diff options
Diffstat (limited to 'usr/src')
| -rw-r--r-- | usr/src/cmd/fm/modules/common/zfs-diagnosis/zfs_de.c | 122 | ||||
| -rw-r--r-- | usr/src/cmd/fm/modules/common/zfs-retire/zfs_retire.c | 24 | ||||
| -rw-r--r-- | usr/src/lib/libzfs/common/libzfs_pool.c | 11 | ||||
| -rw-r--r-- | usr/src/pkg/manifests/system-test-zfstest.mf | 3 | ||||
| -rw-r--r-- | usr/src/test/zfs-tests/include/libtest.shlib | 7 | ||||
| -rw-r--r-- | usr/src/test/zfs-tests/runfiles/delphix.run | 3 | ||||
| -rw-r--r-- | usr/src/test/zfs-tests/runfiles/omnios.run | 3 | ||||
| -rw-r--r-- | usr/src/test/zfs-tests/runfiles/openindiana.run | 3 | ||||
| -rw-r--r-- | usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/setup.ksh | 8 | ||||
| -rw-r--r-- | usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/zpool_replace_002_neg.ksh | 51 | ||||
| -rwxr-xr-x | usr/src/test/zfs-tests/tests/functional/rsend/send_encrypted_props.ksh | 2 | ||||
| -rw-r--r-- | usr/src/uts/common/fs/zfs/spa.c | 19 | ||||
| -rw-r--r-- | usr/src/uts/common/fs/zfs/spa_misc.c | 36 | ||||
| -rw-r--r-- | usr/src/uts/common/fs/zfs/sys/spa.h | 5 | ||||
| -rw-r--r-- | usr/src/uts/common/fs/zfs/sys/spa_impl.h | 2 |
15 files changed, 268 insertions, 31 deletions
diff --git a/usr/src/cmd/fm/modules/common/zfs-diagnosis/zfs_de.c b/usr/src/cmd/fm/modules/common/zfs-diagnosis/zfs_de.c index d1206a34d3..5e70b01308 100644 --- a/usr/src/cmd/fm/modules/common/zfs-diagnosis/zfs_de.c +++ b/usr/src/cmd/fm/modules/common/zfs-diagnosis/zfs_de.c @@ -22,6 +22,7 @@ /* * Copyright 2015 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. + * Copyright 2019 Joyent, Inc. */ #include <assert.h> @@ -38,9 +39,9 @@ #include <sys/fm/fs/zfs.h> /* - * Our serd engines are named 'zfs_<pool_guid>_<vdev_guid>_{checksum,io}'. This - * #define reserves enough space for two 64-bit hex values plus the length of - * the longest string. + * Our serd engines are named 'zfs_<pool_guid>_<vdev_guid>_{checksum,io,probe}'. + * This #define reserves enough space for two 64-bit hex values plus the length + * of the longest string. */ #define MAX_SERDLEN (16 * 2 + sizeof ("zfs___checksum")) @@ -59,6 +60,7 @@ typedef struct zfs_case_data { char zc_serd_checksum[MAX_SERDLEN]; char zc_serd_io[MAX_SERDLEN]; int zc_has_remove_timer; + char zc_serd_probe[MAX_SERDLEN]; } zfs_case_data_t; /* @@ -88,12 +90,16 @@ typedef struct zfs_case { #define CASE_DATA_VERSION_INITIAL 1 #define CASE_DATA_VERSION_SERD 2 +/* The length of the maximum uint64 rendered as a decimal string. */ +#define MAX_ULL_STR 21 + typedef struct zfs_de_stats { fmd_stat_t old_drops; fmd_stat_t dev_drops; fmd_stat_t vdev_drops; fmd_stat_t import_drops; fmd_stat_t resource_drops; + fmd_stat_t pool_drops; } zfs_de_stats_t; zfs_de_stats_t zfs_stats = { @@ -101,7 +107,8 @@ zfs_de_stats_t zfs_stats = { { "dev_drops", FMD_TYPE_UINT64, "ereports dropped (dev during open)"}, { "vdev_drops", FMD_TYPE_UINT64, "ereports dropped (weird vdev types)"}, { "import_drops", FMD_TYPE_UINT64, "ereports dropped (during import)" }, - { "resource_drops", FMD_TYPE_UINT64, "resource related ereports" } + { "resource_drops", FMD_TYPE_UINT64, "resource related ereports" }, + { "pool_drops", FMD_TYPE_UINT64, "ereports dropped (pool iter failed)"}, }; static hrtime_t zfs_remove_timeout; @@ -279,6 +286,29 @@ zfs_mark_pool(zpool_handle_t *zhp, void *unused) return (0); } +/* + * Find a pool with a matching GUID. + */ +typedef struct find_cbdata { + uint64_t cb_guid; + zpool_handle_t *cb_zhp; +} find_cbdata_t; + +static int +find_pool(zpool_handle_t *zhp, void *data) +{ + find_cbdata_t *cbp = data; + + if (cbp->cb_guid == + zpool_get_prop_int(zhp, ZPOOL_PROP_GUID, NULL)) { + cbp->cb_zhp = zhp; + return (0); + } + + zpool_close(zhp); + return (0); +} + struct load_time_arg { uint64_t lt_guid; er_timeval_t *lt_time; @@ -370,8 +400,8 @@ zfs_purge_cases(fmd_hdl_t *hdl) } /* - * Construct the name of a serd engine given the pool/vdev GUID and type (io or - * checksum). + * Construct the name of a serd engine given the pool/vdev GUID and type (io, + * checksum, or probe). */ static void zfs_serd_name(char *buf, uint64_t pool_guid, uint64_t vdev_guid, @@ -527,6 +557,9 @@ static void zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) { zfs_case_t *zcp, *dcp; + libzfs_handle_t *zhdl; + zpool_handle_t *zhp; + int32_t pool_state; uint64_t ena, pool_guid, vdev_guid; er_timeval_t pool_load; @@ -534,7 +567,10 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) nvlist_t *detector; boolean_t pool_found = B_FALSE; boolean_t isresource; - char *fru, *type; + boolean_t is_inactive_spare, islog, iscache; + nvlist_t *vd_nvl = NULL; + char *fru, *type, *vdg; + find_cbdata_t cb; /* * We subscribe to notifications for vdev or pool removal. In these @@ -627,7 +663,8 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) pool_found = B_TRUE; pool_load = zcp->zc_when; } - if (zcp->zc_data.zc_vdev_guid == vdev_guid) + if (zcp->zc_data.zc_vdev_guid == vdev_guid && + zcp->zc_data.zc_pool_guid == pool_guid) break; } @@ -775,6 +812,8 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) if (zcp->zc_data.zc_serd_checksum[0] != '\0') fmd_serd_reset(hdl, zcp->zc_data.zc_serd_checksum); + if (zcp->zc_data.zc_serd_probe[0] != '\0') + fmd_serd_reset(hdl, zcp->zc_data.zc_serd_probe); } zfs_stats.resource_drops.fmds_value.ui64++; return; @@ -791,12 +830,48 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) if (fmd_case_solved(hdl, zcp->zc_case)) return; + zhdl = fmd_hdl_getspecific(hdl); + + /* + * Find the corresponding pool. + */ + cb.cb_guid = pool_guid; + cb.cb_zhp = NULL; + if (zhdl != NULL && zpool_iter(zhdl, find_pool, &cb) != 0) { + zfs_stats.pool_drops.fmds_value.ui64++; + return; + } + + zhp = cb.cb_zhp; /* NULL if pool was not found. */ + if (zhp != NULL) { + /* + * The libzfs API takes a string representation of a base-10 + * guid here instead of a number, likely because the primary + * libzfs consumers are the CLI tools. + */ + vdg = fmd_hdl_zalloc(hdl, MAX_ULL_STR, FMD_SLEEP); + (void) snprintf(vdg, MAX_ULL_STR, "%" PRIx64, vdev_guid); + + /* + * According to libzfs the 'spare' bit is set when the spare is + * unused, and unset when in use. + * + * We don't really care about the returned nvlist. We're only + * interested in the boolean flags. + */ + if ((vd_nvl = zpool_find_vdev(zhp, vdg, + &is_inactive_spare, &islog, &iscache)) != NULL) { + nvlist_free(vd_nvl); + } + fmd_hdl_free(hdl, vdg, MAX_ULL_STR); + } + /* * Determine if we should solve the case and generate a fault. We solve * a case if: * - * a. A pool failed to open (ereport.fs.zfs.pool) - * b. A device failed to open (ereport.fs.zfs.pool) while a pool + * a. A pool failed to open (ereport.fs.zfs.pool) + * b. A device failed to open (ereport.fs.zfs.pool) while a pool * was up and running. * * We may see a series of ereports associated with a pool open, all @@ -843,8 +918,8 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) boolean_t checkremove = B_FALSE; /* - * If this is a checksum or I/O error, then toss it into the - * appropriate SERD engine and check to see if it has fired. + * If this is a checksum, I/O, or probe error, then toss it into + * the appropriate SERD engine and check to see if it has fired. * Ideally, we want to do something more sophisticated, * (persistent errors for a single data block, etc). For now, * a single SERD engine is sufficient. @@ -894,7 +969,24 @@ zfs_fm_recv(fmd_hdl_t *hdl, fmd_event_t *ep, nvlist_t *nvl, const char *class) } } else if (fmd_nvl_class_match(hdl, nvl, ZFS_MAKE_EREPORT(FM_EREPORT_ZFS_PROBE_FAILURE))) { - checkremove = B_TRUE; + if (zcp->zc_data.zc_serd_probe[0] == '\0') { + zfs_serd_name(zcp->zc_data.zc_serd_probe, + pool_guid, vdev_guid, "probe"); + fmd_serd_create(hdl, zcp->zc_data.zc_serd_probe, + fmd_prop_get_int32(hdl, "probe_N"), + fmd_prop_get_int64(hdl, "probe_T")); + zfs_case_serialize(hdl, zcp); + } + + /* + * We only want to wait for SERD triggers for spare + * vdevs. Normal pool vdevs should be diagnosed + * immediately if a probe failure is received. + */ + if (!is_inactive_spare || fmd_serd_record(hdl, + zcp->zc_data.zc_serd_probe, ep)) { + checkremove = B_TRUE; + } } /* @@ -938,6 +1030,8 @@ zfs_fm_close(fmd_hdl_t *hdl, fmd_case_t *cs) fmd_serd_destroy(hdl, zcp->zc_data.zc_serd_checksum); if (zcp->zc_data.zc_serd_io[0] != '\0') fmd_serd_destroy(hdl, zcp->zc_data.zc_serd_io); + if (zcp->zc_data.zc_serd_probe[0] != '\0') + fmd_serd_destroy(hdl, zcp->zc_data.zc_serd_probe); if (zcp->zc_data.zc_has_remove_timer) fmd_timer_remove(hdl, zcp->zc_remove_timer); uu_list_remove(zfs_cases, zcp); @@ -967,6 +1061,8 @@ static const fmd_prop_t fmd_props[] = { { "checksum_T", FMD_TYPE_TIME, "10min" }, { "io_N", FMD_TYPE_UINT32, "10" }, { "io_T", FMD_TYPE_TIME, "10min" }, + { "probe_N", FMD_TYPE_UINT32, "5" }, + { "probe_T", FMD_TYPE_TIME, "24hour" }, { "remove_timeout", FMD_TYPE_TIME, "15sec" }, { NULL, 0, NULL } }; diff --git a/usr/src/cmd/fm/modules/common/zfs-retire/zfs_retire.c b/usr/src/cmd/fm/modules/common/zfs-retire/zfs_retire.c index 3c70122067..fa46bc51b3 100644 --- a/usr/src/cmd/fm/modules/common/zfs-retire/zfs_retire.c +++ b/usr/src/cmd/fm/modules/common/zfs-retire/zfs_retire.c @@ -124,13 +124,21 @@ find_vdev(libzfs_handle_t *zhdl, nvlist_t *nv, const char *search_fru, } if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_L2CACHE, - &child, &children) != 0) - return (NULL); + &child, &children) == 0) { + for (c = 0; c < children; c++) { + if ((ret = find_vdev(zhdl, child[c], search_fru, + search_guid)) != NULL) + return (ret); + } + } - for (c = 0; c < children; c++) { - if ((ret = find_vdev(zhdl, child[c], search_fru, - search_guid)) != NULL) - return (ret); + if (nvlist_lookup_nvlist_array(nv, ZPOOL_CONFIG_SPARES, + &child, &children) == 0) { + for (c = 0; c < children; c++) { + if ((ret = find_vdev(zhdl, child[c], search_fru, + search_guid)) != NULL) + return (ret); + } } return (NULL); @@ -227,6 +235,8 @@ replace_with_spare(fmd_hdl_t *hdl, zpool_handle_t *zhp, nvlist_t *vdev) char *dev_name; zprop_source_t source; int ashift; + zfs_retire_data_t *zdp = fmd_hdl_getspecific(hdl); + libzfs_handle_t *zhdl = zdp->zrd_hdl; config = zpool_get_config(zhp, NULL); if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, @@ -250,7 +260,7 @@ replace_with_spare(fmd_hdl_t *hdl, zpool_handle_t *zhp, nvlist_t *vdev) (void) nvlist_add_string(replacement, ZPOOL_CONFIG_TYPE, VDEV_TYPE_ROOT); - dev_name = zpool_vdev_name(NULL, zhp, vdev, B_FALSE); + dev_name = zpool_vdev_name(zhdl, zhp, vdev, B_FALSE); /* * Try to replace each spare, ending when we successfully diff --git a/usr/src/lib/libzfs/common/libzfs_pool.c b/usr/src/lib/libzfs/common/libzfs_pool.c index bdb7f93cab..b7c95489c8 100644 --- a/usr/src/lib/libzfs/common/libzfs_pool.c +++ b/usr/src/lib/libzfs/common/libzfs_pool.c @@ -3046,7 +3046,7 @@ zpool_vdev_attach(zpool_handle_t *zhp, zfs_cmd_t zc = { 0 }; char msg[1024]; int ret; - nvlist_t *tgt; + nvlist_t *tgt, *newvd; boolean_t avail_spare, l2cache, islog; uint64_t val; char *newname; @@ -3090,14 +3090,14 @@ zpool_vdev_attach(zpool_handle_t *zhp, if ((newname = zpool_vdev_name(NULL, NULL, child[0], 0)) == NULL) return (-1); + newvd = zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, NULL); /* * If the target is a hot spare that has been swapped in, we can only * replace it with another hot spare. */ if (replacing && nvlist_lookup_uint64(tgt, ZPOOL_CONFIG_IS_SPARE, &val) == 0 && - (zpool_find_vdev(zhp, newname, &avail_spare, &l2cache, - NULL) == NULL || !avail_spare) && + (newvd == NULL || !avail_spare) && is_replacing_spare(config_root, tgt, 1)) { zfs_error_aux(hdl, dgettext(TEXT_DOMAIN, "can only be replaced by another hot spare")); @@ -3107,6 +3107,11 @@ zpool_vdev_attach(zpool_handle_t *zhp, free(newname); + if (replacing && avail_spare && !vdev_is_online(newvd)) { + (void) zpool_standard_error(hdl, ENXIO, msg); + return (-1); + } + if (zcmd_write_conf_nvlist(hdl, &zc, nvroot) != 0) return (-1); diff --git a/usr/src/pkg/manifests/system-test-zfstest.mf b/usr/src/pkg/manifests/system-test-zfstest.mf index 3c5cbfbd42..0a95e4d759 100644 --- a/usr/src/pkg/manifests/system-test-zfstest.mf +++ b/usr/src/pkg/manifests/system-test-zfstest.mf @@ -1941,6 +1941,9 @@ file path=opt/zfs-tests/tests/functional/cli_root/zpool_replace/setup \ file \ path=opt/zfs-tests/tests/functional/cli_root/zpool_replace/zpool_replace_001_neg \ mode=0555 +file \ + path=opt/zfs-tests/tests/functional/cli_root/zpool_replace/zpool_replace_002_neg \ + mode=0555 file path=opt/zfs-tests/tests/functional/cli_root/zpool_resilver/cleanup \ mode=0555 file path=opt/zfs-tests/tests/functional/cli_root/zpool_resilver/setup \ diff --git a/usr/src/test/zfs-tests/include/libtest.shlib b/usr/src/test/zfs-tests/include/libtest.shlib index 363f674f03..725c971a4c 100644 --- a/usr/src/test/zfs-tests/include/libtest.shlib +++ b/usr/src/test/zfs-tests/include/libtest.shlib @@ -323,6 +323,13 @@ function default_mirror_setup log_pass } +function default_mirror_2way_setup +{ + default_mirror_setup_noexit $1 $2 + + log_pass +} + # # Given a pair of disks, set up a storage pool and dataset for the mirror # @parameters: $1 the primary side of the mirror diff --git a/usr/src/test/zfs-tests/runfiles/delphix.run b/usr/src/test/zfs-tests/runfiles/delphix.run index e97d0e6e0b..beb23d9d82 100644 --- a/usr/src/test/zfs-tests/runfiles/delphix.run +++ b/usr/src/test/zfs-tests/runfiles/delphix.run @@ -363,7 +363,8 @@ tests = ['zpool_remove_001_neg', 'zpool_remove_002_pos', 'zpool_remove_003_pos'] [/opt/zfs-tests/tests/functional/cli_root/zpool_replace] -tests = ['zpool_replace_001_neg', 'replace-o_ashift', 'replace_prop_ashift'] +tests = ['zpool_replace_001_neg', 'zpool_replace_002_neg', 'replace-o_ashift', + 'replace_prop_ashift'] [/opt/zfs-tests/tests/functional/cli_root/zpool_resilver] tests = ['zpool_resilver_bad_args', 'zpool_resilver_restart'] diff --git a/usr/src/test/zfs-tests/runfiles/omnios.run b/usr/src/test/zfs-tests/runfiles/omnios.run index ff7a697d5f..af9f29e8ca 100644 --- a/usr/src/test/zfs-tests/runfiles/omnios.run +++ b/usr/src/test/zfs-tests/runfiles/omnios.run @@ -363,7 +363,8 @@ tests = ['zpool_remove_001_neg', 'zpool_remove_002_pos', 'zpool_remove_003_pos'] [/opt/zfs-tests/tests/functional/cli_root/zpool_replace] -tests = ['zpool_replace_001_neg', 'replace-o_ashift', 'replace_prop_ashift'] +tests = ['zpool_replace_001_neg', 'zpool_replace_002_neg', 'replace-o_ashift', + 'replace_prop_ashift'] [/opt/zfs-tests/tests/functional/cli_root/zpool_resilver] tests = ['zpool_resilver_bad_args', 'zpool_resilver_restart'] diff --git a/usr/src/test/zfs-tests/runfiles/openindiana.run b/usr/src/test/zfs-tests/runfiles/openindiana.run index 9ab1d42293..78923b6a1f 100644 --- a/usr/src/test/zfs-tests/runfiles/openindiana.run +++ b/usr/src/test/zfs-tests/runfiles/openindiana.run @@ -363,7 +363,8 @@ tests = ['zpool_remove_001_neg', 'zpool_remove_002_pos', 'zpool_remove_003_pos'] [/opt/zfs-tests/tests/functional/cli_root/zpool_replace] -tests = ['zpool_replace_001_neg', 'replace-o_ashift', 'replace_prop_ashift'] +tests = ['zpool_replace_001_neg', 'zpool_replace_002_neg', 'replace-o_ashift', + 'replace_prop_ashift'] [/opt/zfs-tests/tests/functional/cli_root/zpool_resilver] tests = ['zpool_resilver_bad_args', 'zpool_resilver_restart'] diff --git a/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/setup.ksh b/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/setup.ksh index 51dfd8107b..9a45a363bf 100644 --- a/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/setup.ksh +++ b/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/setup.ksh @@ -25,11 +25,15 @@ # Use is subject to license terms. # +# +# Copyright 2019 Joyent, Inc. +# + . $STF_SUITE/include/libtest.shlib verify_runnable "global" -verify_disk_count "$DISKS" 2 +verify_disk_count "$DISKS" 3 DISK=${DISKS%% *} -default_mirror_setup $DISKS +default_mirror_2way_setup $DISKS diff --git a/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/zpool_replace_002_neg.ksh b/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/zpool_replace_002_neg.ksh new file mode 100644 index 0000000000..fde508254c --- /dev/null +++ b/usr/src/test/zfs-tests/tests/functional/cli_root/zpool_replace/zpool_replace_002_neg.ksh @@ -0,0 +1,51 @@ +#!/usr/bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or http://www.opensolaris.org/os/licensing. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright 2019 Joyent, Inc. +# + +. $STF_SUITE/include/libtest.shlib + +# +# DESCRIPTION: +# +# zpool replace returns an error when spare device is faulted. +# +# STRATEGY: +# 1. Add hot spare to pool +# 2. Fault the hot spare device +# 3. Attempt to replace a device in a pool with the faulted spare +# 4. Verify the 'zpool replace' command fails +# + +SPARE=${DISKS##* } +DISK=${DISKS%% *} + +verify_runnable "global" +log_must zpool add $TESTPOOL spare $SPARE +log_assert "zpool replace returns an error when the hot spare is faulted" + +log_must zinject -d $SPARE -A fault $TESTPOOL +log_mustnot zpool replace $TESTPOOL $DISK $SPARE + +log_pass "zpool replace returns an error when the hot spare is faulted" diff --git a/usr/src/test/zfs-tests/tests/functional/rsend/send_encrypted_props.ksh b/usr/src/test/zfs-tests/tests/functional/rsend/send_encrypted_props.ksh index d4d7dc9e63..f0a591c427 100755 --- a/usr/src/test/zfs-tests/tests/functional/rsend/send_encrypted_props.ksh +++ b/usr/src/test/zfs-tests/tests/functional/rsend/send_encrypted_props.ksh @@ -191,7 +191,7 @@ log_must test "$(get_prop 'encryptionroot' $ds)" == "$TESTPOOL/crypt" log_must test "$(get_prop 'encryption' $ds)" == "aes-256-ccm" log_must test "$(get_prop 'keyformat' $ds)" == "passphrase" log_must test "$(get_prop 'mounted' $ds)" == "yes" -recv_cksum=$(md5digest /$ds/$TESTFILE0) +recv_cksum=$(md5sum /$ds/$TESTFILE0 | awk '{ print $1 }') log_must test "$recv_cksum" == "$cksum" log_must zfs destroy -r $ds diff --git a/usr/src/uts/common/fs/zfs/spa.c b/usr/src/uts/common/fs/zfs/spa.c index 4e37942a8b..05fd29810b 100644 --- a/usr/src/uts/common/fs/zfs/spa.c +++ b/usr/src/uts/common/fs/zfs/spa.c @@ -27,7 +27,7 @@ * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright 2016 Toomas Soome <tsoome@me.com> - * Copyright 2018 Joyent, Inc. + * Copyright 2019 Joyent, Inc. * Copyright (c) 2017, Intel Corporation. * Copyright (c) 2017 Datto Inc. * Copyright 2018 OmniOS Community Edition (OmniOSce) Association. @@ -229,6 +229,13 @@ uint64_t zfs_max_missing_tvds_cachefile = SPA_DVAS_PER_BP - 1; uint64_t zfs_max_missing_tvds_scan = 0; /* + * Interval in seconds at which to poll spare vdevs for health. + * Setting this to zero disables spare polling. + * Set to three hours by default. + */ +uint_t spa_spare_poll_interval_seconds = 60 * 60 * 3; + +/* * Debugging aid that pauses spa_sync() towards the end. */ boolean_t zfs_pause_spa_sync = B_FALSE; @@ -7539,6 +7546,8 @@ spa_async_thread(void *arg) if (tasks & SPA_ASYNC_PROBE) { spa_vdev_state_enter(spa, SCL_NONE); spa_async_probe(spa, spa->spa_root_vdev); + for (int i = 0; i < spa->spa_spares.sav_count; i++) + spa_async_probe(spa, spa->spa_spares.sav_vdevs[i]); (void) spa_vdev_state_exit(spa, NULL, 0); } @@ -8624,6 +8633,14 @@ spa_sync(spa_t *spa, uint64_t txg) spa_handle_ignored_writes(spa); + /* Mark unused spares as needing a health check. */ + if (spa_spare_poll_interval_seconds != 0 && + NSEC2SEC(gethrtime() - spa->spa_spares_last_polled) > + spa_spare_poll_interval_seconds) { + spa_spare_poll(spa); + spa->spa_spares_last_polled = gethrtime(); + } + /* * If any async tasks have been requested, kick them off. */ diff --git a/usr/src/uts/common/fs/zfs/spa_misc.c b/usr/src/uts/common/fs/zfs/spa_misc.c index 53a7bb8d23..45e1978803 100644 --- a/usr/src/uts/common/fs/zfs/spa_misc.c +++ b/usr/src/uts/common/fs/zfs/spa_misc.c @@ -26,6 +26,7 @@ * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Integros [integros.com] * Copyright (c) 2017 Datto Inc. + * Copyright 2019 Joyent, Inc. * Copyright (c) 2017, Intel Corporation. */ @@ -1045,6 +1046,41 @@ spa_aux_activate(vdev_t *vd, avl_tree_t *avl) * be completely consistent with respect to other vdev configuration changes. */ +/* + * Poll the spare vdevs to make sure they are not faulty. + * + * The probe operation will raise an ENXIO error and create an FM ereport if the + * probe fails. + */ +void +spa_spare_poll(spa_t *spa) +{ + boolean_t async_request = B_FALSE; + spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); + for (int i = 0; i < spa->spa_spares.sav_count; i++) { + spa_aux_t search, *found; + vdev_t *vd = spa->spa_spares.sav_vdevs[i]; + + search.aux_guid = vd->vdev_guid; + + mutex_enter(&spa_spare_lock); + found = avl_find(&spa_spare_avl, &search, NULL); + /* This spare is in use by a pool. */ + if (found != NULL && found->aux_pool != 0) { + mutex_exit(&spa_spare_lock); + continue; + } + mutex_exit(&spa_spare_lock); + + vd->vdev_probe_wanted = B_TRUE; + async_request = B_TRUE; + } + if (async_request) + spa_async_request(spa, SPA_ASYNC_PROBE); + + spa_config_exit(spa, SCL_STATE, FTAG); +} + static int spa_spare_compare(const void *a, const void *b) { diff --git a/usr/src/uts/common/fs/zfs/sys/spa.h b/usr/src/uts/common/fs/zfs/sys/spa.h index 17c844243d..cb736db5dd 100644 --- a/usr/src/uts/common/fs/zfs/sys/spa.h +++ b/usr/src/uts/common/fs/zfs/sys/spa.h @@ -25,7 +25,7 @@ * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Integros [integros.com] - * Copyright 2017 Joyent, Inc. + * Copyright 2019 Joyent, Inc. * Copyright (c) 2017 Datto Inc. * Copyright (c) 2017, Intel Corporation. */ @@ -819,6 +819,9 @@ extern void spa_spare_remove(vdev_t *vd); extern boolean_t spa_spare_exists(uint64_t guid, uint64_t *pool, int *refcnt); extern void spa_spare_activate(vdev_t *vd); +/* spare polling */ +extern void spa_spare_poll(spa_t *spa); + /* L2ARC state (which is global across all pools) */ extern void spa_l2cache_add(vdev_t *vd); extern void spa_l2cache_remove(vdev_t *vd); diff --git a/usr/src/uts/common/fs/zfs/sys/spa_impl.h b/usr/src/uts/common/fs/zfs/sys/spa_impl.h index 1408d012fc..bc00528fa9 100644 --- a/usr/src/uts/common/fs/zfs/sys/spa_impl.h +++ b/usr/src/uts/common/fs/zfs/sys/spa_impl.h @@ -26,6 +26,7 @@ * Copyright 2013 Saso Kiselkov. All rights reserved. * Copyright (c) 2017 Datto Inc. * Copyright (c) 2017, Intel Corporation. + * Copyright 2019 Joyent, Inc. */ #ifndef _SYS_SPA_IMPL_H @@ -254,6 +255,7 @@ struct spa { spa_aux_vdev_t spa_spares; /* hot spares */ spa_aux_vdev_t spa_l2cache; /* L2ARC cache devices */ + hrtime_t spa_spares_last_polled; /* time spares last polled */ nvlist_t *spa_label_features; /* Features for reading MOS */ uint64_t spa_config_object; /* MOS object for pool config */ uint64_t spa_config_generation; /* config generation number */ |