[PATCH 2/2] nfs-utils: add initial tests for statd that run via "make check"

[PATCH 2/2] nfs-utils: add initial tests for statd that run via "make check"

[Jeff Layton]

Leverage the support that automake already has for running tests via
make check. Add a simple test that just checks that the statd mon and
unmon calls actually work.

Adding more tests should be a simple matter of adding new scripts
exit 0 on success and non-zero on fail, and adding those to the
Makefile.am.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
---
 tests/Makefile.am                    |    2 +
 tests/t0001-statd-basic-mon-unmon.sh |   56 ++++++++++++++++++++++++++++++++++
 tests/test-lib.sh                    |   41 +++++++++++++++++++++++++
 3 files changed, 99 insertions(+), 0 deletions(-)
 create mode 100755 tests/t0001-statd-basic-mon-unmon.sh
 create mode 100755 tests/test-lib.sh

diff --git a/tests/Makefile.am b/tests/Makefile.am
index 997ac51..634fb7b 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -3,3 +3,5 @@
 SUBDIRS = statdtest
 
 MAINTAINERCLEANFILES = Makefile.in
+
+TESTS = t0001-statd-basic-mon-unmon.sh
diff --git a/tests/t0001-statd-basic-mon-unmon.sh b/tests/t0001-statd-basic-mon-unmon.sh
new file mode 100755
index 0000000..fe7439e
--- /dev/null
+++ b/tests/t0001-statd-basic-mon-unmon.sh
@@ -0,0 +1,56 @@
+#!/bin/bash
+#
+# statd_basic_mon_unmon -- test basic mon/unmon functionality with statd
+#
+# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License
+# as published by the Free Software Foundation; either version 2
+# of the License, or (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation, Inc.,
+# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+#
+
+. ./test-lib.sh
+
+start_statd
+if [ $? -ne 0 ]; then
+	echo "FAIL: problem starting statd"
+	exit 1
+fi
+
+COOKIE=`echo $$ | md5sum | cut -d' ' -f1`
+MON_NAME=`hostname`
+
+statdtest mon $MON_NAME $COOKIE
+if [ $? -ne 0 ]; then
+	echo "FAIL: mon failed"
+	kill_statd
+	exit 1
+fi
+
+grep -q $COOKIE /var/lib/nfs/statd/sm/$MON_NAME 
+if [ $? -ne 0 ]; then
+	echo "FAIL: $COOKIE not present in monitor file"
+	kill_statd
+	exit 1
+fi
+
+statdtest unmon $MON_NAME
+if [ $? -ne 0 ]; then
+	echo "FAIL: unmon failed"
+	kill_statd
+	exit 1
+fi
+
+kill_statd
+
+echo "PASS"
diff --git a/tests/test-lib.sh b/tests/test-lib.sh
new file mode 100755
index 0000000..053e46d
--- /dev/null
+++ b/tests/test-lib.sh
@@ -0,0 +1,41 @@
+#!/bin/bash
+#
+# test-lib.sh -- library of functions for nfs-utils tests
+#
+# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License
+# as published by the Free Software Foundation; either version 2
+# of the License, or (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation, Inc.,
+# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+#
+
+export PATH=$PATH:$srcdir/statdtest
+
+lockd_registered() {
+	rpcinfo -p | grep -q nlockmgr
+	return $?
+}
+
+start_statd() {
+	rpcinfo -u 127.0.0.1 status 1 &> /dev/null
+	if [ $? -eq 0 ]; then
+		echo "***ERROR***: statd is already running and should "
+		echo "             be down when starting this test"
+		return 1
+	fi
+	$srcdir/../utils/statd/statd --no-notify
+}
+
+kill_statd() {
+	kill `cat /var/run/rpc.statd.pid`
+}
-- 
1.6.2.5

[PATCH 0/2] nfs-utils: add testing infrastructure to nfs-utils

[Jeff Layton]

This set is the same as the one I sent a month or so ago, with some
small changes to account for header file changes in Chuck's tree. This
patchset depends on the patches that add libnsm.a to nfs-utils.

This patchset is intended as a starting point for an automated test
suite for nfs-utils. The idea here is to start simply and add a suite of
tests that we can run via "make check" -- the standard automake method
for running tests.

Clearly there are limits to what we can test without a multi-host test
harness. My hope is that this should help keep us from breaking basic
functionality by allowing us to test it in a very simple fashion. At
some point in the future we should also consider how to best handle
multi-machine testing, but I see that as complimenting this code rather
than replacing it.

For this set, the focus is on testing statd, which is particularly
susceptible to subtle breakage. Problems with it are often not noticed
until lock recovery breaks, and that may greatly lag the actual
breakage.

To faciitate statd testing, I've added a "statdtest" program that can
serve as a synthetic statd client and an NLM simulator. It's very
loosely based on the old statd simulator code. That program is dependent
on some of Chuck Lever's recent statd patches -- notably the ones that
break out common NSM code into libnsm.a.

For this initial drop, I'm just adding a single test that tests mon and
unmon functionality with statd. Adding more tests should fairly simple
to do.

Jeff Layton (2):
  nfs-utils: introduce new statd testing simulator
  nfs-utils: add initial tests for statd that run via "make check"

 Makefile.am                          |    2 +-
 configure.ac                         |    4 +-
 tests/Makefile.am                    |    7 +
 tests/statdtest/Makefile.am          |   45 ++++
 tests/statdtest/README               |   12 +
 tests/statdtest/nlm_sm_inter.x       |   43 +++
 tests/statdtest/statdtest.c          |  464 ++++++++++++++++++++++++++++++++++
 tests/t0001-statd-basic-mon-unmon.sh |   55 ++++
 tests/test-lib.sh                    |   41 +++
 9 files changed, 671 insertions(+), 2 deletions(-)
 create mode 100644 tests/Makefile.am
 create mode 100644 tests/statdtest/Makefile.am
 create mode 100644 tests/statdtest/README
 create mode 100644 tests/statdtest/nlm_sm_inter.x
 create mode 100644 tests/statdtest/statdtest.c
 create mode 100755 tests/t0001-statd-basic-mon-unmon.sh
 create mode 100755 tests/test-lib.sh

[PATCH 1/2] nfs-utils: introduce new statd testing simulator

[Jeff Layton]

rpc.statd is often prone to subtle, difficult to detect breakage. When
it has problems, they're often invisible and only manifest themselves
as failed lock recovery.

This program is intended to function as part of a test harness for
statd. It's a multicall binary that serves as a synthetic NSM client
program, and a daemon that can simulate lockd for purposes of testing
the NSM to NLM downcall.

A new top level "tests/" directory is also added to nfs-utils to start
as a repository for automated tests of nfs-utils components.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
---
 Makefile.am                    |    2 +-
 configure.ac                   |    4 +-
 tests/Makefile.am              |    5 +
 tests/statdtest/Makefile.am    |   45 ++++
 tests/statdtest/README         |   12 +
 tests/statdtest/nlm_sm_inter.x |   43 ++++
 tests/statdtest/statdtest.c    |  464 ++++++++++++++++++++++++++++++++++++++++
 7 files changed, 573 insertions(+), 2 deletions(-)
 create mode 100644 tests/Makefile.am
 create mode 100644 tests/statdtest/Makefile.am
 create mode 100644 tests/statdtest/README
 create mode 100644 tests/statdtest/nlm_sm_inter.x
 create mode 100644 tests/statdtest/statdtest.c

diff --git a/Makefile.am b/Makefile.am
index b3a6e91..ae7cd16 100644
--- a/Makefile.am
+++ b/Makefile.am
@@ -2,7 +2,7 @@
 
 AUTOMAKE_OPTIONS = foreign
 
-SUBDIRS = tools support utils linux-nfs
+SUBDIRS = tools support utils linux-nfs tests
 
 MAINTAINERCLEANFILES = Makefile.in
 
diff --git a/configure.ac b/configure.ac
index 7dffeaf..e7298c6 100644
--- a/configure.ac
+++ b/configure.ac
@@ -420,6 +420,8 @@ AC_CONFIG_FILES([
 	utils/nfsd/Makefile
 	utils/nfsstat/Makefile
 	utils/showmount/Makefile
-	utils/statd/Makefile])
+	utils/statd/Makefile
+	tests/Makefile
+	tests/statdtest/Makefile])
 AC_OUTPUT
 
diff --git a/tests/Makefile.am b/tests/Makefile.am
new file mode 100644
index 0000000..997ac51
--- /dev/null
+++ b/tests/Makefile.am
@@ -0,0 +1,5 @@
+## Process this file with automake to produce Makefile.in
+
+SUBDIRS = statdtest
+
+MAINTAINERCLEANFILES = Makefile.in
diff --git a/tests/statdtest/Makefile.am b/tests/statdtest/Makefile.am
new file mode 100644
index 0000000..f7fc48f
--- /dev/null
+++ b/tests/statdtest/Makefile.am
@@ -0,0 +1,45 @@
+## Process this file with automake to produce Makefile.in
+
+GENFILES_CLNT	= nlm_sm_inter_clnt.c
+GENFILES_SVC	= nlm_sm_inter_svc.c
+GENFILES_XDR	= nlm_sm_inter_xdr.c
+GENFILES_H	= nlm_sm_inter.h
+
+GENFILES	= $(GENFILES_CLNT) $(GENFILES_SVC) $(GENFILES_XDR) $(GENFILES_H)
+
+
+check_PROGRAMS	= statdtest
+statdtest_SOURCES = $(GENFILES) statdtest.c
+
+BUILT_SOURCES = $(GENFILES)
+statdtest_LDADD = ../../support/nfs/libnfs.a \
+		       ../../support/nsm/libnsm.a
+
+if CONFIG_RPCGEN
+RPCGEN	= $(top_builddir)/tools/rpcgen/rpcgen
+$(RPCGEN):
+	make -C ../../tools/rpcgen all
+else
+RPCGEN = @RPCGEN_PATH@
+endif
+
+$(GENFILES_CLNT): %_clnt.c: %.x $(RPCGEN)
+	test -f $@ && rm -rf $@ || true
+	$(RPCGEN) -l -o $@ $<
+
+$(GENFILES_SVC): %_svc.c: %.x $(RPCGEN)
+	test -f $@ && rm -rf $@ || true
+	$(RPCGEN) -m -o $@ $<
+
+$(GENFILES_XDR): %_xdr.c: %.x $(RPCGEN)
+	test -f $@ && rm -rf $@ || true
+	$(RPCGEN) -c -o $@ $<
+
+$(GENFILES_H): %.h: %.x $(RPCGEN)
+	test -f $@ && rm -rf $@ || true
+	$(RPCGEN) -h -o $@ $<
+
+MAINTAINERCLEANFILES = Makefile.in
+
+CLEANFILES = $(GENFILES)
+
diff --git a/tests/statdtest/README b/tests/statdtest/README
new file mode 100644
index 0000000..034fbba
--- /dev/null
+++ b/tests/statdtest/README
@@ -0,0 +1,12 @@
+The statdtest program is a multicall binary intended for testing statd.
+It has the ability to act as a synthetic NSM client for sending
+artificial NSM calls to any host you choose.
+
+It also has an NLM simulator that implements the call that statd uses to
+communicate with lockd. The daemon simulator will start itself up,
+register as an NLM service and listen for upcalls from statd. When it
+gets one, it will log a message.
+
+Note that lockd will need to be down when using the daemon simulator, it
+also does not implement the entire NLM protocol and is only really
+useful for testing statd.
diff --git a/tests/statdtest/nlm_sm_inter.x b/tests/statdtest/nlm_sm_inter.x
new file mode 100644
index 0000000..d62e09c
--- /dev/null
+++ b/tests/statdtest/nlm_sm_inter.x
@@ -0,0 +1,43 @@
+/*
+ * Copyright (C) 1995, 1997-1999 Jeffrey A. Uphoff
+ * Modified by Olaf Kirch, 1996.
+ * Modified by H.J. Lu, 1998.
+ * Modified by Jeff Layton, 2009.
+ *
+ * NLM similator for Linux
+ */
+
+#ifdef RPC_CLNT
+%#include <string.h>
+#endif
+
+/*
+ * statd rejects monitor registrations for any non-lockd services, so pretend
+ * to be lockd when testing. Furthermore, the only call we care about from
+ * statd is #16, which is the downcall to notify the kernel of a host's status
+ * change.
+ */
+program NLM_SM_PROG { 
+	/* version 3 of the NLM protocol */
+	version NLM_SM_VERS3 {
+		void	 NLM_SM_NOTIFY(struct nlm_sm_notify) = 16;
+	} = 3;
+
+	/* version 2 of NLM protocol */
+	version NLM_SM_VERS4 {
+		void	 NLM_SM_NOTIFY(struct nlm_sm_notify) = 16;
+	} = 4;
+} = 100021;
+
+const  SM_MAXSTRLEN = 1024;
+const  SM_PRIV_SIZE = 16;
+
+/*
+ * structure of the status message sent back by the status monitor
+ * when monitor site status changes
+ */
+struct nlm_sm_notify {
+	string mon_name<SM_MAXSTRLEN>;
+	int state;
+	opaque priv[SM_PRIV_SIZE]; /* stored private information */
+};
diff --git a/tests/statdtest/statdtest.c b/tests/statdtest/statdtest.c
new file mode 100644
index 0000000..ba92d0a
--- /dev/null
+++ b/tests/statdtest/statdtest.c
@@ -0,0 +1,464 @@
+/*
+ * statdtest.c -- synthetic client and lockd simulator for testing statd
+ *
+ * Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor,
+ * Boston, MA 02110-1301, USA.
+ *
+ * Very loosely based on "simulator.c" in the statd directory. Original
+ * copyright for that program follows:
+ *
+ * Copyright (C) 1995-1997, 1999 Jeffrey A. Uphoff
+ */
+
+#ifdef HAVE_CONFIG_H
+#include "config.h"
+#endif
+
+#include <errno.h>
+#include <getopt.h>
+#include <netdb.h>
+#include <signal.h>
+#include <string.h>
+#include <rpc/rpc.h>
+#include <rpc/pmap_clnt.h>
+#include <rpcmisc.h>
+#include <sys/socket.h>
+#include <sys/types.h>
+#include <unistd.h>
+
+#include "nfslib.h"
+#include "nfsrpc.h"
+#include "sm_inter.h"
+#include "nlm_sm_inter.h"
+#include "sockaddr.h"
+#include "xcommon.h"
+
+static void daemon_simulator(void);
+static void sim_killer(int sig);
+static int statdtest_crash(char *);
+static int statdtest_mon(char *, char *, char *, char *, int, int);
+static int statdtest_stat(char *, char *);
+static int statdtest_notify(char *, char *, char *);
+static int statdtest_unmon(char *, char *, char *, int, int);
+static int statdtest_unmon_all(char *, char *, int, int);
+
+extern void nlm_sm_prog_4(struct svc_req *rqstp, register SVCXPRT *transp);
+extern void svc_exit(void);
+
+/*
+ * default to 15 retransmit interval, which seems to be the default for
+ * UDP clients w/ legacy glibc RPC
+ */
+static struct timeval retrans_interval =
+{
+	.tv_sec = 15,
+};
+
+static struct option longopts[] =
+{
+	{ "help", 0, 0, 'h' },
+	{ "host", 0, 0, 'H' },
+	{ "name", 1, 0, 'n' },
+	{ "program", 1, 0, 'P' },
+	{ "version", 1, 0, 'v' },
+	{ NULL, 0, 0, 0 },
+};
+
+static int
+usage(char *program)
+{
+	printf("Usage:\n");
+	printf("%s [options] <command> [arg]...\n", program);
+	printf("where command is one of these with the specified args:\n");
+	printf("crash\t\t\t\ttell host to simulate crash\n");
+	printf("daemon\t\t\t\t\tstart up lockd daemon simulator\n");
+	printf("notify <mon_name> <state>\tsend a reboot notification to host\n");
+	printf("stat <mon_name>\t\t\tget status of <mon_name> on host\n");
+	printf("unmon_all\t\t\ttell host to unmon everything\n");
+	printf("unmon <mon_name>\t\t\ttell host to unmon <mon_name>\n");
+	printf("mon <mon_name> <cookie>\t\ttell host to monitor <mon_name> with private <cookie>\n");
+	return 1;
+}
+
+static int
+hex2bin(char *dst, size_t dstlen, char *src)
+{
+	int i;
+	unsigned int tmp;
+
+	for (i = 0; *src && i < dstlen; i++) {
+		if (sscanf(src, "%2x", &tmp) != 1)
+			return 0;
+		dst[i] = tmp;
+		src++;
+		if (!*src)
+			break;
+		src ++;
+	}
+
+	return 1;
+}
+
+static void
+bin2hex(char *dst, char *src, size_t srclen)
+{
+	int i;
+
+	for (i = 0; i < srclen; i++)
+		dst += sprintf(dst, "%02x", 0xff & src[i]);
+}
+
+int
+main(int argc, char **argv)
+{
+	int arg, err = 0;
+	int remaining_args;
+	char my_name[NI_MAXHOST], host[NI_MAXHOST];
+	char cookie[SM_PRIV_SIZE];
+	int my_prog = NLM_SM_PROG;
+	int my_vers = NLM_SM_VERS4;
+
+	my_name[0] = '\0';
+	host[0] = '\0';
+
+	while ((arg = getopt_long(argc, argv, "hHn:P:v:", longopts,
+				  NULL)) != EOF ) {
+		switch (arg) {
+		case 'H':
+			strncpy(host, optarg, sizeof(host));
+		case 'n':
+			strncpy(my_name, optarg, sizeof(my_name));
+		case 'P':
+			my_prog = atoi(optarg);
+		case 'v':
+			my_vers = atoi(optarg);
+		}
+	}
+
+	remaining_args = argc - optind;
+	if (remaining_args <= 0)
+		usage(argv[0]);
+
+	if (!my_name[0])
+		gethostname(my_name, sizeof(my_name));
+	if (!host[0])
+		strncpy(host, "127.0.0.1", sizeof(host));
+
+	if (!strcasecmp(argv[optind], "daemon")) {
+		daemon_simulator();
+	} else if (!strcasecmp(argv[optind], "crash")) {
+		err = statdtest_crash(host);
+	} else if (!strcasecmp(argv[optind], "stat")) {
+		if (remaining_args < 2)
+			usage(argv[0]);
+		err = statdtest_stat(host, argv[optind + 2]);
+	} else if (!strcasecmp(argv[optind], "unmon_all")) {
+		err = statdtest_unmon_all(host, my_name, my_prog, my_vers);
+	} else if (!strcasecmp(argv[optind], "unmon")) {
+		if (remaining_args < 2)
+			usage(argv[0]);
+		err = statdtest_unmon(host, argv[optind + 1], my_name, my_prog,
+					my_vers);
+	} else if (!strcasecmp(argv[optind], "notify")) {
+		if (remaining_args < 2)
+			usage(argv[0]);
+		err = statdtest_notify(host, argv[optind + 1],
+					argv[optind + 2]);
+	} else if (!strcasecmp(argv[optind], "mon")) {
+		if (remaining_args < 2)
+			usage(argv[0]);
+
+		memset(cookie, '\0', SM_PRIV_SIZE);
+		if (!hex2bin(cookie, sizeof(cookie), argv[optind + 2])) {
+			fprintf(stderr, "SYS:%d\n", EINVAL);
+			printf("Unable to convert hex cookie %s to binary.\n",
+				argv[optind + 2]);
+			return 1;
+		}
+
+		err = statdtest_mon(host, argv[optind + 1], cookie, my_name,
+					my_prog, my_vers);
+	} else {
+		err = usage(argv[0]);
+	}
+
+	return err;
+}
+
+static CLIENT *
+statdtest_get_rpcclient(const char *node)
+{
+	unsigned short		port;
+	struct addrinfo		*ai;
+	struct addrinfo		hints = { .ai_flags	= AI_ADDRCONFIG };
+	int			err;
+	CLIENT			*client = NULL;
+
+#ifndef IPV6_ENABLED
+	hints.ai_family	= AF_INET;
+#endif /* IPV6_ENABLED */
+
+	/* FIXME: allow support for providing port? */
+	err = getaddrinfo(node, NULL, &hints, &ai);
+	if (err) {
+		fprintf(stderr, "EAI:%d\n", err);
+		if (err == EAI_SYSTEM)
+			fprintf(stderr, "SYS:%d\n", errno);
+		printf("Unable to translate host to address: %s\n",
+			err == EAI_SYSTEM ? strerror(errno) :
+			gai_strerror(err));
+		return client;
+	}
+
+	/* FIXME: allow for TCP too? */
+	port = nfs_getport(ai->ai_addr, ai->ai_addrlen, SM_PROG,
+			   SM_VERS, IPPROTO_UDP);
+	if (!port) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("Unable to determine port for service\n");
+		goto out;
+	}
+
+	nfs_set_port(ai->ai_addr, port);
+
+	client = nfs_get_rpcclient(ai->ai_addr, ai->ai_addrlen, IPPROTO_UDP,
+				   SM_PROG, SM_VERS, &retrans_interval);
+	if (!client) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("RPC client creation failed\n");
+	}
+out:
+	freeaddrinfo(ai);
+	return client;
+}
+
+static int
+statdtest_mon(char *calling, char *monitoring, char *cookie, char *my_name,
+		int my_prog, int my_vers)
+{
+	CLIENT *client;
+	sm_stat_res *result;
+	mon mon;
+	int err = 0;
+
+	printf("Calling %s (as %s) to monitor %s\n", calling, my_name,
+		monitoring);
+
+	if ((client = statdtest_get_rpcclient(calling)) == NULL)
+		return 1;
+
+	memcpy(mon.priv, cookie, SM_PRIV_SIZE);
+	mon.mon_id.my_id.my_name = my_name;
+	mon.mon_id.my_id.my_prog = my_prog;
+	mon.mon_id.my_id.my_vers = my_vers;
+	mon.mon_id.my_id.my_proc = NLM_SM_NOTIFY;
+	mon.mon_id.mon_name = monitoring;
+
+	if (!(result = sm_mon_1(&mon, client))) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("%s\n", clnt_sperror(client, "sm_mon_1"));
+		err = 1;
+		goto mon_out;
+	}
+
+	printf("SM_MON request %s, state: %d\n",
+		result->res_stat == stat_succ ? "successful" : "failed",
+		result->state);
+
+	if (result->res_stat != stat_succ) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		err = 1;
+	}
+
+mon_out:
+	clnt_destroy(client);
+	return err;
+}
+
+static int
+statdtest_unmon(char *calling, char *unmonitoring, char *my_name, int my_prog,
+		int my_vers)
+{
+	CLIENT *client;
+	sm_stat *result;
+	mon_id mon_id;
+	int err = 0;
+
+	printf("Calling %s (as %s) to unmonitor %s\n", calling, my_name,
+		unmonitoring);
+
+	if ((client = statdtest_get_rpcclient(calling)) == NULL)
+		return 1;
+
+	mon_id.my_id.my_name = my_name;
+	mon_id.my_id.my_prog = my_prog;
+	mon_id.my_id.my_vers = my_vers;
+	mon_id.my_id.my_proc = NLM_SM_NOTIFY;
+	mon_id.mon_name = unmonitoring;
+
+	if (!(result = sm_unmon_1(&mon_id, client))) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("%s\n", clnt_sperror(client, "sm_unmon_1"));
+		err = 1;
+		goto unmon_out;
+	}
+
+	printf("SM_UNMON state: %d\n", result->state);
+
+unmon_out:
+	clnt_destroy(client);
+	return err;
+}
+
+static int
+statdtest_unmon_all(char *calling, char *my_name, int my_prog, int my_vers)
+{
+	CLIENT *client;
+	sm_stat *result;
+	my_id my_id;
+	int err = 0;
+
+	printf("Calling %s (as %s) to unmonitor all hosts\n", calling, my_name);
+
+	if ((client = statdtest_get_rpcclient(calling)) == NULL) {
+		printf("RPC client creation failed\n");
+		return 1;
+	}
+
+	my_id.my_name = my_name;
+	my_id.my_prog = my_prog;
+	my_id.my_vers = my_vers;
+	my_id.my_proc = NLM_SM_NOTIFY;
+
+	if (!(result = sm_unmon_all_1(&my_id, client))) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("%s\n", clnt_sperror(client, "sm_unmon_all_1"));
+		err = 1;
+		goto unmon_all_out;
+	}
+
+	printf("SM_UNMON_ALL state: %d\n", result->state);
+
+unmon_all_out:
+	return err;
+}
+
+static int
+statdtest_crash(char *host)
+{
+	CLIENT *client;
+
+	if ((client = statdtest_get_rpcclient(host)) == NULL)
+		return 1;
+
+	if (!sm_simu_crash_1(NULL, client)) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("%s\n", clnt_sperror(client, "sm_simu_crash_1"));
+		return 1;
+	}
+
+	return 0;
+}
+
+static int
+statdtest_stat(char *calling, char *monitoring)
+{
+	CLIENT *client;
+	sm_name checking;
+	sm_stat_res *result;
+
+	if ((client = statdtest_get_rpcclient(calling)) == NULL)
+		return 1;
+
+	checking.mon_name = monitoring;
+
+	if (!(result = sm_stat_1(&checking, client))) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("%s\n", clnt_sperror(client, "sm_stat_1"));
+		return 1;
+	}
+
+	if (result->res_stat != stat_succ) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("stat_fail from %s for %s, state: %d\n", calling,
+			monitoring, result->state);
+		return 1;
+	}
+
+	printf("stat_succ from %s for %s, state: %d\n", calling,
+		monitoring, result->state);
+
+	return 0;
+}
+
+static int
+statdtest_notify(char *calling, char *mon_name, char *statestr)
+{
+	CLIENT *client;
+
+	stat_chge stat_chge = { .mon_name	= mon_name };
+
+	stat_chge.state = atoi(statestr);
+
+	if ((client = statdtest_get_rpcclient(calling)) == NULL)
+		return 1;
+
+	if (!sm_notify_1(&stat_chge, client)) {
+		fprintf(stderr, "RPC:%d\n", rpc_createerr.cf_stat);
+		printf("%s\n", clnt_sperror(client, "sm_notify_1"));
+		return 1;
+	}
+
+	return 0;
+}
+
+static void sim_killer(int sig)
+{
+#ifdef HAVE_LIBTIRPC
+	(void)rpcb_unset(NLM_SM_PROG, NLM_SM_VERS4, NULL);
+#else
+	(void)pmap_unset(NLM_SM_PROG, NLM_SM_VERS4);
+#endif
+	exit(0);
+}
+
+static void daemon_simulator(void)
+{
+	signal(SIGHUP, sim_killer);
+	signal(SIGINT, sim_killer);
+	signal(SIGTERM, sim_killer);
+	/* FIXME: allow for different versions? */
+	nfs_svc_create("nlmsim", NLM_SM_PROG, NLM_SM_VERS4, nlm_sm_prog_4, 0);
+	svc_run();
+}
+
+void *nlm_sm_notify_4_svc(struct nlm_sm_notify *argp, struct svc_req *rqstp)
+{
+	static char *result;
+	char	    priv[SM_PRIV_SIZE * 2 + 1];
+
+	bin2hex(priv, argp->priv, SM_PRIV_SIZE);
+
+	printf("state=%d:mon_name=%s:private=%s\n", argp->state,
+		argp->mon_name, priv);
+	return ((void *)&result);
+}
+
+void *nlm_sm_notify_3_svc(struct nlm_sm_notify *argp, struct svc_req *rqstp)
+{
+	return nlm_sm_notify_4_svc(argp, rqstp);
+}
-- 
1.6.5.2

[PATCH 2/2] nfs-utils: add initial tests for statd that run via "make check"

[Jeff Layton]

Leverage the support that automake already has for running tests via
make check. Add a simple test that just checks that the statd mon and
unmon calls actually work.

Adding more tests should be a simple matter of adding new scripts
exit 0 on success and non-zero on fail, and adding those to the
Makefile.am.

Signed-off-by: Jeff Layton <jlayton@redhat.com>
---
 tests/Makefile.am                    |    2 +
 tests/t0001-statd-basic-mon-unmon.sh |   55 ++++++++++++++++++++++++++++++++++
 tests/test-lib.sh                    |   41 +++++++++++++++++++++++++
 3 files changed, 98 insertions(+), 0 deletions(-)
 create mode 100755 tests/t0001-statd-basic-mon-unmon.sh
 create mode 100755 tests/test-lib.sh

diff --git a/tests/Makefile.am b/tests/Makefile.am
index 997ac51..634fb7b 100644
--- a/tests/Makefile.am
+++ b/tests/Makefile.am
@@ -3,3 +3,5 @@
 SUBDIRS = statdtest
 
 MAINTAINERCLEANFILES = Makefile.in
+
+TESTS = t0001-statd-basic-mon-unmon.sh
diff --git a/tests/t0001-statd-basic-mon-unmon.sh b/tests/t0001-statd-basic-mon-unmon.sh
new file mode 100755
index 0000000..d5dffc5
--- /dev/null
+++ b/tests/t0001-statd-basic-mon-unmon.sh
@@ -0,0 +1,55 @@
+#!/bin/bash
+#
+# statd_basic_mon_unmon -- test basic mon/unmon functionality with statd
+#
+# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License
+# as published by the Free Software Foundation; either version 2
+# of the License, or (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation, Inc.,
+# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+#
+
+. ./test-lib.sh
+
+start_statd
+if [ $? -ne 0 ]; then
+	echo "FAIL: problem starting statd"
+	exit 1
+fi
+
+COOKIE=`echo $$ | md5sum | cut -d' ' -f1`
+MON_NAME=`hostname`
+
+statdtest mon $MON_NAME $COOKIE
+if [ $? -ne 0 ]; then
+	echo "FAIL: mon failed"
+	kill_statd
+	exit 1
+fi
+
+grep -q $COOKIE /var/lib/nfs/statd/sm/$MON_NAME 
+if [ $? -ne 0 ]; then
+	echo "FAIL: $COOKIE not present in monitor file"
+	kill_statd
+	exit 1
+fi
+
+statdtest unmon $MON_NAME
+if [ $? -ne 0 ]; then
+	echo "FAIL: unmon failed"
+	kill_statd
+	exit 1
+fi
+
+kill_statd
+
diff --git a/tests/test-lib.sh b/tests/test-lib.sh
new file mode 100755
index 0000000..053e46d
--- /dev/null
+++ b/tests/test-lib.sh
@@ -0,0 +1,41 @@
+#!/bin/bash
+#
+# test-lib.sh -- library of functions for nfs-utils tests
+#
+# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License
+# as published by the Free Software Foundation; either version 2
+# of the License, or (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation, Inc.,
+# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
+#
+
+export PATH=$PATH:$srcdir/statdtest
+
+lockd_registered() {
+	rpcinfo -p | grep -q nlockmgr
+	return $?
+}
+
+start_statd() {
+	rpcinfo -u 127.0.0.1 status 1 &> /dev/null
+	if [ $? -eq 0 ]; then
+		echo "***ERROR***: statd is already running and should "
+		echo "             be down when starting this test"
+		return 1
+	fi
+	$srcdir/../utils/statd/statd --no-notify
+}
+
+kill_statd() {
+	kill `cat /var/run/rpc.statd.pid`
+}
-- 
1.6.5.2

Re: [PATCH 2/2] nfs-utils: add initial tests for statd that run via "make check"

[Steve Dickson]

On 12/16/2009 02:40 PM, Jeff Layton wrote:
> Leverage the support that automake already has for running tests via
> make check. Add a simple test that just checks that the statd mon and
> unmon calls actually work.
> 
> Adding more tests should be a simple matter of adding new scripts
> exit 0 on success and non-zero on fail, and adding those to the
> Makefile.am.
> 
> Signed-off-by: Jeff Layton <jlayton@redhat.com>
> ---
>  tests/Makefile.am                    |    2 +
>  tests/t0001-statd-basic-mon-unmon.sh |   55 ++++++++++++++++++++++++++++++++++
>  tests/test-lib.sh                    |   41 +++++++++++++++++++++++++
>  3 files changed, 98 insertions(+), 0 deletions(-)
>  create mode 100755 tests/t0001-statd-basic-mon-unmon.sh
>  create mode 100755 tests/test-lib.sh
> 
> diff --git a/tests/Makefile.am b/tests/Makefile.am
> index 997ac51..634fb7b 100644
> --- a/tests/Makefile.am
> +++ b/tests/Makefile.am
> @@ -3,3 +3,5 @@
>  SUBDIRS = statdtest
>  
>  MAINTAINERCLEANFILES = Makefile.in
> +
> +TESTS = t0001-statd-basic-mon-unmon.sh
> diff --git a/tests/t0001-statd-basic-mon-unmon.sh b/tests/t0001-statd-basic-mon-unmon.sh
> new file mode 100755
> index 0000000..d5dffc5
> --- /dev/null
> +++ b/tests/t0001-statd-basic-mon-unmon.sh
> @@ -0,0 +1,55 @@
> +#!/bin/bash
> +#
> +# statd_basic_mon_unmon -- test basic mon/unmon functionality with statd
> +#
> +# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
> +#
> +# This program is free software; you can redistribute it and/or
> +# modify it under the terms of the GNU General Public License
> +# as published by the Free Software Foundation; either version 2
> +# of the License, or (at your option) any later version.
> +#
> +# This program is distributed in the hope that it will be useful,
> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> +# GNU General Public License for more details.
> +#
> +# You should have received a copy of the GNU General Public License
> +# along with this program; if not, write to the Free Software Foundation, Inc.,
> +# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
> +#
> +
> +. ./test-lib.sh
> +
> +start_statd
> +if [ $? -ne 0 ]; then
> +	echo "FAIL: problem starting statd"
> +	exit 1
> +fi
> +
> +COOKIE=`echo $$ | md5sum | cut -d' ' -f1`
> +MON_NAME=`hostname`
> +
> +statdtest mon $MON_NAME $COOKIE
> +if [ $? -ne 0 ]; then
> +	echo "FAIL: mon failed"
> +	kill_statd
> +	exit 1
> +fi
> +
> +grep -q $COOKIE /var/lib/nfs/statd/sm/$MON_NAME
This is a distro specific path... The default is /var/lib/nfs/sm
It would be nice if somehow we could pull in the NFS_STATEDIR define
from support/include/config.h file... 

> +if [ $? -ne 0 ]; then
> +	echo "FAIL: $COOKIE not present in monitor file"
> +	kill_statd
> +	exit 1
> +fi
> +
> +statdtest unmon $MON_NAME
> +if [ $? -ne 0 ]; then
> +	echo "FAIL: unmon failed"
> +	kill_statd
> +	exit 1
> +fi
> +
> +kill_statd
> +
> diff --git a/tests/test-lib.sh b/tests/test-lib.sh
> new file mode 100755
> index 0000000..053e46d
> --- /dev/null
> +++ b/tests/test-lib.sh
> @@ -0,0 +1,41 @@
> +#!/bin/bash
> +#
> +# test-lib.sh -- library of functions for nfs-utils tests
> +#
> +# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
> +#
> +# This program is free software; you can redistribute it and/or
> +# modify it under the terms of the GNU General Public License
> +# as published by the Free Software Foundation; either version 2
> +# of the License, or (at your option) any later version.
> +#
> +# This program is distributed in the hope that it will be useful,
> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> +# GNU General Public License for more details.
> +#
> +# You should have received a copy of the GNU General Public License
> +# along with this program; if not, write to the Free Software Foundation, Inc.,
> +# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
> +#
> +
> +export PATH=$PATH:$srcdir/statdtest
> +
> +lockd_registered() {
> +	rpcinfo -p | grep -q nlockmgr
> +	return $?
> +}
> +
> +start_statd() {
> +	rpcinfo -u 127.0.0.1 status 1 &> /dev/null
> +	if [ $? -eq 0 ]; then
> +		echo "***ERROR***: statd is already running and should "
> +		echo "             be down when starting this test"
> +		return 1
> +	fi
> +	$srcdir/../utils/statd/statd --no-notify
the $srcdir variable is not define so the script fails
right out of the box... So maybe something like this...
Initialize srcdir=`dirname $PWD`
then '$srcdir/utils/statd/statd --no-notify' will work.

> +}
> +
> +kill_statd() {
> +	kill `cat /var/run/rpc.statd.pid`
> +}

Overall I like the idea of having a test harness... but I have
to wonder if we start building the harness in shell scripts 
if we are not limiting things right out of the box... 

Maybe we should consider doing the harness in a more up to date
scripting languages like perl or python or even java... Granted 
I'm not as familiar (or comfortable) with those languages as with
shell scripting. Plus its not clear what type of dependencies that
would added to the core nfs-utils code, but we should at least
think about using a language that might give us more 
possibilities down the road... Who knowns, in the end, shell scripting
might be the best way to go... 

steved.

Re: [PATCH 2/2] nfs-utils: add initial tests for statd that run via "make check"

[Jeff Layton]

On Thu, 24 Dec 2009 08:27:01 -0500
Steve Dickson <SteveD@redhat.com> wrote:

> On 12/16/2009 02:40 PM, Jeff Layton wrote:
> > Leverage the support that automake already has for running tests via
> > make check. Add a simple test that just checks that the statd mon and
> > unmon calls actually work.
> > 
> > Adding more tests should be a simple matter of adding new scripts
> > exit 0 on success and non-zero on fail, and adding those to the
> > Makefile.am.
> > 
> > Signed-off-by: Jeff Layton <jlayton@redhat.com>
> > ---
> >  tests/Makefile.am                    |    2 +
> >  tests/t0001-statd-basic-mon-unmon.sh |   55 ++++++++++++++++++++++++++++++++++
> >  tests/test-lib.sh                    |   41 +++++++++++++++++++++++++
> >  3 files changed, 98 insertions(+), 0 deletions(-)
> >  create mode 100755 tests/t0001-statd-basic-mon-unmon.sh
> >  create mode 100755 tests/test-lib.sh
> > 
> > diff --git a/tests/Makefile.am b/tests/Makefile.am
> > index 997ac51..634fb7b 100644
> > --- a/tests/Makefile.am
> > +++ b/tests/Makefile.am
> > @@ -3,3 +3,5 @@
> >  SUBDIRS = statdtest
> >  
> >  MAINTAINERCLEANFILES = Makefile.in
> > +
> > +TESTS = t0001-statd-basic-mon-unmon.sh
> > diff --git a/tests/t0001-statd-basic-mon-unmon.sh b/tests/t0001-statd-basic-mon-unmon.sh
> > new file mode 100755
> > index 0000000..d5dffc5
> > --- /dev/null
> > +++ b/tests/t0001-statd-basic-mon-unmon.sh
> > @@ -0,0 +1,55 @@
> > +#!/bin/bash
> > +#
> > +# statd_basic_mon_unmon -- test basic mon/unmon functionality with statd
> > +#
> > +# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
> > +#
> > +# This program is free software; you can redistribute it and/or
> > +# modify it under the terms of the GNU General Public License
> > +# as published by the Free Software Foundation; either version 2
> > +# of the License, or (at your option) any later version.
> > +#
> > +# This program is distributed in the hope that it will be useful,
> > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > +# GNU General Public License for more details.
> > +#
> > +# You should have received a copy of the GNU General Public License
> > +# along with this program; if not, write to the Free Software Foundation, Inc.,
> > +# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
> > +#
> > +
> > +. ./test-lib.sh
> > +
> > +start_statd
> > +if [ $? -ne 0 ]; then
> > +	echo "FAIL: problem starting statd"
> > +	exit 1
> > +fi
> > +
> > +COOKIE=`echo $$ | md5sum | cut -d' ' -f1`
> > +MON_NAME=`hostname`
> > +
> > +statdtest mon $MON_NAME $COOKIE
> > +if [ $? -ne 0 ]; then
> > +	echo "FAIL: mon failed"
> > +	kill_statd
> > +	exit 1
> > +fi
> > +
> > +grep -q $COOKIE /var/lib/nfs/statd/sm/$MON_NAME
> This is a distro specific path... The default is /var/lib/nfs/sm
> It would be nice if somehow we could pull in the NFS_STATEDIR define
> from support/include/config.h file... 
> 

Good point. I'll need to think on how we can deal with that. Maybe what
I should do is add a new "statdtest" command that prints out this
directory? Unless someone has a better idea, I'll respin this set with
that change soon.

> > +if [ $? -ne 0 ]; then
> > +	echo "FAIL: $COOKIE not present in monitor file"
> > +	kill_statd
> > +	exit 1
> > +fi
> > +
> > +statdtest unmon $MON_NAME
> > +if [ $? -ne 0 ]; then
> > +	echo "FAIL: unmon failed"
> > +	kill_statd
> > +	exit 1
> > +fi
> > +
> > +kill_statd
> > +
> > diff --git a/tests/test-lib.sh b/tests/test-lib.sh
> > new file mode 100755
> > index 0000000..053e46d
> > --- /dev/null
> > +++ b/tests/test-lib.sh
> > @@ -0,0 +1,41 @@
> > +#!/bin/bash
> > +#
> > +# test-lib.sh -- library of functions for nfs-utils tests
> > +#
> > +# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
> > +#
> > +# This program is free software; you can redistribute it and/or
> > +# modify it under the terms of the GNU General Public License
> > +# as published by the Free Software Foundation; either version 2
> > +# of the License, or (at your option) any later version.
> > +#
> > +# This program is distributed in the hope that it will be useful,
> > +# but WITHOUT ANY WARRANTY; without even the implied warranty of
> > +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
> > +# GNU General Public License for more details.
> > +#
> > +# You should have received a copy of the GNU General Public License
> > +# along with this program; if not, write to the Free Software Foundation, Inc.,
> > +# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
> > +#
> > +
> > +export PATH=$PATH:$srcdir/statdtest
> > +
> > +lockd_registered() {
> > +	rpcinfo -p | grep -q nlockmgr
> > +	return $?
> > +}
> > +
> > +start_statd() {
> > +	rpcinfo -u 127.0.0.1 status 1 &> /dev/null
> > +	if [ $? -eq 0 ]; then
> > +		echo "***ERROR***: statd is already running and should "
> > +		echo "             be down when starting this test"
> > +		return 1
> > +	fi
> > +	$srcdir/../utils/statd/statd --no-notify
> the $srcdir variable is not define so the script fails
> right out of the box... So maybe something like this...
> Initialize srcdir=`dirname $PWD`
> then '$srcdir/utils/statd/statd --no-notify' will work.
> 

This script is run via "make check". Automake or autoconf (I forget
which) set $srcdir. There's no need to redefine it here.

> > +}
> > +
> > +kill_statd() {
> > +	kill `cat /var/run/rpc.statd.pid`
> > +}
> 
> Overall I like the idea of having a test harness... but I have
> to wonder if we start building the harness in shell scripts 
> if we are not limiting things right out of the box... 
> 
> Maybe we should consider doing the harness in a more up to date
> scripting languages like perl or python or even java... Granted 
> I'm not as familiar (or comfortable) with those languages as with
> shell scripting. Plus its not clear what type of dependencies that
> would added to the core nfs-utils code, but we should at least
> think about using a language that might give us more 
> possibilities down the road... Who knowns, in the end, shell scripting
> might be the best way to go... 
> 

One of the nice things about this "test harness" is that it's pretty
minimalistic. You can write tests in whatever you like as long as it
can be run from the makefile. While this test is written as a shell
script, the bulk of the test is a C program (statdtest.c).

I have no objection to writing tests in something else, but that does
add additional dependencies. There's something to be said for not
adding those unless they're really needed.

Thanks for the review so far,
-- 
Jeff Layton <jlayton@redhat.com>

Re: [PATCH 2/2] nfs-utils: add initial tests for statd that run via "make check"

[Chuck Lever]

On Dec 24, 2009, at 7:20 PM, Jeff Layton wrote:
> On Thu, 24 Dec 2009 08:27:01 -0500
> Steve Dickson <SteveD@redhat.com> wrote:
>
>> On 12/16/2009 02:40 PM, Jeff Layton wrote:
>>> Leverage the support that automake already has for running tests via
>>> make check. Add a simple test that just checks that the statd mon  
>>> and
>>> unmon calls actually work.
>>>
>>> Adding more tests should be a simple matter of adding new scripts
>>> exit 0 on success and non-zero on fail, and adding those to the
>>> Makefile.am.
>>>
>>> Signed-off-by: Jeff Layton <jlayton@redhat.com>
>>> ---
>>> tests/Makefile.am                    |    2 +
>>> tests/t0001-statd-basic-mon-unmon.sh |   55 +++++++++++++++++++++++ 
>>> +++++++++++
>>> tests/test-lib.sh                    |   41 +++++++++++++++++++++++ 
>>> ++
>>> 3 files changed, 98 insertions(+), 0 deletions(-)
>>> create mode 100755 tests/t0001-statd-basic-mon-unmon.sh
>>> create mode 100755 tests/test-lib.sh
>>>
>>> diff --git a/tests/Makefile.am b/tests/Makefile.am
>>> index 997ac51..634fb7b 100644
>>> --- a/tests/Makefile.am
>>> +++ b/tests/Makefile.am
>>> @@ -3,3 +3,5 @@
>>> SUBDIRS = statdtest
>>>
>>> MAINTAINERCLEANFILES = Makefile.in
>>> +
>>> +TESTS = t0001-statd-basic-mon-unmon.sh
>>> diff --git a/tests/t0001-statd-basic-mon-unmon.sh b/tests/t0001- 
>>> statd-basic-mon-unmon.sh
>>> new file mode 100755
>>> index 0000000..d5dffc5
>>> --- /dev/null
>>> +++ b/tests/t0001-statd-basic-mon-unmon.sh
>>> @@ -0,0 +1,55 @@
>>> +#!/bin/bash
>>> +#
>>> +# statd_basic_mon_unmon -- test basic mon/unmon functionality  
>>> with statd
>>> +#
>>> +# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
>>> +#
>>> +# This program is free software; you can redistribute it and/or
>>> +# modify it under the terms of the GNU General Public License
>>> +# as published by the Free Software Foundation; either version 2
>>> +# of the License, or (at your option) any later version.
>>> +#
>>> +# This program is distributed in the hope that it will be useful,
>>> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
>>> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>>> +# GNU General Public License for more details.
>>> +#
>>> +# You should have received a copy of the GNU General Public License
>>> +# along with this program; if not, write to the Free Software  
>>> Foundation, Inc.,
>>> +# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
>>> +#
>>> +
>>> +. ./test-lib.sh
>>> +
>>> +start_statd
>>> +if [ $? -ne 0 ]; then
>>> +	echo "FAIL: problem starting statd"
>>> +	exit 1
>>> +fi
>>> +
>>> +COOKIE=`echo $$ | md5sum | cut -d' ' -f1`
>>> +MON_NAME=`hostname`
>>> +
>>> +statdtest mon $MON_NAME $COOKIE
>>> +if [ $? -ne 0 ]; then
>>> +	echo "FAIL: mon failed"
>>> +	kill_statd
>>> +	exit 1
>>> +fi
>>> +
>>> +grep -q $COOKIE /var/lib/nfs/statd/sm/$MON_NAME
>> This is a distro specific path... The default is /var/lib/nfs/sm
>> It would be nice if somehow we could pull in the NFS_STATEDIR define
>> from support/include/config.h file...
>>
>
> Good point. I'll need to think on how we can deal with that. Maybe  
> what
> I should do is add a new "statdtest" command that prints out this
> directory? Unless someone has a better idea, I'll respin this set with
> that change soon.

We'd be better off, I think, if you did this test using calls to  
libnsm.a, rather than naked greps.  You're adding yet another  
invisible and undocumented dependency on the "formal API" that is the  
state directory.  I'd rather see the C API tested here, and then  
deprecate direct accesses over time.

>>> +if [ $? -ne 0 ]; then
>>> +	echo "FAIL: $COOKIE not present in monitor file"
>>> +	kill_statd
>>> +	exit 1
>>> +fi
>>> +
>>> +statdtest unmon $MON_NAME
>>> +if [ $? -ne 0 ]; then
>>> +	echo "FAIL: unmon failed"
>>> +	kill_statd
>>> +	exit 1
>>> +fi
>>> +
>>> +kill_statd
>>> +
>>> diff --git a/tests/test-lib.sh b/tests/test-lib.sh
>>> new file mode 100755
>>> index 0000000..053e46d
>>> --- /dev/null
>>> +++ b/tests/test-lib.sh
>>> @@ -0,0 +1,41 @@
>>> +#!/bin/bash
>>> +#
>>> +# test-lib.sh -- library of functions for nfs-utils tests
>>> +#
>>> +# Copyright (C) 2009  Red Hat, Jeff Layton <jlayton@redhat.com>
>>> +#
>>> +# This program is free software; you can redistribute it and/or
>>> +# modify it under the terms of the GNU General Public License
>>> +# as published by the Free Software Foundation; either version 2
>>> +# of the License, or (at your option) any later version.
>>> +#
>>> +# This program is distributed in the hope that it will be useful,
>>> +# but WITHOUT ANY WARRANTY; without even the implied warranty of
>>> +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
>>> +# GNU General Public License for more details.
>>> +#
>>> +# You should have received a copy of the GNU General Public License
>>> +# along with this program; if not, write to the Free Software  
>>> Foundation, Inc.,
>>> +# 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.
>>> +#
>>> +
>>> +export PATH=$PATH:$srcdir/statdtest
>>> +
>>> +lockd_registered() {
>>> +	rpcinfo -p | grep -q nlockmgr
>>> +	return $?
>>> +}
>>> +
>>> +start_statd() {
>>> +	rpcinfo -u 127.0.0.1 status 1 &> /dev/null
>>> +	if [ $? -eq 0 ]; then
>>> +		echo "***ERROR***: statd is already running and should "
>>> +		echo "             be down when starting this test"
>>> +		return 1
>>> +	fi
>>> +	$srcdir/../utils/statd/statd --no-notify
>> the $srcdir variable is not define so the script fails
>> right out of the box... So maybe something like this...
>> Initialize srcdir=`dirname $PWD`
>> then '$srcdir/utils/statd/statd --no-notify' will work.
>>
>
> This script is run via "make check". Automake or autoconf (I forget
> which) set $srcdir. There's no need to redefine it here.
>
>>> +}
>>> +
>>> +kill_statd() {
>>> +	kill `cat /var/run/rpc.statd.pid`
>>> +}
>>
>> Overall I like the idea of having a test harness... but I have
>> to wonder if we start building the harness in shell scripts
>> if we are not limiting things right out of the box...
>>
>> Maybe we should consider doing the harness in a more up to date
>> scripting languages like perl or python or even java... Granted
>> I'm not as familiar (or comfortable) with those languages as with
>> shell scripting. Plus its not clear what type of dependencies that
>> would added to the core nfs-utils code, but we should at least
>> think about using a language that might give us more
>> possibilities down the road... Who knowns, in the end, shell  
>> scripting
>> might be the best way to go...
>>
>
> One of the nice things about this "test harness" is that it's pretty
> minimalistic. You can write tests in whatever you like as long as it
> can be run from the makefile. While this test is written as a shell
> script, the bulk of the test is a C program (statdtest.c).
>
> I have no objection to writing tests in something else, but that does
> add additional dependencies. There's something to be said for not
> adding those unless they're really needed.
>
> Thanks for the review so far,
> -- 
> Jeff Layton <jlayton@redhat.com>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-nfs"  
> in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

-- 
Chuck Lever
chuck[dot]lever[at]oracle[dot]com

Re: [PATCH 2/2] nfs-utils: add initial tests for statd that run via "make check"

[Jeff Layton]

On Tue, 29 Dec 2009 10:38:38 -0500
Chuck Lever <chuck.lever@oracle.com> wrote:

> >>> +grep -q $COOKIE /var/lib/nfs/statd/sm/$MON_NAME
> >> This is a distro specific path... The default is /var/lib/nfs/sm
> >> It would be nice if somehow we could pull in the NFS_STATEDIR define
> >> from support/include/config.h file...
> >>
> >
> > Good point. I'll need to think on how we can deal with that. Maybe  
> > what
> > I should do is add a new "statdtest" command that prints out this
> > directory? Unless someone has a better idea, I'll respin this set with
> > that change soon.
> 
> We'd be better off, I think, if you did this test using calls to  
> libnsm.a, rather than naked greps.  You're adding yet another  
> invisible and undocumented dependency on the "formal API" that is the  
> state directory.  I'd rather see the C API tested here, and then  
> deprecate direct accesses over time.
> 

Ok, maybe I need to sit down and write a "statdb" utility that can
read/manipulate the statd DB via the libnsm.a interfaces? That might be
handy for other purposes too. I'll look at that when I get back from
vacation...

-- 
Jeff Layton <jlayton@redhat.com>