Skip to contentSkip to conversation

perf pmu: Event parsing and listing fixes

PATCH Applied James Clark 4 participants 6 messages Jun 26, 2024 – Jul 10, 2024 perf events

lore
Share & tools
lore
#0 lore raw

v2 cover perf pmu: Event parsing and listing fixes

The second fix is related to the discussion here [1].

The first fix is unrelated but I just noticed it while fixing the listing issue.

[1]: https://lore.kernel.org/all/[email protected]/

Changes since v1:

* Add Ian's suggested-by tag on patch 2
* Fix mistake where info.pmu_name instead of pmu->name should have been used in perf_pmu__for_each_event()

James Clark (2):
perf pmu: Restore full PMU name wildcard support perf pmu: Don't de-duplicate core PMUs

 tools/perf/tests/pmu.c | 78 ++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/pmu.c  | 29 ++++++++++++----
 2 files changed, 100 insertions(+), 7 deletions(-)
Signature · 1 line
2.34.1
#1 lore raw

PATCH v2 1/2 perf pmu: Restore full PMU name wildcard support

Commit b2b9d3a3f021 ("perf pmu: Support wildcards on pmu name in dynamic pmu events") gives the following example for wildcarding a subset of PMUs:

E.g., in a system with the following dynamic pmus:

        mypmu_0
        mypmu_1
        mypmu_2
        mypmu_4

perf stat -e mypmu_[01]/<config>/

Since commit f91fa2ae6360 ("perf pmu: Refactor perf_pmu__match()"), only "*" has been supported, removing the ability to subset PMUs, even though parse-events.l still supports ? and [] characters.

Fix it by using fnmatch() when any glob character is detected and add a test which covers that and other scenarios of perf_pmu__match_ignoring_suffix().

Fixes f91fa2ae6360 ("perf pmu: Refactor perf_pmu__match()")
---
 tools/perf/tests/pmu.c | 78 ++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/pmu.c  |  2 +-
 2 files changed, 79 insertions(+), 1 deletion(-)
tools/perf/tests/pmu.c +78 -0
index 201df4446f7c..40132655ccd1 100644
@@ -453,12 +453,90 @@ static int test__name_cmp(struct test_suite *test __maybe_unused, int subtest __
 	return TEST_OK;
 }
 
+/**
+ * Test perf_pmu__match() that's used to search for a PMU given a name passed
+ * on the command line. The name that's passed may also be a filename type glob
+ * match.
+ */
+static int test__pmu_match(struct test_suite *test __maybe_unused, int subtest __maybe_unused)
+{
+	struct perf_pmu test_pmu;
+
+	test_pmu.name = "pmuname";
+	TEST_ASSERT_EQUAL("Exact match", perf_pmu__match(&test_pmu, "pmuname"),	     true);
+	TEST_ASSERT_EQUAL("Longer token", perf_pmu__match(&test_pmu, "longertoken"), false);
+	TEST_ASSERT_EQUAL("Shorter token", perf_pmu__match(&test_pmu, "pmu"),	     false);
+
+	test_pmu.name = "pmuname_10";
+	TEST_ASSERT_EQUAL("Diff suffix_", perf_pmu__match(&test_pmu, "pmuname_2"),  false);
+	TEST_ASSERT_EQUAL("Sub suffix_",  perf_pmu__match(&test_pmu, "pmuname_1"),  true);
+	TEST_ASSERT_EQUAL("Same suffix_", perf_pmu__match(&test_pmu, "pmuname_10"), true);
+	TEST_ASSERT_EQUAL("No suffix_",   perf_pmu__match(&test_pmu, "pmuname"),    true);
+	TEST_ASSERT_EQUAL("Underscore_",  perf_pmu__match(&test_pmu, "pmuname_"),   true);
+	TEST_ASSERT_EQUAL("Substring_",   perf_pmu__match(&test_pmu, "pmuna"),      false);
+
+	test_pmu.name = "pmuname_ab23";
+	TEST_ASSERT_EQUAL("Diff suffix hex_", perf_pmu__match(&test_pmu, "pmuname_2"),    false);
+	TEST_ASSERT_EQUAL("Sub suffix hex_",  perf_pmu__match(&test_pmu, "pmuname_ab"),   true);
+	TEST_ASSERT_EQUAL("Same suffix hex_", perf_pmu__match(&test_pmu, "pmuname_ab23"), true);
+	TEST_ASSERT_EQUAL("No suffix hex_",   perf_pmu__match(&test_pmu, "pmuname"),      true);
+	TEST_ASSERT_EQUAL("Underscore hex_",  perf_pmu__match(&test_pmu, "pmuname_"),     true);
+	TEST_ASSERT_EQUAL("Substring hex_",   perf_pmu__match(&test_pmu, "pmuna"),	 false);
+
+	test_pmu.name = "pmuname10";
+	TEST_ASSERT_EQUAL("Diff suffix", perf_pmu__match(&test_pmu, "pmuname2"),  false);
+	TEST_ASSERT_EQUAL("Sub suffix",  perf_pmu__match(&test_pmu, "pmuname1"),  true);
+	TEST_ASSERT_EQUAL("Same suffix", perf_pmu__match(&test_pmu, "pmuname10"), true);
+	TEST_ASSERT_EQUAL("No suffix",   perf_pmu__match(&test_pmu, "pmuname"),   true);
+	TEST_ASSERT_EQUAL("Underscore",  perf_pmu__match(&test_pmu, "pmuname_"),  false);
+	TEST_ASSERT_EQUAL("Substring",   perf_pmu__match(&test_pmu, "pmuna"),     false);
+
+	test_pmu.name = "pmunameab23";
+	TEST_ASSERT_EQUAL("Diff suffix hex", perf_pmu__match(&test_pmu, "pmuname2"),    false);
+	TEST_ASSERT_EQUAL("Sub suffix hex",  perf_pmu__match(&test_pmu, "pmunameab"),   true);
+	TEST_ASSERT_EQUAL("Same suffix hex", perf_pmu__match(&test_pmu, "pmunameab23"), true);
+	TEST_ASSERT_EQUAL("No suffix hex",   perf_pmu__match(&test_pmu, "pmuname"),     true);
+	TEST_ASSERT_EQUAL("Underscore hex",  perf_pmu__match(&test_pmu, "pmuname_"),    false);
+	TEST_ASSERT_EQUAL("Substring hex",   perf_pmu__match(&test_pmu, "pmuna"),	false);
+
+	/*
+	 * 2 hex chars or less are not considered suffixes so it shouldn't be
+	 * possible to wildcard by skipping the suffix. Therefore there are more
+	 * false results here than above.
+	 */
+	test_pmu.name = "pmuname_a3";
+	TEST_ASSERT_EQUAL("Diff suffix 2 hex_", perf_pmu__match(&test_pmu, "pmuname_2"),  false);
+	/*
+	 * This one should be false, but because pmuname_a3 ends in 3 which is
+	 * decimal, it's not possible to determine if it's a short hex suffix or
+	 * a normal decimal suffix following text. And we want to match on any
+	 * length of decimal suffix. Run the test anyway and expect the wrong
+	 * result. And slightly fuzzy matching shouldn't do too much harm.
+	 */
+	TEST_ASSERT_EQUAL("Sub suffix 2 hex_",  perf_pmu__match(&test_pmu, "pmuname_a"),  true);
+	TEST_ASSERT_EQUAL("Same suffix 2 hex_", perf_pmu__match(&test_pmu, "pmuname_a3"), true);
+	TEST_ASSERT_EQUAL("No suffix 2 hex_",   perf_pmu__match(&test_pmu, "pmuname"),    false);
+	TEST_ASSERT_EQUAL("Underscore 2 hex_",  perf_pmu__match(&test_pmu, "pmuname_"),   false);
+	TEST_ASSERT_EQUAL("Substring 2 hex_",   perf_pmu__match(&test_pmu, "pmuna"),	  false);
+
+	test_pmu.name = "pmuname_5";
+	TEST_ASSERT_EQUAL("Glob 1", perf_pmu__match(&test_pmu, "pmu*"),		   true);
+	TEST_ASSERT_EQUAL("Glob 2", perf_pmu__match(&test_pmu, "nomatch*"),	   false);
+	TEST_ASSERT_EQUAL("Seq 1",  perf_pmu__match(&test_pmu, "pmuname_[12345]"), true);
+	TEST_ASSERT_EQUAL("Seq 2",  perf_pmu__match(&test_pmu, "pmuname_[67890]"), false);
+	TEST_ASSERT_EQUAL("? 1",    perf_pmu__match(&test_pmu, "pmuname_?"),	   true);
+	TEST_ASSERT_EQUAL("? 2",    perf_pmu__match(&test_pmu, "pmuname_1?"),	   false);
+
+	return TEST_OK;
+}
+
 static struct test_case tests__pmu[] = {
 	TEST_CASE("Parsing with PMU format directory", pmu_format),
 	TEST_CASE("Parsing with PMU event", pmu_events),
 	TEST_CASE("PMU event names", pmu_event_names),
 	TEST_CASE("PMU name combining", name_len),
 	TEST_CASE("PMU name comparison", name_cmp),
+	TEST_CASE("PMU cmdline match", pmu_match),
 	{	.name = NULL, }
 };
 
tools/perf/util/pmu.c +1 -1
index c94a91645b21..97d74fe6d816 100644
@@ -2150,7 +2150,7 @@ void perf_pmu__warn_invalid_config(struct perf_pmu *pmu, __u64 config,
 bool perf_pmu__match(const struct perf_pmu *pmu, const char *tok)
 {
 	const char *name = pmu->name;
-	bool need_fnmatch = strchr(tok, '*') != NULL;
+	bool need_fnmatch = strisglob(tok);
 
 	if (!strncmp(tok, "uncore_", 7))
 		tok += 7;
Signature · 1 line
2.34.1
#2 lore raw

PATCH v2 2/2 perf pmu: Don't de-duplicate core PMUs

In reply to James Clark

Arm PMUs have a suffix, either a single decimal (armv8_pmuv3_0) or 3 hex digits which (armv8_cortex_a53) which Perf assumes are both strippable suffixes for the purposes of deduplication. S390 "cpum_cf" is a similarly suffixed core PMU but is only two characters so is not treated as strippable because the rules are a minimum of 3 hex characters or 1 decimal character.

There are two paths involved in listing PMU events:

* HW/cache event printing assumes core PMUs don't have suffixes so doesn't try to strip.
* Sysfs PMU events share the printing function with uncore PMUs which strips.

This results in slightly inconsistent Perf list behavior if a core PMU has a suffix:

  # perf list
  ...
  armv8_pmuv3_0/branch-load-misses/
  armv8_pmuv3/l3d_cache_wb/          [Kernel PMU event]
  ...

Fix it by partially reverting back to the old list behavior where stripping was only done for uncore PMUs. For example commit 8d9f5146f5da ("perf pmus: Sort pmus by name then suffix") mentions that only PMUs starting 'uncore_' are considered to have a potential suffix. This change doesn't go back that far, but does only strip PMUs that are !is_core. This keeps the desirable behavior where the many possibly duplicated uncore PMUs aren't repeated, but it doesn't break listing for core PMUs.

Searching for a PMU continues to use the new stripped comparison functions, meaning that it's still possible to request an event by specifying the common part of a PMU name, or even open events on multiple similarly named PMUs. For example:

  # perf stat -e armv8_cortex/inst_retired/
  5777173628      armv8_cortex_a53/inst_retired/          (99.93%)
  7469626951      armv8_cortex_a57/inst_retired/          (49.88%)
Fixes 3241d46f5f54 ("perf pmus: Sort/merge/aggregate PMUs like mrvl_ddr_pmu")
---
 tools/perf/util/pmu.c | 27 +++++++++++++++++++++------
 1 file changed, 21 insertions(+), 6 deletions(-)
tools/perf/util/pmu.c +21 -6
index 97d74fe6d816..28af48ea803f 100644
@@ -847,6 +847,23 @@ __weak const struct pmu_metrics_table *pmu_metrics_table__find(void)
 	return perf_pmu__find_metrics_table(NULL);
 }
 
+/**
+ * Return the length of the PMU name not including the suffix for uncore PMUs.
+ *
+ * We want to deduplicate many similar uncore PMUs by stripping their suffixes,
+ * but there are never going to be too many core PMUs and the suffixes might be
+ * interesting. "arm_cortex_a53" vs "arm_cortex_a57" or "cpum_cf" for example.
+ *
+ * @skip_duplicate_pmus: False in verbose mode so all uncore PMUs are visible
+ */
+static size_t pmu_deduped_name_len(const struct perf_pmu *pmu, const char *name,
+				   bool skip_duplicate_pmus)
+{
+	return skip_duplicate_pmus && !pmu->is_core
+		? pmu_name_len_no_suffix(name)
+		: strlen(name);
+}
+
 /**
  * perf_pmu__match_ignoring_suffix - Does the pmu_name match tok ignoring any
  *                                   trailing suffix? The Suffix must be in form
@@ -1796,9 +1813,8 @@ static char *format_alias(char *buf, int len, const struct perf_pmu *pmu,
 			  const struct perf_pmu_alias *alias, bool skip_duplicate_pmus)
 {
 	struct parse_events_term *term;
-	size_t pmu_name_len = skip_duplicate_pmus
-		? pmu_name_len_no_suffix(pmu->name)
-		: strlen(pmu->name);
+	size_t pmu_name_len = pmu_deduped_name_len(pmu, pmu->name,
+						   skip_duplicate_pmus);
 	int used = snprintf(buf, len, "%.*s/%s", (int)pmu_name_len, pmu->name, alias->name);
 
 	list_for_each_entry(term, &alias->terms.terms, list) {
@@ -1839,9 +1855,8 @@ int perf_pmu__for_each_event(struct perf_pmu *pmu, bool skip_duplicate_pmus,
 		size_t buf_used, pmu_name_len;
 
 		info.pmu_name = event->pmu_name ?: pmu->name;
-		pmu_name_len = skip_duplicate_pmus
-			? pmu_name_len_no_suffix(info.pmu_name)
-			: strlen(info.pmu_name);
+		pmu_name_len = pmu_deduped_name_len(pmu, info.pmu_name,
+						    skip_duplicate_pmus);
 		info.alias = NULL;
 		if (event->desc) {
 			info.name = event->name;
Signature · 1 line
2.34.1
reviewer Reviewed-by on the v2 series
#3 lore raw
In reply to James Clark
Quoted 10 lines · James Clark

The second fix is related to the discussion here [1].

The first fix is unrelated but I just noticed it while fixing the listing issue.

[1]: https://lore.kernel.org/all/[email protected]/

Changes since v1:

Quoted 7 lines

* Add Ian's suggested-by tag on patch 2
* Fix mistake where info.pmu_name instead of pmu->name should have been used in perf_pmu__for_each_event()

James Clark (2):
perf pmu: Restore full PMU name wildcard support perf pmu: Don't de-duplicate core PMUs

Looking forward to an update on the bugs in:
https://lore.kernel.org/lkml/CAP-5=fWxdZ1ibv7Pi0L38+NF-vNWZCtcDS9JVf6k3z9nDsrQQw@mail.gmail.com/

Thanks,
Ian

Quoted 7 lines
 tools/perf/tests/pmu.c | 78 ++++++++++++++++++++++++++++++++++++++++++
 tools/perf/util/pmu.c  | 29 ++++++++++++----
 2 files changed, 100 insertions(+), 7 deletions(-)

--
2.34.1

reviewer
#4 lore raw
Quoted 19 lines · James Clark
Quoted 10 lines

The second fix is related to the discussion here [1].

The first fix is unrelated but I just noticed it while fixing the listing issue.

[1]: https://lore.kernel.org/all/[email protected]/

Changes since v1:

Quoted 7 lines

* Add Ian's suggested-by tag on patch 2
* Fix mistake where info.pmu_name instead of pmu->name should have been used in perf_pmu__for_each_event()

James Clark (2):
perf pmu: Restore full PMU name wildcard support perf pmu: Don't de-duplicate core PMUs

Quoted 7 lines

Looking forward to an update on the bugs in:
https://lore.kernel.org/lkml/CAP-5=fWxdZ1ibv7Pi0L38+NF-vNWZCtcDS9JVf6k3z9nDsrQQw@mail.gmail.com/

Thanks,
Ian

Hi Ian,

Thanks for the review. I'm just finishing off one Coresight thing and then I'll start the big.little default events one:
https://lore.kernel.org/lkml/[email protected]/

Thanks
James