intel_pmu 2.0 - bugfix & more verbose warning messages

Fixed issue on HardwareEvents option set to "All".
Warning messages are more verbose when max number of opened files is reached and max number of events per group is reached.

Signed-off-by: Pawel Tomaszewski <pawelx.tomaszewski@intel.com>

diff --git a/src/intel_pmu.c b/src/intel_pmu.c
index c57c05fc415d9e8d74721966a3bf4eeec534cfaa..00358d9afd35c4b2d9e2d8a967e2f0e3d67fcbe7 100644 (file)
--- a/src/intel_pmu.c
+++ b/src/intel_pmu.c
@@ -44,6 +44,7 @@ struct intel_pmu_entity_s {
   size_t first_cgroup;
   size_t cgroups_count;
   bool copied;
+  bool all_events;
   struct eventlist *event_list;
   user_data_t user_data;
   struct intel_pmu_entity_s *next;
@@ -53,7 +54,6 @@ typedef struct intel_pmu_entity_s intel_pmu_entity_t;
 struct intel_pmu_ctx_s {
   char event_list_fn[PATH_MAX];
   bool dispatch_cloned_pmus;
-  bool all_events;
 
   intel_pmu_entity_t *entl;
 };
@@ -184,7 +184,7 @@ static int pmu_config_hw_events(oconfig_item_t *ci, intel_pmu_entity_t *ent) {
 
     if (strcasecmp(ci->values[i].value.string, "All") == 0) {
       INFO(PMU_PLUGIN ": Requested all events.");
-      g_ctx.all_events = true;
+      ent->all_events = true;
       return 0;
     }
   }
@@ -595,7 +595,6 @@ static int pmu_setup_events(core_groups_list_t *cores, struct eventlist *el,
                             bool measure_all, int measure_pid) {
   struct event *e, *leader = NULL;
   int ret = -1;
-
   for (e = el->eventlist; e; e = e->next) {
 
     for (size_t i = 0; i < cores->num_cgroups; i++) {
@@ -614,9 +613,19 @@ static int pmu_setup_events(core_groups_list_t *cores, struct eventlist *el,
             continue;
         }
 
-        if (setup_event(e, core, leader, measure_all, measure_pid) < 0) {
+        int res = setup_event(e, core, leader, measure_all, measure_pid);
+        if (res < 0 && errno == EMFILE) {
+          WARNING(PMU_PLUGIN
+                  ": perf event '%s' is not available (cpu=%d). "
+                  "Max number of open files reached for current process.",
+                  e->event, core);
+        } else if (res < 0) {
           WARNING(PMU_PLUGIN ": perf event '%s' is not available (cpu=%d).",
                   e->event, core);
+        } else if (e->efd[core].fd < 0) {
+          WARNING(PMU_PLUGIN ": max number of events "
+                             "per group reached for event '%s' (cpu=%d).",
+                  e->event, core);
         } else {
           /* success if at least one event was set */
           ret = 0;
@@ -683,6 +692,11 @@ static int pmu_read_all_events(void *data, char *name, char *event,
 
   event_counter++;
 
+  /* zeroing event_counter for next cores events */
+  if (event_counter == ent->hw_events_count) {
+    event_counter = 0;
+  }
+
   return 0;
 }
 
@@ -730,9 +744,14 @@ static int pmu_init(void) {
   }
 
   /* write all events from provided EventList into hw_events */
-  if (g_ctx.all_events) {
-    for (intel_pmu_entity_t *ent = g_ctx.entl; ent != NULL; ent = ent->next) {
-      walk_events(pmu_count_all_events, ent);
+  for (intel_pmu_entity_t *ent = g_ctx.entl; ent != NULL; ent = ent->next) {
+    if (ent->all_events) {
+      ret = walk_events(pmu_count_all_events, ent);
+      if (ret != 0) {
+        ERROR(PMU_PLUGIN ": Invalid core groups configuration.");
+        goto init_error;
+      }
+
       // allocating memory for all events
       ent->hw_events = calloc(ent->hw_events_count, sizeof(*ent->hw_events));
       if (ent->hw_events == NULL) {
@@ -740,7 +759,11 @@ static int pmu_init(void) {
         return -ENOMEM;
       }
 
-      walk_events(pmu_read_all_events, ent);
+      ret = walk_events(pmu_read_all_events, ent);
+      if (ret != 0) {
+        ERROR(PMU_PLUGIN ": Invalid core groups configuration.");
+        goto init_error;
+      }
     }
   }