The branch releng/15.0 has been updated by cperciva: URL: https://cgit.FreeBSD.org/src/commit/?id=6bf158defbcb2f3c948b28394388179c7e95bd8b
commit 6bf158defbcb2f3c948b28394388179c7e95bd8b Author: Martin Matuska <[email protected]> AuthorDate: 2025-10-25 08:14:36 +0000 Commit: Colin Percival <[email protected]> CommitDate: 2025-10-28 21:22:49 +0000 zfs: merge openzfs/zfs@814f9afba (zfs-2.4-release) into stable/15 OpenZFS 2.4.0 rc3 Notable upstream pull request merges: #17750 964dfc317 FreeBSD: Correct _PC_MIN_HOLE_SIZE #17793 b9d1e28a7 ddt prune: Add SCL_ZIO deadlock workaround #17799 1585a10a8 Make mount/share errors non-fatal for zfs create/clone #17801 073b34b3e Fix display of default xattr to show 'sa' #17803 e09c86cb1 zvol: verify IO type is supported #17807 -multiple zpool iostat: fix regressions in "all pools" mode after #17786 #17826 b9356f06e Explicit set ashift for non-leaf vdevs #17830 f0bff230f Suppress some ashift warnings #17833 c1f55bff8 Fix the type of the raidz_outlier_check_interval_ms parameter #17836 799bda73e Fix return value for setting zvol threading #17834 f0c76f8a7 libzpool/cmn_err: remove suppression, add stop option, cleanup #17843 1956417b5 mmap_seek: print error code and text on failure #17847 7987d4deb Update device removal documentation #17851 6f6e1c90a FreeBSD: zfs_getpages: Don't zero freshly allocated pages Approved by: re (cperciva) Obtained from: OpenZFS OpenZFS commit: 814f9afba7d6714bd63f18ac808955b61d37f9f0 OpenZFS tag: zfs-2.4.0-rc3 (cherry picked from commit ce7e1f86a1810038f142c5c68b787c47252b39c5) --- .../.github/ISSUE_TEMPLATE/feature_request.md | 2 +- .../.github/workflows/scripts/qemu-2-start.sh | 11 +- sys/contrib/openzfs/.github/workflows/zfs-qemu.yml | 2 +- sys/contrib/openzfs/META | 2 +- sys/contrib/openzfs/cmd/zdb/zdb.c | 191 +++++++++++++++++++-- sys/contrib/openzfs/cmd/zfs/zfs_main.c | 10 +- sys/contrib/openzfs/cmd/zinject/zinject.c | 81 +++++++-- sys/contrib/openzfs/cmd/zpool/zpool_iter.c | 19 +- sys/contrib/openzfs/cmd/zpool/zpool_main.c | 1 + sys/contrib/openzfs/cmd/zpool/zpool_vdev.c | 64 +++---- .../config/kernel-block-device-operations.m4 | 34 ++++ sys/contrib/openzfs/config/kernel-drop-inode.m4 | 24 +++ sys/contrib/openzfs/config/kernel-namespace.m4 | 31 ++++ .../openzfs/config/kernel-userns-capabilities.m4 | 79 --------- sys/contrib/openzfs/config/kernel-writeback.m4 | 58 +++++++ sys/contrib/openzfs/config/kernel-writepage_t.m4 | 26 --- sys/contrib/openzfs/config/kernel.m4 | 8 +- sys/contrib/openzfs/contrib/intel_qat/readme.md | 2 +- sys/contrib/openzfs/etc/init.d/README.md | 2 +- sys/contrib/openzfs/include/libzfs.h | 2 + .../include/os/linux/kernel/linux/vfs_compat.h | 7 + sys/contrib/openzfs/include/sys/spa.h | 2 +- sys/contrib/openzfs/include/sys/zfs_ioctl.h | 1 + sys/contrib/openzfs/include/sys/zio.h | 1 + sys/contrib/openzfs/lib/libspl/include/sys/uio.h | 1 + sys/contrib/openzfs/lib/libuutil/libuutil.abi | 105 +---------- sys/contrib/openzfs/lib/libzfs/libzfs.abi | 172 ++++--------------- sys/contrib/openzfs/lib/libzfs/libzfs_config.c | 17 ++ sys/contrib/openzfs/lib/libzfs/libzfs_status.c | 78 +++++---- .../openzfs/lib/libzfs_core/libzfs_core.abi | 105 +---------- sys/contrib/openzfs/lib/libzpool/kernel.c | 79 +++++++-- sys/contrib/openzfs/man/man8/zdb.8 | 14 ++ sys/contrib/openzfs/man/man8/zinject.8 | 14 ++ sys/contrib/openzfs/man/man8/zpool-remove.8 | 4 +- .../openzfs/module/icp/algs/sha2/sha2_generic.c | 41 +++-- .../openzfs/module/os/freebsd/zfs/zfs_ctldir.c | 3 +- .../openzfs/module/os/freebsd/zfs/zfs_vnops_os.c | 31 +++- sys/contrib/openzfs/module/os/linux/spl/spl-zone.c | 19 +- sys/contrib/openzfs/module/os/linux/zfs/abd_os.c | 9 + sys/contrib/openzfs/module/os/linux/zfs/zfs_acl.c | 2 +- .../openzfs/module/os/linux/zfs/zfs_vnops_os.c | 5 +- sys/contrib/openzfs/module/os/linux/zfs/zpl_file.c | 74 ++++++++ .../openzfs/module/os/linux/zfs/zpl_super.c | 4 +- sys/contrib/openzfs/module/os/linux/zfs/zvol_os.c | 109 ++++++++---- sys/contrib/openzfs/module/zcommon/zfs_prop.c | 2 +- sys/contrib/openzfs/module/zfs/arc.c | 17 +- sys/contrib/openzfs/module/zfs/mmp.c | 2 +- sys/contrib/openzfs/module/zfs/spa_misc.c | 8 +- sys/contrib/openzfs/module/zfs/vdev_removal.c | 80 ++++++--- sys/contrib/openzfs/module/zfs/zio.c | 35 +++- sys/contrib/openzfs/module/zfs/zio_inject.c | 38 ++++ sys/contrib/openzfs/module/zfs/zvol.c | 2 +- sys/contrib/openzfs/tests/runfiles/common.run | 3 +- sys/contrib/openzfs/tests/runfiles/sanity.run | 2 +- .../openzfs/tests/zfs-tests/cmd/mmap_seek.c | 35 ++-- .../openzfs/tests/zfs-tests/tests/Makefile.am | 1 + .../cli_root/zpool_reopen/zpool_reopen_004_pos.ksh | 2 + .../functional/delegate/delegate_common.kshlib | 6 +- .../tests/functional/xattr/xattr_014_pos.ksh | 53 ++++++ .../functional/zvol/zvol_misc/zvol_misc_trim.ksh | 7 + 60 files changed, 1133 insertions(+), 706 deletions(-) diff --git a/sys/contrib/openzfs/.github/ISSUE_TEMPLATE/feature_request.md b/sys/contrib/openzfs/.github/ISSUE_TEMPLATE/feature_request.md index 9b50a4a3d96e..f3d4316f6f67 100644 --- a/sys/contrib/openzfs/.github/ISSUE_TEMPLATE/feature_request.md +++ b/sys/contrib/openzfs/.github/ISSUE_TEMPLATE/feature_request.md @@ -14,7 +14,7 @@ Please check our issue tracker before opening a new feature request. Filling out the following template will help other contributors better understand your proposed feature. --> -### Describe the feature would like to see added to OpenZFS +### Describe the feature you would like to see added to OpenZFS <!-- Provide a clear and concise description of the feature. diff --git a/sys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh b/sys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh index 1c608348ffcd..422b3e9df388 100755 --- a/sys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh +++ b/sys/contrib/openzfs/.github/workflows/scripts/qemu-2-start.sh @@ -121,7 +121,14 @@ case "$OS" in KSRC="$FREEBSD_SNAP/../amd64/$FreeBSD/src.txz" ;; freebsd15-0c) - FreeBSD="15.0-ALPHA3" + FreeBSD="15.0-ALPHA4" + OSNAME="FreeBSD $FreeBSD" + OSv="freebsd14.0" + URLxz="$FREEBSD_SNAP/$FreeBSD/amd64/Latest/FreeBSD-$FreeBSD-amd64-BASIC-CI-ufs.raw.xz" + KSRC="$FREEBSD_SNAP/../amd64/$FreeBSD/src.txz" + ;; + freebsd16-0c) + FreeBSD="16.0-CURRENT" OSNAME="FreeBSD $FreeBSD" OSv="freebsd14.0" URLxz="$FREEBSD_SNAP/$FreeBSD/amd64/Latest/FreeBSD-$FreeBSD-amd64-BASIC-CI-ufs.raw.xz" @@ -287,7 +294,7 @@ else while pidof /usr/bin/qemu-system-x86_64 >/dev/null; do ssh 2>/dev/null root@vm0 "uname -a" && break done - ssh root@vm0 "pkg install -y bash ca_root_nss git qemu-guest-agent python3 py311-cloud-init" + ssh root@vm0 "env IGNORE_OSVERSION=yes pkg install -y bash ca_root_nss git qemu-guest-agent python3 py311-cloud-init" ssh root@vm0 "chsh -s $BASH root" ssh root@vm0 'sysrc qemu_guest_agent_enable="YES"' ssh root@vm0 'sysrc cloudinit_enable="YES"' diff --git a/sys/contrib/openzfs/.github/workflows/zfs-qemu.yml b/sys/contrib/openzfs/.github/workflows/zfs-qemu.yml index 64277dcca236..f1b189062bb7 100644 --- a/sys/contrib/openzfs/.github/workflows/zfs-qemu.yml +++ b/sys/contrib/openzfs/.github/workflows/zfs-qemu.yml @@ -68,7 +68,7 @@ jobs: # FreeBSD variants of 2025-06: # FreeBSD Release: freebsd13-5r, freebsd14-2r, freebsd14-3r # FreeBSD Stable: freebsd13-5s, freebsd14-3s - # FreeBSD Current: freebsd15-0c + # FreeBSD Current: freebsd15-0c, freebsd16-0c os: ${{ fromJson(needs.test-config.outputs.test_os) }} runs-on: ubuntu-24.04 steps: diff --git a/sys/contrib/openzfs/META b/sys/contrib/openzfs/META index ddb64774e4c0..2bfa51841cc4 100644 --- a/sys/contrib/openzfs/META +++ b/sys/contrib/openzfs/META @@ -2,7 +2,7 @@ Meta: 1 Name: zfs Branch: 1.0 Version: 2.4.0 -Release: rc2 +Release: rc3 Release-Tags: relext License: CDDL Author: OpenZFS diff --git a/sys/contrib/openzfs/cmd/zdb/zdb.c b/sys/contrib/openzfs/cmd/zdb/zdb.c index 70a4ed46f263..2560ad045db3 100644 --- a/sys/contrib/openzfs/cmd/zdb/zdb.c +++ b/sys/contrib/openzfs/cmd/zdb/zdb.c @@ -106,11 +106,15 @@ extern boolean_t spa_mode_readable_spacemaps; extern uint_t zfs_reconstruct_indirect_combinations_max; extern uint_t zfs_btree_verify_intensity; +enum { + ARG_ALLOCATED = 256, + ARG_BLOCK_BIN_MODE, + ARG_BLOCK_CLASSES, +}; + static const char cmdname[] = "zdb"; uint8_t dump_opt[512]; -#define ALLOCATED_OPT 256 - typedef void object_viewer_t(objset_t *, uint64_t, void *data, size_t size); static uint64_t *zopt_metaslab = NULL; @@ -131,6 +135,20 @@ static objset_t *os; static boolean_t kernel_init_done; static boolean_t corruption_found = B_FALSE; +static enum { + BIN_AUTO = 0, + BIN_PSIZE, + BIN_LSIZE, + BIN_ASIZE, +} block_bin_mode = BIN_AUTO; + +static enum { + CLASS_NORMAL = 1 << 1, + CLASS_SPECIAL = 1 << 2, + CLASS_DEDUP = 1 << 3, + CLASS_OTHER = 1 << 4, +} block_classes = 0; + static void snprintf_blkptr_compact(char *, size_t, const blkptr_t *, boolean_t); static void mos_obj_refd(uint64_t); @@ -749,6 +767,12 @@ usage(void) (void) fprintf(stderr, " Options to control amount of output:\n"); (void) fprintf(stderr, " -b --block-stats " "block statistics\n"); + (void) fprintf(stderr, " --bin=(lsize|psize|asize) " + "bin blocks based on this size in all three columns\n"); + (void) fprintf(stderr, + " --class=(normal|special|dedup|other)[,...]\n" + " only consider blocks from " + "these allocation classes\n"); (void) fprintf(stderr, " -B --backup " "backup stream\n"); (void) fprintf(stderr, " -c --checksum " @@ -1694,7 +1718,7 @@ 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[ALLOCATED_OPT] || + if (dump_opt[ARG_ALLOCATED] || (dump_opt['m'] > 2 && !dump_opt['L'])) { mutex_enter(&msp->ms_lock); VERIFY0(metaslab_load(msp)); @@ -1705,7 +1729,7 @@ dump_metaslab(metaslab_t *msp) dump_metaslab_stats(msp); } - if (dump_opt[ALLOCATED_OPT]) { + if (dump_opt[ARG_ALLOCATED]) { uint64_t off = msp->ms_start; zfs_range_tree_walk(msp->ms_allocatable, dump_allocated, &off); @@ -1726,7 +1750,7 @@ dump_metaslab(metaslab_t *msp) SPACE_MAP_HISTOGRAM_SIZE, sm->sm_shift); } - if (dump_opt[ALLOCATED_OPT] || + if (dump_opt[ARG_ALLOCATED] || (dump_opt['m'] > 2 && !dump_opt['L'])) { metaslab_unload(msp); mutex_exit(&msp->ms_lock); @@ -5814,6 +5838,34 @@ dump_size_histograms(zdb_cb_t *zcb) (void) printf("\nBlock Size Histogram\n"); + switch (block_bin_mode) { + case BIN_PSIZE: + printf("(note: all categories are binned by %s)\n", "psize"); + break; + case BIN_LSIZE: + printf("(note: all categories are binned by %s)\n", "lsize"); + break; + case BIN_ASIZE: + printf("(note: all categories are binned by %s)\n", "asize"); + break; + default: + printf("(note: all categories are binned separately)\n"); + break; + } + if (block_classes != 0) { + char buf[256] = ""; + if (block_classes & CLASS_NORMAL) + strlcat(buf, "\"normal\", ", sizeof (buf)); + if (block_classes & CLASS_SPECIAL) + strlcat(buf, "\"special\", ", sizeof (buf)); + if (block_classes & CLASS_DEDUP) + strlcat(buf, "\"dedup\", ", sizeof (buf)); + if (block_classes & CLASS_OTHER) + strlcat(buf, "\"other\", ", sizeof (buf)); + buf[strlen(buf)-2] = '\0'; + printf("(note: only blocks in these classes are counted: %s)\n", + buf); + } /* * Print the first line titles */ @@ -6162,29 +6214,85 @@ skipped: [BPE_GET_PSIZE(bp)]++; return; } + + if (block_classes != 0) { + spa_config_enter(zcb->zcb_spa, SCL_CONFIG, FTAG, RW_READER); + + uint64_t vdev = DVA_GET_VDEV(&bp->blk_dva[0]); + uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[0]); + vdev_t *vd = vdev_lookup_top(zcb->zcb_spa, vdev); + ASSERT(vd != NULL); + metaslab_t *ms = vd->vdev_ms[offset >> vd->vdev_ms_shift]; + ASSERT(ms != NULL); + metaslab_group_t *mg = ms->ms_group; + ASSERT(mg != NULL); + metaslab_class_t *mc = mg->mg_class; + ASSERT(mc != NULL); + + spa_config_exit(zcb->zcb_spa, SCL_CONFIG, FTAG); + + int class; + if (mc == spa_normal_class(zcb->zcb_spa)) { + class = CLASS_NORMAL; + } else if (mc == spa_special_class(zcb->zcb_spa)) { + class = CLASS_SPECIAL; + } else if (mc == spa_dedup_class(zcb->zcb_spa)) { + class = CLASS_DEDUP; + } else { + class = CLASS_OTHER; + } + + if (!(block_classes & class)) { + goto hist_skipped; + } + } + /* * The binning histogram bins by powers of two up to * SPA_MAXBLOCKSIZE rather than creating bins for * every possible blocksize found in the pool. */ - int bin = highbit64(BP_GET_PSIZE(bp)) - 1; + int bin; + + /* + * Binning strategy: each bin includes blocks up to and including + * the given size (excluding blocks that fit into the previous bin). + * This way, the "4K" bin includes blocks within the (2K; 4K] range. + */ +#define BIN(size) (highbit64((size) - 1)) + + switch (block_bin_mode) { + case BIN_PSIZE: bin = BIN(BP_GET_PSIZE(bp)); break; + case BIN_LSIZE: bin = BIN(BP_GET_LSIZE(bp)); break; + case BIN_ASIZE: bin = BIN(BP_GET_ASIZE(bp)); break; + case BIN_AUTO: break; + default: PANIC("bad block_bin_mode"); abort(); + } + + if (block_bin_mode == BIN_AUTO) + bin = BIN(BP_GET_PSIZE(bp)); zcb->zcb_psize_count[bin]++; zcb->zcb_psize_len[bin] += BP_GET_PSIZE(bp); zcb->zcb_psize_total += BP_GET_PSIZE(bp); - bin = highbit64(BP_GET_LSIZE(bp)) - 1; + if (block_bin_mode == BIN_AUTO) + bin = BIN(BP_GET_LSIZE(bp)); zcb->zcb_lsize_count[bin]++; zcb->zcb_lsize_len[bin] += BP_GET_LSIZE(bp); zcb->zcb_lsize_total += BP_GET_LSIZE(bp); - bin = highbit64(BP_GET_ASIZE(bp)) - 1; + if (block_bin_mode == BIN_AUTO) + bin = BIN(BP_GET_ASIZE(bp)); zcb->zcb_asize_count[bin]++; zcb->zcb_asize_len[bin] += BP_GET_ASIZE(bp); zcb->zcb_asize_total += BP_GET_ASIZE(bp); +#undef BIN + +hist_skipped: if (!do_claim) return; @@ -9426,7 +9534,11 @@ main(int argc, char **argv) {"livelist", no_argument, NULL, 'y'}, {"zstd-headers", no_argument, NULL, 'Z'}, {"allocated-map", no_argument, NULL, - ALLOCATED_OPT}, + ARG_ALLOCATED}, + {"bin", required_argument, NULL, + ARG_BLOCK_BIN_MODE}, + {"class", required_argument, NULL, + ARG_BLOCK_CLASSES}, {0, 0, 0, 0} }; @@ -9457,7 +9569,7 @@ main(int argc, char **argv) case 'u': case 'y': case 'Z': - case ALLOCATED_OPT: + case ARG_ALLOCATED: dump_opt[c]++; dump_all = 0; break; @@ -9540,6 +9652,59 @@ main(int argc, char **argv) case 'x': vn_dumpdir = optarg; break; + case ARG_BLOCK_BIN_MODE: + if (strcmp(optarg, "lsize") == 0) { + block_bin_mode = BIN_LSIZE; + } else if (strcmp(optarg, "psize") == 0) { + block_bin_mode = BIN_PSIZE; + } else if (strcmp(optarg, "asize") == 0) { + block_bin_mode = BIN_ASIZE; + } else { + (void) fprintf(stderr, + "--bin=\"%s\" must be one of \"lsize\", " + "\"psize\" or \"asize\"\n", optarg); + usage(); + } + break; + + case ARG_BLOCK_CLASSES: { + char *buf = strdup(optarg), *tok = buf, *next, + *save = NULL; + + while ((next = strtok_r(tok, ",", &save)) != NULL) { + tok = NULL; + + if (strcmp(next, "normal") == 0) { + block_classes |= CLASS_NORMAL; + } else if (strcmp(next, "special") == 0) { + block_classes |= CLASS_SPECIAL; + } else if (strcmp(next, "dedup") == 0) { + block_classes |= CLASS_DEDUP; + } else if (strcmp(next, "other") == 0) { + block_classes |= CLASS_OTHER; + } else { + (void) fprintf(stderr, + "--class=\"%s\" must be a " + "comma-separated list of either " + "\"normal\", \"special\", " + "\"asize\" or \"other\"; " + "got \"%s\"\n", + optarg, next); + usage(); + } + } + + if (block_classes == 0) { + (void) fprintf(stderr, + "--class= must be a comma-separated " + "list of either \"normal\", \"special\", " + "\"asize\" or \"other\"; got empty\n"); + usage(); + } + + free(buf); + break; + } default: usage(); break; @@ -9582,6 +9747,9 @@ main(int argc, char **argv) */ spa_mode_readable_spacemaps = B_TRUE; + libspl_set_assert_ok((dump_opt['A'] == 1) || (dump_opt['A'] > 2)); + zfs_recover = (dump_opt['A'] > 1); + if (dump_all) verbose = MAX(verbose, 1); @@ -9592,9 +9760,6 @@ main(int argc, char **argv) dump_opt[c] += verbose; } - libspl_set_assert_ok((dump_opt['A'] == 1) || (dump_opt['A'] > 2)); - zfs_recover = (dump_opt['A'] > 1); - argc -= optind; argv += optind; if (argc < 2 && dump_opt['R']) diff --git a/sys/contrib/openzfs/cmd/zfs/zfs_main.c b/sys/contrib/openzfs/cmd/zfs/zfs_main.c index b66440c8f9f8..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); @@ -1333,7 +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); 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/zpool_iter.c b/sys/contrib/openzfs/cmd/zpool/zpool_iter.c index 16d4fe5eb1c6..fef602736705 100644 --- a/sys/contrib/openzfs/cmd/zpool/zpool_iter.c +++ b/sys/contrib/openzfs/cmd/zpool/zpool_iter.c @@ -103,6 +103,7 @@ add_pool(zpool_handle_t *zhp, zpool_list_t *zlp) 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(new); @@ -203,31 +204,23 @@ pool_list_refresh(zpool_list_t *zlp) return (navail); } - /* - * Search for any new pools and add them to the list. zpool_iter() - * will call zpool_refresh_stats() as part of its work, so this has - * the side effect of updating all active handles. - */ + /* Search for any new pools and add them to the list. */ (void) zpool_iter(g_zfs, add_pool_cb, zlp); - /* - * Walk the list for any that weren't refreshed, and update and remove - * them. It's not enough to just skip available ones, as zpool_iter() - * won't update them, so they'll still appear active in our list. - */ + /* 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're already - * handled. + * 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; - /* Do the refresh ourselves, just in case. */ + /* Refresh and remove if necessary. */ boolean_t missing; zpool_refresh_stats(node->zn_handle, &missing); if (missing) { diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_main.c b/sys/contrib/openzfs/cmd/zpool/zpool_main.c index 286336cf2730..1feec55c0e8b 100644 --- a/sys/contrib/openzfs/cmd/zpool/zpool_main.c +++ b/sys/contrib/openzfs/cmd/zpool/zpool_main.c @@ -6677,6 +6677,7 @@ zpool_do_iostat(int argc, char **argv) if (skip) { (void) fflush(stdout); (void) fsleep(interval); + last_npools = npools; continue; } diff --git a/sys/contrib/openzfs/cmd/zpool/zpool_vdev.c b/sys/contrib/openzfs/cmd/zpool/zpool_vdev.c index 088c0108e911..222b5524669e 100644 --- a/sys/contrib/openzfs/cmd/zpool/zpool_vdev.c +++ b/sys/contrib/openzfs/cmd/zpool/zpool_vdev.c @@ -270,14 +270,13 @@ is_spare(nvlist_t *config, const char *path) * draid* Virtual dRAID spare */ static nvlist_t * -make_leaf_vdev(nvlist_t *props, const char *arg, boolean_t is_primary) +make_leaf_vdev(const char *arg, boolean_t is_primary, uint64_t ashift) { char path[MAXPATHLEN]; struct stat64 statbuf; nvlist_t *vdev = NULL; const char *type = NULL; boolean_t wholedisk = B_FALSE; - uint64_t ashift = 0; int err; /* @@ -381,31 +380,6 @@ make_leaf_vdev(nvlist_t *props, const char *arg, boolean_t is_primary) verify(nvlist_add_uint64(vdev, ZPOOL_CONFIG_WHOLE_DISK, (uint64_t)wholedisk) == 0); - /* - * Override defaults if custom properties are provided. - */ - if (props != NULL) { - const char *value = NULL; - - if (nvlist_lookup_string(props, - zpool_prop_to_name(ZPOOL_PROP_ASHIFT), &value) == 0) { - if (zfs_nicestrtonum(NULL, value, &ashift) != 0) { - (void) fprintf(stderr, - gettext("ashift must be a number.\n")); - return (NULL); - } - if (ashift != 0 && - (ashift < ASHIFT_MIN || ashift > ASHIFT_MAX)) { - (void) fprintf(stderr, - gettext("invalid 'ashift=%" PRIu64 "' " - "property: only values between %" PRId32 " " - "and %" PRId32 " are allowed.\n"), - ashift, ASHIFT_MIN, ASHIFT_MAX); - return (NULL); - } - } - } - /* * If the device is known to incorrectly report its physical sector * size explicitly provide the known correct value. @@ -1513,6 +1487,29 @@ construct_spec(nvlist_t *props, int argc, char **argv) const char *type, *fulltype; boolean_t is_log, is_special, is_dedup, is_spare; boolean_t seen_logs; + uint64_t ashift = 0; + + if (props != NULL) { + const char *value = NULL; + + if (nvlist_lookup_string(props, + zpool_prop_to_name(ZPOOL_PROP_ASHIFT), &value) == 0) { + if (zfs_nicestrtonum(NULL, value, &ashift) != 0) { + (void) fprintf(stderr, + gettext("ashift must be a number.\n")); + return (NULL); + } + if (ashift != 0 && + (ashift < ASHIFT_MIN || ashift > ASHIFT_MAX)) { + (void) fprintf(stderr, + gettext("invalid 'ashift=%" PRIu64 "' " + "property: only values between %" PRId32 " " + "and %" PRId32 " are allowed.\n"), + ashift, ASHIFT_MIN, ASHIFT_MAX); + return (NULL); + } + } + } top = NULL; toplevels = 0; @@ -1618,9 +1615,9 @@ construct_spec(nvlist_t *props, int argc, char **argv) children * sizeof (nvlist_t *)); if (child == NULL) zpool_no_memory(); - if ((nv = make_leaf_vdev(props, argv[c], + if ((nv = make_leaf_vdev(argv[c], !(is_log || is_special || is_dedup || - is_spare))) == NULL) { + is_spare), ashift)) == NULL) { for (c = 0; c < children - 1; c++) nvlist_free(child[c]); free(child); @@ -1684,6 +1681,10 @@ construct_spec(nvlist_t *props, int argc, char **argv) ZPOOL_CONFIG_ALLOCATION_BIAS, VDEV_ALLOC_BIAS_DEDUP) == 0); } + if (ashift > 0) { + fnvlist_add_uint64(nv, + ZPOOL_CONFIG_ASHIFT, ashift); + } if (strcmp(type, VDEV_TYPE_RAIDZ) == 0) { verify(nvlist_add_uint64(nv, ZPOOL_CONFIG_NPARITY, @@ -1711,8 +1712,9 @@ construct_spec(nvlist_t *props, int argc, char **argv) * We have a device. Pass off to make_leaf_vdev() to * construct the appropriate nvlist describing the vdev. */ - if ((nv = make_leaf_vdev(props, argv[0], !(is_log || - is_special || is_dedup || is_spare))) == NULL) + if ((nv = make_leaf_vdev(argv[0], !(is_log || + is_special || is_dedup || is_spare), + ashift)) == NULL) goto spec_out; verify(nvlist_add_uint64(nv, diff --git a/sys/contrib/openzfs/config/kernel-block-device-operations.m4 b/sys/contrib/openzfs/config/kernel-block-device-operations.m4 index 4ff20b9c413d..1905340a9c7d 100644 --- a/sys/contrib/openzfs/config/kernel-block-device-operations.m4 +++ b/sys/contrib/openzfs/config/kernel-block-device-operations.m4 @@ -119,15 +119,49 @@ AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK], [ ]) ]) +dnl # +dnl # 6.18 API change +dnl # block_device_operation->getgeo takes struct gendisk* as first arg +dnl # +AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK], [ + ZFS_LINUX_TEST_SRC([block_device_operations_getgeo_gendisk], [ + #include <linux/blkdev.h> + + static int blk_getgeo(struct gendisk *disk, struct hd_geometry *geo) + { + (void) disk, (void) geo; + return (0); + } + + static const struct block_device_operations + bops __attribute__ ((unused)) = { + .getgeo = blk_getgeo, + }; + ], [], []) +]) + +AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK], [ + AC_MSG_CHECKING([whether bops->getgeo() takes gendisk as first arg]) + ZFS_LINUX_TEST_RESULT([block_device_operations_getgeo_gendisk], [ + AC_MSG_RESULT(yes) + AC_DEFINE([HAVE_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK], [1], + [Define if getgeo() in block_device_operations takes struct gendisk * as its first arg]) + ],[ + AC_MSG_RESULT(no) + ]) +]) + AC_DEFUN([ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS], [ ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_RELEASE_1ARG ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK + ZFS_AC_KERNEL_SRC_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK ]) AC_DEFUN([ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS], [ ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_CHECK_EVENTS ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_RELEASE_VOID ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_REVALIDATE_DISK + ZFS_AC_KERNEL_BLOCK_DEVICE_OPERATIONS_GETGEO_GENDISK ]) diff --git a/sys/contrib/openzfs/config/kernel-drop-inode.m4 b/sys/contrib/openzfs/config/kernel-drop-inode.m4 new file mode 100644 index 000000000000..6f2b12cadc02 --- /dev/null +++ b/sys/contrib/openzfs/config/kernel-drop-inode.m4 @@ -0,0 +1,24 @@ +dnl # +dnl # 6.18 API change +dnl # - generic_drop_inode() renamed to inode_generic_drop() +dnl # - generic_delete_inode() renamed to inode_just_drop() +dnl # +AC_DEFUN([ZFS_AC_KERNEL_SRC_INODE_GENERIC_DROP], [ + ZFS_LINUX_TEST_SRC([inode_generic_drop], [ + #include <linux/fs.h> + ],[ + struct inode *ip = NULL; + inode_generic_drop(ip); + ]) +]) + +AC_DEFUN([ZFS_AC_KERNEL_INODE_GENERIC_DROP], [ + AC_MSG_CHECKING([whether inode_generic_drop() exists]) + ZFS_LINUX_TEST_RESULT([inode_generic_drop], [ + AC_MSG_RESULT(yes) + AC_DEFINE(HAVE_INODE_GENERIC_DROP, 1, + [inode_generic_drop() exists]) + ],[ + AC_MSG_RESULT(no) + ]) +]) diff --git a/sys/contrib/openzfs/config/kernel-namespace.m4 b/sys/contrib/openzfs/config/kernel-namespace.m4 new file mode 100644 index 000000000000..9b0b12e4eab4 --- /dev/null +++ b/sys/contrib/openzfs/config/kernel-namespace.m4 @@ -0,0 +1,31 @@ +dnl # +dnl # 6.18 API change +dnl # ns->ops->type was moved to ns->ns.ns_type (struct ns_common) +dnl # +AC_DEFUN([ZFS_AC_KERNEL_SRC_NS_COMMON_TYPE], [ + ZFS_LINUX_TEST_SRC([ns_common_type], [ + #include <linux/user_namespace.h> + ],[ + struct user_namespace ns; + ns.ns.ns_type = 0; *** 2486 LINES SKIPPED ***
