id){	sprintf(mbm_total_path, CON_MBM_LOCAL_BYTES_PATH, RESCTRL_PATH,		param->ctrlgrp, domain_id);} /* * Open file to read MBM local bytes from resctrl FS */static FILE *open_mem_bw_resctrl(const char *mbm_bw_file){	FILE *fp; 	fp = fopen(mbm_bw_file, "r");	if (!fp)		ksft_perror("Failed to open total memory bandwidth file"); 	return fp;} /* * Get MBM Local bytes as reported by resctrl FS */static int get_mem_bw_resctrl(FILE *fp, unsigned long *mbm_total){	if (fscanf(fp, "%lu\n", mbm_total) <= 0) {		ksft_perror("Could not get MBM local bytes");		return -1;	}	return 0;} static pid_t bm_pid, ppid; void ctrlc_handler(int signum, siginfo_t *info, void *ptr){	/* Only kill child after bm_pid is set after fork() */	if (bm_pid)		kill(bm_pid, SIGKILL);	umount_resctrlfs();	if (current_test && current_test->cleanup)		current_test->cleanup();	ksft_print_msg("Ending\n\n"); 	exit(EXIT_SUCCESS);} /* * Register CTRL-C handler for parent, as it has to kill * child process before exiting. */int signal_handler_register(const struct resctrl_test *test){	struct sigaction sigact = {};	int ret = 0; 	bm_pid = 0; 	current_test = test;	sigact.sa_sigaction = ctrlc_handler;	sigemptyset(&sigact.sa_mask);	sigact.sa_flags = SA_SIGINFO;	if (sigaction(SIGINT, &sigact, NULL) ||	    sigaction(SIGTERM, &sigact, NULL) ||	    sigaction(SIGHUP, &sigact, NULL)) {		ksft_perror("sigaction");		ret = -1;	}	return ret;} /* * Reset signal handler to SIG_DFL. * Non-Value return because the caller should keep * the error code of other path even if sigaction fails. */void signal_handler_unregister(void){	struct sigaction sigact = {}; 	current_test = NULL;	sigact.sa_handler = SIG_DFL;

Lines 1–500 of 828. Load more