[DEBUG-ONLY] Enhance jb->blocked assert in jnl_write.c to account for…

… defunct processes

If the process id corresponding to jb->blocked was abnormally killed (e.g. kill -9), it is possible
the process is still alive but in a defunct state (when another process comes into jnl_write.c
after having salvaged crit etc.) if the parent process has not yet done a waitpid() on this abnormally
terminated child. In that case, the following assert would fail.

    266     assert((!jb->blocked) || (FALSE == is_proc_alive(jb->blocked, 0)));

This is now addressed by using the new is_defunct_pid() function in the assert.

sr_port/jnl_write.c

@@ -50,6 +50,7 @@
 #include "buddy_list.h"		/* needed for tp.h */
 #include "hashtab_int4.h"	/* needed for tp.h */
 #include "tp.h"
+#include "is_defunct_pid.h"
 #endif
 
 GBLREF	uint4			process_id;
@@ -122,6 +123,7 @@ void	jnl_write(jnl_private_control *jpc, enum jnl_record_type rectype, jnl_recor
 	char			*mumps_node_ptr;
 	struct_jrec_align	*align_rec;
 	uint4			end_freeaddr;
+	int4			blocking_pid;
 #	endif
 
 	assert(MAX_JNL_WRITE_RECURSION_DEPTH > jnl_write_recursion_depth++);
@@ -262,8 +264,11 @@ void	jnl_write(jnl_private_control *jpc, enum jnl_record_type rectype, jnl_recor
 	ADJUST_CHECKSUM(checksum, csd->jnl_checksum, checksum);
 	SET_JREC_CHECKSUM(jnl_rec, rectype, checksum);
 	if (!in_phase2)
-	{
-		assert((!jb->blocked) || (FALSE == is_proc_alive(jb->blocked, 0)));
+	{	/* Check that the pid in jb->blocked is dead if it is not 0. The check handles the case that a pid could be dead
+		 * but its parent could still have not gotten its exit status in which case it could be a defunct.
+		 */
+		DEBUG_ONLY(blocking_pid = jb->blocked);
+		assert(!blocking_pid || is_defunct_pid((uint4)blocking_pid) || (FALSE == is_proc_alive(blocking_pid, 0)));
 		jb->blocked = process_id;
 	}
 	jnl_fs_block_size = jb->fs_block_size;

sr_unix/is_defunct_pid.c

@@ -0,0 +1,63 @@
+/****************************************************************
+ *								*
+ * Copyright (c) 2019 YottaDB LLC and/or its subsidiaries.	*
+ * All rights reserved.						*
+ *								*
+ *	This source code contains the intellectual property	*
+ *	of its copyright holder(s), and is made available	*
+ *	under a license.  If you do not know the terms of	*
+ *	the license, please stop and do not read further.	*
+ *								*
+ ****************************************************************/
+
+#include "mdef.h"
+
+#include "gtm_stdio.h"
+
+#include "is_defunct_pid.h"
+
+#ifdef DEBUG
+GBLREF	uint4			process_id;
+#endif
+
+/* ----------------------------------------------
+ * Check if input pid is a defunct process (zombie)
+ *
+ * Arguments:
+ *	pid	- process ID
+ *
+ * Return:
+ *	TRUE	- If process is a defunct process (i.e. shows up as <defunct> in ps -ef listing)
+ *	FALSE	- Otherwise
+ * ----------------------------------------------
+ */
+
+boolean_t	is_defunct_pid(int4 pid)
+{
+	int		status;
+	char		procfilename[64];
+	boolean_t	is_defunct;
+	FILE		*fp;
+	char		pidstate;
+
+	assert(0 != pid);
+	assert(process_id != pid);
+	is_defunct = FALSE;	/* by default it is not a defunct process */
+#	ifdef __linux__
+	/* open the /proc/<pid>/stat file */
+	SNPRINTF(procfilename, sizeof(procfilename), "/proc/%d/stat", (int)pid);
+	fp = fopen(procfilename, "r");
+	if (NULL != fp)
+	{
+		fscanf(fp, "%*d %*s %c", &pidstate);
+		is_defunct = ('Z' == pidstate);
+		fclose(fp);
+	}
+#	else
+	/* This is likely MacOS or Cygwin. Those ports need to implement this functionality using other methods
+	 * (maybe do a "system" call with a "ps -ef | grep defunct" if nothing else is possible).
+	 */
+#	error unsupported platform
+#	endif
+	return is_defunct;
+}

sr_unix/is_defunct_pid.h

@@ -0,0 +1,18 @@
+/****************************************************************
+ *								*
+ * Copyright (c) 2019 YottaDB LLC and/or its subsidiaries.	*
+ * All rights reserved.						*
+ *								*
+ *	This source code contains the intellectual property	*
+ *	of its copyright holder(s), and is made available	*
+ *	under a license.  If you do not know the terms of	*
+ *	the license, please stop and do not read further.	*
+ *								*
+ ****************************************************************/
+
+#ifndef IS_DEFUNCT_PID_INCLUDED
+#define IS_DEFUNCT_PID_INCLUDED
+
+boolean_t	is_defunct_pid(int4 pid);
+
+#endif /* IS_DEFUNCT_PID_INCLUDED */