zap.c revision 215198a6ad15cf4832370e2f19247abeb36b951a
fa9e4066f08beec538e775443c5be79dd423fcabahrens * CDDL HEADER START
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The contents of this file are subject to the terms of the
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens * Common Development and Distribution License (the "License").
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens * You may not use this file except in compliance with the License.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
fa9e4066f08beec538e775443c5be79dd423fcabahrens * See the License for the specific language governing permissions
fa9e4066f08beec538e775443c5be79dd423fcabahrens * and limitations under the License.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * When distributing Covered Code, include this CDDL HEADER in each
fa9e4066f08beec538e775443c5be79dd423fcabahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * If applicable, add the following below this CDDL HEADER, with the
fa9e4066f08beec538e775443c5be79dd423fcabahrens * fields enclosed by brackets "[]" replaced with your own identifying
fa9e4066f08beec538e775443c5be79dd423fcabahrens * information: Portions Copyright [yyyy] [name of copyright owner]
fa9e4066f08beec538e775443c5be79dd423fcabahrens * CDDL HEADER END
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
215198a6ad15cf4832370e2f19247abeb36b951aJoe Stein * Copyright (c) 2012, 2015 by Delphix. All rights reserved.
bc9014e6a81272073b9854d9f65dd59e18d18c35Justin Gibbs * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * This file contains the top half of the zfs directory structure
fa9e4066f08beec538e775443c5be79dd423fcabahrens * implementation. The bottom half is in zap_leaf.c.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The zdir is an extendable hash data structure. There is a table of
fa9e4066f08beec538e775443c5be79dd423fcabahrens * pointers to buckets (zap_t->zd_data->zd_leafs). The buckets are
fa9e4066f08beec538e775443c5be79dd423fcabahrens * each a constant size and hold a variable number of directory entries.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The buckets (aka "leaf nodes") are implemented in zap_leaf.c.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * The pointer table holds a power of 2 number of pointers.
fa9e4066f08beec538e775443c5be79dd423fcabahrens * (1<<zap_t->zd_data->zd_phys->zd_prefix_len). The bucket pointed to
fa9e4066f08beec538e775443c5be79dd423fcabahrens * by the pointer at index i in the table holds entries whose hash value
fa9e4066f08beec538e775443c5be79dd423fcabahrens * has a zd_prefix_len - bit prefix
f65e61c04bc28ffd6bda04619c84330b420450b5ahrensint fzap_default_block_shift = 14; /* 16k blocksize */
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbsextern inline zap_phys_t *zap_f_phys(zap_t *zap);
66328dd3ff7c6b3566c06478698593e061733076ahrensstatic uint64_t zap_allocate_blocks(zap_t *zap, int nblocks);
5ad820458efd0fdb914baff9c1447c22b819fa23nd if (block_type == ZBT_LEAF || block_type == BSWAP_64(ZBT_LEAF))
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* it's a ptrtbl block */
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwickfzap_upgrade(zap_t *zap, dmu_tx_t *tx, zap_flags_t flags)
fa9e4066f08beec538e775443c5be79dd423fcabahrens mutex_init(&zap->zap_f.zap_num_entries_mtx, 0, 0, 0);
bf16b11e8deb633dd6c4296d46e92399d1582df4Matthew Ahrens zap->zap_f.zap_block_shift = highbit64(zap->zap_dbuf->db_size) - 1;
fa9e4066f08beec538e775443c5be79dd423fcabahrens * explicitly zero it since it might be coming from an
fa9e4066f08beec538e775443c5be79dd423fcabahrens * initialized microzap
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens bzero(zap->zap_dbuf->db_data, zap->zap_dbuf->db_size);
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens zp->zap_ptrtbl.zt_shift = ZAP_EMBEDDED_PTRTBL_SHIFT(zap);
fa9e4066f08beec538e775443c5be79dd423fcabahrens zp->zap_freeblk = 2; /* block 1 will be the first leaf */
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens /* block 1 will be the first leaf */
fa9e4066f08beec538e775443c5be79dd423fcabahrens * set up block 1 - the first leaf
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick 1<<FZAP_BLOCK_SHIFT(zap), FTAG, &db, DMU_READ_NO_PREFETCH));
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (1);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Generic routines for dealing with the pointer & cookie tables.
fa9e4066f08beec538e775443c5be79dd423fcabahrens void (*transfer_func)(const uint64_t *src, uint64_t *dst, int n),
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* hepb = half the number of entries in a block */
66328dd3ff7c6b3566c06478698593e061733076ahrens newblk = zap_allocate_blocks(zap, tbl->zt_numblks * 2);
a2cdcdd260232b58202b11a9bfc0103c9449ed52Paul Dagnelie dmu_prefetch(zap->zap_objset, zap->zap_object, 0,
66328dd3ff7c6b3566c06478698593e061733076ahrens * Copy the ptrtbl from the old to new location.
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick (tbl->zt_blk + b) << bs, FTAG, &db_old, DMU_READ_NO_PREFETCH);
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* first half of entries in old[b] go to new[2*b+0] */
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick (newblk + 2*b+0) << bs, FTAG, &db_new, DMU_READ_NO_PREFETCH));
fa9e4066f08beec538e775443c5be79dd423fcabahrens transfer_func(db_old->db_data, db_new->db_data, hepb);
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* second half of entries in old[b] go to new[2*b+1] */
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick (newblk + 2*b+1) << bs, FTAG, &db_new, DMU_READ_NO_PREFETCH));
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock (void) dmu_free_range(zap->zap_objset, zap->zap_object,
fa9e4066f08beec538e775443c5be79dd423fcabahrens dprintf("finished; numblocks now %llu (%lluk entries)\n",
66328dd3ff7c6b3566c06478698593e061733076ahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrenszap_table_store(zap_t *zap, zap_table_phys_t *tbl, uint64_t idx, uint64_t val,
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick (tbl->zt_blk + blk) << bs, FTAG, &db, DMU_READ_NO_PREFETCH);
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock return (0);
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrockzap_table_load(zap_t *zap, zap_table_phys_t *tbl, uint64_t idx, uint64_t *valp)
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick (tbl->zt_blk + blk) << bs, FTAG, &db, DMU_READ_NO_PREFETCH);
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock * read the nextblk for the sake of i/o error checking,
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock * so that zap_table_load() will catch errors for
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock * zap_table_store.
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Routines for growing the ptrtbl.
fa9e4066f08beec538e775443c5be79dd423fcabahrenszap_ptrtbl_transfer(const uint64_t *src, uint64_t *dst, int n)
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (i = 0; i < n; i++) {
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * The pointer table should never use more hash bits than we
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * have (otherwise we'd be using useless zero bits to index it).
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * If we are within 2 bits of running out, stop growing, since
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick * this is already an aberrant condition.
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_f_phys(zap)->zap_ptrtbl.zt_shift >= zap_hashbits(zap) - 2)
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens * We are outgrowing the "embedded" ptrtbl (the one
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens * stored in the header block). Give it its own entire
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens * block, which will double the size of the ptrtbl.
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ASSERT3U(zap_f_phys(zap)->zap_ptrtbl.zt_shift, ==,
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ASSERT0(zap_f_phys(zap)->zap_ptrtbl.zt_blk);
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick newblk << FZAP_BLOCK_SHIFT(zap), FTAG, &db_new,
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens zap_ptrtbl_transfer(&ZAP_EMBEDDED_PTRTBL_ENT(zap, 0),
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens db_new->db_data, 1 << ZAP_EMBEDDED_PTRTBL_SHIFT(zap));
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zap_f_phys(zap)->zap_ptrtbl.zt_blk = newblk;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ASSERT3U(1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift, ==,
66328dd3ff7c6b3566c06478698593e061733076ahrens return (0);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs return (zap_table_grow(zap, &zap_f_phys(zap)->zap_ptrtbl,
fa9e4066f08beec538e775443c5be79dd423fcabahrenszap_increment_num_entries(zap_t *zap, int delta, dmu_tx_t *tx)
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ASSERT(delta > 0 || zap_f_phys(zap)->zap_num_entries >= -delta);
bc9014e6a81272073b9854d9f65dd59e18d18c35Justin Gibbs zap_leaf_t *l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock VERIFY(0 == dmu_buf_hold(zap->zap_objset, zap->zap_object,
47cb52daa729f19e298c85a84e8df069365c5232Jeff Bonwick l->l_blkid << FZAP_BLOCK_SHIFT(zap), NULL, &l->l_dbuf,
bc9014e6a81272073b9854d9f65dd59e18d18c35Justin Gibbs dmu_buf_init_user(&l->l_dbu, zap_leaf_pageout, &l->l_dbuf);
bc9014e6a81272073b9854d9f65dd59e18d18c35Justin Gibbs winner = dmu_buf_set_user(l->l_dbuf, &l->l_dbu);
66328dd3ff7c6b3566c06478698593e061733076ahrens return (l);
fa9e4066f08beec538e775443c5be79dd423fcabahrens mutex_enter(&zap->zap_f.zap_num_entries_mtx); /* unnecessary */
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Routines for obtaining zap_leaf_t's
bc9014e6a81272073b9854d9f65dd59e18d18c35Justin Gibbs l = kmem_zalloc(sizeof (zap_leaf_t), KM_SLEEP);
bc9014e6a81272073b9854d9f65dd59e18d18c35Justin Gibbs dmu_buf_init_user(&l->l_dbu, zap_leaf_pageout, &l->l_dbuf);
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* someone else set it first */
66328dd3ff7c6b3566c06478698593e061733076ahrens * lhr_pad was previously used for the next leaf in the leaf
66328dd3ff7c6b3566c06478698593e061733076ahrens * chain. There should be no chained leafs (as we have removed
66328dd3ff7c6b3566c06478698593e061733076ahrens * support for them).
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens * There should be more hash entries than there can be
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens * chunks to put in the hash table
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens ASSERT3U(ZAP_LEAF_HASH_NUMENTRIES(l), >, ZAP_LEAF_NUMCHUNKS(l) / 3);
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens /* The chunks should begin at the end of the hash table */
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs &zap_leaf_phys(l)->l_hash[ZAP_LEAF_HASH_NUMENTRIES(l)]);
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens /* The chunks should end at the end of the block */
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens ASSERT3U((uintptr_t)&ZAP_LEAF_CHUNK(l, ZAP_LEAF_NUMCHUNKS(l)) -
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs (uintptr_t)zap_leaf_phys(l), ==, l->l_dbuf->db_size);
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (l);
66328dd3ff7c6b3566c06478698593e061733076ahrenszap_get_leaf_byblk(zap_t *zap, uint64_t blkid, dmu_tx_t *tx, krw_t lt,
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs * Must lock before dirtying, otherwise zap_leaf_phys(l) could change,
fa9e4066f08beec538e775443c5be79dd423fcabahrens * causing ASSERT below to fail.
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_block_type, ==, ZBT_LEAF);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ASSERT3U(zap_leaf_phys(l)->l_hdr.lh_magic, ==, ZAP_LEAF_MAGIC);
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock return (0);
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrockzap_idx_to_blk(zap_t *zap, uint64_t idx, uint64_t *valp)
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs (1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift));
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock return (0);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs return (zap_table_load(zap, &zap_f_phys(zap)->zap_ptrtbl,
fa9e4066f08beec538e775443c5be79dd423fcabahrenszap_set_idx_to_blk(zap_t *zap, uint64_t idx, uint64_t blk, dmu_tx_t *tx)
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_f_phys(zap)->zap_ptrtbl.zt_blk == 0) {
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock return (0);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs return (zap_table_store(zap, &zap_f_phys(zap)->zap_ptrtbl,
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrockzap_deref_leaf(zap_t *zap, uint64_t h, dmu_tx_t *tx, krw_t lt, zap_leaf_t **lp)
02525cd08fb3730fff3a69cb5376443d481f7839Chunwei Chen /* Reality check for corrupt zap objects (leaf or header). */
02525cd08fb3730fff3a69cb5376443d481f7839Chunwei Chen if ((zap_f_phys(zap)->zap_block_type != ZBT_LEAF &&
02525cd08fb3730fff3a69cb5376443d481f7839Chunwei Chen zap_f_phys(zap)->zap_block_type != ZBT_HEADER) ||
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs idx = ZAP_HASH_IDX(h, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ZAP_HASH_IDX(h, zap_leaf_phys(*lp)->l_hdr.lh_prefix_len) ==
ad860c82b2a8ffe69f42630da73ded0c92a7a36ebonwickzap_expand_leaf(zap_name_t *zn, zap_leaf_t *l, dmu_tx_t *tx, zap_leaf_t **lp)
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs int old_prefix_len = zap_leaf_phys(l)->l_hdr.lh_prefix_len;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ASSERT3U(old_prefix_len, <=, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs old_prefix_len == zap_f_phys(zap)->zap_ptrtbl.zt_shift) {
66328dd3ff7c6b3566c06478698593e061733076ahrens /* We failed to upgrade, or need to grow the pointer table */
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_leaf_phys(l)->l_hdr.lh_prefix_len != old_prefix_len) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* it split while our locks were down */
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock return (0);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs ASSERT3U(old_prefix_len, <, zap_f_phys(zap)->zap_ptrtbl.zt_shift);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs prefix_diff = zap_f_phys(zap)->zap_ptrtbl.zt_shift -
66328dd3ff7c6b3566c06478698593e061733076ahrens sibling = (ZAP_HASH_IDX(hash, old_prefix_len + 1) | 1) << prefix_diff;
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock /* check for i/o errors before doing zap_leaf_split */
66328dd3ff7c6b3566c06478698593e061733076ahrens /* set sibling pointers */
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = zap_set_idx_to_blk(zap, sibling+i, nl->l_blkid, tx);
fb09f5aad449c97fe309678f3f604982b563a96fMadhav Suresh ASSERT0(err); /* we checked for i/o errors above */
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (hash & (1ULL << (64 - zap_leaf_phys(l)->l_hdr.lh_prefix_len))) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* we want the sibling */
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock return (0);
ad860c82b2a8ffe69f42630da73ded0c92a7a36ebonwickzap_put_leaf_maybe_grow_ptrtbl(zap_name_t *zn, zap_leaf_t *l, dmu_tx_t *tx)
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs int shift = zap_f_phys(zap)->zap_ptrtbl.zt_shift;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs int leaffull = (zap_leaf_phys(l)->l_hdr.lh_prefix_len == shift &&
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zap_leaf_phys(l)->l_hdr.lh_nfree < ZAP_LEAF_LOW_WATER);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (leaffull || zap_f_phys(zap)->zap_ptrtbl.zt_nextblk) {
66328dd3ff7c6b3566c06478698593e061733076ahrens * We are in the middle of growing the pointer table, or
66328dd3ff7c6b3566c06478698593e061733076ahrens * this leaf will soon make us grow it.
66328dd3ff7c6b3566c06478698593e061733076ahrens /* could have finished growing while our locks were down */
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_f_phys(zap)->zap_ptrtbl.zt_shift == shift)
486ae710326c477d7237ab803b67abb9119c0799Matthew Ahrens if (zn->zn_key_orig_numints * zn->zn_key_intlen > ZAP_MAXNAMELEN)
92241e0b80813d0b83c08e730a29b9d1831794fcTom Ericksonfzap_checksize(uint64_t integer_size, uint64_t num_integers)
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* Only integer sizes supported by C */
fa9e4066f08beec538e775443c5be79dd423fcabahrens return (0);
92241e0b80813d0b83c08e730a29b9d1831794fcTom Ericksonfzap_check(zap_name_t *zn, uint64_t integer_size, uint64_t num_integers)
92241e0b80813d0b83c08e730a29b9d1831794fcTom Erickson return (fzap_checksize(integer_size, num_integers));
da6c28aaf62fa55f0fdb8004aa40f88f23bf53f0amw * Routines for manipulating attributes.
da6c28aaf62fa55f0fdb8004aa40f88f23bf53f0amw uint64_t integer_size, uint64_t num_integers, void *buf,
da6c28aaf62fa55f0fdb8004aa40f88f23bf53f0amw err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, NULL, RW_READER, &l);
da6c28aaf62fa55f0fdb8004aa40f88f23bf53f0amw if (err == 0) {
92241e0b80813d0b83c08e730a29b9d1831794fcTom Erickson if ((err = fzap_checksize(integer_size, num_integers)) != 0) {
66328dd3ff7c6b3566c06478698593e061733076ahrens err = zap_entry_read(&zeh, integer_size, num_integers, buf);
b24ab6762772a3f6a89393947930c7fa61306783Jeff Bonwick (void) zap_entry_read_name(zn->zn_zap, &zeh, rn_len, realname);
92241e0b80813d0b83c08e730a29b9d1831794fcTom Erickson ASSERT(fzap_check(zn, integer_size, num_integers) == 0);
da6c28aaf62fa55f0fdb8004aa40f88f23bf53f0amw err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (err == 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (err == 0) {
ad860c82b2a8ffe69f42630da73ded0c92a7a36ebonwick zap = zn->zn_zap; /* zap_expand_leaf() may change zap */
92241e0b80813d0b83c08e730a29b9d1831794fcTom Erickson int err = fzap_check(zn, integer_size, num_integers);
fa9e4066f08beec538e775443c5be79dd423fcabahrens int integer_size, uint64_t num_integers, const void *val, dmu_tx_t *tx)
92241e0b80813d0b83c08e730a29b9d1831794fcTom Erickson err = fzap_check(zn, integer_size, num_integers);
da6c28aaf62fa55f0fdb8004aa40f88f23bf53f0amw err = zap_deref_leaf(zap, zn->zn_hash, tx, RW_WRITER, &l);
fa9e4066f08beec538e775443c5be79dd423fcabahrens err = zap_entry_update(&zeh, integer_size, num_integers, val);
ad860c82b2a8ffe69f42630da73ded0c92a7a36ebonwick zap = zn->zn_zap; /* zap_expand_leaf() may change zap */
da6c28aaf62fa55f0fdb8004aa40f88f23bf53f0amw err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, NULL, RW_READER, &l);
da6c28aaf62fa55f0fdb8004aa40f88f23bf53f0amw err = zap_deref_leaf(zn->zn_zap, zn->zn_hash, tx, RW_WRITER, &l);
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (err == 0) {
a2cdcdd260232b58202b11a9bfc0103c9449ed52Paul Dagnelie dmu_prefetch(zap->zap_objset, zap->zap_object, 0, blk << bs, 1 << bs,
088f389458728c464569a5506b58070254fa4f7dahrens * Helper functions for consumers.
ad135b5d644628e791c3188a6ecbd9c257961ef8Christopher Sidenzap_create_link(objset_t *os, dmu_object_type_t ot, uint64_t parent_obj,
ad135b5d644628e791c3188a6ecbd9c257961ef8Christopher Siden VERIFY((new_obj = zap_create(os, ot, DMU_OT_NONE, 0, tx)) > 0);
215198a6ad15cf4832370e2f19247abeb36b951aJoe Stein VERIFY0(zap_add(os, parent_obj, name, sizeof (uint64_t), 1, &new_obj,
e7437265dc2a4920c197ed4337665539d358b22cahrenszap_value_search(objset_t *os, uint64_t zapobj, uint64_t value, uint64_t mask,
e7437265dc2a4920c197ed4337665539d358b22cahrens if ((za->za_first_integer & mask) == (value & mask)) {
088f389458728c464569a5506b58070254fa4f7dahrenszap_join(objset_t *os, uint64_t fromobj, uint64_t intoobj, dmu_tx_t *tx)
b287be1ba86043996f49b1cc34c80cc620f9b841Will Andrews if (za.za_integer_length != 8 || za.za_num_integers != 1) {
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Lingzap_join_key(objset_t *os, uint64_t fromobj, uint64_t intoobj,
b287be1ba86043996f49b1cc34c80cc620f9b841Will Andrews if (za.za_integer_length != 8 || za.za_num_integers != 1) {
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Lingzap_join_increment(objset_t *os, uint64_t fromobj, uint64_t intoobj,
b287be1ba86043996f49b1cc34c80cc620f9b841Will Andrews if (za.za_integer_length != 8 || za.za_num_integers != 1) {
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Ling err = zap_lookup(os, intoobj, za.za_name, 8, 1, &delta);
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Ling err = zap_update(os, intoobj, za.za_name, 8, 1, &delta, tx);
088f389458728c464569a5506b58070254fa4f7dahrenszap_add_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx)
088f389458728c464569a5506b58070254fa4f7dahrens (void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
088f389458728c464569a5506b58070254fa4f7dahrenszap_remove_int(objset_t *os, uint64_t obj, uint64_t value, dmu_tx_t *tx)
088f389458728c464569a5506b58070254fa4f7dahrens (void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
088f389458728c464569a5506b58070254fa4f7dahrenszap_lookup_int(objset_t *os, uint64_t obj, uint64_t value)
088f389458728c464569a5506b58070254fa4f7dahrens (void) snprintf(name, sizeof (name), "%llx", (longlong_t)value);
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Ling (void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
f17457368189aa911f774c38c1f21875a568bdcaMatthew Ahrenszap_update_int_key(objset_t *os, uint64_t obj,
f17457368189aa911f774c38c1f21875a568bdcaMatthew Ahrens (void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
f17457368189aa911f774c38c1f21875a568bdcaMatthew Ahrens return (zap_update(os, obj, name, 8, 1, &value, tx));
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Lingzap_lookup_int_key(objset_t *os, uint64_t obj, uint64_t key, uint64_t *valuep)
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Ling (void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Lingzap_increment(objset_t *os, uint64_t obj, const char *name, int64_t delta,
9966ca11f4a1481acce85f690fa59e4084050627Matthew Ahrens err = zap_lookup(os, obj, name, 8, 1, &value);
9966ca11f4a1481acce85f690fa59e4084050627Matthew Ahrens err = zap_update(os, obj, name, 8, 1, &value, tx);
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Lingzap_increment_int(objset_t *os, uint64_t obj, uint64_t key, int64_t delta,
3f9d6ad73e45c6823b409f93b0c8d4f62861d2d5Lin Ling (void) snprintf(name, sizeof (name), "%llx", (longlong_t)key);
fa9e4066f08beec538e775443c5be79dd423fcabahrens * Routines for iterating over the attributes.
fa9e4066f08beec538e775443c5be79dd423fcabahrensfzap_cursor_retrieve(zap_t *zap, zap_cursor_t *zc, zap_attribute_t *za)
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* retrieve the next entry at or after zc_hash/zc_cd */
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* if no entry, return ENOENT */
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zap_leaf_phys(zc->zc_leaf)->l_hdr.lh_prefix_len) !=
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zap_leaf_phys(zc->zc_leaf)->l_hdr.lh_prefix)) {
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = zap_deref_leaf(zap, zc->zc_hash, NULL, RW_READER,
fa9e4066f08beec538e775443c5be79dd423fcabahrens err = zap_leaf_lookup_closest(l, zc->zc_hash, zc->zc_cd, &zeh);
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs (1ULL << (64 - zap_leaf_phys(l)->l_hdr.lh_prefix_len)) - 1;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_leaf_phys(l)->l_hdr.lh_prefix_len == 0 ||
fa9e4066f08beec538e775443c5be79dd423fcabahrens if (err == 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens err = zap_entry_read(&zeh, 8, 1, &za->za_first_integer);
fa9e4066f08beec538e775443c5be79dd423fcabahrenszap_stats_ptrtbl(zap_t *zap, uint64_t *tbl, int len, zap_stats_t *zs)
fa9e4066f08beec538e775443c5be79dd423fcabahrens * NB: if a leaf has more pointers than an entire ptrtbl block
fa9e4066f08beec538e775443c5be79dd423fcabahrens * can hold, then it'll be accounted for more than once, since
fa9e4066f08beec538e775443c5be79dd423fcabahrens * we won't have lastblk.
fa9e4066f08beec538e775443c5be79dd423fcabahrens for (i = 0; i < len; i++) {
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = zap_get_leaf_byblk(zap, tbl[i], NULL, RW_READER, &l);
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock if (err == 0) {
8248818d5849649ef734d62da097e90222a23763nd * Set zap_phys_t fields
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_num_leafs = zap_f_phys(zap)->zap_num_leafs;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_num_entries = zap_f_phys(zap)->zap_num_entries;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_num_blocks = zap_f_phys(zap)->zap_freeblk;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_block_type = zap_f_phys(zap)->zap_block_type;
8248818d5849649ef734d62da097e90222a23763nd * Set zap_ptrtbl fields
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_ptrtbl_len = 1ULL << zap_f_phys(zap)->zap_ptrtbl.zt_shift;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_ptrtbl_nextblk = zap_f_phys(zap)->zap_ptrtbl.zt_nextblk;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_ptrtbl_zt_blk = zap_f_phys(zap)->zap_ptrtbl.zt_blk;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_ptrtbl_zt_numblks = zap_f_phys(zap)->zap_ptrtbl.zt_numblks;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs zs->zs_ptrtbl_zt_shift = zap_f_phys(zap)->zap_ptrtbl.zt_shift;
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_f_phys(zap)->zap_ptrtbl.zt_numblks == 0) {
fa9e4066f08beec538e775443c5be79dd423fcabahrens /* the ptrtbl is entirely in the header block. */
f65e61c04bc28ffd6bda04619c84330b420450b5ahrens zap_stats_ptrtbl(zap, &ZAP_EMBEDDED_PTRTBL_ENT(zap, 0),
a2cdcdd260232b58202b11a9bfc0103c9449ed52Paul Dagnelie dmu_prefetch(zap->zap_objset, zap->zap_object, 0,
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs for (b = 0; b < zap_f_phys(zap)->zap_ptrtbl.zt_numblks;
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock err = dmu_buf_hold(zap->zap_objset, zap->zap_object,
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs (zap_f_phys(zap)->zap_ptrtbl.zt_blk + b) << bs,
ea8dc4b6d2251b437950c0056bc626b311c73c27eschrock if (err == 0) {
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadifzap_count_write(zap_name_t *zn, int add, uint64_t *towrite,
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * Account for the header block of the fatzap.
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi if (!add && dmu_buf_freeable(zap->zap_dbuf)) {
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * Account for the pointer table blocks.
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * If we are adding we need to account for the following cases :
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * - If the pointer table is embedded, this operation could force an
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * external pointer table.
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * - If this already has an external pointer table this operation
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * could extend the table.
c1379625401dfbe1c39b79136dd384a571d47fdeJustin T. Gibbs if (zap_f_phys(zap)->zap_ptrtbl.zt_blk == 0)
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * Now, check if the block containing leaf is freeable
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * and account accordingly.
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi err = zap_deref_leaf(zap, zn->zn_hash, NULL, RW_READER, &l);
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * If this an add operation, the leaf block could split.
3d6926289465757c3da780cea696825b0d730283Sanjeev Bagewadi * Hence, we need to account for an additional leaf block.
12a2833a0260ba374a4e6d8225376193e908784aSanjeev Bagewadi *towrite += (add ? 2 : 1) * l->l_dbuf->db_size;