diff options
Diffstat (limited to 'sys/contrib/openzfs/cmd')
27 files changed, 824 insertions, 263 deletions
diff --git a/sys/contrib/openzfs/cmd/Makefile.am b/sys/contrib/openzfs/cmd/Makefile.am index 96040976e53e..ca94f6b77e06 100644 --- a/sys/contrib/openzfs/cmd/Makefile.am +++ b/sys/contrib/openzfs/cmd/Makefile.am @@ -98,17 +98,16 @@ endif if USING_PYTHON -bin_SCRIPTS += arc_summary arcstat dbufstat zilstat -CLEANFILES += arc_summary arcstat dbufstat zilstat -dist_noinst_DATA += %D%/arc_summary %D%/arcstat.in %D%/dbufstat.in %D%/zilstat.in +bin_SCRIPTS += zarcsummary zarcstat dbufstat zilstat +CLEANFILES += zarcsummary zarcstat dbufstat zilstat +dist_noinst_DATA += %D%/zarcsummary %D%/zarcstat.in %D%/dbufstat.in %D%/zilstat.in -$(call SUBST,arcstat,%D%/) +$(call SUBST,zarcstat,%D%/) $(call SUBST,dbufstat,%D%/) $(call SUBST,zilstat,%D%/) -arc_summary: %D%/arc_summary +zarcsummary: %D%/zarcsummary $(AM_V_at)cp $< $@ endif - PHONY += cmd cmd: $(bin_SCRIPTS) $(bin_PROGRAMS) $(sbin_SCRIPTS) $(sbin_PROGRAMS) $(dist_bin_SCRIPTS) $(zfsexec_PROGRAMS) $(mounthelper_PROGRAMS) diff --git a/sys/contrib/openzfs/cmd/arcstat.in b/sys/contrib/openzfs/cmd/zarcstat.in index 6f9abb39c3fb..8ffd20481166 100755 --- a/sys/contrib/openzfs/cmd/arcstat.in +++ b/sys/contrib/openzfs/cmd/zarcstat.in @@ -2,7 +2,7 @@ # SPDX-License-Identifier: CDDL-1.0 # # Print out ZFS ARC Statistics exported via kstat(1) -# For a definition of fields, or usage, use arcstat -v +# For a definition of fields, or usage, use zarcstat -v # # This script was originally a fork of the original arcstat.pl (0.1) # by Neelakanth Nadgir, originally published on his Sun blog on @@ -56,6 +56,7 @@ import time import getopt import re import copy +import os from signal import signal, SIGINT, SIGWINCH, SIG_DFL @@ -171,7 +172,7 @@ cols = { "zactive": [7, 1000, "zfetch prefetches active per second"], } -# ARC structural breakdown from arc_summary +# ARC structural breakdown from zarcsummary structfields = { "cmp": ["compressed", "Compressed"], "ovh": ["overhead", "Overhead"], @@ -187,7 +188,7 @@ structstats = { # size stats "sz": ["_size", "size"], } -# ARC types breakdown from arc_summary +# ARC types breakdown from zarcsummary typefields = { "data": ["data", "ARC data"], "meta": ["metadata", "ARC metadata"], @@ -198,7 +199,7 @@ typestats = { # size stats "sz": ["_size", "size"], } -# ARC states breakdown from arc_summary +# ARC states breakdown from zarcsummary statefields = { "ano": ["anon", "Anonymous"], "mfu": ["mfu", "MFU"], @@ -261,7 +262,7 @@ hdr_intr = 20 # Print header every 20 lines of output opfile = None sep = " " # Default separator is 2 spaces l2exist = False -cmd = ("Usage: arcstat [-havxp] [-f fields] [-o file] [-s string] [interval " +cmd = ("Usage: zarcstat [-havxp] [-f fields] [-o file] [-s string] [interval " "[count]]\n") cur = {} d = {} @@ -348,10 +349,10 @@ def usage(): "character or string\n") sys.stderr.write("\t -p : Disable auto-scaling of numerical fields\n") sys.stderr.write("\nExamples:\n") - sys.stderr.write("\tarcstat -o /tmp/a.log 2 10\n") - sys.stderr.write("\tarcstat -s \",\" -o /tmp/a.log 2 10\n") - sys.stderr.write("\tarcstat -v\n") - sys.stderr.write("\tarcstat -f time,hit%,dh%,ph%,mh% 1\n") + sys.stderr.write("\tzarcstat -o /tmp/a.log 2 10\n") + sys.stderr.write("\tzarcstat -s \",\" -o /tmp/a.log 2 10\n") + sys.stderr.write("\tzarcstat -v\n") + sys.stderr.write("\tzarcstat -f time,hit%,dh%,ph%,mh% 1\n") sys.stderr.write("\n") sys.exit(1) @@ -366,7 +367,7 @@ def snap_stats(): cur = kstat - # fill in additional values from arc_summary + # fill in additional values from zarcsummary cur["caches_size"] = caches_size = cur["anon_data"]+cur["anon_metadata"]+\ cur["mfu_data"]+cur["mfu_metadata"]+cur["mru_data"]+cur["mru_metadata"]+\ cur["uncached_data"]+cur["uncached_metadata"] @@ -766,6 +767,7 @@ def calculate(): def main(): + global sint global count global hdr_intr diff --git a/sys/contrib/openzfs/cmd/arc_summary b/sys/contrib/openzfs/cmd/zarcsummary index e60c6b64e8a1..24a129d9ca70 100755 --- a/sys/contrib/openzfs/cmd/arc_summary +++ b/sys/contrib/openzfs/cmd/zarcsummary @@ -34,7 +34,7 @@ Provides basic information on the ARC, its efficiency, the L2ARC (if present), the Data Management Unit (DMU), Virtual Devices (VDEVs), and tunables. See the in-source documentation and code at https://github.com/openzfs/zfs/blob/master/module/zfs/arc.c for details. -The original introduction to arc_summary can be found at +The original introduction to zarcsummary can be found at http://cuddletech.com/?p=454 """ @@ -161,7 +161,7 @@ elif sys.platform.startswith('linux'): return get_params(TUNABLES_PATH) def get_version_impl(request): - # The original arc_summary called /sbin/modinfo/{spl,zfs} to get + # The original zarcsummary called /sbin/modinfo/{spl,zfs} to get # the version information. We switch to /sys/module/{spl,zfs}/version # to make sure we get what is really loaded in the kernel try: @@ -439,7 +439,7 @@ def print_header(): """ # datetime is now recommended over time but we keep the exact formatting - # from the older version of arc_summary in case there are scripts + # from the older version of zarcsummary in case there are scripts # that expect it in this way daydate = time.strftime(DATE_FORMAT) spc_date = LINE_LENGTH-len(daydate) diff --git a/sys/contrib/openzfs/cmd/zdb/zdb.c b/sys/contrib/openzfs/cmd/zdb/zdb.c index 66d5fbd6adbe..70a4ed46f263 100644 --- a/sys/contrib/openzfs/cmd/zdb/zdb.c +++ b/sys/contrib/openzfs/cmd/zdb/zdb.c @@ -107,7 +107,9 @@ extern uint_t zfs_reconstruct_indirect_combinations_max; extern uint_t zfs_btree_verify_intensity; static const char cmdname[] = "zdb"; -uint8_t dump_opt[256]; +uint8_t dump_opt[512]; + +#define ALLOCATED_OPT 256 typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size); @@ -127,6 +129,7 @@ static zfs_range_tree_t *mos_refd_objs; static spa_t *spa; static objset_t *os; static boolean_t kernel_init_done; +static boolean_t corruption_found = B_FALSE; static void snprintf_blkptr_compact(char *, size_t, const blkptr_t *, boolean_t); @@ -250,6 +253,7 @@ sublivelist_verify_func(void *args, dsl_deadlist_entry_t *dle) &e->svbr_blk, B_TRUE); (void) printf("\tERROR: %d unmatched FREE(s): %s\n", e->svbr_refcnt, blkbuf); + corruption_found = B_TRUE; } zfs_btree_destroy(&sv->sv_pair); @@ -381,7 +385,7 @@ verify_livelist_allocs(metaslab_verify_t *mv, uint64_t txg, sublivelist_verify_block_t svb = {{{0}}}; DVA_SET_VDEV(&svb.svb_dva, mv->mv_vdid); DVA_SET_OFFSET(&svb.svb_dva, offset); - DVA_SET_ASIZE(&svb.svb_dva, size); + DVA_SET_ASIZE(&svb.svb_dva, 0); zfs_btree_index_t where; uint64_t end_offset = offset + size; @@ -405,6 +409,7 @@ verify_livelist_allocs(metaslab_verify_t *mv, uint64_t txg, (u_longlong_t)DVA_GET_ASIZE(&found->svb_dva), (u_longlong_t)found->svb_allocated_txg, (u_longlong_t)txg); + corruption_found = B_TRUE; } } } @@ -426,6 +431,7 @@ metaslab_spacemap_validation_cb(space_map_entry_t *sme, void *arg) (u_longlong_t)txg, (u_longlong_t)offset, (u_longlong_t)size, (u_longlong_t)mv->mv_vdid, (u_longlong_t)mv->mv_msid); + corruption_found = B_TRUE; } else { zfs_range_tree_add(mv->mv_allocated, offset, size); @@ -439,6 +445,7 @@ metaslab_spacemap_validation_cb(space_map_entry_t *sme, void *arg) (u_longlong_t)txg, (u_longlong_t)offset, (u_longlong_t)size, (u_longlong_t)mv->mv_vdid, (u_longlong_t)mv->mv_msid); + corruption_found = B_TRUE; } else { zfs_range_tree_remove(mv->mv_allocated, offset, size); @@ -526,6 +533,7 @@ mv_populate_livelist_allocs(metaslab_verify_t *mv, sublivelist_verify_t *sv) (u_longlong_t)DVA_GET_VDEV(&svb->svb_dva), (u_longlong_t)DVA_GET_OFFSET(&svb->svb_dva), (u_longlong_t)DVA_GET_ASIZE(&svb->svb_dva)); + corruption_found = B_TRUE; continue; } @@ -542,6 +550,7 @@ mv_populate_livelist_allocs(metaslab_verify_t *mv, sublivelist_verify_t *sv) (u_longlong_t)DVA_GET_VDEV(&svb->svb_dva), (u_longlong_t)DVA_GET_OFFSET(&svb->svb_dva), (u_longlong_t)DVA_GET_ASIZE(&svb->svb_dva)); + corruption_found = B_TRUE; continue; } @@ -655,6 +664,7 @@ livelist_metaslab_validate(spa_t *spa) } (void) printf("ERROR: Found livelist blocks marked as allocated " "for indirect vdevs:\n"); + corruption_found = B_TRUE; zfs_btree_index_t *where = NULL; sublivelist_verify_block_t *svb; @@ -827,7 +837,7 @@ usage(void) (void) fprintf(stderr, "Specify an option more than once (e.g. -bb) " "to make only that option verbose\n"); (void) fprintf(stderr, "Default is to dump everything non-verbosely\n"); - zdb_exit(1); + zdb_exit(2); } static void @@ -1578,9 +1588,8 @@ dump_spacemap(objset_t *os, space_map_t *sm) continue; } - uint8_t words; char entry_type; - uint64_t entry_off, entry_run, entry_vdev = SM_NO_VDEVID; + uint64_t entry_off, entry_run, entry_vdev; if (sm_entry_is_single_word(word)) { entry_type = (SM_TYPE_DECODE(word) == SM_ALLOC) ? @@ -1588,35 +1597,43 @@ dump_spacemap(objset_t *os, space_map_t *sm) entry_off = (SM_OFFSET_DECODE(word) << mapshift) + sm->sm_start; entry_run = SM_RUN_DECODE(word) << mapshift; - words = 1; + + (void) printf("\t [%6llu] %c " + "range: %012llx-%012llx size: %08llx\n", + (u_longlong_t)entry_id, entry_type, + (u_longlong_t)entry_off, + (u_longlong_t)(entry_off + entry_run - 1), + (u_longlong_t)entry_run); } else { /* it is a two-word entry so we read another word */ ASSERT(sm_entry_is_double_word(word)); uint64_t extra_word; offset += sizeof (extra_word); + ASSERT3U(offset, <, space_map_length(sm)); VERIFY0(dmu_read(os, space_map_object(sm), offset, sizeof (extra_word), &extra_word, DMU_READ_PREFETCH)); - ASSERT3U(offset, <=, space_map_length(sm)); - entry_run = SM2_RUN_DECODE(word) << mapshift; entry_vdev = SM2_VDEV_DECODE(word); entry_type = (SM2_TYPE_DECODE(extra_word) == SM_ALLOC) ? 'A' : 'F'; entry_off = (SM2_OFFSET_DECODE(extra_word) << mapshift) + sm->sm_start; - words = 2; - } - (void) printf("\t [%6llu] %c range:" - " %010llx-%010llx size: %06llx vdev: %06llu words: %u\n", - (u_longlong_t)entry_id, - entry_type, (u_longlong_t)entry_off, - (u_longlong_t)(entry_off + entry_run), - (u_longlong_t)entry_run, - (u_longlong_t)entry_vdev, words); + if (zopt_metaslab_args == 0 || + zopt_metaslab[0] == entry_vdev) { + (void) printf("\t [%6llu] %c " + "range: %012llx-%012llx size: %08llx " + "vdev: %llu\n", + (u_longlong_t)entry_id, entry_type, + (u_longlong_t)entry_off, + (u_longlong_t)(entry_off + entry_run - 1), + (u_longlong_t)entry_run, + (u_longlong_t)entry_vdev); + } + } if (entry_type == 'A') alloc += entry_run; @@ -1652,6 +1669,16 @@ dump_metaslab_stats(metaslab_t *msp) } static void +dump_allocated(void *arg, uint64_t start, uint64_t size) +{ + uint64_t *off = arg; + if (*off != start) + (void) printf("ALLOC: %"PRIu64" %"PRIu64"\n", *off, + start - *off); + *off = start + size; +} + +static void dump_metaslab(metaslab_t *msp) { vdev_t *vd = msp->ms_group->mg_vd; @@ -1667,13 +1694,24 @@ dump_metaslab(metaslab_t *msp) (u_longlong_t)msp->ms_id, (u_longlong_t)msp->ms_start, (u_longlong_t)space_map_object(sm), freebuf); - if (dump_opt['m'] > 2 && !dump_opt['L']) { + if (dump_opt[ALLOCATED_OPT] || + (dump_opt['m'] > 2 && !dump_opt['L'])) { mutex_enter(&msp->ms_lock); VERIFY0(metaslab_load(msp)); + } + + if (dump_opt['m'] > 2 && !dump_opt['L']) { zfs_range_tree_stat_verify(msp->ms_allocatable); dump_metaslab_stats(msp); - metaslab_unload(msp); - mutex_exit(&msp->ms_lock); + } + + if (dump_opt[ALLOCATED_OPT]) { + uint64_t off = msp->ms_start; + zfs_range_tree_walk(msp->ms_allocatable, dump_allocated, + &off); + if (off != msp->ms_start + msp->ms_size) + (void) printf("ALLOC: %"PRIu64" %"PRIu64"\n", off, + msp->ms_size - off); } if (dump_opt['m'] > 1 && sm != NULL && @@ -1688,6 +1726,12 @@ dump_metaslab(metaslab_t *msp) SPACE_MAP_HISTOGRAM_SIZE, sm->sm_shift); } + if (dump_opt[ALLOCATED_OPT] || + (dump_opt['m'] > 2 && !dump_opt['L'])) { + metaslab_unload(msp); + mutex_exit(&msp->ms_lock); + } + if (vd->vdev_ops == &vdev_draid_ops) ASSERT3U(msp->ms_size, <=, 1ULL << vd->vdev_ms_shift); else @@ -1724,8 +1768,9 @@ print_vdev_metaslab_header(vdev_t *vd) } } - (void) printf("\tvdev %10llu %s", - (u_longlong_t)vd->vdev_id, bias_str); + (void) printf("\tvdev %10llu\t%s metaslab shift %4llu", + (u_longlong_t)vd->vdev_id, bias_str, + (u_longlong_t)vd->vdev_ms_shift); if (ms_flush_data_obj != 0) { (void) printf(" ms_unflushed_phys object %llu", @@ -1865,7 +1910,7 @@ dump_metaslabs(spa_t *spa) (void) printf("\nMetaslabs:\n"); - if (!dump_opt['d'] && zopt_metaslab_args > 0) { + if (zopt_metaslab_args > 0) { c = zopt_metaslab[0]; if (c >= children) @@ -2583,19 +2628,17 @@ snprintf_blkptr_compact(char *blkbuf, size_t buflen, const blkptr_t *bp, } } -static void +static u_longlong_t print_indirect(spa_t *spa, blkptr_t *bp, const zbookmark_phys_t *zb, const dnode_phys_t *dnp) { char blkbuf[BP_SPRINTF_LEN]; + u_longlong_t offset; int l; - if (!BP_IS_EMBEDDED(bp)) { - ASSERT3U(BP_GET_TYPE(bp), ==, dnp->dn_type); - ASSERT3U(BP_GET_LEVEL(bp), ==, zb->zb_level); - } + offset = (u_longlong_t)blkid2offset(dnp, bp, zb); - (void) printf("%16llx ", (u_longlong_t)blkid2offset(dnp, bp, zb)); + (void) printf("%16llx ", offset); ASSERT(zb->zb_level >= 0); @@ -2610,19 +2653,38 @@ print_indirect(spa_t *spa, blkptr_t *bp, const zbookmark_phys_t *zb, snprintf_blkptr_compact(blkbuf, sizeof (blkbuf), bp, B_FALSE); if (dump_opt['Z'] && BP_GET_COMPRESS(bp) == ZIO_COMPRESS_ZSTD) snprintf_zstd_header(spa, blkbuf, sizeof (blkbuf), bp); - (void) printf("%s\n", blkbuf); + (void) printf("%s", blkbuf); + + if (!BP_IS_EMBEDDED(bp)) { + if (BP_GET_TYPE(bp) != dnp->dn_type) { + (void) printf(" (ERROR: Block pointer type " + "(%llu) does not match dnode type (%hhu))", + BP_GET_TYPE(bp), dnp->dn_type); + corruption_found = B_TRUE; + } + if (BP_GET_LEVEL(bp) != zb->zb_level) { + (void) printf(" (ERROR: Block pointer level " + "(%llu) does not match bookmark level (%lld))", + BP_GET_LEVEL(bp), (longlong_t)zb->zb_level); + corruption_found = B_TRUE; + } + } + (void) printf("\n"); + + return (offset); } static int visit_indirect(spa_t *spa, const dnode_phys_t *dnp, blkptr_t *bp, const zbookmark_phys_t *zb) { + u_longlong_t offset; int err = 0; if (BP_GET_BIRTH(bp) == 0) return (0); - print_indirect(spa, bp, zb, dnp); + offset = print_indirect(spa, bp, zb, dnp); if (BP_GET_LEVEL(bp) > 0 && !BP_IS_HOLE(bp)) { arc_flags_t flags = ARC_FLAG_WAIT; @@ -2652,8 +2714,15 @@ visit_indirect(spa_t *spa, const dnode_phys_t *dnp, break; fill += BP_GET_FILL(cbp); } - if (!err) - ASSERT3U(fill, ==, BP_GET_FILL(bp)); + if (!err) { + if (fill != BP_GET_FILL(bp)) { + (void) printf("%16llx: Block pointer " + "fill (%llu) does not match calculated " + "value (%llu)\n", offset, BP_GET_FILL(bp), + (u_longlong_t)fill); + corruption_found = B_TRUE; + } + } arc_buf_destroy(buf, &buf); } @@ -2909,6 +2978,7 @@ dump_full_bpobj(bpobj_t *bpo, const char *name, int indent) (void) printf("ERROR %u while trying to open " "subobj id %llu\n", error, (u_longlong_t)subobj); + corruption_found = B_TRUE; continue; } dump_full_bpobj(&subbpo, "subobj", indent + 1); @@ -3088,6 +3158,7 @@ bpobj_count_refd(bpobj_t *bpo) (void) printf("ERROR %u while trying to open " "subobj id %llu\n", error, (u_longlong_t)subobj); + corruption_found = B_TRUE; continue; } bpobj_count_refd(&subbpo); @@ -3230,6 +3301,7 @@ zdb_derive_key(dsl_dir_t *dd, uint8_t *key_out) uint64_t keyformat, salt, iters; int i; unsigned char c; + FILE *f; VERIFY0(zap_lookup(dd->dd_pool->dp_meta_objset, dd->dd_crypto_obj, zfs_prop_to_name(ZFS_PROP_KEYFORMAT), sizeof (uint64_t), @@ -3262,6 +3334,25 @@ zdb_derive_key(dsl_dir_t *dd, uint8_t *key_out) break; + case ZFS_KEYFORMAT_RAW: + if ((f = fopen(key_material, "r")) == NULL) + return (B_FALSE); + + if (fread(key_out, 1, WRAPPING_KEY_LEN, f) != + WRAPPING_KEY_LEN) { + (void) fclose(f); + return (B_FALSE); + } + + /* Check the key length */ + if (fgetc(f) != EOF) { + (void) fclose(f); + return (B_FALSE); + } + + (void) fclose(f); + break; + default: fatal("no support for key format %u\n", (unsigned int) keyformat); @@ -9334,6 +9425,8 @@ main(int argc, char **argv) {"all-reconstruction", no_argument, NULL, 'Y'}, {"livelist", no_argument, NULL, 'y'}, {"zstd-headers", no_argument, NULL, 'Z'}, + {"allocated-map", no_argument, NULL, + ALLOCATED_OPT}, {0, 0, 0, 0} }; @@ -9364,6 +9457,7 @@ main(int argc, char **argv) case 'u': case 'y': case 'Z': + case ALLOCATED_OPT: dump_opt[c]++; dump_all = 0; break; @@ -9634,7 +9728,7 @@ main(int argc, char **argv) } else if (objset_str && !zdb_numeric(objset_str + 1) && dump_opt['N']) { printf("Supply a numeric objset ID with -N\n"); - error = 1; + error = 2; goto fini; } } else { @@ -9936,5 +10030,8 @@ fini: if (kernel_init_done) kernel_fini(); + if (corruption_found && error == 0) + error = 3; + return (error); } diff --git a/sys/contrib/openzfs/cmd/zdb/zdb.h b/sys/contrib/openzfs/cmd/zdb/zdb.h index 6b6c9169816b..48b561eb202c 100644 --- a/sys/contrib/openzfs/cmd/zdb/zdb.h +++ b/sys/contrib/openzfs/cmd/zdb/zdb.h @@ -29,6 +29,6 @@ #define _ZDB_H void dump_intent_log(zilog_t *); -extern uint8_t dump_opt[256]; +extern uint8_t dump_opt[512]; #endif /* _ZDB_H */ diff --git a/sys/contrib/openzfs/cmd/zdb/zdb_il.c b/sys/contrib/openzfs/cmd/zdb/zdb_il.c index 62e290cd122c..3d91fb28a4c7 100644 --- a/sys/contrib/openzfs/cmd/zdb/zdb_il.c +++ b/sys/contrib/openzfs/cmd/zdb/zdb_il.c @@ -48,8 +48,6 @@ #include "zdb.h" -extern uint8_t dump_opt[256]; - static char tab_prefix[4] = "\t\t\t"; static void diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/Makefile.am b/sys/contrib/openzfs/cmd/zed/zed.d/Makefile.am index 093a04c4636a..c0b161ecf248 100644 --- a/sys/contrib/openzfs/cmd/zed/zed.d/Makefile.am +++ b/sys/contrib/openzfs/cmd/zed/zed.d/Makefile.am @@ -9,18 +9,18 @@ dist_zedexec_SCRIPTS = \ %D%/all-debug.sh \ %D%/all-syslog.sh \ %D%/data-notify.sh \ - %D%/deadman-slot_off.sh \ + %D%/deadman-sync-slot_off.sh \ %D%/generic-notify.sh \ - %D%/pool_import-led.sh \ + %D%/pool_import-sync-led.sh \ %D%/resilver_finish-notify.sh \ %D%/resilver_finish-start-scrub.sh \ %D%/scrub_finish-notify.sh \ - %D%/statechange-led.sh \ + %D%/statechange-sync-led.sh \ %D%/statechange-notify.sh \ - %D%/statechange-slot_off.sh \ + %D%/statechange-sync-slot_off.sh \ %D%/trim_finish-notify.sh \ - %D%/vdev_attach-led.sh \ - %D%/vdev_clear-led.sh + %D%/vdev_attach-sync-led.sh \ + %D%/vdev_clear-sync-led.sh nodist_zedexec_SCRIPTS = \ %D%/history_event-zfs-list-cacher.sh @@ -30,17 +30,17 @@ SUBSTFILES += $(nodist_zedexec_SCRIPTS) zedconfdefaults = \ all-syslog.sh \ data-notify.sh \ - deadman-slot_off.sh \ + deadman-sync-slot_off.sh \ history_event-zfs-list-cacher.sh \ - pool_import-led.sh \ + pool_import-sync-led.sh \ resilver_finish-notify.sh \ resilver_finish-start-scrub.sh \ scrub_finish-notify.sh \ - statechange-led.sh \ + statechange-sync-led.sh \ statechange-notify.sh \ - statechange-slot_off.sh \ - vdev_attach-led.sh \ - vdev_clear-led.sh + statechange-sync-slot_off.sh \ + vdev_attach-sync-led.sh \ + vdev_clear-sync-led.sh dist_noinst_DATA += %D%/README diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/deadman-slot_off.sh b/sys/contrib/openzfs/cmd/zed/zed.d/deadman-sync-slot_off.sh index 7b339b3add01..7b339b3add01 100755 --- a/sys/contrib/openzfs/cmd/zed/zed.d/deadman-slot_off.sh +++ b/sys/contrib/openzfs/cmd/zed/zed.d/deadman-sync-slot_off.sh diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/pool_import-led.sh b/sys/contrib/openzfs/cmd/zed/zed.d/pool_import-led.sh deleted file mode 120000 index 7d7404398a4a..000000000000 --- a/sys/contrib/openzfs/cmd/zed/zed.d/pool_import-led.sh +++ /dev/null @@ -1 +0,0 @@ -statechange-led.sh
\ No newline at end of file diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/pool_import-sync-led.sh b/sys/contrib/openzfs/cmd/zed/zed.d/pool_import-sync-led.sh new file mode 120000 index 000000000000..8b9c10c11ebb --- /dev/null +++ b/sys/contrib/openzfs/cmd/zed/zed.d/pool_import-sync-led.sh @@ -0,0 +1 @@ +statechange-sync-led.sh
\ No newline at end of file diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/statechange-led.sh b/sys/contrib/openzfs/cmd/zed/zed.d/statechange-sync-led.sh index 40cb61f17307..40cb61f17307 100755 --- a/sys/contrib/openzfs/cmd/zed/zed.d/statechange-led.sh +++ b/sys/contrib/openzfs/cmd/zed/zed.d/statechange-sync-led.sh diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/statechange-slot_off.sh b/sys/contrib/openzfs/cmd/zed/zed.d/statechange-sync-slot_off.sh index 06acce93b8aa..06acce93b8aa 100755 --- a/sys/contrib/openzfs/cmd/zed/zed.d/statechange-slot_off.sh +++ b/sys/contrib/openzfs/cmd/zed/zed.d/statechange-sync-slot_off.sh diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/vdev_attach-led.sh b/sys/contrib/openzfs/cmd/zed/zed.d/vdev_attach-led.sh deleted file mode 120000 index 7d7404398a4a..000000000000 --- a/sys/contrib/openzfs/cmd/zed/zed.d/vdev_attach-led.sh +++ /dev/null @@ -1 +0,0 @@ -statechange-led.sh
\ No newline at end of file diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/vdev_attach-sync-led.sh b/sys/contrib/openzfs/cmd/zed/zed.d/vdev_attach-sync-led.sh new file mode 120000 index 000000000000..8b9c10c11ebb --- /dev/null +++ b/sys/contrib/openzfs/cmd/zed/zed.d/vdev_attach-sync-led.sh @@ -0,0 +1 @@ +statechange-sync-led.sh
\ No newline at end of file diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/vdev_clear-led.sh b/sys/contrib/openzfs/cmd/zed/zed.d/vdev_clear-led.sh deleted file mode 120000 index 7d7404398a4a..000000000000 --- a/sys/contrib/openzfs/cmd/zed/zed.d/vdev_clear-led.sh +++ /dev/null @@ -1 +0,0 @@ -statechange-led.sh
\ No newline at end of file diff --git a/sys/contrib/openzfs/cmd/zed/zed.d/vdev_clear-sync-led.sh b/sys/contrib/openzfs/cmd/zed/zed.d/vdev_clear-sync-led.sh new file mode 120000 index 000000000000..8b9c10c11ebb --- /dev/null +++ b/sys/contrib/openzfs/cmd/zed/zed.d/vdev_clear-sync-led.sh @@ -0,0 +1 @@ +statechange-sync-led.sh
\ No newline at end of file diff --git a/sys/contrib/openzfs/cmd/zed/zed_exec.c b/sys/contrib/openzfs/cmd/zed/zed_exec.c index 036081decd64..a14af4f20a85 100644 --- a/sys/contrib/openzfs/cmd/zed/zed_exec.c +++ b/sys/contrib/openzfs/cmd/zed/zed_exec.c @@ -196,37 +196,29 @@ _nop(int sig) (void) sig; } -static void * -_reap_children(void *arg) +static void +wait_for_children(boolean_t do_pause, boolean_t wait) { - (void) arg; - struct launched_process_node node, *pnode; pid_t pid; - int status; struct rusage usage; - struct sigaction sa = {}; - - (void) sigfillset(&sa.sa_mask); - (void) sigdelset(&sa.sa_mask, SIGCHLD); - (void) pthread_sigmask(SIG_SETMASK, &sa.sa_mask, NULL); - - (void) sigemptyset(&sa.sa_mask); - sa.sa_handler = _nop; - sa.sa_flags = SA_NOCLDSTOP; - (void) sigaction(SIGCHLD, &sa, NULL); + int status; + struct launched_process_node node, *pnode; for (_reap_children_stop = B_FALSE; !_reap_children_stop; ) { (void) pthread_mutex_lock(&_launched_processes_lock); - pid = wait4(0, &status, WNOHANG, &usage); - + pid = wait4(0, &status, wait ? 0 : WNOHANG, &usage); if (pid == 0 || pid == (pid_t)-1) { (void) pthread_mutex_unlock(&_launched_processes_lock); - if (pid == 0 || errno == ECHILD) - pause(); - else if (errno != EINTR) + if ((pid == 0) || (errno == ECHILD)) { + if (do_pause) + pause(); + } else if (errno != EINTR) zed_log_msg(LOG_WARNING, "Failed to wait for children: %s", strerror(errno)); + if (!do_pause) + return; + } else { memset(&node, 0, sizeof (node)); node.pid = pid; @@ -278,6 +270,25 @@ _reap_children(void *arg) } } +} + +static void * +_reap_children(void *arg) +{ + (void) arg; + struct sigaction sa = {}; + + (void) sigfillset(&sa.sa_mask); + (void) sigdelset(&sa.sa_mask, SIGCHLD); + (void) pthread_sigmask(SIG_SETMASK, &sa.sa_mask, NULL); + + (void) sigemptyset(&sa.sa_mask); + sa.sa_handler = _nop; + sa.sa_flags = SA_NOCLDSTOP; + (void) sigaction(SIGCHLD, &sa, NULL); + + wait_for_children(B_TRUE, B_FALSE); + return (NULL); } @@ -307,6 +318,45 @@ zed_exec_fini(void) } /* + * Check if the zedlet name indicates if it is a synchronous zedlet + * + * Synchronous zedlets have a "-sync-" immediately following the event name in + * their zedlet filename, like: + * + * EVENT_NAME-sync-ZEDLETNAME.sh + * + * For example, if you wanted a synchronous statechange script: + * + * statechange-sync-myzedlet.sh + * + * Synchronous zedlets are guaranteed to be the only zedlet running. No other + * zedlets may run in parallel with a synchronous zedlet. A synchronous + * zedlet will wait for all previously spawned zedlets to finish before running. + * Users should be careful to only use synchronous zedlets when needed, since + * they decrease parallelism. + */ +static boolean_t +zedlet_is_sync(const char *zedlet, const char *event) +{ + const char *sync_str = "-sync-"; + size_t sync_str_len; + size_t zedlet_len; + size_t event_len; + + sync_str_len = strlen(sync_str); + zedlet_len = strlen(zedlet); + event_len = strlen(event); + + if (event_len + sync_str_len >= zedlet_len) + return (B_FALSE); + + if (strncmp(&zedlet[event_len], sync_str, sync_str_len) == 0) + return (B_TRUE); + + return (B_FALSE); +} + +/* * Process the event [eid] by synchronously invoking all zedlets with a * matching class prefix. * @@ -368,9 +418,28 @@ zed_exec_process(uint64_t eid, const char *class, const char *subclass, z = zed_strings_next(zcp->zedlets)) { for (csp = class_strings; *csp; csp++) { n = strlen(*csp); - if ((strncmp(z, *csp, n) == 0) && !isalpha(z[n])) + if ((strncmp(z, *csp, n) == 0) && !isalpha(z[n])) { + boolean_t is_sync = zedlet_is_sync(z, *csp); + + if (is_sync) { + /* + * Wait for previous zedlets to + * finish + */ + wait_for_children(B_FALSE, B_TRUE); + } + _zed_exec_fork_child(eid, zcp->zedlet_dir, z, e, zcp->zevent_fd, zcp->do_foreground); + + if (is_sync) { + /* + * Wait for sync zedlet we just launched + * to finish. + */ + wait_for_children(B_FALSE, B_TRUE); + } + } } } free(e); diff --git a/sys/contrib/openzfs/cmd/zfs/zfs_main.c b/sys/contrib/openzfs/cmd/zfs/zfs_main.c index 235f011af953..ccdd5ffef8e6 100644 --- a/sys/contrib/openzfs/cmd/zfs/zfs_main.c +++ b/sys/contrib/openzfs/cmd/zfs/zfs_main.c @@ -914,7 +914,11 @@ zfs_do_clone(int argc, char **argv) log_history = B_FALSE; } - ret = zfs_mount_and_share(g_zfs, argv[1], ZFS_TYPE_DATASET); + /* + * Dataset cloned successfully, mount/share failures are + * non-fatal. + */ + (void) zfs_mount_and_share(g_zfs, argv[1], ZFS_TYPE_DATASET); } zfs_close(zhp); @@ -930,19 +934,15 @@ usage: } /* - * Return a default volblocksize for the pool which always uses more than - * half of the data sectors. This primarily applies to dRAID which always - * writes full stripe widths. + * Calculate the minimum allocation size based on the top-level vdevs. */ static uint64_t -default_volblocksize(zpool_handle_t *zhp, nvlist_t *props) +calculate_volblocksize(nvlist_t *config) { - uint64_t volblocksize, asize = SPA_MINBLOCKSIZE; + uint64_t asize = SPA_MINBLOCKSIZE; nvlist_t *tree, **vdevs; uint_t nvdevs; - nvlist_t *config = zpool_get_config(zhp, NULL); - if (nvlist_lookup_nvlist(config, ZPOOL_CONFIG_VDEV_TREE, &tree) != 0 || nvlist_lookup_nvlist_array(tree, ZPOOL_CONFIG_CHILDREN, &vdevs, &nvdevs) != 0) { @@ -973,6 +973,24 @@ default_volblocksize(zpool_handle_t *zhp, nvlist_t *props) } } + return (asize); +} + +/* + * Return a default volblocksize for the pool which always uses more than + * half of the data sectors. This primarily applies to dRAID which always + * writes full stripe widths. + */ +static uint64_t +default_volblocksize(zpool_handle_t *zhp, nvlist_t *props) +{ + uint64_t volblocksize, asize = SPA_MINBLOCKSIZE; + + nvlist_t *config = zpool_get_config(zhp, NULL); + + if (nvlist_lookup_uint64(config, ZPOOL_CONFIG_MAX_ALLOC, &asize) != 0) + asize = calculate_volblocksize(config); + /* * Calculate the target volblocksize such that more than half * of the asize is used. The following table is for 4k sectors. @@ -1319,7 +1337,9 @@ zfs_do_create(int argc, char **argv) goto error; } - ret = zfs_mount_and_share(g_zfs, argv[0], ZFS_TYPE_DATASET); + /* Dataset created successfully, mount/share failures are non-fatal */ + ret = 0; + (void) zfs_mount_and_share(g_zfs, argv[0], ZFS_TYPE_DATASET); error: nvlist_free(props); return (ret); @@ -5303,6 +5323,7 @@ zfs_do_receive(int argc, char **argv) #define ZFS_DELEG_PERM_MOUNT "mount" #define ZFS_DELEG_PERM_SHARE "share" #define ZFS_DELEG_PERM_SEND "send" +#define ZFS_DELEG_PERM_SEND_RAW "send:raw" #define ZFS_DELEG_PERM_RECEIVE "receive" #define ZFS_DELEG_PERM_RECEIVE_APPEND "receive:append" #define ZFS_DELEG_PERM_ALLOW "allow" @@ -5345,6 +5366,7 @@ static zfs_deleg_perm_tab_t zfs_deleg_perm_tbl[] = { { ZFS_DELEG_PERM_RENAME, ZFS_DELEG_NOTE_RENAME }, { ZFS_DELEG_PERM_ROLLBACK, ZFS_DELEG_NOTE_ROLLBACK }, { ZFS_DELEG_PERM_SEND, ZFS_DELEG_NOTE_SEND }, + { ZFS_DELEG_PERM_SEND_RAW, ZFS_DELEG_NOTE_SEND_RAW }, { ZFS_DELEG_PERM_SHARE, ZFS_DELEG_NOTE_SHARE }, { ZFS_DELEG_PERM_SNAPSHOT, ZFS_DELEG_NOTE_SNAPSHOT }, { ZFS_DELEG_PERM_BOOKMARK, ZFS_DELEG_NOTE_BOOKMARK }, @@ -5929,6 +5951,10 @@ deleg_perm_comment(zfs_deleg_note_t note) case ZFS_DELEG_NOTE_SEND: str = gettext(""); break; + case ZFS_DELEG_NOTE_SEND_RAW: + str = gettext("Allow sending ONLY encrypted (raw) replication" + "\n\t\t\t\tstreams"); + break; case ZFS_DELEG_NOTE_SHARE: str = gettext("Allows sharing file systems over NFS or SMB" "\n\t\t\t\tprotocols"); @@ -6858,17 +6884,17 @@ print_holds(boolean_t scripted, int nwidth, int tagwidth, nvlist_t *nvl, if (scripted) { if (parsable) { - (void) printf("%s\t%s\t%ld\n", zname, - tagname, (unsigned long)time); + (void) printf("%s\t%s\t%lld\n", zname, + tagname, (long long)time); } else { (void) printf("%s\t%s\t%s\n", zname, tagname, tsbuf); } } else { if (parsable) { - (void) printf("%-*s %-*s %ld\n", + (void) printf("%-*s %-*s %lld\n", nwidth, zname, tagwidth, - tagname, (unsigned long)time); + tagname, (long long)time); } else { (void) printf("%-*s %-*s %s\n", nwidth, zname, tagwidth, diff --git a/sys/contrib/openzfs/cmd/zhack.c b/sys/contrib/openzfs/cmd/zhack.c index 2bd3051dce7b..8ffbf91ffb30 100644 --- a/sys/contrib/openzfs/cmd/zhack.c +++ b/sys/contrib/openzfs/cmd/zhack.c @@ -54,6 +54,7 @@ #include <sys/dmu_tx.h> #include <zfeature_common.h> #include <libzutil.h> +#include <sys/metaslab_impl.h> static importargs_t g_importargs; static char *g_pool; @@ -69,7 +70,8 @@ static __attribute__((noreturn)) void usage(void) { (void) fprintf(stderr, - "Usage: zhack [-c cachefile] [-d dir] <subcommand> <args> ...\n" + "Usage: zhack [-o tunable] [-c cachefile] [-d dir] <subcommand> " + "<args> ...\n" "where <subcommand> <args> is one of the following:\n" "\n"); @@ -93,7 +95,10 @@ usage(void) " -c repair corrupted label checksums\n" " -u restore the label on a detached device\n" "\n" - " <device> : path to vdev\n"); + " <device> : path to vdev\n" + "\n" + " metaslab leak <pool>\n" + " apply allocation map from zdb to specified pool\n"); exit(1); } @@ -363,10 +368,12 @@ feature_incr_sync(void *arg, dmu_tx_t *tx) zfeature_info_t *feature = arg; uint64_t refcount; + mutex_enter(&spa->spa_feat_stats_lock); VERIFY0(feature_get_refcount_from_disk(spa, feature, &refcount)); feature_sync(spa, feature, refcount + 1, tx); spa_history_log_internal(spa, "zhack feature incr", tx, "name=%s", feature->fi_guid); + mutex_exit(&spa->spa_feat_stats_lock); } static void @@ -376,10 +383,12 @@ feature_decr_sync(void *arg, dmu_tx_t *tx) zfeature_info_t *feature = arg; uint64_t refcount; + mutex_enter(&spa->spa_feat_stats_lock); VERIFY0(feature_get_refcount_from_disk(spa, feature, &refcount)); feature_sync(spa, feature, refcount - 1, tx); spa_history_log_internal(spa, "zhack feature decr", tx, "name=%s", feature->fi_guid); + mutex_exit(&spa->spa_feat_stats_lock); } static void @@ -496,6 +505,186 @@ zhack_do_feature(int argc, char **argv) return (0); } +static boolean_t +strstarts(const char *a, const char *b) +{ + return (strncmp(a, b, strlen(b)) == 0); +} + +static void +metaslab_force_alloc(metaslab_t *msp, uint64_t start, uint64_t size, + dmu_tx_t *tx) +{ + ASSERT(msp->ms_disabled); + ASSERT(MUTEX_HELD(&msp->ms_lock)); + uint64_t txg = dmu_tx_get_txg(tx); + + uint64_t off = start; + while (off < start + size) { + uint64_t ostart, osize; + boolean_t found = zfs_range_tree_find_in(msp->ms_allocatable, + off, start + size - off, &ostart, &osize); + if (!found) + break; + zfs_range_tree_remove(msp->ms_allocatable, ostart, osize); + + if (zfs_range_tree_is_empty(msp->ms_allocating[txg & TXG_MASK])) + vdev_dirty(msp->ms_group->mg_vd, VDD_METASLAB, msp, + txg); + + zfs_range_tree_add(msp->ms_allocating[txg & TXG_MASK], ostart, + osize); + msp->ms_allocating_total += osize; + off = ostart + osize; + } +} + +static void +zhack_do_metaslab_leak(int argc, char **argv) +{ + int c; + char *target; + spa_t *spa; + + optind = 1; + boolean_t force = B_FALSE; + while ((c = getopt(argc, argv, "f")) != -1) { + switch (c) { + case 'f': + force = B_TRUE; + break; + default: + usage(); + break; + } + } + + argc -= optind; + argv += optind; + + if (argc < 1) { + (void) fprintf(stderr, "error: missing pool name\n"); + usage(); + } + target = argv[0]; + + zhack_spa_open(target, B_FALSE, FTAG, &spa); + spa_config_enter(spa, SCL_VDEV | SCL_ALLOC, FTAG, RW_READER); + + char *line = NULL; + size_t cap = 0; + + vdev_t *vd = NULL; + metaslab_t *prev = NULL; + dmu_tx_t *tx = NULL; + while (getline(&line, &cap, stdin) > 0) { + if (strstarts(line, "\tvdev ")) { + uint64_t vdev_id, ms_shift; + if (sscanf(line, + "\tvdev %10"PRIu64"\t%*s metaslab shift %4"PRIu64, + &vdev_id, &ms_shift) == 1) { + VERIFY3U(sscanf(line, "\tvdev %"PRIu64 + "\t metaslab shift %4"PRIu64, + &vdev_id, &ms_shift), ==, 2); + } + vd = vdev_lookup_top(spa, vdev_id); + if (vd == NULL) { + fprintf(stderr, "error: no such vdev with " + "id %"PRIu64"\n", vdev_id); + break; + } + if (tx) { + dmu_tx_commit(tx); + mutex_exit(&prev->ms_lock); + metaslab_enable(prev, B_FALSE, B_FALSE); + tx = NULL; + prev = NULL; + } + if (vd->vdev_ms_shift != ms_shift) { + fprintf(stderr, "error: ms_shift mismatch: %" + PRIu64" != %"PRIu64"\n", vd->vdev_ms_shift, + ms_shift); + break; + } + } else if (strstarts(line, "\tmetaslabs ")) { + uint64_t ms_count; + VERIFY3U(sscanf(line, "\tmetaslabs %"PRIu64, &ms_count), + ==, 1); + ASSERT(vd); + if (!force && vd->vdev_ms_count != ms_count) { + fprintf(stderr, "error: ms_count mismatch: %" + PRIu64" != %"PRIu64"\n", vd->vdev_ms_count, + ms_count); + break; + } + } else if (strstarts(line, "ALLOC:")) { + uint64_t start, size; + VERIFY3U(sscanf(line, "ALLOC: %"PRIu64" %"PRIu64"\n", + &start, &size), ==, 2); + + ASSERT(vd); + metaslab_t *cur = + vd->vdev_ms[start >> vd->vdev_ms_shift]; + if (prev != cur) { + if (prev) { + dmu_tx_commit(tx); + mutex_exit(&prev->ms_lock); + metaslab_enable(prev, B_FALSE, B_FALSE); + } + ASSERT(cur); + metaslab_disable(cur); + mutex_enter(&cur->ms_lock); + metaslab_load(cur); + prev = cur; + tx = dmu_tx_create_dd( + spa_get_dsl(vd->vdev_spa)->dp_root_dir); + dmu_tx_assign(tx, DMU_TX_WAIT); + } + + metaslab_force_alloc(cur, start, size, tx); + } else { + continue; + } + } + if (tx) { + dmu_tx_commit(tx); + mutex_exit(&prev->ms_lock); + metaslab_enable(prev, B_FALSE, B_FALSE); + tx = NULL; + prev = NULL; + } + if (line) + free(line); + + spa_config_exit(spa, SCL_VDEV | SCL_ALLOC, FTAG); + spa_close(spa, FTAG); +} + +static int +zhack_do_metaslab(int argc, char **argv) +{ + char *subcommand; + + argc--; + argv++; + if (argc == 0) { + (void) fprintf(stderr, + "error: no metaslab operation specified\n"); + usage(); + } + + subcommand = argv[0]; + if (strcmp(subcommand, "leak") == 0) { + zhack_do_metaslab_leak(argc, argv); + } else { + (void) fprintf(stderr, "error: unknown subcommand: %s\n", + subcommand); + usage(); + } + + return (0); +} + #define ASHIFT_UBERBLOCK_SHIFT(ashift) \ MIN(MAX(ashift, UBERBLOCK_SHIFT), \ MAX_UBERBLOCK_SHIFT) @@ -525,6 +714,23 @@ zhack_repair_read_label(const int fd, vdev_label_t *vl, return (0); } +static int +zhack_repair_get_byteswap(const zio_eck_t *vdev_eck, const int l, int *byteswap) +{ + if (vdev_eck->zec_magic == ZEC_MAGIC) { + *byteswap = B_FALSE; + } else if (vdev_eck->zec_magic == BSWAP_64((uint64_t)ZEC_MAGIC)) { + *byteswap = B_TRUE; + } else { + (void) fprintf(stderr, "error: label %d: " + "Expected the nvlist checksum magic number but instead got " + "0x%" PRIx64 "\n", + l, vdev_eck->zec_magic); + return (1); + } + return (0); +} + static void zhack_repair_calc_cksum(const int byteswap, void *data, const uint64_t offset, const uint64_t abdsize, zio_eck_t *eck, zio_cksum_t *cksum) @@ -551,33 +757,10 @@ zhack_repair_calc_cksum(const int byteswap, void *data, const uint64_t offset, } static int -zhack_repair_check_label(uberblock_t *ub, const int l, const char **cfg_keys, - const size_t cfg_keys_len, nvlist_t *cfg, nvlist_t *vdev_tree_cfg, - uint64_t *ashift) +zhack_repair_get_ashift(nvlist_t *cfg, const int l, uint64_t *ashift) { int err; - - if (ub->ub_txg != 0) { - (void) fprintf(stderr, - "error: label %d: UB TXG of 0 expected, but got %" - PRIu64 "\n", - l, ub->ub_txg); - (void) fprintf(stderr, "It would appear the device was not " - "properly removed.\n"); - return (1); - } - - for (int i = 0; i < cfg_keys_len; i++) { - uint64_t val; - err = nvlist_lookup_uint64(cfg, cfg_keys[i], &val); - if (err) { - (void) fprintf(stderr, - "error: label %d, %d: " - "cannot find nvlist key %s\n", - l, i, cfg_keys[i]); - return (err); - } - } + nvlist_t *vdev_tree_cfg; err = nvlist_lookup_nvlist(cfg, ZPOOL_CONFIG_VDEV_TREE, &vdev_tree_cfg); @@ -601,7 +784,7 @@ zhack_repair_check_label(uberblock_t *ub, const int l, const char **cfg_keys, (void) fprintf(stderr, "error: label %d: nvlist key %s is zero\n", l, ZPOOL_CONFIG_ASHIFT); - return (err); + return (1); } return (0); @@ -616,30 +799,35 @@ zhack_repair_undetach(uberblock_t *ub, nvlist_t *cfg, const int l) */ if (BP_GET_LOGICAL_BIRTH(&ub->ub_rootbp) != 0) { const uint64_t txg = BP_GET_LOGICAL_BIRTH(&ub->ub_rootbp); + int err; + ub->ub_txg = txg; - if (nvlist_remove_all(cfg, ZPOOL_CONFIG_CREATE_TXG) != 0) { + err = nvlist_remove_all(cfg, ZPOOL_CONFIG_CREATE_TXG); + if (err) { (void) fprintf(stderr, "error: label %d: " "Failed to remove pool creation TXG\n", l); - return (1); + return (err); } - if (nvlist_remove_all(cfg, ZPOOL_CONFIG_POOL_TXG) != 0) { + err = nvlist_remove_all(cfg, ZPOOL_CONFIG_POOL_TXG); + if (err) { (void) fprintf(stderr, "error: label %d: Failed to remove pool TXG to " "be replaced.\n", l); - return (1); + return (err); } - if (nvlist_add_uint64(cfg, ZPOOL_CONFIG_POOL_TXG, txg) != 0) { + err = nvlist_add_uint64(cfg, ZPOOL_CONFIG_POOL_TXG, txg); + if (err) { (void) fprintf(stderr, "error: label %d: " "Failed to add pool TXG of %" PRIu64 "\n", l, txg); - return (1); + return (err); } } @@ -733,6 +921,7 @@ zhack_repair_test_cksum(const int byteswap, void *vdev_data, BSWAP_64(ZEC_MAGIC) : ZEC_MAGIC; const uint64_t actual_magic = vdev_eck->zec_magic; int err = 0; + if (actual_magic != expected_magic) { (void) fprintf(stderr, "error: label %d: " "Expected " @@ -754,6 +943,36 @@ zhack_repair_test_cksum(const int byteswap, void *vdev_data, return (err); } +static int +zhack_repair_unpack_cfg(vdev_label_t *vl, const int l, nvlist_t **cfg) +{ + const char *cfg_keys[] = { ZPOOL_CONFIG_VERSION, + ZPOOL_CONFIG_POOL_STATE, ZPOOL_CONFIG_GUID }; + int err; + + err = nvlist_unpack(vl->vl_vdev_phys.vp_nvlist, + VDEV_PHYS_SIZE - sizeof (zio_eck_t), cfg, 0); + if (err) { + (void) fprintf(stderr, + "error: cannot unpack nvlist label %d\n", l); + return (err); + } + + for (int i = 0; i < ARRAY_SIZE(cfg_keys); i++) { + uint64_t val; + err = nvlist_lookup_uint64(*cfg, cfg_keys[i], &val); + if (err) { + (void) fprintf(stderr, + "error: label %d, %d: " + "cannot find nvlist key %s\n", + l, i, cfg_keys[i]); + return (err); + } + } + + return (0); +} + static void zhack_repair_one_label(const zhack_repair_op_t op, const int fd, vdev_label_t *vl, const uint64_t label_offset, const int l, @@ -767,10 +986,7 @@ zhack_repair_one_label(const zhack_repair_op_t op, const int fd, (zio_eck_t *)((char *)(vdev_data) + VDEV_PHYS_SIZE) - 1; const uint64_t vdev_phys_offset = label_offset + offsetof(vdev_label_t, vl_vdev_phys); - const char *cfg_keys[] = { ZPOOL_CONFIG_VERSION, - ZPOOL_CONFIG_POOL_STATE, ZPOOL_CONFIG_GUID }; nvlist_t *cfg; - nvlist_t *vdev_tree_cfg = NULL; uint64_t ashift; int byteswap; @@ -778,18 +994,9 @@ zhack_repair_one_label(const zhack_repair_op_t op, const int fd, if (err) return; - if (vdev_eck->zec_magic == 0) { - (void) fprintf(stderr, "error: label %d: " - "Expected the nvlist checksum magic number to not be zero" - "\n", - l); - (void) fprintf(stderr, "There should already be a checksum " - "for the label.\n"); + err = zhack_repair_get_byteswap(vdev_eck, l, &byteswap); + if (err) return; - } - - byteswap = - (vdev_eck->zec_magic == BSWAP_64((uint64_t)ZEC_MAGIC)); if (byteswap) { byteswap_uint64_array(&vdev_eck->zec_cksum, @@ -805,16 +1012,7 @@ zhack_repair_one_label(const zhack_repair_op_t op, const int fd, return; } - err = nvlist_unpack(vl->vl_vdev_phys.vp_nvlist, - VDEV_PHYS_SIZE - sizeof (zio_eck_t), &cfg, 0); - if (err) { - (void) fprintf(stderr, - "error: cannot unpack nvlist label %d\n", l); - return; - } - - err = zhack_repair_check_label(ub, - l, cfg_keys, ARRAY_SIZE(cfg_keys), cfg, vdev_tree_cfg, &ashift); + err = zhack_repair_unpack_cfg(vl, l, &cfg); if (err) return; @@ -822,6 +1020,19 @@ zhack_repair_one_label(const zhack_repair_op_t op, const int fd, char *buf; size_t buflen; + if (ub->ub_txg != 0) { + (void) fprintf(stderr, + "error: label %d: UB TXG of 0 expected, but got %" + PRIu64 "\n", l, ub->ub_txg); + (void) fprintf(stderr, "It would appear the device was " + "not properly detached.\n"); + return; + } + + err = zhack_repair_get_ashift(cfg, l, &ashift); + if (err) + return; + err = zhack_repair_undetach(ub, cfg, l); if (err) return; @@ -981,7 +1192,7 @@ main(int argc, char **argv) dprintf_setup(&argc, argv); zfs_prop_init(); - while ((c = getopt(argc, argv, "+c:d:")) != -1) { + while ((c = getopt(argc, argv, "+c:d:o:")) != -1) { switch (c) { case 'c': g_importargs.cachefile = optarg; @@ -990,6 +1201,10 @@ main(int argc, char **argv) assert(g_importargs.paths < MAX_NUM_PATHS); g_importargs.path[g_importargs.paths++] = optarg; break; + case 'o': + if (handle_tunable_option(optarg, B_FALSE) != 0) + exit(1); + break; default: usage(); break; @@ -1011,6 +1226,8 @@ main(int argc, char **argv) rv = zhack_do_feature(argc, argv); } else if (strcmp(subcommand, "label") == 0) { return (zhack_do_label(argc, argv)); + } else if (strcmp(subcommand, "metaslab") == 0) { + rv = zhack_do_metaslab(argc, argv); } else { (void) fprintf(stderr, "error: unknown subcommand: %s\n", subcommand); diff --git a/sys/contrib/openzfs/cmd/zinject/zinject.c b/sys/contrib/openzfs/cmd/zinject/zinject.c index 113797c878b9..c2f646f2567d 100644 --- a/sys/contrib/openzfs/cmd/zinject/zinject.c +++ b/sys/contrib/openzfs/cmd/zinject/zinject.c @@ -107,6 +107,8 @@ * zinject * zinject <-a | -u pool> * zinject -c <id|all> + * zinject -E <delay> [-a] [-m] [-f freq] [-l level] [-r range] + * [-T iotype] [-t type object | -b bookmark pool] * zinject [-q] <-t type> [-f freq] [-u] [-a] [-m] [-e errno] [-l level] * [-r range] <object> * zinject [-f freq] [-a] [-m] [-u] -b objset:object:level:start:end pool @@ -132,14 +134,18 @@ * The '-f' flag controls the frequency of errors injected, expressed as a * real number percentage between 0.0001 and 100. The default is 100. * - * The this form is responsible for actually injecting the handler into the + * The <object> form is responsible for actually injecting the handler into the * framework. It takes the arguments described above, translates them to the * internal tuple using libzpool, and then issues an ioctl() to register the * handler. * - * The final form can target a specific bookmark, regardless of whether a + * The '-b' option can target a specific bookmark, regardless of whether a * human-readable interface has been designed. It allows developers to specify * a particular block by number. + * + * The '-E' option injects pipeline ready stage delays for the given object or + * bookmark. The delay is specified in milliseconds, and it supports I/O type + * and range filters. */ #include <errno.h> @@ -346,6 +352,13 @@ usage(void) "\t\tsuch that the operation takes a minimum of supplied seconds\n" "\t\tto complete.\n" "\n" + "\tzinject -E <delay> [-a] [-m] [-f freq] [-l level] [-r range]\n" + "\t\t[-T iotype] [-t type object | -b bookmark pool]\n" + "\n" + "\t\tInject pipeline ready stage delays for the given object path\n" + "\t\t(data or dnode) or raw bookmark. The delay is specified in\n" + "\t\tmilliseconds.\n" + "\n" "\tzinject -I [-s <seconds> | -g <txgs>] pool\n" "\t\tCause the pool to stop writing blocks yet not\n" "\t\treport errors for a duration. Simulates buggy hardware\n" @@ -724,12 +737,15 @@ register_handler(const char *pool, int flags, zinject_record_t *record, if (quiet) { (void) printf("%llu\n", (u_longlong_t)zc.zc_guid); } else { + boolean_t show_object = B_FALSE; + boolean_t show_iotype = B_FALSE; (void) printf("Added handler %llu with the following " "properties:\n", (u_longlong_t)zc.zc_guid); (void) printf(" pool: %s\n", pool); if (record->zi_guid) { (void) printf(" vdev: %llx\n", (u_longlong_t)record->zi_guid); + show_iotype = B_TRUE; } else if (record->zi_func[0] != '\0') { (void) printf(" panic function: %s\n", record->zi_func); @@ -742,7 +758,18 @@ register_handler(const char *pool, int flags, zinject_record_t *record, } else if (record->zi_timer > 0) { (void) printf(" timer: %lld ms\n", (u_longlong_t)NSEC2MSEC(record->zi_timer)); + if (record->zi_cmd == ZINJECT_DELAY_READY) { + show_object = B_TRUE; + show_iotype = B_TRUE; + } } else { + show_object = B_TRUE; + } + if (show_iotype) { + (void) printf("iotype: %s\n", + iotype_to_str(record->zi_iotype)); + } + if (show_object) { (void) printf("objset: %llu\n", (u_longlong_t)record->zi_objset); (void) printf("object: %llu\n", @@ -910,6 +937,7 @@ main(int argc, char **argv) int ret; int flags = 0; uint32_t dvas = 0; + hrtime_t ready_delay = -1; if ((g_zfs = libzfs_init()) == NULL) { (void) fprintf(stderr, "%s\n", libzfs_error_init(errno)); @@ -940,7 +968,7 @@ main(int argc, char **argv) } while ((c = getopt(argc, argv, - ":aA:b:C:d:D:f:Fg:qhIc:t:T:l:mr:s:e:uL:p:P:")) != -1) { + ":aA:b:C:d:D:E:f:Fg:qhIc:t:T:l:mr:s:e:uL:p:P:")) != -1) { switch (c) { case 'a': flags |= ZINJECT_FLUSH_ARC; @@ -1113,6 +1141,18 @@ main(int argc, char **argv) case 'u': flags |= ZINJECT_UNLOAD_SPA; break; + case 'E': + ready_delay = MSEC2NSEC(strtol(optarg, &end, 10)); + if (ready_delay <= 0 || *end != '\0') { + (void) fprintf(stderr, "invalid delay '%s': " + "must be a positive duration\n", optarg); + usage(); + libzfs_fini(g_zfs); + return (1); + } + record.zi_cmd = ZINJECT_DELAY_READY; + record.zi_timer = ready_delay; + break; case 'L': if ((label = name_to_type(optarg)) == TYPE_INVAL && !LABEL_TYPE(type)) { @@ -1150,7 +1190,7 @@ main(int argc, char **argv) */ if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED || - record.zi_freq > 0 || dvas != 0) { + record.zi_freq > 0 || dvas != 0 || ready_delay >= 0) { (void) fprintf(stderr, "cancel (-c) incompatible with " "any other options\n"); usage(); @@ -1186,7 +1226,7 @@ main(int argc, char **argv) */ if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || record.zi_cmd != ZINJECT_UNINITIALIZED || - dvas != 0) { + dvas != 0 || ready_delay >= 0) { (void) fprintf(stderr, "device (-d) incompatible with " "data error injection\n"); usage(); @@ -1276,13 +1316,23 @@ main(int argc, char **argv) return (1); } - record.zi_cmd = ZINJECT_DATA_FAULT; + if (record.zi_cmd == ZINJECT_UNINITIALIZED) { + record.zi_cmd = ZINJECT_DATA_FAULT; + if (!error) + error = EIO; + } else if (error != 0) { + (void) fprintf(stderr, "error type -e incompatible " + "with delay injection\n"); + libzfs_fini(g_zfs); + return (1); + } else { + record.zi_iotype = io_type; + } + if (translate_raw(raw, &record) != 0) { libzfs_fini(g_zfs); return (1); } - if (!error) - error = EIO; } else if (record.zi_cmd == ZINJECT_PANIC) { if (raw != NULL || range != NULL || type != TYPE_INVAL || level != 0 || device != NULL || record.zi_freq > 0 || @@ -1410,6 +1460,13 @@ main(int argc, char **argv) record.zi_dvas = dvas; } + if (record.zi_cmd != ZINJECT_UNINITIALIZED && error != 0) { + (void) fprintf(stderr, "error type -e incompatible " + "with delay injection\n"); + libzfs_fini(g_zfs); + return (1); + } + if (error == EACCES) { if (type != TYPE_DATA) { (void) fprintf(stderr, "decryption errors " @@ -1425,8 +1482,12 @@ main(int argc, char **argv) * not found. */ error = ECKSUM; - } else { + } else if (record.zi_cmd == ZINJECT_UNINITIALIZED) { record.zi_cmd = ZINJECT_DATA_FAULT; + if (!error) + error = EIO; + } else { + record.zi_iotype = io_type; } if (translate_record(type, argv[0], range, level, &record, pool, @@ -1434,8 +1495,6 @@ main(int argc, char **argv) libzfs_fini(g_zfs); return (1); } - if (!error) - error = EIO; } /* diff --git a/sys/contrib/openzfs/cmd/zpool/Makefile.am b/sys/contrib/openzfs/cmd/zpool/Makefile.am index 2f962408e5a3..5bb6d8160b18 100644 --- a/sys/contrib/openzfs/cmd/zpool/Makefile.am +++ b/sys/contrib/openzfs/cmd/zpool/Makefile.am @@ -148,6 +148,7 @@ dist_zpoolcompat_DATA = \ %D%/compatibility.d/openzfs-2.1-linux \ %D%/compatibility.d/openzfs-2.2 \ %D%/compatibility.d/openzfs-2.3 \ + %D%/compatibility.d/openzfs-2.4 \ %D%/compatibility.d/openzfsonosx-1.7.0 \ %D%/compatibility.d/openzfsonosx-1.8.1 \ %D%/compatibility.d/openzfsonosx-1.9.3 \ @@ -187,7 +188,9 @@ zpoolcompatlinks = \ "openzfs-2.2 openzfs-2.2-linux" \ "openzfs-2.2 openzfs-2.2-freebsd" \ "openzfs-2.3 openzfs-2.3-linux" \ - "openzfs-2.3 openzfs-2.3-freebsd" + "openzfs-2.3 openzfs-2.3-freebsd" \ + "openzfs-2.4 openzfs-2.4-linux" \ + "openzfs-2.4 openzfs-2.4-freebsd" zpoolconfdir = $(sysconfdir)/zfs/zpool.d INSTALL_DATA_HOOKS += zpool-install-data-hook diff --git a/sys/contrib/openzfs/cmd/zpool/compatibility.d/openzfs-2.4 b/sys/contrib/openzfs/cmd/zpool/compatibility.d/openzfs-2.4 new file mode 100644 index 000000000000..3fbd91014c95 --- /dev/null +++ b/sys/contrib/openzfs/cmd/zpool/compatibility.d/openzfs-2.4 @@ -0,0 +1,48 @@ +# Features supported by OpenZFS 2.4 on Linux and FreeBSD +allocation_classes +async_destroy +blake3 +block_cloning +block_cloning_endian +bookmark_v2 +bookmark_written +bookmarks +device_rebuild +device_removal +draid +dynamic_gang_header +edonr +embedded_data +empty_bpobj +enabled_txg +encryption +extensible_dataset +fast_dedup +filesystem_limits +head_errlog +hole_birth +large_blocks +large_dnode +large_microzap +livelist +log_spacemap +longname +lz4_compress +multi_vdev_crash_dump +obsolete_counts +physical_rewrite +project_quota +raidz_expansion +redacted_datasets +redaction_bookmarks +redaction_list_spill +resilver_defer +sha512 +skein +spacemap_histogram +spacemap_v2 +userobj_accounting +vdev_zaps_v2 +zilsaxattr +zpool_checkpoint +zstd_compress diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_iter.c b/sys/contrib/openzfs/cmd/zpool/zpool_iter.c index 2eec9a95e24c..fef602736705 100644 --- a/sys/contrib/openzfs/cmd/zpool/zpool_iter.c +++ b/sys/contrib/openzfs/cmd/zpool/zpool_iter.c @@ -26,6 +26,7 @@ /* * Copyright 2016 Igor Kozhukhov <ikozhukhov@gmail.com>. + * Copyright (c) 2025, Klara, Inc. */ #include <libintl.h> @@ -52,7 +53,7 @@ typedef struct zpool_node { zpool_handle_t *zn_handle; uu_avl_node_t zn_avlnode; - int zn_mark; + hrtime_t zn_last_refresh; } zpool_node_t; struct zpool_list { @@ -62,6 +63,7 @@ struct zpool_list { uu_avl_pool_t *zl_pool; zprop_list_t **zl_proplist; zfs_type_t zl_type; + hrtime_t zl_last_refresh; }; static int @@ -81,26 +83,30 @@ zpool_compare(const void *larg, const void *rarg, void *unused) * of known pools. */ static int -add_pool(zpool_handle_t *zhp, void *data) +add_pool(zpool_handle_t *zhp, zpool_list_t *zlp) { - zpool_list_t *zlp = data; - zpool_node_t *node = safe_malloc(sizeof (zpool_node_t)); + zpool_node_t *node, *new = safe_malloc(sizeof (zpool_node_t)); uu_avl_index_t idx; - node->zn_handle = zhp; - uu_avl_node_init(node, &node->zn_avlnode, zlp->zl_pool); - if (uu_avl_find(zlp->zl_avl, node, NULL, &idx) == NULL) { + new->zn_handle = zhp; + uu_avl_node_init(new, &new->zn_avlnode, zlp->zl_pool); + + node = uu_avl_find(zlp->zl_avl, new, NULL, &idx); + if (node == NULL) { if (zlp->zl_proplist && zpool_expand_proplist(zhp, zlp->zl_proplist, zlp->zl_type, zlp->zl_literal) != 0) { zpool_close(zhp); - free(node); + free(new); return (-1); } - uu_avl_insert(zlp->zl_avl, node, idx); + new->zn_last_refresh = zlp->zl_last_refresh; + uu_avl_insert(zlp->zl_avl, new, idx); } else { + zpool_refresh_stats_from_handle(node->zn_handle, zhp); + node->zn_last_refresh = zlp->zl_last_refresh; zpool_close(zhp); - free(node); + free(new); return (-1); } @@ -108,6 +114,18 @@ add_pool(zpool_handle_t *zhp, void *data) } /* + * add_pool(), but always returns 0. This allows zpool_iter() to continue + * even if a pool exists in the tree, or we fail to get the properties for + * a new one. + */ +static int +add_pool_cb(zpool_handle_t *zhp, void *data) +{ + (void) add_pool(zhp, data); + return (0); +} + +/* * Create a list of pools based on the given arguments. If we're given no * arguments, then iterate over all pools in the system and add them to the AVL * tree. Otherwise, add only those pool explicitly specified on the command @@ -135,9 +153,10 @@ pool_list_get(int argc, char **argv, zprop_list_t **proplist, zfs_type_t type, zlp->zl_type = type; zlp->zl_literal = literal; + zlp->zl_last_refresh = gethrtime(); if (argc == 0) { - (void) zpool_iter(g_zfs, add_pool, zlp); + (void) zpool_iter(g_zfs, add_pool_cb, zlp); zlp->zl_findall = B_TRUE; } else { int i; @@ -159,15 +178,61 @@ pool_list_get(int argc, char **argv, zprop_list_t **proplist, zfs_type_t type, } /* - * Search for any new pools, adding them to the list. We only add pools when no - * options were given on the command line. Otherwise, we keep the list fixed as - * those that were explicitly specified. + * Refresh the state of all pools on the list. Additionally, if no options were + * given on the command line, add any new pools and remove any that are no + * longer available. */ -void -pool_list_update(zpool_list_t *zlp) +int +pool_list_refresh(zpool_list_t *zlp) { - if (zlp->zl_findall) - (void) zpool_iter(g_zfs, add_pool, zlp); + zlp->zl_last_refresh = gethrtime(); + + if (!zlp->zl_findall) { + /* + * This list is a fixed list of pools, so we must not add + * or remove any. Just walk over them and refresh their + * state. + */ + int navail = 0; + for (zpool_node_t *node = uu_avl_first(zlp->zl_avl); + node != NULL; node = uu_avl_next(zlp->zl_avl, node)) { + boolean_t missing; + zpool_refresh_stats(node->zn_handle, &missing); + navail += !missing; + node->zn_last_refresh = zlp->zl_last_refresh; + } + return (navail); + } + + /* Search for any new pools and add them to the list. */ + (void) zpool_iter(g_zfs, add_pool_cb, zlp); + + /* Walk the list of existing pools, and update or remove them. */ + zpool_node_t *node, *next; + for (node = uu_avl_first(zlp->zl_avl); node != NULL; node = next) { + next = uu_avl_next(zlp->zl_avl, node); + + /* + * Skip any that were refreshed and are online; they were added + * by zpool_iter() and are already up to date. + */ + if (node->zn_last_refresh == zlp->zl_last_refresh && + zpool_get_state(node->zn_handle) != POOL_STATE_UNAVAIL) + continue; + + /* Refresh and remove if necessary. */ + boolean_t missing; + zpool_refresh_stats(node->zn_handle, &missing); + if (missing) { + uu_avl_remove(zlp->zl_avl, node); + zpool_close(node->zn_handle); + free(node); + } else { + node->zn_last_refresh = zlp->zl_last_refresh; + } + } + + return (uu_avl_numnodes(zlp->zl_avl)); } /* @@ -191,23 +256,6 @@ pool_list_iter(zpool_list_t *zlp, int unavail, zpool_iter_f func, } /* - * Remove the given pool from the list. When running iostat, we want to remove - * those pools that no longer exist. - */ -void -pool_list_remove(zpool_list_t *zlp, zpool_handle_t *zhp) -{ - zpool_node_t search, *node; - - search.zn_handle = zhp; - if ((node = uu_avl_find(zlp->zl_avl, &search, NULL, NULL)) != NULL) { - uu_avl_remove(zlp->zl_avl, node); - zpool_close(node->zn_handle); - free(node); - } -} - -/* * Free all the handles associated with this list. */ void diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_main.c b/sys/contrib/openzfs/cmd/zpool/zpool_main.c index 237e558da65b..1feec55c0e8b 100644 --- a/sys/contrib/openzfs/cmd/zpool/zpool_main.c +++ b/sys/contrib/openzfs/cmd/zpool/zpool_main.c @@ -33,7 +33,7 @@ * Copyright (c) 2017, Intel Corporation. * Copyright (c) 2019, loli10K <ezomori.nozomu@gmail.com> * Copyright (c) 2021, Colm Buckley <colm@tuatha.org> - * Copyright (c) 2021, 2023, Klara Inc. + * Copyright (c) 2021, 2023, 2025, Klara, Inc. * Copyright (c) 2021, 2025 Hewlett Packard Enterprise Development LP. */ @@ -456,7 +456,7 @@ get_usage(zpool_help_t idx) "<pool> <vdev> ...\n")); case HELP_ATTACH: return (gettext("\tattach [-fsw] [-o property=value] " - "<pool> <device> <new-device>\n")); + "<pool> <vdev> <new-device>\n")); case HELP_CLEAR: return (gettext("\tclear [[--power]|[-nF]] <pool> [device]\n")); case HELP_CREATE: @@ -5761,24 +5761,6 @@ children: return (ret); } -static int -refresh_iostat(zpool_handle_t *zhp, void *data) -{ - iostat_cbdata_t *cb = data; - boolean_t missing; - - /* - * If the pool has disappeared, remove it from the list and continue. - */ - if (zpool_refresh_stats(zhp, &missing) != 0) - return (-1); - - if (missing) - pool_list_remove(cb->cb_list, zhp); - - return (0); -} - /* * Callback to print out the iostats for the given pool. */ @@ -6359,15 +6341,14 @@ get_namewidth_iostat(zpool_handle_t *zhp, void *data) * This command can be tricky because we want to be able to deal with pool * creation/destruction as well as vdev configuration changes. The bulk of this * processing is handled by the pool_list_* routines in zpool_iter.c. We rely - * on pool_list_update() to detect the addition of new pools. Configuration - * changes are all handled within libzfs. + * on pool_list_refresh() to detect the addition and removal of pools. + * Configuration changes are all handled within libzfs. */ int zpool_do_iostat(int argc, char **argv) { int c; int ret; - int npools; float interval = 0; unsigned long count = 0; zpool_list_t *list; @@ -6618,10 +6599,24 @@ zpool_do_iostat(int argc, char **argv) return (1); } + int last_npools = 0; for (;;) { - if ((npools = pool_list_count(list)) == 0) + /* + * Refresh all pools in list, adding or removing pools as + * necessary. + */ + int npools = pool_list_refresh(list); + if (npools == 0) { (void) fprintf(stderr, gettext("no pools available\n")); - else { + } else { + /* + * If the list of pools has changed since last time + * around, reset the iteration count to force the + * header to be redisplayed. + */ + if (last_npools != npools) + cb.cb_iteration = 0; + /* * If this is the first iteration and -y was supplied * we skip any printing. @@ -6630,15 +6625,6 @@ zpool_do_iostat(int argc, char **argv) cb.cb_iteration == 0); /* - * Refresh all statistics. This is done as an - * explicit step before calculating the maximum name - * width, so that any * configuration changes are - * properly accounted for. - */ - (void) pool_list_iter(list, B_FALSE, refresh_iostat, - &cb); - - /* * Iterate over all pools to determine the maximum width * for the pool / device name column across all pools. */ @@ -6691,6 +6677,7 @@ zpool_do_iostat(int argc, char **argv) if (skip) { (void) fflush(stdout); (void) fsleep(interval); + last_npools = npools; continue; } @@ -6728,6 +6715,8 @@ zpool_do_iostat(int argc, char **argv) (void) fflush(stdout); (void) fsleep(interval); + + last_npools = npools; } pool_list_free(list); @@ -7644,7 +7633,7 @@ zpool_do_replace(int argc, char **argv) } /* - * zpool attach [-fsw] [-o property=value] <pool> <device>|<vdev> <new_device> + * zpool attach [-fsw] [-o property=value] <pool> <vdev> <new_device> * * -f Force attach, even if <new_device> appears to be in use. * -s Use sequential instead of healing reconstruction for resilver. @@ -7652,9 +7641,9 @@ zpool_do_replace(int argc, char **argv) * -w Wait for resilvering (mirror) or expansion (raidz) to complete * before returning. * - * Attach <new_device> to a <device> or <vdev>, where the vdev can be of type - * mirror or raidz. If <device> is not part of a mirror, then <device> will - * be transformed into a mirror of <device> and <new_device>. When a mirror + * Attach <new_device> to a <vdev>, where the vdev can be of type + * device, mirror or raidz. If <vdev> is not part of a mirror, then <vdev> will + * be transformed into a mirror of <vdev> and <new_device>. When a mirror * is involved, <new_device> will begin life with a DTL of [0, now], and will * immediately begin to resilver itself. For the raidz case, a expansion will * commence and reflow the raidz data across all the disks including the diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_util.h b/sys/contrib/openzfs/cmd/zpool/zpool_util.h index 5ab7cb9750f1..3af23c52bd45 100644 --- a/sys/contrib/openzfs/cmd/zpool/zpool_util.h +++ b/sys/contrib/openzfs/cmd/zpool/zpool_util.h @@ -76,11 +76,10 @@ typedef struct zpool_list zpool_list_t; zpool_list_t *pool_list_get(int, char **, zprop_list_t **, zfs_type_t, boolean_t, int *); -void pool_list_update(zpool_list_t *); +int pool_list_refresh(zpool_list_t *); int pool_list_iter(zpool_list_t *, int unavail, zpool_iter_f, void *); void pool_list_free(zpool_list_t *); int pool_list_count(zpool_list_t *); -void pool_list_remove(zpool_list_t *, zpool_handle_t *); extern libzfs_handle_t *g_zfs; diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_vdev.c b/sys/contrib/openzfs/cmd/zpool/zpool_vdev.c index 684b46a2d673..088c0108e911 100644 --- a/sys/contrib/openzfs/cmd/zpool/zpool_vdev.c +++ b/sys/contrib/openzfs/cmd/zpool/zpool_vdev.c @@ -609,22 +609,28 @@ get_replication(nvlist_t *nvroot, boolean_t fatal) ZPOOL_CONFIG_PATH, &path) == 0); /* + * Skip active spares they should never cause + * the pool to be evaluated as inconsistent. + */ + if (is_spare(NULL, path)) + continue; + + /* * If we have a raidz/mirror that combines disks - * with files, report it as an error. + * with files, only report it as an error when + * fatal is set to ensure all the replication + * checks aren't skipped in check_replication(). */ - if (!dontreport && type != NULL && + if (fatal && !dontreport && type != NULL && strcmp(type, childtype) != 0) { if (ret != NULL) free(ret); ret = NULL; - if (fatal) - vdev_error(gettext( - "mismatched replication " - "level: %s contains both " - "files and devices\n"), - rep.zprl_type); - else - return (NULL); + vdev_error(gettext( + "mismatched replication " + "level: %s contains both " + "files and devices\n"), + rep.zprl_type); dontreport = B_TRUE; } diff --git a/sys/contrib/openzfs/cmd/zstream/Makefile.am b/sys/contrib/openzfs/cmd/zstream/Makefile.am index be3539fe905d..80ef1ea7ca11 100644 --- a/sys/contrib/openzfs/cmd/zstream/Makefile.am +++ b/sys/contrib/openzfs/cmd/zstream/Makefile.am @@ -18,6 +18,7 @@ zstream_LDADD = \ libzpool.la \ libnvpair.la -PHONY += install-exec-hook -install-exec-hook: +cmd-zstream-install-exec-hook: cd $(DESTDIR)$(sbindir) && $(LN_S) -f zstream zstreamdump + +INSTALL_EXEC_HOOKS += cmd-zstream-install-exec-hook |