1#!/bin/ksh -p 2# SPDX-License-Identifier: CDDL-1.0 3# 4# CDDL HEADER START 5# 6# The contents of this file are subject to the terms of the 7# Common Development and Distribution License (the "License"). 8# You may not use this file except in compliance with the License. 9# 10# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 11# or https://opensource.org/licenses/CDDL-1.0. 12# See the License for the specific language governing permissions 13# and limitations under the License. 14# 15# When distributing Covered Code, include this CDDL HEADER in each 16# file and include the License file at usr/src/OPENSOLARIS.LICENSE. 17# If applicable, add the following below this CDDL HEADER, with the 18# fields enclosed by brackets "[]" replaced with your own identifying 19# information: Portions Copyright [yyyy] [name of copyright owner] 20# 21# CDDL HEADER END 22 23# 24# Copyright (c) 2018 by Lawrence Livermore National Security, LLC. 25# 26 27# 28# DESCRIPTION: 29# Test /proc/spl/kstat/zfs/<pool>/state kstat 30# 31# STRATEGY: 32# 1. Create a mirrored pool 33# 2. Check that pool is ONLINE 34# 3. Fault one disk 35# 4. Check that pool is DEGRADED 36# 5. Create a new pool with a single scsi_debug disk 37# 6. Remove the disk 38# 7. Check that pool is SUSPENDED 39# 8. Add the disk back in 40# 9. Clear errors and destroy the pools 41 42. $STF_SUITE/include/libtest.shlib 43 44verify_runnable "both" 45 46function cleanup 47{ 48 # Destroy the scsi_debug pool 49 if [ -n "$TESTPOOL2" ] ; then 50 if [ -n "$host" ] ; then 51 # Re-enable the disk 52 scan_scsi_hosts $host 53 54 # Device may have changed names after being inserted 55 SDISK=$(get_debug_device) 56 log_must ln $DEV_RDSKDIR/$SDISK $REALDISK 57 fi 58 59 # Restore our working pool image 60 if [ -n "$BACKUP" ] ; then 61 gunzip -c $BACKUP > $REALDISK 62 log_must rm -f $BACKUP 63 fi 64 65 if poolexists $TESTPOOL2 ; then 66 # Our disk is back. Now we can clear errors and destroy the 67 # pool cleanly. 68 log_must zpool clear $TESTPOOL2 69 70 # Now that the disk is back and errors cleared, wait for our 71 # hung 'zpool scrub' to finish. 72 wait 73 74 destroy_pool $TESTPOOL2 75 fi 76 log_must rm -f $REALDISK 77 unload_scsi_debug 78 fi 79} 80 81# Check that our pool state values match what's expected 82# 83# $1: pool name 84# $2: expected state ("ONLINE", "DEGRADED", "SUSPENDED", etc) 85function check_all 86{ 87 pool=$1 88 expected=$2 89 90 state1=$(zpool status $pool | awk '/state: /{print $2}'); 91 state2=$(zpool list -H -o health $pool) 92 state3=$(</proc/spl/kstat/zfs/$pool/state) 93 log_note "Checking $expected = $state1 = $state2 = $state3" 94 if [[ "$expected" == "$state1" && "$expected" == "$state2" && \ 95 "$expected" == "$state3" ]] ; then 96 true 97 else 98 false 99 fi 100} 101 102log_onexit cleanup 103 104log_assert "Testing /proc/spl/kstat/zfs/<pool>/state kstat" 105 106# Test that the initial pool is healthy 107check_all $TESTPOOL "ONLINE" 108 109# Fault one of the disks, and check that pool is degraded 110read -r DISK1 _ <<<"$DISKS" 111log_must zpool offline -tf $TESTPOOL $DISK1 112check_all $TESTPOOL "DEGRADED" 113log_must zpool online $TESTPOOL $DISK1 114log_must zpool clear $TESTPOOL 115 116# Create a new pool out of a scsi_debug disk 117TESTPOOL2=testpool2 118MINVDEVSIZE_MB=$((MINVDEVSIZE / 1048576)) 119load_scsi_debug $MINVDEVSIZE_MB 1 1 1 '512b' 120 121SDISK=$(get_debug_device) 122host=$(get_scsi_host $SDISK) 123 124# Use $REALDISK instead of $SDISK in our pool because $SDISK can change names 125# as we remove/add the disk (i.e. /dev/sdf -> /dev/sdg). 126REALDISK=/dev/kstat-state-realdisk 127log_must [ ! -e $REALDISK ] 128ln $DEV_RDSKDIR/$SDISK $REALDISK 129 130log_must zpool create $TESTPOOL2 $REALDISK 131 132# Backup the contents of the disk image 133BACKUP=$TEST_BASE_DIR/kstat-state-realdisk.gz 134log_must [ ! -e $BACKUP ] 135gzip -c $REALDISK > $BACKUP 136 137# Yank out the disk from under the pool 138log_must rm $REALDISK 139remove_disk $SDISK 140 141# Run a 'zpool scrub' in the background to suspend the pool. We run it in the 142# background since the command will hang when the pool gets suspended. The 143# command will resume and exit after we restore the missing disk later on. 144zpool scrub $TESTPOOL2 & 145# Once we trigger the zpool scrub, all zpool/zfs command gets stuck for 180 seconds. 146# Post 180 seconds zpool/zfs commands gets start executing however few more seconds(10s) 147# it take to update the status. 148# hence sleeping for 200 seconds so that we get the correct status. 149sleep 200 # Give the scrub some time to run before we check if it fails 150 151log_must check_all $TESTPOOL2 "SUSPENDED" 152 153log_pass "/proc/spl/kstat/zfs/<pool>/state test successful" 154