1fa9e4066Sahrens /* 2fa9e4066Sahrens * CDDL HEADER START 3fa9e4066Sahrens * 4fa9e4066Sahrens * The contents of this file are subject to the terms of the 5ea8dc4b6Seschrock * Common Development and Distribution License (the "License"). 6ea8dc4b6Seschrock * You may not use this file except in compliance with the License. 7fa9e4066Sahrens * 8fa9e4066Sahrens * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE 9fa9e4066Sahrens * or http://www.opensolaris.org/os/licensing. 10fa9e4066Sahrens * See the License for the specific language governing permissions 11fa9e4066Sahrens * and limitations under the License. 12fa9e4066Sahrens * 13fa9e4066Sahrens * When distributing Covered Code, include this CDDL HEADER in each 14fa9e4066Sahrens * file and include the License file at usr/src/OPENSOLARIS.LICENSE. 15fa9e4066Sahrens * If applicable, add the following below this CDDL HEADER, with the 16fa9e4066Sahrens * fields enclosed by brackets "[]" replaced with your own identifying 17fa9e4066Sahrens * information: Portions Copyright [yyyy] [name of copyright owner] 18fa9e4066Sahrens * 19fa9e4066Sahrens * CDDL HEADER END 20fa9e4066Sahrens */ 21ad135b5dSChristopher Siden 22fa9e4066Sahrens /* 2394d1a210STim Haley * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. 24edf345e6SMatthew Ahrens * Copyright (c) 2011, 2014 by Delphix. All rights reserved. 25ec5cf9d5SAlexander Stetsenko * Copyright 2011 Nexenta Systems, Inc. All rights reserved. 264e3c9f44SBill Pijewski * Copyright (c) 2012, Joyent, Inc. All rights reserved. 27e4be62a2SSaso Kiselkov * Copyright 2013 DEY Storage Systems, Inc. 28e77d42eaSMatthew Ahrens * Copyright 2014 HybridCluster. All rights reserved. 29bc9014e6SJustin Gibbs * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. 30ec5cf9d5SAlexander Stetsenko */ 31fa9e4066Sahrens 3255da60b9SMark J Musante /* Portions Copyright 2010 Robert Milkowski */ 3355da60b9SMark J Musante 34fa9e4066Sahrens #ifndef _SYS_DMU_H 35fa9e4066Sahrens #define _SYS_DMU_H 36fa9e4066Sahrens 37fa9e4066Sahrens /* 38fa9e4066Sahrens * This file describes the interface that the DMU provides for its 39fa9e4066Sahrens * consumers. 40fa9e4066Sahrens * 41fa9e4066Sahrens * The DMU also interacts with the SPA. That interface is described in 42fa9e4066Sahrens * dmu_spa.h. 43fa9e4066Sahrens */ 44fa9e4066Sahrens 45bc9014e6SJustin Gibbs #include <sys/zfs_context.h> 46fa9e4066Sahrens #include <sys/inttypes.h> 47ecd6cf80Smarks #include <sys/cred.h> 484445fffbSMatthew Ahrens #include <sys/fs/zfs.h> 49fa9e4066Sahrens 50fa9e4066Sahrens #ifdef __cplusplus 51fa9e4066Sahrens extern "C" { 52fa9e4066Sahrens #endif 53fa9e4066Sahrens 54fa9e4066Sahrens struct uio; 55c242f9a0Schunli zhang - Sun Microsystems - Irvine United States struct xuio; 5644eda4d7Smaybee struct page; 57fa9e4066Sahrens struct vnode; 58fa9e4066Sahrens struct spa; 59fa9e4066Sahrens struct zilog; 60fa9e4066Sahrens struct zio; 61fa9e4066Sahrens struct blkptr; 62fa9e4066Sahrens struct zap_cursor; 63fa9e4066Sahrens struct dsl_dataset; 64fa9e4066Sahrens struct dsl_pool; 65fa9e4066Sahrens struct dnode; 66fa9e4066Sahrens struct drr_begin; 67fa9e4066Sahrens struct drr_end; 687802d7bfSMatthew Ahrens struct zbookmark_phys; 6944cd46caSbillm struct spa; 70a2eea2e1Sahrens struct nvlist; 712fdbea25SAleksandr Guzovskiy struct arc_buf; 72b24ab676SJeff Bonwick struct zio_prop; 730a586ceaSMark Shellenbaum struct sa_handle; 74fa9e4066Sahrens 75fa9e4066Sahrens typedef struct objset objset_t; 76fa9e4066Sahrens typedef struct dmu_tx dmu_tx_t; 77fa9e4066Sahrens typedef struct dsl_dir dsl_dir_t; 78fa9e4066Sahrens 79ad135b5dSChristopher Siden typedef enum dmu_object_byteswap { 80ad135b5dSChristopher Siden DMU_BSWAP_UINT8, 81ad135b5dSChristopher Siden DMU_BSWAP_UINT16, 82ad135b5dSChristopher Siden DMU_BSWAP_UINT32, 83ad135b5dSChristopher Siden DMU_BSWAP_UINT64, 84ad135b5dSChristopher Siden DMU_BSWAP_ZAP, 85ad135b5dSChristopher Siden DMU_BSWAP_DNODE, 86ad135b5dSChristopher Siden DMU_BSWAP_OBJSET, 87ad135b5dSChristopher Siden DMU_BSWAP_ZNODE, 88ad135b5dSChristopher Siden DMU_BSWAP_OLDACL, 89ad135b5dSChristopher Siden DMU_BSWAP_ACL, 90ad135b5dSChristopher Siden /* 91ad135b5dSChristopher Siden * Allocating a new byteswap type number makes the on-disk format 92ad135b5dSChristopher Siden * incompatible with any other format that uses the same number. 93ad135b5dSChristopher Siden * 94ad135b5dSChristopher Siden * Data can usually be structured to work with one of the 95ad135b5dSChristopher Siden * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 96ad135b5dSChristopher Siden */ 97ad135b5dSChristopher Siden DMU_BSWAP_NUMFUNCS 98ad135b5dSChristopher Siden } dmu_object_byteswap_t; 99ad135b5dSChristopher Siden 100ad135b5dSChristopher Siden #define DMU_OT_NEWTYPE 0x80 101ad135b5dSChristopher Siden #define DMU_OT_METADATA 0x40 102ad135b5dSChristopher Siden #define DMU_OT_BYTESWAP_MASK 0x3f 103ad135b5dSChristopher Siden 104ad135b5dSChristopher Siden /* 105ad135b5dSChristopher Siden * Defines a uint8_t object type. Object types specify if the data 106ad135b5dSChristopher Siden * in the object is metadata (boolean) and how to byteswap the data 107ad135b5dSChristopher Siden * (dmu_object_byteswap_t). 108ad135b5dSChristopher Siden */ 109ad135b5dSChristopher Siden #define DMU_OT(byteswap, metadata) \ 110ad135b5dSChristopher Siden (DMU_OT_NEWTYPE | \ 111ad135b5dSChristopher Siden ((metadata) ? DMU_OT_METADATA : 0) | \ 112ad135b5dSChristopher Siden ((byteswap) & DMU_OT_BYTESWAP_MASK)) 113ad135b5dSChristopher Siden 114ad135b5dSChristopher Siden #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 115ad135b5dSChristopher Siden ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ 116ad135b5dSChristopher Siden (ot) < DMU_OT_NUMTYPES) 117ad135b5dSChristopher Siden 118ad135b5dSChristopher Siden #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 119ad135b5dSChristopher Siden ((ot) & DMU_OT_METADATA) : \ 120ad135b5dSChristopher Siden dmu_ot[(ot)].ot_metadata) 121ad135b5dSChristopher Siden 1225d7b4d43SMatthew Ahrens /* 1235d7b4d43SMatthew Ahrens * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't 1245d7b4d43SMatthew Ahrens * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill 1255d7b4d43SMatthew Ahrens * is repurposed for embedded BPs. 1265d7b4d43SMatthew Ahrens */ 1275d7b4d43SMatthew Ahrens #define DMU_OT_HAS_FILL(ot) \ 1285d7b4d43SMatthew Ahrens ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) 1295d7b4d43SMatthew Ahrens 130ad135b5dSChristopher Siden #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 131ad135b5dSChristopher Siden ((ot) & DMU_OT_BYTESWAP_MASK) : \ 132ad135b5dSChristopher Siden dmu_ot[(ot)].ot_byteswap) 133ad135b5dSChristopher Siden 134fa9e4066Sahrens typedef enum dmu_object_type { 135fa9e4066Sahrens DMU_OT_NONE, 136fa9e4066Sahrens /* general: */ 137fa9e4066Sahrens DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 138fa9e4066Sahrens DMU_OT_OBJECT_ARRAY, /* UINT64 */ 139fa9e4066Sahrens DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 140fa9e4066Sahrens DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 141cde58dbcSMatthew Ahrens DMU_OT_BPOBJ, /* UINT64 */ 142cde58dbcSMatthew Ahrens DMU_OT_BPOBJ_HDR, /* UINT64 */ 143fa9e4066Sahrens /* spa: */ 144fa9e4066Sahrens DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 145fa9e4066Sahrens DMU_OT_SPACE_MAP, /* UINT64 */ 146fa9e4066Sahrens /* zil: */ 147fa9e4066Sahrens DMU_OT_INTENT_LOG, /* UINT64 */ 148fa9e4066Sahrens /* dmu: */ 149fa9e4066Sahrens DMU_OT_DNODE, /* DNODE */ 150fa9e4066Sahrens DMU_OT_OBJSET, /* OBJSET */ 151fa9e4066Sahrens /* dsl: */ 1521649cd4bStabriz DMU_OT_DSL_DIR, /* UINT64 */ 15387e5029aSahrens DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 15487e5029aSahrens DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 155fa9e4066Sahrens DMU_OT_DSL_PROPS, /* ZAP */ 1561649cd4bStabriz DMU_OT_DSL_DATASET, /* UINT64 */ 157fa9e4066Sahrens /* zpl: */ 158fa9e4066Sahrens DMU_OT_ZNODE, /* ZNODE */ 159da6c28aaSamw DMU_OT_OLDACL, /* Old ACL */ 160fa9e4066Sahrens DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 161fa9e4066Sahrens DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 162fa9e4066Sahrens DMU_OT_MASTER_NODE, /* ZAP */ 163893a6d32Sahrens DMU_OT_UNLINKED_SET, /* ZAP */ 164fa9e4066Sahrens /* zvol: */ 165fa9e4066Sahrens DMU_OT_ZVOL, /* UINT8 */ 166fa9e4066Sahrens DMU_OT_ZVOL_PROP, /* ZAP */ 167fa9e4066Sahrens /* other; for testing only! */ 168fa9e4066Sahrens DMU_OT_PLAIN_OTHER, /* UINT8 */ 169fa9e4066Sahrens DMU_OT_UINT64_OTHER, /* UINT64 */ 170fa9e4066Sahrens DMU_OT_ZAP_OTHER, /* ZAP */ 171ea8dc4b6Seschrock /* new object types: */ 172ea8dc4b6Seschrock DMU_OT_ERROR_LOG, /* ZAP */ 17306eeb2adSek DMU_OT_SPA_HISTORY, /* UINT8 */ 17406eeb2adSek DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 175b1b8ab34Slling DMU_OT_POOL_PROPS, /* ZAP */ 176ecd6cf80Smarks DMU_OT_DSL_PERMS, /* ZAP */ 177da6c28aaSamw DMU_OT_ACL, /* ACL */ 178da6c28aaSamw DMU_OT_SYSACL, /* SYSACL */ 179da6c28aaSamw DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 180da6c28aaSamw DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 181088f3894Sahrens DMU_OT_NEXT_CLONES, /* ZAP */ 1823f9d6ad7SLin Ling DMU_OT_SCAN_QUEUE, /* ZAP */ 18314843421SMatthew Ahrens DMU_OT_USERGROUP_USED, /* ZAP */ 18414843421SMatthew Ahrens DMU_OT_USERGROUP_QUOTA, /* ZAP */ 185842727c2SChris Kirby DMU_OT_USERREFS, /* ZAP */ 186b24ab676SJeff Bonwick DMU_OT_DDT_ZAP, /* ZAP */ 187b24ab676SJeff Bonwick DMU_OT_DDT_STATS, /* ZAP */ 1880a586ceaSMark Shellenbaum DMU_OT_SA, /* System attr */ 1890a586ceaSMark Shellenbaum DMU_OT_SA_MASTER_NODE, /* ZAP */ 1900a586ceaSMark Shellenbaum DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1910a586ceaSMark Shellenbaum DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1923f9d6ad7SLin Ling DMU_OT_SCAN_XLATE, /* ZAP */ 1933f9d6ad7SLin Ling DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 194cde58dbcSMatthew Ahrens DMU_OT_DEADLIST, /* ZAP */ 195cde58dbcSMatthew Ahrens DMU_OT_DEADLIST_HDR, /* UINT64 */ 196cde58dbcSMatthew Ahrens DMU_OT_DSL_CLONES, /* ZAP */ 197cde58dbcSMatthew Ahrens DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 198ad135b5dSChristopher Siden /* 199ad135b5dSChristopher Siden * Do not allocate new object types here. Doing so makes the on-disk 200ad135b5dSChristopher Siden * format incompatible with any other format that uses the same object 201ad135b5dSChristopher Siden * type number. 202ad135b5dSChristopher Siden * 203ad135b5dSChristopher Siden * When creating an object which does not have one of the above types 204ad135b5dSChristopher Siden * use the DMU_OTN_* type with the correct byteswap and metadata 205ad135b5dSChristopher Siden * values. 206ad135b5dSChristopher Siden * 207ad135b5dSChristopher Siden * The DMU_OTN_* types do not have entries in the dmu_ot table, 208ad135b5dSChristopher Siden * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead 209ad135b5dSChristopher Siden * of indexing into dmu_ot directly (this works for both DMU_OT_* types 210ad135b5dSChristopher Siden * and DMU_OTN_* types). 211ad135b5dSChristopher Siden */ 212ad135b5dSChristopher Siden DMU_OT_NUMTYPES, 213ad135b5dSChristopher Siden 214ad135b5dSChristopher Siden /* 215ad135b5dSChristopher Siden * Names for valid types declared with DMU_OT(). 216ad135b5dSChristopher Siden */ 217ad135b5dSChristopher Siden DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 218ad135b5dSChristopher Siden DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 219ad135b5dSChristopher Siden DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 220ad135b5dSChristopher Siden DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 221ad135b5dSChristopher Siden DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 222ad135b5dSChristopher Siden DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 223ad135b5dSChristopher Siden DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 224ad135b5dSChristopher Siden DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 225ad135b5dSChristopher Siden DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 226ad135b5dSChristopher Siden DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE), 227fa9e4066Sahrens } dmu_object_type_t; 228fa9e4066Sahrens 2293b2aab18SMatthew Ahrens typedef enum txg_how { 2303b2aab18SMatthew Ahrens TXG_WAIT = 1, 2313b2aab18SMatthew Ahrens TXG_NOWAIT, 23269962b56SMatthew Ahrens TXG_WAITED, 2333b2aab18SMatthew Ahrens } txg_how_t; 2343b2aab18SMatthew Ahrens 235fa9e4066Sahrens void byteswap_uint64_array(void *buf, size_t size); 236fa9e4066Sahrens void byteswap_uint32_array(void *buf, size_t size); 237fa9e4066Sahrens void byteswap_uint16_array(void *buf, size_t size); 238fa9e4066Sahrens void byteswap_uint8_array(void *buf, size_t size); 239fa9e4066Sahrens void zap_byteswap(void *buf, size_t size); 240da6c28aaSamw void zfs_oldacl_byteswap(void *buf, size_t size); 241fa9e4066Sahrens void zfs_acl_byteswap(void *buf, size_t size); 242fa9e4066Sahrens void zfs_znode_byteswap(void *buf, size_t size); 243fa9e4066Sahrens 2440b69c2f0Sahrens #define DS_FIND_SNAPSHOTS (1<<0) 2450b69c2f0Sahrens #define DS_FIND_CHILDREN (1<<1) 246*12380e1eSArne Jansen #define DS_FIND_SERIALIZE (1<<2) 247fa9e4066Sahrens 248fa9e4066Sahrens /* 249fa9e4066Sahrens * The maximum number of bytes that can be accessed as part of one 250fa9e4066Sahrens * operation, including metadata. 251fa9e4066Sahrens */ 252b5152584SMatthew Ahrens #define DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */ 253cdb0ab79Smaybee #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ 254fa9e4066Sahrens 25514843421SMatthew Ahrens #define DMU_USERUSED_OBJECT (-1ULL) 25614843421SMatthew Ahrens #define DMU_GROUPUSED_OBJECT (-2ULL) 25714843421SMatthew Ahrens 2580a586ceaSMark Shellenbaum /* 2590a586ceaSMark Shellenbaum * artificial blkids for bonus buffer and spill blocks 2600a586ceaSMark Shellenbaum */ 2610a586ceaSMark Shellenbaum #define DMU_BONUS_BLKID (-1ULL) 2620a586ceaSMark Shellenbaum #define DMU_SPILL_BLKID (-2ULL) 263fa9e4066Sahrens /* 264fa9e4066Sahrens * Public routines to create, destroy, open, and close objsets. 265fa9e4066Sahrens */ 266503ad85cSMatthew Ahrens int dmu_objset_hold(const char *name, void *tag, objset_t **osp); 267503ad85cSMatthew Ahrens int dmu_objset_own(const char *name, dmu_objset_type_t type, 268503ad85cSMatthew Ahrens boolean_t readonly, void *tag, objset_t **osp); 269503ad85cSMatthew Ahrens void dmu_objset_rele(objset_t *os, void *tag); 270503ad85cSMatthew Ahrens void dmu_objset_disown(objset_t *os, void *tag); 271503ad85cSMatthew Ahrens int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 272503ad85cSMatthew Ahrens 2733b2aab18SMatthew Ahrens void dmu_objset_evict_dbufs(objset_t *os); 274ae46e4c7SMatthew Ahrens int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 275ecd6cf80Smarks void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); 2763b2aab18SMatthew Ahrens int dmu_objset_clone(const char *name, const char *origin); 2773b2aab18SMatthew Ahrens int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, 2784445fffbSMatthew Ahrens struct nvlist *errlist); 2794445fffbSMatthew Ahrens int dmu_objset_snapshot_one(const char *fsname, const char *snapname); 2804445fffbSMatthew Ahrens int dmu_objset_snapshot_tmp(const char *, const char *, int); 281fd136879SMatthew Ahrens int dmu_objset_find(char *name, int func(const char *, void *), void *arg, 282fa9e4066Sahrens int flags); 283fa9e4066Sahrens void dmu_objset_byteswap(void *buf, size_t size); 2843b2aab18SMatthew Ahrens int dsl_dataset_rename_snapshot(const char *fsname, 2853b2aab18SMatthew Ahrens const char *oldsnapname, const char *newsnapname, boolean_t recursive); 286fa9e4066Sahrens 287fa9e4066Sahrens typedef struct dmu_buf { 288fa9e4066Sahrens uint64_t db_object; /* object that this buffer is part of */ 289fa9e4066Sahrens uint64_t db_offset; /* byte offset in this object */ 290fa9e4066Sahrens uint64_t db_size; /* size of buffer in bytes */ 291fa9e4066Sahrens void *db_data; /* data in buffer */ 292fa9e4066Sahrens } dmu_buf_t; 293fa9e4066Sahrens 29499653d4eSeschrock /* 29599653d4eSeschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 29699653d4eSeschrock */ 297fa9e4066Sahrens #define DMU_POOL_DIRECTORY_OBJECT 1 298fa9e4066Sahrens #define DMU_POOL_CONFIG "config" 299ad135b5dSChristopher Siden #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 300ad135b5dSChristopher Siden #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 301ad135b5dSChristopher Siden #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 30243466aaeSMax Grossman #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" 303fa9e4066Sahrens #define DMU_POOL_ROOT_DATASET "root_dataset" 304cde58dbcSMatthew Ahrens #define DMU_POOL_SYNC_BPOBJ "sync_bplist" 305ea8dc4b6Seschrock #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 306ea8dc4b6Seschrock #define DMU_POOL_ERRLOG_LAST "errlog_last" 30799653d4eSeschrock #define DMU_POOL_SPARES "spares" 30899653d4eSeschrock #define DMU_POOL_DEFLATE "deflate" 30906eeb2adSek #define DMU_POOL_HISTORY "history" 310b1b8ab34Slling #define DMU_POOL_PROPS "pool_props" 311fa94a07fSbrendan #define DMU_POOL_L2CACHE "l2cache" 312ca45db41SChris Kirby #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 313b24ab676SJeff Bonwick #define DMU_POOL_DDT "DDT-%s-%s-%s" 314b24ab676SJeff Bonwick #define DMU_POOL_DDT_STATS "DDT-statistics" 3153f9d6ad7SLin Ling #define DMU_POOL_CREATION_VERSION "creation_version" 3163f9d6ad7SLin Ling #define DMU_POOL_SCAN "scan" 317cde58dbcSMatthew Ahrens #define DMU_POOL_FREE_BPOBJ "free_bpobj" 318ad135b5dSChristopher Siden #define DMU_POOL_BPTREE_OBJ "bptree_obj" 319f1745736SMatthew Ahrens #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 320088f3894Sahrens 321fa9e4066Sahrens /* 322fa9e4066Sahrens * Allocate an object from this objset. The range of object numbers 323fa9e4066Sahrens * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 324fa9e4066Sahrens * 325fa9e4066Sahrens * The transaction must be assigned to a txg. The newly allocated 326fa9e4066Sahrens * object will be "held" in the transaction (ie. you can modify the 327fa9e4066Sahrens * newly allocated object in this transaction). 328fa9e4066Sahrens * 329fa9e4066Sahrens * dmu_object_alloc() chooses an object and returns it in *objectp. 330fa9e4066Sahrens * 331fa9e4066Sahrens * dmu_object_claim() allocates a specific object number. If that 332fa9e4066Sahrens * number is already allocated, it fails and returns EEXIST. 333fa9e4066Sahrens * 334fa9e4066Sahrens * Return 0 on success, or ENOSPC or EEXIST as specified above. 335fa9e4066Sahrens */ 336fa9e4066Sahrens uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 337fa9e4066Sahrens int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 338fa9e4066Sahrens int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 339fa9e4066Sahrens int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 340fa9e4066Sahrens int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 341e77d42eaSMatthew Ahrens int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); 342fa9e4066Sahrens 343fa9e4066Sahrens /* 344fa9e4066Sahrens * Free an object from this objset. 345fa9e4066Sahrens * 346fa9e4066Sahrens * The object's data will be freed as well (ie. you don't need to call 347fa9e4066Sahrens * dmu_free(object, 0, -1, tx)). 348fa9e4066Sahrens * 349fa9e4066Sahrens * The object need not be held in the transaction. 350fa9e4066Sahrens * 351fa9e4066Sahrens * If there are any holds on this object's buffers (via dmu_buf_hold()), 352fa9e4066Sahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not 353fa9e4066Sahrens * free it; it fails and returns EBUSY. 354fa9e4066Sahrens * 355fa9e4066Sahrens * If the object is not allocated, it fails and returns ENOENT. 356fa9e4066Sahrens * 357fa9e4066Sahrens * Return 0 on success, or EBUSY or ENOENT as specified above. 358fa9e4066Sahrens */ 359fa9e4066Sahrens int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 360fa9e4066Sahrens 361fa9e4066Sahrens /* 362fa9e4066Sahrens * Find the next allocated or free object. 363fa9e4066Sahrens * 364fa9e4066Sahrens * The objectp parameter is in-out. It will be updated to be the next 3656754306eSahrens * object which is allocated. Ignore objects which have not been 3666754306eSahrens * modified since txg. 367fa9e4066Sahrens * 368fa9e4066Sahrens * XXX Can only be called on a objset with no dirty data. 369fa9e4066Sahrens * 370fa9e4066Sahrens * Returns 0 on success, or ENOENT if there are no more objects. 371fa9e4066Sahrens */ 3726754306eSahrens int dmu_object_next(objset_t *os, uint64_t *objectp, 3736754306eSahrens boolean_t hole, uint64_t txg); 374fa9e4066Sahrens 375fa9e4066Sahrens /* 376fa9e4066Sahrens * Set the data blocksize for an object. 377fa9e4066Sahrens * 378fa9e4066Sahrens * The object cannot have any blocks allcated beyond the first. If 379fa9e4066Sahrens * the first block is allocated already, the new size must be greater 380fa9e4066Sahrens * than the current block size. If these conditions are not met, 381fa9e4066Sahrens * ENOTSUP will be returned. 382fa9e4066Sahrens * 383fa9e4066Sahrens * Returns 0 on success, or EBUSY if there are any holds on the object 384fa9e4066Sahrens * contents, or ENOTSUP as described above. 385fa9e4066Sahrens */ 386fa9e4066Sahrens int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 387fa9e4066Sahrens int ibs, dmu_tx_t *tx); 388fa9e4066Sahrens 389fa9e4066Sahrens /* 390fa9e4066Sahrens * Set the checksum property on a dnode. The new checksum algorithm will 391fa9e4066Sahrens * apply to all newly written blocks; existing blocks will not be affected. 392fa9e4066Sahrens */ 393fa9e4066Sahrens void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 394fa9e4066Sahrens dmu_tx_t *tx); 395fa9e4066Sahrens 396fa9e4066Sahrens /* 397fa9e4066Sahrens * Set the compress property on a dnode. The new compression algorithm will 398fa9e4066Sahrens * apply to all newly written blocks; existing blocks will not be affected. 399fa9e4066Sahrens */ 400fa9e4066Sahrens void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 401fa9e4066Sahrens dmu_tx_t *tx); 402fa9e4066Sahrens 4035d7b4d43SMatthew Ahrens void 4045d7b4d43SMatthew Ahrens dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, 4055d7b4d43SMatthew Ahrens void *data, uint8_t etype, uint8_t comp, int uncompressed_size, 4065d7b4d43SMatthew Ahrens int compressed_size, int byteorder, dmu_tx_t *tx); 4075d7b4d43SMatthew Ahrens 40844cd46caSbillm /* 409b24ab676SJeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc. 41044cd46caSbillm */ 411b24ab676SJeff Bonwick #define WP_NOFILL 0x1 412b24ab676SJeff Bonwick #define WP_DMU_SYNC 0x2 4130a586ceaSMark Shellenbaum #define WP_SPILL 0x4 414b24ab676SJeff Bonwick 415b24ab676SJeff Bonwick void dmu_write_policy(objset_t *os, struct dnode *dn, int level, int wp, 416b24ab676SJeff Bonwick struct zio_prop *zp); 417fa9e4066Sahrens /* 418fa9e4066Sahrens * The bonus data is accessed more or less like a regular buffer. 419fa9e4066Sahrens * You must dmu_bonus_hold() to get the buffer, which will give you a 420fa9e4066Sahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 421fa9e4066Sahrens * data. As with any normal buffer, you must call dmu_buf_read() to 422fa9e4066Sahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the 423fa9e4066Sahrens * object must be held in an assigned transaction before calling 424fa9e4066Sahrens * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 425fa9e4066Sahrens * buffer as well. You must release your hold with dmu_buf_rele(). 4263e30c24aSWill Andrews * 4273e30c24aSWill Andrews * Returns ENOENT, EIO, or 0. 428fa9e4066Sahrens */ 429ea8dc4b6Seschrock int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); 430fa9e4066Sahrens int dmu_bonus_max(void); 4311934e92fSmaybee int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 4320a586ceaSMark Shellenbaum int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 433744947dcSTom Erickson dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); 4340a586ceaSMark Shellenbaum int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 4350a586ceaSMark Shellenbaum 4360a586ceaSMark Shellenbaum /* 4370a586ceaSMark Shellenbaum * Special spill buffer support used by "SA" framework 4380a586ceaSMark Shellenbaum */ 4390a586ceaSMark Shellenbaum 4400a586ceaSMark Shellenbaum int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 4410a586ceaSMark Shellenbaum int dmu_spill_hold_by_dnode(struct dnode *dn, uint32_t flags, 4420a586ceaSMark Shellenbaum void *tag, dmu_buf_t **dbp); 4430a586ceaSMark Shellenbaum int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 444fa9e4066Sahrens 445fa9e4066Sahrens /* 446fa9e4066Sahrens * Obtain the DMU buffer from the specified object which contains the 447fa9e4066Sahrens * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 448fa9e4066Sahrens * that it will remain in memory. You must release the hold with 449fa9e4066Sahrens * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your 450fa9e4066Sahrens * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 451fa9e4066Sahrens * 452fa9e4066Sahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 453fa9e4066Sahrens * on the returned buffer before reading or writing the buffer's 454fa9e4066Sahrens * db_data. The comments for those routines describe what particular 455fa9e4066Sahrens * operations are valid after calling them. 456fa9e4066Sahrens * 457fa9e4066Sahrens * The object number must be a valid, allocated object number. 458fa9e4066Sahrens */ 459ea8dc4b6Seschrock int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 46047cb52daSJeff Bonwick void *tag, dmu_buf_t **, int flags); 461e57a022bSJustin T. Gibbs 462e57a022bSJustin T. Gibbs /* 463e57a022bSJustin T. Gibbs * Add a reference to a dmu buffer that has already been held via 464e57a022bSJustin T. Gibbs * dmu_buf_hold() in the current context. 465e57a022bSJustin T. Gibbs */ 466fa9e4066Sahrens void dmu_buf_add_ref(dmu_buf_t *db, void* tag); 467e57a022bSJustin T. Gibbs 468e57a022bSJustin T. Gibbs /* 469e57a022bSJustin T. Gibbs * Attempt to add a reference to a dmu buffer that is in an unknown state, 470e57a022bSJustin T. Gibbs * using a pointer that may have been invalidated by eviction processing. 471e57a022bSJustin T. Gibbs * The request will succeed if the passed in dbuf still represents the 472e57a022bSJustin T. Gibbs * same os/object/blkid, is ineligible for eviction, and has at least 473e57a022bSJustin T. Gibbs * one hold by a user other than the syncer. 474e57a022bSJustin T. Gibbs */ 475e57a022bSJustin T. Gibbs boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, 476e57a022bSJustin T. Gibbs uint64_t blkid, void *tag); 477e57a022bSJustin T. Gibbs 478ea8dc4b6Seschrock void dmu_buf_rele(dmu_buf_t *db, void *tag); 479fa9e4066Sahrens uint64_t dmu_buf_refcount(dmu_buf_t *db); 480fa9e4066Sahrens 481fa9e4066Sahrens /* 482fa9e4066Sahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 483fa9e4066Sahrens * range of an object. A pointer to an array of dmu_buf_t*'s is 484fa9e4066Sahrens * returned (in *dbpp). 485fa9e4066Sahrens * 486fa9e4066Sahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 487fa9e4066Sahrens * frees the array. The hold on the array of buffers MUST be released 488fa9e4066Sahrens * with dmu_buf_rele_array. You can NOT release the hold on each buffer 489fa9e4066Sahrens * individually with dmu_buf_rele. 490fa9e4066Sahrens */ 49113506d1eSmaybee int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 49213506d1eSmaybee uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp); 493ea8dc4b6Seschrock void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); 494fa9e4066Sahrens 495bc9014e6SJustin Gibbs typedef void dmu_buf_evict_func_t(void *user_ptr); 496bc9014e6SJustin Gibbs 497fa9e4066Sahrens /* 498bc9014e6SJustin Gibbs * A DMU buffer user object may be associated with a dbuf for the 499bc9014e6SJustin Gibbs * duration of its lifetime. This allows the user of a dbuf (client) 500bc9014e6SJustin Gibbs * to attach private data to a dbuf (e.g. in-core only data such as a 501bc9014e6SJustin Gibbs * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified 502bc9014e6SJustin Gibbs * when that dbuf has been evicted. Clients typically respond to the 503bc9014e6SJustin Gibbs * eviction notification by freeing their private data, thus ensuring 504bc9014e6SJustin Gibbs * the same lifetime for both dbuf and private data. 505fa9e4066Sahrens * 506bc9014e6SJustin Gibbs * The mapping from a dmu_buf_user_t to any client private data is the 507bc9014e6SJustin Gibbs * client's responsibility. All current consumers of the API with private 508bc9014e6SJustin Gibbs * data embed a dmu_buf_user_t as the first member of the structure for 509bc9014e6SJustin Gibbs * their private data. This allows conversions between the two types 510bc9014e6SJustin Gibbs * with a simple cast. Since the DMU buf user API never needs access 511bc9014e6SJustin Gibbs * to the private data, other strategies can be employed if necessary 512bc9014e6SJustin Gibbs * or convenient for the client (e.g. using container_of() to do the 513bc9014e6SJustin Gibbs * conversion for private data that cannot have the dmu_buf_user_t as 514bc9014e6SJustin Gibbs * its first member). 515fa9e4066Sahrens * 516bc9014e6SJustin Gibbs * Eviction callbacks are executed without the dbuf mutex held or any 517bc9014e6SJustin Gibbs * other type of mechanism to guarantee that the dbuf is still available. 518bc9014e6SJustin Gibbs * For this reason, users must assume the dbuf has already been freed 519bc9014e6SJustin Gibbs * and not reference the dbuf from the callback context. 520fa9e4066Sahrens * 521bc9014e6SJustin Gibbs * Users requesting "immediate eviction" are notified as soon as the dbuf 522bc9014e6SJustin Gibbs * is only referenced by dirty records (dirties == holds). Otherwise the 523bc9014e6SJustin Gibbs * notification occurs after eviction processing for the dbuf begins. 524fa9e4066Sahrens */ 525bc9014e6SJustin Gibbs typedef struct dmu_buf_user { 526bc9014e6SJustin Gibbs /* 527bc9014e6SJustin Gibbs * Asynchronous user eviction callback state. 528bc9014e6SJustin Gibbs */ 529bc9014e6SJustin Gibbs taskq_ent_t dbu_tqent; 530bc9014e6SJustin Gibbs 531bc9014e6SJustin Gibbs /* This instance's eviction function pointer. */ 532bc9014e6SJustin Gibbs dmu_buf_evict_func_t *dbu_evict_func; 533bc9014e6SJustin Gibbs #ifdef ZFS_DEBUG 534bc9014e6SJustin Gibbs /* 535bc9014e6SJustin Gibbs * Pointer to user's dbuf pointer. NULL for clients that do 536bc9014e6SJustin Gibbs * not associate a dbuf with their user data. 537bc9014e6SJustin Gibbs * 538bc9014e6SJustin Gibbs * The dbuf pointer is cleared upon eviction so as to catch 539bc9014e6SJustin Gibbs * use-after-evict bugs in clients. 540bc9014e6SJustin Gibbs */ 541bc9014e6SJustin Gibbs dmu_buf_t **dbu_clear_on_evict_dbufp; 542bc9014e6SJustin Gibbs #endif 543bc9014e6SJustin Gibbs } dmu_buf_user_t; 544bc9014e6SJustin Gibbs 545fa9e4066Sahrens /* 546bc9014e6SJustin Gibbs * Initialize the given dmu_buf_user_t instance with the eviction function 547bc9014e6SJustin Gibbs * evict_func, to be called when the user is evicted. 548bc9014e6SJustin Gibbs * 549bc9014e6SJustin Gibbs * NOTE: This function should only be called once on a given dmu_buf_user_t. 550bc9014e6SJustin Gibbs * To allow enforcement of this, dbu must already be zeroed on entry. 551fa9e4066Sahrens */ 552bc9014e6SJustin Gibbs #ifdef __lint 553bc9014e6SJustin Gibbs /* Very ugly, but it beats issuing suppression directives in many Makefiles. */ 554bc9014e6SJustin Gibbs extern void 555bc9014e6SJustin Gibbs dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func, 556bc9014e6SJustin Gibbs dmu_buf_t **clear_on_evict_dbufp); 557bc9014e6SJustin Gibbs #else /* __lint */ 558bc9014e6SJustin Gibbs inline void 559bc9014e6SJustin Gibbs dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func, 560bc9014e6SJustin Gibbs dmu_buf_t **clear_on_evict_dbufp) 561bc9014e6SJustin Gibbs { 562bc9014e6SJustin Gibbs ASSERT(dbu->dbu_evict_func == NULL); 563bc9014e6SJustin Gibbs ASSERT(evict_func != NULL); 564bc9014e6SJustin Gibbs dbu->dbu_evict_func = evict_func; 565bc9014e6SJustin Gibbs #ifdef ZFS_DEBUG 566bc9014e6SJustin Gibbs dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; 567bc9014e6SJustin Gibbs #endif 568bc9014e6SJustin Gibbs } 569bc9014e6SJustin Gibbs #endif /* __lint */ 570fa9e4066Sahrens 571fa9e4066Sahrens /* 572bc9014e6SJustin Gibbs * Attach user data to a dbuf and mark it for normal (when the dbuf's 573bc9014e6SJustin Gibbs * data is cleared or its reference count goes to zero) eviction processing. 574bc9014e6SJustin Gibbs * 575bc9014e6SJustin Gibbs * Returns NULL on success, or the existing user if another user currently 576bc9014e6SJustin Gibbs * owns the buffer. 577bc9014e6SJustin Gibbs */ 578bc9014e6SJustin Gibbs void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); 579bc9014e6SJustin Gibbs 580bc9014e6SJustin Gibbs /* 581bc9014e6SJustin Gibbs * Attach user data to a dbuf and mark it for immediate (its dirty and 582bc9014e6SJustin Gibbs * reference counts are equal) eviction processing. 583bc9014e6SJustin Gibbs * 584bc9014e6SJustin Gibbs * Returns NULL on success, or the existing user if another user currently 585bc9014e6SJustin Gibbs * owns the buffer. 586bc9014e6SJustin Gibbs */ 587bc9014e6SJustin Gibbs void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); 588bc9014e6SJustin Gibbs 589bc9014e6SJustin Gibbs /* 590bc9014e6SJustin Gibbs * Replace the current user of a dbuf. 591bc9014e6SJustin Gibbs * 592bc9014e6SJustin Gibbs * If given the current user of a dbuf, replaces the dbuf's user with 593bc9014e6SJustin Gibbs * "new_user" and returns the user data pointer that was replaced. 594bc9014e6SJustin Gibbs * Otherwise returns the current, and unmodified, dbuf user pointer. 595bc9014e6SJustin Gibbs */ 596bc9014e6SJustin Gibbs void *dmu_buf_replace_user(dmu_buf_t *db, 597bc9014e6SJustin Gibbs dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); 598bc9014e6SJustin Gibbs 599bc9014e6SJustin Gibbs /* 600bc9014e6SJustin Gibbs * Remove the specified user data for a DMU buffer. 601bc9014e6SJustin Gibbs * 602bc9014e6SJustin Gibbs * Returns the user that was removed on success, or the current user if 603bc9014e6SJustin Gibbs * another user currently owns the buffer. 604bc9014e6SJustin Gibbs */ 605bc9014e6SJustin Gibbs void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); 606bc9014e6SJustin Gibbs 607bc9014e6SJustin Gibbs /* 608bc9014e6SJustin Gibbs * Returns the user data (dmu_buf_user_t *) associated with this dbuf. 609fa9e4066Sahrens */ 610fa9e4066Sahrens void *dmu_buf_get_user(dmu_buf_t *db); 611fa9e4066Sahrens 612bc9014e6SJustin Gibbs /* Block until any in-progress dmu buf user evictions complete. */ 613bc9014e6SJustin Gibbs void dmu_buf_user_evict_wait(void); 614bc9014e6SJustin Gibbs 61580901aeaSGeorge Wilson /* 61680901aeaSGeorge Wilson * Returns the blkptr associated with this dbuf, or NULL if not set. 61780901aeaSGeorge Wilson */ 61880901aeaSGeorge Wilson struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); 61980901aeaSGeorge Wilson 620fa9e4066Sahrens /* 621fa9e4066Sahrens * Indicate that you are going to modify the buffer's data (db_data). 622fa9e4066Sahrens * 623fa9e4066Sahrens * The transaction (tx) must be assigned to a txg (ie. you've called 624fa9e4066Sahrens * dmu_tx_assign()). The buffer's object must be held in the tx 625fa9e4066Sahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 626fa9e4066Sahrens */ 627fa9e4066Sahrens void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 628fa9e4066Sahrens 6293d692628SSanjeev Bagewadi /* 6303d692628SSanjeev Bagewadi * Tells if the given dbuf is freeable. 6313d692628SSanjeev Bagewadi */ 6323d692628SSanjeev Bagewadi boolean_t dmu_buf_freeable(dmu_buf_t *); 6333d692628SSanjeev Bagewadi 634fa9e4066Sahrens /* 635fa9e4066Sahrens * You must create a transaction, then hold the objects which you will 636fa9e4066Sahrens * (or might) modify as part of this transaction. Then you must assign 637fa9e4066Sahrens * the transaction to a transaction group. Once the transaction has 638fa9e4066Sahrens * been assigned, you can modify buffers which belong to held objects as 639fa9e4066Sahrens * part of this transaction. You can't modify buffers before the 640fa9e4066Sahrens * transaction has been assigned; you can't modify buffers which don't 641fa9e4066Sahrens * belong to objects which this transaction holds; you can't hold 642fa9e4066Sahrens * objects once the transaction has been assigned. You may hold an 643fa9e4066Sahrens * object which you are going to free (with dmu_object_free()), but you 644fa9e4066Sahrens * don't have to. 645fa9e4066Sahrens * 646fa9e4066Sahrens * You can abort the transaction before it has been assigned. 647fa9e4066Sahrens * 648fa9e4066Sahrens * Note that you may hold buffers (with dmu_buf_hold) at any time, 649fa9e4066Sahrens * regardless of transaction state. 650fa9e4066Sahrens */ 651fa9e4066Sahrens 652fa9e4066Sahrens #define DMU_NEW_OBJECT (-1ULL) 653fa9e4066Sahrens #define DMU_OBJECT_END (-1ULL) 654fa9e4066Sahrens 655fa9e4066Sahrens dmu_tx_t *dmu_tx_create(objset_t *os); 656fa9e4066Sahrens void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); 657fa9e4066Sahrens void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 658fa9e4066Sahrens uint64_t len); 65914843421SMatthew Ahrens void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 660fa9e4066Sahrens void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 6610a586ceaSMark Shellenbaum void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 6620a586ceaSMark Shellenbaum void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 6630a586ceaSMark Shellenbaum void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 664fa9e4066Sahrens void dmu_tx_abort(dmu_tx_t *tx); 6653b2aab18SMatthew Ahrens int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how); 6668a2f1b91Sahrens void dmu_tx_wait(dmu_tx_t *tx); 667fa9e4066Sahrens void dmu_tx_commit(dmu_tx_t *tx); 6684bb73804SMatthew Ahrens void dmu_tx_mark_netfree(dmu_tx_t *tx); 669fa9e4066Sahrens 670d20e665cSRicardo M. Correia /* 671d20e665cSRicardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called. 672d20e665cSRicardo M. Correia * 673d20e665cSRicardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a 674d20e665cSRicardo M. Correia * callback parameter. The caller is responsible for properly allocating and 675d20e665cSRicardo M. Correia * freeing it. 676d20e665cSRicardo M. Correia * 677d20e665cSRicardo M. Correia * When registering a callback, the transaction must be already created, but 678d20e665cSRicardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not. 679d20e665cSRicardo M. Correia * 680d20e665cSRicardo M. Correia * The callback will be called after the transaction has been safely written 681d20e665cSRicardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted. 682d20e665cSRicardo M. Correia * If there is any error which prevents the transaction from being committed to 683d20e665cSRicardo M. Correia * disk, the callback will be called with a value of error != 0. 684d20e665cSRicardo M. Correia */ 685d20e665cSRicardo M. Correia typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 686d20e665cSRicardo M. Correia 687d20e665cSRicardo M. Correia void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 688d20e665cSRicardo M. Correia void *dcb_data); 689d20e665cSRicardo M. Correia 690fa9e4066Sahrens /* 691fa9e4066Sahrens * Free up the data blocks for a defined range of a file. If size is 692ad135b5dSChristopher Siden * -1, the range from offset to end-of-file is freed. 693fa9e4066Sahrens */ 694ea8dc4b6Seschrock int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 695fa9e4066Sahrens uint64_t size, dmu_tx_t *tx); 696cdb0ab79Smaybee int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 697cdb0ab79Smaybee uint64_t size); 698713d6c20SMatthew Ahrens int dmu_free_long_object(objset_t *os, uint64_t object); 699fa9e4066Sahrens 700fa9e4066Sahrens /* 701fa9e4066Sahrens * Convenience functions. 702fa9e4066Sahrens * 703fa9e4066Sahrens * Canfail routines will return 0 on success, or an errno if there is a 704fa9e4066Sahrens * nonrecoverable I/O error. 705fa9e4066Sahrens */ 7067bfdf011SNeil Perrin #define DMU_READ_PREFETCH 0 /* prefetch */ 7077bfdf011SNeil Perrin #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 708ea8dc4b6Seschrock int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 7097bfdf011SNeil Perrin void *buf, uint32_t flags); 710fa9e4066Sahrens void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 711fa9e4066Sahrens const void *buf, dmu_tx_t *tx); 71282c9918fSTim Haley void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 71382c9918fSTim Haley dmu_tx_t *tx); 714feb08c6bSbillm int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); 715f8554bb9SMatthew Ahrens int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size); 716feb08c6bSbillm int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, 717feb08c6bSbillm dmu_tx_t *tx); 71894d1a210STim Haley int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, 71994d1a210STim Haley dmu_tx_t *tx); 72044eda4d7Smaybee int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, 72144eda4d7Smaybee uint64_t size, struct page *pp, dmu_tx_t *tx); 7222fdbea25SAleksandr Guzovskiy struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 7232fdbea25SAleksandr Guzovskiy void dmu_return_arcbuf(struct arc_buf *buf); 7242fdbea25SAleksandr Guzovskiy void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, 7252fdbea25SAleksandr Guzovskiy dmu_tx_t *tx); 726c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_init(struct xuio *uio, int niov); 727c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void dmu_xuio_fini(struct xuio *uio); 728c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, 729c242f9a0Schunli zhang - Sun Microsystems - Irvine United States size_t n); 730c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_cnt(struct xuio *uio); 731c242f9a0Schunli zhang - Sun Microsystems - Irvine United States struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); 732c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void dmu_xuio_clear(struct xuio *uio, int i); 733c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void xuio_stat_wbuf_copied(); 734c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void xuio_stat_wbuf_nocopy(); 735fa9e4066Sahrens 736416e0cd8Sek extern int zfs_prefetch_disable; 737b5152584SMatthew Ahrens extern int zfs_max_recordsize; 738416e0cd8Sek 739fa9e4066Sahrens /* 740fa9e4066Sahrens * Asynchronously try to read in the data. 741fa9e4066Sahrens */ 742fa9e4066Sahrens void dmu_prefetch(objset_t *os, uint64_t object, uint64_t offset, 743fa9e4066Sahrens uint64_t len); 744fa9e4066Sahrens 745fa9e4066Sahrens typedef struct dmu_object_info { 746b24ab676SJeff Bonwick /* All sizes are in bytes unless otherwise indicated. */ 747fa9e4066Sahrens uint32_t doi_data_block_size; 748fa9e4066Sahrens uint32_t doi_metadata_block_size; 749fa9e4066Sahrens dmu_object_type_t doi_type; 750fa9e4066Sahrens dmu_object_type_t doi_bonus_type; 751b24ab676SJeff Bonwick uint64_t doi_bonus_size; 752fa9e4066Sahrens uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 753fa9e4066Sahrens uint8_t doi_checksum; 754fa9e4066Sahrens uint8_t doi_compress; 755e77d42eaSMatthew Ahrens uint8_t doi_nblkptr; 756e77d42eaSMatthew Ahrens uint8_t doi_pad[4]; 757b24ab676SJeff Bonwick uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 758b24ab676SJeff Bonwick uint64_t doi_max_offset; 759b24ab676SJeff Bonwick uint64_t doi_fill_count; /* number of non-empty blocks */ 760fa9e4066Sahrens } dmu_object_info_t; 761fa9e4066Sahrens 762e45ce728Sahrens typedef void arc_byteswap_func_t(void *buf, size_t size); 763e45ce728Sahrens 764fa9e4066Sahrens typedef struct dmu_object_type_info { 765ad135b5dSChristopher Siden dmu_object_byteswap_t ot_byteswap; 766fa9e4066Sahrens boolean_t ot_metadata; 767fa9e4066Sahrens char *ot_name; 768fa9e4066Sahrens } dmu_object_type_info_t; 769fa9e4066Sahrens 770ad135b5dSChristopher Siden typedef struct dmu_object_byteswap_info { 771ad135b5dSChristopher Siden arc_byteswap_func_t *ob_func; 772ad135b5dSChristopher Siden char *ob_name; 773ad135b5dSChristopher Siden } dmu_object_byteswap_info_t; 774ad135b5dSChristopher Siden 775fa9e4066Sahrens extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 776ad135b5dSChristopher Siden extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; 777fa9e4066Sahrens 778fa9e4066Sahrens /* 779fa9e4066Sahrens * Get information on a DMU object. 780fa9e4066Sahrens * 781fa9e4066Sahrens * Return 0 on success or ENOENT if object is not allocated. 782fa9e4066Sahrens * 783fa9e4066Sahrens * If doi is NULL, just indicates whether the object exists. 784fa9e4066Sahrens */ 785fa9e4066Sahrens int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 7863e30c24aSWill Andrews /* Like dmu_object_info, but faster if you have a held dnode in hand. */ 787fa9e4066Sahrens void dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); 7883e30c24aSWill Andrews /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ 789fa9e4066Sahrens void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 7903e30c24aSWill Andrews /* 7913e30c24aSWill Andrews * Like dmu_object_info_from_db, but faster still when you only care about 7923e30c24aSWill Andrews * the size. This is specifically optimized for zfs_getattr(). 7933e30c24aSWill Andrews */ 794fa9e4066Sahrens void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 795fa9e4066Sahrens u_longlong_t *nblk512); 796fa9e4066Sahrens 797fa9e4066Sahrens typedef struct dmu_objset_stats { 798a2eea2e1Sahrens uint64_t dds_num_clones; /* number of clones of this */ 799a2eea2e1Sahrens uint64_t dds_creation_txg; 8003cb34c60Sahrens uint64_t dds_guid; 801fa9e4066Sahrens dmu_objset_type_t dds_type; 802fa9e4066Sahrens uint8_t dds_is_snapshot; 80331fd60d3Sahrens uint8_t dds_inconsistent; 8043cb34c60Sahrens char dds_origin[MAXNAMELEN]; 805a2eea2e1Sahrens } dmu_objset_stats_t; 806fa9e4066Sahrens 807a2eea2e1Sahrens /* 808a2eea2e1Sahrens * Get stats on a dataset. 809a2eea2e1Sahrens */ 810a2eea2e1Sahrens void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 811fa9e4066Sahrens 812a2eea2e1Sahrens /* 813a2eea2e1Sahrens * Add entries to the nvlist for all the objset's properties. See 814a2eea2e1Sahrens * zfs_prop_table[] and zfs(1m) for details on the properties. 815a2eea2e1Sahrens */ 816a2eea2e1Sahrens void dmu_objset_stats(objset_t *os, struct nvlist *nv); 817fa9e4066Sahrens 818a2eea2e1Sahrens /* 819a2eea2e1Sahrens * Get the space usage statistics for statvfs(). 820a2eea2e1Sahrens * 821a2eea2e1Sahrens * refdbytes is the amount of space "referenced" by this objset. 822a2eea2e1Sahrens * availbytes is the amount of space available to this objset, taking 823a2eea2e1Sahrens * into account quotas & reservations, assuming that no other objsets 824a2eea2e1Sahrens * use the space first. These values correspond to the 'referenced' and 825a2eea2e1Sahrens * 'available' properties, described in the zfs(1m) manpage. 826a2eea2e1Sahrens * 827a2eea2e1Sahrens * usedobjs and availobjs are the number of objects currently allocated, 828a2eea2e1Sahrens * and available. 829a2eea2e1Sahrens */ 830a2eea2e1Sahrens void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 831a2eea2e1Sahrens uint64_t *usedobjsp, uint64_t *availobjsp); 832fa9e4066Sahrens 833fa9e4066Sahrens /* 834a2eea2e1Sahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions. 835a2eea2e1Sahrens * (Contrast with the ds_guid which is a 64-bit ID that will never 836a2eea2e1Sahrens * change, so there is a small probability that it will collide.) 837fa9e4066Sahrens */ 838a2eea2e1Sahrens uint64_t dmu_objset_fsid_guid(objset_t *os); 839fa9e4066Sahrens 84071eb0538SChris Kirby /* 84171eb0538SChris Kirby * Get the [cm]time for an objset's snapshot dir 84271eb0538SChris Kirby */ 84371eb0538SChris Kirby timestruc_t dmu_objset_snap_cmtime(objset_t *os); 84471eb0538SChris Kirby 845fa9e4066Sahrens int dmu_objset_is_snapshot(objset_t *os); 846fa9e4066Sahrens 847fa9e4066Sahrens extern struct spa *dmu_objset_spa(objset_t *os); 848fa9e4066Sahrens extern struct zilog *dmu_objset_zil(objset_t *os); 849fa9e4066Sahrens extern struct dsl_pool *dmu_objset_pool(objset_t *os); 850fa9e4066Sahrens extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 851fa9e4066Sahrens extern void dmu_objset_name(objset_t *os, char *buf); 852fa9e4066Sahrens extern dmu_objset_type_t dmu_objset_type(objset_t *os); 853fa9e4066Sahrens extern uint64_t dmu_objset_id(objset_t *os); 854edf345e6SMatthew Ahrens extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); 855edf345e6SMatthew Ahrens extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); 856fa9e4066Sahrens extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, 857b38f0970Sck uint64_t *id, uint64_t *offp, boolean_t *case_conflict); 858ab04eb8eStimh extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, 859ab04eb8eStimh int maxlen, boolean_t *conflict); 86087e5029aSahrens extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, 86187e5029aSahrens uint64_t *idp, uint64_t *offp); 86214843421SMatthew Ahrens 8639966ca11SMatthew Ahrens typedef int objset_used_cb_t(dmu_object_type_t bonustype, 8649966ca11SMatthew Ahrens void *bonus, uint64_t *userp, uint64_t *groupp); 86514843421SMatthew Ahrens extern void dmu_objset_register_type(dmu_objset_type_t ost, 86614843421SMatthew Ahrens objset_used_cb_t *cb); 867f18faf3fSek extern void dmu_objset_set_user(objset_t *os, void *user_ptr); 868f18faf3fSek extern void *dmu_objset_get_user(objset_t *os); 869fa9e4066Sahrens 870fa9e4066Sahrens /* 871fa9e4066Sahrens * Return the txg number for the given assigned transaction. 872fa9e4066Sahrens */ 87387e5029aSahrens uint64_t dmu_tx_get_txg(dmu_tx_t *tx); 874fa9e4066Sahrens 875fa9e4066Sahrens /* 876fa9e4066Sahrens * Synchronous write. 877c5c6ffa0Smaybee * If a parent zio is provided this function initiates a write on the 878c5c6ffa0Smaybee * provided buffer as a child of the parent zio. 879da6c28aaSamw * In the absence of a parent zio, the write is completed synchronously. 880c5c6ffa0Smaybee * At write completion, blk is filled with the bp of the written block. 881fa9e4066Sahrens * Note that while the data covered by this function will be on stable 882c5c6ffa0Smaybee * storage when the write completes this new data does not become a 883fa9e4066Sahrens * permanent part of the file until the associated transaction commits. 884fa9e4066Sahrens */ 885b24ab676SJeff Bonwick 886b24ab676SJeff Bonwick /* 887b24ab676SJeff Bonwick * {zfs,zvol,ztest}_get_done() args 888b24ab676SJeff Bonwick */ 889b24ab676SJeff Bonwick typedef struct zgd { 890b24ab676SJeff Bonwick struct zilog *zgd_zilog; 891b24ab676SJeff Bonwick struct blkptr *zgd_bp; 892b24ab676SJeff Bonwick dmu_buf_t *zgd_db; 893b24ab676SJeff Bonwick struct rl *zgd_rl; 894b24ab676SJeff Bonwick void *zgd_private; 895b24ab676SJeff Bonwick } zgd_t; 896b24ab676SJeff Bonwick 897b24ab676SJeff Bonwick typedef void dmu_sync_cb_t(zgd_t *arg, int error); 898b24ab676SJeff Bonwick int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); 899fa9e4066Sahrens 900fa9e4066Sahrens /* 901fa9e4066Sahrens * Find the next hole or data block in file starting at *off 902fa9e4066Sahrens * Return found offset in *off. Return ESRCH for end of file. 903fa9e4066Sahrens */ 904fa9e4066Sahrens int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, 905fa9e4066Sahrens uint64_t *off); 906fa9e4066Sahrens 9072bcf0248SMax Grossman /* 9082bcf0248SMax Grossman * Check if a DMU object has any dirty blocks. If so, sync out 9092bcf0248SMax Grossman * all pending transaction groups. Otherwise, this function 9102bcf0248SMax Grossman * does not alter DMU state. This could be improved to only sync 9112bcf0248SMax Grossman * out the necessary transaction groups for this particular 9122bcf0248SMax Grossman * object. 9132bcf0248SMax Grossman */ 9142bcf0248SMax Grossman int dmu_object_wait_synced(objset_t *os, uint64_t object); 9152bcf0248SMax Grossman 916fa9e4066Sahrens /* 917fa9e4066Sahrens * Initial setup and final teardown. 918fa9e4066Sahrens */ 919fa9e4066Sahrens extern void dmu_init(void); 920fa9e4066Sahrens extern void dmu_fini(void); 921fa9e4066Sahrens 922fa9e4066Sahrens typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, 923fa9e4066Sahrens uint64_t object, uint64_t offset, int len); 924fa9e4066Sahrens void dmu_traverse_objset(objset_t *os, uint64_t txg_start, 925fa9e4066Sahrens dmu_traverse_cb_t cb, void *arg); 926fa9e4066Sahrens 9273b2aab18SMatthew Ahrens int dmu_diff(const char *tosnap_name, const char *fromsnap_name, 9283b2aab18SMatthew Ahrens struct vnode *vp, offset_t *offp); 92999d5e173STim Haley 930fa9e4066Sahrens /* CRC64 table */ 931fa9e4066Sahrens #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 932fa9e4066Sahrens extern uint64_t zfs_crc64_table[256]; 933fa9e4066Sahrens 934e4be62a2SSaso Kiselkov extern int zfs_mdcomp_disable; 935e4be62a2SSaso Kiselkov 936fa9e4066Sahrens #ifdef __cplusplus 937fa9e4066Sahrens } 938fa9e4066Sahrens #endif 939fa9e4066Sahrens 940fa9e4066Sahrens #endif /* _SYS_DMU_H */ 941