[PATCH 0/4] perf tests: Add test to check backward ring buffer

[PATCH 0/4] perf tests: Add test to check backward ring buffer

[Wang Nan]

This 4 patches add a new test case, test reading from backward ring
bfufer implemented by [1].

[1]: http://lkml.kernel.org/g/1459865478-53413-1-git-send-email-wangnan0@huawei.com

Wang Nan (4):
  perf tools: Enforce ring buffer reading
  perf tools: Extract __perf_evlist__mmap_read()
  perf tools: Support reading from backward ring buffer
  perf tests: Add test to check backward ring buffer

 tools/perf/tests/Build                  |   1 +
 tools/perf/tests/backward-ring-buffer.c | 151 ++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c         |   4 +
 tools/perf/tests/tests.h                |   1 +
 tools/perf/util/evlist.c                |  87 ++++++++++++++----
 tools/perf/util/evlist.h                |   4 +
 6 files changed, 233 insertions(+), 15 deletions(-)
 create mode 100644 tools/perf/tests/backward-ring-buffer.c

-- 
1.8.3.4

[PATCH 1/4] perf tools: Enforce ring buffer reading

[Wang Nan]

Don't even consider reading data after 'head' pointer.

Following commits will feed perf_evlist__mmap_read() with some 'head'
pointers not maintained by kernel. If 'head' pointer breaks an event,
we should avoid reading from the broken event. This can happen in
backward ring buffer.

For example:

                              old     head
                                |     |
                                V     V
     +---+------+----------+----+-----+--+
     |..E|D....D|C........C|B..B|A....|E.|
     +---+------+----------+----+-----+--+

'old' pointer points to the beginning of 'A' and trying read from it,
but 'A' has been overwritten. In this case, don't try to read from 'A',
simply return NULL.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
---
 tools/perf/util/evlist.c | 12 ++++++++++--
 1 file changed, 10 insertions(+), 2 deletions(-)

diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 86a0383..692fd05 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -684,6 +684,7 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
 	struct perf_mmap *md = &evlist->mmap[idx];
 	u64 head;
 	u64 old = md->prev;
+	int diff;
 	unsigned char *data = md->base + page_size;
 	union perf_event *event = NULL;
 
@@ -694,6 +695,7 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
 		return NULL;
 
 	head = perf_mmap__read_head(md);
+	diff = head - old;
 	if (evlist->overwrite) {
 		/*
 		 * If we're further behind than half the buffer, there's a chance
@@ -703,7 +705,6 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
 		 *
 		 * In either case, truncate and restart at head.
 		 */
-		int diff = head - old;
 		if (diff > md->mask / 2 || diff < 0) {
 			fprintf(stderr, "WARNING: failed to keep up with mmap data.\n");
 
@@ -711,15 +712,21 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
 			 * head points to a known good entry, start there.
 			 */
 			old = head;
+			diff = 0;
 		}
 	}
 
-	if (old != head) {
+	if (diff >= (int)sizeof(event->header)) {
 		size_t size;
 
 		event = (union perf_event *)&data[old & md->mask];
 		size = event->header.size;
 
+		if (size < sizeof(event->header) || diff < (int)size) {
+			event = NULL;
+			goto broken_event;
+		}
+
 		/*
 		 * Event straddles the mmap boundary -- header should always
 		 * be inside due to u64 alignment of output.
@@ -743,6 +750,7 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
 		old += size;
 	}
 
+broken_event:
 	md->prev = old;
 
 	return event;
-- 
1.8.3.4

[PATCH 2/4] perf tools: Extract __perf_evlist__mmap_read()

[Wang Nan]

Extract event reader to __perf_evlist__mmap_read(). Future commit will
feed it with manually computed 'head' and 'old' pointers.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
---
 tools/perf/util/evlist.c | 40 +++++++++++++++++++++++++---------------
 1 file changed, 25 insertions(+), 15 deletions(-)

diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 692fd05..119bf27 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -679,24 +679,15 @@ static struct perf_evsel *perf_evlist__event2evsel(struct perf_evlist *evlist,
 	return NULL;
 }
 
-union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
+static union perf_event *
+__perf_evlist__mmap_read(struct perf_mmap *md, bool overwrite, u64 head,
+			 u64 old, u64 *prev)
 {
-	struct perf_mmap *md = &evlist->mmap[idx];
-	u64 head;
-	u64 old = md->prev;
-	int diff;
 	unsigned char *data = md->base + page_size;
 	union perf_event *event = NULL;
+	int diff = head - old;
 
-	/*
-	 * Check if event was unmapped due to a POLLHUP/POLLERR.
-	 */
-	if (!atomic_read(&md->refcnt))
-		return NULL;
-
-	head = perf_mmap__read_head(md);
-	diff = head - old;
-	if (evlist->overwrite) {
+	if (overwrite) {
 		/*
 		 * If we're further behind than half the buffer, there's a chance
 		 * the writer will bite our tail and mess up the samples under us.
@@ -751,11 +742,30 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
 	}
 
 broken_event:
-	md->prev = old;
+	if (prev)
+		*prev = old;
 
 	return event;
 }
 
+union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
+{
+	struct perf_mmap *md = &evlist->mmap[idx];
+	u64 head;
+	u64 old = md->prev;
+
+	/*
+	 * Check if event was unmapped due to a POLLHUP/POLLERR.
+	 */
+	if (!atomic_read(&md->refcnt))
+		return NULL;
+
+	head = perf_mmap__read_head(md);
+
+	return __perf_evlist__mmap_read(md, evlist->overwrite, head,
+					old, &md->prev);
+}
+
 static bool perf_mmap__empty(struct perf_mmap *md)
 {
 	return perf_mmap__read_head(md) == md->prev && !md->auxtrace_mmap.base;
-- 
1.8.3.4

[PATCH 3/4] perf tools: Support reading from backward ring buffer

[Wang Nan]

perf_evlist__mmap_read_backward() is introduced for reading backward
ring buffer. Different from reading forward, before reading caller
needs to call perf_evlist__mmap_read_catchup() first.

Backward ring buffer should be read from 'head' pointer, not '0'.
perf_evlist__mmap_read_catchup() saves 'head' to 'md->prev', then
use it to remember the start position after each reading.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
---
 tools/perf/util/evlist.c | 45 ++++++++++++++++++++++++++++++++++++++++++---
 tools/perf/util/evlist.h |  4 ++++
 2 files changed, 46 insertions(+), 3 deletions(-)

diff --git a/tools/perf/util/evlist.c b/tools/perf/util/evlist.c
index 119bf27..94d9085 100644
--- a/tools/perf/util/evlist.c
+++ b/tools/perf/util/evlist.c
@@ -680,14 +680,14 @@ static struct perf_evsel *perf_evlist__event2evsel(struct perf_evlist *evlist,
 }
 
 static union perf_event *
-__perf_evlist__mmap_read(struct perf_mmap *md, bool overwrite, u64 head,
-			 u64 old, u64 *prev)
+__perf_evlist__mmap_read(struct perf_mmap *md, bool check_messup,
+			 u64 head, u64 old, u64 *prev)
 {
 	unsigned char *data = md->base + page_size;
 	union perf_event *event = NULL;
 	int diff = head - old;
 
-	if (overwrite) {
+	if (check_messup) {
 		/*
 		 * If we're further behind than half the buffer, there's a chance
 		 * the writer will bite our tail and mess up the samples under us.
@@ -766,6 +766,45 @@ union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx)
 					old, &md->prev);
 }
 
+union perf_event *
+perf_evlist__mmap_read_backward(struct perf_evlist *evlist, int idx)
+{
+	struct perf_mmap *md = &evlist->mmap[idx];
+	u64 head, end;
+	u64 start = md->prev;
+
+	/*
+	 * Check if event was unmapped due to a POLLHUP/POLLERR.
+	 */
+	if (!atomic_read(&md->refcnt))
+		return NULL;
+
+	/* NOTE: head is negative in this case */
+	head = perf_mmap__read_head(md);
+
+	if (!head)
+		return NULL;
+
+	end = head + md->mask + 1;
+
+	if ((end - head) > -head)
+		end = 0;
+
+	return __perf_evlist__mmap_read(md, false, end, start, &md->prev);
+}
+
+void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx)
+{
+	struct perf_mmap *md = &evlist->mmap[idx];
+	u64 head;
+
+	if (!atomic_read(&md->refcnt))
+		return;
+
+	head = perf_mmap__read_head(md);
+	md->prev = head;
+}
+
 static bool perf_mmap__empty(struct perf_mmap *md)
 {
 	return perf_mmap__read_head(md) == md->prev && !md->auxtrace_mmap.base;
diff --git a/tools/perf/util/evlist.h b/tools/perf/util/evlist.h
index a0d1522..7c10481 100644
--- a/tools/perf/util/evlist.h
+++ b/tools/perf/util/evlist.h
@@ -118,6 +118,10 @@ struct perf_sample_id *perf_evlist__id2sid(struct perf_evlist *evlist, u64 id);
 
 union perf_event *perf_evlist__mmap_read(struct perf_evlist *evlist, int idx);
 
+union perf_event *perf_evlist__mmap_read_backward(struct perf_evlist *evlist,
+						  int idx);
+void perf_evlist__mmap_read_catchup(struct perf_evlist *evlist, int idx);
+
 void perf_evlist__mmap_consume(struct perf_evlist *evlist, int idx);
 
 int perf_evlist__open(struct perf_evlist *evlist);
-- 
1.8.3.4

[PATCH 4/4] perf tests: Add test to check backward ring buffer

[Wang Nan]

This test checks reading from backward ring buffer.

Test result:

 # ~/perf test 'ring buffer'
 45: Test reading from ring buffer                            : Ok

Test case is a while loop which calls prctl(PR_SET_NAME) multiple
times. Each prctl should issue 2 events: one PERF_RECORD_SAMPLE,
one PERF_RECORD_COMM.

The first round creates a relative large ring buffer (256 pages). It
can afford all events. Read from it and check the count of each type of
events.

The second round creates a small ring buffer (1 page) and make it
overwritable. Check the correctness of the buffer.

Signed-off-by: Wang Nan <wangnan0@huawei.com>
Cc: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Zefan Li <lizefan@huawei.com>
Cc: pi3orama@163.com
---
 tools/perf/tests/Build                  |   1 +
 tools/perf/tests/backward-ring-buffer.c | 151 ++++++++++++++++++++++++++++++++
 tools/perf/tests/builtin-test.c         |   4 +
 tools/perf/tests/tests.h                |   1 +
 4 files changed, 157 insertions(+)
 create mode 100644 tools/perf/tests/backward-ring-buffer.c

diff --git a/tools/perf/tests/Build b/tools/perf/tests/Build
index 449fe97..66a2898 100644
--- a/tools/perf/tests/Build
+++ b/tools/perf/tests/Build
@@ -38,6 +38,7 @@ perf-y += cpumap.o
 perf-y += stat.o
 perf-y += event_update.o
 perf-y += event-times.o
+perf-y += backward-ring-buffer.o
 
 $(OUTPUT)tests/llvm-src-base.c: tests/bpf-script-example.c tests/Build
 	$(call rule_mkdir)
diff --git a/tools/perf/tests/backward-ring-buffer.c b/tools/perf/tests/backward-ring-buffer.c
new file mode 100644
index 0000000..d1a1f75
--- /dev/null
+++ b/tools/perf/tests/backward-ring-buffer.c
@@ -0,0 +1,151 @@
+/*
+ * Test backward bit in event attribute, read ring buffer from end to
+ * beginning
+ */
+
+#include <perf.h>
+#include <evlist.h>
+#include <sys/prctl.h>
+#include "tests.h"
+#include "debug.h"
+
+#define NR_ITERS 111
+
+static void testcase(void)
+{
+	int i;
+
+	for (i = 0; i < NR_ITERS; i++) {
+		char proc_name[10];
+
+		snprintf(proc_name, sizeof(proc_name), "p:%d\n", i);
+		prctl(PR_SET_NAME, proc_name);
+	}
+}
+
+static int count_samples(struct perf_evlist *evlist, int *sample_count,
+			 int *comm_count)
+{
+	int i;
+
+	for (i = 0; i < evlist->nr_mmaps; i++) {
+		union perf_event *event;
+
+		perf_evlist__mmap_read_catchup(evlist, i);
+		while ((event = perf_evlist__mmap_read_backward(evlist, i)) != NULL) {
+			const u32 type = event->header.type;
+
+			switch (type) {
+			case PERF_RECORD_SAMPLE:
+				(*sample_count)++;
+				break;
+			case PERF_RECORD_COMM:
+				(*comm_count)++;
+				break;
+			default:
+				pr_err("Unexpected record of type %d\n", type);
+				return TEST_FAIL;
+			}
+		}
+	}
+	return TEST_OK;
+}
+
+static int do_test(struct perf_evlist *evlist, int mmap_pages,
+		   int *sample_count, int *comm_count)
+{
+	int err;
+	char sbuf[STRERR_BUFSIZE];
+
+	err = perf_evlist__mmap(evlist, mmap_pages, true);
+	if (err < 0) {
+		pr_debug("perf_evlist__mmap: %s\n",
+			 strerror_r(errno, sbuf, sizeof(sbuf)));
+		return TEST_FAIL;
+	}
+
+	perf_evlist__enable(evlist);
+	testcase();
+	perf_evlist__disable(evlist);
+
+	err = count_samples(evlist, sample_count, comm_count);
+	perf_evlist__munmap(evlist);
+	return err;
+}
+
+
+int test__ring_buffer(int subtest __maybe_unused)
+{
+	int ret = TEST_SKIP, err, sample_count = 0, comm_count = 0;
+	char pid[16], sbuf[STRERR_BUFSIZE];
+	struct perf_evlist *evlist;
+	struct perf_evsel *evsel __maybe_unused;
+	struct parse_events_error parse_error;
+	struct record_opts opts = {
+		.target = {
+			.uid = UINT_MAX,
+			.uses_mmap = true,
+		},
+		.freq	      = 0,
+		.mmap_pages   = 256,
+		.default_interval = 1,
+	};
+
+	snprintf(pid, sizeof(pid), "%d", getpid());
+	pid[sizeof(pid) - 1] = '\0';
+	opts.target.tid = opts.target.pid = pid;
+
+	evlist = perf_evlist__new();
+	if (!evlist) {
+		pr_debug("No ehough memory to create evlist\n");
+		return TEST_FAIL;
+	}
+
+	err = perf_evlist__create_maps(evlist, &opts.target);
+	if (err < 0) {
+		pr_debug("Not enough memory to create thread/cpu maps\n");
+		goto out_delete_evlist;
+	}
+
+	bzero(&parse_error, sizeof(parse_error));
+	err = parse_events(evlist, "syscalls:sys_enter_prctl", &parse_error);
+	if (err) {
+		pr_debug("Failed to parse tracepoint event, try use root\n");
+		ret = TEST_SKIP;
+		goto out_delete_evlist;
+	}
+
+	perf_evlist__config(evlist, &opts);
+
+	/* Set backward bit, ring buffer should be writing from end */
+	evlist__for_each(evlist, evsel)
+		evsel->attr.write_backward = 1;
+
+	err = perf_evlist__open(evlist);
+	if (err < 0) {
+		pr_debug("perf_evlist__open: %s\n",
+			 strerror_r(errno, sbuf, sizeof(sbuf)));
+		goto out_delete_evlist;
+	}
+
+	ret = TEST_FAIL;
+	err = do_test(evlist, opts.mmap_pages, &sample_count,
+		      &comm_count);
+	if (err != TEST_OK)
+		goto out_delete_evlist;
+
+	if ((sample_count != NR_ITERS) || (comm_count != NR_ITERS)) {
+		pr_err("Unexpected counter: sample_count=%d, comm_count=%d\n",
+		       sample_count, comm_count);
+		goto out_delete_evlist;
+	}
+
+	err = do_test(evlist, 1, &sample_count, &comm_count);
+	if (err != TEST_OK)
+		goto out_delete_evlist;
+
+	ret = TEST_OK;
+out_delete_evlist:
+	perf_evlist__delete(evlist);
+	return ret;
+}
diff --git a/tools/perf/tests/builtin-test.c b/tools/perf/tests/builtin-test.c
index 93c4670..b025681 100644
--- a/tools/perf/tests/builtin-test.c
+++ b/tools/perf/tests/builtin-test.c
@@ -208,6 +208,10 @@ static struct test generic_tests[] = {
 		.func = test__event_times,
 	},
 	{
+		.desc = "Test reading from ring buffer",
+		.func = test__ring_buffer,
+	},
+	{
 		.func = NULL,
 	},
 };
diff --git a/tools/perf/tests/tests.h b/tools/perf/tests/tests.h
index 0fc9469..bf68bda 100644
--- a/tools/perf/tests/tests.h
+++ b/tools/perf/tests/tests.h
@@ -86,6 +86,7 @@ int test__synthesize_stat(int subtest);
 int test__synthesize_stat_round(int subtest);
 int test__event_update(int subtest);
 int test__event_times(int subtest);
+int test__ring_buffer(int subtest);
 
 #if defined(__arm__) || defined(__aarch64__)
 #ifdef HAVE_DWARF_UNWIND_SUPPORT
-- 
1.8.3.4