Add scrub after resilver zed script

* Add a zed script to kick off a scrub after a resilver.  The script is
disabled by default.

* Add a optional $PATH (-P) option to zed to allow it to use a custom
$PATH for its zedlets.  This is needed when you're running zed under
the ZTS in a local workspace.

* Update test scripts to not copy in all-debug.sh and all-syslog.sh by
default.  They can be optionally copied in as part of zed_setup().
These scripts slow down zed considerably under heavy events loads and
can cause events to be dropped or their delivery delayed. This was
causing some sporadic failures in the 'fault' tests.

Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Reviewed-by: Richard Laager <rlaager@wiktel.com>
Signed-off-by: Tony Hutter <hutter2@llnl.gov>
Closes #4662 
Closes #7086

cmd/zed/Makefile.am

@@ -66,7 +66,8 @@ dist_zedexec_SCRIPTS = \
 	zed.d/statechange-notify.sh \
 	zed.d/vdev_clear-led.sh \
 	zed.d/vdev_attach-led.sh \
-	zed.d/pool_import-led.sh
+	zed.d/pool_import-led.sh \
+	zed.d/resilver_finish-start-scrub.sh
 
 zedconfdefaults = \
 	all-syslog.sh \
@@ -77,7 +78,8 @@ zedconfdefaults = \
 	statechange-notify.sh \
 	vdev_clear-led.sh \
 	vdev_attach-led.sh \
-	pool_import-led.sh
+	pool_import-led.sh \
+	resilver_finish-start-scrub.sh
 
 install-data-hook:
 	$(MKDIR_P) "$(DESTDIR)$(zedconfdir)"

cmd/zed/zed.d/resilver_finish-start-scrub.sh

@@ -0,0 +1,17 @@
+#!/bin/sh
+# resilver_finish-start-scrub.sh
+# Run a scrub after a resilver
+#
+# Exit codes:
+# 1: Internal error
+# 2: Script wasn't enabled in zed.rc
+[ -f "${ZED_ZEDLET_DIR}/zed.rc" ] && . "${ZED_ZEDLET_DIR}/zed.rc"
+. "${ZED_ZEDLET_DIR}/zed-functions.sh"
+
+[ "${ZED_SCRUB_AFTER_RESILVER}" = "1" ] || exit 2
+[ -n "${ZEVENT_POOL}" ] || exit 1
+[ -n "${ZEVENT_SUBCLASS}" ] || exit 1
+zed_check_cmd "${ZPOOL}" || exit 1
+
+zed_log_msg "Starting scrub after resilver on ${ZEVENT_POOL}"
+"${ZPOOL}" scrub "${ZEVENT_POOL}"

cmd/zed/zed.d/zed.rc

@@ -86,6 +86,9 @@
 #
 ZED_USE_ENCLOSURE_LEDS=1
 
+##
+# Run a scrub after every resilver
+#ZED_SCRUB_AFTER_RESILVER=1
 
 ##
 # The syslog priority (e.g., specified as a "facility.level" pair).

cmd/zed/zed_conf.c

@@ -155,6 +155,8 @@ _zed_conf_display_help(const char *prog, int got_err)
 	    "Run daemon in the foreground.");
 	fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-M",
 	    "Lock all pages in memory.");
+	fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-P",
+	    "$PATH for ZED to use (only used by ZTS).");
 	fprintf(fp, "%*c%*s %s\n", w1, 0x20, -w2, "-Z",
 	    "Zero state file.");
 	fprintf(fp, "\n");
@@ -247,7 +249,7 @@ _zed_conf_parse_path(char **resultp, const char *path)
 void
 zed_conf_parse_opts(struct zed_conf *zcp, int argc, char **argv)
 {
-	const char * const opts = ":hLVc:d:p:s:vfFMZ";
+	const char * const opts = ":hLVc:d:p:P:s:vfFMZ";
 	int opt;
 
 	if (!zcp || !argv || !argv[0])
@@ -275,6 +277,9 @@ zed_conf_parse_opts(struct zed_conf *zcp, int argc, char **argv)
 		case 'p':
 			_zed_conf_parse_path(&zcp->pid_file, optarg);
 			break;
+		case 'P':
+			_zed_conf_parse_path(&zcp->path, optarg);
+			break;
 		case 's':
 			_zed_conf_parse_path(&zcp->state_file, optarg);
 			break;

cmd/zed/zed_conf.h

@@ -37,6 +37,7 @@ struct zed_conf {
 	int		state_fd;		/* fd to state file */
 	libzfs_handle_t	*zfs_hdl;		/* handle to libzfs */
 	int		zevent_fd;		/* fd for access to zevents */
+	char		*path;		/* custom $PATH for zedlets to use */
 };
 
 struct zed_conf *zed_conf_create(void);

cmd/zed/zed_event.c

@@ -733,12 +733,14 @@ _zed_event_add_nvpair(uint64_t eid, zed_strings_t *zsp, nvpair_t *nvp)
 
 /*
  * Restrict various environment variables to safe and sane values
- * when constructing the environment for the child process.
+ * when constructing the environment for the child process, unless
+ * we're running with a custom $PATH (like under the ZFS test suite).
  *
  * Reference: Secure Programming Cookbook by Viega & Messier, Section 1.1.
  */
 static void
-_zed_event_add_env_restrict(uint64_t eid, zed_strings_t *zsp)
+_zed_event_add_env_restrict(uint64_t eid, zed_strings_t *zsp,
+    const char *path)
 {
 	const char *env_restrict[][2] = {
 		{ "IFS",		" \t\n" },
@@ -753,11 +755,35 @@ _zed_event_add_env_restrict(uint64_t eid, zed_strings_t *zsp)
 		{ "ZFS_RELEASE",	ZFS_META_RELEASE },
 		{ NULL,			NULL }
 	};
+
+	/*
+	 * If we have a custom $PATH, use the default ZFS binary locations
+	 * instead of the hard-coded ones.
+	 */
+	const char *env_path[][2] = {
+		{ "IFS",		" \t\n" },
+		{ "PATH",		NULL }, /* $PATH copied in later on */
+		{ "ZDB",		"zdb" },
+		{ "ZED",		"zed" },
+		{ "ZFS",		"zfs" },
+		{ "ZINJECT",		"zinject" },
+		{ "ZPOOL",		"zpool" },
+		{ "ZFS_ALIAS",		ZFS_META_ALIAS },
+		{ "ZFS_VERSION",	ZFS_META_VERSION },
+		{ "ZFS_RELEASE",	ZFS_META_RELEASE },
+		{ NULL,			NULL }
+	};
 	const char *(*pa)[2];
 
 	assert(zsp != NULL);
 
-	for (pa = env_restrict; *(*pa); pa++) {
+	pa = path != NULL ? env_path : env_restrict;
+
+	for (; *(*pa); pa++) {
+		/* Use our custom $PATH if we have one */
+		if (path != NULL && strcmp((*pa)[0], "PATH") == 0)
+			(*pa)[1] = path;
+
 		_zed_event_add_var(eid, zsp, NULL, (*pa)[0], "%s", (*pa)[1]);
 	}
 }
@@ -902,7 +928,7 @@ zed_event_service(struct zed_conf *zcp)
 		while ((nvp = nvlist_next_nvpair(nvl, nvp)))
 			_zed_event_add_nvpair(eid, zsp, nvp);
 
-		_zed_event_add_env_restrict(eid, zsp);
+		_zed_event_add_env_restrict(eid, zsp, zcp->path);
 		_zed_event_add_env_preserve(eid, zsp);
 
 		_zed_event_add_var(eid, zsp, ZED_VAR_PREFIX, "PID",

man/man8/zed.8.in

@@ -27,6 +27,7 @@ ZED \- ZFS Event Daemon
 [\fB\-L\fR]
 [\fB\-M\fR]
 [\fB\-p\fR \fIpidfile\fR]
+[\fB\-P\fR \fIpath\fR]
 [\fB\-s\fR \fIstatefile\fR]
 [\fB\-v\fR]
 [\fB\-V\fR]
@@ -78,9 +79,16 @@ Read the enabled ZEDLETs from the specified directory.
 .BI \-p\  pidfile
 Write the daemon's process ID to the specified file.
 .TP
+.BI \-P\  path
+Custom $PATH for zedlets to use.  Normally zedlets run in a locked-down
+environment, with hardcoded paths to the ZFS commands ($ZFS, $ZPOOL, $ZED, ...),
+and a hardcoded $PATH.  This is done for security reasons.  However, the
+ZFS test suite uses a custom PATH for its ZFS commands, and passes it to zed
+with -P.  In short, -P is only to be used by the ZFS test suite; never use
+it in production!
+.TP
 .BI \-s\  statefile
 Write the daemon's state to the specified file.
-
 .SH ZEVENTS
 .PP
 A zevent is comprised of a list of nvpairs (name/value pairs).  Each zevent

tests/runfiles/linux.run

@@ -498,7 +498,8 @@ tags = ['functional', 'exec']
 
 [tests/functional/fault]
 tests = ['auto_online_001_pos', 'auto_replace_001_pos', 'auto_spare_001_pos',
-    'auto_spare_002_pos', 'auto_spare_ashift', 'auto_spare_multiple']
+    'auto_spare_002_pos', 'auto_spare_ashift', 'auto_spare_multiple',
+    'scrub_after_resilver']
 tags = ['functional', 'fault']
 
 [tests/functional/features/async_destroy]

tests/zfs-tests/include/commands.cfg

@@ -86,6 +86,7 @@ export SYSTEM_FILES='arp
     pgrep
     ping
     pkill
+    printenv
     printf
     ps
     pwd

tests/zfs-tests/include/libtest.shlib

@@ -3053,9 +3053,32 @@ function wait_replacing #pool
 	done
 }
 
+#
+# Wait for a pool to be scrubbed
+#
+# $1 pool name
+# $2 number of seconds to wait (optional)
+#
+# Returns true when pool has been scrubbed, or false if there's a timeout or if
+# no scrub was done.
+#
+function wait_scrubbed
+{
+	typeset pool=${1:-$TESTPOOL}
+	typeset iter=${2:-10}
+	for i in {1..$iter} ; do
+		if is_pool_scrubbed $pool ; then
+			return 0
+		fi
+		sleep 1
+	done
+	return 1
+}
+
 #
 # Setup custom environment for the ZED.
 #
+# $@ Optional list of zedlets to run under zed.
 function zed_setup
 {
 	if ! is_linux; then
@@ -3073,6 +3096,7 @@ function zed_setup
 	if [[ -e $VDEVID_CONF_ETC ]]; then
 		log_fail "Must not have $VDEVID_CONF_ETC file present on system"
 	fi
+	EXTRA_ZEDLETS=$@
 
 	# Create a symlink for /etc/zfs/vdev_id.conf file.
 	log_must ln -s $VDEVID_CONF $VDEVID_CONF_ETC
@@ -3082,32 +3106,44 @@ function zed_setup
 	log_must cp ${ZEDLET_ETC_DIR}/zed.rc $ZEDLET_DIR
 	log_must cp ${ZEDLET_ETC_DIR}/zed-functions.sh $ZEDLET_DIR
 
+	# Scripts must only be user writable.
+	if [[ -n "$EXTRA_ZEDLETS" ]] ; then
+		saved_umask=$(umask)
+		log_must umask 0022
+		for i in $EXTRA_ZEDLETS ; do
+			log_must cp ${ZEDLET_LIBEXEC_DIR}/$i $ZEDLET_DIR
+		done
+		log_must umask $saved_umask
+	fi
+
 	# Customize the zed.rc file to enable the full debug log.
 	log_must sed -i '/\#ZED_DEBUG_LOG=.*/d' $ZEDLET_DIR/zed.rc
 	echo "ZED_DEBUG_LOG=$ZED_DEBUG_LOG" >>$ZEDLET_DIR/zed.rc
 
-	# Scripts must only be user writable.
-	saved_umask=$(umask)
-	log_must umask 0022
-	log_must cp ${ZEDLET_LIBEXEC_DIR}/all-syslog.sh $ZEDLET_DIR
-	log_must cp ${ZEDLET_LIBEXEC_DIR}/all-debug.sh $ZEDLET_DIR
-	log_must umask $saved_umask
 }
 
 #
 # Cleanup custom ZED environment.
 #
+# $@ Optional list of zedlets to remove from our test zed.d directory.
 function zed_cleanup
 {
 	if ! is_linux; then
 		return
 	fi
+	EXTRA_ZEDLETS=$@
 
 	log_must rm -f ${ZEDLET_DIR}/zed.rc
 	log_must rm -f ${ZEDLET_DIR}/zed-functions.sh
 	log_must rm -f ${ZEDLET_DIR}/all-syslog.sh
 	log_must rm -f ${ZEDLET_DIR}/all-debug.sh
 	log_must rm -f ${ZEDLET_DIR}/state
+
+	if [[ -n "$EXTRA_ZEDLETS" ]] ; then
+		for i in $EXTRA_ZEDLETS ; do
+			log_must rm -f ${ZEDLET_DIR}/$i
+		done
+	fi
 	log_must rm -f $ZED_LOG
 	log_must rm -f $ZED_DEBUG_LOG
 	log_must rm -f $VDEVID_CONF_ETC
@@ -3139,7 +3175,7 @@ function zed_start
 	# run ZED in the background and redirect foreground logging
 	# output to $ZED_LOG.
 	log_must truncate -s 0 $ZED_DEBUG_LOG
-	log_must eval "zed -vF -d $ZEDLET_DIR -p $ZEDLET_DIR/zed.pid" \
+	log_must eval "zed -vF -d $ZEDLET_DIR -p $ZEDLET_DIR/zed.pid -P $PATH" \
 	    "-s $ZEDLET_DIR/state 2>$ZED_LOG &"
 
 	return 0

tests/zfs-tests/tests/functional/events/cleanup.ksh

@@ -26,6 +26,6 @@
 
 . $STF_SUITE/include/libtest.shlib
 
-zed_cleanup
+zed_cleanup all-debug.sh all-syslog.sh
 
 default_cleanup

tests/zfs-tests/tests/functional/events/setup.ksh

@@ -28,6 +28,6 @@
 
 DISK=${DISKS%% *}
 
-zed_setup
+zed_setup all-debug.sh all-syslog.sh
 
 default_setup $DISK

tests/zfs-tests/tests/functional/fault/Makefile.am

@@ -8,4 +8,5 @@ dist_pkgdata_SCRIPTS = \
 	auto_spare_001_pos.ksh \
 	auto_spare_002_pos.ksh \
 	auto_spare_ashift.ksh \
-	auto_spare_multiple.ksh
+	auto_spare_multiple.ksh \
+	scrub_after_resilver.ksh

tests/zfs-tests/tests/functional/fault/cleanup.ksh

@@ -31,6 +31,6 @@ verify_runnable "global"
 cleanup_devices $DISKS
 
 zed_stop
-zed_cleanup
+zed_cleanup resilver_finish-start-scrub.sh
 
 log_pass

tests/zfs-tests/tests/functional/fault/scrub_after_resilver.ksh

@@ -0,0 +1,65 @@
+#!/bin/ksh -p
+#
+# This file and its contents are supplied under the terms of the
+# Common Development and Distribution License ("CDDL"), version 1.0.
+# You may only use this file in accordance with the terms of version
+# 1.0 of the CDDL.
+#
+# A full copy of the text of the CDDL should have accompanied this
+# source.  A copy of the CDDL is also available via the Internet at
+# http://www.illumos.org/license/CDDL.
+#
+
+#
+# Copyright (c) 2018 by Lawrence Livermore National Security, LLC.
+# All rights reserved.
+#
+
+. $STF_SUITE/include/libtest.shlib
+. $STF_SUITE/tests/functional/fault/fault.cfg
+
+#
+# DESCRIPTION:
+# Test the scrub after resilver zedlet
+#
+# STRATEGY:
+# 1. Create a mirrored pool
+# 2. Fault a disk
+# 3. Replace the disk, starting a resilver
+# 4. Verify that a scrub happens after the resilver finishes
+#
+
+log_assert "Testing the scrub after resilver zedlet"
+
+# Backup our zed.rc
+zedrc_backup="$(mktemp)"
+log_must cp $ZEDLET_DIR/zed.rc $zedrc_backup
+
+# Enable ZED_SCRUB_AFTER_RESILVER
+eval "sed -i 's/\#ZED_SCRUB_AFTER_RESILVER/ZED_SCRUB_AFTER_RESILVER/g' $ZEDLET_DIR/zed.rc"
+
+function cleanup
+{
+	# Restore our zed.rc
+	log_must mv $zedrc_backup $ZEDLET_DIR/zed.rc
+	default_cleanup_noexit
+}
+
+log_onexit cleanup
+
+verify_disk_count "$DISKS" 3
+default_mirror_setup_noexit $DISK1 $DISK2
+
+log_must zpool offline -f $TESTPOOL $DISK1
+
+# Write to our degraded pool so we have some data to resilver
+log_must mkfile 16M $TESTDIR/file1
+
+# Replace the failed disks, forcing a resilver
+log_must zpool replace $TESTPOOL $DISK1 $DISK3
+
+# Wait for the resilver to finish, and then the subsequent scrub to finish.
+# Waiting for the scrub has the effect of waiting for both.  Timeout after 10
+# seconds if nothing is happening.
+log_must wait_scrubbed $TESTPOOL 10
+log_pass "Successfully ran the scrub after resilver zedlet"

tests/zfs-tests/tests/functional/fault/setup.ksh

@@ -28,7 +28,7 @@
 
 verify_runnable "global"
 
-zed_setup
+zed_setup resilver_finish-start-scrub.sh
 zed_start
 
 log_pass