[PATCH v6 01/28] perf stat: Allow retry for default events

[Ian Rogers <irogers@google.com>]

Default events are marked skippable. Checking skippable first means
retrying and adding modifiers like exclude kernel isn't
performed. Push the skippable checking after fallbacks are tried and
avoid warning multiple times for the event.

Fixes: 9eac5612da1c ("perf stat: Don't skip failing group events")
Signed-off-by: Ian Rogers <irogers@google.com>
---
 tools/perf/builtin-stat.c | 38 +++++++++++++++++++++++---------------
 1 file changed, 23 insertions(+), 15 deletions(-)

diff --git a/tools/perf/builtin-stat.c b/tools/perf/builtin-stat.c
index ab567919b89a..303628189004 100644
--- a/tools/perf/builtin-stat.c
+++ b/tools/perf/builtin-stat.c
@@ -616,16 +616,7 @@ enum counter_recovery {
 static enum counter_recovery stat_handle_error(struct evsel *counter, int err)
 {
 	char msg[BUFSIZ];
-
-	if (counter->skippable) {
-		if (verbose > 0) {
-			ui__warning("skipping event %s that kernel failed to open .\n",
-				    evsel__name(counter));
-		}
-		counter->supported = false;
-		counter->errored = true;
-		return COUNTER_SKIP;
-	}
+	bool warned = false;
 
 	/*
 	 * PPC returns ENXIO for HW counters until 2.6.37
@@ -635,6 +626,7 @@ static enum counter_recovery stat_handle_error(struct evsel *counter, int err)
 		if (verbose > 0) {
 			ui__warning("%s event is not supported by the kernel.\n",
 				    evsel__name(counter));
+			warned = true;
 		}
 		counter->supported = false;
 		/*
@@ -642,13 +634,15 @@ static enum counter_recovery stat_handle_error(struct evsel *counter, int err)
 		 * cpu event had a problem and needs to be reexamined.
 		 */
 		counter->errored = true;
-	} else if (evsel__fallback(counter, &target, err, msg, sizeof(msg))) {
+		goto skip_or_fatal;
+	}
+	if (evsel__fallback(counter, &target, err, msg, sizeof(msg))) {
 		if (verbose > 0)
 			ui__warning("%s\n", msg);
 		return COUNTER_RETRY;
-	} else if (target__has_per_thread(&target) && err != EOPNOTSUPP &&
-		   evsel_list->core.threads &&
-		   evsel_list->core.threads->err_thread != -1) {
+	}
+	if (target__has_per_thread(&target) && err != EOPNOTSUPP &&
+	    evsel_list->core.threads && evsel_list->core.threads->err_thread != -1) {
 		/*
 		 * For global --per-thread case, skip current
 		 * error thread.
@@ -658,15 +652,29 @@ static enum counter_recovery stat_handle_error(struct evsel *counter, int err)
 			evsel_list->core.threads->err_thread = -1;
 			return COUNTER_RETRY;
 		}
-	} else if (err == EOPNOTSUPP) {
+		goto skip_or_fatal;
+	}
+	if (err == EOPNOTSUPP) {
 		if (verbose > 0) {
 			ui__warning("%s event is not supported by the kernel.\n",
 				    evsel__name(counter));
+			warned = true;
 		}
 		counter->supported = false;
 		counter->errored = true;
 	}
 
+skip_or_fatal:
+	if (counter->skippable) {
+		if (verbose > 0 && !warned) {
+			ui__warning("skipping event %s that kernel failed to open .\n",
+				    evsel__name(counter));
+		}
+		counter->supported = false;
+		counter->errored = true;
+		return COUNTER_SKIP;
+	}
+
 	evsel__open_strerror(counter, &target, err, msg, sizeof(msg));
 	ui__error("%s\n", msg);
 
-- 
2.51.0.534.gc79095c0ca-goog