diff mbox series

[RFC] ftrace/selftests: Add binary test to verify ring buffer timestamps

Message ID 20201201160656.7cc6a5e8@gandalf.local.home (mailing list archive)
State Accepted
Headers show
Series [RFC] ftrace/selftests: Add binary test to verify ring buffer timestamps | expand

Commit Message

Steven Rostedt Dec. 1, 2020, 9:06 p.m. UTC
From: Steven Rostedt (VMware) <rostedt@goodmis.org>

A bug was reported about the ftrace ring buffer going backwards:

Link: https://lore.kernel.org/r/20201124223917.795844-1-elavila@google.com

In debugging this code, I wrote a C program that uses libtracefs to enable
all events and function tracing (if it exits), and then read the raw ring
buffer binary data and make sure that all the events never go backwards. If
they do, then it does a dump of the ring buffer sub buffer page and shows
the layout of the events and their deltas.

This was a very useful tool, and can be used to make sure that the ring
buffer's timestamps are consistently monotonic.

Adding this to the ftrace selftests seems to be a way that this can be
tested more often. But this would introduce the first binary code to the
ftracetests.

To make sure that the tests still work on embedded devices (where a
compiler may not even exist), and also since this binary incorporates the
yet-to-be-released libtracefs library, if the make fails, the test exits
with UNTESTED. The UNTESTED is documented as being a place holder which
this would be if the make does not work.

Thoughts?

Soon-to-be-signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
---

Comments

Steven Rostedt Dec. 8, 2020, 2:55 a.m. UTC | #1
Masami?

-- Steve


On Tue, 1 Dec 2020 16:06:56 -0500
Steven Rostedt <rostedt@goodmis.org> wrote:

> From: Steven Rostedt (VMware) <rostedt@goodmis.org>
> 
> A bug was reported about the ftrace ring buffer going backwards:
> 
> Link: https://lore.kernel.org/r/20201124223917.795844-1-elavila@google.com
> 
> In debugging this code, I wrote a C program that uses libtracefs to enable
> all events and function tracing (if it exits), and then read the raw ring
> buffer binary data and make sure that all the events never go backwards. If
> they do, then it does a dump of the ring buffer sub buffer page and shows
> the layout of the events and their deltas.
> 
> This was a very useful tool, and can be used to make sure that the ring
> buffer's timestamps are consistently monotonic.
> 
> Adding this to the ftrace selftests seems to be a way that this can be
> tested more often. But this would introduce the first binary code to the
> ftracetests.
> 
> To make sure that the tests still work on embedded devices (where a
> compiler may not even exist), and also since this binary incorporates the
> yet-to-be-released libtracefs library, if the make fails, the test exits
> with UNTESTED. The UNTESTED is documented as being a place holder which
> this would be if the make does not work.
> 
> Thoughts?
> 
> Soon-to-be-signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
> ---
> diff --git a/tools/testing/selftests/ftrace/test-ring-buffer/Makefile b/tools/testing/selftests/ftrace/test-ring-buffer/Makefile
> new file mode 100644
> index 000000000000..86f96b6fed9f
> --- /dev/null
> +++ b/tools/testing/selftests/ftrace/test-ring-buffer/Makefile
> @@ -0,0 +1,32 @@
> +
> +LD_PATH := -L/usr/local/lib/tracefs/ -L/usr/local/lib/traceevent
> +
> +TARGETS = test-ring-buffer
> +
> +PKG_CONFIG = pkg-config
> +TRACEFS_INCLUDES = $(shell $(PKG_CONFIG) --cflags libtracefs 2>/dev/null)
> +TRACEFS_LIBS = $(shell $(PKG_CONFIG) --libs libtracefs 2>/dev/null)
> +
> +TRACEEVENT_INCLUDES = $(shell $(PKG_CONFIG) --cflags libtraceevent 2>/dev/null)
> +TRACEEVENT_LIBS = $(shell $(PKG_CONFIG) --libs libtraceevent 2>/dev/null)
> +
> +ifeq ("$(TRACEFS_INCLUDES)","")
> +$(error no libtracefs found)
> +endif
> +
> +ifeq ("$(TRACEEVENT_INCLUDES)","")
> +$(error no libtraceevent found)
> +endif
> +
> +LIBS = $(TRACEFS_LIBS) $(TRACEEVENT_LIBS) -ldl
> +
> +CFLAGS := $(TRACEFS_INCLUDES) $(TRACEEVENT_INCLUDES)
> +
> +all: $(TARGETS)
> +
> +test-ring-buffer: test-ring-buffer.c
> +	gcc -g -Wall -o $@ $(CFLAGS) $^ $(LIBS)
> +
> +clean:
> +	$(RM) $(TARGETS) *.o
> +
> diff --git a/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c b/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c
> new file mode 100644
> index 000000000000..af823e2102ff
> --- /dev/null
> +++ b/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c
> @@ -0,0 +1,399 @@
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <stdarg.h>
> +#include <string.h>
> +#include <errno.h>
> +#include <unistd.h>
> +#include <signal.h>
> +#include <stdbool.h>
> +#include <sys/types.h>
> +#include <fcntl.h>
> +
> +#include <tracefs.h>
> +#include <kbuffer.h>
> +
> +typedef long long u64;
> +
> +static char *THIS_NAME;
> +static struct tep_handle *tep;
> +
> +static void __vdie(const char *fmt, va_list ap, int err)
> +{
> +	int ret = errno;
> +
> +	if (err && errno)
> +		perror(THIS_NAME);
> +	else
> +		ret = -1;
> +
> +	fprintf(stderr, "  ");
> +	vfprintf(stderr, fmt, ap);
> +
> +	fprintf(stderr, "\n");
> +	exit(ret);
> +}
> +
> +void die(const char *fmt, ...)
> +{
> +	va_list ap;
> +
> +	va_start(ap, fmt);
> +	__vdie(fmt, ap, 0);
> +	va_end(ap);
> +}
> +
> +void pdie(const char *fmt, ...)
> +{
> +	va_list ap;
> +
> +	va_start(ap, fmt);
> +	__vdie(fmt, ap, 1);
> +	va_end(ap);
> +}
> +
> +static void usage(char **argv)
> +{
> +	char *arg = argv[0];
> +	char *p = arg+strlen(arg);
> +
> +	while (p >= arg && *p != '/')
> +		p--;
> +	p++;
> +
> +	printf("\nusage: %s timeout\n"
> +	       "\n",p);
> +	exit(-1);
> +}
> +
> +static int tfd;
> +static int result;
> +static int done = 1;
> +
> +static void stop(int sig)
> +{
> +	write(tfd, "0", 1);
> +	done = 0;
> +}
> +
> +static struct last_info {
> +	u64			timestamp;
> +	void 			*content;
> +	void 			*content_next;
> +	unsigned long		page;
> +	int			index;
> +	int			irq;
> +	int			sirq;
> +} *info;
> +
> +static int page_size;
> +
> +static void *get_meta_data(struct tep_record *record)
> +{
> +	void *p = record->data;
> +	int type_len;
> +
> +	if (record->size > 120) {
> +		int size;
> +
> +		p -= 4;
> +		size = *(unsigned int *)p;
> +		p -= 4;
> +		type_len = *(unsigned int *)p;
> +		if (size != record->size + 4)
> +			die("Expected size %d but found %d (%d)\n",
> +			    record->size + 4,
> +			    size, type_len);
> +	} else
> +		p -= 4;
> +
> +	return p;
> +}
> +
> +static void init_content(struct last_info *info, int cpu,
> +			 struct tep_record *record)
> +{
> +	int type_len;
> +	void *p;
> +
> +	p = get_meta_data(record);
> +	p -= 8;
> +
> +	type_len = *(unsigned int *)p;
> +	/* FIXME for big endian */
> +	type_len = type_len & ((1 << 5) - 1);
> +	switch (type_len) {
> +	case KBUFFER_TYPE_TIME_EXTEND:
> +	case KBUFFER_TYPE_TIME_STAMP:
> +		p -= 8;
> +		break;
> +	}
> +	p -= sizeof(long);
> +	memcpy(info[cpu].content, p, (record->data - p) + record->size);
> +	info[cpu].page = (unsigned long)p;
> +	info[cpu].content_next = record->data + record->size;
> +}
> +
> +static void update_content(struct last_info *info, int cpu,
> +			   struct tep_record *record)
> +{
> +	int size = (record->data + record->size) - info[cpu].content_next;
> +	int start = info[cpu].content_next - (void *)info[cpu].page;
> +
> +	memcpy(info[cpu].content + start, (void *)info[cpu].page + start, size);
> +	info[cpu].content_next = record->data + record->size;
> +}
> +
> +static int read_kbuf_record(struct kbuffer *kbuf, struct tep_record *record)
> +{
> +	unsigned long long ts;
> +	void *ptr;
> +
> +	ptr = kbuffer_read_event(kbuf, &ts);
> +	if (!ptr || !record)
> +		return -1;
> +
> +	memset(record, 0, sizeof(*record));
> +	record->ts = ts;
> +	record->size = kbuffer_event_size(kbuf);
> +	record->record_size = kbuffer_curr_size(kbuf);
> +	record->cpu = 0;
> +	record->data = ptr;
> +	record->ref_count = 1;
> +
> +	kbuffer_next_event(kbuf, NULL);
> +
> +	return 0;
> +}
> +
> +static unsigned get_delta(void *p)
> +{
> +	return (*(unsigned *)p & ((1<<27)-1) << 5) >> 5;
> +}
> +
> +static unsigned long long get_extend_delta(void *p)
> +{
> +	unsigned long long delta;
> +	unsigned long long extend;
> +
> +	delta = get_delta(p);
> +	p += 4;
> +	extend = *(unsigned *)p;
> +
> +	return (extend << 27) + delta;
> +}
> +
> +static void check_meta(void *p, bool show)
> +{
> +	int type_len;
> +
> +	type_len = *(unsigned *)p & ((1<<5)-1);
> +	if (show)
> +		printf(" type_len=%d\n", type_len);
> +	switch (type_len) {
> +	case KBUFFER_TYPE_TIME_EXTEND:
> +		printf("   TIME_EXTEND: delta: %lld\n", get_extend_delta(p));
> +		break;
> +	case KBUFFER_TYPE_TIME_STAMP:
> +		printf("   TIME_STAMP: stamp: %lld\n", get_extend_delta(p));
> +		break;
> +	}
> +}
> +
> +static void dump_content(struct last_info *info, int cpu,
> +			 struct tep_record *this_record)
> +{
> +	struct tep_record record;
> +	struct kbuffer *kbuf;
> +	enum kbuffer_long_size kls;
> +	int last = info[cpu].content_next - (void *)info[cpu].page;
> +	int ret;
> +	void *p;
> +
> +	switch (sizeof(long)) {
> +	case 4:
> +		kls = KBUFFER_LSIZE_4;
> +		break;
> +	default:
> +		kls = KBUFFER_LSIZE_8;
> +		break;
> +	}
> +
> +
> +	kbuf = kbuffer_alloc(kls, KBUFFER_ENDIAN_LITTLE);
> +	if (!kbuf)
> +		die("Allocating kbuffer");
> +	kbuffer_load_subbuffer(kbuf, info[cpu].content);
> +
> +	p = info[cpu].content;
> +
> +	printf(" [%lld] PAGE TIME STAMP\n", *(unsigned long long *)p);
> +
> +	p += 8 + sizeof(long);
> +	check_meta(p, true);
> +
> +	while (!(ret = read_kbuf_record(kbuf, &record))) {
> +		struct tep_event *event;
> +		int id;
> +
> +		id = tep_data_type(tep, &record);
> +		event = tep_find_event(tep, id);
> +		if (!event) {
> +			printf("Failed to find event!\n");
> +			continue;
> +		}
> +
> +		p = get_meta_data(&record);
> +
> +		printf(" [%lld] delta:%d (%d)\n", record.ts, get_delta(p), record.size);
> +
> +		if (info[cpu].content + last == record.data + record.size)
> +			break;
> +
> +		check_meta(record.data + record.size, false);
> +	}
> +	kbuffer_free(kbuf);
> +
> +	if (!this_record)
> +		return;
> +
> +	printf("\nDumping record buffer:\n");
> +
> +	init_content(info, cpu, this_record);
> +	dump_content(info, cpu, NULL);
> +}
> +
> +static int callback(struct tep_event *event,
> +		    struct tep_record *record,
> +		    int cpu, void *context)
> +{
> +	static int done;
> +	int index;
> +	int irq, sirq;
> +	int flags;
> +
> +	if (done)
> +		return 1;
> +
> +	if (!info[cpu].content) {
> +		info[cpu].content = calloc(1, page_size);
> +		if (!info[cpu].content)
> +			die("Allocating content");
> +	}
> +	if (!info[cpu].page)
> +		init_content(info, cpu, record);
> +
> +	index = (unsigned long)record->data - info[cpu].page;
> +
> +	if (index < 0) {
> +		printf("HERE!\n");
> +		info[cpu].page = (unsigned long)record->data;
> +		info[cpu].index += index;
> +		index = 0;
> +	}
> +
> +	flags = tep_data_flags(event->tep, record);
> +	irq = !!(flags & TRACE_FLAG_HARDIRQ);
> +	sirq = !!(flags & TRACE_FLAG_SOFTIRQ);
> +
> +	if (info[cpu].timestamp > record->ts) {
> +		stop(0);
> +		printf("Record on cpu %d went backwards: %lld to %lld\n"
> +		       "  last index = %d : this index = %d\n"
> +		       "  last softirq = %d : last hardirq = %d\n",
> +		       cpu, info[cpu].timestamp, record->ts,
> +		       info[cpu].index, index, info[cpu].sirq, info[cpu].irq);
> +		dump_content(info, cpu, record);
> +		done = 1;
> +		result = -1;
> +		return 1;
> +	}
> +
> +	if (info[cpu].content_next > record->data)
> +		init_content(info, cpu, record);
> +	else
> +		update_content(info, cpu, record);
> +
> +	info[cpu].timestamp = record->ts;
> +	if (info[cpu].index > index) {
> +		info[cpu].sirq = sirq;
> +		info[cpu].irq = irq;
> +	} else {
> +		info[cpu].sirq |= sirq;
> +		info[cpu].irq |= irq;
> +	}
> +	info[cpu].index = index;
> +	return 0;
> +}
> +
> +static void test_buffer(struct tep_handle *tep, int timeout)
> +{
> +	signal(SIGALRM, stop);
> +	alarm(timeout);
> +
> +	tracefs_iterate_raw_events(tep, NULL, callback, NULL);
> +}
> +
> +int main (int argc, char **argv)
> +{
> +	char *current_tracer;
> +	char *events_enable;
> +	char *tracing_on;
> +	int time;
> +	int cpus;
> +	int efd;
> +	int cfd;
> +
> +	THIS_NAME = argv[0];
> +
> +	if (argc < 2)
> +		usage(argv);
> +
> +	page_size = getpagesize();
> +
> +	time = atoi(argv[1]);
> +	if (time <= 0)
> +		usage(argv);
> +
> +	tep = tracefs_local_events(NULL);
> +	if (!tep)
> +		pdie("Error loading events");
> +
> +	events_enable = tracefs_instance_get_file(NULL, "events/enable");
> +	current_tracer = tracefs_instance_get_file(NULL, "current_tracer");
> +	tracing_on = tracefs_instance_get_file(NULL, "tracing_on");
> +
> +	if (!events_enable || !current_tracer || !tracing_on)
> +		pdie("Allocating strings");
> +
> +	cpus = sysconf(_SC_NPROCESSORS_CONF);
> +
> +	info = calloc(cpus, sizeof(*info));
> +	if (!info)
> +		pdie("Allocating per cpu info");
> +
> +	efd = open(events_enable, O_WRONLY);
> +	if (efd < 0)
> +		pdie(events_enable);
> +
> +	cfd = open(current_tracer, O_WRONLY);
> +	if (cfd < 0)
> +		pdie(current_tracer);
> +
> +	tfd = open(tracing_on, O_WRONLY);
> +	if (tfd < 0)
> +		pdie(current_tracer);
> +
> +	write(efd, "1", 1);
> +	write(cfd, "function", 8);
> +
> +	test_buffer(tep, time);
> +
> +	write(efd, "0", 1);
> +	write(cfd, "nop", 3);
> +	write(tfd, "1", 1);
> +
> +	close(efd);
> +	close(cfd);
> +	close(tfd);
> +	exit(result);
> +}
> diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc b/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc
> new file mode 100644
> index 000000000000..05f628124774
> --- /dev/null
> +++ b/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc
> @@ -0,0 +1,16 @@
> +#!/bin/sh
> +# SPDX-License-Identifier: GPL-2.0
> +# description: ftrace - test ring buffer timestamps
> +
> +cd $TOP_DIR/test-ring-buffer
> +if ! make ; then
> +	cd $TRACING_DIR
> +	exit_untested
> +fi
> +
> +./test-ring-buffer 10
> +
> +make clean
> +
> +cd $TRACING_DIR
> +
Masami Hiramatsu (Google) Dec. 8, 2020, 12:14 p.m. UTC | #2
On Tue, 1 Dec 2020 16:06:56 -0500
Steven Rostedt <rostedt@goodmis.org> wrote:

> From: Steven Rostedt (VMware) <rostedt@goodmis.org>
> 
> A bug was reported about the ftrace ring buffer going backwards:
> 
> Link: https://lore.kernel.org/r/20201124223917.795844-1-elavila@google.com
> 
> In debugging this code, I wrote a C program that uses libtracefs to enable
> all events and function tracing (if it exits), and then read the raw ring
> buffer binary data and make sure that all the events never go backwards. If
> they do, then it does a dump of the ring buffer sub buffer page and shows
> the layout of the events and their deltas.
> 
> This was a very useful tool, and can be used to make sure that the ring
> buffer's timestamps are consistently monotonic.

Yeah, this sounds good to me.

> 
> Adding this to the ftrace selftests seems to be a way that this can be
> tested more often. But this would introduce the first binary code to the
> ftracetests.

No problem. I think it is better to be placed under ftracetest/bin/ and
add a PATH before running testcases.

> To make sure that the tests still work on embedded devices (where a
> compiler may not even exist), and also since this binary incorporates the
> yet-to-be-released libtracefs library, if the make fails, the test exits
> with UNTESTED. The UNTESTED is documented as being a place holder which
> this would be if the make does not work.

Hmm, in some embedded environment, we may not have make nor gcc.
So it would be better to be build in the kselftest build process as similar
to the other kselftests.

What about putting the source code under ftracetest/src/. For the embedded
devices, it can be built with cross-tools (and make it static binary if
needed) and install it under ftracetest/bin. If we have no cross-build
tool etc. we just skip building the binary under ftracetest/bin. And if 
the testcase finds there is no binary, it just returns UNRESOLVED or UNTESTED.

(currently I returns UNRESOLVED when the test target kmodule is not found)

Thank you,

> 
> Thoughts?
> 
> Soon-to-be-signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
> ---
> diff --git a/tools/testing/selftests/ftrace/test-ring-buffer/Makefile b/tools/testing/selftests/ftrace/test-ring-buffer/Makefile
> new file mode 100644
> index 000000000000..86f96b6fed9f
> --- /dev/null
> +++ b/tools/testing/selftests/ftrace/test-ring-buffer/Makefile
> @@ -0,0 +1,32 @@
> +
> +LD_PATH := -L/usr/local/lib/tracefs/ -L/usr/local/lib/traceevent
> +
> +TARGETS = test-ring-buffer
> +
> +PKG_CONFIG = pkg-config
> +TRACEFS_INCLUDES = $(shell $(PKG_CONFIG) --cflags libtracefs 2>/dev/null)
> +TRACEFS_LIBS = $(shell $(PKG_CONFIG) --libs libtracefs 2>/dev/null)
> +
> +TRACEEVENT_INCLUDES = $(shell $(PKG_CONFIG) --cflags libtraceevent 2>/dev/null)
> +TRACEEVENT_LIBS = $(shell $(PKG_CONFIG) --libs libtraceevent 2>/dev/null)
> +
> +ifeq ("$(TRACEFS_INCLUDES)","")
> +$(error no libtracefs found)
> +endif
> +
> +ifeq ("$(TRACEEVENT_INCLUDES)","")
> +$(error no libtraceevent found)
> +endif
> +
> +LIBS = $(TRACEFS_LIBS) $(TRACEEVENT_LIBS) -ldl
> +
> +CFLAGS := $(TRACEFS_INCLUDES) $(TRACEEVENT_INCLUDES)
> +
> +all: $(TARGETS)
> +
> +test-ring-buffer: test-ring-buffer.c
> +	gcc -g -Wall -o $@ $(CFLAGS) $^ $(LIBS)
> +
> +clean:
> +	$(RM) $(TARGETS) *.o
> +
> diff --git a/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c b/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c
> new file mode 100644
> index 000000000000..af823e2102ff
> --- /dev/null
> +++ b/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c
> @@ -0,0 +1,399 @@
> +#include <stdio.h>
> +#include <stdlib.h>
> +#include <stdarg.h>
> +#include <string.h>
> +#include <errno.h>
> +#include <unistd.h>
> +#include <signal.h>
> +#include <stdbool.h>
> +#include <sys/types.h>
> +#include <fcntl.h>
> +
> +#include <tracefs.h>
> +#include <kbuffer.h>
> +
> +typedef long long u64;
> +
> +static char *THIS_NAME;
> +static struct tep_handle *tep;
> +
> +static void __vdie(const char *fmt, va_list ap, int err)
> +{
> +	int ret = errno;
> +
> +	if (err && errno)
> +		perror(THIS_NAME);
> +	else
> +		ret = -1;
> +
> +	fprintf(stderr, "  ");
> +	vfprintf(stderr, fmt, ap);
> +
> +	fprintf(stderr, "\n");
> +	exit(ret);
> +}
> +
> +void die(const char *fmt, ...)
> +{
> +	va_list ap;
> +
> +	va_start(ap, fmt);
> +	__vdie(fmt, ap, 0);
> +	va_end(ap);
> +}
> +
> +void pdie(const char *fmt, ...)
> +{
> +	va_list ap;
> +
> +	va_start(ap, fmt);
> +	__vdie(fmt, ap, 1);
> +	va_end(ap);
> +}
> +
> +static void usage(char **argv)
> +{
> +	char *arg = argv[0];
> +	char *p = arg+strlen(arg);
> +
> +	while (p >= arg && *p != '/')
> +		p--;
> +	p++;
> +
> +	printf("\nusage: %s timeout\n"
> +	       "\n",p);
> +	exit(-1);
> +}
> +
> +static int tfd;
> +static int result;
> +static int done = 1;
> +
> +static void stop(int sig)
> +{
> +	write(tfd, "0", 1);
> +	done = 0;
> +}
> +
> +static struct last_info {
> +	u64			timestamp;
> +	void 			*content;
> +	void 			*content_next;
> +	unsigned long		page;
> +	int			index;
> +	int			irq;
> +	int			sirq;
> +} *info;
> +
> +static int page_size;
> +
> +static void *get_meta_data(struct tep_record *record)
> +{
> +	void *p = record->data;
> +	int type_len;
> +
> +	if (record->size > 120) {
> +		int size;
> +
> +		p -= 4;
> +		size = *(unsigned int *)p;
> +		p -= 4;
> +		type_len = *(unsigned int *)p;
> +		if (size != record->size + 4)
> +			die("Expected size %d but found %d (%d)\n",
> +			    record->size + 4,
> +			    size, type_len);
> +	} else
> +		p -= 4;
> +
> +	return p;
> +}
> +
> +static void init_content(struct last_info *info, int cpu,
> +			 struct tep_record *record)
> +{
> +	int type_len;
> +	void *p;
> +
> +	p = get_meta_data(record);
> +	p -= 8;
> +
> +	type_len = *(unsigned int *)p;
> +	/* FIXME for big endian */
> +	type_len = type_len & ((1 << 5) - 1);
> +	switch (type_len) {
> +	case KBUFFER_TYPE_TIME_EXTEND:
> +	case KBUFFER_TYPE_TIME_STAMP:
> +		p -= 8;
> +		break;
> +	}
> +	p -= sizeof(long);
> +	memcpy(info[cpu].content, p, (record->data - p) + record->size);
> +	info[cpu].page = (unsigned long)p;
> +	info[cpu].content_next = record->data + record->size;
> +}
> +
> +static void update_content(struct last_info *info, int cpu,
> +			   struct tep_record *record)
> +{
> +	int size = (record->data + record->size) - info[cpu].content_next;
> +	int start = info[cpu].content_next - (void *)info[cpu].page;
> +
> +	memcpy(info[cpu].content + start, (void *)info[cpu].page + start, size);
> +	info[cpu].content_next = record->data + record->size;
> +}
> +
> +static int read_kbuf_record(struct kbuffer *kbuf, struct tep_record *record)
> +{
> +	unsigned long long ts;
> +	void *ptr;
> +
> +	ptr = kbuffer_read_event(kbuf, &ts);
> +	if (!ptr || !record)
> +		return -1;
> +
> +	memset(record, 0, sizeof(*record));
> +	record->ts = ts;
> +	record->size = kbuffer_event_size(kbuf);
> +	record->record_size = kbuffer_curr_size(kbuf);
> +	record->cpu = 0;
> +	record->data = ptr;
> +	record->ref_count = 1;
> +
> +	kbuffer_next_event(kbuf, NULL);
> +
> +	return 0;
> +}
> +
> +static unsigned get_delta(void *p)
> +{
> +	return (*(unsigned *)p & ((1<<27)-1) << 5) >> 5;
> +}
> +
> +static unsigned long long get_extend_delta(void *p)
> +{
> +	unsigned long long delta;
> +	unsigned long long extend;
> +
> +	delta = get_delta(p);
> +	p += 4;
> +	extend = *(unsigned *)p;
> +
> +	return (extend << 27) + delta;
> +}
> +
> +static void check_meta(void *p, bool show)
> +{
> +	int type_len;
> +
> +	type_len = *(unsigned *)p & ((1<<5)-1);
> +	if (show)
> +		printf(" type_len=%d\n", type_len);
> +	switch (type_len) {
> +	case KBUFFER_TYPE_TIME_EXTEND:
> +		printf("   TIME_EXTEND: delta: %lld\n", get_extend_delta(p));
> +		break;
> +	case KBUFFER_TYPE_TIME_STAMP:
> +		printf("   TIME_STAMP: stamp: %lld\n", get_extend_delta(p));
> +		break;
> +	}
> +}
> +
> +static void dump_content(struct last_info *info, int cpu,
> +			 struct tep_record *this_record)
> +{
> +	struct tep_record record;
> +	struct kbuffer *kbuf;
> +	enum kbuffer_long_size kls;
> +	int last = info[cpu].content_next - (void *)info[cpu].page;
> +	int ret;
> +	void *p;
> +
> +	switch (sizeof(long)) {
> +	case 4:
> +		kls = KBUFFER_LSIZE_4;
> +		break;
> +	default:
> +		kls = KBUFFER_LSIZE_8;
> +		break;
> +	}
> +
> +
> +	kbuf = kbuffer_alloc(kls, KBUFFER_ENDIAN_LITTLE);
> +	if (!kbuf)
> +		die("Allocating kbuffer");
> +	kbuffer_load_subbuffer(kbuf, info[cpu].content);
> +
> +	p = info[cpu].content;
> +
> +	printf(" [%lld] PAGE TIME STAMP\n", *(unsigned long long *)p);
> +
> +	p += 8 + sizeof(long);
> +	check_meta(p, true);
> +
> +	while (!(ret = read_kbuf_record(kbuf, &record))) {
> +		struct tep_event *event;
> +		int id;
> +
> +		id = tep_data_type(tep, &record);
> +		event = tep_find_event(tep, id);
> +		if (!event) {
> +			printf("Failed to find event!\n");
> +			continue;
> +		}
> +
> +		p = get_meta_data(&record);
> +
> +		printf(" [%lld] delta:%d (%d)\n", record.ts, get_delta(p), record.size);
> +
> +		if (info[cpu].content + last == record.data + record.size)
> +			break;
> +
> +		check_meta(record.data + record.size, false);
> +	}
> +	kbuffer_free(kbuf);
> +
> +	if (!this_record)
> +		return;
> +
> +	printf("\nDumping record buffer:\n");
> +
> +	init_content(info, cpu, this_record);
> +	dump_content(info, cpu, NULL);
> +}
> +
> +static int callback(struct tep_event *event,
> +		    struct tep_record *record,
> +		    int cpu, void *context)
> +{
> +	static int done;
> +	int index;
> +	int irq, sirq;
> +	int flags;
> +
> +	if (done)
> +		return 1;
> +
> +	if (!info[cpu].content) {
> +		info[cpu].content = calloc(1, page_size);
> +		if (!info[cpu].content)
> +			die("Allocating content");
> +	}
> +	if (!info[cpu].page)
> +		init_content(info, cpu, record);
> +
> +	index = (unsigned long)record->data - info[cpu].page;
> +
> +	if (index < 0) {
> +		printf("HERE!\n");
> +		info[cpu].page = (unsigned long)record->data;
> +		info[cpu].index += index;
> +		index = 0;
> +	}
> +
> +	flags = tep_data_flags(event->tep, record);
> +	irq = !!(flags & TRACE_FLAG_HARDIRQ);
> +	sirq = !!(flags & TRACE_FLAG_SOFTIRQ);
> +
> +	if (info[cpu].timestamp > record->ts) {
> +		stop(0);
> +		printf("Record on cpu %d went backwards: %lld to %lld\n"
> +		       "  last index = %d : this index = %d\n"
> +		       "  last softirq = %d : last hardirq = %d\n",
> +		       cpu, info[cpu].timestamp, record->ts,
> +		       info[cpu].index, index, info[cpu].sirq, info[cpu].irq);
> +		dump_content(info, cpu, record);
> +		done = 1;
> +		result = -1;
> +		return 1;
> +	}
> +
> +	if (info[cpu].content_next > record->data)
> +		init_content(info, cpu, record);
> +	else
> +		update_content(info, cpu, record);
> +
> +	info[cpu].timestamp = record->ts;
> +	if (info[cpu].index > index) {
> +		info[cpu].sirq = sirq;
> +		info[cpu].irq = irq;
> +	} else {
> +		info[cpu].sirq |= sirq;
> +		info[cpu].irq |= irq;
> +	}
> +	info[cpu].index = index;
> +	return 0;
> +}
> +
> +static void test_buffer(struct tep_handle *tep, int timeout)
> +{
> +	signal(SIGALRM, stop);
> +	alarm(timeout);
> +
> +	tracefs_iterate_raw_events(tep, NULL, callback, NULL);
> +}
> +
> +int main (int argc, char **argv)
> +{
> +	char *current_tracer;
> +	char *events_enable;
> +	char *tracing_on;
> +	int time;
> +	int cpus;
> +	int efd;
> +	int cfd;
> +
> +	THIS_NAME = argv[0];
> +
> +	if (argc < 2)
> +		usage(argv);
> +
> +	page_size = getpagesize();
> +
> +	time = atoi(argv[1]);
> +	if (time <= 0)
> +		usage(argv);
> +
> +	tep = tracefs_local_events(NULL);
> +	if (!tep)
> +		pdie("Error loading events");
> +
> +	events_enable = tracefs_instance_get_file(NULL, "events/enable");
> +	current_tracer = tracefs_instance_get_file(NULL, "current_tracer");
> +	tracing_on = tracefs_instance_get_file(NULL, "tracing_on");
> +
> +	if (!events_enable || !current_tracer || !tracing_on)
> +		pdie("Allocating strings");
> +
> +	cpus = sysconf(_SC_NPROCESSORS_CONF);
> +
> +	info = calloc(cpus, sizeof(*info));
> +	if (!info)
> +		pdie("Allocating per cpu info");
> +
> +	efd = open(events_enable, O_WRONLY);
> +	if (efd < 0)
> +		pdie(events_enable);
> +
> +	cfd = open(current_tracer, O_WRONLY);
> +	if (cfd < 0)
> +		pdie(current_tracer);
> +
> +	tfd = open(tracing_on, O_WRONLY);
> +	if (tfd < 0)
> +		pdie(current_tracer);
> +
> +	write(efd, "1", 1);
> +	write(cfd, "function", 8);
> +
> +	test_buffer(tep, time);
> +
> +	write(efd, "0", 1);
> +	write(cfd, "nop", 3);
> +	write(tfd, "1", 1);
> +
> +	close(efd);
> +	close(cfd);
> +	close(tfd);
> +	exit(result);
> +}
> diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc b/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc
> new file mode 100644
> index 000000000000..05f628124774
> --- /dev/null
> +++ b/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc
> @@ -0,0 +1,16 @@
> +#!/bin/sh
> +# SPDX-License-Identifier: GPL-2.0
> +# description: ftrace - test ring buffer timestamps
> +
> +cd $TOP_DIR/test-ring-buffer
> +if ! make ; then
> +	cd $TRACING_DIR
> +	exit_untested
> +fi
> +
> +./test-ring-buffer 10
> +
> +make clean
> +
> +cd $TRACING_DIR
> +
Steven Rostedt Dec. 8, 2020, 3 p.m. UTC | #3
On Tue, 8 Dec 2020 21:14:11 +0900
Masami Hiramatsu <mhiramat@kernel.org> wrote:

> On Tue, 1 Dec 2020 16:06:56 -0500
> Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> > From: Steven Rostedt (VMware) <rostedt@goodmis.org>
> > 
> > A bug was reported about the ftrace ring buffer going backwards:
> > 
> > Link: https://lore.kernel.org/r/20201124223917.795844-1-elavila@google.com
> > 
> > In debugging this code, I wrote a C program that uses libtracefs to enable
> > all events and function tracing (if it exits), and then read the raw ring
> > buffer binary data and make sure that all the events never go backwards. If
> > they do, then it does a dump of the ring buffer sub buffer page and shows
> > the layout of the events and their deltas.
> > 
> > This was a very useful tool, and can be used to make sure that the ring
> > buffer's timestamps are consistently monotonic.  
> 
> Yeah, this sounds good to me.
> 
> > 
> > Adding this to the ftrace selftests seems to be a way that this can be
> > tested more often. But this would introduce the first binary code to the
> > ftracetests.  
> 
> No problem. I think it is better to be placed under ftracetest/bin/ and
> add a PATH before running testcases.

OK.

> 
> > To make sure that the tests still work on embedded devices (where a
> > compiler may not even exist), and also since this binary incorporates the
> > yet-to-be-released libtracefs library, if the make fails, the test exits
> > with UNTESTED. The UNTESTED is documented as being a place holder which
> > this would be if the make does not work.  
> 
> Hmm, in some embedded environment, we may not have make nor gcc.
> So it would be better to be build in the kselftest build process as similar
> to the other kselftests.
> 
> What about putting the source code under ftracetest/src/. For the embedded
> devices, it can be built with cross-tools (and make it static binary if
> needed) and install it under ftracetest/bin. If we have no cross-build
> tool etc. we just skip building the binary under ftracetest/bin. And if 
> the testcase finds there is no binary, it just returns UNRESOLVED or UNTESTED.
> 

OK. I'll look at how to make this for both cases (embedded and not).
Because, my current case is to copy the selftests to the machine and run
them there. So my use case requires the build to happen at test time. But I
can make it where it wont build if the binary already exists.

> (currently I returns UNRESOLVED when the test target kmodule is not found)

I used UNTESTED for a couple of reasons. I figured "UNRESOLVED" was for
lack of kernel features or modules. But this is not a lack of the kernel,
but a lack of user space. If something in user space is lacking (a tool,
library, or binary), then I thought UNTESTED would be a better option. But
if you have a strong opinion on it, I'll change it to UNRESOLVED, otherwise
I'll keep UNTESTED.

Thanks!

-- Steve
Masami Hiramatsu (Google) Dec. 8, 2020, 11:39 p.m. UTC | #4
On Tue, 8 Dec 2020 10:00:24 -0500
Steven Rostedt <rostedt@goodmis.org> wrote:

> On Tue, 8 Dec 2020 21:14:11 +0900
> Masami Hiramatsu <mhiramat@kernel.org> wrote:
> 
> > On Tue, 1 Dec 2020 16:06:56 -0500
> > Steven Rostedt <rostedt@goodmis.org> wrote:
> > 
> > > From: Steven Rostedt (VMware) <rostedt@goodmis.org>
> > > 
> > > A bug was reported about the ftrace ring buffer going backwards:
> > > 
> > > Link: https://lore.kernel.org/r/20201124223917.795844-1-elavila@google.com
> > > 
> > > In debugging this code, I wrote a C program that uses libtracefs to enable
> > > all events and function tracing (if it exits), and then read the raw ring
> > > buffer binary data and make sure that all the events never go backwards. If
> > > they do, then it does a dump of the ring buffer sub buffer page and shows
> > > the layout of the events and their deltas.
> > > 
> > > This was a very useful tool, and can be used to make sure that the ring
> > > buffer's timestamps are consistently monotonic.  
> > 
> > Yeah, this sounds good to me.
> > 
> > > 
> > > Adding this to the ftrace selftests seems to be a way that this can be
> > > tested more often. But this would introduce the first binary code to the
> > > ftracetests.  
> > 
> > No problem. I think it is better to be placed under ftracetest/bin/ and
> > add a PATH before running testcases.
> 
> OK.
> 
> > 
> > > To make sure that the tests still work on embedded devices (where a
> > > compiler may not even exist), and also since this binary incorporates the
> > > yet-to-be-released libtracefs library, if the make fails, the test exits
> > > with UNTESTED. The UNTESTED is documented as being a place holder which
> > > this would be if the make does not work.  
> > 
> > Hmm, in some embedded environment, we may not have make nor gcc.
> > So it would be better to be build in the kselftest build process as similar
> > to the other kselftests.
> > 
> > What about putting the source code under ftracetest/src/. For the embedded
> > devices, it can be built with cross-tools (and make it static binary if
> > needed) and install it under ftracetest/bin. If we have no cross-build
> > tool etc. we just skip building the binary under ftracetest/bin. And if 
> > the testcase finds there is no binary, it just returns UNRESOLVED or UNTESTED.
> > 
> 
> OK. I'll look at how to make this for both cases (embedded and not).
> Because, my current case is to copy the selftests to the machine and run
> them there. So my use case requires the build to happen at test time. But I
> can make it where it wont build if the binary already exists.

For that case, what about just "make clean" before copy, then the binary
will be removed? 

> 
> > (currently I returns UNRESOLVED when the test target kmodule is not found)
> 
> I used UNTESTED for a couple of reasons. I figured "UNRESOLVED" was for
> lack of kernel features or modules. But this is not a lack of the kernel,
> but a lack of user space. If something in user space is lacking (a tool,
> library, or binary), then I thought UNTESTED would be a better option. But
> if you have a strong opinion on it, I'll change it to UNRESOLVED, otherwise
> I'll keep UNTESTED.

The idea of UNTESTED/UNRESOLVED (and UNSUPPORTED) came from POSIX standard,
it is expained in dejagnu's manual:

https://www.gnu.org/software/dejagnu/manual/A-POSIX-Conforming-Test-Framework.html

In this case, user can build the user space binary for the environment but
does't, so I think UNRESOLVED will fit to that case.

So strictly speaking, UNTESTED is just a placeholder which will be implemented
in the future. (hmm, it will be a good chance to write a document for it)

Thank you,
Steven Rostedt Dec. 9, 2020, 2:49 p.m. UTC | #5
On Wed, 9 Dec 2020 08:39:54 +0900
Masami Hiramatsu <mhiramat@kernel.org> wrote:

> > OK. I'll look at how to make this for both cases (embedded and not).
> > Because, my current case is to copy the selftests to the machine and run
> > them there. So my use case requires the build to happen at test time. But I
> > can make it where it wont build if the binary already exists.  
> 
> For that case, what about just "make clean" before copy, then the binary
> will be removed? 

I just meant that the binary build should not depend on anything outside
the directory. It should be able to be built if you just copy the
ftracetest directory and run there.

> 
> >   
> > > (currently I returns UNRESOLVED when the test target kmodule is not found)  
> > 
> > I used UNTESTED for a couple of reasons. I figured "UNRESOLVED" was for
> > lack of kernel features or modules. But this is not a lack of the kernel,
> > but a lack of user space. If something in user space is lacking (a tool,
> > library, or binary), then I thought UNTESTED would be a better option. But
> > if you have a strong opinion on it, I'll change it to UNRESOLVED, otherwise
> > I'll keep UNTESTED.  
> 
> The idea of UNTESTED/UNRESOLVED (and UNSUPPORTED) came from POSIX standard,
> it is expained in dejagnu's manual:
> 
> https://www.gnu.org/software/dejagnu/manual/A-POSIX-Conforming-Test-Framework.html
> 
> In this case, user can build the user space binary for the environment but
> does't, so I think UNRESOLVED will fit to that case.
> 
> So strictly speaking, UNTESTED is just a placeholder which will be implemented
> in the future. (hmm, it will be a good chance to write a document for it)

OK, I'll change it to UNRESOLVED. One reason I used UNTESTED, is that the
build wont happen until libtracefs is packaged and released (or you
download and build it yourself). And it's just getting ready now. So in
essence, this is currently just a "placeholder" ;-)

-- Steve
Masami Hiramatsu (Google) Dec. 10, 2020, 2:08 a.m. UTC | #6
On Wed, 9 Dec 2020 09:49:19 -0500
Steven Rostedt <rostedt@goodmis.org> wrote:

> On Wed, 9 Dec 2020 08:39:54 +0900
> Masami Hiramatsu <mhiramat@kernel.org> wrote:
> 
> > > OK. I'll look at how to make this for both cases (embedded and not).
> > > Because, my current case is to copy the selftests to the machine and run
> > > them there. So my use case requires the build to happen at test time. But I
> > > can make it where it wont build if the binary already exists.  
> > 
> > For that case, what about just "make clean" before copy, then the binary
> > will be removed? 
> 
> I just meant that the binary build should not depend on anything outside
> the directory. It should be able to be built if you just copy the
> ftracetest directory and run there.

Oh I thought you considered the case that the library version differences
between the binary and execute environment.

> > > > (currently I returns UNRESOLVED when the test target kmodule is not found)  
> > > 
> > > I used UNTESTED for a couple of reasons. I figured "UNRESOLVED" was for
> > > lack of kernel features or modules. But this is not a lack of the kernel,
> > > but a lack of user space. If something in user space is lacking (a tool,
> > > library, or binary), then I thought UNTESTED would be a better option. But
> > > if you have a strong opinion on it, I'll change it to UNRESOLVED, otherwise
> > > I'll keep UNTESTED.  
> > 
> > The idea of UNTESTED/UNRESOLVED (and UNSUPPORTED) came from POSIX standard,
> > it is expained in dejagnu's manual:
> > 
> > https://www.gnu.org/software/dejagnu/manual/A-POSIX-Conforming-Test-Framework.html
> > 
> > In this case, user can build the user space binary for the environment but
> > does't, so I think UNRESOLVED will fit to that case.
> > 
> > So strictly speaking, UNTESTED is just a placeholder which will be implemented
> > in the future. (hmm, it will be a good chance to write a document for it)
> 
> OK, I'll change it to UNRESOLVED. One reason I used UNTESTED, is that the
> build wont happen until libtracefs is packaged and released (or you
> download and build it yourself). And it's just getting ready now. So in
> essence, this is currently just a "placeholder" ;-)

Yeah, so now if it can not run, user can solve it by installing libtracefs:)

Thank you,
diff mbox series

Patch

diff --git a/tools/testing/selftests/ftrace/test-ring-buffer/Makefile b/tools/testing/selftests/ftrace/test-ring-buffer/Makefile
new file mode 100644
index 000000000000..86f96b6fed9f
--- /dev/null
+++ b/tools/testing/selftests/ftrace/test-ring-buffer/Makefile
@@ -0,0 +1,32 @@ 
+
+LD_PATH := -L/usr/local/lib/tracefs/ -L/usr/local/lib/traceevent
+
+TARGETS = test-ring-buffer
+
+PKG_CONFIG = pkg-config
+TRACEFS_INCLUDES = $(shell $(PKG_CONFIG) --cflags libtracefs 2>/dev/null)
+TRACEFS_LIBS = $(shell $(PKG_CONFIG) --libs libtracefs 2>/dev/null)
+
+TRACEEVENT_INCLUDES = $(shell $(PKG_CONFIG) --cflags libtraceevent 2>/dev/null)
+TRACEEVENT_LIBS = $(shell $(PKG_CONFIG) --libs libtraceevent 2>/dev/null)
+
+ifeq ("$(TRACEFS_INCLUDES)","")
+$(error no libtracefs found)
+endif
+
+ifeq ("$(TRACEEVENT_INCLUDES)","")
+$(error no libtraceevent found)
+endif
+
+LIBS = $(TRACEFS_LIBS) $(TRACEEVENT_LIBS) -ldl
+
+CFLAGS := $(TRACEFS_INCLUDES) $(TRACEEVENT_INCLUDES)
+
+all: $(TARGETS)
+
+test-ring-buffer: test-ring-buffer.c
+	gcc -g -Wall -o $@ $(CFLAGS) $^ $(LIBS)
+
+clean:
+	$(RM) $(TARGETS) *.o
+
diff --git a/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c b/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c
new file mode 100644
index 000000000000..af823e2102ff
--- /dev/null
+++ b/tools/testing/selftests/ftrace/test-ring-buffer/test-ring-buffer.c
@@ -0,0 +1,399 @@ 
+#include <stdio.h>
+#include <stdlib.h>
+#include <stdarg.h>
+#include <string.h>
+#include <errno.h>
+#include <unistd.h>
+#include <signal.h>
+#include <stdbool.h>
+#include <sys/types.h>
+#include <fcntl.h>
+
+#include <tracefs.h>
+#include <kbuffer.h>
+
+typedef long long u64;
+
+static char *THIS_NAME;
+static struct tep_handle *tep;
+
+static void __vdie(const char *fmt, va_list ap, int err)
+{
+	int ret = errno;
+
+	if (err && errno)
+		perror(THIS_NAME);
+	else
+		ret = -1;
+
+	fprintf(stderr, "  ");
+	vfprintf(stderr, fmt, ap);
+
+	fprintf(stderr, "\n");
+	exit(ret);
+}
+
+void die(const char *fmt, ...)
+{
+	va_list ap;
+
+	va_start(ap, fmt);
+	__vdie(fmt, ap, 0);
+	va_end(ap);
+}
+
+void pdie(const char *fmt, ...)
+{
+	va_list ap;
+
+	va_start(ap, fmt);
+	__vdie(fmt, ap, 1);
+	va_end(ap);
+}
+
+static void usage(char **argv)
+{
+	char *arg = argv[0];
+	char *p = arg+strlen(arg);
+
+	while (p >= arg && *p != '/')
+		p--;
+	p++;
+
+	printf("\nusage: %s timeout\n"
+	       "\n",p);
+	exit(-1);
+}
+
+static int tfd;
+static int result;
+static int done = 1;
+
+static void stop(int sig)
+{
+	write(tfd, "0", 1);
+	done = 0;
+}
+
+static struct last_info {
+	u64			timestamp;
+	void 			*content;
+	void 			*content_next;
+	unsigned long		page;
+	int			index;
+	int			irq;
+	int			sirq;
+} *info;
+
+static int page_size;
+
+static void *get_meta_data(struct tep_record *record)
+{
+	void *p = record->data;
+	int type_len;
+
+	if (record->size > 120) {
+		int size;
+
+		p -= 4;
+		size = *(unsigned int *)p;
+		p -= 4;
+		type_len = *(unsigned int *)p;
+		if (size != record->size + 4)
+			die("Expected size %d but found %d (%d)\n",
+			    record->size + 4,
+			    size, type_len);
+	} else
+		p -= 4;
+
+	return p;
+}
+
+static void init_content(struct last_info *info, int cpu,
+			 struct tep_record *record)
+{
+	int type_len;
+	void *p;
+
+	p = get_meta_data(record);
+	p -= 8;
+
+	type_len = *(unsigned int *)p;
+	/* FIXME for big endian */
+	type_len = type_len & ((1 << 5) - 1);
+	switch (type_len) {
+	case KBUFFER_TYPE_TIME_EXTEND:
+	case KBUFFER_TYPE_TIME_STAMP:
+		p -= 8;
+		break;
+	}
+	p -= sizeof(long);
+	memcpy(info[cpu].content, p, (record->data - p) + record->size);
+	info[cpu].page = (unsigned long)p;
+	info[cpu].content_next = record->data + record->size;
+}
+
+static void update_content(struct last_info *info, int cpu,
+			   struct tep_record *record)
+{
+	int size = (record->data + record->size) - info[cpu].content_next;
+	int start = info[cpu].content_next - (void *)info[cpu].page;
+
+	memcpy(info[cpu].content + start, (void *)info[cpu].page + start, size);
+	info[cpu].content_next = record->data + record->size;
+}
+
+static int read_kbuf_record(struct kbuffer *kbuf, struct tep_record *record)
+{
+	unsigned long long ts;
+	void *ptr;
+
+	ptr = kbuffer_read_event(kbuf, &ts);
+	if (!ptr || !record)
+		return -1;
+
+	memset(record, 0, sizeof(*record));
+	record->ts = ts;
+	record->size = kbuffer_event_size(kbuf);
+	record->record_size = kbuffer_curr_size(kbuf);
+	record->cpu = 0;
+	record->data = ptr;
+	record->ref_count = 1;
+
+	kbuffer_next_event(kbuf, NULL);
+
+	return 0;
+}
+
+static unsigned get_delta(void *p)
+{
+	return (*(unsigned *)p & ((1<<27)-1) << 5) >> 5;
+}
+
+static unsigned long long get_extend_delta(void *p)
+{
+	unsigned long long delta;
+	unsigned long long extend;
+
+	delta = get_delta(p);
+	p += 4;
+	extend = *(unsigned *)p;
+
+	return (extend << 27) + delta;
+}
+
+static void check_meta(void *p, bool show)
+{
+	int type_len;
+
+	type_len = *(unsigned *)p & ((1<<5)-1);
+	if (show)
+		printf(" type_len=%d\n", type_len);
+	switch (type_len) {
+	case KBUFFER_TYPE_TIME_EXTEND:
+		printf("   TIME_EXTEND: delta: %lld\n", get_extend_delta(p));
+		break;
+	case KBUFFER_TYPE_TIME_STAMP:
+		printf("   TIME_STAMP: stamp: %lld\n", get_extend_delta(p));
+		break;
+	}
+}
+
+static void dump_content(struct last_info *info, int cpu,
+			 struct tep_record *this_record)
+{
+	struct tep_record record;
+	struct kbuffer *kbuf;
+	enum kbuffer_long_size kls;
+	int last = info[cpu].content_next - (void *)info[cpu].page;
+	int ret;
+	void *p;
+
+	switch (sizeof(long)) {
+	case 4:
+		kls = KBUFFER_LSIZE_4;
+		break;
+	default:
+		kls = KBUFFER_LSIZE_8;
+		break;
+	}
+
+
+	kbuf = kbuffer_alloc(kls, KBUFFER_ENDIAN_LITTLE);
+	if (!kbuf)
+		die("Allocating kbuffer");
+	kbuffer_load_subbuffer(kbuf, info[cpu].content);
+
+	p = info[cpu].content;
+
+	printf(" [%lld] PAGE TIME STAMP\n", *(unsigned long long *)p);
+
+	p += 8 + sizeof(long);
+	check_meta(p, true);
+
+	while (!(ret = read_kbuf_record(kbuf, &record))) {
+		struct tep_event *event;
+		int id;
+
+		id = tep_data_type(tep, &record);
+		event = tep_find_event(tep, id);
+		if (!event) {
+			printf("Failed to find event!\n");
+			continue;
+		}
+
+		p = get_meta_data(&record);
+
+		printf(" [%lld] delta:%d (%d)\n", record.ts, get_delta(p), record.size);
+
+		if (info[cpu].content + last == record.data + record.size)
+			break;
+
+		check_meta(record.data + record.size, false);
+	}
+	kbuffer_free(kbuf);
+
+	if (!this_record)
+		return;
+
+	printf("\nDumping record buffer:\n");
+
+	init_content(info, cpu, this_record);
+	dump_content(info, cpu, NULL);
+}
+
+static int callback(struct tep_event *event,
+		    struct tep_record *record,
+		    int cpu, void *context)
+{
+	static int done;
+	int index;
+	int irq, sirq;
+	int flags;
+
+	if (done)
+		return 1;
+
+	if (!info[cpu].content) {
+		info[cpu].content = calloc(1, page_size);
+		if (!info[cpu].content)
+			die("Allocating content");
+	}
+	if (!info[cpu].page)
+		init_content(info, cpu, record);
+
+	index = (unsigned long)record->data - info[cpu].page;
+
+	if (index < 0) {
+		printf("HERE!\n");
+		info[cpu].page = (unsigned long)record->data;
+		info[cpu].index += index;
+		index = 0;
+	}
+
+	flags = tep_data_flags(event->tep, record);
+	irq = !!(flags & TRACE_FLAG_HARDIRQ);
+	sirq = !!(flags & TRACE_FLAG_SOFTIRQ);
+
+	if (info[cpu].timestamp > record->ts) {
+		stop(0);
+		printf("Record on cpu %d went backwards: %lld to %lld\n"
+		       "  last index = %d : this index = %d\n"
+		       "  last softirq = %d : last hardirq = %d\n",
+		       cpu, info[cpu].timestamp, record->ts,
+		       info[cpu].index, index, info[cpu].sirq, info[cpu].irq);
+		dump_content(info, cpu, record);
+		done = 1;
+		result = -1;
+		return 1;
+	}
+
+	if (info[cpu].content_next > record->data)
+		init_content(info, cpu, record);
+	else
+		update_content(info, cpu, record);
+
+	info[cpu].timestamp = record->ts;
+	if (info[cpu].index > index) {
+		info[cpu].sirq = sirq;
+		info[cpu].irq = irq;
+	} else {
+		info[cpu].sirq |= sirq;
+		info[cpu].irq |= irq;
+	}
+	info[cpu].index = index;
+	return 0;
+}
+
+static void test_buffer(struct tep_handle *tep, int timeout)
+{
+	signal(SIGALRM, stop);
+	alarm(timeout);
+
+	tracefs_iterate_raw_events(tep, NULL, callback, NULL);
+}
+
+int main (int argc, char **argv)
+{
+	char *current_tracer;
+	char *events_enable;
+	char *tracing_on;
+	int time;
+	int cpus;
+	int efd;
+	int cfd;
+
+	THIS_NAME = argv[0];
+
+	if (argc < 2)
+		usage(argv);
+
+	page_size = getpagesize();
+
+	time = atoi(argv[1]);
+	if (time <= 0)
+		usage(argv);
+
+	tep = tracefs_local_events(NULL);
+	if (!tep)
+		pdie("Error loading events");
+
+	events_enable = tracefs_instance_get_file(NULL, "events/enable");
+	current_tracer = tracefs_instance_get_file(NULL, "current_tracer");
+	tracing_on = tracefs_instance_get_file(NULL, "tracing_on");
+
+	if (!events_enable || !current_tracer || !tracing_on)
+		pdie("Allocating strings");
+
+	cpus = sysconf(_SC_NPROCESSORS_CONF);
+
+	info = calloc(cpus, sizeof(*info));
+	if (!info)
+		pdie("Allocating per cpu info");
+
+	efd = open(events_enable, O_WRONLY);
+	if (efd < 0)
+		pdie(events_enable);
+
+	cfd = open(current_tracer, O_WRONLY);
+	if (cfd < 0)
+		pdie(current_tracer);
+
+	tfd = open(tracing_on, O_WRONLY);
+	if (tfd < 0)
+		pdie(current_tracer);
+
+	write(efd, "1", 1);
+	write(cfd, "function", 8);
+
+	test_buffer(tep, time);
+
+	write(efd, "0", 1);
+	write(cfd, "nop", 3);
+	write(tfd, "1", 1);
+
+	close(efd);
+	close(cfd);
+	close(tfd);
+	exit(result);
+}
diff --git a/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc b/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc
new file mode 100644
index 000000000000..05f628124774
--- /dev/null
+++ b/tools/testing/selftests/ftrace/test.d/ftrace/ring-buffer.tc
@@ -0,0 +1,16 @@ 
+#!/bin/sh
+# SPDX-License-Identifier: GPL-2.0
+# description: ftrace - test ring buffer timestamps
+
+cd $TOP_DIR/test-ring-buffer
+if ! make ; then
+	cd $TRACING_DIR
+	exit_untested
+fi
+
+./test-ring-buffer 10
+
+make clean
+
+cd $TRACING_DIR
+