diff options
19 files changed, 368 insertions, 101 deletions
diff --git a/cmd/zed/agents/zfs_agents.c b/cmd/zed/agents/zfs_agents.c index a40d59497..47e251a5e 100644 --- a/cmd/zed/agents/zfs_agents.c +++ b/cmd/zed/agents/zfs_agents.c @@ -350,19 +350,3 @@ zfs_agent_fini(void) g_zfs_hdl = NULL; } - -/* - * In ZED context, all the FMA agents run in the same thread - * and do not require a unique libzfs instance. Modules should - * use these stubs. - */ -libzfs_handle_t * -__libzfs_init(void) -{ - return (g_zfs_hdl); -} - -void -__libzfs_fini(libzfs_handle_t *hdl) -{ -} diff --git a/cmd/zed/agents/zfs_agents.h b/cmd/zed/agents/zfs_agents.h index 3c9af54c9..d1a459139 100644 --- a/cmd/zed/agents/zfs_agents.h +++ b/cmd/zed/agents/zfs_agents.h @@ -39,13 +39,6 @@ extern int zfs_slm_init(void); extern void zfs_slm_fini(void); extern void zfs_slm_event(const char *, const char *, nvlist_t *); -/* - * In ZED context, all the FMA agents run in the same thread - * and do not require a unique libzfs instance. - */ -extern libzfs_handle_t *__libzfs_init(void); -extern void __libzfs_fini(libzfs_handle_t *); - #ifdef __cplusplus } #endif diff --git a/cmd/zed/agents/zfs_diagnosis.c b/cmd/zed/agents/zfs_diagnosis.c index 49e3e1660..6f1f4d017 100644 --- a/cmd/zed/agents/zfs_diagnosis.c +++ b/cmd/zed/agents/zfs_diagnosis.c @@ -919,27 +919,27 @@ _zfs_diagnosis_init(fmd_hdl_t *hdl) { libzfs_handle_t *zhdl; - if ((zhdl = __libzfs_init()) == NULL) + if ((zhdl = libzfs_init()) == NULL) return; if ((zfs_case_pool = uu_list_pool_create("zfs_case_pool", sizeof (zfs_case_t), offsetof(zfs_case_t, zc_node), NULL, UU_LIST_POOL_DEBUG)) == NULL) { - __libzfs_fini(zhdl); + libzfs_fini(zhdl); return; } if ((zfs_cases = uu_list_create(zfs_case_pool, NULL, UU_LIST_DEBUG)) == NULL) { uu_list_pool_destroy(zfs_case_pool); - __libzfs_fini(zhdl); + libzfs_fini(zhdl); return; } if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) { uu_list_destroy(zfs_cases); uu_list_pool_destroy(zfs_case_pool); - __libzfs_fini(zhdl); + libzfs_fini(zhdl); return; } @@ -975,5 +975,5 @@ _zfs_diagnosis_fini(fmd_hdl_t *hdl) uu_list_pool_destroy(zfs_case_pool); zhdl = fmd_hdl_getspecific(hdl); - __libzfs_fini(zhdl); + libzfs_fini(zhdl); } diff --git a/cmd/zed/agents/zfs_mod.c b/cmd/zed/agents/zfs_mod.c index 5fa74d0ce..54568e75a 100644 --- a/cmd/zed/agents/zfs_mod.c +++ b/cmd/zed/agents/zfs_mod.c @@ -64,7 +64,6 @@ * trigger the FMA fault that we skipped earlier. * * ZFS on Linux porting notes: - * In lieu of a thread pool, just spawn a thread on demmand. * Linux udev provides a disk insert for both the disk and the partition * */ @@ -83,6 +82,7 @@ #include <sys/sunddi.h> #include <sys/sysevent/eventdefs.h> #include <sys/sysevent/dev.h> +#include <thread_pool.h> #include <pthread.h> #include <unistd.h> #include "zfs_agents.h" @@ -97,12 +97,12 @@ typedef void (*zfs_process_func_t)(zpool_handle_t *, nvlist_t *, boolean_t); libzfs_handle_t *g_zfshdl; list_t g_pool_list; /* list of unavailable pools at initialization */ list_t g_device_list; /* list of disks with asynchronous label request */ +tpool_t *g_tpool; boolean_t g_enumeration_done; -pthread_t g_zfs_tid; +pthread_t g_zfs_tid; /* zfs_enum_pools() thread */ typedef struct unavailpool { zpool_handle_t *uap_zhp; - pthread_t uap_enable_tid; /* dataset enable thread if activated */ list_node_t uap_node; } unavailpool_t; @@ -135,7 +135,6 @@ zfs_unavail_pool(zpool_handle_t *zhp, void *data) unavailpool_t *uap; uap = malloc(sizeof (unavailpool_t)); uap->uap_zhp = zhp; - uap->uap_enable_tid = 0; list_insert_tail((list_t *)data, uap); } else { zpool_close(zhp); @@ -512,19 +511,14 @@ zfs_iter_vdev(zpool_handle_t *zhp, nvlist_t *nvl, void *data) (dp->dd_func)(zhp, nvl, dp->dd_islabeled); } -static void * +void zfs_enable_ds(void *arg) { unavailpool_t *pool = (unavailpool_t *)arg; - assert(pool->uap_enable_tid = pthread_self()); - (void) zpool_enable_datasets(pool->uap_zhp, NULL, 0); zpool_close(pool->uap_zhp); - pool->uap_zhp = NULL; - - /* Note: zfs_slm_fini() will cleanup this pool entry on exit */ - return (NULL); + free(pool); } static int @@ -559,15 +553,13 @@ zfs_iter_pool(zpool_handle_t *zhp, void *data) for (pool = list_head(&g_pool_list); pool != NULL; pool = list_next(&g_pool_list, pool)) { - if (pool->uap_enable_tid != 0) - continue; /* entry already processed */ if (strcmp(zpool_get_name(zhp), zpool_get_name(pool->uap_zhp))) continue; if (zfs_toplevel_state(zhp) >= VDEV_STATE_DEGRADED) { - /* send to a background thread; keep on list */ - (void) pthread_create(&pool->uap_enable_tid, - NULL, zfs_enable_ds, pool); + list_remove(&g_pool_list, pool); + (void) tpool_dispatch(g_tpool, zfs_enable_ds, + pool); break; } } @@ -857,7 +849,7 @@ zfs_enum_pools(void *arg) int zfs_slm_init() { - if ((g_zfshdl = __libzfs_init()) == NULL) + if ((g_zfshdl = libzfs_init()) == NULL) return (-1); /* @@ -869,7 +861,7 @@ zfs_slm_init() if (pthread_create(&g_zfs_tid, NULL, zfs_enum_pools, NULL) != 0) { list_destroy(&g_pool_list); - __libzfs_fini(g_zfshdl); + libzfs_fini(g_zfshdl); return (-1); } @@ -887,19 +879,15 @@ zfs_slm_fini() /* wait for zfs_enum_pools thread to complete */ (void) pthread_join(g_zfs_tid, NULL); + /* destroy the thread pool */ + if (g_tpool != NULL) { + tpool_wait(g_tpool); + tpool_destroy(g_tpool); + } while ((pool = (list_head(&g_pool_list))) != NULL) { - /* - * each pool entry has two possibilities - * 1. was made available (so wait for zfs_enable_ds thread) - * 2. still unavailable (just close the pool) - */ - if (pool->uap_enable_tid) - (void) pthread_join(pool->uap_enable_tid, NULL); - else if (pool->uap_zhp != NULL) - zpool_close(pool->uap_zhp); - list_remove(&g_pool_list, pool); + zpool_close(pool->uap_zhp); free(pool); } list_destroy(&g_pool_list); @@ -910,7 +898,7 @@ zfs_slm_fini() } list_destroy(&g_device_list); - __libzfs_fini(g_zfshdl); + libzfs_fini(g_zfshdl); } void diff --git a/cmd/zed/agents/zfs_retire.c b/cmd/zed/agents/zfs_retire.c index f69c583f0..5a090e32f 100644 --- a/cmd/zed/agents/zfs_retire.c +++ b/cmd/zed/agents/zfs_retire.c @@ -176,6 +176,8 @@ replace_with_spare(fmd_hdl_t *hdl, zpool_handle_t *zhp, nvlist_t *vdev) nvlist_t **spares; uint_t s, nspares; char *dev_name; + zprop_source_t source; + int ashift; config = zpool_get_config(zhp, NULL); if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, @@ -189,6 +191,11 @@ replace_with_spare(fmd_hdl_t *hdl, zpool_handle_t *zhp, nvlist_t *vdev) &spares, &nspares) != 0) return; + /* + * lookup "ashift" pool property, we may need it for the replacement + */ + ashift = zpool_get_prop_int(zhp, ZPOOL_PROP_ASHIFT, &source); + replacement = fmd_nvl_alloc(hdl, FMD_SLEEP); (void) nvlist_add_string(replacement, ZPOOL_CONFIG_TYPE, @@ -207,6 +214,11 @@ replace_with_spare(fmd_hdl_t *hdl, zpool_handle_t *zhp, nvlist_t *vdev) &spare_name) != 0) continue; + /* if set, add the "ashift" pool property to the spare nvlist */ + if (source != ZPROP_SRC_DEFAULT) + (void) nvlist_add_uint64(spares[s], + ZPOOL_CONFIG_ASHIFT, ashift); + (void) nvlist_add_nvlist_array(replacement, ZPOOL_CONFIG_CHILDREN, &spares[s], 1); @@ -483,7 +495,7 @@ _zfs_retire_init(fmd_hdl_t *hdl) zfs_retire_data_t *zdp; libzfs_handle_t *zhdl; - if ((zhdl = __libzfs_init()) == NULL) + if ((zhdl = libzfs_init()) == NULL) return; if (fmd_hdl_register(hdl, FMD_API_VERSION, &fmd_info) != 0) { @@ -504,7 +516,7 @@ _zfs_retire_fini(fmd_hdl_t *hdl) if (zdp != NULL) { zfs_retire_clear_data(hdl, zdp); - __libzfs_fini(zdp->zrd_hdl); + libzfs_fini(zdp->zrd_hdl); fmd_hdl_free(hdl, zdp, sizeof (zfs_retire_data_t)); } } diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run index 9424c80a0..0c8f298c6 100644 --- a/tests/runfiles/linux.run +++ b/tests/runfiles/linux.run @@ -464,7 +464,7 @@ tags = ['functional', 'exec'] [tests/functional/fault] tests = ['auto_online_001_pos', 'auto_replace_001_pos', 'auto_spare_001_pos', - 'auto_spare_002_pos.ksh'] + 'auto_spare_002_pos', 'auto_spare_ashift', 'auto_spare_multiple'] tags = ['functional', 'fault'] [tests/functional/features/async_destroy] diff --git a/tests/zfs-tests/include/blkdev.shlib b/tests/zfs-tests/include/blkdev.shlib index 876c84356..28ac1052c 100644 --- a/tests/zfs-tests/include/blkdev.shlib +++ b/tests/zfs-tests/include/blkdev.shlib @@ -353,16 +353,35 @@ function insert_disk #disk scsi_host # # Load scsi_debug module with specified parameters +# $blksz can be either one of: < 512b | 512e | 4Kn > # -function load_scsi_debug # dev_size_mb add_host num_tgts max_luns +function load_scsi_debug # dev_size_mb add_host num_tgts max_luns blksz { typeset devsize=$1 typeset hosts=$2 typeset tgts=$3 typeset luns=$4 + typeset blksz=$5 [[ -z $devsize ]] || [[ -z $hosts ]] || [[ -z $tgts ]] || \ - [[ -z $luns ]] && log_fail "Arguments invalid or missing" + [[ -z $luns ]] || [[ -z $blksz ]] && \ + log_fail "Arguments invalid or missing" + + case "$5" in + '512b') + typeset sector=512 + typeset blkexp=0 + ;; + '512e') + typeset sector=512 + typeset blkexp=3 + ;; + '4Kn') + typeset sector=4096 + typeset blkexp=0 + ;; + *) log_fail "Unsupported blksz value: $5" ;; + esac if is_linux; then modprobe -n scsi_debug @@ -375,7 +394,8 @@ function load_scsi_debug # dev_size_mb add_host num_tgts max_luns log_fail "scsi_debug module already installed" else log_must modprobe scsi_debug dev_size_mb=$devsize \ - add_host=$hosts num_tgts=$tgts max_luns=$luns + add_host=$hosts num_tgts=$tgts max_luns=$luns \ + sector_size=$sector physblk_exp=$blkexp block_device_wait lsscsi | egrep scsi_debug > /dev/null if (($? == 1)); then @@ -386,6 +406,16 @@ function load_scsi_debug # dev_size_mb add_host num_tgts max_luns } # +# Unload scsi_debug module, if needed. +# +function unload_scsi_debug +{ + if lsmod | grep scsi_debug >/dev/null; then + log_must modprobe -r scsi_debug + fi +} + +# # Get scsi_debug device name. # Returns basename of scsi_debug device (for example "sdb"). # diff --git a/tests/zfs-tests/include/libtest.shlib b/tests/zfs-tests/include/libtest.shlib index 6b1658c1b..cc9fddd41 100644 --- a/tests/zfs-tests/include/libtest.shlib +++ b/tests/zfs-tests/include/libtest.shlib @@ -3158,14 +3158,26 @@ function zed_stop if [[ -f ${ZEDLET_DIR}/zed.pid ]]; then zedpid=$(cat ${ZEDLET_DIR}/zed.pid) kill $zedpid - wait $zedpid + while ps -p $zedpid > /dev/null; do + sleep 1 + done rm -f ${ZEDLET_DIR}/zed.pid fi - return 0 } # +# Drain all zevents +# +function zed_events_drain +{ + while [ $(zpool events -H | wc -l) -ne 0 ]; do + sleep 1 + zpool events -c >/dev/null + done +} + +# # Check is provided device is being active used as a swap device. # function is_swap_inuse diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_reopen/cleanup.ksh b/tests/zfs-tests/tests/functional/cli_root/zpool_reopen/cleanup.ksh index 4477e5402..99c51351c 100755 --- a/tests/zfs-tests/tests/functional/cli_root/zpool_reopen/cleanup.ksh +++ b/tests/zfs-tests/tests/functional/cli_root/zpool_reopen/cleanup.ksh @@ -27,7 +27,7 @@ if is_linux; then for SDDEVICE in $(get_debug_device); do unplug $SDDEVICE done - modprobe -r scsi_debug + unload_scsi_debug fi log_pass diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_reopen/setup.ksh b/tests/zfs-tests/tests/functional/cli_root/zpool_reopen/setup.ksh index 4dbf8965d..59b8764ce 100755 --- a/tests/zfs-tests/tests/functional/cli_root/zpool_reopen/setup.ksh +++ b/tests/zfs-tests/tests/functional/cli_root/zpool_reopen/setup.ksh @@ -22,7 +22,7 @@ verify_runnable "global" # Create scsi_debug devices for the reopen tests if is_linux; then - load_scsi_debug $SDSIZE $SDHOSTS $SDTGTS $SDLUNS + load_scsi_debug $SDSIZE $SDHOSTS $SDTGTS $SDLUNS '512b' else log_unsupported "scsi debug module unsupported" fi diff --git a/tests/zfs-tests/tests/functional/fault/Makefile.am b/tests/zfs-tests/tests/functional/fault/Makefile.am index 436f3e8be..ef4380835 100644 --- a/tests/zfs-tests/tests/functional/fault/Makefile.am +++ b/tests/zfs-tests/tests/functional/fault/Makefile.am @@ -6,4 +6,6 @@ dist_pkgdata_SCRIPTS = \ auto_online_001_pos.ksh \ auto_replace_001_pos.ksh \ auto_spare_001_pos.ksh \ - auto_spare_002_pos.ksh + auto_spare_002_pos.ksh \ + auto_spare_ashift.ksh \ + auto_spare_multiple.ksh diff --git a/tests/zfs-tests/tests/functional/fault/auto_online_001_pos.ksh b/tests/zfs-tests/tests/functional/fault/auto_online_001_pos.ksh index 0f6e38ac2..beff5dc87 100755 --- a/tests/zfs-tests/tests/functional/fault/auto_online_001_pos.ksh +++ b/tests/zfs-tests/tests/functional/fault/auto_online_001_pos.ksh @@ -54,9 +54,8 @@ fi function cleanup { - #online last disk before fail - insert_disk $offline_disk $host - poolexists $TESTPOOL && destroy_pool $TESTPOOL + destroy_pool $TESTPOOL + unload_scsi_debug } log_assert "Testing automated auto-online FMA test" @@ -65,8 +64,8 @@ log_onexit cleanup # If using the default loop devices, need a scsi_debug device for auto-online if is_loop_device $DISK1; then - SD=$(lsscsi | nawk '/scsi_debug/ {print $6; exit}') - SDDEVICE=$(echo $SD | nawk -F / '{print $3}') + load_scsi_debug $SDSIZE $SDHOSTS $SDTGTS $SDLUNS '512b' + SDDEVICE=$(get_debug_device) SDDEVICE_ID=$(get_persistent_disk_name $SDDEVICE) autoonline_disks="$SDDEVICE" else diff --git a/tests/zfs-tests/tests/functional/fault/auto_replace_001_pos.ksh b/tests/zfs-tests/tests/functional/fault/auto_replace_001_pos.ksh index 40a680a9b..8e48b2ab4 100755 --- a/tests/zfs-tests/tests/functional/fault/auto_replace_001_pos.ksh +++ b/tests/zfs-tests/tests/functional/fault/auto_replace_001_pos.ksh @@ -57,27 +57,23 @@ fi function setup { - lsmod | egrep scsi_debug > /dev/null - if (($? == 1)); then - load_scsi_debug $SDSIZE $SDHOSTS $SDTGTS $SDLUNS - fi + load_scsi_debug $SDSIZE $SDHOSTS $SDTGTS $SDLUNS '512b' + SD=$(get_debug_device) + SDDEVICE_ID=$(get_persistent_disk_name $SD) # Register vdev_id alias rule for scsi_debug device to create a # persistent path - SD=$(lsscsi | nawk '/scsi_debug/ {print $6; exit}' \ - | nawk -F / '{print $3}') - SDDEVICE_ID=$(get_persistent_disk_name $SD) log_must eval "echo "alias scsidebug /dev/disk/by-id/$SDDEVICE_ID" \ >> $VDEVID_CONF" block_device_wait - - SDDEVICE=$(udevadm info -q all -n $DEV_DSKDIR/$SD | egrep ID_VDEV \ - | nawk '{print $2; exit}' | nawk -F = '{print $2; exit}') + SDDEVICE=$(udevadm info -q all -n $DEV_DSKDIR/$SD \ + | awk -F'=' '/ID_VDEV=/{print $2; exit}') [[ -z $SDDEVICE ]] && log_fail "vdev rule was not registered properly" } function cleanup { - poolexists $TESTPOOL && destroy_pool $TESTPOOL + destroy_pool $TESTPOOL + unload_scsi_debug } log_assert "Testing automated auto-replace FMA test" @@ -112,7 +108,7 @@ log_must zpool export -F $TESTPOOL # Offline disk remove_disk $SD block_device_wait -log_must modprobe -r scsi_debug +unload_scsi_debug # Reimport pool with drive missing log_must zpool import $TESTPOOL diff --git a/tests/zfs-tests/tests/functional/fault/auto_spare_001_pos.ksh b/tests/zfs-tests/tests/functional/fault/auto_spare_001_pos.ksh index 82f7f4834..b6af1a3f4 100755 --- a/tests/zfs-tests/tests/functional/fault/auto_spare_001_pos.ksh +++ b/tests/zfs-tests/tests/functional/fault/auto_spare_001_pos.ksh @@ -42,7 +42,7 @@ verify_runnable "both" function cleanup { log_must zinject -c all - poolexists $TESTPOOL && destroy_pool $TESTPOOL + destroy_pool $TESTPOOL rm -f $VDEV_FILES $SPARE_FILE } @@ -50,6 +50,9 @@ log_assert "Testing automated auto-spare FMA test" log_onexit cleanup +# Clear events from previous runs +zed_events_drain + TESTFILE="/$TESTPOOL/$TESTFS/testfile" for type in "mirror" "raidz" "raidz2"; do diff --git a/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh b/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh index f0ddac35c..cfa748d30 100755 --- a/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh +++ b/tests/zfs-tests/tests/functional/fault/auto_spare_002_pos.ksh @@ -42,7 +42,7 @@ verify_runnable "both" function cleanup { log_must zinject -c all - poolexists $TESTPOOL && destroy_pool $TESTPOOL + destroy_pool $TESTPOOL rm -f $VDEV_FILES $SPARE_FILE } @@ -50,6 +50,9 @@ log_assert "Testing automated auto-spare FMA test" log_onexit cleanup +# Clear events from previous runs +zed_events_drain + TESTFILE="/$TESTPOOL/$TESTFS/testfile" for type in "mirror" "raidz" "raidz2"; do @@ -65,8 +68,14 @@ for type in "mirror" "raidz" "raidz2"; do log_must dd if=/dev/urandom of=$TESTFILE bs=1M count=16 # 4. Inject CHECKSUM ERRORS on read with a zinject error handler + # NOTE: checksum events are ratelimited to max 5 per second, ZED needs + # 10 to kick in a spare log_must zinject -d $FAULT_FILE -e corrupt -f 50 -T read $TESTPOOL log_must cp $TESTFILE /dev/null + log_must sleep 1 + log_must cp $TESTFILE /dev/null + log_must sleep 1 + log_must cp $TESTFILE /dev/null # 5. Verify the ZED kicks in a hot spare and expected pool/device status log_note "Wait for ZED to auto-spare" diff --git a/tests/zfs-tests/tests/functional/fault/auto_spare_ashift.ksh b/tests/zfs-tests/tests/functional/fault/auto_spare_ashift.ksh new file mode 100755 index 000000000..e9857518e --- /dev/null +++ b/tests/zfs-tests/tests/functional/fault/auto_spare_ashift.ksh @@ -0,0 +1,101 @@ +#!/bin/ksh -p + +# +# CDDL HEADER START +# +# 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. +# +# CDDL HEADER END +# + +# +# Copyright (c) 2017 by Intel Corporation. All rights reserved. +# Copyright 2017, loli10K <[email protected]>. All rights reserved. +# + +. $STF_SUITE/include/libtest.shlib +. $STF_SUITE/include/math.shlib +. $STF_SUITE/tests/functional/fault/fault.cfg + +# +# DESCRIPTION: +# Testing Fault Management Agent ZED Logic - Automated Auto-Spare Test when +# drive is faulted and a custom ashift value needs to be provided to replace it. +# +# STRATEGY: +# 1. Create a pool from 512b devices and set "ashift" pool property accordingly +# 2. Add one 512e spare device (4Kn would generate IO errors on replace) +# 3. Inject IO errors with a zinject error handler +# 4. Start a scrub +# 5. Verify the ZED kicks in the hot spare and expected pool/device status +# 6. Clear the fault +# 7. Verify the hot spare is available and expected pool/device status +# + +verify_runnable "both" + +function cleanup +{ + log_must zinject -c all + destroy_pool $TESTPOOL + unload_scsi_debug + rm -f $SAFE_DEVICE $FAIL_DEVICE +} + +log_assert "ZED should replace a device using the configured ashift property" +log_onexit cleanup + +# Clear events from previous runs +zed_events_drain + +SAFE_DEVICE="$TEST_BASE_DIR/safe-dev" +FAIL_DEVICE="$TEST_BASE_DIR/fail-dev" + +# 1. Create a pool from 512b devices and set "ashift" pool property accordingly +for vdev in $SAFE_DEVICE $FAIL_DEVICE; do + truncate -s $SPA_MINDEVSIZE $vdev +done +log_must zpool create -f $TESTPOOL mirror $SAFE_DEVICE $FAIL_DEVICE +# NOTE: file VDEVs should be added as 512b devices, verify this "just in case" +for vdev in $SAFE_DEVICE $FAIL_DEVICE; do + verify_eq "9" "$(zdb -e -l $vdev | awk '/ashift: /{print $2}')" "ashift" +done +log_must zpool set ashift=9 $TESTPOOL + +# 2. Add one 512e spare device (4Kn would generate IO errors on replace) +# NOTE: must be larger than the existing 512b devices, add 32m of fudge +load_scsi_debug $(($SPA_MINDEVSIZE/1024/1024+32)) $SDHOSTS $SDTGTS $SDLUNS '512e' +SPARE_DEVICE=$(get_debug_device) +log_must_busy zpool add $TESTPOOL spare $SPARE_DEVICE + +# 3. Inject IO errors with a zinject error handler +log_must zinject -d $FAIL_DEVICE -e io -T all -f 100 $TESTPOOL + +# 4. Start a scrub +log_must zpool scrub $TESTPOOL + +# 5. Verify the ZED kicks in a hot spare and expected pool/device status +log_note "Wait for ZED to auto-spare" +log_must wait_vdev_state $TESTPOOL $FAIL_DEVICE "FAULTED" 60 +log_must wait_vdev_state $TESTPOOL $SPARE_DEVICE "ONLINE" 60 +log_must wait_hotspare_state $TESTPOOL $SPARE_DEVICE "INUSE" +log_must check_state $TESTPOOL "" "DEGRADED" + +# 6. Clear the fault +log_must zinject -c all +log_must zpool clear $TESTPOOL $FAIL_DEVICE + +# 7. Verify the hot spare is available and expected pool/device status +log_must wait_vdev_state $TESTPOOL $FAIL_DEVICE "ONLINE" 60 +log_must wait_hotspare_state $TESTPOOL $SPARE_DEVICE "AVAIL" +log_must is_pool_resilvered $TESTPOOL +log_must check_state $TESTPOOL "" "ONLINE" + +log_pass "ZED successfully replaces a device using the configured ashift property" diff --git a/tests/zfs-tests/tests/functional/fault/auto_spare_multiple.ksh b/tests/zfs-tests/tests/functional/fault/auto_spare_multiple.ksh new file mode 100755 index 000000000..8650ceff7 --- /dev/null +++ b/tests/zfs-tests/tests/functional/fault/auto_spare_multiple.ksh @@ -0,0 +1,152 @@ +#!/bin/ksh -p + +# +# CDDL HEADER START +# +# 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. +# +# CDDL HEADER END +# + +# +# Copyright (c) 2017 by Intel Corporation. All rights reserved. +# Copyright 2017, loli10K <[email protected]>. All rights reserved. +# + +. $STF_SUITE/include/libtest.shlib +. $STF_SUITE/tests/functional/fault/fault.cfg + +# +# DESCRIPTION: +# Testing Fault Management Agent ZED Logic - Automated Auto-Spare Test when +# multiple drives are faulted. +# +# STRATEGY: +# 1. Create a pool with two hot spares +# 2. Inject IO ERRORS with a zinject error handler on the first device +# 3. Start a scrub +# 4. Verify the ZED kicks in a hot spare and expected pool/device status +# 5. Inject IO ERRORS on a second device +# 6. Start a scrub +# 7. Verify the ZED kicks in a second hot spare +# 8. Clear the fault on both devices +# 9. Verify the hot spares are available and expected pool/device status +# 10. Rinse and repeat, this time faulting both devices at the same time +# + +verify_runnable "both" + +function cleanup +{ + log_must zinject -c all + destroy_pool $TESTPOOL + rm -f $DATA_DEVS $SPARE_DEVS +} + +log_assert "ZED should be able to handle multiple faulted devices" +log_onexit cleanup + +# Clear events from previous runs +zed_events_drain + +FAULT_DEV1="$TEST_BASE_DIR/fault-dev1" +FAULT_DEV2="$TEST_BASE_DIR/fault-dev2" +SAFE_DEV1="$TEST_BASE_DIR/safe-dev1" +SAFE_DEV2="$TEST_BASE_DIR/safe-dev2" +DATA_DEVS="$FAULT_DEV1 $FAULT_DEV2 $SAFE_DEV1 $SAFE_DEV2" +SPARE_DEV1="$TEST_BASE_DIR/spare-dev1" +SPARE_DEV2="$TEST_BASE_DIR/spare-dev2" +SPARE_DEVS="$SPARE_DEV1 $SPARE_DEV2" + +for type in "mirror" "raidz" "raidz2" "raidz3"; do + # 1. Create a pool with two hot spares + truncate -s $SPA_MINDEVSIZE $DATA_DEVS $SPARE_DEVS + log_must zpool create -f $TESTPOOL $type $DATA_DEVS spare $SPARE_DEVS + + # 2. Inject IO ERRORS with a zinject error handler on the first device + log_must zinject -d $FAULT_DEV1 -e io -T all -f 100 $TESTPOOL + + # 3. Start a scrub + log_must zpool scrub $TESTPOOL + + # 4. Verify the ZED kicks in a hot spare and expected pool/device status + log_note "Wait for ZED to auto-spare" + log_must wait_vdev_state $TESTPOOL $FAULT_DEV1 "FAULTED" 60 + log_must wait_vdev_state $TESTPOOL $SPARE_DEV1 "ONLINE" 60 + log_must wait_hotspare_state $TESTPOOL $SPARE_DEV1 "INUSE" + log_must check_state $TESTPOOL "" "DEGRADED" + + # 5. Inject IO ERRORS on a second device + log_must zinject -d $FAULT_DEV2 -e io -T all -f 100 $TESTPOOL + + # 6. Start a scrub + while is_pool_scrubbing $TESTPOOL || is_pool_resilvering $TESTPOOL; do + sleep 1 + done + log_must zpool scrub $TESTPOOL + + # 7. Verify the ZED kicks in a second hot spare + log_note "Wait for ZED to auto-spare" + log_must wait_vdev_state $TESTPOOL $FAULT_DEV2 "FAULTED" 60 + log_must wait_vdev_state $TESTPOOL $SPARE_DEV2 "ONLINE" 60 + log_must wait_hotspare_state $TESTPOOL $SPARE_DEV2 "INUSE" + log_must check_state $TESTPOOL "" "DEGRADED" + + # 8. Clear the fault on both devices + log_must zinject -c all + log_must zpool clear $TESTPOOL $FAULT_DEV1 + log_must zpool clear $TESTPOOL $FAULT_DEV2 + + # 9. Verify the hot spares are available and expected pool/device status + log_must wait_vdev_state $TESTPOOL $FAULT_DEV1 "ONLINE" 60 + log_must wait_vdev_state $TESTPOOL $FAULT_DEV2 "ONLINE" 60 + log_must wait_hotspare_state $TESTPOOL $SPARE_DEV1 "AVAIL" + log_must wait_hotspare_state $TESTPOOL $SPARE_DEV2 "AVAIL" + log_must check_state $TESTPOOL "" "ONLINE" + + # Cleanup + cleanup +done + +# Rinse and repeat, this time faulting both devices at the same time +# NOTE: "raidz" is exluded since it cannot survive 2 faulted devices +# NOTE: "mirror" is a 4-way mirror here and should survive this test +for type in "mirror" "raidz2" "raidz3"; do + # 1. Create a pool with two hot spares + truncate -s $SPA_MINDEVSIZE $DATA_DEVS $SPARE_DEVS + log_must zpool create -f $TESTPOOL $type $DATA_DEVS spare $SPARE_DEVS + + # 2. Inject IO ERRORS with a zinject error handler on two devices + log_must eval "zinject -d $FAULT_DEV1 -e io -T all -f 100 $TESTPOOL &" + log_must eval "zinject -d $FAULT_DEV2 -e io -T all -f 100 $TESTPOOL &" + + # 3. Start a scrub + log_must zpool scrub $TESTPOOL + + # 4. Verify the ZED kicks in two hot spares and expected pool/device status + log_note "Wait for ZED to auto-spare" + log_must wait_vdev_state $TESTPOOL $FAULT_DEV1 "FAULTED" 60 + log_must wait_vdev_state $TESTPOOL $FAULT_DEV2 "FAULTED" 60 + log_must wait_vdev_state $TESTPOOL $SPARE_DEV1 "ONLINE" 60 + log_must wait_vdev_state $TESTPOOL $SPARE_DEV2 "ONLINE" 60 + log_must wait_hotspare_state $TESTPOOL $SPARE_DEV1 "INUSE" + log_must wait_hotspare_state $TESTPOOL $SPARE_DEV2 "INUSE" + log_must check_state $TESTPOOL "" "DEGRADED" + + # 5. Clear the fault on both devices + log_must zinject -c all + log_must zpool clear $TESTPOOL $FAULT_DEV1 + log_must zpool clear $TESTPOOL $FAULT_DEV2 + + # Cleanup + cleanup +done + +log_pass "ZED successfully handles multiple faulted devices" diff --git a/tests/zfs-tests/tests/functional/fault/cleanup.ksh b/tests/zfs-tests/tests/functional/fault/cleanup.ksh index 82e379b0d..9d354f30e 100755 --- a/tests/zfs-tests/tests/functional/fault/cleanup.ksh +++ b/tests/zfs-tests/tests/functional/fault/cleanup.ksh @@ -33,14 +33,4 @@ cleanup_devices $DISKS zed_stop zed_cleanup -SDDEVICE=$(get_debug_device) - -# Offline disk and remove scsi_debug module -if is_linux; then - if [ -n "$SDDEVICE" ]; then - remove_disk $SDDEVICE - fi - modprobe -r scsi_debug -fi - log_pass diff --git a/tests/zfs-tests/tests/functional/fault/setup.ksh b/tests/zfs-tests/tests/functional/fault/setup.ksh index 3d54d4f21..3d3cbc9e5 100755 --- a/tests/zfs-tests/tests/functional/fault/setup.ksh +++ b/tests/zfs-tests/tests/functional/fault/setup.ksh @@ -31,8 +31,4 @@ verify_runnable "global" zed_setup zed_start -# Create a scsi_debug device to be used with auto-online (if using loop devices) -# and auto-replace regardless of other devices -load_scsi_debug $SDSIZE $SDHOSTS $SDTGTS $SDLUNS - log_pass |