Add --valgringing flag instead of --nosetsched

The do_setsched was being tested for whether to mmap tdbs: let's make it
explicit.  We can also happily move the kill-child eventscript hack under
this flag.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au> 


(This used to be ctdb commit 2ee86cc1f311d7b7504c7b14d142b9c4f6f4b469)

8 files changed, 19 insertions, 10 deletions

diff --git a/ctdb/client/ctdb_client.c b/ctdb/client/ctdb_client.c
index 0469ba7280..75bc0a45fb 100644
--- a/ctdb/client/ctdb_client.c
+++ b/ctdb/client/ctdb_client.c
@@ -1683,7 +1683,7 @@ struct ctdb_db_context *ctdb_attach(struct ctdb_context *ctdb, const char *name,
 	}
 
 	tdb_flags = persistent?TDB_DEFAULT:TDB_NOSYNC;
-	if (!ctdb->do_setsched) {
+	if (ctdb->valgrinding) {
 		tdb_flags |= TDB_NOMMAP;
 	}
 
diff --git a/ctdb/config/ctdb.init b/ctdb/config/ctdb.init
index 67747fd149..1bd329ea71 100755
--- a/ctdb/config/ctdb.init
+++ b/ctdb/config/ctdb.init
@@ -153,7 +153,7 @@ start() {
     case $init_style in
 	valgrind)
 	    valgrind -q --log-file=/var/log/ctdb_valgrind \
-		$ctdbd --nosetsched $CTDB_OPTIONS 
+		$ctdbd --valgrinding $CTDB_OPTIONS
 	    RETVAL=$?
 	    echo
 	    ;;
diff --git a/ctdb/include/ctdb_private.h b/ctdb/include/ctdb_private.h
index feff0bc24e..e577abc9f8 100644
--- a/ctdb/include/ctdb_private.h
+++ b/ctdb/include/ctdb_private.h
@@ -447,6 +447,7 @@ struct ctdb_context {
 	struct ctdb_log_state *log;
 	int start_as_disabled;
 	int start_as_stopped;
+	bool valgrinding;
 	uint32_t event_script_timeouts; /* counting how many consecutive times an eventscript has timedout */
 	uint32_t *recd_ping_count;
 	TALLOC_CTX *release_ips_ctx; /* a context used to automatically drop all IPs if we fail to recover the node */
diff --git a/ctdb/server/ctdb_ltdb_server.c b/ctdb/server/ctdb_ltdb_server.c
index 37abe116da..6c204821dc 100644
--- a/ctdb/server/ctdb_ltdb_server.c
+++ b/ctdb/server/ctdb_ltdb_server.c
@@ -248,7 +248,7 @@ static int ctdb_local_attach(struct ctdb_context *ctdb, const char *db_name, boo
 					   db_name, ctdb->pnn);
 
 	tdb_flags = persistent? TDB_DEFAULT : TDB_CLEAR_IF_FIRST | TDB_NOSYNC;
-	if (!ctdb->do_setsched) {
+	if (ctdb->valgrinding) {
 		tdb_flags |= TDB_NOMMAP;
 	}
 
diff --git a/ctdb/server/ctdb_recoverd.c b/ctdb/server/ctdb_recoverd.c
index 3e596da9ec..45b59d128a 100644
--- a/ctdb/server/ctdb_recoverd.c
+++ b/ctdb/server/ctdb_recoverd.c
@@ -1017,7 +1017,7 @@ static struct tdb_wrap *create_recdb(struct ctdb_context *ctdb, TALLOC_CTX *mem_
 	unlink(name);
 
 	tdb_flags = TDB_NOLOCK;
-	if (!ctdb->do_setsched) {
+	if (ctdb->valgrinding) {
 		tdb_flags |= TDB_NOMMAP;
 	}
 
diff --git a/ctdb/server/ctdb_vacuum.c b/ctdb/server/ctdb_vacuum.c
index d09c07d9ff..655ecca6cf 100644
--- a/ctdb/server/ctdb_vacuum.c
+++ b/ctdb/server/ctdb_vacuum.c
@@ -514,6 +514,7 @@ static int update_tuning_db(struct ctdb_db_context *ctdb_db, struct vacuum_data
 	struct vacuum_tuning_data tdata;
 	struct vacuum_tuning_data *tptr;
 	char *vac_dbname;
+	int flags;
 
 	vac_dbname = talloc_asprintf(tmp_ctx, "%s/%s.%u",
 					ctdb_db->ctdb->db_directory, 
@@ -524,7 +525,8 @@ static int update_tuning_db(struct ctdb_db_context *ctdb_db, struct vacuum_data
 		return -1;
 	}
 
-	tune_tdb = tdb_open(vac_dbname, 0, 0, O_RDWR|O_CREAT, 0644);
+	flags = ctdb_db->ctdb->valgrinding ? TDB_NOMMAP : 0;
+	tune_tdb = tdb_open(vac_dbname, 0, flags, O_RDWR|O_CREAT, 0644);
 	if (tune_tdb == NULL) {
 		DEBUG(DEBUG_ERR,(__location__ " Failed to create/open %s\n", TUNINGDBNAME));
 		talloc_free(tmp_ctx);
@@ -677,6 +679,7 @@ static int get_vacuum_interval(struct ctdb_db_context *ctdb_db)
 	char *vac_dbname;
 	uint interval = ctdb_db->ctdb->tunable.vacuum_default_interval;
 	struct ctdb_context *ctdb = ctdb_db->ctdb;
+	int flags;
 
 	vac_dbname = talloc_asprintf(tmp_ctx, "%s/%s.%u", ctdb->db_directory, TUNINGDBNAME, ctdb->pnn);
 	if (vac_dbname == NULL) {
@@ -685,7 +688,8 @@ static int get_vacuum_interval(struct ctdb_db_context *ctdb_db)
 		return interval;
 	}
 
-	tdb = tdb_open(vac_dbname, 0, 0, O_RDWR|O_CREAT, 0644);
+	flags = ctdb_db->ctdb->valgrinding ? TDB_NOMMAP : 0;
+	tdb = tdb_open(vac_dbname, 0, flags, O_RDWR|O_CREAT, 0644);
 	if (!tdb) {
 		DEBUG(DEBUG_ERR,("Unable to open/create database %s using default interval\n", vac_dbname));
 		talloc_free(tmp_ctx);
diff --git a/ctdb/server/ctdbd.c b/ctdb/server/ctdbd.c
index 8647aad189..f38ed66d83 100644
--- a/ctdb/server/ctdbd.c
+++ b/ctdb/server/ctdbd.c
@@ -41,7 +41,7 @@ static struct {
 	const char *public_interface;
 	const char *single_public_ip;
 	const char *node_ip;
-	int         no_setsched;
+	int         valgrinding;
 	int         use_syslog;
 	int         start_as_disabled;
 	int         start_as_stopped;
@@ -127,7 +127,7 @@ int main(int argc, const char *argv[])
 		{ "dbdir", 0, POPT_ARG_STRING, &options.db_dir, 0, "directory for the tdb files", NULL },
 		{ "dbdir-persistent", 0, POPT_ARG_STRING, &options.db_dir_persistent, 0, "directory for persistent tdb files", NULL },
 		{ "reclock", 0, POPT_ARG_STRING, &options.recovery_lock_file, 0, "location of recovery lock file", "filename" },
-		{ "nosetsched", 0, POPT_ARG_NONE, &options.no_setsched, 0, "disable setscheduler SCHED_FIFO call", NULL },
+		{ "valgrinding", 0, POPT_ARG_NONE, &options.valgrinding, 0, "disable setscheduler SCHED_FIFO call, use mmap for tdbs", NULL },
 		{ "syslog", 0, POPT_ARG_NONE, &options.use_syslog, 0, "log messages to syslog", NULL },
 		{ "start-as-disabled", 0, POPT_ARG_NONE, &options.start_as_disabled, 0, "Node starts in disabled state", NULL },
 		{ "start-as-stopped", 0, POPT_ARG_NONE, &options.start_as_stopped, 0, "Node starts in stopped state", NULL },
@@ -311,7 +311,8 @@ int main(int argc, const char *argv[])
 		}
 	}
 
-	ctdb->do_setsched = !options.no_setsched;
+	ctdb->valgrinding = options.valgrinding;
+	ctdb->do_setsched = !ctdb->valgrinding;
 
 	ctdb->do_checkpublicip = !options.no_publicipcheck;
 
diff --git a/ctdb/server/eventscript.c b/ctdb/server/eventscript.c
index 59cb6a3006..803ac1daa0 100644
--- a/ctdb/server/eventscript.c
+++ b/ctdb/server/eventscript.c
@@ -452,7 +452,10 @@ static void ctdb_event_script_handler(struct event_context *ev, struct fd_event
 
 	/* valgrind gets overloaded if we run next script as it's still doing
 	 * post-execution analysis, so kill finished child here. */
-	kill(state->child, SIGKILL);
+	if (ctdb->valgrinding) {
+		kill(state->child, SIGKILL);
+	}
+
 	state->child = 0;
 
 	/* Aborted or finished all scripts?  We're done. */