metrics: Expose more errors in tetragon_bpf_missed_events_total counter

When testing I saw a bunch of "unknown" errors. To further investigate, let's split out ENOENT, E2BIG and EINVAL into separate label values. This should be almost all errors returned by perf_event_output, I left out only EOPNOTSUPP. Signed-off-by: Anna Kapuscinska <anna@isovalent.com>
cilium · Sep 2, 2024 · 4166a3f · 4166a3f
1 parent b7ef6aa
commit 4166a3f
Show file tree

Hide file tree

Showing 4 changed files with 20 additions and 5 deletions.
diff --git a/bpf/lib/process.h b/bpf/lib/process.h
@@ -565,9 +565,12 @@ _Static_assert(sizeof(struct execve_map_value) % 8 == 0,
 	       "struct execve_map_value should have size multiple of 8 bytes");
 
 #define SENT_FAILED_UNKNOWN 0 // unknown error
-#define SENT_FAILED_EBUSY   1 // EBUSY
-#define SENT_FAILED_ENOSPC  2 // ENOSPC
-#define SENT_FAILED_MAX	    3
+#define SENT_FAILED_ENOENT  1 // ENOENT
+#define SENT_FAILED_E2BIG   2 // E2BIG
+#define SENT_FAILED_EBUSY   3 // EBUSY
+#define SENT_FAILED_EINVAL  4 // EINVAL
+#define SENT_FAILED_ENOSPC  5 // ENOSPC
+#define SENT_FAILED_MAX	    6
 
 struct kernel_stats {
 	__u64 sent_failed[256][SENT_FAILED_MAX];
@@ -591,8 +594,14 @@ perf_event_output_metric(void *ctx, u8 msg_op, void *map, u64 flags, void *data,
 	if (err < 0) {
 		valp = map_lookup_elem(&tg_stats_map, &zero);
 		if (valp) {
-			if (err == -16) // EBUSY
+			if (err == -2) // ENOENT
+				__sync_fetch_and_add(&valp->sent_failed[msg_op][SENT_FAILED_ENOENT], 1);
+			else if (err == -7) // E2BIG
+				__sync_fetch_and_add(&valp->sent_failed[msg_op][SENT_FAILED_E2BIG], 1);
+			else if (err == -16) // EBUSY
 				__sync_fetch_and_add(&valp->sent_failed[msg_op][SENT_FAILED_EBUSY], 1);
+			else if (err == -22) // EINVAL
+				__sync_fetch_and_add(&valp->sent_failed[msg_op][SENT_FAILED_EINVAL], 1);
 			else if (err == -28) // ENOSPC
 				__sync_fetch_and_add(&valp->sent_failed[msg_op][SENT_FAILED_ENOSPC], 1);
 			else

diff --git a/docs/content/en/docs/reference/metrics.md b/docs/content/en/docs/reference/metrics.md
diff --git a/pkg/api/processapi/processapi.go b/pkg/api/processapi/processapi.go
@@ -51,7 +51,10 @@ const (
 
 const (
 	SentFailedUnknown = iota
+	SentFailedEnoent
+	SentFailedE2big
 	SentFailedEbusy
+	SentFailedEinval
 	SentFailedEnospc
 	SentFailedMax
 )

diff --git a/pkg/metrics/eventmetrics/eventmetrics.go b/pkg/metrics/eventmetrics/eventmetrics.go
@@ -25,7 +25,10 @@ import (
 var (
 	perfEventErrors = map[int]string{
 		processapi.SentFailedUnknown: "unknown",
+		processapi.SentFailedEnoent:  "ENOENT",
+		processapi.SentFailedE2big:   "E2BIG",
 		processapi.SentFailedEbusy:   "EBUSY",
+		processapi.SentFailedEinval:  "EINVAL",
 		processapi.SentFailedEnospc:  "ENOSPC",
 	}
 	perfEventErrorLabel = metrics.ConstrainedLabel{