From 6fa5e3a87cd7838453be66c3a69c2236a1680504 Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 18 Feb 2025 15:58:52 +0100 Subject: rtla/timerlat: Unify params struct Instead of having separate structs timerlat_top_params and timerlat_hist_params, use one struct timerlat_params for both. This allows code using the structs to be shared between timerlat-top and timerlat-hist. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Cc: Clark Williams Link: https://lore.kernel.org/20250218145859.27762-2-tglozar@redhat.com Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/tracing/rtla/src/osnoise.h | 2 ++ tools/tracing/rtla/src/timerlat.h | 54 +++++++++++++++++++++++++++++ tools/tracing/rtla/src/timerlat_aa.c | 2 -- tools/tracing/rtla/src/timerlat_hist.c | 62 ++++++---------------------------- tools/tracing/rtla/src/timerlat_top.c | 57 ++++++------------------------- 5 files changed, 78 insertions(+), 99 deletions(-) (limited to 'tools') diff --git a/tools/tracing/rtla/src/osnoise.h b/tools/tracing/rtla/src/osnoise.h index 91835a7d8c2b..056c8b113dee 100644 --- a/tools/tracing/rtla/src/osnoise.h +++ b/tools/tracing/rtla/src/osnoise.h @@ -1,4 +1,6 @@ // SPDX-License-Identifier: GPL-2.0 +#pragma once + #include "trace.h" /* diff --git a/tools/tracing/rtla/src/timerlat.h b/tools/tracing/rtla/src/timerlat.h index 88561bfd14f3..e452d385cb0f 100644 --- a/tools/tracing/rtla/src/timerlat.h +++ b/tools/tracing/rtla/src/timerlat.h @@ -1,4 +1,58 @@ // SPDX-License-Identifier: GPL-2.0 +#include "utils.h" +#include "osnoise.h" + +struct timerlat_params { + /* Common params */ + char *cpus; + cpu_set_t monitored_cpus; + char *trace_output; + char *cgroup_name; + unsigned long long runtime; + long long stop_us; + long long stop_total_us; + long long timerlat_period_us; + long long print_stack; + int sleep_time; + int output_divisor; + int duration; + int quiet; + int set_sched; + int dma_latency; + int no_aa; + int aa_only; + int dump_tasks; + int cgroup; + int hk_cpus; + int user_workload; + int kernel_workload; + int pretty_output; + int warmup; + int buffer_size; + int deepest_idle_state; + cpu_set_t hk_cpu_set; + struct sched_attr sched_param; + struct trace_events *events; + union { + struct { + /* top only */ + int user_top; + }; + struct { + /* hist only */ + int user_hist; + char no_irq; + char no_thread; + char no_header; + char no_summary; + char no_index; + char with_zeros; + int bucket_size; + int entries; + }; + }; +}; + int timerlat_hist_main(int argc, char *argv[]); int timerlat_top_main(int argc, char *argv[]); int timerlat_main(int argc, char *argv[]); diff --git a/tools/tracing/rtla/src/timerlat_aa.c b/tools/tracing/rtla/src/timerlat_aa.c index 7bd80ee2a5b4..31e66ea2b144 100644 --- a/tools/tracing/rtla/src/timerlat_aa.c +++ b/tools/tracing/rtla/src/timerlat_aa.c @@ -5,8 +5,6 @@ #include #include -#include "utils.h" -#include "osnoise.h" #include "timerlat.h" #include diff --git a/tools/tracing/rtla/src/timerlat_hist.c b/tools/tracing/rtla/src/timerlat_hist.c index 6d7d0a2d45b4..fe683a20b0e2 100644 --- a/tools/tracing/rtla/src/timerlat_hist.c +++ b/tools/tracing/rtla/src/timerlat_hist.c @@ -14,50 +14,10 @@ #include #include -#include "utils.h" -#include "osnoise.h" #include "timerlat.h" #include "timerlat_aa.h" #include "timerlat_u.h" -struct timerlat_hist_params { - char *cpus; - cpu_set_t monitored_cpus; - char *trace_output; - char *cgroup_name; - unsigned long long runtime; - long long stop_us; - long long stop_total_us; - long long timerlat_period_us; - long long print_stack; - int sleep_time; - int output_divisor; - int duration; - int set_sched; - int dma_latency; - int cgroup; - int hk_cpus; - int no_aa; - int dump_tasks; - int user_workload; - int kernel_workload; - int user_hist; - cpu_set_t hk_cpu_set; - struct sched_attr sched_param; - struct trace_events *events; - char no_irq; - char no_thread; - char no_header; - char no_summary; - char no_index; - char with_zeros; - int bucket_size; - int entries; - int warmup; - int buffer_size; - int deepest_idle_state; -}; - struct timerlat_hist_cpu { int *irq; int *thread; @@ -174,7 +134,7 @@ timerlat_hist_update(struct osnoise_tool *tool, int cpu, unsigned long long context, unsigned long long latency) { - struct timerlat_hist_params *params = tool->params; + struct timerlat_params *params = tool->params; struct timerlat_hist_data *data = tool->data; int entries = data->entries; int bucket; @@ -238,7 +198,7 @@ timerlat_hist_handler(struct trace_seq *s, struct tep_record *record, */ static void timerlat_hist_header(struct osnoise_tool *tool) { - struct timerlat_hist_params *params = tool->params; + struct timerlat_params *params = tool->params; struct timerlat_hist_data *data = tool->data; struct trace_seq *s = tool->trace.seq; char duration[26]; @@ -300,7 +260,7 @@ static void format_summary_value(struct trace_seq *seq, * timerlat_print_summary - print the summary of the hist data to the output */ static void -timerlat_print_summary(struct timerlat_hist_params *params, +timerlat_print_summary(struct timerlat_params *params, struct trace_instance *trace, struct timerlat_hist_data *data) { @@ -427,7 +387,7 @@ timerlat_print_summary(struct timerlat_hist_params *params, } static void -timerlat_print_stats_all(struct timerlat_hist_params *params, +timerlat_print_stats_all(struct timerlat_params *params, struct trace_instance *trace, struct timerlat_hist_data *data) { @@ -575,7 +535,7 @@ timerlat_print_stats_all(struct timerlat_hist_params *params, * timerlat_print_stats - print data for each CPUs */ static void -timerlat_print_stats(struct timerlat_hist_params *params, struct osnoise_tool *tool) +timerlat_print_stats(struct timerlat_params *params, struct osnoise_tool *tool) { struct timerlat_hist_data *data = tool->data; struct trace_instance *trace = &tool->trace; @@ -734,10 +694,10 @@ static void timerlat_hist_usage(char *usage) /* * timerlat_hist_parse_args - allocs, parse and fill the cmd line parameters */ -static struct timerlat_hist_params +static struct timerlat_params *timerlat_hist_parse_args(int argc, char *argv[]) { - struct timerlat_hist_params *params; + struct timerlat_params *params; struct trace_events *tevent; int auto_thresh; int retval; @@ -1017,7 +977,7 @@ static struct timerlat_hist_params * timerlat_hist_apply_config - apply the hist configs to the initialized tool */ static int -timerlat_hist_apply_config(struct osnoise_tool *tool, struct timerlat_hist_params *params) +timerlat_hist_apply_config(struct osnoise_tool *tool, struct timerlat_params *params) { int retval, i; @@ -1122,7 +1082,7 @@ out_err: * timerlat_init_hist - initialize a timerlat hist tool with parameters */ static struct osnoise_tool -*timerlat_init_hist(struct timerlat_hist_params *params) +*timerlat_init_hist(struct timerlat_params *params) { struct osnoise_tool *tool; int nr_cpus; @@ -1170,7 +1130,7 @@ static void stop_hist(int sig) * timerlat_hist_set_signals - handles the signal to stop the tool */ static void -timerlat_hist_set_signals(struct timerlat_hist_params *params) +timerlat_hist_set_signals(struct timerlat_params *params) { signal(SIGINT, stop_hist); if (params->duration) { @@ -1181,7 +1141,7 @@ timerlat_hist_set_signals(struct timerlat_hist_params *params) int timerlat_hist_main(int argc, char *argv[]) { - struct timerlat_hist_params *params; + struct timerlat_params *params; struct osnoise_tool *record = NULL; struct timerlat_u_params params_u; struct osnoise_tool *tool = NULL; diff --git a/tools/tracing/rtla/src/timerlat_top.c b/tools/tracing/rtla/src/timerlat_top.c index 05a9403b01d2..97eead91decc 100644 --- a/tools/tracing/rtla/src/timerlat_top.c +++ b/tools/tracing/rtla/src/timerlat_top.c @@ -15,45 +15,10 @@ #include #include -#include "utils.h" -#include "osnoise.h" #include "timerlat.h" #include "timerlat_aa.h" #include "timerlat_u.h" -struct timerlat_top_params { - char *cpus; - cpu_set_t monitored_cpus; - char *trace_output; - char *cgroup_name; - unsigned long long runtime; - long long stop_us; - long long stop_total_us; - long long timerlat_period_us; - long long print_stack; - int sleep_time; - int output_divisor; - int duration; - int quiet; - int set_sched; - int dma_latency; - int no_aa; - int aa_only; - int dump_tasks; - int cgroup; - int hk_cpus; - int user_top; - int user_workload; - int kernel_workload; - int pretty_output; - int warmup; - int buffer_size; - int deepest_idle_state; - cpu_set_t hk_cpu_set; - struct sched_attr sched_param; - struct trace_events *events; -}; - struct timerlat_top_cpu { unsigned long long irq_count; unsigned long long thread_count; @@ -194,7 +159,7 @@ 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; + struct timerlat_params *params; unsigned long long latency, thread; struct osnoise_tool *top; int cpu = record->cpu; @@ -215,7 +180,7 @@ timerlat_top_handler(struct trace_seq *s, struct tep_record *record, /* * timerlat_top_header - print the header of the tool output */ -static void timerlat_top_header(struct timerlat_top_params *params, struct osnoise_tool *top) +static void timerlat_top_header(struct timerlat_params *params, struct osnoise_tool *top) { struct trace_seq *s = top->trace.seq; char duration[26]; @@ -263,7 +228,7 @@ static const char *no_value = " -"; static void timerlat_top_print(struct osnoise_tool *top, int cpu) { - struct timerlat_top_params *params = top->params; + struct timerlat_params *params = top->params; struct timerlat_top_data *data = top->data; struct timerlat_top_cpu *cpu_data = &data->cpu_data[cpu]; int divisor = params->output_divisor; @@ -327,7 +292,7 @@ static void timerlat_top_print_sum(struct osnoise_tool *top, struct timerlat_top_cpu *summary) { const char *split = "----------------------------------------"; - struct timerlat_top_params *params = top->params; + struct timerlat_params *params = top->params; unsigned long long count = summary->irq_count; int divisor = params->output_divisor; struct trace_seq *s = top->trace.seq; @@ -404,7 +369,7 @@ static void clear_terminal(struct trace_seq *seq) * timerlat_print_stats - print data for all cpus */ static void -timerlat_print_stats(struct timerlat_top_params *params, struct osnoise_tool *top) +timerlat_print_stats(struct timerlat_params *params, struct osnoise_tool *top) { struct trace_instance *trace = &top->trace; struct timerlat_top_cpu summary; @@ -505,10 +470,10 @@ static void timerlat_top_usage(char *usage) /* * timerlat_top_parse_args - allocs, parse and fill the cmd line parameters */ -static struct timerlat_top_params +static struct timerlat_params *timerlat_top_parse_args(int argc, char **argv) { - struct timerlat_top_params *params; + struct timerlat_params *params; struct trace_events *tevent; long long auto_thresh; int retval; @@ -765,7 +730,7 @@ static struct timerlat_top_params * timerlat_top_apply_config - apply the top configs to the initialized tool */ static int -timerlat_top_apply_config(struct osnoise_tool *top, struct timerlat_top_params *params) +timerlat_top_apply_config(struct osnoise_tool *top, struct timerlat_params *params) { int retval; int i; @@ -876,7 +841,7 @@ out_err: * timerlat_init_top - initialize a timerlat top tool with parameters */ static struct osnoise_tool -*timerlat_init_top(struct timerlat_top_params *params) +*timerlat_init_top(struct timerlat_params *params) { struct osnoise_tool *top; int nr_cpus; @@ -924,7 +889,7 @@ static void stop_top(int sig) * timerlat_top_set_signals - handles the signal to stop the tool */ static void -timerlat_top_set_signals(struct timerlat_top_params *params) +timerlat_top_set_signals(struct timerlat_params *params) { signal(SIGINT, stop_top); if (params->duration) { @@ -935,7 +900,7 @@ timerlat_top_set_signals(struct timerlat_top_params *params) int timerlat_top_main(int argc, char *argv[]) { - struct timerlat_top_params *params; + struct timerlat_params *params; struct osnoise_tool *record = NULL; struct timerlat_u_params params_u; struct osnoise_tool *top = NULL; -- cgit v1.2.3 From 8a635c3856ddb74ed3fe7c856b271cdfeb65f293 Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 18 Feb 2025 15:58:53 +0100 Subject: tools/build: Add bpftool-skeletons feature test Add bpftool-skeletons feature test, testing the presence of a bpftool capable of generating skeletons. This is to be used for tools that do not require building their own bootstrap bpftool from the kernel source tree. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Cc: Clark Williams Link: https://lore.kernel.org/20250218145859.27762-3-tglozar@redhat.com Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/build/Makefile.feature | 3 ++- tools/build/feature/Makefile | 3 +++ tools/scripts/Makefile.include | 3 +++ 3 files changed, 8 insertions(+), 1 deletion(-) (limited to 'tools') diff --git a/tools/build/Makefile.feature b/tools/build/Makefile.feature index 1931b6321314..1f44ca677ad3 100644 --- a/tools/build/Makefile.feature +++ b/tools/build/Makefile.feature @@ -135,7 +135,8 @@ FEATURE_TESTS_EXTRA := \ libbpf-bpf_create_map \ libpfm4 \ libdebuginfod \ - clang-bpf-co-re + clang-bpf-co-re \ + bpftool-skeletons FEATURE_TESTS ?= $(FEATURE_TESTS_BASIC) diff --git a/tools/build/feature/Makefile b/tools/build/feature/Makefile index cb1e3e2feedf..4f9c1d950f5d 100644 --- a/tools/build/feature/Makefile +++ b/tools/build/feature/Makefile @@ -418,6 +418,9 @@ $(OUTPUT)test-file-handle.bin: $(OUTPUT)test-libpfm4.bin: $(BUILD) -lpfm +$(OUTPUT)test-bpftool-skeletons.bin: + $(BPFTOOL) version | grep '^features:.*skeletons' \ + > $(@:.bin=.make.output) 2>&1 ############################### clean: diff --git a/tools/scripts/Makefile.include b/tools/scripts/Makefile.include index 0aa4005017c7..71bbe52721b3 100644 --- a/tools/scripts/Makefile.include +++ b/tools/scripts/Makefile.include @@ -91,6 +91,9 @@ LLVM_CONFIG ?= llvm-config LLVM_OBJCOPY ?= llvm-objcopy LLVM_STRIP ?= llvm-strip +# Some tools require bpftool +BPFTOOL ?= bpftool + ifeq ($(CC_NO_CLANG), 1) EXTRA_WARNINGS += -Wstrict-aliasing=3 -- cgit v1.2.3 From 9dc3766ed07c95c9a77fa98dcbc83dcb7f49df3d Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 18 Feb 2025 15:58:54 +0100 Subject: rtla: Add optional dependency on BPF tooling If tooling required for building BPF CO-RE skeletons is present (that is, libbpf, clang with BPF CO-RE support, and bpftool), turn on HAVE_BPF_SKEL flag. Those requirements are similar to what perf requires, with the difference of using system libbpf and bpftool instead of in-tree versions. rtla can be forcefully built without BPF skeleton support by setting BUILD_BPF_SKEL=0 manually; in that case, a warning is displayed. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Cc: Clark Williams Link: https://lore.kernel.org/20250218145859.27762-4-tglozar@redhat.com Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/tracing/rtla/Makefile | 6 ++++++ tools/tracing/rtla/Makefile.config | 42 ++++++++++++++++++++++++++++++++++++++ 2 files changed, 48 insertions(+) (limited to 'tools') diff --git a/tools/tracing/rtla/Makefile b/tools/tracing/rtla/Makefile index 8b5101457c70..0aa116f7b355 100644 --- a/tools/tracing/rtla/Makefile +++ b/tools/tracing/rtla/Makefile @@ -33,9 +33,15 @@ DOCSRC := ../../../Documentation/tools/rtla/ FEATURE_TESTS := libtraceevent FEATURE_TESTS += libtracefs FEATURE_TESTS += libcpupower +FEATURE_TESTS += libbpf +FEATURE_TESTS += clang-bpf-co-re +FEATURE_TESTS += bpftool-skeletons FEATURE_DISPLAY := libtraceevent FEATURE_DISPLAY += libtracefs FEATURE_DISPLAY += libcpupower +FEATURE_DISPLAY += libbpf +FEATURE_DISPLAY += clang-bpf-co-re +FEATURE_DISPLAY += bpftool-skeletons ifeq ($(V),1) Q = diff --git a/tools/tracing/rtla/Makefile.config b/tools/tracing/rtla/Makefile.config index 92a6e12e42d3..5f2231d8d626 100644 --- a/tools/tracing/rtla/Makefile.config +++ b/tools/tracing/rtla/Makefile.config @@ -53,6 +53,48 @@ else $(info Please install libcpupower-dev/kernel-tools-libs-devel) endif +ifndef BUILD_BPF_SKEL + # BPF skeletons are used to implement improved sample collection, enable + # them by default. + BUILD_BPF_SKEL := 1 +endif + +ifeq ($(BUILD_BPF_SKEL),0) + $(info BPF skeleton support disabled, building without BPF skeleton support.) +endif + +$(call feature_check,libbpf) +ifeq ($(feature-libbpf), 1) + $(call detected,CONFIG_LIBBPF) +else + $(info libbpf is missing, building without BPF skeleton support.) + $(info Please install libbpf-dev/libbpf-devel) + BUILD_BPF_SKEL := 0 +endif + +$(call feature_check,clang-bpf-co-re) +ifeq ($(feature-clang-bpf-co-re), 1) + $(call detected,CONFIG_CLANG_BPF_CO_RE) +else + $(info clang is missing or does not support BPF CO-RE, building without BPF skeleton support.) + $(info Please install clang) + BUILD_BPF_SKEL := 0 +endif + +$(call feature_check,bpftool-skeletons) +ifeq ($(feature-bpftool-skeletons), 1) + $(call detected,CONFIG_BPFTOOL_SKELETONS) +else + $(info bpftool is missing or not supporting skeletons, building without BPF skeleton support.) + $(info Please install bpftool) + BUILD_BPF_SKEL := 0 +endif + +ifeq ($(BUILD_BPF_SKEL),1) + CFLAGS += -DHAVE_BPF_SKEL + EXTLIBS += -lbpf +endif + ifeq ($(STOP_ERROR),1) $(error Please, check the errors above.) endif -- cgit v1.2.3 From e34293ddcebd6bd39ff04666f010fce8d1f5e15a Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 18 Feb 2025 15:58:55 +0100 Subject: rtla/timerlat: Add BPF skeleton to collect samples Add BPF program that attaches to the osnoise:timerlat_sample tracepoint and collects both the summary and the histogram (if requested) into BPF maps (one map of each kind per context). The program is designed to be used for both timerlat-top and timerlat-hist. If using with timerlat-top, the "entries" parameter is set to zero, which prevents the BPF program from recording histogram entries. In that case, the maps for histograms do not have to be created, as the BPF verifier will identify the code using them as unreachable. An IRQ or thread latency threshold might be supplied to stop recording if hit, similar to the timerlat tracer threshold, which stops ftrace tracing if hit. A BPF ringbuffer is used to signal threshold overflow to userspace. In aa-only mode, this is the only function of the BPF program. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Cc: Clark Williams Link: https://lore.kernel.org/20250218145859.27762-5-tglozar@redhat.com Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/tracing/rtla/.gitignore | 1 + tools/tracing/rtla/Makefile | 14 ++- tools/tracing/rtla/src/Build | 1 + tools/tracing/rtla/src/timerlat.bpf.c | 149 ++++++++++++++++++++++++++++++ tools/tracing/rtla/src/timerlat_bpf.c | 166 ++++++++++++++++++++++++++++++++++ tools/tracing/rtla/src/timerlat_bpf.h | 59 ++++++++++++ 6 files changed, 389 insertions(+), 1 deletion(-) create mode 100644 tools/tracing/rtla/src/timerlat.bpf.c create mode 100644 tools/tracing/rtla/src/timerlat_bpf.c create mode 100644 tools/tracing/rtla/src/timerlat_bpf.h (limited to 'tools') diff --git a/tools/tracing/rtla/.gitignore b/tools/tracing/rtla/.gitignore index 293f0dbb0ca2..1a394ad26cc1 100644 --- a/tools/tracing/rtla/.gitignore +++ b/tools/tracing/rtla/.gitignore @@ -4,3 +4,4 @@ rtla-static fixdep feature FEATURE-DUMP +*.skel.h diff --git a/tools/tracing/rtla/Makefile b/tools/tracing/rtla/Makefile index 0aa116f7b355..557af322be61 100644 --- a/tools/tracing/rtla/Makefile +++ b/tools/tracing/rtla/Makefile @@ -73,6 +73,17 @@ CFLAGS += $(INCLUDES) $(LIB_INCLUDES) export CFLAGS OUTPUT srctree +ifeq ($(BUILD_BPF_SKEL),1) +src/timerlat.bpf.o: src/timerlat.bpf.c + $(QUIET_CLANG)$(CLANG) -g -O2 -target bpf -c $(filter %.c,$^) -o $@ + +src/timerlat.skel.h: src/timerlat.bpf.o + $(QUIET_GENSKEL)$(BPFTOOL) gen skeleton $< > $@ +else +src/timerlat.skel.h: + $(Q)echo '/* BPF skeleton is disabled */' > src/timerlat.skel.h +endif + $(RTLA): $(RTLA_IN) $(QUIET_LINK)$(CC) $(LDFLAGS) -o $(RTLA) $(RTLA_IN) $(EXTLIBS) @@ -83,7 +94,7 @@ static: $(RTLA_IN) rtla.%: fixdep FORCE make -f $(srctree)/tools/build/Makefile.build dir=. $@ -$(RTLA_IN): fixdep FORCE +$(RTLA_IN): fixdep FORCE src/timerlat.skel.h make $(build)=rtla clean: doc_clean fixdep-clean @@ -91,6 +102,7 @@ clean: doc_clean fixdep-clean $(Q)find . -name '*.o' -delete -o -name '\.*.cmd' -delete -o -name '\.*.d' -delete $(Q)rm -f rtla rtla-static fixdep FEATURE-DUMP rtla-* $(Q)rm -rf feature + $(Q)rm -f src/timerlat.bpf.o src/timerlat.skel.h check: $(RTLA) RTLA=$(RTLA) prove -o -f tests/ .PHONY: FORCE clean check diff --git a/tools/tracing/rtla/src/Build b/tools/tracing/rtla/src/Build index dbed9e31829b..7bb7e39e391a 100644 --- a/tools/tracing/rtla/src/Build +++ b/tools/tracing/rtla/src/Build @@ -8,4 +8,5 @@ rtla-y += timerlat_top.o rtla-y += timerlat_hist.o rtla-y += timerlat_u.o rtla-y += timerlat_aa.o +rtla-y += timerlat_bpf.o rtla-y += rtla.o diff --git a/tools/tracing/rtla/src/timerlat.bpf.c b/tools/tracing/rtla/src/timerlat.bpf.c new file mode 100644 index 000000000000..96196d46e170 --- /dev/null +++ b/tools/tracing/rtla/src/timerlat.bpf.c @@ -0,0 +1,149 @@ +// SPDX-License-Identifier: GPL-2.0 +#include +#include +#include +#include "timerlat_bpf.h" + +#define nosubprog __always_inline +#define MAX_ENTRIES_DEFAULT 4096 + +char LICENSE[] SEC("license") = "GPL"; + +struct trace_event_raw_timerlat_sample { + unsigned long long timer_latency; + int context; +} __attribute__((preserve_access_index)); + +struct { + __uint(type, BPF_MAP_TYPE_PERCPU_ARRAY); + __uint(max_entries, MAX_ENTRIES_DEFAULT); + __type(key, unsigned int); + __type(value, unsigned long long); +} hist_irq SEC(".maps"), hist_thread SEC(".maps"), hist_user SEC(".maps"); + +struct { + __uint(type, BPF_MAP_TYPE_PERCPU_ARRAY); + __uint(max_entries, SUMMARY_FIELD_N); + __type(key, unsigned int); + __type(value, unsigned long long); +} summary_irq SEC(".maps"), summary_thread SEC(".maps"), summary_user SEC(".maps"); + +struct { + __uint(type, BPF_MAP_TYPE_RINGBUF); + __uint(max_entries, 1); +} signal_stop_tracing SEC(".maps"); + +/* Params to be set by rtla */ +const volatile int bucket_size = 1; +const volatile int output_divisor = 1000; +const volatile int entries = 256; +const volatile int irq_threshold; +const volatile int thread_threshold; +const volatile bool aa_only; + +int stop_tracing; + +nosubprog unsigned long long map_get(void *map, + unsigned int key) +{ + unsigned long long *value_ptr; + + value_ptr = bpf_map_lookup_elem(map, &key); + + return !value_ptr ? 0 : *value_ptr; +} + +nosubprog void map_set(void *map, + unsigned int key, + unsigned long long value) +{ + bpf_map_update_elem(map, &key, &value, BPF_ANY); +} + +nosubprog void map_increment(void *map, + unsigned int key) +{ + map_set(map, key, map_get(map, key) + 1); +} + +nosubprog void update_main_hist(void *map, + int bucket) +{ + if (entries == 0) + /* No histogram */ + return; + + if (bucket >= entries) + /* Overflow */ + return; + + map_increment(map, bucket); +} + +nosubprog void update_summary(void *map, + unsigned long long latency, + int bucket) +{ + if (aa_only) + /* Auto-analysis only, nothing to be done here */ + return; + + map_set(map, SUMMARY_CURRENT, latency); + + if (bucket >= entries) + /* Overflow */ + map_increment(map, SUMMARY_OVERFLOW); + + if (latency > map_get(map, SUMMARY_MAX)) + map_set(map, SUMMARY_MAX, latency); + + if (latency < map_get(map, SUMMARY_MIN) || map_get(map, SUMMARY_COUNT) == 0) + map_set(map, SUMMARY_MIN, latency); + + map_increment(map, SUMMARY_COUNT); + map_set(map, SUMMARY_SUM, map_get(map, SUMMARY_SUM) + latency); +} + +nosubprog void set_stop_tracing(void) +{ + int value = 0; + + /* Suppress further sample processing */ + stop_tracing = 1; + + /* Signal to userspace */ + bpf_ringbuf_output(&signal_stop_tracing, &value, sizeof(value), 0); +} + +SEC("tp/osnoise/timerlat_sample") +int handle_timerlat_sample(struct trace_event_raw_timerlat_sample *tp_args) +{ + unsigned long long latency, latency_us; + int bucket; + + if (stop_tracing) + return 0; + + latency = tp_args->timer_latency / output_divisor; + latency_us = tp_args->timer_latency / 1000; + bucket = latency / bucket_size; + + if (tp_args->context == 0) { + update_main_hist(&hist_irq, bucket); + update_summary(&summary_irq, latency, bucket); + + if (irq_threshold != 0 && latency_us >= irq_threshold) + set_stop_tracing(); + } else if (tp_args->context == 1) { + update_main_hist(&hist_thread, bucket); + update_summary(&summary_thread, latency, bucket); + + if (thread_threshold != 0 && latency_us >= thread_threshold) + set_stop_tracing(); + } else { + update_main_hist(&hist_user, bucket); + update_summary(&summary_user, latency, bucket); + } + + return 0; +} diff --git a/tools/tracing/rtla/src/timerlat_bpf.c b/tools/tracing/rtla/src/timerlat_bpf.c new file mode 100644 index 000000000000..5abee884037a --- /dev/null +++ b/tools/tracing/rtla/src/timerlat_bpf.c @@ -0,0 +1,166 @@ +// SPDX-License-Identifier: GPL-2.0 +#ifdef HAVE_BPF_SKEL +#include "timerlat.h" +#include "timerlat_bpf.h" +#include "timerlat.skel.h" + +static struct timerlat_bpf *bpf; + +/* + * timerlat_bpf_init - load and initialize BPF program to collect timerlat data + */ +int timerlat_bpf_init(struct timerlat_params *params) +{ + int err; + + debug_msg("Loading BPF program\n"); + + bpf = timerlat_bpf__open(); + if (!bpf) + return 1; + + /* Pass common options */ + bpf->rodata->output_divisor = params->output_divisor; + bpf->rodata->entries = params->entries; + bpf->rodata->irq_threshold = params->stop_us; + bpf->rodata->thread_threshold = params->stop_total_us; + bpf->rodata->aa_only = params->aa_only; + + if (params->entries != 0) { + /* Pass histogram options */ + bpf->rodata->bucket_size = params->bucket_size; + + /* Set histogram array sizes */ + bpf_map__set_max_entries(bpf->maps.hist_irq, params->entries); + bpf_map__set_max_entries(bpf->maps.hist_thread, params->entries); + bpf_map__set_max_entries(bpf->maps.hist_user, params->entries); + } else { + /* No entries, disable histogram */ + bpf_map__set_autocreate(bpf->maps.hist_irq, false); + bpf_map__set_autocreate(bpf->maps.hist_thread, false); + bpf_map__set_autocreate(bpf->maps.hist_user, false); + } + + if (params->aa_only) { + /* Auto-analysis only, disable summary */ + bpf_map__set_autocreate(bpf->maps.summary_irq, false); + bpf_map__set_autocreate(bpf->maps.summary_thread, false); + bpf_map__set_autocreate(bpf->maps.summary_user, false); + } + + /* Load and verify BPF program */ + err = timerlat_bpf__load(bpf); + if (err) { + timerlat_bpf__destroy(bpf); + return err; + } + + return 0; +} + +/* + * timerlat_bpf_attach - attach BPF program to collect timerlat data + */ +int timerlat_bpf_attach(void) +{ + debug_msg("Attaching BPF program\n"); + + return timerlat_bpf__attach(bpf); +} + +/* + * timerlat_bpf_detach - detach BPF program to collect timerlat data + */ +void timerlat_bpf_detach(void) +{ + timerlat_bpf__detach(bpf); +} + +/* + * timerlat_bpf_detach - destroy BPF program to collect timerlat data + */ +void timerlat_bpf_destroy(void) +{ + timerlat_bpf__destroy(bpf); +} + +static int handle_rb_event(void *ctx, void *data, size_t data_sz) +{ + return 0; +} + +/* + * timerlat_bpf_wait - wait until tracing is stopped or signal + */ +int timerlat_bpf_wait(int timeout) +{ + struct ring_buffer *rb; + int retval; + + rb = ring_buffer__new(bpf_map__fd(bpf->maps.signal_stop_tracing), + handle_rb_event, NULL, NULL); + retval = ring_buffer__poll(rb, timeout * 1000); + ring_buffer__free(rb); + + return retval; +} + +static int get_value(struct bpf_map *map_irq, + struct bpf_map *map_thread, + struct bpf_map *map_user, + int key, + long long *value_irq, + long long *value_thread, + long long *value_user, + int cpus) +{ + int err; + + err = bpf_map__lookup_elem(map_irq, &key, + sizeof(unsigned int), value_irq, + sizeof(long long) * cpus, 0); + if (err) + return err; + err = bpf_map__lookup_elem(map_thread, &key, + sizeof(unsigned int), value_thread, + sizeof(long long) * cpus, 0); + if (err) + return err; + err = bpf_map__lookup_elem(map_user, &key, + sizeof(unsigned int), value_user, + sizeof(long long) * cpus, 0); + if (err) + return err; + return 0; +} + +/* + * timerlat_bpf_get_hist_value - get value from BPF hist map + */ +int timerlat_bpf_get_hist_value(int key, + long long *value_irq, + long long *value_thread, + long long *value_user, + int cpus) +{ + return get_value(bpf->maps.hist_irq, + bpf->maps.hist_thread, + bpf->maps.hist_user, + key, value_irq, value_thread, value_user, cpus); +} + +/* + * timerlat_bpf_get_summary_value - get value from BPF summary map + */ +int timerlat_bpf_get_summary_value(enum summary_field key, + long long *value_irq, + long long *value_thread, + long long *value_user, + int cpus) +{ + return get_value(bpf->maps.summary_irq, + bpf->maps.summary_thread, + bpf->maps.summary_user, + key, value_irq, value_thread, value_user, cpus); +} +#endif /* HAVE_BPF_SKEL */ diff --git a/tools/tracing/rtla/src/timerlat_bpf.h b/tools/tracing/rtla/src/timerlat_bpf.h new file mode 100644 index 000000000000..f1b54dbddb0e --- /dev/null +++ b/tools/tracing/rtla/src/timerlat_bpf.h @@ -0,0 +1,59 @@ +/* SPDX-License-Identifier: GPL-2.0 */ +#pragma once + +enum summary_field { + SUMMARY_CURRENT, + SUMMARY_MIN, + SUMMARY_MAX, + SUMMARY_COUNT, + SUMMARY_SUM, + SUMMARY_OVERFLOW, + SUMMARY_FIELD_N +}; + +#ifndef __bpf__ +#ifdef HAVE_BPF_SKEL +int timerlat_bpf_init(struct timerlat_params *params); +int timerlat_bpf_attach(void); +void timerlat_bpf_detach(void); +void timerlat_bpf_destroy(void); +int timerlat_bpf_wait(int timeout); +int timerlat_bpf_get_hist_value(int key, + long long *value_irq, + long long *value_thread, + long long *value_user, + int cpus); +int timerlat_bpf_get_summary_value(enum summary_field key, + long long *value_irq, + long long *value_thread, + long long *value_user, + int cpus); +static inline int have_libbpf_support(void) { return 1; } +#else +static inline int timerlat_bpf_init(struct timerlat_params *params) +{ + return -1; +} +static inline int timerlat_bpf_attach(void) { return -1; } +static inline void timerlat_bpf_detach(void) { }; +static inline void timerlat_bpf_destroy(void) { }; +static inline int timerlat_bpf_wait(int timeout) { return -1; } +static inline int timerlat_bpf_get_hist_value(int key, + long long *value_irq, + long long *value_thread, + long long *value_user, + int cpus) +{ + return -1; +} +static inline int timerlat_bpf_get_summary_value(enum summary_field key, + long long *value_irq, + long long *value_thread, + long long *value_user, + int cpus) +{ + return -1; +} +static inline int have_libbpf_support(void) { return 0; } +#endif /* HAVE_BPF_SKEL */ +#endif /* __bpf__ */ -- cgit v1.2.3 From fd7925cbb729c7744ffbac95928be4598354c540 Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 18 Feb 2025 15:58:56 +0100 Subject: rtla/timerlat_hist: Use BPF to collect samples Collect samples using BPF program instead of pulling them from tracefs. If the osnoise:timerlat_sample tracepoint is unavailable or the BPF program fails to load for whatever reason, rtla falls back to the old implementation. The collection of samples using the BPF program is fully self-contained and requires no activity of the userspace part of rtla during the measurement. Thus, instead of waking up every second to collect samples, rtla simply sleeps until woken up by a signal or threshold overflow. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Cc: Clark Williams Link: https://lore.kernel.org/20250218145859.27762-6-tglozar@redhat.com Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/tracing/rtla/src/timerlat_hist.c | 167 ++++++++++++++++++++++++++++----- 1 file changed, 146 insertions(+), 21 deletions(-) (limited to 'tools') diff --git a/tools/tracing/rtla/src/timerlat_hist.c b/tools/tracing/rtla/src/timerlat_hist.c index fe683a20b0e2..202d99a598ba 100644 --- a/tools/tracing/rtla/src/timerlat_hist.c +++ b/tools/tracing/rtla/src/timerlat_hist.c @@ -17,6 +17,7 @@ #include "timerlat.h" #include "timerlat_aa.h" #include "timerlat_u.h" +#include "timerlat_bpf.h" struct timerlat_hist_cpu { int *irq; @@ -193,6 +194,89 @@ timerlat_hist_handler(struct trace_seq *s, struct tep_record *record, return 0; } +/* + * timerlat_hist_bpf_pull_data - copy data from BPF maps into userspace + */ +static int timerlat_hist_bpf_pull_data(struct osnoise_tool *tool) +{ + struct timerlat_hist_data *data = tool->data; + int i, j, err; + long long value_irq[data->nr_cpus], + value_thread[data->nr_cpus], + value_user[data->nr_cpus]; + + /* Pull histogram */ + for (i = 0; i < data->entries; i++) { + err = timerlat_bpf_get_hist_value(i, value_irq, value_thread, + value_user, data->nr_cpus); + if (err) + return err; + for (j = 0; j < data->nr_cpus; j++) { + data->hist[j].irq[i] = value_irq[j]; + data->hist[j].thread[i] = value_thread[j]; + data->hist[j].user[i] = value_user[j]; + } + } + + /* Pull summary */ + err = timerlat_bpf_get_summary_value(SUMMARY_COUNT, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->hist[i].irq_count = value_irq[i]; + data->hist[i].thread_count = value_thread[i]; + data->hist[i].user_count = value_user[i]; + } + + err = timerlat_bpf_get_summary_value(SUMMARY_MIN, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->hist[i].min_irq = value_irq[i]; + data->hist[i].min_thread = value_thread[i]; + data->hist[i].min_user = value_user[i]; + } + + err = timerlat_bpf_get_summary_value(SUMMARY_MAX, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->hist[i].max_irq = value_irq[i]; + data->hist[i].max_thread = value_thread[i]; + data->hist[i].max_user = value_user[i]; + } + + err = timerlat_bpf_get_summary_value(SUMMARY_SUM, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->hist[i].sum_irq = value_irq[i]; + data->hist[i].sum_thread = value_thread[i]; + data->hist[i].sum_user = value_user[i]; + } + + err = timerlat_bpf_get_summary_value(SUMMARY_OVERFLOW, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->hist[i].irq[data->entries] = value_irq[i]; + data->hist[i].thread[data->entries] = value_thread[i]; + data->hist[i].user[data->entries] = value_user[i]; + } + + return 0; +} + /* * timerlat_hist_header - print the header of the tracer to the output */ @@ -1152,6 +1236,7 @@ int timerlat_hist_main(int argc, char *argv[]) pthread_t timerlat_u; int retval; int nr_cpus, i; + bool no_bpf = false; params = timerlat_hist_parse_args(argc, argv); if (!params) @@ -1177,6 +1262,24 @@ int timerlat_hist_main(int argc, char *argv[]) */ hist_inst = trace; + if (getenv("RTLA_NO_BPF") && strncmp(getenv("RTLA_NO_BPF"), "1", 2) == 0) { + debug_msg("RTLA_NO_BPF set, disabling BPF\n"); + no_bpf = true; + } + + if (!no_bpf && !tep_find_event_by_name(trace->tep, "osnoise", "timerlat_sample")) { + debug_msg("osnoise:timerlat_sample missing, disabling BPF\n"); + no_bpf = true; + } + + if (!no_bpf) { + retval = timerlat_bpf_init(params); + if (retval) { + debug_msg("Could not enable BPF\n"); + no_bpf = true; + } + } + retval = enable_timerlat(trace); if (retval) { err_msg("Failed to enable timerlat tracer\n"); @@ -1304,35 +1407,55 @@ int timerlat_hist_main(int argc, char *argv[]) trace_instance_start(&record->trace); if (!params->no_aa) trace_instance_start(&aa->trace); - trace_instance_start(trace); + if (no_bpf) { + trace_instance_start(trace); + } else { + retval = timerlat_bpf_attach(); + if (retval) { + err_msg("Error attaching BPF program\n"); + goto out_hist; + } + } tool->start_time = time(NULL); timerlat_hist_set_signals(params); - while (!stop_tracing) { - sleep(params->sleep_time); - - retval = tracefs_iterate_raw_events(trace->tep, - trace->inst, - NULL, - 0, - collect_registered_events, - trace); - if (retval < 0) { - err_msg("Error iterating on events\n"); - goto out_hist; - } - - if (osnoise_trace_is_off(tool, record)) - break; + if (no_bpf) { + while (!stop_tracing) { + sleep(params->sleep_time); + + retval = tracefs_iterate_raw_events(trace->tep, + trace->inst, + NULL, + 0, + collect_registered_events, + trace); + if (retval < 0) { + err_msg("Error iterating on events\n"); + goto out_hist; + } - /* is there still any user-threads ? */ - if (params->user_workload) { - if (params_u.stopped_running) { - debug_msg("timerlat user-space threads stopped!\n"); + if (osnoise_trace_is_off(tool, record)) break; + + /* is there still any user-threads ? */ + if (params->user_workload) { + if (params_u.stopped_running) { + debug_msg("timerlat user-space threads stopped!\n"); + break; + } } } + } else + timerlat_bpf_wait(-1); + + if (!no_bpf) { + timerlat_bpf_detach(); + retval = timerlat_hist_bpf_pull_data(tool); + if (retval) { + err_msg("Error pulling BPF data\n"); + goto out_hist; + } } if (params->user_workload && !params_u.stopped_running) { @@ -1376,6 +1499,8 @@ out_free: osnoise_destroy_tool(tool); free(params); free_cpu_idle_disable_states(); + if (!no_bpf) + timerlat_bpf_destroy(); out_exit: exit(return_value); } -- cgit v1.2.3 From 18923806b1291102cad3a6b713006c7e7f563534 Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 18 Feb 2025 15:58:57 +0100 Subject: rtla/timerlat_top: Move divisor to update Unlike timerlat-hist, timerlat-top applies the output divisor used to set ns/us mode when printing results instead of applying it when collecting the samples. Move the application of the divisor from timerlat_top_print into timerlat_top_update to make it consistent with timerlat-hist. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Cc: Clark Williams Link: https://lore.kernel.org/20250218145859.27762-7-tglozar@redhat.com Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/tracing/rtla/src/timerlat_top.c | 54 ++++++++++++++++------------------- 1 file changed, 25 insertions(+), 29 deletions(-) (limited to 'tools') diff --git a/tools/tracing/rtla/src/timerlat_top.c b/tools/tracing/rtla/src/timerlat_top.c index 97eead91decc..a3d9e8f67a4f 100644 --- a/tools/tracing/rtla/src/timerlat_top.c +++ b/tools/tracing/rtla/src/timerlat_top.c @@ -127,9 +127,13 @@ timerlat_top_update(struct osnoise_tool *tool, int cpu, unsigned long long thread, unsigned long long latency) { + struct timerlat_params *params = tool->params; struct timerlat_top_data *data = tool->data; struct timerlat_top_cpu *cpu_data = &data->cpu_data[cpu]; + if (params->output_divisor) + latency = latency / params->output_divisor; + if (!thread) { cpu_data->irq_count++; cpu_data->cur_irq = latency; @@ -231,12 +235,8 @@ static void timerlat_top_print(struct osnoise_tool *top, int cpu) struct timerlat_params *params = top->params; struct timerlat_top_data *data = top->data; struct timerlat_top_cpu *cpu_data = &data->cpu_data[cpu]; - int divisor = params->output_divisor; struct trace_seq *s = top->trace.seq; - if (divisor == 0) - return; - /* * Skip if no data is available: is this cpu offline? */ @@ -251,20 +251,20 @@ static void timerlat_top_print(struct osnoise_tool *top, int cpu) if (!cpu_data->irq_count) { trace_seq_printf(s, "%s %s %s %s |", no_value, no_value, no_value, no_value); } else { - trace_seq_printf(s, "%9llu ", cpu_data->cur_irq / params->output_divisor); - trace_seq_printf(s, "%9llu ", cpu_data->min_irq / params->output_divisor); - trace_seq_printf(s, "%9llu ", (cpu_data->sum_irq / cpu_data->irq_count) / divisor); - trace_seq_printf(s, "%9llu |", cpu_data->max_irq / divisor); + trace_seq_printf(s, "%9llu ", cpu_data->cur_irq); + trace_seq_printf(s, "%9llu ", cpu_data->min_irq); + trace_seq_printf(s, "%9llu ", cpu_data->sum_irq / cpu_data->irq_count); + trace_seq_printf(s, "%9llu |", cpu_data->max_irq); } if (!cpu_data->thread_count) { trace_seq_printf(s, "%s %s %s %s", no_value, no_value, no_value, no_value); } else { - trace_seq_printf(s, "%9llu ", cpu_data->cur_thread / divisor); - trace_seq_printf(s, "%9llu ", cpu_data->min_thread / divisor); + trace_seq_printf(s, "%9llu ", cpu_data->cur_thread); + trace_seq_printf(s, "%9llu ", cpu_data->min_thread); trace_seq_printf(s, "%9llu ", - (cpu_data->sum_thread / cpu_data->thread_count) / divisor); - trace_seq_printf(s, "%9llu", cpu_data->max_thread / divisor); + cpu_data->sum_thread / cpu_data->thread_count); + trace_seq_printf(s, "%9llu", cpu_data->max_thread); } if (!params->user_top) { @@ -277,11 +277,11 @@ static void timerlat_top_print(struct osnoise_tool *top, int cpu) if (!cpu_data->user_count) { trace_seq_printf(s, "%s %s %s %s\n", no_value, no_value, no_value, no_value); } else { - trace_seq_printf(s, "%9llu ", cpu_data->cur_user / divisor); - trace_seq_printf(s, "%9llu ", cpu_data->min_user / divisor); + trace_seq_printf(s, "%9llu ", cpu_data->cur_user); + trace_seq_printf(s, "%9llu ", cpu_data->min_user); trace_seq_printf(s, "%9llu ", - (cpu_data->sum_user / cpu_data->user_count) / divisor); - trace_seq_printf(s, "%9llu\n", cpu_data->max_user / divisor); + cpu_data->sum_user / cpu_data->user_count); + trace_seq_printf(s, "%9llu\n", cpu_data->max_user); } } @@ -294,13 +294,9 @@ timerlat_top_print_sum(struct osnoise_tool *top, struct timerlat_top_cpu *summar const char *split = "----------------------------------------"; struct timerlat_params *params = top->params; unsigned long long count = summary->irq_count; - int divisor = params->output_divisor; struct trace_seq *s = top->trace.seq; int e = 0; - if (divisor == 0) - return; - /* * Skip if no data is available: is this cpu offline? */ @@ -323,19 +319,19 @@ timerlat_top_print_sum(struct osnoise_tool *top, struct timerlat_top_cpu *summar trace_seq_printf(s, " %s %s %s |", no_value, no_value, no_value); } else { trace_seq_printf(s, " "); - trace_seq_printf(s, "%9llu ", summary->min_irq / params->output_divisor); - trace_seq_printf(s, "%9llu ", (summary->sum_irq / summary->irq_count) / divisor); - trace_seq_printf(s, "%9llu |", summary->max_irq / divisor); + trace_seq_printf(s, "%9llu ", summary->min_irq); + trace_seq_printf(s, "%9llu ", summary->sum_irq / summary->irq_count); + trace_seq_printf(s, "%9llu |", summary->max_irq); } if (!summary->thread_count) { trace_seq_printf(s, "%s %s %s %s", no_value, no_value, no_value, no_value); } else { trace_seq_printf(s, " "); - trace_seq_printf(s, "%9llu ", summary->min_thread / divisor); + trace_seq_printf(s, "%9llu ", summary->min_thread); trace_seq_printf(s, "%9llu ", - (summary->sum_thread / summary->thread_count) / divisor); - trace_seq_printf(s, "%9llu", summary->max_thread / divisor); + summary->sum_thread / summary->thread_count); + trace_seq_printf(s, "%9llu", summary->max_thread); } if (!params->user_top) { @@ -349,10 +345,10 @@ timerlat_top_print_sum(struct osnoise_tool *top, struct timerlat_top_cpu *summar trace_seq_printf(s, " %s %s %s |", no_value, no_value, no_value); } else { trace_seq_printf(s, " "); - trace_seq_printf(s, "%9llu ", summary->min_user / divisor); + trace_seq_printf(s, "%9llu ", summary->min_user); trace_seq_printf(s, "%9llu ", - (summary->sum_user / summary->user_count) / divisor); - trace_seq_printf(s, "%9llu\n", summary->max_user / divisor); + summary->sum_user / summary->user_count); + trace_seq_printf(s, "%9llu\n", summary->max_user); } } -- cgit v1.2.3 From 9a82a3fd9ef2afaafbd6f8059609cdead0f97f15 Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 18 Feb 2025 15:58:58 +0100 Subject: rtla/timerlat_top: Use BPF to collect samples Collect samples using BPF program instead of pulling them from tracefs. If the osnoise:timerlat_sample tracepoint is unavailable or the BPF program fails to load for whatever reason, rtla falls back to the old implementation. The collection of samples using the BPF program is fully self-contained and requires no activity of the userspace part of rtla during the measurement. Thus, rtla only pulls the summary from the BPF map and displays it every second, improving the performance. In --aa-only mode, the BPF program does not collect any data and only signalizes the end of tracing to userspace. An optimization that re-used the main trace instance for auto-analysis in aa-only mode was dropped, as rtla no longer turns tracing on in the main trace instance, making it useless for auto-analysis. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Cc: Clark Williams Link: https://lore.kernel.org/20250218145859.27762-8-tglozar@redhat.com Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/tracing/rtla/src/timerlat_top.c | 255 ++++++++++++++++++++++++++++------ 1 file changed, 215 insertions(+), 40 deletions(-) (limited to 'tools') diff --git a/tools/tracing/rtla/src/timerlat_top.c b/tools/tracing/rtla/src/timerlat_top.c index a3d9e8f67a4f..1b6455b9e093 100644 --- a/tools/tracing/rtla/src/timerlat_top.c +++ b/tools/tracing/rtla/src/timerlat_top.c @@ -18,6 +18,7 @@ #include "timerlat.h" #include "timerlat_aa.h" #include "timerlat_u.h" +#include "timerlat_bpf.h" struct timerlat_top_cpu { unsigned long long irq_count; @@ -181,6 +182,76 @@ timerlat_top_handler(struct trace_seq *s, struct tep_record *record, return 0; } +/* + * timerlat_top_bpf_pull_data - copy data from BPF maps into userspace + */ +static int timerlat_top_bpf_pull_data(struct osnoise_tool *tool) +{ + struct timerlat_top_data *data = tool->data; + int i, err; + long long value_irq[data->nr_cpus], + value_thread[data->nr_cpus], + value_user[data->nr_cpus]; + + /* Pull summary */ + err = timerlat_bpf_get_summary_value(SUMMARY_CURRENT, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->cpu_data[i].cur_irq = value_irq[i]; + data->cpu_data[i].cur_thread = value_thread[i]; + data->cpu_data[i].cur_user = value_user[i]; + } + + err = timerlat_bpf_get_summary_value(SUMMARY_COUNT, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->cpu_data[i].irq_count = value_irq[i]; + data->cpu_data[i].thread_count = value_thread[i]; + data->cpu_data[i].user_count = value_user[i]; + } + + err = timerlat_bpf_get_summary_value(SUMMARY_MIN, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->cpu_data[i].min_irq = value_irq[i]; + data->cpu_data[i].min_thread = value_thread[i]; + data->cpu_data[i].min_user = value_user[i]; + } + + err = timerlat_bpf_get_summary_value(SUMMARY_MAX, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->cpu_data[i].max_irq = value_irq[i]; + data->cpu_data[i].max_thread = value_thread[i]; + data->cpu_data[i].max_user = value_user[i]; + } + + err = timerlat_bpf_get_summary_value(SUMMARY_SUM, + value_irq, value_thread, value_user, + data->nr_cpus); + if (err) + return err; + for (i = 0; i < data->nr_cpus; i++) { + data->cpu_data[i].sum_irq = value_irq[i]; + data->cpu_data[i].sum_thread = value_thread[i]; + data->cpu_data[i].sum_user = value_user[i]; + } + + return 0; +} + /* * timerlat_top_header - print the header of the tool output */ @@ -894,6 +965,111 @@ timerlat_top_set_signals(struct timerlat_params *params) } } +/* + * timerlat_top_main_loop - main loop to process events + */ +static int +timerlat_top_main_loop(struct osnoise_tool *top, + struct osnoise_tool *record, + struct timerlat_params *params, + struct timerlat_u_params *params_u) +{ + struct trace_instance *trace = &top->trace; + int retval; + + while (!stop_tracing) { + sleep(params->sleep_time); + + if (params->aa_only && !osnoise_trace_is_off(top, record)) + continue; + + retval = tracefs_iterate_raw_events(trace->tep, + trace->inst, + NULL, + 0, + collect_registered_events, + trace); + if (retval < 0) { + err_msg("Error iterating on events\n"); + return retval; + } + + if (!params->quiet) + timerlat_print_stats(params, top); + + if (osnoise_trace_is_off(top, record)) + break; + + /* is there still any user-threads ? */ + if (params->user_workload) { + if (params_u->stopped_running) { + debug_msg("timerlat user space threads stopped!\n"); + break; + } + } + } + + return 0; +} + +/* + * timerlat_top_bpf_main_loop - main loop to process events (BPF variant) + */ +static int +timerlat_top_bpf_main_loop(struct osnoise_tool *top, + struct osnoise_tool *record, + struct timerlat_params *params, + struct timerlat_u_params *params_u) +{ + int retval, wait_retval; + + if (params->aa_only) { + /* Auto-analysis only, just wait for stop tracing */ + timerlat_bpf_wait(-1); + return 0; + } + + if (params->quiet) { + /* Quiet mode: wait for stop and then, print results */ + timerlat_bpf_wait(-1); + + retval = timerlat_top_bpf_pull_data(top); + if (retval) { + err_msg("Error pulling BPF data\n"); + return retval; + } + + return 0; + } + + /* Pull and display data in a loop */ + while (!stop_tracing) { + wait_retval = timerlat_bpf_wait(params->sleep_time); + + retval = timerlat_top_bpf_pull_data(top); + if (retval) { + err_msg("Error pulling BPF data\n"); + return retval; + } + + timerlat_print_stats(params, top); + + if (wait_retval == 1) + /* Stopping requested by tracer */ + break; + + /* is there still any user-threads ? */ + if (params->user_workload) { + if (params_u->stopped_running) { + debug_msg("timerlat user space threads stopped!\n"); + break; + } + } + } + + return 0; +} + int timerlat_top_main(int argc, char *argv[]) { struct timerlat_params *params; @@ -908,6 +1084,7 @@ int timerlat_top_main(int argc, char *argv[]) char *max_lat; int retval; int nr_cpus, i; + bool no_bpf = false; params = timerlat_top_parse_args(argc, argv); if (!params) @@ -933,6 +1110,23 @@ int timerlat_top_main(int argc, char *argv[]) */ top_inst = trace; + if (getenv("RTLA_NO_BPF") && strncmp(getenv("RTLA_NO_BPF"), "1", 2) == 0) { + debug_msg("RTLA_NO_BPF set, disabling BPF\n"); + no_bpf = true; + } + + if (!no_bpf && !tep_find_event_by_name(trace->tep, "osnoise", "timerlat_sample")) { + debug_msg("osnoise:timerlat_sample missing, disabling BPF\n"); + no_bpf = true; + } + + if (!no_bpf) { + retval = timerlat_bpf_init(params); + if (retval) { + debug_msg("Could not enable BPF\n"); + no_bpf = true; + } + } retval = enable_timerlat(trace); if (retval) { @@ -1007,15 +1201,9 @@ int timerlat_top_main(int argc, char *argv[]) } if (!params->no_aa) { - if (params->aa_only) { - /* as top is not used for display, use it for aa */ - aa = top; - } else { - /* otherwise, a new instance is needed */ - aa = osnoise_init_tool("timerlat_aa"); - if (!aa) - goto out_top; - } + aa = osnoise_init_tool("timerlat_aa"); + if (!aa) + goto out_top; retval = timerlat_aa_init(aa, params->dump_tasks); if (retval) { @@ -1066,44 +1254,31 @@ int timerlat_top_main(int argc, char *argv[]) */ if (params->trace_output) trace_instance_start(&record->trace); - if (!params->no_aa && aa != top) + if (!params->no_aa) trace_instance_start(&aa->trace); - trace_instance_start(trace); + if (no_bpf) { + trace_instance_start(trace); + } else { + retval = timerlat_bpf_attach(); + if (retval) { + err_msg("Error attaching BPF program\n"); + goto out_top; + } + } top->start_time = time(NULL); timerlat_top_set_signals(params); - while (!stop_tracing) { - sleep(params->sleep_time); + if (no_bpf) + retval = timerlat_top_main_loop(top, record, params, ¶ms_u); + else + retval = timerlat_top_bpf_main_loop(top, record, params, ¶ms_u); - if (params->aa_only && !osnoise_trace_is_off(top, record)) - continue; + if (retval) + goto out_top; - retval = tracefs_iterate_raw_events(trace->tep, - trace->inst, - NULL, - 0, - collect_registered_events, - trace); - if (retval < 0) { - err_msg("Error iterating on events\n"); - goto out_top; - } - - if (!params->quiet) - timerlat_print_stats(params, top); - - if (osnoise_trace_is_off(top, record)) - break; - - /* is there still any user-threads ? */ - if (params->user_workload) { - if (params_u.stopped_running) { - debug_msg("timerlat user space threads stopped!\n"); - break; - } - } - } + if (!no_bpf) + timerlat_bpf_detach(); if (params->user_workload && !params_u.stopped_running) { params_u.should_run = 0; -- cgit v1.2.3 From 005682b403c5ecf7e1eb965e8dc9d6058061ee30 Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 18 Feb 2025 15:58:59 +0100 Subject: rtla/timerlat: Test BPF mode Using the RTLA_NO_BPF environmental variable, execute rtla-timerlat tests both with and without BPF support to cover both paths. If rtla is built without BPF or the osnoise:timerlat_sample trace event is not available, test only the non-BPF path. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Cc: Clark Williams Link: https://lore.kernel.org/20250218145859.27762-9-tglozar@redhat.com Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/tracing/rtla/tests/timerlat.t | 14 ++++++++++++++ 1 file changed, 14 insertions(+) (limited to 'tools') diff --git a/tools/tracing/rtla/tests/timerlat.t b/tools/tracing/rtla/tests/timerlat.t index e86f40e5749e..e939ff71d6be 100644 --- a/tools/tracing/rtla/tests/timerlat.t +++ b/tools/tracing/rtla/tests/timerlat.t @@ -4,7 +4,20 @@ source tests/engine.sh test_begin set_timeout 2m +timerlat_sample_event='/sys/kernel/tracing/events/osnoise/timerlat_sample' +if ldd $RTLA | grep libbpf >/dev/null && [ -d "$timerlat_sample_event" ] +then + # rtla build with BPF and system supports BPF mode + no_bpf_options='0 1' +else + no_bpf_options='1' +fi + +# Do every test with and without BPF +for option in $no_bpf_options +do +export RTLA_NO_BPF=$option check "verify help page" \ "timerlat --help" check "verify -s/--stack" \ @@ -23,5 +36,6 @@ check "verify -c/--cpus" \ "timerlat hist -c 0 -d 30s" check "hist test in nanoseconds" \ "timerlat hist -i 2 -c 0 -n -d 30s" +done test_end -- cgit v1.2.3 From e82c78afa3d48f6512570e7d39258cd68e7bae0a Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Tue, 4 Mar 2025 15:22:28 +0100 Subject: tools/rv: Keep user LDFLAGS in build rv, unlike rtla and perf, drops LDFLAGS supplied by the user and honors only EXTRA_LDFLAGS. This is inconsistent with both perf and rtla and can lead to all kinds of unexpected behavior. For example, on Fedora and RHEL, it causes rv to be build without PIE, unlike the aforementioned perf and rtla: $ file /usr/bin/{rv,rtla,perf} /usr/bin/rv: ELF 64-bit LSB executable, ... /usr/bin/rtla: ELF 64-bit LSB pie executable, ... /usr/bin/perf: ELF 64-bit LSB pie executable, ... Keep both LDFLAGS and EXTRA_LDFLAGS for the build. Cc: John Kacur Cc: Luis Goncalves Cc: Gabriele Monaco Link: https://lore.kernel.org/20250304142228.767658-1-tglozar@redhat.com Fixes: 012e4e77df73 ("tools/verification: Use tools/build makefiles on rv") Signed-off-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/verification/rv/Makefile.rv | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'tools') diff --git a/tools/verification/rv/Makefile.rv b/tools/verification/rv/Makefile.rv index 161baa29eb86..2497fb96c83d 100644 --- a/tools/verification/rv/Makefile.rv +++ b/tools/verification/rv/Makefile.rv @@ -27,7 +27,7 @@ endif INCLUDE := -Iinclude/ CFLAGS := -g -DVERSION=\"$(VERSION)\" $(FOPTS) $(WOPTS) $(EXTRA_CFLAGS) $(INCLUDE) -LDFLAGS := -ggdb $(EXTRA_LDFLAGS) +LDFLAGS := -ggdb $(LDFLAGS) $(EXTRA_LDFLAGS) INSTALL := install MKDIR := mkdir -- cgit v1.2.3 From dc4d4e7c72d17e6f6bdd9ce0d7a72a085828987f Mon Sep 17 00:00:00 2001 From: Costa Shulyupin Date: Wed, 19 Feb 2025 13:51:10 +0200 Subject: rtla: Refactor save_trace_to_file The functions osnoise_hist_main(), osnoise_top_main(), timerlat_hist_main(), and timerlat_top_main() are lengthy and contain duplicated code. Refactor by consolidating the duplicate lines into the save_trace_to_file() function. Cc: Daniel Bristot de Oliveira Cc: John Kacur Cc: "Luis Claudio R. Goncalves" Cc: Eder Zulian Cc: Dan Carpenter Cc: Gabriele Monaco Link: https://lore.kernel.org/20250219115138.406075-1-costa.shul@redhat.com Signed-off-by: Costa Shulyupin Reviewed-by: Tomas Glozar Tested-by: Tomas Glozar Signed-off-by: Steven Rostedt (Google) --- tools/tracing/rtla/src/osnoise_hist.c | 5 +---- tools/tracing/rtla/src/osnoise_top.c | 5 +---- tools/tracing/rtla/src/timerlat_hist.c | 5 +---- tools/tracing/rtla/src/timerlat_top.c | 5 +---- tools/tracing/rtla/src/trace.c | 4 ++++ 5 files changed, 8 insertions(+), 16 deletions(-) (limited to 'tools') diff --git a/tools/tracing/rtla/src/osnoise_hist.c b/tools/tracing/rtla/src/osnoise_hist.c index b4930b835b0a..7c6ef67ef3e6 100644 --- a/tools/tracing/rtla/src/osnoise_hist.c +++ b/tools/tracing/rtla/src/osnoise_hist.c @@ -983,10 +983,7 @@ int osnoise_hist_main(int argc, char *argv[]) if (osnoise_trace_is_off(tool, record)) { printf("rtla osnoise hit stop tracing\n"); - if (params->trace_output) { - printf(" Saving trace to %s\n", params->trace_output); - save_trace_to_file(record->trace.inst, params->trace_output); - } + save_trace_to_file(record->trace.inst, params->trace_output); } out_hist: diff --git a/tools/tracing/rtla/src/osnoise_top.c b/tools/tracing/rtla/src/osnoise_top.c index 4772677ac762..0eeefbbbf317 100644 --- a/tools/tracing/rtla/src/osnoise_top.c +++ b/tools/tracing/rtla/src/osnoise_top.c @@ -813,10 +813,7 @@ int osnoise_top_main(int argc, char **argv) if (osnoise_trace_is_off(tool, record)) { printf("osnoise hit stop tracing\n"); - if (params->trace_output) { - printf(" Saving trace to %s\n", params->trace_output); - save_trace_to_file(record->trace.inst, params->trace_output); - } + save_trace_to_file(record->trace.inst, params->trace_output); } out_top: diff --git a/tools/tracing/rtla/src/timerlat_hist.c b/tools/tracing/rtla/src/timerlat_hist.c index 202d99a598ba..93d0c9e45020 100644 --- a/tools/tracing/rtla/src/timerlat_hist.c +++ b/tools/tracing/rtla/src/timerlat_hist.c @@ -1473,10 +1473,7 @@ int timerlat_hist_main(int argc, char *argv[]) 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); - } + save_trace_to_file(record->trace.inst, params->trace_output); } out_hist: diff --git a/tools/tracing/rtla/src/timerlat_top.c b/tools/tracing/rtla/src/timerlat_top.c index 1b6455b9e093..3894ac37d81c 100644 --- a/tools/tracing/rtla/src/timerlat_top.c +++ b/tools/tracing/rtla/src/timerlat_top.c @@ -1295,10 +1295,7 @@ int timerlat_top_main(int argc, char *argv[]) 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); - } + save_trace_to_file(record->trace.inst, params->trace_output); } else if (params->aa_only) { /* * If the trace did not stop with --aa-only, at least print the diff --git a/tools/tracing/rtla/src/trace.c b/tools/tracing/rtla/src/trace.c index 728f5029d533..74ed2f6208ba 100644 --- a/tools/tracing/rtla/src/trace.c +++ b/tools/tracing/rtla/src/trace.c @@ -75,12 +75,16 @@ int save_trace_to_file(struct tracefs_instance *inst, const char *filename) int out_fd, in_fd; int retval = -1; + if (!filename) + return 0; + in_fd = tracefs_instance_file_open(inst, file, O_RDONLY); if (in_fd < 0) { err_msg("Failed to open trace file\n"); return -1; } + printf(" Saving trace to %s\n", filename); out_fd = creat(filename, mode); if (out_fd < 0) { err_msg("Failed to create output file %s\n", filename); -- cgit v1.2.3 From 814d051ebed40b27285ab3c5e2454bd01a0f9631 Mon Sep 17 00:00:00 2001 From: Tomas Glozar Date: Wed, 26 Mar 2025 01:40:18 +0100 Subject: tools/build: Use SYSTEM_BPFTOOL for system bpftool The feature test for system bpftool uses BPFTOOL as the variable to set its path, defaulting to just "bpftool" if not set by the user. This conflicts with selftests and a few other utilities, which expect BPFTOOL to be set to the in-tree bpftool path by default. For example, bpftool selftests fail to build: $ make -C tools/testing/selftests/bpf/ make: Entering directory '/home/tglozar/dev/linux/tools/testing/selftests/bpf' make: *** No rule to make target 'bpftool', needed by '/home/tglozar/dev/linux/tools/testing/selftests/bpf/tools/include/vmlinux.h'. Stop. make: Leaving directory '/home/tglozar/dev/linux/tools/testing/selftests/bpf' Fix the problem by renaming the variable used for system bpftool from BPFTOOL to SYSTEM_BPFTOOL, so that the new usage does not conflict with the existing one of BPFTOOL. Cc: John Kacur Cc: Luis Goncalves Link: https://lore.kernel.org/20250326004018.248357-1-tglozar@redhat.com Fixes: 8a635c3856dd ("tools/b