003 File Manager
Current Path:
/usr/src/sys/contrib/openzfs/module/zfs
usr
/
src
/
sys
/
contrib
/
openzfs
/
module
/
zfs
/
📁
..
📄
Makefile.in
(4.57 KB)
📄
THIRDPARTYLICENSE.cityhash
(1.03 KB)
📄
THIRDPARTYLICENSE.cityhash.descrip
(39 B)
📄
abd.c
(31.72 KB)
📄
aggsum.c
(8.46 KB)
📄
arc.c
(330.91 KB)
📄
blkptr.c
(4.26 KB)
📄
bplist.c
(2.38 KB)
📄
bpobj.c
(27.82 KB)
📄
bptree.c
(8.2 KB)
📄
bqueue.c
(4.7 KB)
📄
btree.c
(65.19 KB)
📄
dataset_kstats.c
(6.32 KB)
📄
dbuf.c
(141.84 KB)
📄
dbuf_stats.c
(6.05 KB)
📄
ddt.c
(28.01 KB)
📄
ddt_zap.c
(4.38 KB)
📄
dmu.c
(60.44 KB)
📄
dmu_diff.c
(6.25 KB)
📄
dmu_object.c
(14.73 KB)
📄
dmu_objset.c
(79.38 KB)
📄
dmu_recv.c
(98.69 KB)
📄
dmu_redact.c
(37.09 KB)
📄
dmu_send.c
(95.51 KB)
📄
dmu_traverse.c
(21.98 KB)
📄
dmu_tx.c
(40.13 KB)
📄
dmu_zfetch.c
(13.56 KB)
📄
dnode.c
(71.32 KB)
📄
dnode_sync.c
(24.93 KB)
📄
dsl_bookmark.c
(51.22 KB)
📄
dsl_crypt.c
(78.06 KB)
📄
dsl_dataset.c
(141.61 KB)
📄
dsl_deadlist.c
(27.94 KB)
📄
dsl_deleg.c
(19.92 KB)
📄
dsl_destroy.c
(36.73 KB)
📄
dsl_dir.c
(65.45 KB)
📄
dsl_pool.c
(42.92 KB)
📄
dsl_prop.c
(32.88 KB)
📄
dsl_scan.c
(134.47 KB)
📄
dsl_synctask.c
(7.94 KB)
📄
dsl_userhold.c
(18.26 KB)
📄
edonr_zfs.c
(3.21 KB)
📄
fm.c
(39.75 KB)
📄
gzip.c
(2.57 KB)
📄
hkdf.c
(4.5 KB)
📄
lz4.c
(26.46 KB)
📄
lzjb.c
(3.85 KB)
📄
metaslab.c
(190.62 KB)
📄
mmp.c
(24.93 KB)
📄
multilist.c
(12.21 KB)
📄
objlist.c
(2.58 KB)
📄
pathname.c
(2.52 KB)
📄
range_tree.c
(24.9 KB)
📄
refcount.c
(7.82 KB)
📄
rrwlock.c
(10.82 KB)
📄
sa.c
(59.23 KB)
📄
sha256.c
(2.89 KB)
📄
skein_zfs.c
(2.92 KB)
📄
spa.c
(272.81 KB)
📄
spa_boot.c
(1.28 KB)
📄
spa_checkpoint.c
(21.54 KB)
📄
spa_config.c
(16.97 KB)
📄
spa_errlog.c
(11.35 KB)
📄
spa_history.c
(17.75 KB)
📄
spa_log_spacemap.c
(47.13 KB)
📄
spa_misc.c
(75.4 KB)
📄
spa_stats.c
(27.08 KB)
📄
space_map.c
(31.45 KB)
📄
space_reftree.c
(4.21 KB)
📄
txg.c
(27.49 KB)
📄
uberblock.c
(2.17 KB)
📄
unique.c
(2.53 KB)
📄
vdev.c
(143.91 KB)
📄
vdev_cache.c
(11.71 KB)
📄
vdev_draid.c
(96.86 KB)
📄
vdev_draid_rand.c
(1.16 KB)
📄
vdev_indirect.c
(60.83 KB)
📄
vdev_indirect_births.c
(5.88 KB)
📄
vdev_indirect_mapping.c
(18.06 KB)
📄
vdev_initialize.c
(22.38 KB)
📄
vdev_label.c
(57.52 KB)
📄
vdev_mirror.c
(26.76 KB)
📄
vdev_missing.c
(3.74 KB)
📄
vdev_queue.c
(37.61 KB)
📄
vdev_raidz.c
(76.09 KB)
📄
vdev_raidz_math.c
(17.18 KB)
📄
vdev_raidz_math_aarch64_neon.c
(112.13 KB)
📄
vdev_raidz_math_aarch64_neon_common.h
(23.64 KB)
📄
vdev_raidz_math_aarch64_neonx2.c
(5.39 KB)
📄
vdev_raidz_math_avx2.c
(10.64 KB)
📄
vdev_raidz_math_avx512bw.c
(10.93 KB)
📄
vdev_raidz_math_avx512f.c
(17.9 KB)
📄
vdev_raidz_math_impl.h
(34.78 KB)
📄
vdev_raidz_math_powerpc_altivec.c
(213.46 KB)
📄
vdev_raidz_math_powerpc_altivec_common.h
(22.62 KB)
📄
vdev_raidz_math_scalar.c
(9.54 KB)
📄
vdev_raidz_math_sse2.c
(25.84 KB)
📄
vdev_raidz_math_ssse3.c
(118 KB)
📄
vdev_rebuild.c
(34.65 KB)
📄
vdev_removal.c
(72.08 KB)
📄
vdev_root.c
(4.28 KB)
📄
vdev_trim.c
(50.92 KB)
📄
zap.c
(35.35 KB)
📄
zap_leaf.c
(22.75 KB)
📄
zap_micro.c
(42.18 KB)
📄
zcp.c
(42.29 KB)
📄
zcp_get.c
(21.01 KB)
📄
zcp_global.c
(1.9 KB)
📄
zcp_iter.c
(18.21 KB)
📄
zcp_set.c
(2.37 KB)
📄
zcp_synctask.c
(13.69 KB)
📄
zfeature.c
(17.6 KB)
📄
zfs_byteswap.c
(5.64 KB)
📄
zfs_fm.c
(40.58 KB)
📄
zfs_fuid.c
(19.69 KB)
📄
zfs_ioctl.c
(195.48 KB)
📄
zfs_log.c
(20.6 KB)
📄
zfs_onexit.c
(5.09 KB)
📄
zfs_quota.c
(12.93 KB)
📄
zfs_ratelimit.c
(2.35 KB)
📄
zfs_replay.c
(26.11 KB)
📄
zfs_rlock.c
(20.38 KB)
📄
zfs_sa.c
(13.09 KB)
📄
zfs_vnops.c
(22 KB)
📄
zil.c
(111.32 KB)
📄
zio.c
(144.72 KB)
📄
zio_checksum.c
(17.28 KB)
📄
zio_compress.c
(5.92 KB)
📄
zio_inject.c
(25.79 KB)
📄
zle.c
(2.52 KB)
📄
zrlock.c
(4.45 KB)
📄
zthr.c
(16.41 KB)
📄
zvol.c
(42.96 KB)
Editing: dmu_object.c
/* * 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) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2013, 2017 by Delphix. All rights reserved. * Copyright 2014 HybridCluster. All rights reserved. */ #include <sys/dbuf.h> #include <sys/dmu.h> #include <sys/dmu_impl.h> #include <sys/dmu_objset.h> #include <sys/dmu_tx.h> #include <sys/dnode.h> #include <sys/zap.h> #include <sys/zfeature.h> #include <sys/dsl_dataset.h> /* * Each of the concurrent object allocators will grab * 2^dmu_object_alloc_chunk_shift dnode slots at a time. The default is to * grab 128 slots, which is 4 blocks worth. This was experimentally * determined to be the lowest value that eliminates the measurable effect * of lock contention from this code path. */ int dmu_object_alloc_chunk_shift = 7; static uint64_t dmu_object_alloc_impl(objset_t *os, dmu_object_type_t ot, int blocksize, int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen, int dnodesize, dnode_t **allocated_dnode, void *tag, dmu_tx_t *tx) { uint64_t object; uint64_t L1_dnode_count = DNODES_PER_BLOCK << (DMU_META_DNODE(os)->dn_indblkshift - SPA_BLKPTRSHIFT); dnode_t *dn = NULL; int dn_slots = dnodesize >> DNODE_SHIFT; boolean_t restarted = B_FALSE; uint64_t *cpuobj = NULL; int dnodes_per_chunk = 1 << dmu_object_alloc_chunk_shift; int error; cpuobj = &os->os_obj_next_percpu[CPU_SEQID_UNSTABLE % os->os_obj_next_percpu_len]; if (dn_slots == 0) { dn_slots = DNODE_MIN_SLOTS; } else { ASSERT3S(dn_slots, >=, DNODE_MIN_SLOTS); ASSERT3S(dn_slots, <=, DNODE_MAX_SLOTS); } /* * The "chunk" of dnodes that is assigned to a CPU-specific * allocator needs to be at least one block's worth, to avoid * lock contention on the dbuf. It can be at most one L1 block's * worth, so that the "rescan after polishing off a L1's worth" * logic below will be sure to kick in. */ if (dnodes_per_chunk < DNODES_PER_BLOCK) dnodes_per_chunk = DNODES_PER_BLOCK; if (dnodes_per_chunk > L1_dnode_count) dnodes_per_chunk = L1_dnode_count; /* * The caller requested the dnode be returned as a performance * optimization in order to avoid releasing the hold only to * immediately reacquire it. Since they caller is responsible * for releasing the hold they must provide the tag. */ if (allocated_dnode != NULL) { ASSERT3P(tag, !=, NULL); } else { ASSERT3P(tag, ==, NULL); tag = FTAG; } object = *cpuobj; for (;;) { /* * If we finished a chunk of dnodes, get a new one from * the global allocator. */ if ((P2PHASE(object, dnodes_per_chunk) == 0) || (P2PHASE(object + dn_slots - 1, dnodes_per_chunk) < dn_slots)) { DNODE_STAT_BUMP(dnode_alloc_next_chunk); mutex_enter(&os->os_obj_lock); ASSERT0(P2PHASE(os->os_obj_next_chunk, dnodes_per_chunk)); object = os->os_obj_next_chunk; /* * Each time we polish off a L1 bp worth of dnodes * (2^12 objects), move to another L1 bp that's * still reasonably sparse (at most 1/4 full). Look * from the beginning at most once per txg. If we * still can't allocate from that L1 block, search * for an empty L0 block, which will quickly skip * to the end of the metadnode if no nearby L0 * blocks are empty. This fallback avoids a * pathology where full dnode blocks containing * large dnodes appear sparse because they have a * low blk_fill, leading to many failed allocation * attempts. In the long term a better mechanism to * search for sparse metadnode regions, such as * spacemaps, could be implemented. * * os_scan_dnodes is set during txg sync if enough * objects have been freed since the previous * rescan to justify backfilling again. * * Note that dmu_traverse depends on the behavior * that we use multiple blocks of the dnode object * before going back to reuse objects. Any change * to this algorithm should preserve that property * or find another solution to the issues described * in traverse_visitbp. */ if (P2PHASE(object, L1_dnode_count) == 0) { uint64_t offset; uint64_t blkfill; int minlvl; if (os->os_rescan_dnodes) { offset = 0; os->os_rescan_dnodes = B_FALSE; } else { offset = object << DNODE_SHIFT; } blkfill = restarted ? 1 : DNODES_PER_BLOCK >> 2; minlvl = restarted ? 1 : 2; restarted = B_TRUE; error = dnode_next_offset(DMU_META_DNODE(os), DNODE_FIND_HOLE, &offset, minlvl, blkfill, 0); if (error == 0) { object = offset >> DNODE_SHIFT; } } /* * Note: if "restarted", we may find a L0 that * is not suitably aligned. */ os->os_obj_next_chunk = P2ALIGN(object, dnodes_per_chunk) + dnodes_per_chunk; (void) atomic_swap_64(cpuobj, object); mutex_exit(&os->os_obj_lock); } /* * The value of (*cpuobj) before adding dn_slots is the object * ID assigned to us. The value afterwards is the object ID * assigned to whoever wants to do an allocation next. */ object = atomic_add_64_nv(cpuobj, dn_slots) - dn_slots; /* * XXX We should check for an i/o error here and return * up to our caller. Actually we should pre-read it in * dmu_tx_assign(), but there is currently no mechanism * to do so. */ error = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE, dn_slots, tag, &dn); if (error == 0) { rw_enter(&dn->dn_struct_rwlock, RW_WRITER); /* * Another thread could have allocated it; check * again now that we have the struct lock. */ if (dn->dn_type == DMU_OT_NONE) { dnode_allocate(dn, ot, blocksize, indirect_blockshift, bonustype, bonuslen, dn_slots, tx); rw_exit(&dn->dn_struct_rwlock); dmu_tx_add_new_object(tx, dn); /* * Caller requested the allocated dnode be * returned and is responsible for the hold. */ if (allocated_dnode != NULL) *allocated_dnode = dn; else dnode_rele(dn, tag); return (object); } rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, tag); DNODE_STAT_BUMP(dnode_alloc_race); } /* * Skip to next known valid starting point on error. This * is the start of the next block of dnodes. */ if (dmu_object_next(os, &object, B_TRUE, 0) != 0) { object = P2ROUNDUP(object + 1, DNODES_PER_BLOCK); DNODE_STAT_BUMP(dnode_alloc_next_block); } (void) atomic_swap_64(cpuobj, object); } } uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx) { return dmu_object_alloc_impl(os, ot, blocksize, 0, bonustype, bonuslen, 0, NULL, NULL, tx); } uint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize, int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx) { return dmu_object_alloc_impl(os, ot, blocksize, indirect_blockshift, bonustype, bonuslen, 0, NULL, NULL, tx); } uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx) { return (dmu_object_alloc_impl(os, ot, blocksize, 0, bonustype, bonuslen, dnodesize, NULL, NULL, tx)); } /* * Allocate a new object and return a pointer to the newly allocated dnode * via the allocated_dnode argument. The returned dnode will be held and * the caller is responsible for releasing the hold by calling dnode_rele(). */ uint64_t dmu_object_alloc_hold(objset_t *os, dmu_object_type_t ot, int blocksize, int indirect_blockshift, dmu_object_type_t bonustype, int bonuslen, int dnodesize, dnode_t **allocated_dnode, void *tag, dmu_tx_t *tx) { return (dmu_object_alloc_impl(os, ot, blocksize, indirect_blockshift, bonustype, bonuslen, dnodesize, allocated_dnode, tag, tx)); } int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx) { return (dmu_object_claim_dnsize(os, object, ot, blocksize, bonustype, bonuslen, 0, tx)); } int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *tx) { dnode_t *dn; int dn_slots = dnodesize >> DNODE_SHIFT; int err; if (dn_slots == 0) dn_slots = DNODE_MIN_SLOTS; ASSERT3S(dn_slots, >=, DNODE_MIN_SLOTS); ASSERT3S(dn_slots, <=, DNODE_MAX_SLOTS); if (object == DMU_META_DNODE_OBJECT && !dmu_tx_private_ok(tx)) return (SET_ERROR(EBADF)); err = dnode_hold_impl(os, object, DNODE_MUST_BE_FREE, dn_slots, FTAG, &dn); if (err) return (err); dnode_allocate(dn, ot, blocksize, 0, bonustype, bonuslen, dn_slots, tx); dmu_tx_add_new_object(tx, dn); dnode_rele(dn, FTAG); return (0); } int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx) { return (dmu_object_reclaim_dnsize(os, object, ot, blocksize, bonustype, bonuslen, DNODE_MIN_SIZE, B_FALSE, tx)); } int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, int dnodesize, boolean_t keep_spill, dmu_tx_t *tx) { dnode_t *dn; int dn_slots = dnodesize >> DNODE_SHIFT; int err; if (dn_slots == 0) dn_slots = DNODE_MIN_SLOTS; if (object == DMU_META_DNODE_OBJECT) return (SET_ERROR(EBADF)); err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0, FTAG, &dn); if (err) return (err); dnode_reallocate(dn, ot, blocksize, bonustype, bonuslen, dn_slots, keep_spill, tx); dnode_rele(dn, FTAG); return (err); } int dmu_object_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx) { dnode_t *dn; int err; err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0, FTAG, &dn); if (err) return (err); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); if (dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR) { dbuf_rm_spill(dn, tx); dnode_rm_spill(dn, tx); } rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); return (err); } int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx) { dnode_t *dn; int err; ASSERT(object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx)); err = dnode_hold_impl(os, object, DNODE_MUST_BE_ALLOCATED, 0, FTAG, &dn); if (err) return (err); ASSERT(dn->dn_type != DMU_OT_NONE); /* * If we don't create this free range, we'll leak indirect blocks when * we get to freeing the dnode in syncing context. */ dnode_free_range(dn, 0, DMU_OBJECT_END, tx); dnode_free(dn, tx); dnode_rele(dn, FTAG); return (0); } /* * Return (in *objectp) the next object which is allocated (or a hole) * after *object, taking into account only objects that may have been modified * after the specified txg. */ int dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole, uint64_t txg) { uint64_t offset; uint64_t start_obj; struct dsl_dataset *ds = os->os_dsl_dataset; int error; if (*objectp == 0) { start_obj = 1; } else if (ds && dsl_dataset_feature_is_active(ds, SPA_FEATURE_LARGE_DNODE)) { uint64_t i = *objectp + 1; uint64_t last_obj = *objectp | (DNODES_PER_BLOCK - 1); dmu_object_info_t doi; /* * Scan through the remaining meta dnode block. The contents * of each slot in the block are known so it can be quickly * checked. If the block is exhausted without a match then * hand off to dnode_next_offset() for further scanning. */ while (i <= last_obj) { error = dmu_object_info(os, i, &doi); if (error == ENOENT) { if (hole) { *objectp = i; return (0); } else { i++; } } else if (error == EEXIST) { i++; } else if (error == 0) { if (hole) { i += doi.doi_dnodesize >> DNODE_SHIFT; } else { *objectp = i; return (0); } } else { return (error); } } start_obj = i; } else { start_obj = *objectp + 1; } offset = start_obj << DNODE_SHIFT; error = dnode_next_offset(DMU_META_DNODE(os), (hole ? DNODE_FIND_HOLE : 0), &offset, 0, DNODES_PER_BLOCK, txg); *objectp = offset >> DNODE_SHIFT; return (error); } /* * Turn this object from old_type into DMU_OTN_ZAP_METADATA, and bump the * refcount on SPA_FEATURE_EXTENSIBLE_DATASET. * * Only for use from syncing context, on MOS objects. */ void dmu_object_zapify(objset_t *mos, uint64_t object, dmu_object_type_t old_type, dmu_tx_t *tx) { dnode_t *dn; ASSERT(dmu_tx_is_syncing(tx)); VERIFY0(dnode_hold(mos, object, FTAG, &dn)); if (dn->dn_type == DMU_OTN_ZAP_METADATA) { dnode_rele(dn, FTAG); return; } ASSERT3U(dn->dn_type, ==, old_type); ASSERT0(dn->dn_maxblkid); /* * We must initialize the ZAP data before changing the type, * so that concurrent calls to *_is_zapified() can determine if * the object has been completely zapified by checking the type. */ mzap_create_impl(dn, 0, 0, tx); dn->dn_next_type[tx->tx_txg & TXG_MASK] = dn->dn_type = DMU_OTN_ZAP_METADATA; dnode_setdirty(dn, tx); dnode_rele(dn, FTAG); spa_feature_incr(dmu_objset_spa(mos), SPA_FEATURE_EXTENSIBLE_DATASET, tx); } void dmu_object_free_zapified(objset_t *mos, uint64_t object, dmu_tx_t *tx) { dnode_t *dn; dmu_object_type_t t; ASSERT(dmu_tx_is_syncing(tx)); VERIFY0(dnode_hold(mos, object, FTAG, &dn)); t = dn->dn_type; dnode_rele(dn, FTAG); if (t == DMU_OTN_ZAP_METADATA) { spa_feature_decr(dmu_objset_spa(mos), SPA_FEATURE_EXTENSIBLE_DATASET, tx); } VERIFY0(dmu_object_free(mos, object, tx)); } EXPORT_SYMBOL(dmu_object_alloc); EXPORT_SYMBOL(dmu_object_alloc_ibs); EXPORT_SYMBOL(dmu_object_alloc_dnsize); EXPORT_SYMBOL(dmu_object_alloc_hold); EXPORT_SYMBOL(dmu_object_claim); EXPORT_SYMBOL(dmu_object_claim_dnsize); EXPORT_SYMBOL(dmu_object_reclaim); EXPORT_SYMBOL(dmu_object_reclaim_dnsize); EXPORT_SYMBOL(dmu_object_rm_spill); EXPORT_SYMBOL(dmu_object_free); EXPORT_SYMBOL(dmu_object_next); EXPORT_SYMBOL(dmu_object_zapify); EXPORT_SYMBOL(dmu_object_free_zapified); /* BEGIN CSTYLED */ ZFS_MODULE_PARAM(zfs, , dmu_object_alloc_chunk_shift, INT, ZMOD_RW, "CPU-specific allocator grabs 2^N objects at once"); /* END CSTYLED */
Upload File
Create Folder