* Changed signal handling

* Moved clixon-proc sigchild handling	from handler to clixon_events

CHANGELOG.md

@@ -53,6 +53,8 @@ Developers may need to change their code
 
 ### Minor features
 
+* Changed signal handling
+  * Moved clixon-proc sigchild handling	from handler to clixon_events
 * The base capability has been changed to "urn:ietf:params:netconf:base:1.1" following RFC6241.
 * Made a separate Clixon datastore XML/JSON top-level symbol
   * Replaces the hardcoded "config" keyword.

apps/backend/backend_main.c

@@ -166,13 +166,8 @@ backend_sig_term(int arg)
 static void
 backend_sig_child(int arg)
 {
-    int status;
-    int pid;
-
     clicon_debug(1, "%s", __FUNCTION__);
-    if ((pid = waitpid(-1, &status, 0)) != -1 && WIFEXITED(status)){
+    clicon_sig_child_set(1);
-    }
-    clicon_sig_ignore_set(1);
 }
 
 /*! Create backend server socket and register callback
@@ -1038,7 +1033,7 @@ main(int    argc,
 	goto done;
     if (stream_timer_setup(0, h) < 0)
 	goto done;
-    if (clixon_event_loop() < 0)
+    if (clixon_event_loop(h) < 0)
 	goto done;
  ok:
     retval = 0;

apps/netconf/netconf_main.c

@@ -873,7 +873,7 @@ main(int    argc,
 	if (clixon_event_reg_timeout(t, timeout_fn, NULL, "timeout") < 0)
 	    goto done;
     }
-    if (clixon_event_loop() < 0)
+    if (clixon_event_loop(h) < 0)
 	goto done;
     retval = 0;
   done:

apps/restconf/restconf_lib.c

@@ -256,6 +256,7 @@ restconf_terminate(clicon_handle h)
     xpath_optimize_exit();
     restconf_handle_exit(h);
     clicon_log_exit();
+    clicon_debug(1, "%s done", __FUNCTION__);
     return 0;
 }
 

apps/restconf/restconf_main_evhtp.c

@@ -160,17 +160,6 @@ restconf_sig_term(int arg)
     exit(-1);
 }
 
-static void
-restconf_sig_child(int arg)
-{
-    int status;
-    int pid;
-
-    if ((pid = waitpid(-1, &status, 0)) != -1 && WIFEXITED(status)){
-    }
-    clicon_sig_ignore_set(1);
-}
-
 static char*
 evhtp_method2str(enum htp_method m)
 {
@@ -1286,11 +1275,6 @@ main(int    argc,
 	clicon_err(OE_DAEMON, errno, "Setting signal");
 	goto done;
     }
-    if (set_signal(SIGCHLD, restconf_sig_child, NULL) < 0){
-	clicon_err(OE_DAEMON, errno, "Setting signal");
-	goto done;
-    }
-    
     /* Find and read configfile */
     if (clicon_options_main(h) < 0)
 	goto done;

apps/restconf/restconf_main_fcgi.c

@@ -63,7 +63,6 @@
 #include <pwd.h>
 #include <time.h>
 #include <limits.h>
-#include <signal.h>
 #include <sys/time.h>
 #include <sys/wait.h>
 #include <libgen.h>
@@ -146,9 +145,13 @@ restconf_sig_term(int arg)
 	restconf_terminate(_CLICON_HANDLE);
     }
     clicon_exit_set(); /* checked in clixon_event_loop() */
+    clicon_debug(1, "%s done", __FUNCTION__);
     exit(-1);
 }
 
+/*! Reap stream child
+ * XXX The -1 should be changed to proper pid, see eg clixon_process_waitpid
+ */
 static void
 restconf_sig_child(int arg)
 {

apps/restconf/restconf_stream_fcgi.c

@@ -463,7 +463,7 @@ api_stream(clicon_handle h,
 	    /* Poll upstream errors */
 	    stream_timeout(0, req);
 	    /* Start loop */
-	    clixon_event_loop();
+	    clixon_event_loop(h);
 	    close(s);
 	    clixon_event_unreg_fd(s, restconf_stream_cb);
 	    clixon_event_unreg_fd(rfcgi->listen_sock,

lib/clixon/clixon_event.h

@@ -49,6 +49,10 @@ int clicon_exit_reset(void);
 
 int clicon_exit_get(void);
 
+int clicon_sig_child_set(int val);
+
+int clicon_sig_child_get(void);
+
 int clicon_sig_ignore_set(int val);
 
 int clicon_sig_ignore_get(void);
@@ -64,7 +68,7 @@ int clixon_event_unreg_timeout(int (*fn)(int, void*), void *arg);
 
 int clixon_event_poll(int fd);
 
-int clixon_event_loop(void);
+int clixon_event_loop(clicon_handle h);
 
 int clixon_event_exit(void);
 

lib/clixon/clixon_proc.h

@@ -67,5 +67,6 @@ int clixon_process_delete_all(clicon_handle h);
 int clixon_process_operation(clicon_handle h, const char *name, proc_operation op, const int wrapit, uint32_t *pid);
 int clixon_process_start_all(clicon_handle h);
 int clixon_process_sched_register(clicon_handle h);
+int clixon_process_waitpid(clicon_handle h);
 
 #endif  /* _CLIXON_PROC_H_ */

lib/src/clixon_event.c

@@ -54,8 +54,11 @@
 
 #include "clixon_queue.h"
 #include "clixon_log.h"
+#include "clixon_hash.h"
+#include "clixon_handle.h"
 #include "clixon_err.h"
 #include "clixon_sig.h"
+#include "clixon_proc.h"
 #include "clixon_event.h"
 
 /*
@@ -89,6 +92,9 @@ static int _ee_unreg = 0;
 /* If set (eg by signal handler) exit select loop on next run and return 0 */
 static int _clicon_exit = 0;
 
+/* If set (eg by signal handler) call waitpid on waiting processes, ignore EINTR, continue select loop */
+static int _clicon_sig_child = 0;
+
 /* If set (eg by signal handler) ignore EINTR and continue select loop */
 static int _clicon_sig_ignore = 0;
 
@@ -121,6 +127,19 @@ clicon_exit_get(void)
     return _clicon_exit;
 }
 
+int
+clicon_sig_child_set(int val)
+{
+    _clicon_sig_child = val;
+    return 0;
+}
+
+int
+clicon_sig_child_get(void)
+{
+    return _clicon_sig_child;
+}
+
 int
 clicon_sig_ignore_set(int val)
 {
@@ -308,7 +327,7 @@ clixon_event_poll(int fd)
  * @retval -1  Error: eg select, callback, timer, 
  */
 int
-clixon_event_loop(void)
+clixon_event_loop(clicon_handle h)
 {
     struct event_data *e;
     struct event_data *e_next;
@@ -321,6 +340,12 @@ clixon_event_loop(void)
 
     while (!clicon_exit_get()){
 	FD_ZERO(&fdset);
+	if (clicon_sig_child_get()){
+	    /* Go through processes and wait for child processes */
+	    if (clixon_process_waitpid(h) < 0)
+		goto err;
+	    clicon_sig_child_set(0);
+	}
 	for (e=ee; e; e=e->e_next)
 	    if (e->e_type == EVENT_FD)
 		FD_SET(e->e_fd, &fdset);
@@ -333,7 +358,7 @@ clixon_event_loop(void)
 		n = select(FD_SETSIZE, &fdset, NULL, NULL, &t); 
 	}
 	else
-	    n = select(FD_SETSIZE, &fdset, NULL, NULL, NULL); 
+	    n = select(FD_SETSIZE, &fdset, NULL, NULL, NULL);
 	if (clicon_exit_get())
 	    break;
 	if (n == -1) {
@@ -342,7 +367,9 @@ clixon_event_loop(void)
 		 * (1) Signals that exit gracefully, the function returns 0
 		 *     Must be registered such as by set_signal() of SIGTERM,SIGINT, etc with a handler that calls
 		 *     clicon_exit_set().
-		 * (2) Signals that are ignored, and the select is rerun, eg SIGCHLD if handler calls clicon_sig_ignore()
+		 * (2) SIGCHILD Childs that exit(), go through clixon_proc list and cal waitpid
+		 *     New select loop is called
+		 * (2) Signals are ignored, and the select is rerun, ie handler calls clicon_sig_ignore_get
 		 *     New select loop is called
 		 * (3) Other signals result in an error and return -1.
 		 */
@@ -351,6 +378,13 @@ clixon_event_loop(void)
 		    clicon_err(OE_EVENTS, errno, "select");
 		    retval = 0;
 		}
+		else if (clicon_sig_child_get()){
+		    /* Go through processes and wait for child processes */
+		    if (clixon_process_waitpid(h) < 0)
+			goto err;
+		    clicon_sig_child_set(0);
+		    continue;
+		}
 		else if (clicon_sig_ignore_get()){
 		    clicon_sig_ignore_set(0);
 		    continue;

lib/src/clixon_proc.c

@@ -69,9 +69,9 @@
 #include "clixon_err.h"
 #include "clixon_log.h"
 #include "clixon_queue.h"
-#include "clixon_event.h"
 #include "clixon_hash.h"
 #include "clixon_handle.h"
+#include "clixon_event.h"
 #include "clixon_sig.h"
 #include "clixon_string.h"
 #include "clixon_queue.h"
@@ -83,13 +83,17 @@
 
 /* Process entry list */
 struct process_entry_t {
-    qelem_t    pe_qelem;   /* List header */
+    qelem_t        pe_qelem;     /* List header */
-    char      *pe_name;    /* Name of process used for internal use */
+    char          *pe_name;     /* Name of process used for internal use. Unique with exiting=0 */
-    char      *pe_netns;   /* Network namespace */
+    char          *pe_netns;    /* Network namespace */
-    char     **pe_argv;    /* argv with command as element 0 and NULL-terminated */
+    char         **pe_argv;     /* argv with command as element 0 and NULL-terminated */
-    pid_t      pe_pid;     /* Running process id (state) or 0 if dead */
+    int            pe_argc;     /* Length of argc */
-    proc_operation pe_op;  /* Operation pending? */
+    pid_t          pe_pid;      /* Running process id (state) or 0 if dead (pid is set if exiting=1) */
-    proc_cb_t *pe_callback; /* Wrapper function, may be called from process_operation  */
+    int            pe_exiting;  /* If set process is in the process of dying needs reaping */
+    int            pe_clone;    /* Duplicate when restarting, delete when reaped */
+    pid_t          pe_status;   /* Status on exit as defined in waitpid */
+    proc_operation pe_op;       /* Operation pending? */
+    proc_cb_t     *pe_callback; /* Wrapper function, may be called from process_operation  */
 };
 
 static void
@@ -209,7 +213,7 @@ clixon_proc_background(char       **argv,
     sigset_t      oset;
     struct rlimit rlim = {0, };
 
-    clicon_debug(1, "%s netns:%s", __FUNCTION__, netns);
+    clicon_debug(1, "%s", __FUNCTION__);
     if (argv == NULL){
 	clicon_err(OE_UNIX, EINVAL, "argv is NULL");
 	goto quit;
@@ -303,6 +307,58 @@ clixon_process_op_str2int(char *opstr)
     return clicon_str2int(proc_operation_map, opstr);
 }
 
+/*! Make a copy of process-entry struct */
+static int
+clixon_process_register_dup(process_entry_t  *pe0,
+			    process_entry_t **pnew)
+{
+    int              retval = -1;
+    process_entry_t *pe1 = NULL;
+    int              i;
+    
+    if (pe0 == NULL){
+	clicon_err(OE_DB, EINVAL, "pe0 is NULL");
+	goto done;
+    }
+    if (pnew == NULL){
+	clicon_err(OE_DB, EINVAL, "pnew is NULL");
+	goto done;
+    }
+    if ((pe1 = malloc(sizeof(process_entry_t))) == NULL) {
+	clicon_err(OE_DB, errno, "malloc");
+	goto done;
+    }
+    memset(pe1, 0, sizeof(*pe1));
+    memcpy(pe1, pe0, sizeof(process_entry_t)); /* Note lots of malloced memory that needs to be handled after this copy*/
+    pe1->pe_exiting = 0;
+    pe1->pe_clone = 0;
+    if ((pe1->pe_name = strdup(pe0->pe_name)) == NULL){
+	clicon_err(OE_DB, errno, "strdup name");
+	goto done;
+    }
+    if (pe0->pe_netns && (pe1->pe_netns = strdup(pe0->pe_netns)) == NULL){
+	clicon_err(OE_DB, errno, "strdup netns");
+	goto done;
+    }
+    if ((pe1->pe_argv = calloc(pe0->pe_argc, sizeof(char *))) == NULL){
+	clicon_err(OE_UNIX, errno, "calloc");
+	goto done;
+    }
+    for (i=0; i<pe0->pe_argc; i++){
+	if (pe0->pe_argv[i] != NULL &&
+	    (pe1->pe_argv[i] = strdup(pe0->pe_argv[i])) == NULL){
+	    clicon_err(OE_UNIX, errno, "strdup");
+	    goto done;
+	}
+    }
+    ADDQ(pe1, _proc_entry_list);
+    *pnew = pe1;
+    retval = 0;
+ done:
+    /* dealloc pe1 on error */
+    return retval;
+}
+
 /*! Register an internal process
  *
  * @param[in]  h        Clixon handle
@@ -348,6 +404,7 @@ clixon_process_register(clicon_handle h,
 	clicon_err(OE_DB, errno, "strdup netns");
 	goto done;
     }
+    pe->pe_argc = argc;
     if ((pe->pe_argv = calloc(argc, sizeof(char *))) == NULL){
 	clicon_err(OE_UNIX, errno, "calloc");
 	goto done;
@@ -356,6 +413,7 @@ clixon_process_register(clicon_handle h,
 	if (argv[i] != NULL &&
 	    (pe->pe_argv[i] = strdup(argv[i])) == NULL){
 	    clicon_err(OE_UNIX, errno, "strdup");
+	    goto done;
 	}
     }
     pe->pe_callback = callback;
@@ -365,32 +423,42 @@ clixon_process_register(clicon_handle h,
     return retval;
 }
 
+static int
+clixon_process_delete_only(process_entry_t *pe)
+{
+    char           **pa;
+
+    if (pe->pe_name)
+	free(pe->pe_name);
+    if (pe->pe_netns)
+	free(pe->pe_netns);
+    if (pe->pe_argv){
+	for (pa = pe->pe_argv; *pa != NULL; pa++){
+	    if (*pa)
+		free(*pa);
+	}
+	free(pe->pe_argv);
+    }
+    free(pe);
+    return 0;
+}
+    
 /*! Delete all Upgrade callbacks
  */
 int
 clixon_process_delete_all(clicon_handle h)
 {
     process_entry_t *pe;
-    char           **pa;
 
     while((pe = _proc_entry_list) != NULL) {
 	DELQ(pe, _proc_entry_list, process_entry_t *);
-	if (pe->pe_name)
+	clixon_process_delete_only(pe);
-	    free(pe->pe_name);
-	if (pe->pe_netns)
-	    free(pe->pe_netns);
-	if (pe->pe_argv){
-	    for (pa = pe->pe_argv; *pa != NULL; pa++){
-		if (*pa)
-		    free(*pa);
-	    }
-	    free(pe->pe_argv);
-	}
-	free(pe);
     }
     return 0;
 }
 
+/*!
+ */
 static int
 proc_op_run(pid_t pid0,
 	    int  *runp)
@@ -401,7 +469,7 @@ proc_op_run(pid_t pid0,
 
     run = 0;
     if ((pid = pid0) != 0){ /* if 0 stopped */
-	/* Check if lives */
+	/* Check if alive */
 	run = 1;
 	if ((kill(pid, 0)) < 0){
 	    if (errno == ESRCH){
@@ -420,49 +488,6 @@ proc_op_run(pid_t pid0,
     return retval;
 }
 
-/*! Perform process operation
- *
- * @param[in]  op      One of: start, stop, restart, status
- * @param[in]  netns   Network namespace
- * @param[in]  argv    NULL-terminated Argument vector
- * @param[out] pidp    Process-id
- */
-static int
-clixon_process_operation_one(const proc_operation op,
-			     const char          *netns,
-			     char               **argv,
-			     const char          *name,
-			     pid_t               *pidp)
-{
-    int retval = -1;
-    int run = 0;
-    
-    /* Check if running */
-    if (proc_op_run(*pidp, &run) < 0)
-	goto done;
-    if (op == PROC_OP_STOP || op == PROC_OP_RESTART){
-	if (run){
-	    clicon_log(LOG_NOTICE, "Killing old daemon %s with pid: %d", name, *pidp);
-	    kill(*pidp, SIGTERM);
-	}
-	*pidp = 0; /* mark as dead */
-	run = 0;
-    }
-    if (op == PROC_OP_START || op == PROC_OP_RESTART){
-	if (run == 1){
-	    ; /* Already runs */
-	}
-	else{
-	    if (clixon_proc_background(argv, netns, pidp) < 0)
-		goto done;
-	    clicon_debug(1, "%s started pid:%d", __FUNCTION__, *pidp);
-	}
-    }
-    retval = 0;
- done:
-     return retval;
-}
-
 /*! Find process operation entry given name and op and perform operation if found
  *
  * @param[in]  h       clicon handle
@@ -495,8 +520,11 @@ clixon_process_operation(clicon_handle  h,
 	    if (wrapit && pe->pe_callback != NULL)
 		if (pe->pe_callback(h, pe, &op) < 0)
 		    goto done;
+	    clicon_debug(1, "%s name: %s pid:%d op: %s", __FUNCTION__,
+			 name, pe->pe_pid, clicon_int2str(proc_operation_map, op));
 	    if (op == PROC_OP_START || op == PROC_OP_STOP || op == PROC_OP_RESTART){
 		pe->pe_op = op;
+		clicon_debug(1, "%s scheduling %s pid:%d", __FUNCTION__, name, pe->pe_pid);
 		sched++;
 	    }
 	    if (pid)
@@ -510,7 +538,7 @@ clixon_process_operation(clicon_handle  h,
  ok:
     retval = 0;
  done:
-    clicon_debug(2, "%s retval:%d", __FUNCTION__, retval);
+    clicon_debug(1, "%s retval:%d", __FUNCTION__, retval);
     return retval;
 }
 
@@ -558,6 +586,7 @@ clixon_process_start_all(clicon_handle h)
  * (1) at startup, if started before deamoninization, process will get as child of 1
  * (2) edit changes or rpc restart especially of restconf where you may saw of your arm and terminate
  *     return socket.
+ * A special complexity is restarting processes, where the old is killed, but state must be kept until it is reaped
  */
 static int
 clixon_process_sched(int           fd,
@@ -565,25 +594,68 @@ clixon_process_sched(int           fd,
 {
     int              retval = -1;
     process_entry_t *pe;
+    process_entry_t *pe1;
     proc_operation   op;
+    pid_t            newpid;
+    int              run;
 
-    clicon_debug(2, "%s",__FUNCTION__);
+    clicon_debug(1, "%s",__FUNCTION__);
     if (_proc_entry_list == NULL)
 	goto ok;
     pe = _proc_entry_list;
     do {
-	if ((op = pe->pe_op) != PROC_OP_NONE){
+	clicon_debug(1, "%s name: %s pid:%d op: %s", __FUNCTION__,
-	    if (clixon_process_operation_one(op, pe->pe_netns, pe->pe_argv, pe->pe_name, &pe->pe_pid) < 0)
+		     pe->pe_name, pe->pe_pid, clicon_int2str(proc_operation_map, pe->pe_op));
+	/* Execute pending operations and not already exiting */
+	if ((op = pe->pe_op) != PROC_OP_NONE &&
+	    pe->pe_exiting == 0){
+	    /* Check if running */
+	    run = 0;
+	    clicon_debug(1, "%s run: %d", __FUNCTION__, run);
+	    if (proc_op_run(pe->pe_pid, &run) < 0)
 		goto done;
-	    clicon_debug(1, "%s op:%s pid:%d", __FUNCTION__, clicon_int2str(proc_operation_map, op), pe->pe_pid);
+	    switch (op){
-	    pe->pe_op = PROC_OP_NONE;
+	    case PROC_OP_STOP:
+		clicon_debug(1, "%s stop pid:%d", __FUNCTION__, pe->pe_pid);
+	    case PROC_OP_RESTART:
+		if (run){
+		    clicon_log(LOG_NOTICE, "Killing old process %s with pid: %d", pe->pe_name, pe->pe_pid);
+		    kill(pe->pe_pid, SIGTERM);
+		    /* Cant wait here because it would block the backend and terminating may involve
+		     * some protocol handling, instead SIGCHLD is receoved and 
+		     * clixon_process_waitpid is called that for waits/reaps the dead process */
+		    pe->pe_exiting = 1;
+		}
+		if (op == PROC_OP_STOP)
+		    break;
+		if (clixon_proc_background(pe->pe_argv, pe->pe_netns, &newpid) < 0)
+		    goto done;
+		clicon_debug(1, "%s restart pid:%d -> %d", __FUNCTION__, pe->pe_pid, newpid);
+		/* Create a new pe */
+		if (clixon_process_register_dup(pe, &pe1) < 0)
+		    goto done;
+		pe->pe_clone = 1; /* Delete when reaped */
+		pe1->pe_op = PROC_OP_NONE; /* Dont restart again */
+		pe1->pe_pid = newpid;
+		break;
+	    case PROC_OP_START:
+		if (run) /* Already runs */
+		    break;
+		if (clixon_proc_background(pe->pe_argv, pe->pe_netns, &pe->pe_pid) < 0)
+		    goto done;
+		clicon_debug(1, "%s started pid:%d", __FUNCTION__, pe->pe_pid);
+		break;
+	    default:
+		break;
+	    }
 	}
+	pe->pe_op = PROC_OP_NONE;
 	pe = NEXTQ(process_entry_t *, pe);
     } while (pe != _proc_entry_list);
  ok:
     retval = 0;
  done:
-    clicon_debug(2, "%s retval:%d", __FUNCTION__, retval);
+    clicon_debug(1, "%s retval:%d", __FUNCTION__, retval);
     return retval;
 }
 
@@ -602,6 +674,7 @@ clixon_process_sched_register(clicon_handle h)
     struct timeval t;
     struct timeval t1 = {0, 1500}; /* See discussion ^*/
 
+    clicon_debug(2, "%s", __FUNCTION__);
     gettimeofday(&t, NULL);
     timeradd(&t, &t1, &t);
     if (clixon_event_reg_timeout(t, clixon_process_sched, h, "process") < 0)
@@ -611,3 +684,44 @@ clixon_process_sched_register(clicon_handle h)
     clicon_debug(2, "%s retval:%d", __FUNCTION__, retval);
     return retval;
 }
+
+/*! Go through processes and wait for child processes
+ * Typically we know a child has been killed by SIGCHLD, but we do not know which process it is
+ * Traverse all known processes and reap them, eg call waitpid() to avoid zombies.
+ * @param[in]  h  Clixon handle
+ */
+int
+clixon_process_waitpid(clicon_handle h)
+{
+    int              retval = -1;
+    process_entry_t *pe;
+    int              status = 0;
+    pid_t            wpid;
+
+    clicon_debug(1, "%s", __FUNCTION__);
+    pe = _proc_entry_list;
+    do {
+	if (pe->pe_pid != 0){
+	    clicon_debug(1, "%s waitpid(%d)", __FUNCTION__, pe->pe_pid);
+	    if ((wpid = waitpid(pe->pe_pid, &status, WNOHANG)) == pe->pe_pid){
+		clicon_debug(1, "%s waitpid(%d) waited", __FUNCTION__, pe->pe_pid);
+		pe->pe_exiting = 0;
+		pe->pe_pid = 0;       /* mark as dead */
+		pe->pe_status = status;   
+		if (pe->pe_clone){
+		    /* Delete it */
+		    DELQ(pe, _proc_entry_list, process_entry_t *);
+		    clixon_process_delete_only(pe);
+		}
+		break; /* pid is unique */
+	    }
+	    else
+		clicon_debug(1, "%s waitpid(%d) nomatch:%d", __FUNCTION__, pe->pe_pid, wpid);
+	}
+    	pe = NEXTQ(process_entry_t *, pe);
+    } while (pe != _proc_entry_list);
+    retval = 0;
+    // done:
+    clicon_debug(1, "%s retval:%d", __FUNCTION__, retval);
+    return retval;
+}

lib/src/clixon_proto.c

@@ -67,9 +67,9 @@
 /* clicon */
 #include "clixon_err.h"
 #include "clixon_queue.h"
-#include "clixon_event.h"
 #include "clixon_hash.h"
 #include "clixon_handle.h"
+#include "clixon_event.h"
 #include "clixon_log.h"
 #include "clixon_yang.h"
 #include "clixon_sig.h"

lib/src/clixon_stream.c

@@ -73,10 +73,10 @@
 #include "clixon_queue.h"
 #include "clixon_err.h"
 #include "clixon_log.h"
-#include "clixon_event.h"
 #include "clixon_string.h"
 #include "clixon_hash.h"
 #include "clixon_handle.h"
+#include "clixon_event.h"
 #include "clixon_yang.h"
 #include "clixon_xml.h"
 #include "clixon_xml_io.h"

test/test_restconf_rpc.sh

@@ -7,6 +7,7 @@
 # - on enable change, make the state as configured
 # - No restconf config means enable: false (extra rule)
 # See test_restconf_netns for network namespaces
+# XXX Lots of sleeps to remove race conditions. I am sure there are others way to fix this
 
 # Magic line must be first in script (see README.md)
 s="$_" ; . ./lib.sh || if [ "$s" = $0 ]; then exit 0; else return 0; fi
@@ -62,6 +63,7 @@ function testrpc()
     operation=$1
     expectret=$2
     
+    sleep $DEMSLEEP
     new "send rpc $operation"
     ret=$($clixon_netconf -qf $cfg<<EOF
 $DEFAULTHELLO
@@ -106,7 +108,7 @@ EOF
 	    fi
 	fi
     fi
-
+    sleep $DEMSLEEP
     echo "$pid" # cant use return that only uses 0-255
 }
 
@@ -165,7 +167,7 @@ wait_restconf
 new "try restconf rpc status"
 expectpart "$(curl $CURLOPTS -X POST -H "Content-Type: application/yang-data+json" $RCPROTO://localhost/restconf/operations/clixon-lib:process-control -d '{"clixon-lib:input":{"name":"restconf","operation":"status"}}')" 0 "HTTP/1.1 200 OK" '{"clixon-lib:output":{"pid":'
 
-new "1.1. Get status"
+new "2. Get status"
 pid1=$(testrpc status 1)
 if [ $? -ne 0 ]; then echo "$pid1";exit -1; fi
 
@@ -177,7 +179,7 @@ fi
 new "try restconf rpc restart"
 expectpart "$(curl $CURLOPTS -X POST -H "Content-Type: application/yang-data+json" $RCPROTO://localhost/restconf/operations/clixon-lib:process-control -d '{"clixon-lib:input":{"name":"restconf","operation":"restart"}}')" 0 "HTTP/1.1 200 OK" '{"clixon-lib:output":{"pid":'
 
-new "1.1. Get status"
+new "3. Get status"
 pid1=$(testrpc status 1)
 if [ $? -ne 0 ]; then echo "$pid1";exit -1; fi
 
@@ -186,19 +188,19 @@ if [ "$pid0" -eq "$pid1" ]; then
     err "not $pid0"
 fi
 
-new "2. stop restconf RPC"
+new "4. stop restconf RPC"
 testrpc stop 0
 if [ $? -ne 0 ]; then exit -1; fi
 
-new "3. Get rpc status stopped"
+new "5. Get rpc status stopped"
 pid=$(testrpc status 0)
 if [ $? -ne 0 ]; then echo "$pid";exit -1; fi
 
-new "4. Start rpc again"
+new "6. Start rpc again"
 testrpc start 0
 if [ $? -ne 0 ]; then exit -1; fi
 
-new "4.1. Get rpc status"
+new "7. Get rpc status"
 pid3=$(testrpc status 1)
 if [ $? -ne 0 ]; then echo "$pid3";exit -1; fi
 
@@ -215,19 +217,19 @@ fi
 new "kill restconf"
 stop_restconf_pre
 
-new "5. start restconf RPC"
+new "8. start restconf RPC"
 testrpc start 0
 if [ $? -ne 0 ]; then exit -1; fi
 
-new "6. check status RPC on"
+new "9. check status RPC on"
 pid5=$(testrpc status 1) # Save pid5
 if [ $? -ne 0 ]; then echo "$pid5";exit -1; fi
 
-new "7. restart restconf RPC"
+new "10. restart restconf RPC"
 testrpc restart 0
 if [ $? -ne 0 ]; then exit -1; fi
 
-new "8. Get restconf status rpc"
+new "11. Get restconf status rpc"
 pid7=$(testrpc status 1) # Save pid7
 if [ $? -ne 0 ]; then echo "$pid7";exit -1; fi
 
@@ -264,7 +266,7 @@ if [ $BE -ne 0 ]; then
     wait_backend
 fi
 
-new "9. Get restconf (running) after restart"
+new "12. Get restconf (running) after restart"
 pid=$(testrpc status 1)
 if [ $? -ne 0 ]; then echo "$pid"; exit -1; fi
 
@@ -309,15 +311,15 @@ if [ $BE -ne 0 ]; then
     wait_backend
 fi
 
-new "10. check status RPC off"
+new "13. check status RPC off"
 pid=$(testrpc status 0)
 if [ $? -ne 0 ]; then echo "$pid";exit -1; fi
 
-new "11. start restconf RPC"
+new "14. start restconf RPC"
 testrpc start 0
 if [ $? -ne 0 ]; then exit -1; fi
 
-new "12. check status RPC off"
+new "15. check status RPC off"
 pid=$(testrpc status 0)
 if [ $? -ne 0 ]; then echo "$pid";exit -1; fi
 
@@ -327,7 +329,9 @@ expecteof "$clixon_netconf -qf $cfg" 0 "$DEFAULTHELLO<rpc $DEFAULTNS><edit-confi
 new "netconf commit"
 expecteof "$clixon_netconf -qf $cfg" 0 "$DEFAULTHELLO<rpc $DEFAULTNS><commit/></rpc>]]>]]>" "^<rpc-reply $DEFAULTNS><ok/></rpc-reply>]]>]]>$"
 
-new "13. check status RPC on"
+sleep $DEMSLEEP
+
+new "16. check status RPC on"
 pid=$(testrpc status 1)
 if [ $? -ne 0 ]; then echo "$pid";exit -1; fi
 
@@ -343,9 +347,9 @@ if [ $pid -eq $pid1 ]; then
     err "A different pid" "Same pid: $pid"
 fi
 
-new "Edit a non-restconf field via restconf"
+sleep $DEMSLEEP
-echo "curl $CURLOPTS -X POST -H \"Content-Type: application/yang-data+json\" $RCPROTO://localhost/restconf/data -d '{\"example:val\":\"xyz\"}'"
 
+new "Edit a non-restconf field via restconf"
 expectpart "$(curl $CURLOPTS -X POST -H "Content-Type: application/yang-data+json" $RCPROTO://localhost/restconf/data -d '{"example:val":"xyz"}' )" 0 "HTTP/1.1 201 Created"
 
 new "check status RPC same pid"
@@ -361,7 +365,7 @@ expecteof "$clixon_netconf -qf $cfg" 0 "$DEFAULTHELLO<rpc $DEFAULTNS><edit-confi
 new "netconf commit"
 expecteof "$clixon_netconf -qf $cfg" 0 "$DEFAULTHELLO<rpc $DEFAULTNS><commit/></rpc>]]>]]>" "^<rpc-reply $DEFAULTNS><ok/></rpc-reply>]]>]]>$"
 
-new "14. check status RPC off"
+new "17. check status RPC off"
 pid=$(testrpc status 0)
 if [ $? -ne 0 ]; then echo "$pid";exit -1; fi
 

util/clixon_util_grpc.c

@@ -541,7 +541,7 @@ main(int    argc,
     }
     if (clixon_event_reg_fd(ss, ssl_input_cb, sd, "ssl socket") < 0)
 	goto done;
-    if (clixon_event_loop() < 0)
+    if (clixon_event_loop(h) < 0)
 	goto done;
     retval = 0;
  done:

util/clixon_util_ssl.c

@@ -539,7 +539,7 @@ main(int    argc,
     }
     if (clixon_event_reg_fd(ss, ssl_input_cb, sd, "ssl socket") < 0)
 	goto done;
-    if (clixon_event_loop() < 0)
+    if (clixon_event_loop(h) < 0)
 	goto done;
     retval = 0;
  done: