summaryrefslogtreecommitdiffstats
path: root/tools/tracing/rtla/src
diff options
context:
space:
mode:
Diffstat (limited to 'tools/tracing/rtla/src')
-rw-r--r--tools/tracing/rtla/src/osnoise.c117
-rw-r--r--tools/tracing/rtla/src/osnoise.h7
-rw-r--r--tools/tracing/rtla/src/osnoise_hist.c9
-rw-r--r--tools/tracing/rtla/src/osnoise_top.c84
-rw-r--r--tools/tracing/rtla/src/rtla.c4
-rw-r--r--tools/tracing/rtla/src/timerlat_aa.c990
-rw-r--r--tools/tracing/rtla/src/timerlat_aa.h12
-rw-r--r--tools/tracing/rtla/src/timerlat_top.c46
-rw-r--r--tools/tracing/rtla/src/utils.h3
9 files changed, 1251 insertions, 21 deletions
diff --git a/tools/tracing/rtla/src/osnoise.c b/tools/tracing/rtla/src/osnoise.c
index 4dee343909b1..3ca7a3853943 100644
--- a/tools/tracing/rtla/src/osnoise.c
+++ b/tools/tracing/rtla/src/osnoise.c
@@ -734,6 +734,113 @@ void osnoise_put_tracing_thresh(struct osnoise_context *context)
context->orig_tracing_thresh = OSNOISE_OPTION_INIT_VAL;
}
+static int osnoise_options_get_option(char *option)
+{
+ char *options = tracefs_instance_file_read(NULL, "osnoise/options", NULL);
+ char no_option[128];
+ int retval = 0;
+ char *opt;
+
+ if (!options)
+ return OSNOISE_OPTION_INIT_VAL;
+
+ /*
+ * Check first if the option is disabled.
+ */
+ snprintf(no_option, sizeof(no_option), "NO_%s", option);
+
+ opt = strstr(options, no_option);
+ if (opt)
+ goto out_free;
+
+ /*
+ * Now that it is not disabled, if the string is there, it is
+ * enabled. If the string is not there, the option does not exist.
+ */
+ opt = strstr(options, option);
+ if (opt)
+ retval = 1;
+ else
+ retval = OSNOISE_OPTION_INIT_VAL;
+
+out_free:
+ free(options);
+ return retval;
+}
+
+static int osnoise_options_set_option(char *option, bool onoff)
+{
+ char no_option[128];
+
+ if (onoff)
+ return tracefs_instance_file_write(NULL, "osnoise/options", option);
+
+ snprintf(no_option, sizeof(no_option), "NO_%s", option);
+
+ return tracefs_instance_file_write(NULL, "osnoise/options", no_option);
+}
+
+static int osnoise_get_irq_disable(struct osnoise_context *context)
+{
+ if (context->opt_irq_disable != OSNOISE_OPTION_INIT_VAL)
+ return context->opt_irq_disable;
+
+ if (context->orig_opt_irq_disable != OSNOISE_OPTION_INIT_VAL)
+ return context->orig_opt_irq_disable;
+
+ context->orig_opt_irq_disable = osnoise_options_get_option("OSNOISE_IRQ_DISABLE");
+
+ return context->orig_opt_irq_disable;
+}
+
+int osnoise_set_irq_disable(struct osnoise_context *context, bool onoff)
+{
+ int opt_irq_disable = osnoise_get_irq_disable(context);
+ int retval;
+
+ if (opt_irq_disable == OSNOISE_OPTION_INIT_VAL)
+ return -1;
+
+ if (opt_irq_disable == onoff)
+ return 0;
+
+ retval = osnoise_options_set_option("OSNOISE_IRQ_DISABLE", onoff);
+ if (retval < 0)
+ return -1;
+
+ context->opt_irq_disable = onoff;
+
+ return 0;
+}
+
+static void osnoise_restore_irq_disable(struct osnoise_context *context)
+{
+ int retval;
+
+ if (context->orig_opt_irq_disable == OSNOISE_OPTION_INIT_VAL)
+ return;
+
+ if (context->orig_opt_irq_disable == context->opt_irq_disable)
+ goto out_done;
+
+ retval = osnoise_options_set_option("OSNOISE_IRQ_DISABLE", context->orig_opt_irq_disable);
+ if (retval < 0)
+ err_msg("Could not restore original OSNOISE_IRQ_DISABLE option\n");
+
+out_done:
+ context->orig_opt_irq_disable = OSNOISE_OPTION_INIT_VAL;
+}
+
+static void osnoise_put_irq_disable(struct osnoise_context *context)
+{
+ osnoise_restore_irq_disable(context);
+
+ if (context->orig_opt_irq_disable == OSNOISE_OPTION_INIT_VAL)
+ return;
+
+ context->orig_opt_irq_disable = OSNOISE_OPTION_INIT_VAL;
+}
+
/*
* enable_osnoise - enable osnoise tracer in the trace_instance
*/
@@ -798,6 +905,9 @@ struct osnoise_context *osnoise_context_alloc(void)
context->orig_tracing_thresh = OSNOISE_OPTION_INIT_VAL;
context->tracing_thresh = OSNOISE_OPTION_INIT_VAL;
+ context->orig_opt_irq_disable = OSNOISE_OPTION_INIT_VAL;
+ context->opt_irq_disable = OSNOISE_OPTION_INIT_VAL;
+
osnoise_get_context(context);
return context;
@@ -824,6 +934,7 @@ void osnoise_put_context(struct osnoise_context *context)
osnoise_put_timerlat_period_us(context);
osnoise_put_print_stack(context);
osnoise_put_tracing_thresh(context);
+ osnoise_put_irq_disable(context);
free(context);
}
@@ -958,3 +1069,9 @@ usage:
osnoise_usage(1);
exit(1);
}
+
+int hwnoise_main(int argc, char *argv[])
+{
+ osnoise_top_main(argc, argv);
+ exit(0);
+}
diff --git a/tools/tracing/rtla/src/osnoise.h b/tools/tracing/rtla/src/osnoise.h
index 04a4384cc544..4dcf22ccd704 100644
--- a/tools/tracing/rtla/src/osnoise.h
+++ b/tools/tracing/rtla/src/osnoise.h
@@ -38,6 +38,10 @@ struct osnoise_context {
/* -1 as init value because 0 is disabled */
long long orig_print_stack;
long long print_stack;
+
+ /* -1 as init value because 0 is off */
+ int orig_opt_irq_disable;
+ int opt_irq_disable;
};
/*
@@ -79,6 +83,8 @@ void osnoise_restore_print_stack(struct osnoise_context *context);
int osnoise_set_print_stack(struct osnoise_context *context,
long long print_stack);
+int osnoise_set_irq_disable(struct osnoise_context *context, bool onoff);
+
/*
* osnoise_tool - osnoise based tool definition.
*/
@@ -97,3 +103,4 @@ struct osnoise_tool *osnoise_init_trace_tool(char *tracer);
int osnoise_hist_main(int argc, char *argv[]);
int osnoise_top_main(int argc, char **argv);
int osnoise_main(int argc, char **argv);
+int hwnoise_main(int argc, char **argv);
diff --git a/tools/tracing/rtla/src/osnoise_hist.c b/tools/tracing/rtla/src/osnoise_hist.c
index 5d7ea479ac89..13e1233690bb 100644
--- a/tools/tracing/rtla/src/osnoise_hist.c
+++ b/tools/tracing/rtla/src/osnoise_hist.c
@@ -121,6 +121,7 @@ static void osnoise_hist_update_multiple(struct osnoise_tool *tool, int cpu,
{
struct osnoise_hist_params *params = tool->params;
struct osnoise_hist_data *data = tool->data;
+ unsigned long long total_duration;
int entries = data->entries;
int bucket;
int *hist;
@@ -131,10 +132,12 @@ static void osnoise_hist_update_multiple(struct osnoise_tool *tool, int cpu,
if (data->bucket_size)
bucket = duration / data->bucket_size;
+ total_duration = duration * count;
+
hist = data->hist[cpu].samples;
data->hist[cpu].count += count;
update_min(&data->hist[cpu].min_sample, &duration);
- update_sum(&data->hist[cpu].sum_sample, &duration);
+ update_sum(&data->hist[cpu].sum_sample, &total_duration);
update_max(&data->hist[cpu].max_sample, &duration);
if (bucket < entries)
@@ -332,8 +335,8 @@ osnoise_print_summary(struct osnoise_hist_params *params,
continue;
if (data->hist[cpu].count)
- trace_seq_printf(trace->seq, "%9llu ",
- data->hist[cpu].sum_sample / data->hist[cpu].count);
+ trace_seq_printf(trace->seq, "%9.2f ",
+ ((double) data->hist[cpu].sum_sample) / data->hist[cpu].count);
else
trace_seq_printf(trace->seq, " - ");
}
diff --git a/tools/tracing/rtla/src/osnoise_top.c b/tools/tracing/rtla/src/osnoise_top.c
index 76479bfb2922..562f2e4b18c5 100644
--- a/tools/tracing/rtla/src/osnoise_top.c
+++ b/tools/tracing/rtla/src/osnoise_top.c
@@ -14,6 +14,11 @@
#include "osnoise.h"
#include "utils.h"
+enum osnoise_mode {
+ MODE_OSNOISE = 0,
+ MODE_HWNOISE
+};
+
/*
* osnoise top parameters
*/
@@ -32,6 +37,7 @@ struct osnoise_top_params {
int set_sched;
struct sched_attr sched_param;
struct trace_events *events;
+ enum osnoise_mode mode;
};
struct osnoise_top_cpu {
@@ -143,15 +149,23 @@ osnoise_top_handler(struct trace_seq *s, struct tep_record *record,
*/
static void osnoise_top_header(struct osnoise_tool *top)
{
+ struct osnoise_top_params *params = top->params;
struct trace_seq *s = top->trace.seq;
char duration[26];
get_duration(top->start_time, duration, sizeof(duration));
trace_seq_printf(s, "\033[2;37;40m");
- trace_seq_printf(s, " Operating System Noise");
- trace_seq_printf(s, " ");
- trace_seq_printf(s, " ");
+ trace_seq_printf(s, " ");
+
+ if (params->mode == MODE_OSNOISE) {
+ trace_seq_printf(s, "Operating System Noise");
+ trace_seq_printf(s, " ");
+ } else if (params->mode == MODE_HWNOISE) {
+ trace_seq_printf(s, "Hardware-related Noise");
+ }
+
+ trace_seq_printf(s, " ");
trace_seq_printf(s, "\033[0;0;0m");
trace_seq_printf(s, "\n");
@@ -162,7 +176,14 @@ static void osnoise_top_header(struct osnoise_tool *top)
trace_seq_printf(s, " Noise ");
trace_seq_printf(s, " %% CPU Aval ");
trace_seq_printf(s, " Max Noise Max Single ");
- trace_seq_printf(s, " HW NMI IRQ Softirq Thread");
+ trace_seq_printf(s, " HW NMI");
+
+ if (params->mode == MODE_HWNOISE)
+ goto eol;
+
+ trace_seq_printf(s, " IRQ Softirq Thread");
+
+eol:
trace_seq_printf(s, "\033[0;0;0m");
trace_seq_printf(s, "\n");
}
@@ -181,6 +202,7 @@ static void clear_terminal(struct trace_seq *seq)
*/
static void osnoise_top_print(struct osnoise_tool *tool, int cpu)
{
+ struct osnoise_top_params *params = tool->params;
struct trace_seq *s = tool->trace.seq;
struct osnoise_top_cpu *cpu_data;
struct osnoise_top_data *data;
@@ -205,6 +227,12 @@ static void osnoise_top_print(struct osnoise_tool *tool, int cpu)
trace_seq_printf(s, "%12llu ", cpu_data->hw_count);
trace_seq_printf(s, "%12llu ", cpu_data->nmi_count);
+
+ if (params->mode == MODE_HWNOISE) {
+ trace_seq_printf(s, "\n");
+ return;
+ }
+
trace_seq_printf(s, "%12llu ", cpu_data->irq_count);
trace_seq_printf(s, "%12llu ", cpu_data->softirq_count);
trace_seq_printf(s, "%12llu\n", cpu_data->thread_count);
@@ -241,12 +269,12 @@ osnoise_print_stats(struct osnoise_top_params *params, struct osnoise_tool *top)
/*
* osnoise_top_usage - prints osnoise top usage message
*/
-void osnoise_top_usage(char *usage)
+static void osnoise_top_usage(struct osnoise_top_params *params, char *usage)
{
int i;
static const char * const msg[] = {
- " usage: rtla osnoise [top] [-h] [-q] [-D] [-d s] [-a us] [-p us] [-r us] [-s us] [-S us] \\",
+ " [-h] [-q] [-D] [-d s] [-a us] [-p us] [-r us] [-s us] [-S us] \\",
" [-T us] [-t[=file]] [-e sys[:event]] [--filter <filter>] [--trigger <trigger>] \\",
" [-c cpu-list] [-P priority]",
"",
@@ -277,9 +305,22 @@ void osnoise_top_usage(char *usage)
if (usage)
fprintf(stderr, "%s\n", usage);
- fprintf(stderr, "rtla osnoise top: a per-cpu summary of the OS noise (version %s)\n",
+ if (params->mode == MODE_OSNOISE) {
+ fprintf(stderr,
+ "rtla osnoise top: a per-cpu summary of the OS noise (version %s)\n",
VERSION);
+ fprintf(stderr, " usage: rtla osnoise [top]");
+ }
+
+ if (params->mode == MODE_HWNOISE) {
+ fprintf(stderr,
+ "rtla hwnoise: a summary of hardware-related noise (version %s)\n",
+ VERSION);
+
+ fprintf(stderr, " usage: rtla hwnoise");
+ }
+
for (i = 0; msg[i]; i++)
fprintf(stderr, "%s\n", msg[i]);
exit(1);
@@ -299,6 +340,9 @@ struct osnoise_top_params *osnoise_top_parse_args(int argc, char **argv)
if (!params)
exit(1);
+ if (strcmp(argv[0], "hwnoise") == 0)
+ params->mode = MODE_HWNOISE;
+
while (1) {
static struct option long_options[] = {
{"auto", required_argument, 0, 'a'},
@@ -345,7 +389,7 @@ struct osnoise_top_params *osnoise_top_parse_args(int argc, char **argv)
case 'c':
retval = parse_cpu_list(optarg, &params->monitored_cpus);
if (retval)
- osnoise_top_usage("\nInvalid -c cpu list\n");
+ osnoise_top_usage(params, "\nInvalid -c cpu list\n");
params->cpus = optarg;
break;
case 'D':
@@ -354,7 +398,7 @@ struct osnoise_top_params *osnoise_top_parse_args(int argc, char **argv)
case 'd':
params->duration = parse_seconds_duration(optarg);
if (!params->duration)
- osnoise_top_usage("Invalid -D duration\n");
+ osnoise_top_usage(params, "Invalid -D duration\n");
break;
case 'e':
tevent = trace_event_alloc(optarg);
@@ -370,17 +414,17 @@ struct osnoise_top_params *osnoise_top_parse_args(int argc, char **argv)
break;
case 'h':
case '?':
- osnoise_top_usage(NULL);
+ osnoise_top_usage(params, NULL);
break;
case 'p':
params->period = get_llong_from_str(optarg);
if (params->period > 10000000)
- osnoise_top_usage("Period longer than 10 s\n");
+ osnoise_top_usage(params, "Period longer than 10 s\n");
break;
case 'P':
retval = parse_prio(optarg, &params->sched_param);
if (retval == -1)
- osnoise_top_usage("Invalid -P priority");
+ osnoise_top_usage(params, "Invalid -P priority");
params->set_sched = 1;
break;
case 'q':
@@ -389,7 +433,7 @@ struct osnoise_top_params *osnoise_top_parse_args(int argc, char **argv)
case 'r':
params->runtime = get_llong_from_str(optarg);
if (params->runtime < 100)
- osnoise_top_usage("Runtime shorter than 100 us\n");
+ osnoise_top_usage(params, "Runtime shorter than 100 us\n");
break;
case 's':
params->stop_us = get_llong_from_str(optarg);
@@ -415,7 +459,7 @@ struct osnoise_top_params *osnoise_top_parse_args(int argc, char **argv)
exit(EXIT_FAILURE);
}
} else {
- osnoise_top_usage("--trigger requires a previous -e\n");
+ osnoise_top_usage(params, "--trigger requires a previous -e\n");
}
break;
case '1': /* filter */
@@ -426,11 +470,11 @@ struct osnoise_top_params *osnoise_top_parse_args(int argc, char **argv)
exit(EXIT_FAILURE);
}
} else {
- osnoise_top_usage("--filter requires a previous -e\n");
+ osnoise_top_usage(params, "--filter requires a previous -e\n");
}
break;
default:
- osnoise_top_usage("Invalid option");
+ osnoise_top_usage(params, "Invalid option");
}
}
@@ -495,6 +539,14 @@ osnoise_top_apply_config(struct osnoise_tool *tool, struct osnoise_top_params *p
}
}
+ if (params->mode == MODE_HWNOISE) {
+ retval = osnoise_set_irq_disable(tool->context, 1);
+ if (retval) {
+ err_msg("Failed to set OSNOISE_IRQ_DISABLE option\n");
+ goto out_err;
+ }
+ }
+
return 0;
out_err:
diff --git a/tools/tracing/rtla/src/rtla.c b/tools/tracing/rtla/src/rtla.c
index 52e8f1825281..7635c70123ab 100644
--- a/tools/tracing/rtla/src/rtla.c
+++ b/tools/tracing/rtla/src/rtla.c
@@ -26,6 +26,7 @@ static void rtla_usage(int err)
"",
" commands:",
" osnoise - gives information about the operating system noise (osnoise)",
+ " hwnoise - gives information about hardware-related noise",
" timerlat - measures the timer irq and thread latency",
"",
NULL,
@@ -47,6 +48,9 @@ int run_command(int argc, char **argv, int start_position)
if (strcmp(argv[start_position], "osnoise") == 0) {
osnoise_main(argc-start_position, &argv[start_position]);
goto ran;
+ } else if (strcmp(argv[start_position], "hwnoise") == 0) {
+ hwnoise_main(argc-start_position, &argv[start_position]);
+ goto ran;
} else if (strcmp(argv[start_position], "timerlat") == 0) {
timerlat_main(argc-start_position, &argv[start_position]);
goto ran;
diff --git a/tools/tracing/rtla/src/timerlat_aa.c b/tools/tracing/rtla/src/timerlat_aa.c
new file mode 100644
index 000000000000..ec4e0f4b0e6c
--- /dev/null
+++ b/tools/tracing/rtla/src/timerlat_aa.c
@@ -0,0 +1,990 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2023 Red Hat Inc, Daniel Bristot de Oliveira <bristot@kernel.org>
+ */
+
+#include <stdlib.h>
+#include <errno.h>
+#include "utils.h"
+#include "osnoise.h"
+#include "timerlat.h"
+
+enum timelat_state {
+ TIMERLAT_INIT = 0,
+ TIMERLAT_WAITING_IRQ,
+ TIMERLAT_WAITING_THREAD,
+};
+
+#define MAX_COMM 24
+
+/*
+ * Per-cpu data statistics and data.
+ */
+struct timerlat_aa_data {
+ /* Current CPU state */
+ int curr_state;
+
+ /* timerlat IRQ latency */
+ unsigned long long tlat_irq_seqnum;
+ unsigned long long tlat_irq_latency;
+ unsigned long long tlat_irq_timstamp;
+
+ /* timerlat Thread latency */
+ unsigned long long tlat_thread_seqnum;
+ unsigned long long tlat_thread_latency;
+ unsigned long long tlat_thread_timstamp;
+
+ /*
+ * Information about the thread running when the IRQ
+ * arrived.
+ *
+ * This can be blocking or interference, depending on the
+ * priority of the thread. Assuming timerlat is the highest
+ * prio, it is blocking. If timerlat has a lower prio, it is
+ * interference.
+ * note: "unsigned long long" because they are fetch using tep_get_field_val();
+ */
+ unsigned long long run_thread_pid;
+ char run_thread_comm[MAX_COMM];
+ unsigned long long thread_blocking_duration;
+ unsigned long long max_exit_idle_latency;
+
+ /* Information about the timerlat timer irq */
+ unsigned long long timer_irq_start_time;
+ unsigned long long timer_irq_start_delay;
+ unsigned long long timer_irq_duration;
+ unsigned long long timer_exit_from_idle;
+
+ /*
+ * Information about the last IRQ before the timerlat irq
+ * arrived.
+ *
+ * If now - timestamp is <= latency, it might have influenced
+ * in the timerlat irq latency. Otherwise, ignore it.
+ */
+ unsigned long long prev_irq_duration;
+ unsigned long long prev_irq_timstamp;
+
+ /*
+ * Interference sum.
+ */
+ unsigned long long thread_nmi_sum;
+ unsigned long long thread_irq_sum;
+ unsigned long long thread_softirq_sum;
+ unsigned long long thread_thread_sum;
+
+ /*
+ * Interference task information.
+ */
+ struct trace_seq *prev_irqs_seq;
+ struct trace_seq *nmi_seq;
+ struct trace_seq *irqs_seq;
+ struct trace_seq *softirqs_seq;
+ struct trace_seq *threads_seq;
+ struct trace_seq *stack_seq;
+
+ /*
+ * Current thread.
+ */
+ char current_comm[MAX_COMM];
+ unsigned long long current_pid;
+
+ /*
+ * Is the system running a kworker?
+ */
+ unsigned long long kworker;
+ unsigned long long kworker_func;
+};
+
+/*
+ * The analysis context and system wide view
+ */
+struct timerlat_aa_context {
+ int nr_cpus;
+ int dump_tasks;
+
+ /* per CPU data */
+ struct timerlat_aa_data *taa_data;
+
+ /*
+ * required to translate function names and register
+ * events.
+ */
+ struct osnoise_tool *tool;
+};
+
+/*
+ * The data is stored as a local variable, but accessed via a helper function.
+ *
+ * It could be stored inside the trace context. But every access would
+ * require container_of() + a series of pointers. Do we need it? Not sure.
+ *
+ * For now keep it simple. If needed, store it in the tool, add the *context
+ * as a parameter in timerlat_aa_get_ctx() and do the magic there.
+ */
+static struct timerlat_aa_context *__timerlat_aa_ctx;
+
+static struct timerlat_aa_context *timerlat_aa_get_ctx(void)
+{
+ return __timerlat_aa_ctx;
+}
+
+/*
+ * timerlat_aa_get_data - Get the per-cpu data from the timerlat context
+ */
+static struct timerlat_aa_data
+*timerlat_aa_get_data(struct timerlat_aa_context *taa_ctx, int cpu)
+{
+ return &taa_ctx->taa_data[cpu];
+}
+
+/*
+ * timerlat_aa_irq_latency - Handles timerlat IRQ event
+ */
+static int timerlat_aa_irq_latency(struct timerlat_aa_data *taa_data,
+ struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event)
+{
+ /*
+ * For interference, we start now looking for things that can delay
+ * the thread.
+ */
+ taa_data->curr_state = TIMERLAT_WAITING_THREAD;
+ taa_data->tlat_irq_timstamp = record->ts;
+
+ /*
+ * Zero values.
+ */
+ taa_data->thread_nmi_sum = 0;
+ taa_data->thread_irq_sum = 0;
+ taa_data->thread_softirq_sum = 0;
+ taa_data->thread_blocking_duration = 0;
+ taa_data->timer_irq_start_time = 0;
+ taa_data->timer_irq_duration = 0;
+ taa_data->timer_exit_from_idle = 0;
+
+ /*
+ * Zero interference tasks.
+ */
+ trace_seq_reset(taa_data->nmi_seq);
+ trace_seq_reset(taa_data->irqs_seq);
+ trace_seq_reset(taa_data->softirqs_seq);
+ trace_seq_reset(taa_data->threads_seq);
+
+ /* IRQ latency values */
+ tep_get_field_val(s, event, "timer_latency", record, &taa_data->tlat_irq_latency, 1);
+ tep_get_field_val(s, event, "seqnum", record, &taa_data->tlat_irq_seqnum, 1);
+
+ /* The thread that can cause blocking */
+ tep_get_common_field_val(s, event, "common_pid", record, &taa_data->run_thread_pid, 1);
+
+ /*
+ * Get exit from idle case.
+ *
+ * If it is not idle thread:
+ */
+ if (taa_data->run_thread_pid)
+ return 0;
+
+ /*
+ * if the latency is shorter than the known exit from idle:
+ */
+ if (taa_data->tlat_irq_latency < taa_data->max_exit_idle_latency)
+ return 0;
+
+ /*
+ * To be safe, ignore the cases in which an IRQ/NMI could have
+ * interfered with the timerlat IRQ.
+ */
+ if (taa_data->tlat_irq_timstamp - taa_data->tlat_irq_latency
+ < taa_data->prev_irq_timstamp + taa_data->prev_irq_duration)
+ return 0;
+
+ taa_data->max_exit_idle_latency = taa_data->tlat_irq_latency;
+
+ return 0;
+}
+
+/*
+ * timerlat_aa_thread_latency - Handles timerlat thread event
+ */
+static int timerlat_aa_thread_latency(struct timerlat_aa_data *taa_data,
+ struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event)
+{
+ /*
+ * For interference, we start now looking for things that can delay
+ * the IRQ of the next cycle.
+ */
+ taa_data->curr_state = TIMERLAT_WAITING_IRQ;
+ taa_data->tlat_thread_timstamp = record->ts;
+
+ /* Thread latency values */
+ tep_get_field_val(s, event, "timer_latency", record, &taa_data->tlat_thread_latency, 1);
+ tep_get_field_val(s, event, "seqnum", record, &taa_data->tlat_thread_seqnum, 1);
+
+ return 0;
+}
+
+/*
+ * timerlat_aa_handler - Handle timerlat events
+ *
+ * This function is called to handle timerlat events recording statistics.
+ *
+ * Returns 0 on success, -1 otherwise.
+ */
+int timerlat_aa_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ struct timerlat_aa_data *taa_data = timerlat_aa_get_data(taa_ctx, record->cpu);
+ unsigned long long thread;
+
+ if (!taa_data)
+ return -1;
+
+ tep_get_field_val(s, event, "context", record, &thread, 1);
+ if (!thread)
+ return timerlat_aa_irq_latency(taa_data, s, record, event);
+ else
+ return timerlat_aa_thread_latency(taa_data, s, record, event);
+}
+
+/*
+ * timerlat_aa_nmi_handler - Handles NMI noise
+ *
+ * It is used to collect information about interferences from NMI. It is
+ * hooked to the osnoise:nmi_noise event.
+ */
+static int timerlat_aa_nmi_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ struct timerlat_aa_data *taa_data = timerlat_aa_get_data(taa_ctx, record->cpu);
+ unsigned long long duration;
+ unsigned long long start;
+
+ tep_get_field_val(s, event, "duration", record, &duration, 1);
+ tep_get_field_val(s, event, "start", record, &start, 1);
+
+ if (taa_data->curr_state == TIMERLAT_WAITING_IRQ) {
+ taa_data->prev_irq_duration = duration;
+ taa_data->prev_irq_timstamp = start;
+
+ trace_seq_reset(taa_data->prev_irqs_seq);
+ trace_seq_printf(taa_data->prev_irqs_seq, "\t%24s \t\t\t%9.2f us\n",
+ "nmi", ns_to_usf(duration));
+ return 0;
+ }
+
+ taa_data->thread_nmi_sum += duration;
+ trace_seq_printf(taa_data->nmi_seq, " %24s \t\t\t%9.2f us\n",
+ "nmi", ns_to_usf(duration));
+
+ return 0;
+}
+
+/*
+ * timerlat_aa_irq_handler - Handles IRQ noise
+ *
+ * It is used to collect information about interferences from IRQ. It is
+ * hooked to the osnoise:irq_noise event.
+ *
+ * It is a little bit more complex than the other because it measures:
+ * - The IRQs that can delay the timer IRQ before it happened.
+ * - The Timerlat IRQ handler
+ * - The IRQs that happened between the timerlat IRQ and the timerlat thread
+ * (IRQ interference).
+ */
+static int timerlat_aa_irq_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ struct timerlat_aa_data *taa_data = timerlat_aa_get_data(taa_ctx, record->cpu);
+ unsigned long long expected_start;
+ unsigned long long duration;
+ unsigned long long vector;
+ unsigned long long start;
+ char *desc;
+ int val;
+
+ tep_get_field_val(s, event, "duration", record, &duration, 1);
+ tep_get_field_val(s, event, "start", record, &start, 1);
+ tep_get_field_val(s, event, "vector", record, &vector, 1);
+ desc = tep_get_field_raw(s, event, "desc", record, &val, 1);
+
+ /*
+ * Before the timerlat IRQ.
+ */
+ if (taa_data->curr_state == TIMERLAT_WAITING_IRQ) {
+ taa_data->prev_irq_duration = duration;
+ taa_data->prev_irq_timstamp = start;
+
+ trace_seq_reset(taa_data->prev_irqs_seq);
+ trace_seq_printf(taa_data->prev_irqs_seq, "\t%24s:%-3llu \t\t%9.2f us\n",
+ desc, vector, ns_to_usf(duration));
+ return 0;
+ }
+
+ /*
+ * The timerlat IRQ: taa_data->timer_irq_start_time is zeroed at
+ * the timerlat irq handler.
+ */
+ if (!taa_data->timer_irq_start_time) {
+ expected_start = taa_data->tlat_irq_timstamp - taa_data->tlat_irq_latency;
+
+ taa_data->timer_irq_start_time = start;
+ taa_data->timer_irq_duration = duration;
+
+ taa_data->timer_irq_start_delay = taa_data->timer_irq_start_time - expected_start;
+
+ /*
+ * not exit from idle.
+ */
+ if (taa_data->run_thread_pid)
+ return 0;
+
+ if (expected_start > taa_data->prev_irq_timstamp + taa_data->prev_irq_duration)
+ taa_data->timer_exit_from_idle = taa_data->timer_irq_start_delay;
+
+ return 0;
+ }
+
+ /*
+ * IRQ interference.
+ */
+ taa_data->thread_irq_sum += duration;
+ trace_seq_printf(taa_data->irqs_seq, " %24s:%-3llu \t %9.2f us\n",
+ desc, vector, ns_to_usf(duration));
+
+ return 0;
+}
+
+static char *softirq_name[] = { "HI", "TIMER", "NET_TX", "NET_RX", "BLOCK",
+ "IRQ_POLL", "TASKLET", "SCHED", "HRTIMER", "RCU" };
+
+
+/*
+ * timerlat_aa_softirq_handler - Handles Softirq noise
+ *
+ * It is used to collect information about interferences from Softirq. It is
+ * hooked to the osnoise:softirq_noise event.
+ *
+ * It is only printed in the non-rt kernel, as softirqs become thread on RT.
+ */
+static int timerlat_aa_softirq_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ struct timerlat_aa_data *taa_data = timerlat_aa_get_data(taa_ctx, record->cpu);
+ unsigned long long duration;
+ unsigned long long vector;
+ unsigned long long start;
+
+ if (taa_data->curr_state == TIMERLAT_WAITING_IRQ)
+ return 0;
+
+ tep_get_field_val(s, event, "duration", record, &duration, 1);
+ tep_get_field_val(s, event, "start", record, &start, 1);
+ tep_get_field_val(s, event, "vector", record, &vector, 1);
+
+ taa_data->thread_softirq_sum += duration;
+
+ trace_seq_printf(taa_data->softirqs_seq, "\t%24s:%-3llu \t %9.2f us\n",
+ softirq_name[vector], vector, ns_to_usf(duration));
+ return 0;
+}
+
+/*
+ * timerlat_aa_softirq_handler - Handles thread noise
+ *
+ * It is used to collect information about interferences from threads. It is
+ * hooked to the osnoise:thread_noise event.
+ *
+ * Note: if you see thread noise, your timerlat thread was not the highest prio one.
+ */
+static int timerlat_aa_thread_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ struct timerlat_aa_data *taa_data = timerlat_aa_get_data(taa_ctx, record->cpu);
+ unsigned long long duration;
+ unsigned long long start;
+ unsigned long long pid;
+ const char *comm;
+ int val;
+
+ if (taa_data->curr_state == TIMERLAT_WAITING_IRQ)
+ return 0;
+
+ tep_get_field_val(s, event, "duration", record, &duration, 1);
+ tep_get_field_val(s, event, "start", record, &start, 1);
+
+ tep_get_common_field_val(s, event, "common_pid", record, &pid, 1);
+ comm = tep_get_field_raw(s, event, "comm", record, &val, 1);
+
+ if (pid == taa_data->run_thread_pid && !taa_data->thread_blocking_duration) {
+ taa_data->thread_blocking_duration = duration;
+
+ if (comm)
+ strncpy(taa_data->run_thread_comm, comm, MAX_COMM);
+ else
+ sprintf(taa_data->run_thread_comm, "<...>");
+
+ } else {
+ taa_data->thread_thread_sum += duration;
+
+ trace_seq_printf(taa_data->threads_seq, "\t%24s:%-3llu \t\t%9.2f us\n",
+ comm, pid, ns_to_usf(duration));
+ }
+
+ return 0;
+}
+
+/*
+ * timerlat_aa_stack_handler - Handles timerlat IRQ stack trace
+ *
+ * Saves and parse the stack trace generated by the timerlat IRQ.
+ */
+static int timerlat_aa_stack_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ struct timerlat_aa_data *taa_data = timerlat_aa_get_data(taa_ctx, record->cpu);
+ unsigned long *caller;
+ const char *function;
+ int val, i;
+
+ trace_seq_reset(taa_data->stack_seq);
+
+ trace_seq_printf(taa_data->stack_seq, " Blocking thread stack trace\n");
+ caller = tep_get_field_raw(s, event, "caller", record, &val, 1);
+ if (caller) {
+ for (i = 0; ; i++) {
+ function = tep_find_function(taa_ctx->tool->trace.tep, caller[i]);
+ if (!function)
+ break;
+ trace_seq_printf(taa_data->stack_seq, "\t\t-> %s\n", function);
+ }
+ }
+ return 0;
+}
+
+/*
+ * timerlat_aa_sched_switch_handler - Tracks the current thread running on the CPU
+ *
+ * Handles the sched:sched_switch event to trace the current thread running on the
+ * CPU. It is used to display the threads running on the other CPUs when the trace
+ * stops.
+ */
+static int timerlat_aa_sched_switch_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ struct timerlat_aa_data *taa_data = timerlat_aa_get_data(taa_ctx, record->cpu);
+ const char *comm;
+ int val;
+
+ tep_get_field_val(s, event, "next_pid", record, &taa_data->current_pid, 1);
+ comm = tep_get_field_raw(s, event, "next_comm", record, &val, 1);
+
+ strncpy(taa_data->current_comm, comm, MAX_COMM);
+
+ /*
+ * If this was a kworker, clean the last kworkers that ran.
+ */
+ taa_data->kworker = 0;
+ taa_data->kworker_func = 0;
+
+ return 0;
+}
+
+/*
+ * timerlat_aa_kworker_start_handler - Tracks a kworker running on the CPU
+ *
+ * Handles workqueue:workqueue_execute_start event, keeping track of
+ * the job that a kworker could be doing in the CPU.
+ *
+ * We already catch problems of hardware related latencies caused by work queues
+ * running driver code that causes hardware stall. For example, with DRM drivers.
+ */
+static int timerlat_aa_kworker_start_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ struct timerlat_aa_data *taa_data = timerlat_aa_get_data(taa_ctx, record->cpu);
+
+ tep_get_field_val(s, event, "work", record, &taa_data->kworker, 1);
+ tep_get_field_val(s, event, "function", record, &taa_data->kworker_func, 1);
+ return 0;
+}
+
+/*
+ * timerlat_thread_analysis - Prints the analysis of a CPU that hit a stop tracing
+ *
+ * This is the core of the analysis.
+ */
+static void timerlat_thread_analysis(struct timerlat_aa_data *taa_data, int cpu,
+ int irq_thresh, int thread_thresh)
+{
+ unsigned long long exp_irq_ts;
+ int total;
+ int irq;
+
+ /*
+ * IRQ latency or Thread latency?
+ */
+ if (taa_data->tlat_irq_seqnum > taa_data->tlat_thread_seqnum) {
+ irq = 1;
+ total = taa_data->tlat_irq_latency;
+ } else {
+ irq = 0;
+ total = taa_data->tlat_thread_latency;
+ }
+
+ /*
+ * Expected IRQ arrival time using the trace clock as the base.
+ */
+ exp_irq_ts = taa_data->timer_irq_start_time - taa_data->timer_irq_start_delay;
+
+ if (exp_irq_ts < taa_data->prev_irq_timstamp + taa_data->prev_irq_duration)
+ printf(" Previous IRQ interference: \t up to %9.2f us",
+ ns_to_usf(taa_data->prev_irq_duration));
+
+ /*
+ * The delay that the IRQ suffered before starting.
+ */
+ printf(" IRQ handler delay: %16s %9.2f us (%.2f %%)\n",
+ (ns_to_usf(taa_data->timer_exit_from_idle) > 10) ? "(exit from idle)" : "",
+ ns_to_usf(taa_data->timer_irq_start_delay),
+ ns_to_per(total, taa_data->timer_irq_start_delay));
+
+ /*
+ * Timerlat IRQ.
+ */
+ printf(" IRQ latency: \t\t\t\t %9.2f us\n",
+ ns_to_usf(taa_data->tlat_irq_latency));
+
+ if (irq) {
+ /*
+ * If the trace stopped due to IRQ, the other events will not happen
+ * because... the trace stopped :-).
+ *
+ * That is all folks, the stack trace was printed before the stop,
+ * so it will be displayed, it is the key.
+ */
+ printf(" Blocking thread:\n");
+ printf(" %24s:%-9llu\n",
+ taa_data->run_thread_comm, taa_data->run_thread_pid);
+ } else {
+ /*
+ * The duration of the IRQ handler that handled the timerlat IRQ.
+ */
+ printf(" Timerlat IRQ duration: \t\t %9.2f us (%.2f %%)\n",
+ ns_to_usf(taa_data->timer_irq_duration),
+ ns_to_per(total, taa_data->timer_irq_duration));
+
+ /*
+ * The amount of time that the current thread postponed the scheduler.
+ *
+ * Recalling that it is net from NMI/IRQ/Softirq interference, so there
+ * is no need to compute values here.
+ */
+ printf(" Blocking thread: \t\t\t %9.2f us (%.2f %%)\n",
+ ns_to_usf(taa_data->thread_blocking_duration),
+ ns_to_per(total, taa_data->thread_blocking_duration));
+
+ printf(" %24s:%-9llu %9.2f us\n",
+ taa_data->run_thread_comm, taa_data->run_thread_pid,
+ ns_to_usf(taa_data->thread_blocking_duration));
+ }
+
+ /*
+ * Print the stack trace!
+ */
+ trace_seq_do_printf(taa_data->stack_seq);
+
+ /*
+ * NMIs can happen during the IRQ, so they are always possible.
+ */
+ if (taa_data->thread_nmi_sum)
+ printf(" NMI interference \t\t\t %9.2f us (%.2f %%)\n",
+ ns_to_usf(taa_data->thread_nmi_sum),
+ ns_to_per(total, taa_data->thread_nmi_sum));
+
+ /*
+ * If it is an IRQ latency, the other factors can be skipped.
+ */
+ if (irq)
+ goto print_total;
+
+ /*
+ * Prints the interference caused by IRQs to the thread latency.
+ */
+ if (taa_data->thread_irq_sum) {
+ printf(" IRQ interference \t\t\t %9.2f us (%.2f %%)\n",
+ ns_to_usf(taa_data->thread_irq_sum),
+ ns_to_per(total, taa_data->thread_irq_sum));
+
+ trace_seq_do_printf(taa_data->irqs_seq);
+ }
+
+ /*
+ * Prints the interference caused by Softirqs to the thread latency.
+ */
+ if (taa_data->thread_softirq_sum) {
+ printf(" Softirq interference \t\t\t %9.2f us (%.2f %%)\n",
+ ns_to_usf(taa_data->thread_softirq_sum),
+ ns_to_per(total, taa_data->thread_softirq_sum));
+
+ trace_seq_do_printf(taa_data->softirqs_seq);
+ }
+
+ /*
+ * Prints the interference caused by other threads to the thread latency.
+ *
+ * If this happens, your timerlat is not the highest prio. OK, migration
+ * thread can happen. But otherwise, you are not measuring the "scheduling
+ * latency" only, and here is the difference from scheduling latency and
+ * timer handling latency.
+ */
+ if (taa_data->thread_thread_sum) {
+ printf(" Thread interference \t\t\t %9.2f us (%.2f %%)\n",
+ ns_to_usf(taa_data->thread_thread_sum),
+ ns_to_per(total, taa_data->thread_thread_sum));
+
+ trace_seq_do_printf(taa_data->threads_seq);
+ }
+
+ /*
+ * Done.
+ */
+print_total:
+ printf("------------------------------------------------------------------------\n");
+ printf(" %s latency: \t\t\t %9.2f us (100%%)\n", irq ? "IRQ" : "Thread",
+ ns_to_usf(total));
+}
+
+/**
+ * timerlat_auto_analysis - Analyze the collected data
+ */
+void timerlat_auto_analysis(int irq_thresh, int thread_thresh)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+ unsigned long long max_exit_from_idle = 0;
+ struct timerlat_aa_data *taa_data;
+ int max_exit_from_idle_cpu;
+ struct tep_handle *tep;
+ int cpu;
+
+ /* bring stop tracing to the ns scale */
+ irq_thresh = irq_thresh * 1000;
+ thread_thresh = thread_thresh * 1000;
+
+ for (cpu = 0; cpu < taa_ctx->nr_cpus; cpu++) {
+ taa_data = timerlat_aa_get_data(taa_ctx, cpu);
+
+ if (irq_thresh && taa_data->tlat_irq_latency >= irq_thresh) {
+ printf("## CPU %d hit stop tracing, analyzing it ##\n", cpu);
+ timerlat_thread_analysis(taa_data, cpu, irq_thresh, thread_thresh);
+ } else if (thread_thresh && (taa_data->tlat_thread_latency) >= thread_thresh) {
+ printf("## CPU %d hit stop tracing, analyzing it ##\n", cpu);
+ timerlat_thread_analysis(taa_data, cpu, irq_thresh, thread_thresh);
+ }
+
+ if (taa_data->max_exit_idle_latency > max_exit_from_idle) {
+ max_exit_from_idle = taa_data->max_exit_idle_latency;
+ max_exit_from_idle_cpu = cpu;
+ }
+
+ }
+
+ if (max_exit_from_idle) {
+ printf("\n");
+ printf("Max timerlat IRQ latency from idle: %.2f us in cpu %d\n",
+ ns_to_usf(max_exit_from_idle), max_exit_from_idle_cpu);
+ }
+ if (!taa_ctx->dump_tasks)
+ return;
+
+ printf("\n");
+ printf("Printing CPU tasks:\n");
+ for (cpu = 0; cpu < taa_ctx->nr_cpus; cpu++) {
+ taa_data = timerlat_aa_get_data(taa_ctx, cpu);
+ tep = taa_ctx->tool->trace.tep;
+
+ printf(" [%.3d] %24s:%llu", cpu, taa_data->current_comm, taa_data->current_pid);
+
+ if (taa_data->kworker_func)
+ printf(" kworker:%s:%s",
+ tep_find_function(tep, taa_data->kworker) ? : "<...>",
+ tep_find_function(tep, taa_data->kworker_func));
+ printf("\n");
+ }
+
+}
+
+/*
+ * timerlat_aa_destroy_seqs - Destroy seq files used to store parsed data
+ */
+static void timerlat_aa_destroy_seqs(struct timerlat_aa_context *taa_ctx)
+{
+ struct timerlat_aa_data *taa_data;
+ int i;
+
+ if (!taa_ctx->taa_data)
+ return;
+
+ for (i = 0; i < taa_ctx->nr_cpus; i++) {
+ taa_data = timerlat_aa_get_data(taa_ctx, i);
+
+ if (taa_data->prev_irqs_seq) {
+ trace_seq_destroy(taa_data->prev_irqs_seq);
+ free(taa_data->prev_irqs_seq);
+ }
+
+ if (taa_data->nmi_seq) {
+ trace_seq_destroy(taa_data->nmi_seq);
+ free(taa_data->nmi_seq);
+ }
+
+ if (taa_data->irqs_seq) {
+ trace_seq_destroy(taa_data->irqs_seq);
+ free(taa_data->irqs_seq);
+ }
+
+ if (taa_data->softirqs_seq) {
+ trace_seq_destroy(taa_data->softirqs_seq);
+ free(taa_data->softirqs_seq);
+ }
+
+ if (taa_data->threads_seq) {
+ trace_seq_destroy(taa_data->threads_seq);
+ free(taa_data->threads_seq);
+ }
+
+ if (taa_data->stack_seq) {
+ trace_seq_destroy(taa_data->stack_seq);
+ free(taa_data->stack_seq);
+ }
+ }
+}
+
+/*
+ * timerlat_aa_init_seqs - Init seq files used to store parsed information
+ *
+ * Instead of keeping data structures to store raw data, use seq files to
+ * store parsed data.
+ *
+ * Allocates and initialize seq files.
+ *
+ * Returns 0 on success, -1 otherwise.
+ */
+static int timerlat_aa_init_seqs(struct timerlat_aa_context *taa_ctx)
+{
+ struct timerlat_aa_data *taa_data;
+ int i;
+
+ for (i = 0; i < taa_ctx->nr_cpus; i++) {
+
+ taa_data = timerlat_aa_get_data(taa_ctx, i);
+
+ taa_data->prev_irqs_seq = calloc(1, sizeof(*taa_data->prev_irqs_seq));
+ if (!taa_data->prev_irqs_seq)
+ goto out_err;
+
+ trace_seq_init(taa_data->prev_irqs_seq);
+
+ taa_data->nmi_seq = calloc(1, sizeof(*taa_data->nmi_seq));
+ if (!taa_data->nmi_seq)
+ goto out_err;
+
+ trace_seq_init(taa_data->nmi_seq);
+
+ taa_data->irqs_seq = calloc(1, sizeof(*taa_data->irqs_seq));
+ if (!taa_data->irqs_seq)
+ goto out_err;
+
+ trace_seq_init(taa_data->irqs_seq);
+
+ taa_data->softirqs_seq = calloc(1, sizeof(*taa_data->softirqs_seq));
+ if (!taa_data->softirqs_seq)
+ goto out_err;
+
+ trace_seq_init(taa_data->softirqs_seq);
+
+ taa_data->threads_seq = calloc(1, sizeof(*taa_data->threads_seq));
+ if (!taa_data->threads_seq)
+ goto out_err;
+
+ trace_seq_init(taa_data->threads_seq);
+
+ taa_data->stack_seq = calloc(1, sizeof(*taa_data->stack_seq));
+ if (!taa_data->stack_seq)
+ goto out_err;
+
+ trace_seq_init(taa_data->stack_seq);
+ }
+
+ return 0;
+
+out_err:
+ timerlat_aa_destroy_seqs(taa_ctx);
+ return -1;
+}
+
+/*
+ * timerlat_aa_unregister_events - Unregister events used in the auto-analysis
+ */
+static void timerlat_aa_unregister_events(struct osnoise_tool *tool, int dump_tasks)
+{
+ tracefs_event_disable(tool->trace.inst, "osnoise", NULL);
+
+ tep_unregister_event_handler(tool->trace.tep, -1, "osnoise", "nmi_noise",
+ timerlat_aa_nmi_handler, tool);
+
+ tep_unregister_event_handler(tool->trace.tep, -1, "osnoise", "irq_noise",
+ timerlat_aa_irq_handler, tool);
+
+ tep_unregister_event_handler(tool->trace.tep, -1, "osnoise", "softirq_noise",
+ timerlat_aa_softirq_handler, tool);
+
+ tep_unregister_event_handler(tool->trace.tep, -1, "osnoise", "thread_noise",
+ timerlat_aa_thread_handler, tool);
+
+ tep_unregister_event_handler(tool->trace.tep, -1, "ftrace", "kernel_stack",
+ timerlat_aa_stack_handler, tool);
+ if (!dump_tasks)
+ return;
+
+ tracefs_event_disable(tool->trace.inst, "sched", "sched_switch");
+ tep_unregister_event_handler(tool->trace.tep, -1, "sched", "sched_switch",
+ timerlat_aa_sched_switch_handler, tool);
+
+ tracefs_event_disable(tool->trace.inst, "workqueue", "workqueue_execute_start");
+ tep_unregister_event_handler(tool->trace.tep, -1, "workqueue", "workqueue_execute_start",
+ timerlat_aa_kworker_start_handler, tool);
+}
+
+/*
+ * timerlat_aa_register_events - Register events used in the auto-analysis
+ *
+ * Returns 0 on success, -1 otherwise.
+ */
+static int timerlat_aa_register_events(struct osnoise_tool *tool, int dump_tasks)
+{
+ int retval;
+
+ /*
+ * register auto-analysis handlers.
+ */
+ retval = tracefs_event_enable(tool->trace.inst, "osnoise", NULL);
+ if (retval < 0 && !errno) {
+ err_msg("Could not find osnoise events\n");
+ goto out_err;
+ }
+
+ tep_register_event_handler(tool->trace.tep, -1, "osnoise", "nmi_noise",
+ timerlat_aa_nmi_handler, tool);
+
+ tep_register_event_handler(tool->trace.tep, -1, "osnoise", "irq_noise",
+ timerlat_aa_irq_handler, tool);
+
+ tep_register_event_handler(tool->trace.tep, -1, "osnoise", "softirq_noise",
+ timerlat_aa_softirq_handler, tool);
+
+ tep_register_event_handler(tool->trace.tep, -1, "osnoise", "thread_noise",
+ timerlat_aa_thread_handler, tool);
+
+ tep_register_event_handler(tool->trace.tep, -1, "ftrace", "kernel_stack",
+ timerlat_aa_stack_handler, tool);
+
+ if (!dump_tasks)
+ return 0;
+
+ /*
+ * Dump task events.
+ */
+ retval = tracefs_event_enable(tool->trace.inst, "sched", "sched_switch");
+ if (retval < 0 && !errno) {
+ err_msg("Could not find sched_switch\n");
+ goto out_err;
+ }
+
+ tep_register_event_handler(tool->trace.tep, -1, "sched", "sched_switch",
+ timerlat_aa_sched_switch_handler, tool);
+
+ retval = tracefs_event_enable(tool->trace.inst, "workqueue", "workqueue_execute_start");
+ if (retval < 0 && !errno) {
+ err_msg("Could not find workqueue_execute_start\n");
+ goto out_err;
+ }
+
+ tep_register_event_handler(tool->trace.tep, -1, "workqueue", "workqueue_execute_start",
+ timerlat_aa_kworker_start_handler, tool);
+
+ return 0;
+
+out_err:
+ timerlat_aa_unregister_events(tool, dump_tasks);
+ return -1;
+}
+
+/**
+ * timerlat_aa_destroy - Destroy timerlat auto-analysis
+ */
+void timerlat_aa_destroy(void)
+{
+ struct timerlat_aa_context *taa_ctx = timerlat_aa_get_ctx();
+
+ if (!taa_ctx)
+ return;
+
+ if (!taa_ctx->taa_data)
+ goto out_ctx;
+
+ timerlat_aa_unregister_events(taa_ctx->tool, taa_ctx->dump_tasks);
+ timerlat_aa_destroy_seqs(taa_ctx);
+ free(taa_ctx->taa_data);
+out_ctx:
+ free(taa_ctx);
+}
+
+/**
+ * timerlat_aa_init - Initialize timerlat auto-analysis
+ *
+ * Returns 0 on success, -1 otherwise.
+ */
+int timerlat_aa_init(struct osnoise_tool *tool, int nr_cpus, int dump_tasks)
+{
+ struct timerlat_aa_context *taa_ctx;
+ int retval;
+
+ taa_ctx = calloc(1, sizeof(*taa_ctx));
+ if (!taa_ctx)
+ return -1;
+
+ __timerlat_aa_ctx = taa_ctx;
+
+ taa_ctx->nr_cpus = nr_cpus;
+ taa_ctx->tool = tool;
+ taa_ctx->dump_tasks = dump_tasks;
+
+ taa_ctx->taa_data = calloc(nr_cpus, sizeof(*taa_ctx->taa_data));
+ if (!taa_ctx->taa_data)
+ goto out_err;
+
+ retval = timerlat_aa_init_seqs(taa_ctx);
+ if (retval)
+ goto out_err;
+
+ retval = timerlat_aa_register_events(tool, dump_tasks);
+ if (retval)
+ goto out_err;
+
+ return 0;
+
+out_err:
+ timerlat_aa_destroy();
+ return -1;
+}
diff --git a/tools/tracing/rtla/src/timerlat_aa.h b/tools/tracing/rtla/src/timerlat_aa.h
new file mode 100644
index 000000000000..d4f6ca7e342a
--- /dev/null
+++ b/tools/tracing/rtla/src/timerlat_aa.h
@@ -0,0 +1,12 @@
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * Copyright (C) 2023 Red Hat Inc, Daniel Bristot de Oliveira <bristot@kernel.org>
+ */
+
+int timerlat_aa_init(struct osnoise_tool *tool, int nr_cpus, int dump_task);
+void timerlat_aa_destroy(void);
+
+int timerlat_aa_handler(struct trace_seq *s, struct tep_record *record,
+ struct tep_event *event, void *context);
+
+void timerlat_auto_analysis(int irq_thresh, int thread_thresh);
diff --git a/tools/tracing/rtla/src/timerlat_top.c b/tools/tracing/rtla/src/timerlat_top.c
index 334271935222..eea5b3357e27 100644
--- a/tools/tracing/rtla/src/timerlat_top.c
+++ b/tools/tracing/rtla/src/timerlat_top.c
@@ -10,10 +10,12 @@
#include <unistd.h>
#include <stdio.h>
#include <time.h>
+#include <errno.h>
#include "utils.h"
#include "osnoise.h"
#include "timerlat.h"
+#include "timerlat_aa.h"
struct timerlat_top_params {
char *cpus;
@@ -30,6 +32,8 @@ struct timerlat_top_params {
int quiet;
int set_sched;
int dma_latency;
+ int no_aa;
+ int dump_tasks;
struct sched_attr sched_param;
struct trace_events *events;
};
@@ -130,17 +134,22 @@ timerlat_top_handler(struct trace_seq *s, struct tep_record *record,
struct tep_event *event, void *context)
{
struct trace_instance *trace = context;
+ struct timerlat_top_params *params;
unsigned long long latency, thread;
struct osnoise_tool *top;
int cpu = record->cpu;
top = container_of(trace, struct osnoise_tool, trace);
+ params = top->params;
tep_get_field_val(s, event, "context", record, &thread, 1);
tep_get_field_val(s, event, "timer_latency", record, &latency, 1);
timerlat_top_update(top, cpu, thread, latency);
+ if (!params->no_aa)
+ timerlat_aa_handler(s, record, event, context);
+
return 0;
}
@@ -281,11 +290,13 @@ static void timerlat_top_usage(char *usage)
" -c/--cpus cpus: run the tracer only on the given cpus",
" -d/--duration time[m|h|d]: duration of the session in seconds",
" -D/--debug: print debug info",
+ " --dump-tasks: prints the task running on all CPUs if stop conditions are met (depends on !--no-aa)",
" -t/--trace[=file]: save the stopped trace to [file|timerlat_trace.txt]",
" -e/--event <sys:event>: enable the <sys:event> in the trace instance, multiple -e are allowed",
" --filter <command>: enable a trace event filter to the previous -e event",
" --trigger <command>: enable a trace event trigger to the previous -e event",
" -n/--nano: display data in nanoseconds",
+ " --no-aa: disable auto-analysis, reducing rtla timerlat cpu usage",
" -q/--quiet print only a summary at the end",
" --dma-latency us: set /dev/cpu_dma_latency latency <us> to reduce exit from idle latency",
" -P/--priority o:prio|r:prio|f:prio|d:runtime:period : set scheduling parameters",
@@ -349,13 +360,15 @@ static struct timerlat_top_params
{"trigger", required_argument, 0, '0'},
{"filter", required_argument, 0, '1'},
{"dma-latency", required_argument, 0, '2'},
+ {"no-aa", no_argument, 0, '3'},
+ {"dump-tasks", no_argument, 0, '4'},
{0, 0, 0, 0}
};
/* getopt_long stores the option index here. */
int option_index = 0;
- c = getopt_long(argc, argv, "a:c:d:De:hi:np:P:qs:t::T:0:1:2:",
+ c = getopt_long(argc, argv, "a:c:d:De:hi:np:P:qs:t::T:0:1:2:34",
long_options, &option_index);
/* detect the end of the options. */
@@ -368,13 +381,13 @@ static struct timerlat_top_params
/* set thread stop to auto_thresh */
params->stop_total_us = auto_thresh;
+ params->stop_us = auto_thresh;
/* get stack trace */
params->print_stack = auto_thresh;
/* set trace */
params->trace_output = "timerlat_trace.txt";
-
break;
case 'c':
retval = parse_cpu_list(optarg, &params->monitored_cpus);
@@ -437,6 +450,7 @@ static struct timerlat_top_params
params->trace_output = &optarg[1];
else
params->trace_output = "timerlat_trace.txt";
+
break;
case '0': /* trigger */
if (params->events) {
@@ -467,6 +481,12 @@ static struct timerlat_top_params
exit(EXIT_FAILURE);
}
break;
+ case '3': /* no-aa */
+ params->no_aa = 1;
+ break;
+ case '4':
+ params->dump_tasks = 1;
+ break;
default:
timerlat_top_usage("Invalid option");
}
@@ -477,6 +497,12 @@ static struct timerlat_top_params
exit(EXIT_FAILURE);
}
+ /*
+ * Auto analysis only happens if stop tracing, thus:
+ */
+ if (!params->stop_us && !params->stop_total_us)
+ params->no_aa = 1;
+
return params;
}
@@ -547,6 +573,7 @@ static struct osnoise_tool
{
struct osnoise_tool *top;
int nr_cpus;
+ int retval;
nr_cpus = sysconf(_SC_NPROCESSORS_CONF);
@@ -563,6 +590,16 @@ static struct osnoise_tool
tep_register_event_handler(top->trace.tep, -1, "ftrace", "timerlat",
timerlat_top_handler, top);
+ /*
+ * If no auto analysis, we are ready.
+ */
+ if (params->no_aa)
+ return top;
+
+ retval = timerlat_aa_init(top, nr_cpus, params->dump_tasks);
+ if (retval)
+ goto out_err;
+
return top;
out_err:
@@ -688,6 +725,10 @@ int timerlat_top_main(int argc, char *argv[])
if (trace_is_off(&top->trace, &record->trace)) {
printf("rtla timerlat hit stop tracing\n");
+
+ if (!params->no_aa)
+ timerlat_auto_analysis(params->stop_us, params->stop_total_us);
+
if (params->trace_output) {
printf(" Saving trace to %s\n", params->trace_output);
save_trace_to_file(record->trace.inst, params->trace_output);
@@ -701,6 +742,7 @@ out_top:
params->events = NULL;
out_free:
timerlat_free_top(top->data);
+ timerlat_aa_destroy();
osnoise_destroy_tool(record);
osnoise_destroy_tool(top);
free(params);
diff --git a/tools/tracing/rtla/src/utils.h b/tools/tracing/rtla/src/utils.h
index 5571afd3b549..90e4f52a030b 100644
--- a/tools/tracing/rtla/src/utils.h
+++ b/tools/tracing/rtla/src/utils.h
@@ -56,3 +56,6 @@ struct sched_attr {
int parse_prio(char *arg, struct sched_attr *sched_param);
int set_comm_sched_attr(const char *comm_prefix, struct sched_attr *attr);
int set_cpu_dma_latency(int32_t latency);
+
+#define ns_to_usf(x) (((double)x/1000))
+#define ns_to_per(total, part) ((part * 100) / (double)total)