From 647bac0e8fc5b05d15fefca05813b748e10cd1ef Mon Sep 17 00:00:00 2001 From: Sanjeev Bagewadi Date: Tue, 2 Feb 2021 13:54:15 +0000 Subject: [PATCH 1/2] Allocate zap_attribute_t from kmem instead of stack This patch is preparatory work for long name feature. It changes all users of zap_attribute_t to allocate it from kmem instead of stack. It also make zap_attribute_t and zap_name_t structure variable length. Signed-off-by: Chunwei Chen --- cmd/zdb/zdb.c | 129 +++++++++++++++------------ cmd/zhack.c | 17 ++-- include/sys/zap.h | 12 ++- include/sys/zap_impl.h | 3 +- module/os/freebsd/zfs/zfs_dir.c | 22 +++-- module/os/freebsd/zfs/zfs_vnops_os.c | 30 ++++--- module/os/linux/zfs/zfs_dir.c | 20 +++-- module/os/linux/zfs/zfs_vnops_os.c | 42 +++++---- module/os/linux/zfs/zpl_xattr.c | 10 ++- module/zfs/ddt_zap.c | 14 +-- module/zfs/dmu_objset.c | 40 +++++---- module/zfs/dmu_redact.c | 8 +- module/zfs/dsl_bookmark.c | 8 +- module/zfs/dsl_crypt.c | 4 +- module/zfs/dsl_dataset.c | 18 ++-- module/zfs/dsl_deadlist.c | 38 ++++---- module/zfs/dsl_deleg.c | 36 ++++---- module/zfs/dsl_destroy.c | 4 +- module/zfs/dsl_dir.c | 4 +- module/zfs/dsl_pool.c | 12 +-- module/zfs/dsl_prop.c | 37 ++++---- module/zfs/dsl_scan.c | 26 +++--- module/zfs/dsl_userhold.c | 4 +- module/zfs/sa.c | 34 +++---- module/zfs/spa.c | 72 ++++++++------- module/zfs/spa_errlog.c | 84 +++++++++-------- module/zfs/spa_log_spacemap.c | 9 +- module/zfs/spa_misc.c | 2 + module/zfs/vdev.c | 19 ++-- module/zfs/zap.c | 18 ++-- module/zfs/zap_micro.c | 50 ++++++++++- module/zfs/zcp_iter.c | 29 +++--- module/zfs/zfeature.c | 4 +- module/zfs/zfs_quota.c | 12 +-- module/zfs/zvol.c | 4 +- 35 files changed, 511 insertions(+), 364 deletions(-) diff --git a/cmd/zdb/zdb.c b/cmd/zdb/zdb.c index a009942401..f29f3f9393 100644 --- a/cmd/zdb/zdb.c +++ b/cmd/zdb/zdb.c @@ -1117,7 +1117,7 @@ dump_zap(objset_t *os, uint64_t object, void *data, size_t size) { (void) data, (void) size; zap_cursor_t zc; - zap_attribute_t attr; + zap_attribute_t *attrp = zap_attribute_alloc(); void *prop; unsigned i; @@ -1125,53 +1125,54 @@ dump_zap(objset_t *os, uint64_t object, void *data, size_t size) (void) printf("\n"); for (zap_cursor_init(&zc, os, object); - zap_cursor_retrieve(&zc, &attr) == 0; + zap_cursor_retrieve(&zc, attrp) == 0; zap_cursor_advance(&zc)) { boolean_t key64 = !!(zap_getflags(zc.zc_zap) & ZAP_FLAG_UINT64_KEY); if (key64) (void) printf("\t\t0x%010lx = ", - *(uint64_t *)attr.za_name); + *(uint64_t *)attrp->za_name); else - (void) printf("\t\t%s = ", attr.za_name); + (void) printf("\t\t%s = ", attrp->za_name); - if (attr.za_num_integers == 0) { + if (attrp->za_num_integers == 0) { (void) printf("\n"); continue; } - prop = umem_zalloc(attr.za_num_integers * - attr.za_integer_length, UMEM_NOFAIL); + prop = umem_zalloc(attrp->za_num_integers * + attrp->za_integer_length, UMEM_NOFAIL); if (key64) (void) zap_lookup_uint64(os, object, - (const uint64_t *)attr.za_name, 1, - attr.za_integer_length, attr.za_num_integers, + (const uint64_t *)attrp->za_name, 1, + attrp->za_integer_length, attrp->za_num_integers, prop); else - (void) zap_lookup(os, object, attr.za_name, - attr.za_integer_length, attr.za_num_integers, + (void) zap_lookup(os, object, attrp->za_name, + attrp->za_integer_length, attrp->za_num_integers, prop); - if (attr.za_integer_length == 1 && !key64) { - if (strcmp(attr.za_name, + if (attrp->za_integer_length == 1 && !key64) { + if (strcmp(attrp->za_name, DSL_CRYPTO_KEY_MASTER_KEY) == 0 || - strcmp(attr.za_name, + strcmp(attrp->za_name, DSL_CRYPTO_KEY_HMAC_KEY) == 0 || - strcmp(attr.za_name, DSL_CRYPTO_KEY_IV) == 0 || - strcmp(attr.za_name, DSL_CRYPTO_KEY_MAC) == 0 || - strcmp(attr.za_name, DMU_POOL_CHECKSUM_SALT) == 0) { + strcmp(attrp->za_name, DSL_CRYPTO_KEY_IV) == 0 || + strcmp(attrp->za_name, DSL_CRYPTO_KEY_MAC) == 0 || + strcmp(attrp->za_name, + DMU_POOL_CHECKSUM_SALT) == 0) { uint8_t *u8 = prop; - for (i = 0; i < attr.za_num_integers; i++) { + for (i = 0; i < attrp->za_num_integers; i++) { (void) printf("%02x", u8[i]); } } else { (void) printf("%s", (char *)prop); } } else { - for (i = 0; i < attr.za_num_integers; i++) { - switch (attr.za_integer_length) { + for (i = 0; i < attrp->za_num_integers; i++) { + switch (attrp->za_integer_length) { case 1: (void) printf("%u ", ((uint8_t *)prop)[i]); @@ -1192,9 +1193,11 @@ dump_zap(objset_t *os, uint64_t object, void *data, size_t size) } } (void) printf("\n"); - umem_free(prop, attr.za_num_integers * attr.za_integer_length); + umem_free(prop, + attrp->za_num_integers * attrp->za_integer_length); } zap_cursor_fini(&zc); + zap_attribute_free(attrp); } static void @@ -1295,26 +1298,27 @@ dump_sa_attrs(objset_t *os, uint64_t object, void *data, size_t size) { (void) data, (void) size; zap_cursor_t zc; - zap_attribute_t attr; + zap_attribute_t *attrp = zap_attribute_alloc(); dump_zap_stats(os, object); (void) printf("\n"); for (zap_cursor_init(&zc, os, object); - zap_cursor_retrieve(&zc, &attr) == 0; + zap_cursor_retrieve(&zc, attrp) == 0; zap_cursor_advance(&zc)) { - (void) printf("\t\t%s = ", attr.za_name); - if (attr.za_num_integers == 0) { + (void) printf("\t\t%s = ", attrp->za_name); + if (attrp->za_num_integers == 0) { (void) printf("\n"); continue; } (void) printf(" %llx : [%d:%d:%d]\n", - (u_longlong_t)attr.za_first_integer, - (int)ATTR_LENGTH(attr.za_first_integer), - (int)ATTR_BSWAP(attr.za_first_integer), - (int)ATTR_NUM(attr.za_first_integer)); + (u_longlong_t)attrp->za_first_integer, + (int)ATTR_LENGTH(attrp->za_first_integer), + (int)ATTR_BSWAP(attrp->za_first_integer), + (int)ATTR_NUM(attrp->za_first_integer)); } zap_cursor_fini(&zc); + zap_attribute_free(attrp); } static void @@ -1322,7 +1326,7 @@ dump_sa_layouts(objset_t *os, uint64_t object, void *data, size_t size) { (void) data, (void) size; zap_cursor_t zc; - zap_attribute_t attr; + zap_attribute_t *attrp = zap_attribute_alloc(); uint16_t *layout_attrs; unsigned i; @@ -1330,29 +1334,30 @@ dump_sa_layouts(objset_t *os, uint64_t object, void *data, size_t size) (void) printf("\n"); for (zap_cursor_init(&zc, os, object); - zap_cursor_retrieve(&zc, &attr) == 0; + zap_cursor_retrieve(&zc, attrp) == 0; zap_cursor_advance(&zc)) { - (void) printf("\t\t%s = [", attr.za_name); - if (attr.za_num_integers == 0) { + (void) printf("\t\t%s = [", attrp->za_name); + if (attrp->za_num_integers == 0) { (void) printf("\n"); continue; } - VERIFY(attr.za_integer_length == 2); - layout_attrs = umem_zalloc(attr.za_num_integers * - attr.za_integer_length, UMEM_NOFAIL); + VERIFY(attrp->za_integer_length == 2); + layout_attrs = umem_zalloc(attrp->za_num_integers * + attrp->za_integer_length, UMEM_NOFAIL); - VERIFY(zap_lookup(os, object, attr.za_name, - attr.za_integer_length, - attr.za_num_integers, layout_attrs) == 0); + VERIFY(zap_lookup(os, object, attrp->za_name, + attrp->za_integer_length, + attrp->za_num_integers, layout_attrs) == 0); - for (i = 0; i != attr.za_num_integers; i++) + for (i = 0; i != attrp->za_num_integers; i++) (void) printf(" %d ", (int)layout_attrs[i]); (void) printf("]\n"); umem_free(layout_attrs, - attr.za_num_integers * attr.za_integer_length); + attrp->za_num_integers * attrp->za_integer_length); } zap_cursor_fini(&zc); + zap_attribute_free(attrp); } static void @@ -1360,7 +1365,7 @@ dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size) { (void) data, (void) size; zap_cursor_t zc; - zap_attribute_t attr; + zap_attribute_t *attrp = zap_attribute_alloc(); const char *typenames[] = { /* 0 */ "not specified", /* 1 */ "FIFO", @@ -1384,13 +1389,14 @@ dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size) (void) printf("\n"); for (zap_cursor_init(&zc, os, object); - zap_cursor_retrieve(&zc, &attr) == 0; + zap_cursor_retrieve(&zc, attrp) == 0; zap_cursor_advance(&zc)) { (void) printf("\t\t%s = %lld (type: %s)\n", - attr.za_name, ZFS_DIRENT_OBJ(attr.za_first_integer), - typenames[ZFS_DIRENT_TYPE(attr.za_first_integer)]); + attrp->za_name, ZFS_DIRENT_OBJ(attrp->za_first_integer), + typenames[ZFS_DIRENT_TYPE(attrp->za_first_integer)]); } zap_cursor_fini(&zc); + zap_attribute_free(attrp); } static int @@ -2100,23 +2106,24 @@ dump_brt(spa_t *spa) continue; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, brt->brt_mos, brtvd->bv_mos_entries); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { uint64_t refcnt; VERIFY0(zap_lookup_uint64(brt->brt_mos, brtvd->bv_mos_entries, - (const uint64_t *)za.za_name, 1, - za.za_integer_length, za.za_num_integers, &refcnt)); + (const uint64_t *)za->za_name, 1, + za->za_integer_length, za->za_num_integers, &refcnt)); - uint64_t offset = *(const uint64_t *)za.za_name; + uint64_t offset = *(const uint64_t *)za->za_name; snprintf(dva, sizeof (dva), "%" PRIu64 ":%llx", vdevid, (u_longlong_t)offset); printf("%-16s %-10llu\n", dva, (u_longlong_t)refcnt); } zap_cursor_fini(&zc); + zap_attribute_free(za); } } @@ -2898,28 +2905,30 @@ static void dump_bookmarks(objset_t *os, int verbosity) { zap_cursor_t zc; - zap_attribute_t attr; + zap_attribute_t *attrp; dsl_dataset_t *ds = dmu_objset_ds(os); dsl_pool_t *dp = spa_get_dsl(os->os_spa); objset_t *mos = os->os_spa->spa_meta_objset; if (verbosity < 4) return; + attrp = zap_attribute_alloc(); dsl_pool_config_enter(dp, FTAG); for (zap_cursor_init(&zc, mos, ds->ds_bookmarks_obj); - zap_cursor_retrieve(&zc, &attr) == 0; + zap_cursor_retrieve(&zc, attrp) == 0; zap_cursor_advance(&zc)) { char osname[ZFS_MAX_DATASET_NAME_LEN]; char buf[ZFS_MAX_DATASET_NAME_LEN]; int len; dmu_objset_name(os, osname); len = snprintf(buf, sizeof (buf), "%s#%s", osname, - attr.za_name); + attrp->za_name); VERIFY3S(len, <, ZFS_MAX_DATASET_NAME_LEN); (void) dump_bookmark(dp, buf, verbosity >= 5, verbosity >= 6); } zap_cursor_fini(&zc); dsl_pool_config_exit(dp, FTAG); + zap_attribute_free(attrp); } static void @@ -6710,18 +6719,19 @@ iterate_deleted_livelists(spa_t *spa, ll_iter_t func, void *arg) ASSERT0(err); zap_cursor_t zc; - zap_attribute_t attr; + zap_attribute_t *attrp = zap_attribute_alloc(); dsl_deadlist_t ll; /* NULL out os prior to dsl_deadlist_open in case it's garbage */ ll.dl_os = NULL; for (zap_cursor_init(&zc, mos, zap_obj); - zap_cursor_retrieve(&zc, &attr) == 0; + zap_cursor_retrieve(&zc, attrp) == 0; (void) zap_cursor_advance(&zc)) { - dsl_deadlist_open(&ll, mos, attr.za_first_integer); + dsl_deadlist_open(&ll, mos, attrp->za_first_integer); func(&ll, arg); dsl_deadlist_close(&ll); } zap_cursor_fini(&zc); + zap_attribute_free(attrp); } static int @@ -7936,13 +7946,14 @@ static void errorlog_count_refd(objset_t *mos, uint64_t errlog) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, errlog); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { - mos_obj_refd(za.za_first_integer); + mos_obj_refd(za->za_first_integer); } zap_cursor_fini(&zc); + zap_attribute_free(za); } static int diff --git a/cmd/zhack.c b/cmd/zhack.c index f15a6ece53..77284b8955 100644 --- a/cmd/zhack.c +++ b/cmd/zhack.c @@ -203,26 +203,27 @@ static void dump_obj(objset_t *os, uint64_t obj, const char *name) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); (void) printf("%s_obj:\n", name); for (zap_cursor_init(&zc, os, obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { - if (za.za_integer_length == 8) { - ASSERT(za.za_num_integers == 1); + if (za->za_integer_length == 8) { + ASSERT(za->za_num_integers == 1); (void) printf("\t%s = %llu\n", - za.za_name, (u_longlong_t)za.za_first_integer); + za->za_name, (u_longlong_t)za->za_first_integer); } else { - ASSERT(za.za_integer_length == 1); + ASSERT(za->za_integer_length == 1); char val[1024]; - VERIFY(zap_lookup(os, obj, za.za_name, + VERIFY(zap_lookup(os, obj, za->za_name, 1, sizeof (val), val) == 0); - (void) printf("\t%s = %s\n", za.za_name, val); + (void) printf("\t%s = %s\n", za->za_name, val); } } zap_cursor_fini(&zc); + zap_attribute_free(za); } static void diff --git a/include/sys/zap.h b/include/sys/zap.h index 0027f7c510..ec37c34471 100644 --- a/include/sys/zap.h +++ b/include/sys/zap.h @@ -376,9 +376,19 @@ typedef struct { boolean_t za_normalization_conflict; uint64_t za_num_integers; uint64_t za_first_integer; /* no sign extension for <8byte ints */ - char za_name[ZAP_MAXNAMELEN]; + uint32_t za_name_len; + char za_name[]; } zap_attribute_t; +void zap_init(void); +void zap_fini(void); + +/* + * Alloc and free zap_attribute_t. + */ +zap_attribute_t *zap_attribute_alloc(void); +void zap_attribute_free(zap_attribute_t *attrp); + /* * The interface for listing all the attributes of a zapobj can be * thought of as cursor moving down a list of the attributes one by diff --git a/include/sys/zap_impl.h b/include/sys/zap_impl.h index 2959aa9b2c..0c72c6881f 100644 --- a/include/sys/zap_impl.h +++ b/include/sys/zap_impl.h @@ -191,7 +191,8 @@ typedef struct zap_name { uint64_t zn_hash; matchtype_t zn_matchtype; int zn_normflags; - char zn_normbuf[ZAP_MAXNAMELEN]; + int zn_normbuf_len; + char zn_normbuf[]; } zap_name_t; #define zap_f zap_u.zap_fat diff --git a/module/os/freebsd/zfs/zfs_dir.c b/module/os/freebsd/zfs/zfs_dir.c index 3cdb94d6cd..79d08d09ae 100644 --- a/module/os/freebsd/zfs/zfs_dir.c +++ b/module/os/freebsd/zfs/zfs_dir.c @@ -287,7 +287,7 @@ void zfs_unlinked_drain(zfsvfs_t *zfsvfs) { zap_cursor_t zc; - zap_attribute_t zap; + zap_attribute_t *zap; dmu_object_info_t doi; znode_t *zp; dmu_tx_t *tx; @@ -296,8 +296,9 @@ zfs_unlinked_drain(zfsvfs_t *zfsvfs) /* * Iterate over the contents of the unlinked set. */ + zap = zap_attribute_alloc(); for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj); - zap_cursor_retrieve(&zc, &zap) == 0; + zap_cursor_retrieve(&zc, zap) == 0; zap_cursor_advance(&zc)) { /* @@ -305,7 +306,7 @@ zfs_unlinked_drain(zfsvfs_t *zfsvfs) */ error = dmu_object_info(zfsvfs->z_os, - zap.za_first_integer, &doi); + zap->za_first_integer, &doi); if (error != 0) continue; @@ -315,7 +316,7 @@ zfs_unlinked_drain(zfsvfs_t *zfsvfs) * We need to re-mark these list entries for deletion, * so we pull them back into core and set zp->z_unlinked. */ - error = zfs_zget(zfsvfs, zap.za_first_integer, &zp); + error = zfs_zget(zfsvfs, zap->za_first_integer, &zp); /* * We may pick up znodes that are already marked for deletion. @@ -351,6 +352,7 @@ zfs_unlinked_drain(zfsvfs_t *zfsvfs) vput(ZTOV(zp)); } zap_cursor_fini(&zc); + zap_attribute_free(zap); } /* @@ -368,18 +370,19 @@ static int zfs_purgedir(znode_t *dzp) { zap_cursor_t zc; - zap_attribute_t zap; + zap_attribute_t *zap; znode_t *xzp; dmu_tx_t *tx; zfsvfs_t *zfsvfs = dzp->z_zfsvfs; int skipped = 0; int error; + zap = zap_attribute_alloc(); for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); - (error = zap_cursor_retrieve(&zc, &zap)) == 0; + (error = zap_cursor_retrieve(&zc, zap)) == 0; zap_cursor_advance(&zc)) { error = zfs_zget(zfsvfs, - ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); + ZFS_DIRENT_OBJ(zap->za_first_integer), &xzp); if (error) { skipped += 1; continue; @@ -391,7 +394,7 @@ zfs_purgedir(znode_t *dzp) tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); - dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); + dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap->za_name); dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); /* Is this really needed ? */ @@ -405,7 +408,7 @@ zfs_purgedir(znode_t *dzp) continue; } - error = zfs_link_destroy(dzp, zap.za_name, xzp, tx, 0, NULL); + error = zfs_link_destroy(dzp, zap->za_name, xzp, tx, 0, NULL); if (error) skipped += 1; dmu_tx_commit(tx); @@ -413,6 +416,7 @@ zfs_purgedir(znode_t *dzp) vput(ZTOV(xzp)); } zap_cursor_fini(&zc); + zap_attribute_free(zap); if (error != ENOENT) skipped += 1; return (skipped); diff --git a/module/os/freebsd/zfs/zfs_vnops_os.c b/module/os/freebsd/zfs/zfs_vnops_os.c index b9b332434b..d1c1be02e2 100644 --- a/module/os/freebsd/zfs/zfs_vnops_os.c +++ b/module/os/freebsd/zfs/zfs_vnops_os.c @@ -1719,7 +1719,7 @@ zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp, caddr_t outbuf; size_t bufsize; zap_cursor_t zc; - zap_attribute_t zap; + zap_attribute_t *zap; uint_t bytes_wanted; uint64_t offset; /* must be unsigned; checks for < 1 */ uint64_t parent; @@ -1767,6 +1767,7 @@ zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp, os = zfsvfs->z_os; offset = zfs_uio_offset(uio); prefetch = zp->z_zn_prefetch; + zap = zap_attribute_alloc(); /* * Initialize the iterator cursor. @@ -1822,18 +1823,18 @@ zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp, * Special case `.', `..', and `.zfs'. */ if (offset == 0) { - (void) strcpy(zap.za_name, "."); - zap.za_normalization_conflict = 0; + (void) strcpy(zap->za_name, "."); + zap->za_normalization_conflict = 0; objnum = zp->z_id; type = DT_DIR; } else if (offset == 1) { - (void) strcpy(zap.za_name, ".."); - zap.za_normalization_conflict = 0; + (void) strcpy(zap->za_name, ".."); + zap->za_normalization_conflict = 0; objnum = parent; type = DT_DIR; } else if (offset == 2 && zfs_show_ctldir(zp)) { - (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); - zap.za_normalization_conflict = 0; + (void) strcpy(zap->za_name, ZFS_CTLDIR_NAME); + zap->za_normalization_conflict = 0; objnum = ZFSCTL_INO_ROOT; type = DT_DIR; } else { @@ -1847,8 +1848,8 @@ zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp, goto update; } - if (zap.za_integer_length != 8 || - zap.za_num_integers != 1) { + if (zap->za_integer_length != 8 || + zap->za_num_integers != 1) { cmn_err(CE_WARN, "zap_readdir: bad directory " "entry, obj = %lld, offset = %lld\n", (u_longlong_t)zp->z_id, @@ -1857,15 +1858,15 @@ zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp, goto update; } - objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); + objnum = ZFS_DIRENT_OBJ(zap->za_first_integer); /* * MacOS X can extract the object type here such as: * uint8_t type = ZFS_DIRENT_TYPE(zap.za_first_integer); */ - type = ZFS_DIRENT_TYPE(zap.za_first_integer); + type = ZFS_DIRENT_TYPE(zap->za_first_integer); } - reclen = DIRENT64_RECLEN(strlen(zap.za_name)); + reclen = DIRENT64_RECLEN(strlen(zap->za_name)); /* * Will this entry fit in the buffer? @@ -1885,10 +1886,10 @@ zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp, */ odp->d_ino = objnum; odp->d_reclen = reclen; - odp->d_namlen = strlen(zap.za_name); + odp->d_namlen = strlen(zap->za_name); /* NOTE: d_off is the offset for the *next* entry. */ next = &odp->d_off; - strlcpy(odp->d_name, zap.za_name, odp->d_namlen + 1); + strlcpy(odp->d_name, zap->za_name, odp->d_namlen + 1); odp->d_type = type; dirent_terminate(odp); odp = (dirent64_t *)((intptr_t)odp + reclen); @@ -1939,6 +1940,7 @@ zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp, update: zap_cursor_fini(&zc); + zap_attribute_free(zap); if (zfs_uio_segflg(uio) != UIO_SYSSPACE || zfs_uio_iovcnt(uio) != 1) kmem_free(outbuf, bufsize); diff --git a/module/os/linux/zfs/zfs_dir.c b/module/os/linux/zfs/zfs_dir.c index 1eeabe53d2..ad2ca15e29 100644 --- a/module/os/linux/zfs/zfs_dir.c +++ b/module/os/linux/zfs/zfs_dir.c @@ -476,7 +476,7 @@ zfs_unlinked_drain_task(void *arg) { zfsvfs_t *zfsvfs = arg; zap_cursor_t zc; - zap_attribute_t zap; + zap_attribute_t *zap = zap_attribute_alloc(); dmu_object_info_t doi; znode_t *zp; int error; @@ -487,7 +487,7 @@ zfs_unlinked_drain_task(void *arg) * Iterate over the contents of the unlinked set. */ for (zap_cursor_init(&zc, zfsvfs->z_os, zfsvfs->z_unlinkedobj); - zap_cursor_retrieve(&zc, &zap) == 0 && !zfsvfs->z_drain_cancel; + zap_cursor_retrieve(&zc, zap) == 0 && !zfsvfs->z_drain_cancel; zap_cursor_advance(&zc)) { /* @@ -495,7 +495,7 @@ zfs_unlinked_drain_task(void *arg) */ error = dmu_object_info(zfsvfs->z_os, - zap.za_first_integer, &doi); + zap->za_first_integer, &doi); if (error != 0) continue; @@ -505,7 +505,7 @@ zfs_unlinked_drain_task(void *arg) * We need to re-mark these list entries for deletion, * so we pull them back into core and set zp->z_unlinked. */ - error = zfs_zget(zfsvfs, zap.za_first_integer, &zp); + error = zfs_zget(zfsvfs, zap->za_first_integer, &zp); /* * We may pick up znodes that are already marked for deletion. @@ -532,6 +532,7 @@ zfs_unlinked_drain_task(void *arg) zfsvfs->z_draining = B_FALSE; zfsvfs->z_drain_task = TASKQID_INVALID; + zap_attribute_free(zap); } /* @@ -589,7 +590,7 @@ static int zfs_purgedir(znode_t *dzp) { zap_cursor_t zc; - zap_attribute_t zap; + zap_attribute_t *zap = zap_attribute_alloc(); znode_t *xzp; dmu_tx_t *tx; zfsvfs_t *zfsvfs = ZTOZSB(dzp); @@ -598,10 +599,10 @@ zfs_purgedir(znode_t *dzp) int error; for (zap_cursor_init(&zc, zfsvfs->z_os, dzp->z_id); - (error = zap_cursor_retrieve(&zc, &zap)) == 0; + (error = zap_cursor_retrieve(&zc, zap)) == 0; zap_cursor_advance(&zc)) { error = zfs_zget(zfsvfs, - ZFS_DIRENT_OBJ(zap.za_first_integer), &xzp); + ZFS_DIRENT_OBJ(zap->za_first_integer), &xzp); if (error) { skipped += 1; continue; @@ -612,7 +613,7 @@ zfs_purgedir(znode_t *dzp) tx = dmu_tx_create(zfsvfs->z_os); dmu_tx_hold_sa(tx, dzp->z_sa_hdl, B_FALSE); - dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap.za_name); + dmu_tx_hold_zap(tx, dzp->z_id, FALSE, zap->za_name); dmu_tx_hold_sa(tx, xzp->z_sa_hdl, B_FALSE); dmu_tx_hold_zap(tx, zfsvfs->z_unlinkedobj, FALSE, NULL); /* Is this really needed ? */ @@ -627,7 +628,7 @@ zfs_purgedir(znode_t *dzp) } memset(&dl, 0, sizeof (dl)); dl.dl_dzp = dzp; - dl.dl_name = zap.za_name; + dl.dl_name = zap->za_name; error = zfs_link_destroy(&dl, xzp, tx, 0, NULL); if (error) @@ -637,6 +638,7 @@ zfs_purgedir(znode_t *dzp) zfs_zrele_async(xzp); } zap_cursor_fini(&zc); + zap_attribute_free(zap); if (error != ENOENT) skipped += 1; return (skipped); diff --git a/module/os/linux/zfs/zfs_vnops_os.c b/module/os/linux/zfs/zfs_vnops_os.c index 1cecad9f77..259176391d 100644 --- a/module/os/linux/zfs/zfs_vnops_os.c +++ b/module/os/linux/zfs/zfs_vnops_os.c @@ -1512,7 +1512,7 @@ zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) zfsvfs_t *zfsvfs = ITOZSB(ip); objset_t *os; zap_cursor_t zc; - zap_attribute_t zap; + zap_attribute_t *zap; int error; uint8_t prefetch; uint8_t type; @@ -1537,6 +1537,7 @@ zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) os = zfsvfs->z_os; offset = ctx->pos; prefetch = zp->z_zn_prefetch; + zap = zap_attribute_alloc(); /* * Initialize the iterator cursor. @@ -1562,25 +1563,25 @@ zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) * Special case `.', `..', and `.zfs'. */ if (offset == 0) { - (void) strcpy(zap.za_name, "."); - zap.za_normalization_conflict = 0; + (void) strcpy(zap->za_name, "."); + zap->za_normalization_conflict = 0; objnum = zp->z_id; type = DT_DIR; } else if (offset == 1) { - (void) strcpy(zap.za_name, ".."); - zap.za_normalization_conflict = 0; + (void) strcpy(zap->za_name, ".."); + zap->za_normalization_conflict = 0; objnum = parent; type = DT_DIR; } else if (offset == 2 && zfs_show_ctldir(zp)) { - (void) strcpy(zap.za_name, ZFS_CTLDIR_NAME); - zap.za_normalization_conflict = 0; + (void) strcpy(zap->za_name, ZFS_CTLDIR_NAME); + zap->za_normalization_conflict = 0; objnum = ZFSCTL_INO_ROOT; type = DT_DIR; } else { /* * Grab next entry. */ - if ((error = zap_cursor_retrieve(&zc, &zap))) { + if ((error = zap_cursor_retrieve(&zc, zap))) { if (error == ENOENT) break; else @@ -1594,24 +1595,24 @@ zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) * * XXX: This should be a feature flag for compatibility */ - if (zap.za_integer_length != 8 || - zap.za_num_integers == 0) { + if (zap->za_integer_length != 8 || + zap->za_num_integers == 0) { cmn_err(CE_WARN, "zap_readdir: bad directory " "entry, obj = %lld, offset = %lld, " "length = %d, num = %lld\n", (u_longlong_t)zp->z_id, (u_longlong_t)offset, - zap.za_integer_length, - (u_longlong_t)zap.za_num_integers); + zap->za_integer_length, + (u_longlong_t)zap->za_num_integers); error = SET_ERROR(ENXIO); goto update; } - objnum = ZFS_DIRENT_OBJ(zap.za_first_integer); - type = ZFS_DIRENT_TYPE(zap.za_first_integer); + objnum = ZFS_DIRENT_OBJ(zap->za_first_integer); + type = ZFS_DIRENT_TYPE(zap->za_first_integer); } - done = !zpl_dir_emit(ctx, zap.za_name, strlen(zap.za_name), + done = !zpl_dir_emit(ctx, zap->za_name, strlen(zap->za_name), objnum, type); if (done) break; @@ -1634,6 +1635,7 @@ zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) update: zap_cursor_fini(&zc); + zap_attribute_free(zap); if (error == ENOENT) error = 0; out: @@ -1732,7 +1734,7 @@ zfs_setattr_dir(znode_t *dzp) zfsvfs_t *zfsvfs = ZTOZSB(dzp); objset_t *os = zfsvfs->z_os; zap_cursor_t zc; - zap_attribute_t zap; + zap_attribute_t *zap; zfs_dirlock_t *dl; znode_t *zp = NULL; dmu_tx_t *tx = NULL; @@ -1741,15 +1743,16 @@ zfs_setattr_dir(znode_t *dzp) int count; int err; + zap = zap_attribute_alloc(); zap_cursor_init(&zc, os, dzp->z_id); - while ((err = zap_cursor_retrieve(&zc, &zap)) == 0) { + while ((err = zap_cursor_retrieve(&zc, zap)) == 0) { count = 0; - if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { + if (zap->za_integer_length != 8 || zap->za_num_integers != 1) { err = ENXIO; break; } - err = zfs_dirent_lock(&dl, dzp, (char *)zap.za_name, &zp, + err = zfs_dirent_lock(&dl, dzp, (char *)zap->za_name, &zp, ZEXISTS, NULL, NULL); if (err == ENOENT) goto next; @@ -1829,6 +1832,7 @@ next: zfs_dirent_unlock(dl); } zap_cursor_fini(&zc); + zap_attribute_free(zap); return (err == ENOENT ? 0 : err); } diff --git a/module/os/linux/zfs/zpl_xattr.c b/module/os/linux/zfs/zpl_xattr.c index 4e4f5210f8..9744a5d0c7 100644 --- a/module/os/linux/zfs/zpl_xattr.c +++ b/module/os/linux/zfs/zpl_xattr.c @@ -153,19 +153,20 @@ static int zpl_xattr_readdir(struct inode *dxip, xattr_filldir_t *xf) { zap_cursor_t zc; - zap_attribute_t zap; + zap_attribute_t *zap = zap_attribute_alloc(); int error; zap_cursor_init(&zc, ITOZSB(dxip)->z_os, ITOZ(dxip)->z_id); - while ((error = -zap_cursor_retrieve(&zc, &zap)) == 0) { + while ((error = -zap_cursor_retrieve(&zc, zap)) == 0) { - if (zap.za_integer_length != 8 || zap.za_num_integers != 1) { + if (zap->za_integer_length != 8 || zap->za_num_integers != 1) { error = -ENXIO; break; } - error = zpl_xattr_filldir(xf, zap.za_name, strlen(zap.za_name)); + error = zpl_xattr_filldir(xf, zap->za_name, + strlen(zap->za_name)); if (error) break; @@ -173,6 +174,7 @@ zpl_xattr_readdir(struct inode *dxip, xattr_filldir_t *xf) } zap_cursor_fini(&zc); + zap_attribute_free(zap); if (error == -ENOENT) error = 0; diff --git a/module/zfs/ddt_zap.c b/module/zfs/ddt_zap.c index 7ce7461a2b..51201d2b1d 100644 --- a/module/zfs/ddt_zap.c +++ b/module/zfs/ddt_zap.c @@ -184,9 +184,10 @@ ddt_zap_walk(objset_t *os, uint64_t object, uint64_t *walk, ddt_key_t *ddk, ddt_phys_t *phys, size_t psize) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; int error; + za = zap_attribute_alloc(); if (*walk == 0) { /* * We don't want to prefetch the entire ZAP object, because @@ -199,20 +200,20 @@ ddt_zap_walk(objset_t *os, uint64_t object, uint64_t *walk, ddt_key_t *ddk, } else { zap_cursor_init_serialized(&zc, os, object, *walk); } - if ((error = zap_cursor_retrieve(&zc, &za)) == 0) { - uint64_t csize = za.za_num_integers; + if ((error = zap_cursor_retrieve(&zc, za)) == 0) { + uint64_t csize = za->za_num_integers; - ASSERT3U(za.za_integer_length, ==, 1); + ASSERT3U(za->za_integer_length, ==, 1); ASSERT3U(csize, <=, psize + 1); uchar_t *cbuf = kmem_alloc(csize, KM_SLEEP); - error = zap_lookup_uint64(os, object, (uint64_t *)za.za_name, + error = zap_lookup_uint64(os, object, (uint64_t *)za->za_name, DDT_KEY_WORDS, 1, csize, cbuf); ASSERT0(error); if (error == 0) { ddt_zap_decompress(cbuf, phys, csize, psize); - *ddk = *(ddt_key_t *)za.za_name; + *ddk = *(ddt_key_t *)za->za_name; } kmem_free(cbuf, csize); @@ -221,6 +222,7 @@ ddt_zap_walk(objset_t *os, uint64_t object, uint64_t *walk, ddt_key_t *ddk, *walk = zap_cursor_serialize(&zc); } zap_cursor_fini(&zc); + zap_attribute_free(za); return (error); } diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c index 8f4fefa4f4..c42dba401c 100644 --- a/module/zfs/dmu_objset.c +++ b/module/zfs/dmu_objset.c @@ -2615,35 +2615,39 @@ dmu_snapshot_list_next(objset_t *os, int namelen, char *name, { dsl_dataset_t *ds = os->os_dsl_dataset; zap_cursor_t cursor; - zap_attribute_t attr; + zap_attribute_t *attr; ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) return (SET_ERROR(ENOENT)); + attr = zap_attribute_alloc(); zap_cursor_init_serialized(&cursor, ds->ds_dir->dd_pool->dp_meta_objset, dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); - if (zap_cursor_retrieve(&cursor, &attr) != 0) { + if (zap_cursor_retrieve(&cursor, attr) != 0) { zap_cursor_fini(&cursor); + zap_attribute_free(attr); return (SET_ERROR(ENOENT)); } - if (strlen(attr.za_name) + 1 > namelen) { + if (strlen(attr->za_name) + 1 > namelen) { zap_cursor_fini(&cursor); + zap_attribute_free(attr); return (SET_ERROR(ENAMETOOLONG)); } - (void) strlcpy(name, attr.za_name, namelen); + (void) strlcpy(name, attr->za_name, namelen); if (idp) - *idp = attr.za_first_integer; + *idp = attr->za_first_integer; if (case_conflict) - *case_conflict = attr.za_normalization_conflict; + *case_conflict = attr->za_normalization_conflict; zap_cursor_advance(&cursor); *offp = zap_cursor_serialize(&cursor); zap_cursor_fini(&cursor); + zap_attribute_free(attr); return (0); } @@ -2660,33 +2664,37 @@ dmu_dir_list_next(objset_t *os, int namelen, char *name, { dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; zap_cursor_t cursor; - zap_attribute_t attr; + zap_attribute_t *attr; /* there is no next dir on a snapshot! */ if (os->os_dsl_dataset->ds_object != dsl_dir_phys(dd)->dd_head_dataset_obj) return (SET_ERROR(ENOENT)); + attr = zap_attribute_alloc(); zap_cursor_init_serialized(&cursor, dd->dd_pool->dp_meta_objset, dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); - if (zap_cursor_retrieve(&cursor, &attr) != 0) { + if (zap_cursor_retrieve(&cursor, attr) != 0) { zap_cursor_fini(&cursor); + zap_attribute_free(attr); return (SET_ERROR(ENOENT)); } - if (strlen(attr.za_name) + 1 > namelen) { + if (strlen(attr->za_name) + 1 > namelen) { zap_cursor_fini(&cursor); + zap_attribute_free(attr); return (SET_ERROR(ENAMETOOLONG)); } - (void) strlcpy(name, attr.za_name, namelen); + (void) strlcpy(name, attr->za_name, namelen); if (idp) - *idp = attr.za_first_integer; + *idp = attr->za_first_integer; zap_cursor_advance(&cursor); *offp = zap_cursor_serialize(&cursor); zap_cursor_fini(&cursor); + zap_attribute_free(attr); return (0); } @@ -2734,7 +2742,7 @@ dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) } thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; - attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + attr = zap_attribute_alloc(); /* * Iterate over all children. @@ -2795,7 +2803,7 @@ dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) } } - kmem_free(attr, sizeof (zap_attribute_t)); + zap_attribute_free(attr); if (err != 0) { dsl_dir_rele(dd, FTAG); @@ -2969,7 +2977,7 @@ dmu_objset_find_impl(spa_t *spa, const char *name, } thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; - attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + attr = zap_attribute_alloc(); /* * Iterate over all children. @@ -2997,7 +3005,7 @@ dmu_objset_find_impl(spa_t *spa, const char *name, if (err != 0) { dsl_dir_rele(dd, FTAG); dsl_pool_config_exit(dp, FTAG); - kmem_free(attr, sizeof (zap_attribute_t)); + zap_attribute_free(attr); return (err); } } @@ -3035,7 +3043,7 @@ dmu_objset_find_impl(spa_t *spa, const char *name, } dsl_dir_rele(dd, FTAG); - kmem_free(attr, sizeof (zap_attribute_t)); + zap_attribute_free(attr); dsl_pool_config_exit(dp, FTAG); if (err != 0) diff --git a/module/zfs/dmu_redact.c b/module/zfs/dmu_redact.c index 1feba0ba83..651461caff 100644 --- a/module/zfs/dmu_redact.c +++ b/module/zfs/dmu_redact.c @@ -189,7 +189,7 @@ zfs_get_deleteq(objset_t *os) objlist_t *deleteq_objlist = objlist_create(); uint64_t deleteq_obj; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; dmu_object_info_t doi; ASSERT3U(os->os_phys->os_type, ==, DMU_OST_ZFS); @@ -208,13 +208,15 @@ zfs_get_deleteq(objset_t *os) avl_create(&at, objnode_compare, sizeof (struct objnode), offsetof(struct objnode, node)); + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, os, deleteq_obj); - zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { struct objnode *obj = kmem_zalloc(sizeof (*obj), KM_SLEEP); - obj->obj = za.za_first_integer; + obj->obj = za->za_first_integer; avl_add(&at, obj); } zap_cursor_fini(&zc); + zap_attribute_free(za); struct objnode *next, *found = avl_first(&at); while (found != NULL) { diff --git a/module/zfs/dsl_bookmark.c b/module/zfs/dsl_bookmark.c index 5fd8bc2a26..a65aded4ef 100644 --- a/module/zfs/dsl_bookmark.c +++ b/module/zfs/dsl_bookmark.c @@ -870,13 +870,14 @@ dsl_bookmark_init_ds(dsl_dataset_t *ds) int err = 0; zap_cursor_t zc; - zap_attribute_t attr; + zap_attribute_t *attr; + attr = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, ds->ds_bookmarks_obj); - (err = zap_cursor_retrieve(&zc, &attr)) == 0; + (err = zap_cursor_retrieve(&zc, attr)) == 0; zap_cursor_advance(&zc)) { dsl_bookmark_node_t *dbn = - dsl_bookmark_node_alloc(attr.za_name); + dsl_bookmark_node_alloc(attr->za_name); err = dsl_bookmark_lookup_impl(ds, dbn->dbn_name, &dbn->dbn_phys); @@ -888,6 +889,7 @@ dsl_bookmark_init_ds(dsl_dataset_t *ds) avl_add(&ds->ds_bookmarks, dbn); } zap_cursor_fini(&zc); + zap_attribute_free(attr); if (err == ENOENT) err = 0; return (err); diff --git a/module/zfs/dsl_crypt.c b/module/zfs/dsl_crypt.c index 8e1055d9bc..475f4f5c1e 100644 --- a/module/zfs/dsl_crypt.c +++ b/module/zfs/dsl_crypt.c @@ -1497,7 +1497,7 @@ spa_keystore_change_key_sync_impl(uint64_t rddobj, uint64_t ddobj, } zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP); - za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); /* Recurse into all child dsl dirs. */ for (zap_cursor_init(zc, dp->dp_meta_objset, @@ -1529,7 +1529,7 @@ spa_keystore_change_key_sync_impl(uint64_t rddobj, uint64_t ddobj, } zap_cursor_fini(zc); - kmem_free(za, sizeof (zap_attribute_t)); + zap_attribute_free(za); kmem_free(zc, sizeof (zap_cursor_t)); dsl_dir_rele(dd, FTAG); diff --git a/module/zfs/dsl_dataset.c b/module/zfs/dsl_dataset.c index 45d8a290d6..237cf8a478 100644 --- a/module/zfs/dsl_dataset.c +++ b/module/zfs/dsl_dataset.c @@ -2290,7 +2290,7 @@ get_clones_stat_impl(dsl_dataset_t *ds, nvlist_t *val) uint64_t count = 0; objset_t *mos = ds->ds_dir->dd_pool->dp_meta_objset; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); @@ -2306,19 +2306,22 @@ get_clones_stat_impl(dsl_dataset_t *ds, nvlist_t *val) if (count != dsl_dataset_phys(ds)->ds_num_children - 1) { return (SET_ERROR(ENOENT)); } + + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, dsl_dataset_phys(ds)->ds_next_clones_obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { dsl_dataset_t *clone; char buf[ZFS_MAX_DATASET_NAME_LEN]; VERIFY0(dsl_dataset_hold_obj(ds->ds_dir->dd_pool, - za.za_first_integer, FTAG, &clone)); + za->za_first_integer, FTAG, &clone)); dsl_dir_name(clone->ds_dir, buf); fnvlist_add_boolean(val, buf); dsl_dataset_rele(clone, FTAG); } zap_cursor_fini(&zc); + zap_attribute_free(za); return (0); } @@ -3640,16 +3643,16 @@ dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx) if (dsl_dataset_phys(ds)->ds_next_clones_obj && spa_version(dp->dp_spa) >= SPA_VERSION_DIR_CLONES) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, dp->dp_meta_objset, dsl_dataset_phys(ds)->ds_next_clones_obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { dsl_dataset_t *cnds; uint64_t o; - if (za.za_first_integer == oldnext_obj) { + if (za->za_first_integer == oldnext_obj) { /* * We've already moved the * origin's reference. @@ -3658,7 +3661,7 @@ dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx) } VERIFY0(dsl_dataset_hold_obj(dp, - za.za_first_integer, FTAG, &cnds)); + za->za_first_integer, FTAG, &cnds)); o = dsl_dir_phys(cnds->ds_dir)-> dd_head_dataset_obj; @@ -3669,6 +3672,7 @@ dsl_dataset_promote_sync(void *arg, dmu_tx_t *tx) dsl_dataset_rele(cnds, FTAG); } zap_cursor_fini(&zc); + zap_attribute_free(za); } ASSERT(!dsl_prop_hascb(ds)); diff --git a/module/zfs/dsl_deadlist.c b/module/zfs/dsl_deadlist.c index eff1f7de77..e457e2fd86 100644 --- a/module/zfs/dsl_deadlist.c +++ b/module/zfs/dsl_deadlist.c @@ -133,7 +133,7 @@ static void dsl_deadlist_load_tree(dsl_deadlist_t *dl) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; int error; ASSERT(MUTEX_HELD(&dl->dl_lock)); @@ -159,20 +159,21 @@ dsl_deadlist_load_tree(dsl_deadlist_t *dl) if (dl->dl_havetree) return; + za = zap_attribute_alloc(); avl_create(&dl->dl_tree, dsl_deadlist_compare, sizeof (dsl_deadlist_entry_t), offsetof(dsl_deadlist_entry_t, dle_node)); for (zap_cursor_init(&zc, dl->dl_os, dl->dl_object); - (error = zap_cursor_retrieve(&zc, &za)) == 0; + (error = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { dsl_deadlist_entry_t *dle = kmem_alloc(sizeof (*dle), KM_SLEEP); - dle->dle_mintxg = zfs_strtonum(za.za_name, NULL); + dle->dle_mintxg = zfs_strtonum(za->za_name, NULL); /* * Prefetch all the bpobj's so that we do that i/o * in parallel. Then open them all in a second pass. */ - dle->dle_bpobj.bpo_object = za.za_first_integer; + dle->dle_bpobj.bpo_object = za->za_first_integer; dmu_prefetch_dnode(dl->dl_os, dle->dle_bpobj.bpo_object, ZIO_PRIORITY_SYNC_READ); @@ -180,6 +181,7 @@ dsl_deadlist_load_tree(dsl_deadlist_t *dl) } VERIFY3U(error, ==, ENOENT); zap_cursor_fini(&zc); + zap_attribute_free(za); for (dsl_deadlist_entry_t *dle = avl_first(&dl->dl_tree); dle != NULL; dle = AVL_NEXT(&dl->dl_tree, dle)) { @@ -207,7 +209,7 @@ static void dsl_deadlist_load_cache(dsl_deadlist_t *dl) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; int error; ASSERT(MUTEX_HELD(&dl->dl_lock)); @@ -221,26 +223,28 @@ dsl_deadlist_load_cache(dsl_deadlist_t *dl) avl_create(&dl->dl_cache, dsl_deadlist_cache_compare, sizeof (dsl_deadlist_cache_entry_t), offsetof(dsl_deadlist_cache_entry_t, dlce_node)); + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, dl->dl_os, dl->dl_object); - (error = zap_cursor_retrieve(&zc, &za)) == 0; + (error = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { - if (za.za_first_integer == empty_bpobj) + if (za->za_first_integer == empty_bpobj) continue; dsl_deadlist_cache_entry_t *dlce = kmem_zalloc(sizeof (*dlce), KM_SLEEP); - dlce->dlce_mintxg = zfs_strtonum(za.za_name, NULL); + dlce->dlce_mintxg = zfs_strtonum(za->za_name, NULL); /* * Prefetch all the bpobj's so that we do that i/o * in parallel. Then open them all in a second pass. */ - dlce->dlce_bpobj = za.za_first_integer; + dlce->dlce_bpobj = za->za_first_integer; dmu_prefetch_dnode(dl->dl_os, dlce->dlce_bpobj, ZIO_PRIORITY_SYNC_READ); avl_add(&dl->dl_cache, dlce); } VERIFY3U(error, ==, ENOENT); zap_cursor_fini(&zc); + zap_attribute_free(za); for (dsl_deadlist_cache_entry_t *dlce = avl_first(&dl->dl_cache); dlce != NULL; dlce = AVL_NEXT(&dl->dl_cache, dlce)) { @@ -381,7 +385,7 @@ dsl_deadlist_free(objset_t *os, uint64_t dlobj, dmu_tx_t *tx) { dmu_object_info_t doi; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; int error; VERIFY0(dmu_object_info(os, dlobj, &doi)); @@ -390,10 +394,11 @@ dsl_deadlist_free(objset_t *os, uint64_t dlobj, dmu_tx_t *tx) return; } + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, os, dlobj); - (error = zap_cursor_retrieve(&zc, &za)) == 0; + (error = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { - uint64_t obj = za.za_first_integer; + uint64_t obj = za->za_first_integer; if (obj == dmu_objset_pool(os)->dp_empty_bpobj) bpobj_decr_empty(os, tx); else @@ -401,6 +406,7 @@ dsl_deadlist_free(objset_t *os, uint64_t dlobj, dmu_tx_t *tx) } VERIFY3U(error, ==, ENOENT); zap_cursor_fini(&zc); + zap_attribute_free(za); VERIFY0(dmu_object_free(os, dlobj, tx)); } @@ -875,8 +881,8 @@ dsl_deadlist_merge(dsl_deadlist_t *dl, uint64_t obj, dmu_tx_t *tx) return; } - za = kmem_alloc(sizeof (*za), KM_SLEEP); - pza = kmem_alloc(sizeof (*pza), KM_SLEEP); + za = zap_attribute_alloc(); + pza = zap_attribute_alloc(); mutex_enter(&dl->dl_lock); /* @@ -913,8 +919,8 @@ dsl_deadlist_merge(dsl_deadlist_t *dl, uint64_t obj, dmu_tx_t *tx) dmu_buf_rele(bonus, FTAG); mutex_exit(&dl->dl_lock); - kmem_free(za, sizeof (*za)); - kmem_free(pza, sizeof (*pza)); + zap_attribute_free(za); + zap_attribute_free(pza); } /* diff --git a/module/zfs/dsl_deleg.c b/module/zfs/dsl_deleg.c index 645ad8e5b8..5e7a2b3fa5 100644 --- a/module/zfs/dsl_deleg.c +++ b/module/zfs/dsl_deleg.c @@ -327,9 +327,9 @@ dsl_deleg_get(const char *ddname, nvlist_t **nvp) mos = dp->dp_meta_objset; zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP); - za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); basezc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP); - baseza = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + baseza = zap_attribute_alloc(); source = kmem_alloc(ZFS_MAX_DATASET_NAME_LEN, KM_SLEEP); VERIFY(nvlist_alloc(nvp, NV_UNIQUE_NAME, KM_SLEEP) == 0); @@ -371,9 +371,9 @@ dsl_deleg_get(const char *ddname, nvlist_t **nvp) } kmem_free(source, ZFS_MAX_DATASET_NAME_LEN); - kmem_free(baseza, sizeof (zap_attribute_t)); + zap_attribute_free(baseza); kmem_free(basezc, sizeof (zap_cursor_t)); - kmem_free(za, sizeof (zap_attribute_t)); + zap_attribute_free(za); kmem_free(zc, sizeof (zap_cursor_t)); dsl_dir_rele(startdd, FTAG); @@ -482,7 +482,7 @@ dsl_load_sets(objset_t *mos, uint64_t zapobj, char type, char checkflag, void *valp, avl_tree_t *avl) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; perm_set_t *permnode; avl_index_t idx; uint64_t jumpobj; @@ -495,11 +495,12 @@ dsl_load_sets(objset_t *mos, uint64_t zapobj, if (error != 0) return (error); + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, jumpobj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { permnode = kmem_alloc(sizeof (perm_set_t), KM_SLEEP); - (void) strlcpy(permnode->p_setname, za.za_name, + (void) strlcpy(permnode->p_setname, za->za_name, sizeof (permnode->p_setname)); permnode->p_matched = B_FALSE; @@ -510,6 +511,7 @@ dsl_load_sets(objset_t *mos, uint64_t zapobj, } } zap_cursor_fini(&zc); + zap_attribute_free(za); return (0); } @@ -683,7 +685,7 @@ copy_create_perms(dsl_dir_t *dd, uint64_t pzapobj, uint64_t jumpobj, pjumpobj; uint64_t zapobj = dsl_dir_phys(dd)->dd_deleg_zapobj; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; char whokey[ZFS_MAX_DELEG_NAME]; zfs_deleg_whokey(whokey, @@ -706,16 +708,18 @@ copy_create_perms(dsl_dir_t *dd, uint64_t pzapobj, VERIFY(zap_add(mos, zapobj, whokey, 8, 1, &jumpobj, tx) == 0); } + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, pjumpobj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { uint64_t zero = 0; - ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1); + ASSERT(za->za_integer_length == 8 && za->za_num_integers == 1); - VERIFY(zap_update(mos, jumpobj, za.za_name, + VERIFY(zap_update(mos, jumpobj, za->za_name, 8, 1, &zero, tx) == 0); } zap_cursor_fini(&zc); + zap_attribute_free(za); } /* @@ -746,19 +750,21 @@ int dsl_deleg_destroy(objset_t *mos, uint64_t zapobj, dmu_tx_t *tx) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; if (zapobj == 0) return (0); + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, zapobj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { - ASSERT(za.za_integer_length == 8 && za.za_num_integers == 1); - VERIFY(0 == zap_destroy(mos, za.za_first_integer, tx)); + ASSERT(za->za_integer_length == 8 && za->za_num_integers == 1); + VERIFY(0 == zap_destroy(mos, za->za_first_integer, tx)); } zap_cursor_fini(&zc); VERIFY(0 == zap_destroy(mos, zapobj, tx)); + zap_attribute_free(za); return (0); } diff --git a/module/zfs/dsl_destroy.c b/module/zfs/dsl_destroy.c index d4a6e5b6e9..b2b925b135 100644 --- a/module/zfs/dsl_destroy.c +++ b/module/zfs/dsl_destroy.c @@ -216,7 +216,7 @@ dsl_dir_remove_clones_key_impl(dsl_dir_t *dd, uint64_t mintxg, dmu_tx_t *tx, return; zap_cursor_t *zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP); - zap_attribute_t *za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(zc, mos, dsl_dir_phys(dd)->dd_clones); zap_cursor_retrieve(zc, za) == 0; @@ -242,7 +242,7 @@ dsl_dir_remove_clones_key_impl(dsl_dir_t *dd, uint64_t mintxg, dmu_tx_t *tx, } zap_cursor_fini(zc); - kmem_free(za, sizeof (zap_attribute_t)); + zap_attribute_free(za); kmem_free(zc, sizeof (zap_cursor_t)); } diff --git a/module/zfs/dsl_dir.c b/module/zfs/dsl_dir.c index baf970121a..5aeff86b94 100644 --- a/module/zfs/dsl_dir.c +++ b/module/zfs/dsl_dir.c @@ -586,7 +586,7 @@ dsl_dir_init_fs_ss_count(dsl_dir_t *dd, dmu_tx_t *tx) return; zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP); - za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); /* Iterate my child dirs */ for (zap_cursor_init(zc, os, dsl_dir_phys(dd)->dd_child_dir_zapobj); @@ -635,7 +635,7 @@ dsl_dir_init_fs_ss_count(dsl_dir_t *dd, dmu_tx_t *tx) dsl_dataset_rele(ds, FTAG); kmem_free(zc, sizeof (zap_cursor_t)); - kmem_free(za, sizeof (zap_attribute_t)); + zap_attribute_free(za); /* we're in a sync task, update counts */ dmu_buf_will_dirty(dd->dd_dbuf, tx); diff --git a/module/zfs/dsl_pool.c b/module/zfs/dsl_pool.c index 342ec5c15c..5ae9688293 100644 --- a/module/zfs/dsl_pool.c +++ b/module/zfs/dsl_pool.c @@ -1196,7 +1196,7 @@ dsl_pool_unlinked_drain_taskq(dsl_pool_t *dp) void dsl_pool_clean_tmp_userrefs(dsl_pool_t *dp) { - zap_attribute_t za; + zap_attribute_t *za; zap_cursor_t zc; objset_t *mos = dp->dp_meta_objset; uint64_t zapobj = dp->dp_tmp_userrefs_obj; @@ -1208,19 +1208,20 @@ dsl_pool_clean_tmp_userrefs(dsl_pool_t *dp) holds = fnvlist_alloc(); + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, zapobj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { char *htag; nvlist_t *tags; - htag = strchr(za.za_name, '-'); + htag = strchr(za->za_name, '-'); *htag = '\0'; ++htag; - if (nvlist_lookup_nvlist(holds, za.za_name, &tags) != 0) { + if (nvlist_lookup_nvlist(holds, za->za_name, &tags) != 0) { tags = fnvlist_alloc(); fnvlist_add_boolean(tags, htag); - fnvlist_add_nvlist(holds, za.za_name, tags); + fnvlist_add_nvlist(holds, za->za_name, tags); fnvlist_free(tags); } else { fnvlist_add_boolean(tags, htag); @@ -1229,6 +1230,7 @@ dsl_pool_clean_tmp_userrefs(dsl_pool_t *dp) dsl_dataset_user_release_tmp(dp, holds); fnvlist_free(holds); zap_cursor_fini(&zc); + zap_attribute_free(za); } /* diff --git a/module/zfs/dsl_prop.c b/module/zfs/dsl_prop.c index 99f931cd86..3b8683593f 100644 --- a/module/zfs/dsl_prop.c +++ b/module/zfs/dsl_prop.c @@ -662,7 +662,7 @@ dsl_prop_changed_notify(dsl_pool_t *dp, uint64_t ddobj, } mutex_exit(&dd->dd_lock); - za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, dsl_dir_phys(dd)->dd_child_dir_zapobj); zap_cursor_retrieve(&zc, za) == 0; @@ -670,7 +670,7 @@ dsl_prop_changed_notify(dsl_pool_t *dp, uint64_t ddobj, dsl_prop_changed_notify(dp, za->za_first_integer, propname, value, FALSE); } - kmem_free(za, sizeof (zap_attribute_t)); + zap_attribute_free(za); zap_cursor_fini(&zc); dsl_dir_rele(dd, FTAG); } @@ -1061,11 +1061,11 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, const char *setpoint, dsl_prop_getflags_t flags, nvlist_t *nv) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); int err = 0; for (zap_cursor_init(&zc, mos, propobj); - (err = zap_cursor_retrieve(&zc, &za)) == 0; + (err = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { nvlist_t *propval; zfs_prop_t prop; @@ -1075,7 +1075,7 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, const char *propname; const char *source; - suffix = strchr(za.za_name, '$'); + suffix = strchr(za->za_name, '$'); if (suffix == NULL) { /* @@ -1085,7 +1085,7 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, if (flags & DSL_PROP_GET_RECEIVED) continue; - propname = za.za_name; + propname = za->za_name; source = setpoint; /* Skip if iuv entries are preset. */ @@ -1102,8 +1102,8 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, if (flags & DSL_PROP_GET_LOCAL) continue; - (void) strlcpy(buf, za.za_name, - MIN(sizeof (buf), suffix - za.za_name + 1)); + (void) strlcpy(buf, za->za_name, + MIN(sizeof (buf), suffix - za->za_name + 1)); propname = buf; if (!(flags & DSL_PROP_GET_RECEIVED)) { @@ -1128,14 +1128,14 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, source = ((flags & DSL_PROP_GET_INHERITING) ? setpoint : ZPROP_SOURCE_VAL_RECVD); } else if (strcmp(suffix, ZPROP_IUV_SUFFIX) == 0) { - (void) strlcpy(buf, za.za_name, - MIN(sizeof (buf), suffix - za.za_name + 1)); + (void) strlcpy(buf, za->za_name, + MIN(sizeof (buf), suffix - za->za_name + 1)); propname = buf; source = setpoint; prop = zfs_name_to_prop(propname); if (dsl_prop_known_index(prop, - za.za_first_integer) != 1) + za->za_first_integer) != 1) continue; } else { /* @@ -1162,28 +1162,28 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, continue; VERIFY(nvlist_alloc(&propval, NV_UNIQUE_NAME, KM_SLEEP) == 0); - if (za.za_integer_length == 1) { + if (za->za_integer_length == 1) { /* * String property */ - char *tmp = kmem_alloc(za.za_num_integers, + char *tmp = kmem_alloc(za->za_num_integers, KM_SLEEP); err = zap_lookup(mos, propobj, - za.za_name, 1, za.za_num_integers, tmp); + za->za_name, 1, za->za_num_integers, tmp); if (err != 0) { - kmem_free(tmp, za.za_num_integers); + kmem_free(tmp, za->za_num_integers); break; } VERIFY(nvlist_add_string(propval, ZPROP_VALUE, tmp) == 0); - kmem_free(tmp, za.za_num_integers); + kmem_free(tmp, za->za_num_integers); } else { /* * Integer property */ - ASSERT(za.za_integer_length == 8); + ASSERT(za->za_integer_length == 8); (void) nvlist_add_uint64(propval, ZPROP_VALUE, - za.za_first_integer); + za->za_first_integer); } VERIFY(nvlist_add_string(propval, ZPROP_SOURCE, source) == 0); @@ -1191,6 +1191,7 @@ dsl_prop_get_all_impl(objset_t *mos, uint64_t propobj, nvlist_free(propval); } zap_cursor_fini(&zc); + zap_attribute_free(za); if (err == ENOENT) err = 0; return (err); diff --git a/module/zfs/dsl_scan.c b/module/zfs/dsl_scan.c index 085cfd3c56..38eb4fab19 100644 --- a/module/zfs/dsl_scan.c +++ b/module/zfs/dsl_scan.c @@ -617,17 +617,18 @@ dsl_scan_init(dsl_pool_t *dp, uint64_t txg) /* reload the queue into the in-core state */ if (scn->scn_phys.scn_queue_obj != 0) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, dp->dp_meta_objset, scn->scn_phys.scn_queue_obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; (void) zap_cursor_advance(&zc)) { scan_ds_queue_insert(scn, - zfs_strtonum(za.za_name, NULL), - za.za_first_integer); + zfs_strtonum(za->za_name, NULL), + za->za_first_integer); } zap_cursor_fini(&zc); + zap_attribute_free(za); } spa_scan_stat_init(spa); @@ -2865,16 +2866,17 @@ dsl_scan_visitds(dsl_scan_t *scn, uint64_t dsobj, dmu_tx_t *tx) if (usenext) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, dp->dp_meta_objset, dsl_dataset_phys(ds)->ds_next_clones_obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; (void) zap_cursor_advance(&zc)) { scan_ds_queue_insert(scn, - zfs_strtonum(za.za_name, NULL), + zfs_strtonum(za->za_name, NULL), dsl_dataset_phys(ds)->ds_creation_txg); } zap_cursor_fini(&zc); + zap_attribute_free(za); } else { VERIFY0(dmu_objset_find_dp(dp, dp->dp_root_dir_obj, enqueue_clones_cb, &ds->ds_object, @@ -4156,7 +4158,7 @@ dsl_errorscrub_sync(dsl_pool_t *dp, dmu_tx_t *tx) zbookmark_phys_t *zb; boolean_t limit_exceeded = B_FALSE; - za = kmem_zalloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); zb = kmem_zalloc(sizeof (zbookmark_phys_t), KM_SLEEP); if (!spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) { @@ -4187,7 +4189,7 @@ dsl_errorscrub_sync(dsl_pool_t *dp, dmu_tx_t *tx) dsl_errorscrub_done(scn, B_TRUE, tx); dsl_errorscrub_sync_state(scn, tx); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); kmem_free(zb, sizeof (*zb)); return; } @@ -4201,7 +4203,7 @@ dsl_errorscrub_sync(dsl_pool_t *dp, dmu_tx_t *tx) zbookmark_err_phys_t head_ds_block; head_ds_cursor = kmem_zalloc(sizeof (zap_cursor_t), KM_SLEEP); - head_ds_attr = kmem_zalloc(sizeof (zap_attribute_t), KM_SLEEP); + head_ds_attr = zap_attribute_alloc(); uint64_t head_ds_err_obj = za->za_first_integer; uint64_t head_ds; @@ -4240,13 +4242,13 @@ dsl_errorscrub_sync(dsl_pool_t *dp, dmu_tx_t *tx) zap_cursor_fini(head_ds_cursor); kmem_free(head_ds_cursor, sizeof (*head_ds_cursor)); - kmem_free(head_ds_attr, sizeof (*head_ds_attr)); + zap_attribute_free(head_ds_attr); if (config_held) dsl_pool_config_exit(dp, FTAG); } - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); kmem_free(zb, sizeof (*zb)); if (!limit_exceeded) dsl_errorscrub_done(scn, B_TRUE, tx); diff --git a/module/zfs/dsl_userhold.c b/module/zfs/dsl_userhold.c index 75953f70f9..ffe208736b 100644 --- a/module/zfs/dsl_userhold.c +++ b/module/zfs/dsl_userhold.c @@ -674,7 +674,7 @@ dsl_dataset_get_holds(const char *dsname, nvlist_t *nvl) zap_attribute_t *za; zap_cursor_t zc; - za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, ds->ds_dir->dd_pool->dp_meta_objset, dsl_dataset_phys(ds)->ds_userrefs_obj); zap_cursor_retrieve(&zc, za) == 0; @@ -683,7 +683,7 @@ dsl_dataset_get_holds(const char *dsname, nvlist_t *nvl) za->za_first_integer); } zap_cursor_fini(&zc); - kmem_free(za, sizeof (zap_attribute_t)); + zap_attribute_free(za); } dsl_dataset_rele(ds, FTAG); dsl_pool_rele(dp, FTAG); diff --git a/module/zfs/sa.c b/module/zfs/sa.c index a1b2ffa870..92deb3c038 100644 --- a/module/zfs/sa.c +++ b/module/zfs/sa.c @@ -840,7 +840,7 @@ sa_attr_table_setup(objset_t *os, const sa_attr_reg_t *reg_attrs, int count) uint64_t attr_value; sa_attr_table_t *tb; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; int registered_count = 0; int i; dmu_objset_type_t ostype = dmu_objset_type(os); @@ -914,11 +914,12 @@ sa_attr_table_setup(objset_t *os, const sa_attr_reg_t *reg_attrs, int count) */ if (sa->sa_reg_attr_obj) { + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, os, sa->sa_reg_attr_obj); - (error = zap_cursor_retrieve(&zc, &za)) == 0; + (error = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { uint64_t value; - value = za.za_first_integer; + value = za->za_first_integer; registered_count++; tb[ATTR_NUM(value)].sa_attr = ATTR_NUM(value); @@ -930,11 +931,12 @@ sa_attr_table_setup(objset_t *os, const sa_attr_reg_t *reg_attrs, int count) continue; } tb[ATTR_NUM(value)].sa_name = - kmem_zalloc(strlen(za.za_name) +1, KM_SLEEP); - (void) strlcpy(tb[ATTR_NUM(value)].sa_name, za.za_name, - strlen(za.za_name) +1); + kmem_zalloc(strlen(za->za_name) +1, KM_SLEEP); + (void) strlcpy(tb[ATTR_NUM(value)].sa_name, za->za_name, + strlen(za->za_name) +1); } zap_cursor_fini(&zc); + zap_attribute_free(za); /* * Make sure we processed the correct number of registered * attributes @@ -996,7 +998,7 @@ sa_setup(objset_t *os, uint64_t sa_obj, const sa_attr_reg_t *reg_attrs, int count, sa_attr_type_t **user_table) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; sa_os_t *sa; dmu_objset_type_t ostype = dmu_objset_type(os); sa_attr_type_t *tb; @@ -1053,33 +1055,35 @@ sa_setup(objset_t *os, uint64_t sa_obj, const sa_attr_reg_t *reg_attrs, goto fail; } + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, os, sa->sa_layout_attr_obj); - (error = zap_cursor_retrieve(&zc, &za)) == 0; + (error = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { sa_attr_type_t *lot_attrs; uint64_t lot_num; lot_attrs = kmem_zalloc(sizeof (sa_attr_type_t) * - za.za_num_integers, KM_SLEEP); + za->za_num_integers, KM_SLEEP); if ((error = (zap_lookup(os, sa->sa_layout_attr_obj, - za.za_name, 2, za.za_num_integers, + za->za_name, 2, za->za_num_integers, lot_attrs))) != 0) { kmem_free(lot_attrs, sizeof (sa_attr_type_t) * - za.za_num_integers); + za->za_num_integers); break; } - VERIFY0(ddi_strtoull(za.za_name, NULL, 10, + VERIFY0(ddi_strtoull(za->za_name, NULL, 10, (unsigned long long *)&lot_num)); (void) sa_add_layout_entry(os, lot_attrs, - za.za_num_integers, lot_num, + za->za_num_integers, lot_num, sa_layout_info_hash(lot_attrs, - za.za_num_integers), B_FALSE, NULL); + za->za_num_integers), B_FALSE, NULL); kmem_free(lot_attrs, sizeof (sa_attr_type_t) * - za.za_num_integers); + za->za_num_integers); } zap_cursor_fini(&zc); + zap_attribute_free(za); /* * Make sure layout count matches number of entries added diff --git a/module/zfs/spa.c b/module/zfs/spa.c index 66f766cf9b..c5d466d952 100644 --- a/module/zfs/spa.c +++ b/module/zfs/spa.c @@ -548,7 +548,7 @@ spa_prop_get(spa_t *spa, nvlist_t **nvp) { objset_t *mos = spa->spa_meta_objset; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; dsl_pool_t *dp; int err; @@ -560,6 +560,7 @@ spa_prop_get(spa_t *spa, nvlist_t **nvp) dp = spa_get_dsl(spa); dsl_pool_config_enter(dp, FTAG); + za = zap_attribute_alloc(); mutex_enter(&spa->spa_props_lock); /* @@ -575,21 +576,21 @@ spa_prop_get(spa_t *spa, nvlist_t **nvp) * Get properties from the MOS pool property object. */ for (zap_cursor_init(&zc, mos, spa->spa_pool_props_object); - (err = zap_cursor_retrieve(&zc, &za)) == 0; + (err = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { uint64_t intval = 0; char *strval = NULL; zprop_source_t src = ZPROP_SRC_DEFAULT; zpool_prop_t prop; - if ((prop = zpool_name_to_prop(za.za_name)) == - ZPOOL_PROP_INVAL && !zfs_prop_user(za.za_name)) + if ((prop = zpool_name_to_prop(za->za_name)) == + ZPOOL_PROP_INVAL && !zfs_prop_user(za->za_name)) continue; - switch (za.za_integer_length) { + switch (za->za_integer_length) { case 8: /* integer property */ - if (za.za_first_integer != + if (za->za_first_integer != zpool_prop_default_numeric(prop)) src = ZPROP_SRC_LOCAL; @@ -597,7 +598,7 @@ spa_prop_get(spa_t *spa, nvlist_t **nvp) dsl_dataset_t *ds = NULL; err = dsl_dataset_hold_obj(dp, - za.za_first_integer, FTAG, &ds); + za->za_first_integer, FTAG, &ds); if (err != 0) break; @@ -607,7 +608,7 @@ spa_prop_get(spa_t *spa, nvlist_t **nvp) dsl_dataset_rele(ds, FTAG); } else { strval = NULL; - intval = za.za_first_integer; + intval = za->za_first_integer; } spa_prop_add_list(*nvp, prop, strval, intval, src); @@ -619,21 +620,21 @@ spa_prop_get(spa_t *spa, nvlist_t **nvp) case 1: /* string property */ - strval = kmem_alloc(za.za_num_integers, KM_SLEEP); + strval = kmem_alloc(za->za_num_integers, KM_SLEEP); err = zap_lookup(mos, spa->spa_pool_props_object, - za.za_name, 1, za.za_num_integers, strval); + za->za_name, 1, za->za_num_integers, strval); if (err) { - kmem_free(strval, za.za_num_integers); + kmem_free(strval, za->za_num_integers); break; } if (prop != ZPOOL_PROP_INVAL) { spa_prop_add_list(*nvp, prop, strval, 0, src); } else { src = ZPROP_SRC_LOCAL; - spa_prop_add_user(*nvp, za.za_name, strval, + spa_prop_add_user(*nvp, za->za_name, strval, src); } - kmem_free(strval, za.za_num_integers); + kmem_free(strval, za->za_num_integers); break; default: @@ -644,6 +645,7 @@ spa_prop_get(spa_t *spa, nvlist_t **nvp) out: mutex_exit(&spa->spa_props_lock); dsl_pool_config_exit(dp, FTAG); + zap_attribute_free(za); if (err && err != ENOENT) { nvlist_free(*nvp); *nvp = NULL; @@ -2967,12 +2969,13 @@ dsl_get_next_livelist_obj(objset_t *os, uint64_t zap_obj, uint64_t *llp) { int err; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); zap_cursor_init(&zc, os, zap_obj); - err = zap_cursor_retrieve(&zc, &za); + err = zap_cursor_retrieve(&zc, za); zap_cursor_fini(&zc); if (err == 0) - *llp = za.za_first_integer; + *llp = za->za_first_integer; + zap_attribute_free(za); return (err); } @@ -6043,17 +6046,17 @@ static void spa_feature_stats_from_disk(spa_t *spa, nvlist_t *features) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); if (spa->spa_feat_for_read_obj != 0) { for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_feat_for_read_obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { - ASSERT(za.za_integer_length == sizeof (uint64_t) && - za.za_num_integers == 1); - VERIFY0(nvlist_add_uint64(features, za.za_name, - za.za_first_integer)); + ASSERT(za->za_integer_length == sizeof (uint64_t) && + za->za_num_integers == 1); + VERIFY0(nvlist_add_uint64(features, za->za_name, + za->za_first_integer)); } zap_cursor_fini(&zc); } @@ -6061,15 +6064,16 @@ spa_feature_stats_from_disk(spa_t *spa, nvlist_t *features) if (spa->spa_feat_for_write_obj != 0) { for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_feat_for_write_obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { - ASSERT(za.za_integer_length == sizeof (uint64_t) && - za.za_num_integers == 1); - VERIFY0(nvlist_add_uint64(features, za.za_name, - za.za_first_integer)); + ASSERT(za->za_integer_length == sizeof (uint64_t) && + za->za_num_integers == 1); + VERIFY0(nvlist_add_uint64(features, za->za_name, + za->za_first_integer)); } zap_cursor_fini(&zc); } + zap_attribute_free(za); } static void @@ -9425,13 +9429,13 @@ spa_sync_config_object(spa_t *spa, dmu_tx_t *tx) /* Diff old AVZ with new one */ zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_all_vdev_zaps); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { - uint64_t vdzap = za.za_first_integer; + uint64_t vdzap = za->za_first_integer; if (zap_lookup_int(spa->spa_meta_objset, new_avz, vdzap) == ENOENT) { /* @@ -9444,6 +9448,7 @@ spa_sync_config_object(spa_t *spa, dmu_tx_t *tx) } zap_cursor_fini(&zc); + zap_attribute_free(za); /* Destroy the old AVZ */ VERIFY0(zap_destroy(spa->spa_meta_objset, @@ -9457,18 +9462,19 @@ spa_sync_config_object(spa_t *spa, dmu_tx_t *tx) spa->spa_all_vdev_zaps = new_avz; } else if (spa->spa_avz_action == AVZ_ACTION_DESTROY) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); /* Walk through the AVZ and destroy all listed ZAPs */ for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_all_vdev_zaps); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { - uint64_t zap = za.za_first_integer; + uint64_t zap = za->za_first_integer; VERIFY0(zap_destroy(spa->spa_meta_objset, zap, tx)); } zap_cursor_fini(&zc); + zap_attribute_free(za); /* Destroy and unlink the AVZ itself */ VERIFY0(zap_destroy(spa->spa_meta_objset, diff --git a/module/zfs/spa_errlog.c b/module/zfs/spa_errlog.c index 62d7b4fa2d..a49e28ee7a 100644 --- a/module/zfs/spa_errlog.c +++ b/module/zfs/spa_errlog.c @@ -429,7 +429,7 @@ check_filesystem(spa_t *spa, uint64_t head_ds, zbookmark_err_phys_t *zep, zap_attribute_t *za; zc = kmem_zalloc(sizeof (zap_cursor_t), KM_SLEEP); - za = kmem_zalloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); for (zap_cursor_init(zc, spa->spa_meta_objset, zap_clone); zap_cursor_retrieve(zc, za) == 0; @@ -463,7 +463,7 @@ check_filesystem(spa_t *spa, uint64_t head_ds, zbookmark_err_phys_t *zep, } zap_cursor_fini(zc); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); kmem_free(zc, sizeof (*zc)); out: @@ -617,10 +617,10 @@ spa_add_healed_error(spa_t *spa, uint64_t obj, zbookmark_phys_t *healed_zb, errphys_to_name(&healed_zep, name, sizeof (name)); zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_errlog_last); - zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { - if (zap_contains(spa->spa_meta_objset, za.za_first_integer, + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { + if (zap_contains(spa->spa_meta_objset, za->za_first_integer, name) == 0) { if (!MUTEX_HELD(&spa->spa_errlog_lock)) { mutex_enter(&spa->spa_errlog_lock); @@ -657,6 +657,7 @@ spa_add_healed_error(spa_t *spa, uint64_t obj, zbookmark_phys_t *healed_zb, } } zap_cursor_fini(&zc); + zap_attribute_free(za); } /* @@ -704,24 +705,25 @@ spa_remove_healed_errors(spa_t *spa, avl_tree_t *s, avl_tree_t *l, dmu_tx_t *tx) } else { errphys_to_name(&se->se_zep, name, sizeof (name)); zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_errlog_last); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { zap_remove(spa->spa_meta_objset, - za.za_first_integer, name, tx); + za->za_first_integer, name, tx); } zap_cursor_fini(&zc); for (zap_cursor_init(&zc, spa->spa_meta_objset, spa->spa_errlog_scrub); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { zap_remove(spa->spa_meta_objset, - za.za_first_integer, name, tx); + za->za_first_integer, name, tx); } zap_cursor_fini(&zc); + zap_attribute_free(za); } kmem_free(se, sizeof (spa_error_entry_t)); } @@ -746,15 +748,16 @@ approx_errlog_size_impl(spa_t *spa, uint64_t spa_err_obj) uint64_t total = 0; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, spa->spa_meta_objset, spa_err_obj); - zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { uint64_t count; - if (zap_count(spa->spa_meta_objset, za.za_first_integer, + if (zap_count(spa->spa_meta_objset, za->za_first_integer, &count) == 0) total += count; } zap_cursor_fini(&zc); + zap_attribute_free(za); return (total); } @@ -806,7 +809,7 @@ sync_upgrade_errlog(spa_t *spa, uint64_t spa_err_obj, uint64_t *newobj, dmu_tx_t *tx) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; zbookmark_phys_t zb; uint64_t count; @@ -822,14 +825,15 @@ sync_upgrade_errlog(spa_t *spa, uint64_t spa_err_obj, uint64_t *newobj, return; } + za = zap_attribute_alloc(); for (zap_cursor_init(&zc, spa->spa_meta_objset, spa_err_obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { if (spa_upgrade_errlog_limit != 0 && zc.zc_cd == spa_upgrade_errlog_limit) break; - name_to_bookmark(za.za_name, &zb); + name_to_bookmark(za->za_name, &zb); zbookmark_err_phys_t zep; zep.zb_object = zb.zb_object; @@ -909,6 +913,7 @@ sync_upgrade_errlog(spa_t *spa, uint64_t spa_err_obj, uint64_t *newobj, buf, 1, strlen(name) + 1, name, tx); } zap_cursor_fini(&zc); + zap_attribute_free(za); VERIFY0(dmu_object_free(spa->spa_meta_objset, spa_err_obj, tx)); } @@ -954,7 +959,7 @@ process_error_log(spa_t *spa, uint64_t obj, void *uaddr, uint64_t *count) zap_attribute_t *za; zc = kmem_zalloc(sizeof (zap_cursor_t), KM_SLEEP); - za = kmem_zalloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); if (!spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) { for (zap_cursor_init(zc, spa->spa_meta_objset, obj); @@ -963,7 +968,7 @@ process_error_log(spa_t *spa, uint64_t obj, void *uaddr, uint64_t *count) if (*count == 0) { zap_cursor_fini(zc); kmem_free(zc, sizeof (*zc)); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); return (SET_ERROR(ENOMEM)); } @@ -974,13 +979,13 @@ process_error_log(spa_t *spa, uint64_t obj, void *uaddr, uint64_t *count) if (error != 0) { zap_cursor_fini(zc); kmem_free(zc, sizeof (*zc)); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); return (error); } } zap_cursor_fini(zc); kmem_free(zc, sizeof (*zc)); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); return (0); } @@ -992,7 +997,7 @@ process_error_log(spa_t *spa, uint64_t obj, void *uaddr, uint64_t *count) zap_attribute_t *head_ds_attr; head_ds_cursor = kmem_zalloc(sizeof (zap_cursor_t), KM_SLEEP); - head_ds_attr = kmem_zalloc(sizeof (zap_attribute_t), KM_SLEEP); + head_ds_attr = zap_attribute_alloc(); uint64_t head_ds_err_obj = za->za_first_integer; uint64_t head_ds; @@ -1010,20 +1015,20 @@ process_error_log(spa_t *spa, uint64_t obj, void *uaddr, uint64_t *count) zap_cursor_fini(head_ds_cursor); kmem_free(head_ds_cursor, sizeof (*head_ds_cursor)); - kmem_free(head_ds_attr, sizeof (*head_ds_attr)); + zap_attribute_free(head_ds_attr); zap_cursor_fini(zc); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); kmem_free(zc, sizeof (*zc)); return (error); } } zap_cursor_fini(head_ds_cursor); kmem_free(head_ds_cursor, sizeof (*head_ds_cursor)); - kmem_free(head_ds_attr, sizeof (*head_ds_attr)); + zap_attribute_free(head_ds_attr); } zap_cursor_fini(zc); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); kmem_free(zc, sizeof (*zc)); return (0); } @@ -1230,14 +1235,15 @@ delete_errlog(spa_t *spa, uint64_t spa_err_obj, dmu_tx_t *tx) { if (spa_feature_is_enabled(spa, SPA_FEATURE_HEAD_ERRLOG)) { zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, spa->spa_meta_objset, spa_err_obj); - zap_cursor_retrieve(&zc, &za) == 0; + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { VERIFY0(dmu_object_free(spa->spa_meta_objset, - za.za_first_integer, tx)); + za->za_first_integer, tx)); } zap_cursor_fini(&zc); + zap_attribute_free(za); } VERIFY0(dmu_object_free(spa->spa_meta_objset, spa_err_obj, tx)); } @@ -1339,20 +1345,21 @@ delete_dataset_errlog(spa_t *spa, uint64_t spa_err_obj, uint64_t ds, return; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, spa->spa_meta_objset, spa_err_obj); - zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { uint64_t head_ds; - name_to_object(za.za_name, &head_ds); + name_to_object(za->za_name, &head_ds); if (head_ds == ds) { (void) zap_remove(spa->spa_meta_objset, spa_err_obj, - za.za_name, tx); + za->za_name, tx); VERIFY0(dmu_object_free(spa->spa_meta_objset, - za.za_first_integer, tx)); + za->za_first_integer, tx)); break; } } zap_cursor_fini(&zc); + zap_attribute_free(za); } void @@ -1435,22 +1442,23 @@ swap_errlog(spa_t *spa, uint64_t spa_err_obj, uint64_t new_head, uint64_t } zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); zbookmark_err_phys_t err_block; for (zap_cursor_init(&zc, spa->spa_meta_objset, old_head_errlog); - zap_cursor_retrieve(&zc, &za) == 0; zap_cursor_advance(&zc)) { + zap_cursor_retrieve(&zc, za) == 0; zap_cursor_advance(&zc)) { const char *name = ""; - name_to_errphys(za.za_name, &err_block); + name_to_errphys(za->za_name, &err_block); if (err_block.zb_birth < txg) { (void) zap_update(spa->spa_meta_objset, new_head_errlog, - za.za_name, 1, strlen(name) + 1, name, tx); + za->za_name, 1, strlen(name) + 1, name, tx); (void) zap_remove(spa->spa_meta_objset, old_head_errlog, - za.za_name, tx); + za->za_name, tx); } } zap_cursor_fini(&zc); + zap_attribute_free(za); } void diff --git a/module/zfs/spa_log_spacemap.c b/module/zfs/spa_log_spacemap.c index 32158e8c59..f55218e357 100644 --- a/module/zfs/spa_log_spacemap.c +++ b/module/zfs/spa_log_spacemap.c @@ -1020,16 +1020,17 @@ spa_ld_log_sm_metadata(spa_t *spa) } zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, spa_meta_objset(spa), spacemap_zap); - (error = zap_cursor_retrieve(&zc, &za)) == 0; + (error = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { - uint64_t log_txg = zfs_strtonum(za.za_name, NULL); + uint64_t log_txg = zfs_strtonum(za->za_name, NULL); spa_log_sm_t *sls = - spa_log_sm_alloc(za.za_first_integer, log_txg); + spa_log_sm_alloc(za->za_first_integer, log_txg); avl_add(&spa->spa_sm_logs_by_txg, sls); } zap_cursor_fini(&zc); + zap_attribute_free(za); if (error != ENOENT) { spa_load_failed(spa, "spa_ld_log_sm_metadata(): failed at " "zap_cursor_retrieve(spacemap_zap) [error %d]", diff --git a/module/zfs/spa_misc.c b/module/zfs/spa_misc.c index 97191e7685..f486513fca 100644 --- a/module/zfs/spa_misc.c +++ b/module/zfs/spa_misc.c @@ -2590,6 +2590,7 @@ spa_init(spa_mode_t mode) scan_init(); qat_init(); spa_import_progress_init(); + zap_init(); } void @@ -2616,6 +2617,7 @@ spa_fini(void) scan_fini(); qat_fini(); spa_import_progress_destroy(); + zap_fini(); avl_destroy(&spa_namespace_avl); avl_destroy(&spa_spare_avl); diff --git a/module/zfs/vdev.c b/module/zfs/vdev.c index 6ae0a14127..5098154491 100644 --- a/module/zfs/vdev.c +++ b/module/zfs/vdev.c @@ -6432,33 +6432,33 @@ vdev_prop_get(vdev_t *vd, nvlist_t *innvl, nvlist_t *outnvl) * Get all properties from the MOS vdev property object. */ zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, mos, objid); - (err = zap_cursor_retrieve(&zc, &za)) == 0; + (err = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { intval = 0; strval = NULL; zprop_source_t src = ZPROP_SRC_DEFAULT; - propname = za.za_name; + propname = za->za_name; - switch (za.za_integer_length) { + switch (za->za_integer_length) { case 8: /* We do not allow integer user properties */ /* This is likely an internal value */ break; case 1: /* string property */ - strval = kmem_alloc(za.za_num_integers, + strval = kmem_alloc(za->za_num_integers, KM_SLEEP); - err = zap_lookup(mos, objid, za.za_name, 1, - za.za_num_integers, strval); + err = zap_lookup(mos, objid, za->za_name, 1, + za->za_num_integers, strval); if (err) { - kmem_free(strval, za.za_num_integers); + kmem_free(strval, za->za_num_integers); break; } vdev_prop_add_list(outnvl, propname, strval, 0, src); - kmem_free(strval, za.za_num_integers); + kmem_free(strval, za->za_num_integers); break; default: @@ -6466,6 +6466,7 @@ vdev_prop_get(vdev_t *vd, nvlist_t *innvl, nvlist_t *outnvl) } } zap_cursor_fini(&zc); + zap_attribute_free(za); } mutex_exit(&spa->spa_props_lock); diff --git a/module/zfs/zap.c b/module/zfs/zap.c index 03b76ea1b7..603c76a2cd 100644 --- a/module/zfs/zap.c +++ b/module/zfs/zap.c @@ -1110,7 +1110,7 @@ zap_value_search(objset_t *os, uint64_t zapobj, uint64_t value, uint64_t mask, if (mask == 0) mask = -1ULL; - zap_attribute_t *za = kmem_alloc(sizeof (*za), KM_SLEEP); + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, os, zapobj); (err = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { @@ -1120,7 +1120,7 @@ zap_value_search(objset_t *os, uint64_t zapobj, uint64_t value, uint64_t mask, } } zap_cursor_fini(&zc); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); return (err); } @@ -1130,7 +1130,7 @@ zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx) zap_cursor_t zc; int err = 0; - zap_attribute_t *za = kmem_alloc(sizeof (*za), KM_SLEEP); + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, os, fromobj); zap_cursor_retrieve(&zc, za) == 0; (void) zap_cursor_advance(&zc)) { @@ -1144,7 +1144,7 @@ zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx) break; } zap_cursor_fini(&zc); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); return (err); } @@ -1155,7 +1155,7 @@ zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj, zap_cursor_t zc; int err = 0; - zap_attribute_t *za = kmem_alloc(sizeof (*za), KM_SLEEP); + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, os, fromobj); zap_cursor_retrieve(&zc, za) == 0; (void) zap_cursor_advance(&zc)) { @@ -1169,7 +1169,7 @@ zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj, break; } zap_cursor_fini(&zc); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); return (err); } @@ -1180,7 +1180,7 @@ zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj, zap_cursor_t zc; int err = 0; - zap_attribute_t *za = kmem_alloc(sizeof (*za), KM_SLEEP); + zap_attribute_t *za = zap_attribute_alloc(); for (zap_cursor_init(&zc, os, fromobj); zap_cursor_retrieve(&zc, za) == 0; (void) zap_cursor_advance(&zc)) { @@ -1200,7 +1200,7 @@ zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj, break; } zap_cursor_fini(&zc); - kmem_free(za, sizeof (*za)); + zap_attribute_free(za); return (err); } @@ -1378,7 +1378,7 @@ again: ASSERT(err == 0 || err == EOVERFLOW); } err = zap_entry_read_name(zap, &zeh, - sizeof (za->za_name), za->za_name); + za->za_name_len, za->za_name); ASSERT(err == 0); za->za_normalization_conflict = diff --git a/module/zfs/zap_micro.c b/module/zfs/zap_micro.c index 0d6533b0e1..e52680dd40 100644 --- a/module/zfs/zap_micro.c +++ b/module/zfs/zap_micro.c @@ -164,10 +164,32 @@ zap_match(zap_name_t *zn, const char *matchname) } } +static kmem_cache_t *zap_name_cache; +static kmem_cache_t *zap_attr_cache; + +void +zap_init(void) +{ + zap_name_cache = kmem_cache_create("zap_name", + sizeof (zap_name_t) + ZAP_MAXNAMELEN, 0, NULL, NULL, + NULL, NULL, NULL, 0); + + zap_attr_cache = kmem_cache_create("zap_attr_cache", + sizeof (zap_attribute_t) + ZAP_MAXNAMELEN, 0, NULL, + NULL, NULL, NULL, NULL, 0); +} + +void +zap_fini(void) +{ + kmem_cache_destroy(zap_name_cache); + kmem_cache_destroy(zap_attr_cache); +} + static zap_name_t * zap_name_alloc(zap_t *zap) { - zap_name_t *zn = kmem_alloc(sizeof (zap_name_t), KM_SLEEP); + zap_name_t *zn = kmem_cache_alloc(zap_name_cache, KM_SLEEP); zn->zn_zap = zap; return (zn); } @@ -175,7 +197,7 @@ zap_name_alloc(zap_t *zap) void zap_name_free(zap_name_t *zn) { - kmem_free(zn, sizeof (zap_name_t)); + kmem_cache_free(zap_name_cache, zn); } static int @@ -188,6 +210,7 @@ zap_name_init_str(zap_name_t *zn, const char *key, matchtype_t mt) zn->zn_key_orig_numints = strlen(zn->zn_key_orig) + 1; zn->zn_matchtype = mt; zn->zn_normflags = zap->zap_normflags; + zn->zn_normbuf_len = ZAP_MAXNAMELEN; /* * If we're dealing with a case sensitive lookup on a mixed or @@ -244,7 +267,7 @@ zap_name_alloc_str(zap_t *zap, const char *key, matchtype_t mt) static zap_name_t * zap_name_alloc_uint64(zap_t *zap, const uint64_t *key, int numints) { - zap_name_t *zn = kmem_alloc(sizeof (zap_name_t), KM_SLEEP); + zap_name_t *zn = kmem_cache_alloc(zap_name_cache, KM_SLEEP); ASSERT(zap->zap_normflags == 0); zn->zn_zap = zap; @@ -252,6 +275,7 @@ zap_name_alloc_uint64(zap_t *zap, const uint64_t *key, int numints) zn->zn_key_orig = zn->zn_key_norm = key; zn->zn_key_orig_numints = zn->zn_key_norm_numints = numints; zn->zn_matchtype = 0; + zn->zn_normbuf_len = ZAP_MAXNAMELEN; zn->zn_hash = zap_hash(zn); return (zn); @@ -1600,6 +1624,24 @@ zap_remove_uint64_by_dnode(dnode_t *dn, const uint64_t *key, int key_numints, return (err); } + +zap_attribute_t * +zap_attribute_alloc(void) +{ + uint32_t len = ZAP_MAXNAMELEN; + zap_attribute_t *za; + + za = kmem_cache_alloc(zap_attr_cache, KM_SLEEP); + za->za_name_len = len; + return (za); +} + +void +zap_attribute_free(zap_attribute_t *za) +{ + kmem_cache_free(zap_attr_cache, za); +} + /* * Routines for iterating over the attributes. */ @@ -1736,7 +1778,7 @@ zap_cursor_retrieve(zap_cursor_t *zc, zap_attribute_t *za) za->za_num_integers = 1; za->za_first_integer = mzep->mze_value; (void) strlcpy(za->za_name, mzep->mze_name, - sizeof (za->za_name)); + za->za_name_len); zc->zc_hash = (uint64_t)mze->mze_hash << 32; zc->zc_cd = mze->mze_cd; err = 0; diff --git a/module/zfs/zcp_iter.c b/module/zfs/zcp_iter.c index 2da0bf9740..b4aa7a1317 100644 --- a/module/zfs/zcp_iter.c +++ b/module/zfs/zcp_iter.c @@ -54,7 +54,7 @@ zcp_clones_iter(lua_State *state) uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2)); dsl_pool_t *dp = zcp_run_info(state)->zri_pool; dsl_dataset_t *ds, *clone; - zap_attribute_t za; + zap_attribute_t *za; zap_cursor_t zc; err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds); @@ -75,9 +75,11 @@ zcp_clones_iter(lua_State *state) dsl_dataset_phys(ds)->ds_next_clones_obj, cursor); dsl_dataset_rele(ds, FTAG); - err = zap_cursor_retrieve(&zc, &za); + za = zap_attribute_alloc(); + err = zap_cursor_retrieve(&zc, za); if (err != 0) { zap_cursor_fini(&zc); + zap_attribute_free(za); if (err != ENOENT) { return (luaL_error(state, "unexpected error %d from zap_cursor_retrieve()", @@ -89,7 +91,8 @@ zcp_clones_iter(lua_State *state) cursor = zap_cursor_serialize(&zc); zap_cursor_fini(&zc); - err = dsl_dataset_hold_obj(dp, za.za_first_integer, FTAG, &clone); + err = dsl_dataset_hold_obj(dp, za->za_first_integer, FTAG, &clone); + zap_attribute_free(za); if (err != 0) { return (luaL_error(state, "unexpected error %d from " @@ -499,7 +502,7 @@ zcp_bookmarks_iter(lua_State *state) uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2)); dsl_pool_t *dp = zcp_run_info(state)->zri_pool; dsl_dataset_t *ds; - zap_attribute_t za; + zap_attribute_t *za; zap_cursor_t zc; int err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds); @@ -536,9 +539,11 @@ zcp_bookmarks_iter(lua_State *state) ds->ds_bookmarks_obj, cursor); dsl_dataset_rele(ds, FTAG); - err = zap_cursor_retrieve(&zc, &za); + za = zap_attribute_alloc(); + err = zap_cursor_retrieve(&zc, za); if (err != 0) { zap_cursor_fini(&zc); + zap_attribute_free(za); if (err != ENOENT) { return (luaL_error(state, "unexpected error %d from zap_cursor_retrieve()", @@ -552,7 +557,8 @@ zcp_bookmarks_iter(lua_State *state) /* Create the full "pool/fs#bookmark" string to return */ int n = snprintf(bookmark_name, ZFS_MAX_DATASET_NAME_LEN, "%s#%s", - ds_name, za.za_name); + ds_name, za->za_name); + zap_attribute_free(za); if (n >= ZFS_MAX_DATASET_NAME_LEN) { return (luaL_error(state, "unexpected error %d from snprintf()", ENAMETOOLONG)); @@ -610,7 +616,7 @@ zcp_holds_iter(lua_State *state) uint64_t cursor = lua_tonumber(state, lua_upvalueindex(2)); dsl_pool_t *dp = zcp_run_info(state)->zri_pool; dsl_dataset_t *ds; - zap_attribute_t za; + zap_attribute_t *za; zap_cursor_t zc; int err = dsl_dataset_hold_obj(dp, dsobj, FTAG, &ds); @@ -631,9 +637,11 @@ zcp_holds_iter(lua_State *state) dsl_dataset_phys(ds)->ds_userrefs_obj, cursor); dsl_dataset_rele(ds, FTAG); - err = zap_cursor_retrieve(&zc, &za); + za = zap_attribute_alloc(); + err = zap_cursor_retrieve(&zc, za); if (err != 0) { zap_cursor_fini(&zc); + zap_attribute_free(za); if (err != ENOENT) { return (luaL_error(state, "unexpected error %d from zap_cursor_retrieve()", @@ -648,8 +656,9 @@ zcp_holds_iter(lua_State *state) lua_pushnumber(state, cursor); lua_replace(state, lua_upvalueindex(2)); - (void) lua_pushstring(state, za.za_name); - (void) lua_pushnumber(state, za.za_first_integer); + (void) lua_pushstring(state, za->za_name); + (void) lua_pushnumber(state, za->za_first_integer); + zap_attribute_free(za); return (2); } diff --git a/module/zfs/zfeature.c b/module/zfs/zfeature.c index 1d25bc4068..38479c7225 100644 --- a/module/zfs/zfeature.c +++ b/module/zfs/zfeature.c @@ -183,7 +183,7 @@ spa_features_check(spa_t *spa, boolean_t for_write, char *buf; zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP); - za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + za = zap_attribute_alloc(); buf = kmem_alloc(MAXPATHLEN, KM_SLEEP); supported = B_TRUE; @@ -217,7 +217,7 @@ spa_features_check(spa_t *spa, boolean_t for_write, zap_cursor_fini(zc); kmem_free(buf, MAXPATHLEN); - kmem_free(za, sizeof (zap_attribute_t)); + zap_attribute_free(za); kmem_free(zc, sizeof (zap_cursor_t)); return (supported); diff --git a/module/zfs/zfs_quota.c b/module/zfs/zfs_quota.c index 9b351eefc0..0fe152bc4d 100644 --- a/module/zfs/zfs_quota.c +++ b/module/zfs/zfs_quota.c @@ -165,7 +165,7 @@ zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, { int error; zap_cursor_t zc; - zap_attribute_t za; + zap_attribute_t *za; zfs_useracct_t *buf = vbuf; uint64_t obj; int offset = 0; @@ -196,8 +196,9 @@ zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, type == ZFS_PROP_PROJECTOBJUSED) offset = DMU_OBJACCT_PREFIX_LEN; + za = zap_attribute_alloc(); for (zap_cursor_init_serialized(&zc, zfsvfs->z_os, obj, *cookiep); - (error = zap_cursor_retrieve(&zc, &za)) == 0; + (error = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { if ((uintptr_t)buf - (uintptr_t)vbuf + sizeof (zfs_useracct_t) > *bufsizep) @@ -207,14 +208,14 @@ zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, * skip object quota (with zap name prefix DMU_OBJACCT_PREFIX) * when dealing with block quota and vice versa. */ - if ((offset > 0) != (strncmp(za.za_name, DMU_OBJACCT_PREFIX, + if ((offset > 0) != (strncmp(za->za_name, DMU_OBJACCT_PREFIX, DMU_OBJACCT_PREFIX_LEN) == 0)) continue; - fuidstr_to_sid(zfsvfs, za.za_name + offset, + fuidstr_to_sid(zfsvfs, za->za_name + offset, buf->zu_domain, sizeof (buf->zu_domain), &buf->zu_rid); - buf->zu_space = za.za_first_integer; + buf->zu_space = za->za_first_integer; buf++; } if (error == ENOENT) @@ -224,6 +225,7 @@ zfs_userspace_many(zfsvfs_t *zfsvfs, zfs_userquota_prop_t type, *bufsizep = (uintptr_t)buf - (uintptr_t)vbuf; *cookiep = zap_cursor_serialize(&zc); zap_cursor_fini(&zc); + zap_attribute_free(za); return (error); } diff --git a/module/zfs/zvol.c b/module/zfs/zvol.c index 5b6a3f5cb4..6f5edb2dd6 100644 --- a/module/zfs/zvol.c +++ b/module/zfs/zvol.c @@ -1027,7 +1027,7 @@ zvol_add_clones(const char *dsname, list_t *minors_list) goto out; zap_cursor_t *zc = kmem_alloc(sizeof (zap_cursor_t), KM_SLEEP); - zap_attribute_t *za = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); + zap_attribute_t *za = zap_attribute_alloc(); objset_t *mos = dd->dd_pool->dp_meta_objset; for (zap_cursor_init(zc, mos, dsl_dir_phys(dd)->dd_clones); @@ -1053,7 +1053,7 @@ zvol_add_clones(const char *dsname, list_t *minors_list) } } zap_cursor_fini(zc); - kmem_free(za, sizeof (zap_attribute_t)); + zap_attribute_free(za); kmem_free(zc, sizeof (zap_cursor_t)); out: From f54bcc9c40427a15c283acbb9602205f932b69e0 Mon Sep 17 00:00:00 2001 From: Sanjeev Bagewadi Date: Fri, 18 Jun 2021 08:55:01 +0000 Subject: [PATCH 2/2] Support for longnames for files/directories (Linux part) This patch adds the ability for zfs to support file/dir name up to 1023 bytes. This number is chosen so we can support up to 255 4-byte characters. This new feature is represented by the new feature flag feature@longname. A new dataset property "longname" is also introduced to toggle longname support for each dataset individually. This property can be disabled, even if it contains longname files. In such case, new file cannot be created with longname but existing longname files can still be looked up. Note that, to my knowledge native Linux filesystems don't support name longer than 255 bytes. So there might be programs not able to work with longname. Note that NFS server may needs to use exportfs_get_name to reconnect dentries, and the buffer being passed is limit to NAME_MAX+1 (256). So NFS may not work when longname is enabled. Signed-off-by: Chunwei Chen --- cmd/zdb/zdb.c | 4 +- cmd/zhack.c | 2 +- include/os/linux/zfs/sys/zfs_vfsops_os.h | 1 + include/os/linux/zfs/sys/zfs_vnops_os.h | 1 + include/sys/fs/zfs.h | 2 + include/sys/zap.h | 3 +- include/sys/zfs_ioctl.h | 4 +- include/zfeature_common.h | 1 + lib/libzfs/libzfs.abi | 73 ++++----- module/os/freebsd/zfs/zfs_vnops_os.c | 2 +- module/os/freebsd/zfs/zfs_znode.c | 4 +- module/os/linux/zfs/zfs_dir.c | 9 ++ module/os/linux/zfs/zfs_vfsops.c | 12 +- module/os/linux/zfs/zfs_vnops_os.c | 42 +++++- module/os/linux/zfs/zfs_znode.c | 8 +- module/os/linux/zfs/zpl_export.c | 31 ++++ module/os/linux/zfs/zpl_inode.c | 56 ++++++- module/zcommon/zfeature_common.c | 13 ++ module/zcommon/zfs_prop.c | 4 + module/zfs/dmu_recv.c | 17 +++ module/zfs/dmu_send.c | 4 + module/zfs/dsl_dataset.c | 3 +- module/zfs/dsl_dir.c | 3 +- module/zfs/zap.c | 16 +- module/zfs/zap_micro.c | 104 ++++++++++--- module/zfs/zfs_ioctl.c | 38 +++++ tests/runfiles/linux.run | 6 +- tests/zfs-tests/tests/Makefile.am | 6 + .../cli_root/zpool_get/zpool_get.cfg | 1 + .../tests/functional/longname/cleanup.ksh | 34 +++++ .../functional/longname/longname_001_pos.ksh | 139 ++++++++++++++++++ .../functional/longname/longname_002_pos.ksh | 115 +++++++++++++++ .../functional/longname/longname_003_pos.ksh | 113 ++++++++++++++ .../tests/functional/longname/setup.ksh | 35 +++++ .../tests/functional/rsend/rsend.kshlib | 1 + .../functional/rsend/send-c_longname.ksh | 98 ++++++++++++ 36 files changed, 915 insertions(+), 90 deletions(-) create mode 100755 tests/zfs-tests/tests/functional/longname/cleanup.ksh create mode 100755 tests/zfs-tests/tests/functional/longname/longname_001_pos.ksh create mode 100755 tests/zfs-tests/tests/functional/longname/longname_002_pos.ksh create mode 100755 tests/zfs-tests/tests/functional/longname/longname_003_pos.ksh create mode 100755 tests/zfs-tests/tests/functional/longname/setup.ksh create mode 100755 tests/zfs-tests/tests/functional/rsend/send-c_longname.ksh diff --git a/cmd/zdb/zdb.c b/cmd/zdb/zdb.c index f29f3f9393..940d2a8474 100644 --- a/cmd/zdb/zdb.c +++ b/cmd/zdb/zdb.c @@ -1117,7 +1117,7 @@ dump_zap(objset_t *os, uint64_t object, void *data, size_t size) { (void) data, (void) size; zap_cursor_t zc; - zap_attribute_t *attrp = zap_attribute_alloc(); + zap_attribute_t *attrp = zap_attribute_long_alloc(); void *prop; unsigned i; @@ -1365,7 +1365,7 @@ dump_zpldir(objset_t *os, uint64_t object, void *data, size_t size) { (void) data, (void) size; zap_cursor_t zc; - zap_attribute_t *attrp = zap_attribute_alloc(); + zap_attribute_t *attrp = zap_attribute_long_alloc(); const char *typenames[] = { /* 0 */ "not specified", /* 1 */ "FIFO", diff --git a/cmd/zhack.c b/cmd/zhack.c index 77284b8955..f297afb65d 100644 --- a/cmd/zhack.c +++ b/cmd/zhack.c @@ -203,7 +203,7 @@ static void dump_obj(objset_t *os, uint64_t obj, const char *name) { zap_cursor_t zc; - zap_attribute_t *za = zap_attribute_alloc(); + zap_attribute_t *za = zap_attribute_long_alloc(); (void) printf("%s_obj:\n", name); diff --git a/include/os/linux/zfs/sys/zfs_vfsops_os.h b/include/os/linux/zfs/sys/zfs_vfsops_os.h index b4d5db21f5..e742e8dc39 100644 --- a/include/os/linux/zfs/sys/zfs_vfsops_os.h +++ b/include/os/linux/zfs/sys/zfs_vfsops_os.h @@ -118,6 +118,7 @@ struct zfsvfs { boolean_t z_xattr_sa; /* allow xattrs to be stores as SA */ boolean_t z_draining; /* is true when drain is active */ boolean_t z_drain_cancel; /* signal the unlinked drain to stop */ + boolean_t z_longname; /* Dataset supports long names */ uint64_t z_version; /* ZPL version */ uint64_t z_shares_dir; /* hidden shares dir */ dataset_kstats_t z_kstat; /* fs kstats */ diff --git a/include/os/linux/zfs/sys/zfs_vnops_os.h b/include/os/linux/zfs/sys/zfs_vnops_os.h index 830c76e574..7298e9d3a1 100644 --- a/include/os/linux/zfs/sys/zfs_vnops_os.h +++ b/include/os/linux/zfs/sys/zfs_vnops_os.h @@ -44,6 +44,7 @@ extern int zfs_write_simple(znode_t *zp, const void *data, size_t len, loff_t pos, size_t *resid); extern int zfs_lookup(znode_t *dzp, char *nm, znode_t **zpp, int flags, cred_t *cr, int *direntflags, pathname_t *realpnp); +extern int zfs_get_name(znode_t *dzp, char *name, znode_t *zp); extern int zfs_create(znode_t *dzp, char *name, vattr_t *vap, int excl, int mode, znode_t **zpp, cred_t *cr, int flag, vsecattr_t *vsecp, zidmap_t *mnt_ns); diff --git a/include/sys/fs/zfs.h b/include/sys/fs/zfs.h index c7e48d1edc..73c8f60ad4 100644 --- a/include/sys/fs/zfs.h +++ b/include/sys/fs/zfs.h @@ -81,6 +81,7 @@ typedef enum dmu_objset_type { * All of these include the terminating NUL byte. */ #define ZAP_MAXNAMELEN 256 +#define ZAP_MAXNAMELEN_NEW 1024 #define ZAP_MAXVALUELEN (1024 * 8) #define ZAP_OLDMAXVALUELEN 1024 #define ZFS_MAX_DATASET_NAME_LEN 256 @@ -193,6 +194,7 @@ typedef enum { ZFS_PROP_SNAPSHOTS_CHANGED, ZFS_PROP_PREFETCH, ZFS_PROP_VOLTHREADING, + ZFS_PROP_LONGNAME, ZFS_NUM_PROPS } zfs_prop_t; diff --git a/include/sys/zap.h b/include/sys/zap.h index ec37c34471..43e8bbea19 100644 --- a/include/sys/zap.h +++ b/include/sys/zap.h @@ -315,7 +315,7 @@ int zap_count(objset_t *ds, uint64_t zapobj, uint64_t *count); * match must be exact (ie, same as mask=-1ULL). */ int zap_value_search(objset_t *os, uint64_t zapobj, - uint64_t value, uint64_t mask, char *name); + uint64_t value, uint64_t mask, char *name, uint64_t namelen); /* * Transfer all the entries from fromobj into intoobj. Only works on @@ -387,6 +387,7 @@ void zap_fini(void); * Alloc and free zap_attribute_t. */ zap_attribute_t *zap_attribute_alloc(void); +zap_attribute_t *zap_attribute_long_alloc(void); void zap_attribute_free(zap_attribute_t *attrp); /* diff --git a/include/sys/zfs_ioctl.h b/include/sys/zfs_ioctl.h index 525d40759f..d177472016 100644 --- a/include/sys/zfs_ioctl.h +++ b/include/sys/zfs_ioctl.h @@ -124,7 +124,7 @@ typedef enum drr_headertype { * default use of "zfs send" won't encounter the bug mentioned above. */ #define DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS (1 << 27) -/* flag #28 is reserved for a Nutanix feature */ +#define DMU_BACKUP_FEATURE_LONGNAME (1 << 28) /* * flag #29 is the last unused bit. It is reserved to indicate a to-be-designed * extension to the stream format which will accomodate more feature flags. @@ -141,7 +141,7 @@ typedef enum drr_headertype { DMU_BACKUP_FEATURE_COMPRESSED | DMU_BACKUP_FEATURE_LARGE_DNODE | \ DMU_BACKUP_FEATURE_RAW | DMU_BACKUP_FEATURE_HOLDS | \ DMU_BACKUP_FEATURE_REDACTED | DMU_BACKUP_FEATURE_SWITCH_TO_LARGE_BLOCKS | \ - DMU_BACKUP_FEATURE_ZSTD) + DMU_BACKUP_FEATURE_ZSTD | DMU_BACKUP_FEATURE_LONGNAME) /* Are all features in the given flag word currently supported? */ #define DMU_STREAM_SUPPORTED(x) (!((x) & ~DMU_BACKUP_FEATURE_MASK)) diff --git a/include/zfeature_common.h b/include/zfeature_common.h index 2515ba3217..ebe3cf5e1e 100644 --- a/include/zfeature_common.h +++ b/include/zfeature_common.h @@ -82,6 +82,7 @@ typedef enum spa_feature { SPA_FEATURE_AVZ_V2, SPA_FEATURE_REDACTION_LIST_SPILL, SPA_FEATURE_RAIDZ_EXPANSION, + SPA_FEATURE_LONGNAME, SPA_FEATURES } spa_feature_t; diff --git a/lib/libzfs/libzfs.abi b/lib/libzfs/libzfs.abi index 4394c94208..94c42e9fcc 100644 --- a/lib/libzfs/libzfs.abi +++ b/lib/libzfs/libzfs.abi @@ -609,7 +609,7 @@ - + @@ -657,6 +657,10 @@ + + + + @@ -1852,7 +1856,8 @@ - + + @@ -3408,18 +3413,9 @@ - - - - - - - - - - - - + + + @@ -3466,11 +3462,6 @@ - - - - - @@ -3483,10 +3474,6 @@ - - - - @@ -3535,10 +3522,6 @@ - - - - @@ -3551,13 +3534,6 @@ - - - - - - - @@ -4364,6 +4340,10 @@ + + + + @@ -4403,6 +4383,13 @@ + + + + + + + @@ -4933,6 +4920,11 @@ + + + + + @@ -5987,7 +5979,8 @@ - + + @@ -8023,10 +8016,6 @@ - - - - @@ -9072,8 +9061,8 @@ - - + + @@ -9150,7 +9139,7 @@ - + diff --git a/module/os/freebsd/zfs/zfs_vnops_os.c b/module/os/freebsd/zfs/zfs_vnops_os.c index d1c1be02e2..8b47a5749b 100644 --- a/module/os/freebsd/zfs/zfs_vnops_os.c +++ b/module/os/freebsd/zfs/zfs_vnops_os.c @@ -1841,7 +1841,7 @@ zfs_readdir(vnode_t *vp, zfs_uio_t *uio, cred_t *cr, int *eofp, /* * Grab next entry. */ - if ((error = zap_cursor_retrieve(&zc, &zap))) { + if ((error = zap_cursor_retrieve(&zc, zap))) { if ((*eofp = (error == ENOENT)) != 0) break; else diff --git a/module/os/freebsd/zfs/zfs_znode.c b/module/os/freebsd/zfs/zfs_znode.c index afbea9798c..c446da63fe 100644 --- a/module/os/freebsd/zfs/zfs_znode.c +++ b/module/os/freebsd/zfs/zfs_znode.c @@ -1978,7 +1978,7 @@ zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, (void) sprintf(component + 1, ""); } else { error = zap_value_search(osp, pobj, obj, - ZFS_DIRENT_OBJ(-1ULL), component + 1); + ZFS_DIRENT_OBJ(-1ULL), component + 1, MAXNAMELEN); if (error != 0) break; } @@ -2191,7 +2191,7 @@ zfs_znode_parent_and_name(znode_t *zp, znode_t **dzpp, char *buf) return (SET_ERROR(EINVAL)); err = zap_value_search(zfsvfs->z_os, parent, zp->z_id, - ZFS_DIRENT_OBJ(-1ULL), buf); + ZFS_DIRENT_OBJ(-1ULL), buf, MAXNAMELEN); if (err != 0) return (err); err = zfs_zget(zfsvfs, parent, dzpp); diff --git a/module/os/linux/zfs/zfs_dir.c b/module/os/linux/zfs/zfs_dir.c index ad2ca15e29..dc03aab2ef 100644 --- a/module/os/linux/zfs/zfs_dir.c +++ b/module/os/linux/zfs/zfs_dir.c @@ -802,6 +802,7 @@ zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) { znode_t *dzp = dl->dl_dzp; zfsvfs_t *zfsvfs = ZTOZSB(zp); + dsl_dataset_t *ds = dmu_objset_ds(zfsvfs->z_os); uint64_t value; int zp_is_dir = S_ISDIR(ZTOI(zp)->i_mode); sa_bulk_attr_t bulk[5]; @@ -847,6 +848,14 @@ zfs_link_create(zfs_dirlock_t *dl, znode_t *zp, dmu_tx_t *tx, int flag) return (error); } + /* + * If we added a longname activate the SPA_FEATURE_LONGNAME. + */ + if (strlen(dl->dl_name) >= ZAP_MAXNAMELEN) { + ds->ds_feature_activation[SPA_FEATURE_LONGNAME] = + (void *)B_TRUE; + } + SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_PARENT(zfsvfs), NULL, &dzp->z_id, sizeof (dzp->z_id)); SA_ADD_BULK_ATTR(bulk, count, SA_ZPL_FLAGS(zfsvfs), NULL, diff --git a/module/os/linux/zfs/zfs_vfsops.c b/module/os/linux/zfs/zfs_vfsops.c index 2015c20d73..55da2d71e8 100644 --- a/module/os/linux/zfs/zfs_vfsops.c +++ b/module/os/linux/zfs/zfs_vfsops.c @@ -57,6 +57,7 @@ #include #include #include +#include #include #include #include "zfs_comutil.h" @@ -448,6 +449,12 @@ acl_inherit_changed_cb(void *arg, uint64_t newval) ((zfsvfs_t *)arg)->z_acl_inherit = newval; } +static void +longname_changed_cb(void *arg, uint64_t newval) +{ + ((zfsvfs_t *)arg)->z_longname = newval; +} + static int zfs_register_callbacks(vfs_t *vfsp) { @@ -508,6 +515,8 @@ zfs_register_callbacks(vfs_t *vfsp) zfsvfs); error = error ? error : dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_NBMAND), nbmand_changed_cb, zfsvfs); + error = error ? error : dsl_prop_register(ds, + zfs_prop_to_name(ZFS_PROP_LONGNAME), longname_changed_cb, zfsvfs); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); if (error) goto unregister; @@ -1139,7 +1148,8 @@ zfs_statvfs(struct inode *ip, struct kstatfs *statp) statp->f_fsid.val[0] = (uint32_t)fsid; statp->f_fsid.val[1] = (uint32_t)(fsid >> 32); statp->f_type = ZFS_SUPER_MAGIC; - statp->f_namelen = MAXNAMELEN - 1; + statp->f_namelen = + zfsvfs->z_longname ? (ZAP_MAXNAMELEN_NEW - 1) : (MAXNAMELEN - 1); /* * We have all of 40 characters to stuff a string here. diff --git a/module/os/linux/zfs/zfs_vnops_os.c b/module/os/linux/zfs/zfs_vnops_os.c index 259176391d..d072650507 100644 --- a/module/os/linux/zfs/zfs_vnops_os.c +++ b/module/os/linux/zfs/zfs_vnops_os.c @@ -531,6 +531,46 @@ zfs_lookup(znode_t *zdp, char *nm, znode_t **zpp, int flags, cred_t *cr, return (error); } +/* + * Perform a linear search in directory for the name of specific inode. + * Note we don't pass in the buffer size of name because it's hardcoded to + * NAME_MAX+1(256) in Linux. + * + * IN: dzp - znode of directory to search. + * zp - znode of the target + * + * OUT: name - dentry name of the target + * + * RETURN: 0 on success, error code on failure. + */ +int +zfs_get_name(znode_t *dzp, char *name, znode_t *zp) +{ + zfsvfs_t *zfsvfs = ZTOZSB(dzp); + int error = 0; + + if ((error = zfs_enter_verify_zp(zfsvfs, dzp, FTAG)) != 0) + return (error); + + if ((error = zfs_verify_zp(zp)) != 0) { + zfs_exit(zfsvfs, FTAG); + return (error); + } + + /* ctldir should have got their name in zfs_vget */ + if (dzp->z_is_ctldir || zp->z_is_ctldir) { + zfs_exit(zfsvfs, FTAG); + return (ENOENT); + } + + /* buffer len is hardcoded to 256 in Linux kernel */ + error = zap_value_search(zfsvfs->z_os, dzp->z_id, zp->z_id, + ZFS_DIRENT_OBJ(-1ULL), name, ZAP_MAXNAMELEN); + + zfs_exit(zfsvfs, FTAG); + return (error); +} + /* * Attempt to create a new entry in a directory. If the entry * already exists, truncate the file if permissible, else return @@ -1537,7 +1577,7 @@ zfs_readdir(struct inode *ip, zpl_dir_context_t *ctx, cred_t *cr) os = zfsvfs->z_os; offset = ctx->pos; prefetch = zp->z_zn_prefetch; - zap = zap_attribute_alloc(); + zap = zap_attribute_long_alloc(); /* * Initialize the iterator cursor. diff --git a/module/os/linux/zfs/zfs_znode.c b/module/os/linux/zfs/zfs_znode.c index 265153e011..bb672bfa73 100644 --- a/module/os/linux/zfs/zfs_znode.c +++ b/module/os/linux/zfs/zfs_znode.c @@ -2145,6 +2145,7 @@ zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, dmu_buf_t *prevdb = NULL; dmu_buf_t *sa_db = NULL; char *path = buf + len - 1; + char *comp_buf; int error; *path = '\0'; @@ -2160,9 +2161,10 @@ zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, return (error); } + comp_buf = kmem_alloc(ZAP_MAXNAMELEN_NEW + 2, KM_SLEEP); for (;;) { uint64_t pobj = 0; - char component[MAXNAMELEN + 2]; + char *component = comp_buf; size_t complen; int is_xattrdir = 0; @@ -2186,7 +2188,8 @@ zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, strcpy(component + 1, ""); } else { error = zap_value_search(osp, pobj, obj, - ZFS_DIRENT_OBJ(-1ULL), component + 1); + ZFS_DIRENT_OBJ(-1ULL), component + 1, + ZAP_MAXNAMELEN_NEW); if (error != 0) break; } @@ -2217,6 +2220,7 @@ zfs_obj_to_path_impl(objset_t *osp, uint64_t obj, sa_handle_t *hdl, if (error == 0) (void) memmove(buf, path, buf + len - path); + kmem_free(comp_buf, ZAP_MAXNAMELEN_NEW +2); return (error); } diff --git a/module/os/linux/zfs/zpl_export.c b/module/os/linux/zfs/zpl_export.c index aa80b72e2d..e9ca7b1349 100644 --- a/module/os/linux/zfs/zpl_export.c +++ b/module/os/linux/zfs/zpl_export.c @@ -24,6 +24,7 @@ */ +#include #include #include #include @@ -109,6 +110,35 @@ zpl_fh_to_dentry(struct super_block *sb, struct fid *fh, return (d_obtain_alias(ip)); } +/* + * In case the filesystem contains name longer than 255, we need to override + * the default get_name so we don't get buffer overflow. Unfortunately, since + * the buffer size is hardcoded in Linux, we will get ESTALE error in this + * case. + */ +static int +zpl_get_name(struct dentry *parent, char *name, struct dentry *child) +{ + cred_t *cr = CRED(); + fstrans_cookie_t cookie; + struct inode *dir = parent->d_inode; + struct inode *ip = child->d_inode; + int error; + + if (!dir || !S_ISDIR(dir->i_mode)) + return (-ENOTDIR); + + crhold(cr); + cookie = spl_fstrans_mark(); + spl_inode_lock_shared(dir); + error = -zfs_get_name(ITOZ(dir), name, ITOZ(ip)); + spl_inode_unlock_shared(dir); + spl_fstrans_unmark(cookie); + crfree(cr); + + return (error); +} + static struct dentry * zpl_get_parent(struct dentry *child) { @@ -153,6 +183,7 @@ zpl_commit_metadata(struct inode *inode) const struct export_operations zpl_export_operations = { .encode_fh = zpl_encode_fh, .fh_to_dentry = zpl_fh_to_dentry, + .get_name = zpl_get_name, .get_parent = zpl_get_parent, .commit_metadata = zpl_commit_metadata, }; diff --git a/module/os/linux/zfs/zpl_inode.c b/module/os/linux/zfs/zpl_inode.c index ad1753f7a0..6d737d9fc6 100644 --- a/module/os/linux/zfs/zpl_inode.c +++ b/module/os/linux/zfs/zpl_inode.c @@ -46,9 +46,29 @@ zpl_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) pathname_t pn; int zfs_flags = 0; zfsvfs_t *zfsvfs = dentry->d_sb->s_fs_info; + dsl_dataset_t *ds = dmu_objset_ds(zfsvfs->z_os); + size_t dlen = dlen(dentry); - if (dlen(dentry) >= ZAP_MAXNAMELEN) + /* + * If z_longname is disabled, disallow create or rename of names + * longer than ZAP_MAXNAMELEN. + * + * This is needed in cases where longname was enabled first and some + * files/dirs with names > ZAP_MAXNAMELEN were created. And later + * longname was disabled. In such a case allow access to existing + * longnames. But disallow creation newer longnamed entities. + */ + if (!zfsvfs->z_longname && (dlen >= ZAP_MAXNAMELEN)) { + /* + * If this is for create or rename fail it. + */ + if (!dsl_dataset_feature_is_active(ds, SPA_FEATURE_LONGNAME) || + (flags & (LOOKUP_CREATE | LOOKUP_RENAME_TARGET))) + return (ERR_PTR(-ENAMETOOLONG)); + } + if (dlen >= ZAP_MAXNAMELEN_NEW) { return (ERR_PTR(-ENAMETOOLONG)); + } crhold(cr); cookie = spl_fstrans_mark(); @@ -131,6 +151,16 @@ zpl_vap_init(vattr_t *vap, struct inode *dir, umode_t mode, cred_t *cr, } } +static inline bool +is_nametoolong(struct dentry *dentry) +{ + zfsvfs_t *zfsvfs = dentry->d_sb->s_fs_info; + size_t dlen = dlen(dentry); + + return ((!zfsvfs->z_longname && dlen >= ZAP_MAXNAMELEN) || + dlen >= ZAP_MAXNAMELEN_NEW); +} + static int #ifdef HAVE_IOPS_CREATE_USERNS zpl_create(struct user_namespace *user_ns, struct inode *dir, @@ -151,6 +181,10 @@ zpl_create(struct inode *dir, struct dentry *dentry, umode_t mode, bool flag) zidmap_t *user_ns = kcred->user_ns; #endif + if (is_nametoolong(dentry)) { + return (-ENAMETOOLONG); + } + crhold(cr); vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP); zpl_vap_init(vap, dir, mode, cr, user_ns); @@ -201,6 +235,10 @@ zpl_mknod(struct inode *dir, struct dentry *dentry, umode_t mode, zidmap_t *user_ns = kcred->user_ns; #endif + if (is_nametoolong(dentry)) { + return (-ENAMETOOLONG); + } + /* * We currently expect Linux to supply rdev=0 for all sockets * and fifos, but we want to know if this behavior ever changes. @@ -355,6 +393,10 @@ zpl_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) zidmap_t *user_ns = kcred->user_ns; #endif + if (is_nametoolong(dentry)) { + return (-ENAMETOOLONG); + } + crhold(cr); vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP); zpl_vap_init(vap, dir, mode | S_IFDIR, cr, user_ns); @@ -570,6 +612,10 @@ zpl_rename2(struct inode *sdip, struct dentry *sdentry, zidmap_t *user_ns = kcred->user_ns; #endif + if (is_nametoolong(tdentry)) { + return (-ENAMETOOLONG); + } + crhold(cr); if (rflags & RENAME_WHITEOUT) { wo_vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP); @@ -621,6 +667,10 @@ zpl_symlink(struct inode *dir, struct dentry *dentry, const char *name) zidmap_t *user_ns = kcred->user_ns; #endif + if (is_nametoolong(dentry)) { + return (-ENAMETOOLONG); + } + crhold(cr); vap = kmem_zalloc(sizeof (vattr_t), KM_SLEEP); zpl_vap_init(vap, dir, S_IFLNK | S_IRWXUGO, cr, user_ns); @@ -773,6 +823,10 @@ zpl_link(struct dentry *old_dentry, struct inode *dir, struct dentry *dentry) int error; fstrans_cookie_t cookie; + if (is_nametoolong(dentry)) { + return (-ENAMETOOLONG); + } + if (ip->i_nlink >= ZFS_LINK_MAX) return (-EMLINK); diff --git a/module/zcommon/zfeature_common.c b/module/zcommon/zfeature_common.c index 309d9bf14c..d9c95b21e7 100644 --- a/module/zcommon/zfeature_common.c +++ b/module/zcommon/zfeature_common.c @@ -754,6 +754,19 @@ zpool_feature_init(void) "Support for raidz expansion", ZFEATURE_FLAG_MOS, ZFEATURE_TYPE_BOOLEAN, NULL, sfeatures); + { + static const spa_feature_t longname_deps[] = { + SPA_FEATURE_EXTENSIBLE_DATASET, + SPA_FEATURE_ENABLED_TXG, + SPA_FEATURE_NONE + }; + zfeature_register(SPA_FEATURE_LONGNAME, + "org.zfsonlinux:longname", "longname", + "support filename upto 1024 bytes", + ZFEATURE_FLAG_READONLY_COMPAT | ZFEATURE_FLAG_PER_DATASET, + ZFEATURE_TYPE_BOOLEAN, longname_deps, sfeatures); + } + zfs_mod_list_supported_free(sfeatures); } diff --git a/module/zcommon/zfs_prop.c b/module/zcommon/zfs_prop.c index 764993b45e..886414de61 100644 --- a/module/zcommon/zfs_prop.c +++ b/module/zcommon/zfs_prop.c @@ -760,6 +760,10 @@ zfs_prop_init(void) ZFS_TYPE_VOLUME, "", "SNAPSHOTS_CHANGED", B_FALSE, B_TRUE, B_TRUE, NULL, sfeatures); + zprop_register_index(ZFS_PROP_LONGNAME, "longname", 0, PROP_INHERIT, + ZFS_TYPE_FILESYSTEM, "on | off", "LONGNAME", boolean_table, + sfeatures); + zfs_mod_list_supported_free(sfeatures); } diff --git a/module/zfs/dmu_recv.c b/module/zfs/dmu_recv.c index 0119191d79..5ed1c952e0 100644 --- a/module/zfs/dmu_recv.c +++ b/module/zfs/dmu_recv.c @@ -602,6 +602,13 @@ recv_begin_check_feature_flags_impl(uint64_t featureflags, spa_t *spa) !spa_feature_is_enabled(spa, SPA_FEATURE_REDACTED_DATASETS)) return (SET_ERROR(ENOTSUP)); + /* + * If the LONGNAME is not enabled on the target, fail that request. + */ + if ((featureflags & DMU_BACKUP_FEATURE_LONGNAME) && + !spa_feature_is_enabled(spa, SPA_FEATURE_LONGNAME)) + return (SET_ERROR(ENOTSUP)); + return (0); } @@ -990,6 +997,16 @@ dmu_recv_begin_sync(void *arg, dmu_tx_t *tx) dmu_buf_will_dirty(newds->ds_dbuf, tx); dsl_dataset_phys(newds)->ds_flags |= DS_FLAG_INCONSISTENT; + /* + * Activate longname feature if received + */ + if (featureflags & DMU_BACKUP_FEATURE_LONGNAME && + !dsl_dataset_feature_is_active(newds, SPA_FEATURE_LONGNAME)) { + dsl_dataset_activate_feature(newds->ds_object, + SPA_FEATURE_LONGNAME, (void *)B_TRUE, tx); + newds->ds_feature[SPA_FEATURE_LONGNAME] = (void *)B_TRUE; + } + /* * If we actually created a non-clone, we need to create the objset * in our new dataset. If this is a raw send we postpone this until diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c index cb2b62fed3..c7d3a5cb6e 100644 --- a/module/zfs/dmu_send.c +++ b/module/zfs/dmu_send.c @@ -2011,6 +2011,10 @@ setup_featureflags(struct dmu_send_params *dspp, objset_t *os, if (dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_LARGE_DNODE)) { *featureflags |= DMU_BACKUP_FEATURE_LARGE_DNODE; } + + if (dsl_dataset_feature_is_active(to_ds, SPA_FEATURE_LONGNAME)) { + *featureflags |= DMU_BACKUP_FEATURE_LONGNAME; + } return (0); } diff --git a/module/zfs/dsl_dataset.c b/module/zfs/dsl_dataset.c index 237cf8a478..4302617982 100644 --- a/module/zfs/dsl_dataset.c +++ b/module/zfs/dsl_dataset.c @@ -494,7 +494,8 @@ dsl_dataset_get_snapname(dsl_dataset_t *ds) return (err); headphys = headdbuf->db_data; err = zap_value_search(dp->dp_meta_objset, - headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname); + headphys->ds_snapnames_zapobj, ds->ds_object, 0, ds->ds_snapname, + sizeof (ds->ds_snapname)); if (err != 0 && zfs_recover == B_TRUE) { err = 0; (void) snprintf(ds->ds_snapname, sizeof (ds->ds_snapname), diff --git a/module/zfs/dsl_dir.c b/module/zfs/dsl_dir.c index 5aeff86b94..1b60fa620b 100644 --- a/module/zfs/dsl_dir.c +++ b/module/zfs/dsl_dir.c @@ -239,7 +239,8 @@ dsl_dir_hold_obj(dsl_pool_t *dp, uint64_t ddobj, err = zap_value_search(dp->dp_meta_objset, dsl_dir_phys(dd->dd_parent)-> dd_child_dir_zapobj, - ddobj, 0, dd->dd_myname); + ddobj, 0, dd->dd_myname, + sizeof (dd->dd_myname)); } if (err != 0) goto errout; diff --git a/module/zfs/zap.c b/module/zfs/zap.c index 603c76a2cd..159c65a0fc 100644 --- a/module/zfs/zap.c +++ b/module/zfs/zap.c @@ -832,7 +832,8 @@ zap_put_leaf_maybe_grow_ptrtbl(zap_name_t *zn, zap_leaf_t *l, static int fzap_checkname(zap_name_t *zn) { - if (zn->zn_key_orig_numints * zn->zn_key_intlen > ZAP_MAXNAMELEN) + uint32_t maxnamelen = zn->zn_normbuf_len; + if (zn->zn_key_orig_numints * zn->zn_key_intlen > maxnamelen) return (SET_ERROR(ENAMETOOLONG)); return (0); } @@ -1102,7 +1103,7 @@ zap_create_link_dnsize(objset_t *os, dmu_object_type_t ot, uint64_t parent_obj, int zap_value_search(objset_t *os, uint64_t zapobj, uint64_t value, uint64_t mask, - char *name) + char *name, uint64_t namelen) { zap_cursor_t zc; int err; @@ -1110,12 +1111,13 @@ zap_value_search(objset_t *os, uint64_t zapobj, uint64_t value, uint64_t mask, if (mask == 0) mask = -1ULL; - zap_attribute_t *za = zap_attribute_alloc(); + zap_attribute_t *za = zap_attribute_long_alloc(); for (zap_cursor_init(&zc, os, zapobj); (err = zap_cursor_retrieve(&zc, za)) == 0; zap_cursor_advance(&zc)) { if ((za->za_first_integer & mask) == (value & mask)) { - (void) strlcpy(name, za->za_name, MAXNAMELEN); + if (strlcpy(name, za->za_name, namelen) >= namelen) + err = SET_ERROR(ENAMETOOLONG); break; } } @@ -1130,7 +1132,7 @@ zap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx) zap_cursor_t zc; int err = 0; - zap_attribute_t *za = zap_attribute_alloc(); + zap_attribute_t *za = zap_attribute_long_alloc(); for (zap_cursor_init(&zc, os, fromobj); zap_cursor_retrieve(&zc, za) == 0; (void) zap_cursor_advance(&zc)) { @@ -1155,7 +1157,7 @@ zap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj, zap_cursor_t zc; int err = 0; - zap_attribute_t *za = zap_attribute_alloc(); + zap_attribute_t *za = zap_attribute_long_alloc(); for (zap_cursor_init(&zc, os, fromobj); zap_cursor_retrieve(&zc, za) == 0; (void) zap_cursor_advance(&zc)) { @@ -1180,7 +1182,7 @@ zap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj, zap_cursor_t zc; int err = 0; - zap_attribute_t *za = zap_attribute_alloc(); + zap_attribute_t *za = zap_attribute_long_alloc(); for (zap_cursor_init(&zc, os, fromobj); zap_cursor_retrieve(&zc, za) == 0; (void) zap_cursor_advance(&zc)) { diff --git a/module/zfs/zap_micro.c b/module/zfs/zap_micro.c index e52680dd40..a428a040a4 100644 --- a/module/zfs/zap_micro.c +++ b/module/zfs/zap_micro.c @@ -131,12 +131,12 @@ zap_hash(zap_name_t *zn) } static int -zap_normalize(zap_t *zap, const char *name, char *namenorm, int normflags) +zap_normalize(zap_t *zap, const char *name, char *namenorm, int normflags, + size_t outlen) { ASSERT(!(zap_getflags(zap) & ZAP_FLAG_UINT64_KEY)); size_t inlen = strlen(name) + 1; - size_t outlen = ZAP_MAXNAMELEN; int err = 0; (void) u8_textprep_str((char *)name, &inlen, namenorm, &outlen, @@ -149,23 +149,39 @@ zap_normalize(zap_t *zap, const char *name, char *namenorm, int normflags) boolean_t zap_match(zap_name_t *zn, const char *matchname) { + boolean_t res = B_FALSE; ASSERT(!(zap_getflags(zn->zn_zap) & ZAP_FLAG_UINT64_KEY)); if (zn->zn_matchtype & MT_NORMALIZE) { - char norm[ZAP_MAXNAMELEN]; + size_t namelen = zn->zn_normbuf_len; + char normbuf[ZAP_MAXNAMELEN]; + char *norm = normbuf; + + /* + * Cannot allocate this on-stack as it exceed the stack-limit of + * 1024. + */ + if (namelen > ZAP_MAXNAMELEN) + norm = kmem_alloc(namelen, KM_SLEEP); if (zap_normalize(zn->zn_zap, matchname, norm, - zn->zn_normflags) != 0) - return (B_FALSE); - - return (strcmp(zn->zn_key_norm, norm) == 0); + zn->zn_normflags, namelen) != 0) { + res = B_FALSE; + } else { + res = (strcmp(zn->zn_key_norm, norm) == 0); + } + if (norm != normbuf) + kmem_free(norm, namelen); } else { - return (strcmp(zn->zn_key_orig, matchname) == 0); + res = (strcmp(zn->zn_key_orig, matchname) == 0); } + return (res); } static kmem_cache_t *zap_name_cache; static kmem_cache_t *zap_attr_cache; +static kmem_cache_t *zap_name_long_cache; +static kmem_cache_t *zap_attr_long_cache; void zap_init(void) @@ -177,6 +193,14 @@ zap_init(void) zap_attr_cache = kmem_cache_create("zap_attr_cache", sizeof (zap_attribute_t) + ZAP_MAXNAMELEN, 0, NULL, NULL, NULL, NULL, NULL, 0); + + zap_name_long_cache = kmem_cache_create("zap_name_long", + sizeof (zap_name_t) + ZAP_MAXNAMELEN_NEW, 0, NULL, NULL, + NULL, NULL, NULL, 0); + + zap_attr_long_cache = kmem_cache_create("zap_attr_long_cache", + sizeof (zap_attribute_t) + ZAP_MAXNAMELEN_NEW, 0, NULL, + NULL, NULL, NULL, NULL, 0); } void @@ -184,33 +208,47 @@ zap_fini(void) { kmem_cache_destroy(zap_name_cache); kmem_cache_destroy(zap_attr_cache); + kmem_cache_destroy(zap_name_long_cache); + kmem_cache_destroy(zap_attr_long_cache); } static zap_name_t * -zap_name_alloc(zap_t *zap) +zap_name_alloc(zap_t *zap, boolean_t longname) { - zap_name_t *zn = kmem_cache_alloc(zap_name_cache, KM_SLEEP); + kmem_cache_t *cache = longname ? zap_name_long_cache : zap_name_cache; + zap_name_t *zn = kmem_cache_alloc(cache, KM_SLEEP); + zn->zn_zap = zap; + zn->zn_normbuf_len = longname ? ZAP_MAXNAMELEN_NEW : ZAP_MAXNAMELEN; return (zn); } void zap_name_free(zap_name_t *zn) { - kmem_cache_free(zap_name_cache, zn); + if (zn->zn_normbuf_len == ZAP_MAXNAMELEN) { + kmem_cache_free(zap_name_cache, zn); + } else { + ASSERT3U(zn->zn_normbuf_len, ==, ZAP_MAXNAMELEN_NEW); + kmem_cache_free(zap_name_long_cache, zn); + } } static int zap_name_init_str(zap_name_t *zn, const char *key, matchtype_t mt) { zap_t *zap = zn->zn_zap; + size_t key_len = strlen(key) + 1; + + /* Make sure zn is allocated for longname if key is long */ + IMPLY(key_len > ZAP_MAXNAMELEN, + zn->zn_normbuf_len == ZAP_MAXNAMELEN_NEW); zn->zn_key_intlen = sizeof (*key); zn->zn_key_orig = key; - zn->zn_key_orig_numints = strlen(zn->zn_key_orig) + 1; + zn->zn_key_orig_numints = key_len; zn->zn_matchtype = mt; zn->zn_normflags = zap->zap_normflags; - zn->zn_normbuf_len = ZAP_MAXNAMELEN; /* * If we're dealing with a case sensitive lookup on a mixed or @@ -226,7 +264,7 @@ zap_name_init_str(zap_name_t *zn, const char *key, matchtype_t mt) * what the hash is computed from. */ if (zap_normalize(zap, key, zn->zn_normbuf, - zap->zap_normflags) != 0) + zap->zap_normflags, zn->zn_normbuf_len) != 0) return (SET_ERROR(ENOTSUP)); zn->zn_key_norm = zn->zn_normbuf; zn->zn_key_norm_numints = strlen(zn->zn_key_norm) + 1; @@ -245,7 +283,7 @@ zap_name_init_str(zap_name_t *zn, const char *key, matchtype_t mt) * what the matching is based on. (Not the hash!) */ if (zap_normalize(zap, key, zn->zn_normbuf, - zn->zn_normflags) != 0) + zn->zn_normflags, zn->zn_normbuf_len) != 0) return (SET_ERROR(ENOTSUP)); zn->zn_key_norm_numints = strlen(zn->zn_key_norm) + 1; } @@ -256,7 +294,8 @@ zap_name_init_str(zap_name_t *zn, const char *key, matchtype_t mt) zap_name_t * zap_name_alloc_str(zap_t *zap, const char *key, matchtype_t mt) { - zap_name_t *zn = zap_name_alloc(zap); + size_t key_len = strlen(key) + 1; + zap_name_t *zn = zap_name_alloc(zap, (key_len > ZAP_MAXNAMELEN)); if (zap_name_init_str(zn, key, mt) != 0) { zap_name_free(zn); return (NULL); @@ -491,7 +530,7 @@ mzap_open(dmu_buf_t *db) zfs_btree_create_custom(&zap->zap_m.zap_tree, mze_compare, mze_find_in_buf, sizeof (mzap_ent_t), 512); - zap_name_t *zn = zap_name_alloc(zap); + zap_name_t *zn = zap_name_alloc(zap, B_FALSE); for (uint16_t i = 0; i < zap->zap_m.zap_num_chunks; i++) { mzap_ent_phys_t *mze = &zap_m_phys(zap)->mz_chunk[i]; @@ -698,7 +737,7 @@ mzap_upgrade(zap_t **zapp, const void *tag, dmu_tx_t *tx, zap_flags_t flags) fzap_upgrade(zap, tx, flags); - zap_name_t *zn = zap_name_alloc(zap); + zap_name_t *zn = zap_name_alloc(zap, B_FALSE); for (int i = 0; i < nchunks; i++) { mzap_ent_phys_t *mze = &mzp->mz_chunk[i]; if (mze->mze_name[0] == 0) @@ -1625,21 +1664,38 @@ zap_remove_uint64_by_dnode(dnode_t *dn, const uint64_t *key, int key_numints, } +static zap_attribute_t * +zap_attribute_alloc_impl(boolean_t longname) +{ + zap_attribute_t *za; + + za = kmem_cache_alloc((longname)? zap_attr_long_cache : zap_attr_cache, + KM_SLEEP); + za->za_name_len = (longname)? ZAP_MAXNAMELEN_NEW : ZAP_MAXNAMELEN; + return (za); +} + zap_attribute_t * zap_attribute_alloc(void) { - uint32_t len = ZAP_MAXNAMELEN; - zap_attribute_t *za; + return (zap_attribute_alloc_impl(B_FALSE)); +} - za = kmem_cache_alloc(zap_attr_cache, KM_SLEEP); - za->za_name_len = len; - return (za); +zap_attribute_t * +zap_attribute_long_alloc(void) +{ + return (zap_attribute_alloc_impl(B_TRUE)); } void zap_attribute_free(zap_attribute_t *za) { - kmem_cache_free(zap_attr_cache, za); + if (za->za_name_len == ZAP_MAXNAMELEN) { + kmem_cache_free(zap_attr_cache, za); + } else { + ASSERT3U(za->za_name_len, ==, ZAP_MAXNAMELEN_NEW); + kmem_cache_free(zap_attr_long_cache, za); + } } /* diff --git a/module/zfs/zfs_ioctl.c b/module/zfs/zfs_ioctl.c index 897335dd4e..8e16a01993 100644 --- a/module/zfs/zfs_ioctl.c +++ b/module/zfs/zfs_ioctl.c @@ -2567,6 +2567,44 @@ zfs_prop_set_special(const char *dsname, zprop_source_t source, } break; } + case ZFS_PROP_LONGNAME: + { + zfsvfs_t *zfsvfs; + dsl_dataset_t *ds; + + /* + * Ignore the checks if the property is being applied as part of + * 'zfs receive'. Because, we already check if the local pool + * has SPA_FEATURE_LONGNAME enabled in dmu_recv_begin_check(). + */ + if (source == ZPROP_SRC_RECEIVED) { + cmn_err(CE_NOTE, "Skipping ZFS_PROP_LONGNAME checks " + "for dsname=%s\n", dsname); + err = -1; + break; + } + + if ((err = zfsvfs_hold(dsname, FTAG, &zfsvfs, B_FALSE)) != 0) { + cmn_err(CE_WARN, "%s:%d Failed to hold for dsname=%s " + "err=%d\n", __FILE__, __LINE__, dsname, err); + break; + } + + ds = dmu_objset_ds(zfsvfs->z_os); + + if (!spa_feature_is_enabled(zfsvfs->z_os->os_spa, + SPA_FEATURE_LONGNAME)) { + err = ENOTSUP; + } else { + /* + * Set err to -1 to force the zfs_set_prop_nvlist code + * down the default path to set the value in the nvlist. + */ + err = -1; + } + zfsvfs_rele(zfsvfs, FTAG); + break; + } default: err = -1; } diff --git a/tests/runfiles/linux.run b/tests/runfiles/linux.run index bd6cc56f35..ee03875941 100644 --- a/tests/runfiles/linux.run +++ b/tests/runfiles/linux.run @@ -138,6 +138,10 @@ pre = post = tags = ['functional', 'largest_pool'] +[tests/functional/longname:Linux] +tests = ['longname_001_pos', 'longname_002_pos', 'longname_003_pos'] +tags = ['functional', 'longname'] + [tests/functional/mmap:Linux] tests = ['mmap_libaio_001_pos', 'mmap_sync_001_pos'] tags = ['functional', 'mmap'] @@ -182,7 +186,7 @@ tests = ['renameat2_noreplace', 'renameat2_exchange', 'renameat2_whiteout'] tags = ['functional', 'renameat2'] [tests/functional/rsend:Linux] -tests = ['send_realloc_dnode_size', 'send_encrypted_files'] +tests = ['send_realloc_dnode_size', 'send_encrypted_files', 'send-c_longname'] tags = ['functional', 'rsend'] [tests/functional/simd:Linux] diff --git a/tests/zfs-tests/tests/Makefile.am b/tests/zfs-tests/tests/Makefile.am index 73a6f00b50..f4721929af 100644 --- a/tests/zfs-tests/tests/Makefile.am +++ b/tests/zfs-tests/tests/Makefile.am @@ -1567,6 +1567,11 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \ functional/link_count/link_count_001.ksh \ functional/link_count/link_count_root_inode.ksh \ functional/link_count/setup.ksh \ + functional/longname/cleanup.ksh \ + functional/longname/longname_001_pos.ksh \ + functional/longname/longname_002_pos.ksh \ + functional/longname/longname_003_pos.ksh \ + functional/longname/setup.ksh \ functional/log_spacemap/log_spacemap_import_logs.ksh \ functional/migration/cleanup.ksh \ functional/migration/migration_001_pos.ksh \ @@ -1900,6 +1905,7 @@ nobase_dist_datadir_zfs_tests_tests_SCRIPTS += \ functional/rsend/rsend_031_pos.ksh \ functional/rsend/send-c_embedded_blocks.ksh \ functional/rsend/send-c_incremental.ksh \ + functional/rsend/send-c_longname.ksh \ functional/rsend/send-c_lz4_disabled.ksh \ functional/rsend/send-c_mixed_compression.ksh \ functional/rsend/send-c_props.ksh \ diff --git a/tests/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg b/tests/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg index e8a94ce209..92d9027960 100644 --- a/tests/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg +++ b/tests/zfs-tests/tests/functional/cli_root/zpool_get/zpool_get.cfg @@ -109,5 +109,6 @@ if is_linux || is_freebsd; then "feature@block_cloning" "feature@vdev_zaps_v2" "feature@raidz_expansion" + "feature@longname" ) fi diff --git a/tests/zfs-tests/tests/functional/longname/cleanup.ksh b/tests/zfs-tests/tests/functional/longname/cleanup.ksh new file mode 100755 index 0000000000..aac8062ec4 --- /dev/null +++ b/tests/zfs-tests/tests/functional/longname/cleanup.ksh @@ -0,0 +1,34 @@ +#!/bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or http://www.opensolaris.org/os/licensing. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright 2007 Sun Microsystems, Inc. All rights reserved. +# Use is subject to license terms. +# + +# +# Copyright (c) 2021 by Nutanix. All rights reserved. +# + +. $STF_SUITE/include/libtest.shlib + +default_cleanup diff --git a/tests/zfs-tests/tests/functional/longname/longname_001_pos.ksh b/tests/zfs-tests/tests/functional/longname/longname_001_pos.ksh new file mode 100755 index 0000000000..e7ecd2351c --- /dev/null +++ b/tests/zfs-tests/tests/functional/longname/longname_001_pos.ksh @@ -0,0 +1,139 @@ +#! /bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or http://www.opensolaris.org/os/licensing. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright (c) 2021 by Nutanix. All rights reserved. +# + +. $STF_SUITE/include/libtest.shlib + +# +# DESCRIPTION: +# Verify the support for long filenames. +# +# STRATEGY: +# 0. On a fresh dataset ensure property "longname" is enabled by default. +# 1. Disable the longname. +# 2. Try to create a filename whose length is > 256 bytes. This should fail. +# 3. Enable "longname" property on the dataset. +# 4. Try to create files and dirs whose names are > 256 bytes. +# 5. Ensure that "ls" is able to list the file. +# 6. Ensure stat(1) is able to stat the file/directory. +# 7. Try to rename the files and directories +# 8. Try to delete the files and directories + +verify_runnable "global" + +WORKDIR=$TESTDIR/workdir +MOVEDIR=$TESTDIR/movedir/level2/level3/level4/level5/level6 + +function cleanup +{ + log_must rm -rf $WORKDIR + log_must rm -rf $TESTDIR/movedir +} + +LONGNAME=$(printf 'a%.0s' {1..512}) +LONGFNAME="file-$LONGNAME" +LONGDNAME="dir-$LONGNAME" +LONGPNAME="mypipe-$LONGNAME" +LONGCNAME="char_dev-$LONGNAME" +LONGLNAME="link-$LONGNAME" +LONGNAME_255=$(printf 'a%.0s' {1..255}) +LONGNAME_1023=$(printf 'a%.0s' {1..1023}) + + +log_assert "Check longname support for directories/files" + +log_onexit cleanup + +log_must mkdir $WORKDIR +log_must mkdir -p $MOVEDIR + +# Disable longname support +log_must zfs set longname=off $TESTPOOL/$TESTFS + +#Ensure a file of length 255bytes can be created +log_must touch $WORKDIR/$LONGNAME_255 + +#Where as file of length 256bytes should fail +long_mustnot touch $WORKDIR/${LONGNAME_255}b + +# Try to create a file with long name with property "longname=off" +log_mustnot touch $WORKDIR/$LONGFNAME +log_mustnot mkdir $WORKDIR/$LONGDNAME + +# Enable longname support +log_must zfs set longname=on $TESTPOOL/$TESTFS + +# Retry the longname creates and that should succeed +log_must mkdir $WORKDIR/$LONGDNAME +log_must touch $WORKDIR/$LONGFNAME + +# Should be able to create a file with name of 1023 chars +log_must touch $WORKDIR/$LONGNAME_1023 + +# And names longer that 1023 should fail +log_mustnot touch $WORKDIR/${LONGNAME_1023}b + +# Ensure the longnamed dir/file can be listed. +name=$(ls $WORKDIR/$LONGFNAME) +if [[ "${name}" != "$WORKDIR/$LONGFNAME" ]]; then + log_fail "Unable to list: $WORKDIR/$LONGFNAME ret:$name" +fi + +name=$(ls -d $WORKDIR/$LONGDNAME) +if [[ "${name}" != "$WORKDIR/$LONGDNAME" ]]; then + log_fail "Unable to list: $WORKDIR/$LONGDNAME ret:$name" +fi + +# Ensure stat works +log_must stat $WORKDIR/$LONGFNAME +log_must stat $WORKDIR/$LONGDNAME + +# Ensure softlinks can be created from a longname to +# another longnamed file. +log_must ln -s $WORKDIR/$LONGFNAME $WORKDIR/$LONGLNAME + +# Ensure a longnamed pipe and character device file +# can be created +log_must mknod $WORKDIR/$LONGPNAME p +log_must mknod $WORKDIR/$LONGCNAME c 92 1 + +# Fetch the inode numbers of the objects created. +INODE_NO_FILE=$(stat -c %i $WORKDIR/$LONGFNAME) +INODE_NO_DIR=$(stat -c %i $WORKDIR/$LONGDNAME) +INODE_NO_LINK=$(stat -c %i $WORKDIR/$LONGLNAME) +INODE_NO_PIPE=$(stat -c %i $WORKDIR/$LONGPNAME) +INODE_NO_CHAR=$(stat -c %i $WORKDIR/$LONGCNAME) + +# Ensure we can rename the longname file +log_must mv $WORKDIR/$LONGFNAME $WORKDIR/file2 + +# Delete the long named dir/file +log_must rmdir $WORKDIR/$LONGDNAME +log_must rm $WORKDIR/file2 +log_must rm $WORKDIR/$LONGPNAME +log_must rm $WORKDIR/$LONGCNAME +log_must rm $WORKDIR/$LONGLNAME + +log_pass diff --git a/tests/zfs-tests/tests/functional/longname/longname_002_pos.ksh b/tests/zfs-tests/tests/functional/longname/longname_002_pos.ksh new file mode 100755 index 0000000000..dd2acabb3e --- /dev/null +++ b/tests/zfs-tests/tests/functional/longname/longname_002_pos.ksh @@ -0,0 +1,115 @@ +#! /bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or http://www.opensolaris.org/os/licensing. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright (c) 2021 by Nutanix. All rights reserved. +# + +. $STF_SUITE/include/libtest.shlib + +# +# DESCRIPTION: +# Check if longname feature is disabled by default and can be enabled. +# +# STRATEGY: +# 1. Create a zpool with longname feature disabled +# 2. Attempt to enable 'longname' property should fail. +# 3. Attempt to create a longnamed (>255) file should fail. +# 4. Enable the feature@longname +# 5. Enable 'longname' property on the dataset. +# 6. Should be able to create long named files and directories. +# 7. Should be able to disable longname property. +# 8. This should disallow creating new longnamed file/dirs. But, should be +# able to access existing longnamed files/dirs. +verify_runnable "global" + +function cleanup +{ + log_must rm -rf $WORKDIR + poolexists $TESTPOOL && zpool destroy $TESTPOOL +} + +log_assert "Check feature@longname and 'longname' dataset propery work correctly" + +log_onexit cleanup + +log_must zpool destroy $TESTPOOL + +log_must zpool create -o feature@longname=disabled $TESTPOOL $DISKS + +log_must zfs create $TESTPOOL/$TESTFS2 + +log_must zfs set mountpoint=$TESTDIR2 $TESTPOOL/$TESTFS2 + +log_mustnot zfs set longname=on $TESTPOOL/$TESTFS2 + +LONGNAME=$(printf 'a%.0s' {1..512}) +LONGFNAME="file-$LONGNAME" +LONGDNAME="dir-$LONGNAME" +SHORTDNAME="dir-short" +SHORTFNAME="file-short" +WORKDIR=$TESTDIR2/workdir + +log_must mkdir $WORKDIR +log_mustnot touch $WORKDIR/$LONGFNAME +log_mustnot mkdir $WORKDIR/$LONGDNAME + +log_must zpool set feature@longname=enabled $TESTPOOL +log_must zfs set longname=on $TESTPOOL/$TESTFS2 + +log_must mkdir $WORKDIR/$LONGDNAME +log_must touch $WORKDIR/$LONGFNAME + +# Ensure the above changes are synced out. +log_must zpool sync $TESTPOOL + +# Ensure that the feature is activated once longnamed files are created. +state=$(zpool get feature@longname -H -o value $TESTPOOL) +log_note "feature@longname on pool: $TESTPOOL : $state" + +if [[ "$state" != "active" ]]; then + log_fail "feature@longname has state $state (expected active)" +fi + +# Set longname=off. +log_must zfs set longname=off $TESTPOOL/$TESTFS2 + +# Ensure no new file/directory with longnames can be created or can be renamed +# to. +log_mustnot mkdir $WORKDIR/${LONGDNAME}.1 +log_mustnot touch $WORKDIR/${LONGFNAME}.1 +log_must mkdir $WORKDIR/$SHORTDNAME +log_mustnot mv $WORKDIR/$SHORTDNAME $WORKDIR/${LONGDNAME}.1 +log_must touch $WORKDIR/$SHORTFNAME +log_mustnot mv $WORKDIR/$SHORTFNAME $WORKDIR/${LONGFNAME}.1 + +#Cleanup shortnames +log_must rmdir $WORKDIR/$SHORTDNAME +log_must rm $WORKDIR/$SHORTFNAME + +# But, should be able to stat and rename existing files +log_must stat $WORKDIR/$LONGDNAME +log_must stat $WORKDIR/$LONGFNAME +log_must mv $WORKDIR/$LONGDNAME $WORKDIR/$SHORTDNAME +log_must mv $WORKDIR/$LONGFNAME $WORKDIR/$SHORTFNAME + +log_pass diff --git a/tests/zfs-tests/tests/functional/longname/longname_003_pos.ksh b/tests/zfs-tests/tests/functional/longname/longname_003_pos.ksh new file mode 100755 index 0000000000..f684b51439 --- /dev/null +++ b/tests/zfs-tests/tests/functional/longname/longname_003_pos.ksh @@ -0,0 +1,113 @@ +#! /bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or http://www.opensolaris.org/os/licensing. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright (c) 2021 by Nutanix. All rights reserved. +# + +. $STF_SUITE/include/libtest.shlib + +# +# DESCRIPTION: +# Check if longnames are handled correctly by ZIL replay and feature is activated. +# +# STRATEGY: +# 1. Create a zpool with longname feature disabled +# 2. Enable the feature@longname +# 3. Enable 'longname' property on the dataset. +# 4. Freeze the zpool +# 5. Create a longname +# 6. Export and import the zpool. +# 7. Replaying of longname create should activate the feature@longname +verify_runnable "global" + +function cleanup +{ + log_must rm -rf $WORKDIR + poolexists $TESTPOOL && zpool destroy $TESTPOOL +} + +log_assert "Check feature@longname and 'longname' dataset propery work correctly" + +log_onexit cleanup + +poolexists $TESTPOOL && zpool destroy $TESTPOOL + +log_must zpool create -o feature@longname=disabled $TESTPOOL $DISKS + +log_must zfs create $TESTPOOL/$TESTFS + +log_must zfs set mountpoint=$TESTDIR $TESTPOOL/$TESTFS + +LONGNAME=$(printf 'a%.0s' {1..512}) +LONGFNAME="file-$LONGNAME" +LONGDNAME="dir-$LONGNAME" +SHORTDNAME="dir-short" +SHORTFNAME="file-short" +WORKDIR=$TESTDIR/workdir + +log_must mkdir $WORKDIR + +log_must zpool set feature@longname=enabled $TESTPOOL +log_must zfs set longname=on $TESTPOOL/$TESTFS + +# Ensure that the feature is NOT activated yet as no longnamed file is created. +state=$(zpool get feature@longname -H -o value $TESTPOOL) +log_note "feature@longname on pool: $TESTPOOL : $state" + +if [[ "$state" != "enabled" ]]; then + log_fail "feature@longname has state $state (expected enabled)" +fi + +# +# This dd command works around an issue where ZIL records aren't created +# after freezing the pool unless a ZIL header already exists. Create a file +# synchronously to force ZFS to write one out. +# +log_must dd if=/dev/zero of=/$WORKDIR/sync conv=fdatasync,fsync bs=1 count=1 + +log_must zpool freeze $TESTPOOL + +log_must mkdir $WORKDIR/$LONGDNAME +log_must touch $WORKDIR/$LONGFNAME + +# Export and re-import the zpool +log_must zpool export $TESTPOOL +log_must zpool import $TESTPOOL + +# Ensure that the feature is activated once longnamed files are created. +state=$(zpool get feature@longname -H -o value $TESTPOOL) +log_note "feature@longname on pool: $TESTPOOL : $state" +if [[ "$state" != "active" ]]; then + log_fail "feature@longname has state $state (expected active)" +fi + +# Destroying the dataset where the feature is activated should put the feature +# back to 'enabled' state +log_must zfs destroy -r $TESTPOOL/$TESTFS +state=$(zpool get feature@longname -H -o value $TESTPOOL) +log_note "feature@longname on pool: $TESTPOOL : $state" +if [[ "$state" != "enabled" ]]; then + log_fail "feature@longname has state $state (expected active)" +fi + +log_pass diff --git a/tests/zfs-tests/tests/functional/longname/setup.ksh b/tests/zfs-tests/tests/functional/longname/setup.ksh new file mode 100755 index 0000000000..3f6759469b --- /dev/null +++ b/tests/zfs-tests/tests/functional/longname/setup.ksh @@ -0,0 +1,35 @@ +#!/bin/ksh -p +# +# CDDL HEADER START +# +# The contents of this file are subject to the terms of the +# Common Development and Distribution License (the "License"). +# You may not use this file except in compliance with the License. +# +# You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE +# or http://www.opensolaris.org/os/licensing. +# See the License for the specific language governing permissions +# and limitations under the License. +# +# When distributing Covered Code, include this CDDL HEADER in each +# file and include the License file at usr/src/OPENSOLARIS.LICENSE. +# If applicable, add the following below this CDDL HEADER, with the +# fields enclosed by brackets "[]" replaced with your own identifying +# information: Portions Copyright [yyyy] [name of copyright owner] +# +# CDDL HEADER END +# + +# +# Copyright 2007 Sun Microsystems, Inc. All rights reserved. +# Use is subject to license terms. +# + +# +# Copyright (c) 2021 by Nutanix. All rights reserved. +# + +. $STF_SUITE/include/libtest.shlib + +DISK=${DISKS%% *} +default_setup $DISK diff --git a/tests/zfs-tests/tests/functional/rsend/rsend.kshlib b/tests/zfs-tests/tests/functional/rsend/rsend.kshlib index 26e7c2cc25..0b2984c59d 100644 --- a/tests/zfs-tests/tests/functional/rsend/rsend.kshlib +++ b/tests/zfs-tests/tests/functional/rsend/rsend.kshlib @@ -720,6 +720,7 @@ function stream_has_features feature[resuming]="100000" feature[redacted]="200000" feature[compressed]="400000" + feature[longname]="10000000" typeset flag known derived=0 for flag in "$@"; do diff --git a/tests/zfs-tests/tests/functional/rsend/send-c_longname.ksh b/tests/zfs-tests/tests/functional/rsend/send-c_longname.ksh new file mode 100755 index 0000000000..3f7edc6810 --- /dev/null +++ b/tests/zfs-tests/tests/functional/rsend/send-c_longname.ksh @@ -0,0 +1,98 @@ +#!/bin/ksh -p + +# +# This file and its contents are supplied under the terms of the +# Common Development and Distribution License ("CDDL"), version 1.0. +# You may only use this file in accordance with the terms of version +# 1.0 of the CDDL. +# +# A full copy of the text of the CDDL should have accompanied this +# source. A copy of the CDDL is also available via the Internet at +# http://www.illumos.org/license/CDDL. +# + +# +# Copyright (c) 2021 by Nutanix. All rights reserved. +# + +. $STF_SUITE/tests/functional/rsend/rsend.kshlib +. $STF_SUITE/include/properties.shlib + +# +# Description: +# Verify that longname featureflag is present in the stream. +# +# Strategy: +# 1. Create a filesystem with longnamed files/directories. +# 2. Verify that the sendstream has the longname featureflag is present in the +# send stream. +# 3. Verify the created streams can be received correctly. +# 4. Verify that the longnamed files/directories are present in the received +# filesystem. +# + +verify_runnable "both" + +log_assert "Verify that longnames are handled correctly in send stream." +log_onexit cleanup_pool $POOL $POOL2 $POOL3 + +typeset sendfs=$POOL/sendfs +typeset recvfs=$POOL2/recvfs +typeset recvfs3=$POOL3/recvfs +typeset stream=$BACKDIR/stream +typeset dump=$BACKDIR/dump + +log_must zfs create -o longname=on $sendfs +typeset dir=$(get_prop mountpoint $sendfs) + +# Create a longnamed dir and a file in the send dataset +LONGNAME=$(printf 'a%.0s' {1..512}) +LONGFNAME="file-$LONGNAME" +LONGDNAME="dir-$LONGNAME" +log_must mkdir $dir/$LONGDNAME +log_must touch $dir/$LONGFNAME + +# When POOL3 is created by rsend.kshlib feature@longname is 'enabled'. +# Recreate the POOL3 with feature@longname disabled. +datasetexists $POOL3 && log_must zpool destroy $POOL3 +log_must zpool create -o feature@longname=disabled $POOL3 $DISK3 + +# Generate the streams and zstreamdump output. +log_must zfs snapshot $sendfs@now +log_must eval "zfs send -p $sendfs@now >$stream" +log_must eval "zstream dump -v <$stream >$dump" +log_must eval "zfs recv $recvfs <$stream" +cmp_ds_cont $sendfs $recvfs +log_must stream_has_features $stream longname + +# Ensure the the receiving pool has feature@longname activated after receiving. +feat_val=$(zpool get -H -o value feature@longname $POOL2) +log_note "Zpool $POOL2 feature@longname=$feat_val" +if [[ "$feat_val" != "active" ]]; then + log_fail "pool $POOL2 feature@longname=$feat_val (expected 'active')" +fi + +# Receiving of the stream on $POOL3 should fail as longname is not enabled +log_mustnot eval "zfs recv $recvfs3 <$stream" + +# Enable feature@longname and retry the receiving the stream. +# It should succeed this time. +log_must eval "zpool set feature@longname=enabled $POOL3" +log_must eval "zfs recv $recvfs3 <$stream" + +log_must zfs get longname $recvfs3 +prop_val=$(zfs get -H -o value longname $recvfs3) +log_note "dataset $recvfs3 has longname=$prop_val" +if [[ "$prop_val" != "on" ]]; then + log_fail "$recvfs3 has longname=$prop_val (expected 'on')" +fi + +# +# TODO: +# - Add a testcase to cover the case where send-stream does not contain +# properties (generated without "-p"). +# In this case the target dataset would have longname files/directories which +# cannot be accessed if the dataset property 'longname=off'. +# + +log_pass "Longnames are handled correctly in send/recv"