Aer support

misc/rasdaemon.env

@@ -88,6 +88,8 @@ TRIGGER_DIR=
 #   MC_UE_TRIGGER=mc_event_trigger
 MC_CE_TRIGGER=
 MC_UE_TRIGGER=
+AER_CE_TRIGGER=
+AER_UE_TRIGGER=
 
 # CE Statistic Threshold
 #

ras-aer-handler.c

@@ -4,9 +4,11 @@
  * Copyright (C) 2013 Mauro Carvalho Chehab <[email protected]>
  */
 
+#define _GNU_SOURCE
 #include <pci/pci.h>
 #include <stdio.h>
 #include <stdlib.h>
+#include <stdio.h>
 #include <string.h>
 #include <traceevent/kbuffer.h>
 #include <unistd.h>
@@ -16,6 +18,7 @@
 #include "ras-logger.h"
 #include "ras-report.h"
 #include "unified-sel.h"
+#include "trigger.h"
 #include "types.h"
 
 /* bit field meaning for correctable error */
@@ -55,6 +58,47 @@ static const char *aer_uncor_errors[32] = {
 
 static bool use_ipmitool = false;
 
+#define MAX_ENV 30
+static const char *aer_ce_trigger = NULL;
+static const char *aer_ue_trigger = NULL;
+
+void aer_event_trigger_setup(void)
+{
+	const char *trigger;
+
+	trigger = getenv("AER_CE_TRIGGER");
+	if (trigger && strcmp(trigger, "")) {
+		aer_ce_trigger = trigger_check(trigger);
+
+		if (!aer_ce_trigger) {
+			log(ALL, LOG_ERR,
+			    "Cannot access aer_event ce trigger `%s`\n",
+			    trigger);
+		} else {
+			log(ALL, LOG_INFO,
+			    "Setup aer_event ce trigger `%s`\n",
+			    trigger);
+		}
+	} else {
+		log(TERM, LOG_ERR, "\t no AER_CE_TRIGGER (%p)\n", trigger);
+	}
+
+	trigger = getenv("AER_UE_TRIGGER");
+	if (trigger && strcmp(trigger, "")) {
+		aer_ue_trigger = trigger_check(trigger);
+
+		if (!aer_ue_trigger) {
+			log(ALL, LOG_ERR,
+			    "Cannot access aer_event ue trigger `%s`\n",
+			    trigger);
+		} else {
+			log(ALL, LOG_INFO,
+			    "Setup aer_event ue trigger `%s`\n",
+			    trigger);
+		}
+	}
+}
+
 void ras_aer_handler_init(int enable_ipmitool)
 {
 #ifdef HAVE_OPENBMC_UNIFIED_SEL
@@ -103,6 +147,32 @@ static void get_pci_dev_name(char *bdf, char *pci_name, ssize_t len, u16 *vendor
 	pci_cleanup(pacc);
 }
 
+static void run_aer_trigger(struct ras_aer_event *ev, const char *aer_trigger)
+{
+	char *env[MAX_ENV];
+	int ei = 0;
+	int i;
+
+	if (asprintf(&env[ei++], "PATH=%s", getenv("PATH") ?: "/sbin:/usr/sbin:/bin:/usr/bin") < 0)
+		goto free;
+	if (asprintf(&env[ei++], "TIMESTAMP=%s", ev->timestamp) < 0)
+		goto free;
+	if (asprintf(&env[ei++], "TYPE=%s", ev->error_type) < 0)
+		goto free;
+	if (asprintf(&env[ei++], "MESSAGE=%s", ev->msg) < 0)
+		goto free;
+	if (asprintf(&env[ei++], "NAME=%s", ev->dev_name) < 0)
+		goto free;
+	env[ei] = NULL;
+	assert(ei < MAX_ENV);
+
+	run_trigger(aer_trigger, NULL, env, "aer_event");
+
+free:
+	for (i = 0; i < ei; i++)
+		free(env[i]);
+}
+
 int ras_aer_event_handler(struct trace_seq *s,
 			  struct tep_record *record,
 			  struct tep_event *event, void *context)
@@ -270,5 +340,11 @@ int ras_aer_event_handler(struct trace_seq *s,
 			return -1;
 #endif
 
+	if (aer_ce_trigger && !strcmp(ev.error_type, "Corrected"))
+		run_aer_trigger(&ev, aer_ce_trigger);
+
+	if (aer_ue_trigger && !strncmp(ev.error_type, "Uncorrected", 11))
+		run_aer_trigger(&ev, aer_ue_trigger);
+
 	return 0;
 }

ras-aer-handler.h

@@ -11,6 +11,7 @@
 
 #include "ras-events.h"
 
+void aer_event_trigger_setup(void);
 int ras_aer_event_handler(struct trace_seq *s,
 			  struct tep_record *record,
 			  struct tep_event *event, void *context);

ras-events.c

@@ -29,6 +29,7 @@
 #include "ras-events.h"
 #include "ras-extlog-handler.h"
 #include "ras-logger.h"
+#include "ras-aer-handler.h"
 #include "ras-mce-handler.h"
 #include "ras-mc-handler.h"
 #include "ras-memory-failure-handler.h"
@@ -47,15 +48,18 @@
 
 /* Test for a little-endian machine */
 #if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
-	#define ENDIAN KBUFFER_ENDIAN_LITTLE
+	#define ENDIAN TEP_LITTLE_ENDIAN
 #else
-	#define ENDIAN KBUFFER_ENDIAN_BIG
+	#define ENDIAN TEP_BIG_ENDIAN
 #endif
 
 char *choices_disable;
 
 static const struct event_trigger event_triggers[] = {
 	{ "mc_event", &mc_event_trigger_setup },
+#ifdef HAVE_AER
+	{ "aer_event", &aer_event_trigger_setup },
+#endif
 #ifdef HAVE_MEMORY_FAILURE
 	{ "memory_failure_event", &mem_fail_event_trigger_setup },
 #endif
@@ -418,6 +422,7 @@ static void parse_ras_data(struct pthread_data *pdata, struct kbuffer *kbuf,
 
 	/* TODO - logging */
 	trace_seq_init(&s);
+	tep_set_file_bigendian(pdata->ras->pevent, ENDIAN);
 	tep_print_event(pdata->ras->pevent, &s, &record,
 			"%16s-%-5d [%03d] %s %6.1000d %s %s",
 			TEP_PRINT_COMM, TEP_PRINT_PID, TEP_PRINT_CPU,
@@ -498,7 +503,7 @@ static int read_ras_event_all_cpus(struct pthread_data *pdata,
 		return -ENOMEM;
 	}
 
-	kbuf = kbuffer_alloc(KBUFFER_LSIZE_8, ENDIAN);
+	kbuf = kbuffer_alloc(KBUFFER_LSIZE_SAME_AS_HOST, KBUFFER_ENDIAN_SAME_AS_HOST);
 	if (!kbuf) {
 		log(TERM, LOG_ERR, "Can't allocate kbuf\n");
 		free(page);
@@ -705,7 +710,7 @@ static void *handle_ras_events_cpu(void *priv)
 		return NULL;
 	}
 
-	kbuf = kbuffer_alloc(KBUFFER_LSIZE_8, ENDIAN);
+	kbuf = kbuffer_alloc(KBUFFER_LSIZE_SAME_AS_HOST, KBUFFER_ENDIAN_SAME_AS_HOST);
 	if (!kbuf) {
 		log(TERM, LOG_ERR, "Can't allocate kbuf");
 		free(page);