tools/power/x86/intel-speed-select: OOB daemon mode

It is possible that some out of band agent changed config level. In this
case CPUs need to be online/offline to support this config change. Add
a command line option --oob, so that this tool can run as daemon and poll
for config level change and take action. The poll interval is configurable
in seconds using config option --poll-interval.
Signed-off-by: Srinivas Pandruvada <srinivas.pandruvada@linux.intel.com>

tools/power/x86/intel-speed-select/Build

-intel-speed-select-y +=  isst-config.o isst-core.o isst-display.o
+intel-speed-select-y +=  isst-config.o isst-core.o isst-display.o isst-daemon.o

tools/power/x86/intel-speed-select/isst-config.c

@@ -368,7 +368,7 @@ int get_topo_max_cpus(void)
 	return topo_max_cpus;
 }
 
-static void set_cpu_online_offline(int cpu, int state)
+void set_cpu_online_offline(int cpu, int state)
 {
 	char buffer[128];
 	int fd, ret;
@@ -409,12 +409,10 @@ static void force_all_cpus_online(void)
 	unlink("/var/run/isst_cpu_topology.dat");
 }
 
-#define MAX_PACKAGE_COUNT 8
-#define MAX_DIE_PER_PACKAGE 2
-static void for_each_online_package_in_set(void (*callback)(int, void *, void *,
-							    void *, void *),
-					   void *arg1, void *arg2, void *arg3,
-					   void *arg4)
+void for_each_online_package_in_set(void (*callback)(int, void *, void *,
+						     void *, void *),
+				    void *arg1, void *arg2, void *arg3,
+				    void *arg4)
 {
 	int max_packages[MAX_PACKAGE_COUNT * MAX_PACKAGE_COUNT];
 	int pkg_index = 0, i;
@@ -2803,7 +2801,9 @@ static void usage(void)
 	printf("\t[-p|--pause] : Delay between two mail box commands in milliseconds\n");
 	printf("\t[-r|--retry] : Retry count for mail box commands on failure, default 3\n");
 	printf("\t[-v|--version] : Print version\n");
-
+	printf("\t[-b|--oob : Start a daemon to process HFI events for perf profile change from Out of Band agent.\n");
+	printf("\t[-n|--no-daemon : Don't run as daemon. By default --oob will turn on daemon mode\n");
+	printf("\t[-w|--delay : Delay for reading config level state change in OOB poll mode.\n");
 	printf("\nResult format\n");
 	printf("\tResult display uses a common format for each command:\n");
 	printf("\tResults are formatted in text/JSON with\n");
@@ -2837,6 +2837,9 @@ static void cmdline(int argc, char **argv)
 	int opt, force_cpus_online = 0;
 	int option_index = 0;
 	int ret;
+	int oob_mode = 0;
+	int poll_interval = -1;
+	int no_daemon = 0;
 
 	static struct option long_options[] = {
 		{ "all-cpus-online", no_argument, 0, 'a' },
@@ -2849,6 +2852,9 @@ static void cmdline(int argc, char **argv)
 		{ "out", required_argument, 0, 'o' },
 		{ "retry", required_argument, 0, 'r' },
 		{ "version", no_argument, 0, 'v' },
+		{ "oob", no_argument, 0, 'b' },
+		{ "no-daemon", no_argument, 0, 'n' },
+		{ "poll-interval", required_argument, 0, 'w' },
 		{ 0, 0, 0, 0 }
 	};
 
@@ -2875,7 +2881,7 @@ static void cmdline(int argc, char **argv)
 	}
 
 	progname = argv[0];
-	while ((opt = getopt_long_only(argc, argv, "+c:df:hio:va", long_options,
+	while ((opt = getopt_long_only(argc, argv, "+c:df:hio:vabw:n", long_options,
 				       &option_index)) != -1) {
 		switch (opt) {
 		case 'a':
@@ -2920,12 +2926,26 @@ static void cmdline(int argc, char **argv)
 		case 'v':
 			print_version();
 			break;
+		case 'b':
+			oob_mode = 1;
+			break;
+		case 'n':
+			no_daemon = 1;
+			break;
+		case 'w':
+			ret = strtol(optarg, &ptr, 10);
+			if (!ret) {
+				fprintf(stderr, "Invalid poll interval count\n");
+				exit(0);
+			}
+			poll_interval = ret;
+			break;
 		default:
 			usage();
 		}
 	}
 
-	if (optind > (argc - 2)) {
+	if (optind > (argc - 2) && !oob_mode) {
 		usage();
 		exit(0);
 	}
@@ -2936,6 +2956,17 @@ static void cmdline(int argc, char **argv)
 	set_cpu_present_cpu_mask();
 	set_cpu_target_cpu_mask();
 
+	if (oob_mode) {
+		create_cpu_map();
+		if (debug_flag)
+			fprintf(stderr, "OOB mode is enabled in debug mode\n");
+
+		ret = isst_daemon(debug_flag, poll_interval, no_daemon);
+		if (ret)
+			fprintf(stderr, "OOB mode enable failed\n");
+		goto out;
+	}
+
 	if (!is_clx_n_platform()) {
 		ret = isst_fill_platform_info();
 		if (ret)

tools/power/x86/intel-speed-select/isst-daemon.c

+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Intel Speed Select -- Allow speed select to daemonize
+ * Copyright (c) 2022 Intel Corporation.
+ */
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <unistd.h>
+#include <fcntl.h>
+#include <sys/file.h>
+#include <sys/types.h>
+#include <sys/stat.h>
+#include <errno.h>
+#include <getopt.h>
+#include <signal.h>
+#include <time.h>
+
+#include "isst.h"
+
+static int per_package_levels_info[MAX_PACKAGE_COUNT][MAX_DIE_PER_PACKAGE];
+static time_t per_package_levels_tm[MAX_PACKAGE_COUNT][MAX_DIE_PER_PACKAGE];
+
+static void init_levels(void)
+{
+	int i, j;
+
+	for (i = 0; i < MAX_PACKAGE_COUNT; ++i)
+		for (j = 0; j < MAX_DIE_PER_PACKAGE; ++j)
+			per_package_levels_info[i][j] = -1;
+}
+
+void process_level_change(int cpu)
+{
+	struct isst_pkg_ctdp_level_info ctdp_level;
+	int pkg_id = get_physical_package_id(cpu);
+	int die_id = get_physical_die_id(cpu);
+	struct isst_pkg_ctdp pkg_dev;
+	time_t tm;
+	int ret;
+
+	if (pkg_id >= MAX_PACKAGE_COUNT || die_id > MAX_DIE_PER_PACKAGE) {
+		debug_printf("Invalid package/die info for cpu:%d\n", cpu);
+		return;
+	}
+
+	tm = time(NULL);
+	if (tm - per_package_levels_tm[pkg_id][die_id] < 2 )
+		return;
+
+	per_package_levels_tm[pkg_id][die_id] = tm;
+
+	ret = isst_get_ctdp_levels(cpu, &pkg_dev);
+	if (ret) {
+		debug_printf("Can't get tdp levels for cpu:%d\n", cpu);
+		return;
+	}
+
+	debug_printf("Get Config level %d pkg:%d die:%d current_level:%d \n", cpu,
+		      pkg_id, die_id, pkg_dev.current_level);
+
+	if (pkg_dev.locked) {
+		debug_printf("config TDP s locked \n");
+		return;
+	}
+
+	if (per_package_levels_info[pkg_id][die_id] == pkg_dev.current_level)
+		return;
+
+	debug_printf("**Config level change for cpu:%d pkg:%d die:%d from %d to %d\n",
+		      cpu, pkg_id, die_id, per_package_levels_info[pkg_id][die_id],
+		      pkg_dev.current_level);
+
+	per_package_levels_info[pkg_id][die_id] = pkg_dev.current_level;
+
+	ctdp_level.core_cpumask_size =
+		alloc_cpu_set(&ctdp_level.core_cpumask);
+	ret = isst_get_coremask_info(cpu, pkg_dev.current_level, &ctdp_level);
+	if (ret) {
+		free_cpu_set(ctdp_level.core_cpumask);
+		debug_printf("Can't get core_mask:%d\n", cpu);
+		return;
+	}
+
+	if (ctdp_level.cpu_count) {
+		int i, max_cpus = get_topo_max_cpus();
+		for (i = 0; i < max_cpus; ++i) {
+			if (pkg_id != get_physical_package_id(i) || die_id != get_physical_die_id(i))
+				continue;
+			if (CPU_ISSET_S(i, ctdp_level.core_cpumask_size, ctdp_level.core_cpumask)) {
+				fprintf(stderr, "online cpu %d\n", i);
+				set_cpu_online_offline(i, 1);
+			} else {
+				fprintf(stderr, "offline cpu %d\n", i);
+				set_cpu_online_offline(i, 0);
+			}
+		}
+	}
+
+	free_cpu_set(ctdp_level.core_cpumask);
+}
+
+static void _poll_for_config_change(int cpu, void *arg1, void *arg2,
+				    void *arg3, void *arg4)
+{
+	process_level_change(cpu);
+}
+
+static void poll_for_config_change(void)
+{
+	for_each_online_package_in_set(_poll_for_config_change, NULL, NULL,
+				       NULL, NULL);
+}
+
+static int done = 0;
+static int pid_file_handle;
+
+static void signal_handler(int sig)
+{
+	switch (sig) {
+	case SIGINT:
+	case SIGTERM:
+		done = 1;
+		exit(0);
+		break;
+	default:
+		break;
+	}
+}
+
+static void daemonize(char *rundir, char *pidfile)
+{
+	int pid, sid, i;
+	char str[10];
+	struct sigaction sig_actions;
+	sigset_t sig_set;
+	int ret;
+
+	if (getppid() == 1)
+		return;
+
+	sigemptyset(&sig_set);
+	sigaddset(&sig_set, SIGCHLD);
+	sigaddset(&sig_set, SIGTSTP);
+	sigaddset(&sig_set, SIGTTOU);
+	sigaddset(&sig_set, SIGTTIN);
+	sigprocmask(SIG_BLOCK, &sig_set, NULL);
+
+	sig_actions.sa_handler = signal_handler;
+	sigemptyset(&sig_actions.sa_mask);
+	sig_actions.sa_flags = 0;
+
+	sigaction(SIGHUP, &sig_actions, NULL);
+	sigaction(SIGTERM, &sig_actions, NULL);
+	sigaction(SIGINT, &sig_actions, NULL);
+
+	pid = fork();
+	if (pid < 0) {
+		/* Could not fork */
+		exit(EXIT_FAILURE);
+	}
+	if (pid > 0)
+		exit(EXIT_SUCCESS);
+
+	umask(027);
+
+	sid = setsid();
+	if (sid < 0)
+		exit(EXIT_FAILURE);
+
+	/* close all descriptors */
+	for (i = getdtablesize(); i >= 0; --i)
+		close(i);
+
+	i = open("/dev/null", O_RDWR);
+	ret = dup(i);
+	if (ret == -1)
+		exit(EXIT_FAILURE);
+
+	ret = dup(i);
+	if (ret == -1)
+		exit(EXIT_FAILURE);
+
+	ret = chdir(rundir);
+	if (ret == -1)
+		exit(EXIT_FAILURE);
+
+	pid_file_handle = open(pidfile, O_RDWR | O_CREAT, 0600);
+	if (pid_file_handle == -1) {
+		/* Couldn't open lock file */
+		exit(1);
+	}
+	/* Try to lock file */
+#ifdef LOCKF_SUPPORT
+	if (lockf(pid_file_handle, F_TLOCK, 0) == -1) {
+#else
+	if (flock(pid_file_handle, LOCK_EX|LOCK_NB) < 0) {
+#endif
+		/* Couldn't get lock on lock file */
+		fprintf(stderr, "Couldn't get lock file %d\n", getpid());
+		exit(1);
+	}
+	snprintf(str, sizeof(str), "%d\n", getpid());
+	ret = write(pid_file_handle, str, strlen(str));
+	if (ret == -1)
+		exit(EXIT_FAILURE);
+
+	close(i);
+}
+
+int isst_daemon(int debug_mode, int poll_interval, int no_daemon)
+{
+	int ret;
+
+	if (!no_daemon && poll_interval < 0 && !debug_mode) {
+		fprintf(stderr, "OOB mode is enabled and will run as daemon\n");
+		daemonize((char *) "/tmp/",
+				(char *)"/tmp/hfi-events.pid");
+	} else {
+		signal(SIGINT, signal_handler);
+	}
+
+	init_levels();
+
+	if (poll_interval < 0) {
+		fprintf(stderr, "Must specify poll-interval\n");
+		return ret;
+	}
+
+	debug_printf("Starting loop\n");
+	while (!done) {
+		sleep(poll_interval);
+		poll_for_config_change();
+	}
+
+	return 0;
+}

tools/power/x86/intel-speed-select/isst.h

@@ -76,6 +76,9 @@
 
 #define DISP_FREQ_MULTIPLIER 100
 
+#define MAX_PACKAGE_COUNT 8
+#define MAX_DIE_PER_PACKAGE 2
+
 struct isst_clos_config {
 	int pkg_id;
 	int die_id;
@@ -260,4 +263,12 @@ extern int is_skx_based_platform(void);
 extern int is_spr_platform(void);
 extern int is_icx_platform(void);
 extern void isst_trl_display_information(int cpu, FILE *outf, unsigned long long trl);
+
+extern void set_cpu_online_offline(int cpu, int state);
+extern void for_each_online_package_in_set(void (*callback)(int, void *, void *,
+							    void *, void *),
+					   void *arg1, void *arg2, void *arg3,
+					   void *arg4);
+extern int isst_daemon(int debug_mode, int poll_interval, int no_daemon);
+extern void process_level_change(int cpu);
 #endif