From 4c0758c36fc189fe33467030dbf3b5438b206d59 Mon Sep 17 00:00:00 2001
From: Mauro Carvalho Chehab <[email protected]>
Date: Fri, 19 Jul 2024 08:24:25 +0200
Subject: [PATCH 64/85] ras-events: make returned error code consistent

- Rework the returned code logic to be more consistent;
  - error codes will be using negative values;
  - positive values indicate special return codes.
- Don't bloat the logs with lots of error messages due to
  unsupported traces;
- Ensure that the number of CPUs will probably retrieved or bail out;
- Don't bail if it can't setup a monotone clock: it is better
  to have a wrong timestamp than no log at all.

Signed-off-by: Mauro Carvalho Chehab <[email protected]>
---
 ras-events.c | 91 +++++++++++++++++++++++++++++++---------------------
 1 file changed, 54 insertions(+), 37 deletions(-)

diff --git a/ras-events.c b/ras-events.c
index acde772..0c4fe60 100644
--- a/ras-events.c
+++ b/ras-events.c
@@ -78,7 +78,7 @@ static int get_debugfs_dir(char *tracing_dir, size_t len)
 	fp = fopen("/proc/mounts", "r");
 	if (!fp) {
 		log(ALL, LOG_INFO, "Can't open /proc/mounts");
-		return errno;
+		return -errno;
 	}
 
 	do {
@@ -124,7 +124,11 @@ static int open_trace(struct ras_events *ras, char *name, int flags)
 	if (rc < 0)
 		return rc;
 
-	return open(fname, flags);
+	rc = open(fname, flags);
+	if (rc)
+		return -errno;
+
+	return 0;
 }
 
 static int get_tracing_dir(struct ras_events *ras)
@@ -145,7 +149,7 @@ static int get_tracing_dir(struct ras_events *ras)
 
 	dir = opendir(fname);
 	if (!dir)
-		return -1;
+		return -EINVAL;
 
 	for (entry = readdir(dir); entry; entry = readdir(dir)) {
 		if (strstr(entry->d_name, "instances")) {
@@ -168,13 +172,13 @@ static int get_tracing_dir(struct ras_events *ras)
 			log(ALL, LOG_INFO,
 			    "Unable to create " TOOL_NAME " instance at %s\n",
 			    ras->tracing);
-			return -1;
+			return -EINVAL;
 		}
 	}
 	return 0;
 }
 
-static int is_disabled_event(char *group, char *event)
+static bool is_disabled_event(char *group, char *event)
 {
 	char ras_event_name[MAX_PATH + 1];
 
@@ -183,9 +187,9 @@ static int is_disabled_event(char *group, char *event)
 
 	if (choices_disable && strlen(choices_disable) != 0 &&
 	    strstr(choices_disable, ras_event_name)) {
-		return 1;
+		return true;
 	}
-	return 0;
+	return false;
 }
 
 /*
@@ -208,7 +212,7 @@ static int __toggle_ras_mc_event(struct ras_events *ras,
 	fd = open_trace(ras, "set_event", O_RDWR | O_APPEND);
 	if (fd < 0) {
 		log(ALL, LOG_WARNING, "Can't open set_event\n");
-		return errno;
+		return -errno;
 	}
 
 	rc = write(fd, fname, strlen(fname));
@@ -220,7 +224,7 @@ static int __toggle_ras_mc_event(struct ras_events *ras,
 	close(fd);
 	if (!rc) {
 		log(ALL, LOG_WARNING, "Nothing was written on set_event\n");
-		return EIO;
+		return -EIO;
 	}
 
 	log(ALL, LOG_INFO, "%s:%s event %s\n",
@@ -238,7 +242,7 @@ int toggle_ras_mc_event(int enable)
 	ras = calloc(1, sizeof(*ras));
 	if (!ras) {
 		log(TERM, LOG_ERR, "Can't allocate memory for ras struct\n");
-		return errno;
+		return -errno;
 	}
 
 	rc = get_tracing_dir(ras);
@@ -283,7 +287,10 @@ int toggle_ras_mc_event(int enable)
 
 free_ras:
 	free(ras);
-	return rc;
+	if (rc)
+		return -EINVAL;
+
+	return 0;
 }
 
 static void setup_event_trigger(char *event)
@@ -421,17 +428,26 @@ static int set_buffer_percent(struct ras_events *ras, int percent)
 		size = write(fd, buf, strlen(buf));
 		if (size <= 0) {
 			log(TERM, LOG_WARNING, "can't write to buffer_percent\n");
-			res = -1;
+			res = -EINVAL;
 		}
 		close(fd);
 	} else {
 		log(TERM, LOG_WARNING, "Can't open buffer_percent\n");
-		res = -1;
+		res = -EINVAL;
 	}
 
 	return res;
 }
 
+/*
+ * Kernel tracepoint had an incompatible change in 2019, causing polling
+ * tracepoints to fail. Rasdaemon can support both legacy and newer versions,
+ * with the help of a backup-compatibility legacy kernel mode.
+ *
+ * The LEGACY_KERNEL flag indicates the need to enable such code.
+ */
+#define LEGACY_KERNEL		255
+
 static int read_ras_event_all_cpus(struct pthread_data *pdata,
 				   unsigned int n_cpus)
 {
@@ -621,9 +637,9 @@ error:
 	}
 
 	if (legacy_kernel)
-		return -255;
-	else
-		return -1;
+		return LEGACY_KERNEL;
+
+	return -EINVAL;
 }
 
 static int read_ras_event(int fd,
@@ -644,7 +660,7 @@ static int read_ras_event(int fd,
 		size = read(fd, page, pdata->ras->page_size);
 		if (size < 0) {
 			log(TERM, LOG_WARNING, "read\n");
-			return -1;
+			return -EINVAL;
 		} else if (size > 0) {
 			kbuffer_load_subbuffer(kbuf, page);
 
@@ -745,13 +761,13 @@ static int select_tracing_timestamp(struct ras_events *ras)
 	fd = open_trace(ras, "trace_clock", O_RDONLY);
 	if (fd < 0) {
 		log(TERM, LOG_ERR, "Can't open trace_clock\n");
-		return -1;
+		return -EINVAL;
 	}
 	size = read(fd, buf, sizeof(buf));
 	close(fd);
 	if (!size) {
 		log(TERM, LOG_ERR, "trace_clock is empty!\n");
-		return -1;
+		return -EINVAL;
 	}
 
 	if (!strstr(buf, UPTIME)) {
@@ -786,7 +802,7 @@ static int select_tracing_timestamp(struct ras_events *ras)
 	fclose(fp);
 	if (rc <= 0) {
 		log(TERM, LOG_ERR, "Can't parse /proc/uptime!\n");
-		return -1;
+		return -EINVAL;
 	}
 	now = time(NULL);
 
@@ -796,6 +812,7 @@ static int select_tracing_timestamp(struct ras_events *ras)
 	return 0;
 }
 
+#define EVENT_DISABLED	1
 static int add_event_handler(struct ras_events *ras, struct pevent *pevent,
 			     unsigned page_size, char *group, char *event,
 			     pevent_event_handler_func func, char *filter_str, int id)
@@ -811,14 +828,14 @@ static int add_event_handler(struct ras_events *ras, struct pevent *pevent,
 		log(TERM, LOG_ERR,
 		    "Can't get %s:%s traces. Perhaps this feature is not supported on your system.\n",
 		    group, event);
-		return errno;
+		return -errno;
 	}
 
 	page = malloc(page_size);
 	if (!page) {
 		log(TERM, LOG_ERR, "Can't allocate page to read %s:%s format\n",
 		    group, event);
-		rc = errno;
+		rc = -errno;
 		close(fd);
 		return rc;
 	}
@@ -837,14 +854,14 @@ static int add_event_handler(struct ras_events *ras, struct pevent *pevent,
 		log(TERM, LOG_ERR, "Can't register event handler for %s:%s\n",
 		    group, event);
 		free(page);
-		return EINVAL;
+		return -EINVAL;
 	}
 
 	rc = pevent_parse_event(pevent, page, size, group);
 	if (rc) {
 		log(TERM, LOG_ERR, "Can't parse event %s:%s\n", group, event);
 		free(page);
-		return EINVAL;
+		return -EINVAL;
 	}
 
 	if (filter_str) {
@@ -855,7 +872,7 @@ static int add_event_handler(struct ras_events *ras, struct pevent *pevent,
 			log(TERM, LOG_ERR,
 			    "Failed to allocate filter for %s/%s.\n", group, event);
 			free(page);
-			return EINVAL;
+			return -EINVAL;
 		}
 		rc = pevent_filter_add_filter_str(filter, filter_str, &error);
 		if (rc) {
@@ -872,7 +889,7 @@ static int add_event_handler(struct ras_events *ras, struct pevent *pevent,
 	if (is_disabled_event(group, event)) {
 		log(ALL, LOG_INFO, "Disabled %s:%s tracing from config\n",
 		    group, event);
-		return -EINVAL;
+		return EVENT_DISABLED;
 	}
 
 	/* Enable RAS events */
@@ -882,7 +899,7 @@ static int add_event_handler(struct ras_events *ras, struct pevent *pevent,
 		log(TERM, LOG_ERR, "Can't enable %s:%s tracing\n",
 		    group, event);
 
-		return EINVAL;
+		return -EINVAL;
 	}
 
 	setup_event_trigger(event);
@@ -907,7 +924,7 @@ int handle_ras_events(int record_events)
 	ras = calloc(1, sizeof(*ras));
 	if (!ras) {
 		log(TERM, LOG_ERR, "Can't allocate memory for ras struct\n");
-		return errno;
+		return -errno;
 	}
 
 	rc = get_tracing_dir(ras);
@@ -925,7 +942,7 @@ int handle_ras_events(int record_events)
 	pevent = pevent_alloc();
 	if (!pevent) {
 		log(TERM, LOG_ERR, "Can't allocate pevent\n");
-		rc = errno;
+		rc = -errno;
 		goto err;
 	}
 
@@ -944,7 +961,7 @@ int handle_ras_events(int record_events)
 			       ras_mc_event_handler, NULL, MC_EVENT);
 	if (!rc)
 		num_events++;
-	else if (rc != -EINVAL)
+	else if (rc != EVENT_DISABLED)
 		log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
 		    "ras", "mc_event");
 
@@ -953,7 +970,7 @@ int handle_ras_events(int record_events)
 			       ras_aer_event_handler, NULL, AER_EVENT);
 	if (!rc)
 		num_events++;
-	else if (rc != -EINVAL)
+	else if (rc != EVENT_DISABLED)
 		log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
 		    "ras", "aer_event");
 #endif
@@ -963,7 +980,7 @@ int handle_ras_events(int record_events)
 			       ras_non_standard_event_handler, NULL, NON_STANDARD_EVENT);
 	if (!rc)
 		num_events++;
-	else if (rc != -EINVAL)
+	else if (rc != EVENT_DISABLED)
 		log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
 		    "ras", "non_standard_event");
 #endif
@@ -973,7 +990,7 @@ int handle_ras_events(int record_events)
 			       ras_arm_event_handler, NULL, ARM_EVENT);
 	if (!rc)
 		num_events++;
-	else if (rc != -EINVAL)
+	else if (rc != EVENT_DISABLED)
 		log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
 		    "ras", "arm_event");
 #endif
@@ -1007,7 +1024,7 @@ int handle_ras_events(int record_events)
 		/* tell kernel we are listening, so don't printk to console */
 		(void)open("/sys/kernel/debug/ras/daemon_active", 0);
 		num_events++;
-	} else if (rc != -EINVAL)
+	} else if (rc != EVENT_DISABLED)
 		log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
 		    "ras", "extlog_mem_event");
 #endif
@@ -1024,7 +1041,7 @@ int handle_ras_events(int record_events)
 			       ras_devlink_event_handler, filter_str, DEVLINK_EVENT);
 	if (!rc)
 		num_events++;
-	else if (rc != -EINVAL)
+	else if (rc != EVENT_DISABLED)
 		log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
 		    "devlink", "devlink_health_report");
 #endif
@@ -1048,7 +1065,7 @@ int handle_ras_events(int record_events)
 			       ras_memory_failure_event_handler, NULL, MF_EVENT);
 	if (!rc)
 		num_events++;
-	else if (rc != -EINVAL)
+	else if (rc != EVENT_DISABLED)
 		log(ALL, LOG_ERR, "Can't get traces from %s:%s\n",
 		    "ras", "memory_failure_event");
 #endif
@@ -1071,7 +1088,7 @@ int handle_ras_events(int record_events)
 	rc = read_ras_event_all_cpus(data, cpus);
 
 	/* Poll doesn't work on this kernel. Fallback to pthread way */
-	if (rc == -255) {
+	if (rc == LEGACY_KERNEL) {
 		if (pthread_mutex_init(&ras->db_lock, NULL) != 0) {
 			log(SYSLOG, LOG_INFO, "sqlite db lock init has failed\n");
 			goto err;
-- 
2.33.1