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. 24ae972795SMatthew Ahrens * Copyright (c) 2011, 2016 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. 3045818ee1SMatthew Ahrens * Copyright 2013 Saso Kiselkov. All rights reserved. 31c3d26abcSMatthew Ahrens * Copyright (c) 2014 Integros [integros.com] 32ec5cf9d5SAlexander Stetsenko */ 33fa9e4066Sahrens 3455da60b9SMark J Musante /* Portions Copyright 2010 Robert Milkowski */ 3555da60b9SMark J Musante 36fa9e4066Sahrens #ifndef _SYS_DMU_H 37fa9e4066Sahrens #define _SYS_DMU_H 38fa9e4066Sahrens 39fa9e4066Sahrens /* 40fa9e4066Sahrens * This file describes the interface that the DMU provides for its 41fa9e4066Sahrens * consumers. 42fa9e4066Sahrens * 43fa9e4066Sahrens * The DMU also interacts with the SPA. That interface is described in 44fa9e4066Sahrens * dmu_spa.h. 45fa9e4066Sahrens */ 46fa9e4066Sahrens 47bc9014e6SJustin Gibbs #include <sys/zfs_context.h> 48fa9e4066Sahrens #include <sys/inttypes.h> 49ecd6cf80Smarks #include <sys/cred.h> 504445fffbSMatthew Ahrens #include <sys/fs/zfs.h> 515602294fSDan Kimmel #include <sys/zio_compress.h> 52a2cdcdd2SPaul Dagnelie #include <sys/zio_priority.h> 53fa9e4066Sahrens 54fa9e4066Sahrens #ifdef __cplusplus 55fa9e4066Sahrens extern "C" { 56fa9e4066Sahrens #endif 57fa9e4066Sahrens 58fa9e4066Sahrens struct uio; 59c242f9a0Schunli zhang - Sun Microsystems - Irvine United States struct xuio; 6044eda4d7Smaybee struct page; 61fa9e4066Sahrens struct vnode; 62fa9e4066Sahrens struct spa; 63fa9e4066Sahrens struct zilog; 64fa9e4066Sahrens struct zio; 65fa9e4066Sahrens struct blkptr; 66fa9e4066Sahrens struct zap_cursor; 67fa9e4066Sahrens struct dsl_dataset; 68fa9e4066Sahrens struct dsl_pool; 69fa9e4066Sahrens struct dnode; 70fa9e4066Sahrens struct drr_begin; 71fa9e4066Sahrens struct drr_end; 727802d7bfSMatthew Ahrens struct zbookmark_phys; 7344cd46caSbillm struct spa; 74a2eea2e1Sahrens struct nvlist; 752fdbea25SAleksandr Guzovskiy struct arc_buf; 76b24ab676SJeff Bonwick struct zio_prop; 770a586ceaSMark Shellenbaum struct sa_handle; 78fa9e4066Sahrens 79fa9e4066Sahrens typedef struct objset objset_t; 80fa9e4066Sahrens typedef struct dmu_tx dmu_tx_t; 81fa9e4066Sahrens typedef struct dsl_dir dsl_dir_t; 8279d72832SMatthew Ahrens typedef struct dnode dnode_t; 83fa9e4066Sahrens 84ad135b5dSChristopher Siden typedef enum dmu_object_byteswap { 85ad135b5dSChristopher Siden DMU_BSWAP_UINT8, 86ad135b5dSChristopher Siden DMU_BSWAP_UINT16, 87ad135b5dSChristopher Siden DMU_BSWAP_UINT32, 88ad135b5dSChristopher Siden DMU_BSWAP_UINT64, 89ad135b5dSChristopher Siden DMU_BSWAP_ZAP, 90ad135b5dSChristopher Siden DMU_BSWAP_DNODE, 91ad135b5dSChristopher Siden DMU_BSWAP_OBJSET, 92ad135b5dSChristopher Siden DMU_BSWAP_ZNODE, 93ad135b5dSChristopher Siden DMU_BSWAP_OLDACL, 94ad135b5dSChristopher Siden DMU_BSWAP_ACL, 95ad135b5dSChristopher Siden /* 96ad135b5dSChristopher Siden * Allocating a new byteswap type number makes the on-disk format 97ad135b5dSChristopher Siden * incompatible with any other format that uses the same number. 98ad135b5dSChristopher Siden * 99ad135b5dSChristopher Siden * Data can usually be structured to work with one of the 100ad135b5dSChristopher Siden * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 101ad135b5dSChristopher Siden */ 102ad135b5dSChristopher Siden DMU_BSWAP_NUMFUNCS 103ad135b5dSChristopher Siden } dmu_object_byteswap_t; 104ad135b5dSChristopher Siden 105ad135b5dSChristopher Siden #define DMU_OT_NEWTYPE 0x80 106ad135b5dSChristopher Siden #define DMU_OT_METADATA 0x40 107ad135b5dSChristopher Siden #define DMU_OT_BYTESWAP_MASK 0x3f 108ad135b5dSChristopher Siden 109ad135b5dSChristopher Siden /* 110ad135b5dSChristopher Siden * Defines a uint8_t object type. Object types specify if the data 111ad135b5dSChristopher Siden * in the object is metadata (boolean) and how to byteswap the data 112ad135b5dSChristopher Siden * (dmu_object_byteswap_t). 113ad135b5dSChristopher Siden */ 114ad135b5dSChristopher Siden #define DMU_OT(byteswap, metadata) \ 115ad135b5dSChristopher Siden (DMU_OT_NEWTYPE | \ 116ad135b5dSChristopher Siden ((metadata) ? DMU_OT_METADATA : 0) | \ 117ad135b5dSChristopher Siden ((byteswap) & DMU_OT_BYTESWAP_MASK)) 118ad135b5dSChristopher Siden 119ad135b5dSChristopher Siden #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 120ad135b5dSChristopher Siden ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ 121ad135b5dSChristopher Siden (ot) < DMU_OT_NUMTYPES) 122ad135b5dSChristopher Siden 123ad135b5dSChristopher Siden #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 124ad135b5dSChristopher Siden ((ot) & DMU_OT_METADATA) : \ 125ad135b5dSChristopher Siden dmu_ot[(ot)].ot_metadata) 126ad135b5dSChristopher Siden 1275d7b4d43SMatthew Ahrens /* 1285d7b4d43SMatthew Ahrens * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't 1295d7b4d43SMatthew Ahrens * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill 1305d7b4d43SMatthew Ahrens * is repurposed for embedded BPs. 1315d7b4d43SMatthew Ahrens */ 1325d7b4d43SMatthew Ahrens #define DMU_OT_HAS_FILL(ot) \ 1335d7b4d43SMatthew Ahrens ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) 1345d7b4d43SMatthew Ahrens 135ad135b5dSChristopher Siden #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 136ad135b5dSChristopher Siden ((ot) & DMU_OT_BYTESWAP_MASK) : \ 137ad135b5dSChristopher Siden dmu_ot[(ot)].ot_byteswap) 138ad135b5dSChristopher Siden 139fa9e4066Sahrens typedef enum dmu_object_type { 140fa9e4066Sahrens DMU_OT_NONE, 141fa9e4066Sahrens /* general: */ 142fa9e4066Sahrens DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 143fa9e4066Sahrens DMU_OT_OBJECT_ARRAY, /* UINT64 */ 144fa9e4066Sahrens DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 145fa9e4066Sahrens DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 146cde58dbcSMatthew Ahrens DMU_OT_BPOBJ, /* UINT64 */ 147cde58dbcSMatthew Ahrens DMU_OT_BPOBJ_HDR, /* UINT64 */ 148fa9e4066Sahrens /* spa: */ 149fa9e4066Sahrens DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 150fa9e4066Sahrens DMU_OT_SPACE_MAP, /* UINT64 */ 151fa9e4066Sahrens /* zil: */ 152fa9e4066Sahrens DMU_OT_INTENT_LOG, /* UINT64 */ 153fa9e4066Sahrens /* dmu: */ 154fa9e4066Sahrens DMU_OT_DNODE, /* DNODE */ 155fa9e4066Sahrens DMU_OT_OBJSET, /* OBJSET */ 156fa9e4066Sahrens /* dsl: */ 1571649cd4bStabriz DMU_OT_DSL_DIR, /* UINT64 */ 15887e5029aSahrens DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 15987e5029aSahrens DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 160fa9e4066Sahrens DMU_OT_DSL_PROPS, /* ZAP */ 1611649cd4bStabriz DMU_OT_DSL_DATASET, /* UINT64 */ 162fa9e4066Sahrens /* zpl: */ 163fa9e4066Sahrens DMU_OT_ZNODE, /* ZNODE */ 164da6c28aaSamw DMU_OT_OLDACL, /* Old ACL */ 165fa9e4066Sahrens DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 166fa9e4066Sahrens DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 167fa9e4066Sahrens DMU_OT_MASTER_NODE, /* ZAP */ 168893a6d32Sahrens DMU_OT_UNLINKED_SET, /* ZAP */ 169fa9e4066Sahrens /* zvol: */ 170fa9e4066Sahrens DMU_OT_ZVOL, /* UINT8 */ 171fa9e4066Sahrens DMU_OT_ZVOL_PROP, /* ZAP */ 172fa9e4066Sahrens /* other; for testing only! */ 173fa9e4066Sahrens DMU_OT_PLAIN_OTHER, /* UINT8 */ 174fa9e4066Sahrens DMU_OT_UINT64_OTHER, /* UINT64 */ 175fa9e4066Sahrens DMU_OT_ZAP_OTHER, /* ZAP */ 176ea8dc4b6Seschrock /* new object types: */ 177ea8dc4b6Seschrock DMU_OT_ERROR_LOG, /* ZAP */ 17806eeb2adSek DMU_OT_SPA_HISTORY, /* UINT8 */ 17906eeb2adSek DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 180b1b8ab34Slling DMU_OT_POOL_PROPS, /* ZAP */ 181ecd6cf80Smarks DMU_OT_DSL_PERMS, /* ZAP */ 182da6c28aaSamw DMU_OT_ACL, /* ACL */ 183da6c28aaSamw DMU_OT_SYSACL, /* SYSACL */ 184da6c28aaSamw DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 185da6c28aaSamw DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 186088f3894Sahrens DMU_OT_NEXT_CLONES, /* ZAP */ 1873f9d6ad7SLin Ling DMU_OT_SCAN_QUEUE, /* ZAP */ 18814843421SMatthew Ahrens DMU_OT_USERGROUP_USED, /* ZAP */ 18914843421SMatthew Ahrens DMU_OT_USERGROUP_QUOTA, /* ZAP */ 190842727c2SChris Kirby DMU_OT_USERREFS, /* ZAP */ 191b24ab676SJeff Bonwick DMU_OT_DDT_ZAP, /* ZAP */ 192b24ab676SJeff Bonwick DMU_OT_DDT_STATS, /* ZAP */ 1930a586ceaSMark Shellenbaum DMU_OT_SA, /* System attr */ 1940a586ceaSMark Shellenbaum DMU_OT_SA_MASTER_NODE, /* ZAP */ 1950a586ceaSMark Shellenbaum DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1960a586ceaSMark Shellenbaum DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1973f9d6ad7SLin Ling DMU_OT_SCAN_XLATE, /* ZAP */ 1983f9d6ad7SLin Ling DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 199cde58dbcSMatthew Ahrens DMU_OT_DEADLIST, /* ZAP */ 200cde58dbcSMatthew Ahrens DMU_OT_DEADLIST_HDR, /* UINT64 */ 201cde58dbcSMatthew Ahrens DMU_OT_DSL_CLONES, /* ZAP */ 202cde58dbcSMatthew Ahrens DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 203ad135b5dSChristopher Siden /* 204ad135b5dSChristopher Siden * Do not allocate new object types here. Doing so makes the on-disk 205ad135b5dSChristopher Siden * format incompatible with any other format that uses the same object 206ad135b5dSChristopher Siden * type number. 207ad135b5dSChristopher Siden * 208ad135b5dSChristopher Siden * When creating an object which does not have one of the above types 209ad135b5dSChristopher Siden * use the DMU_OTN_* type with the correct byteswap and metadata 210ad135b5dSChristopher Siden * values. 211ad135b5dSChristopher Siden * 212ad135b5dSChristopher Siden * The DMU_OTN_* types do not have entries in the dmu_ot table, 213ad135b5dSChristopher Siden * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead 214ad135b5dSChristopher Siden * of indexing into dmu_ot directly (this works for both DMU_OT_* types 215ad135b5dSChristopher Siden * and DMU_OTN_* types). 216ad135b5dSChristopher Siden */ 217ad135b5dSChristopher Siden DMU_OT_NUMTYPES, 218ad135b5dSChristopher Siden 219ad135b5dSChristopher Siden /* 220ad135b5dSChristopher Siden * Names for valid types declared with DMU_OT(). 221ad135b5dSChristopher Siden */ 222ad135b5dSChristopher Siden DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 223ad135b5dSChristopher Siden DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 224ad135b5dSChristopher Siden DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 225ad135b5dSChristopher Siden DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 226ad135b5dSChristopher Siden DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 227ad135b5dSChristopher Siden DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 228ad135b5dSChristopher Siden DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 229ad135b5dSChristopher Siden DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 230ad135b5dSChristopher Siden DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 231ad135b5dSChristopher Siden DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE), 232fa9e4066Sahrens } dmu_object_type_t; 233fa9e4066Sahrens 2343b2aab18SMatthew Ahrens typedef enum txg_how { 2353b2aab18SMatthew Ahrens TXG_WAIT = 1, 2363b2aab18SMatthew Ahrens TXG_NOWAIT, 23769962b56SMatthew Ahrens TXG_WAITED, 2383b2aab18SMatthew Ahrens } txg_how_t; 2393b2aab18SMatthew Ahrens 240fa9e4066Sahrens void byteswap_uint64_array(void *buf, size_t size); 241fa9e4066Sahrens void byteswap_uint32_array(void *buf, size_t size); 242fa9e4066Sahrens void byteswap_uint16_array(void *buf, size_t size); 243fa9e4066Sahrens void byteswap_uint8_array(void *buf, size_t size); 244fa9e4066Sahrens void zap_byteswap(void *buf, size_t size); 245da6c28aaSamw void zfs_oldacl_byteswap(void *buf, size_t size); 246fa9e4066Sahrens void zfs_acl_byteswap(void *buf, size_t size); 247fa9e4066Sahrens void zfs_znode_byteswap(void *buf, size_t size); 248fa9e4066Sahrens 2490b69c2f0Sahrens #define DS_FIND_SNAPSHOTS (1<<0) 2500b69c2f0Sahrens #define DS_FIND_CHILDREN (1<<1) 25112380e1eSArne Jansen #define DS_FIND_SERIALIZE (1<<2) 252fa9e4066Sahrens 253fa9e4066Sahrens /* 254fa9e4066Sahrens * The maximum number of bytes that can be accessed as part of one 255fa9e4066Sahrens * operation, including metadata. 256fa9e4066Sahrens */ 257b5152584SMatthew Ahrens #define DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */ 258cdb0ab79Smaybee #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ 259fa9e4066Sahrens 26014843421SMatthew Ahrens #define DMU_USERUSED_OBJECT (-1ULL) 26114843421SMatthew Ahrens #define DMU_GROUPUSED_OBJECT (-2ULL) 26214843421SMatthew Ahrens 2630a586ceaSMark Shellenbaum /* 2640a586ceaSMark Shellenbaum * artificial blkids for bonus buffer and spill blocks 2650a586ceaSMark Shellenbaum */ 2660a586ceaSMark Shellenbaum #define DMU_BONUS_BLKID (-1ULL) 2670a586ceaSMark Shellenbaum #define DMU_SPILL_BLKID (-2ULL) 268fa9e4066Sahrens /* 269fa9e4066Sahrens * Public routines to create, destroy, open, and close objsets. 270fa9e4066Sahrens */ 271503ad85cSMatthew Ahrens int dmu_objset_hold(const char *name, void *tag, objset_t **osp); 272503ad85cSMatthew Ahrens int dmu_objset_own(const char *name, dmu_objset_type_t type, 273503ad85cSMatthew Ahrens boolean_t readonly, void *tag, objset_t **osp); 274503ad85cSMatthew Ahrens void dmu_objset_rele(objset_t *os, void *tag); 275503ad85cSMatthew Ahrens void dmu_objset_disown(objset_t *os, void *tag); 276503ad85cSMatthew Ahrens int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 277503ad85cSMatthew Ahrens 2783b2aab18SMatthew Ahrens void dmu_objset_evict_dbufs(objset_t *os); 279ae46e4c7SMatthew Ahrens int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 280ecd6cf80Smarks void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); 2813b2aab18SMatthew Ahrens int dmu_objset_clone(const char *name, const char *origin); 2823b2aab18SMatthew Ahrens int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, 2834445fffbSMatthew Ahrens struct nvlist *errlist); 2844445fffbSMatthew Ahrens int dmu_objset_snapshot_one(const char *fsname, const char *snapname); 2854445fffbSMatthew Ahrens int dmu_objset_snapshot_tmp(const char *, const char *, int); 286fd136879SMatthew Ahrens int dmu_objset_find(char *name, int func(const char *, void *), void *arg, 287fa9e4066Sahrens int flags); 288fa9e4066Sahrens void dmu_objset_byteswap(void *buf, size_t size); 2893b2aab18SMatthew Ahrens int dsl_dataset_rename_snapshot(const char *fsname, 2903b2aab18SMatthew Ahrens const char *oldsnapname, const char *newsnapname, boolean_t recursive); 291fa9e4066Sahrens 292fa9e4066Sahrens typedef struct dmu_buf { 293fa9e4066Sahrens uint64_t db_object; /* object that this buffer is part of */ 294fa9e4066Sahrens uint64_t db_offset; /* byte offset in this object */ 295fa9e4066Sahrens uint64_t db_size; /* size of buffer in bytes */ 296fa9e4066Sahrens void *db_data; /* data in buffer */ 297fa9e4066Sahrens } dmu_buf_t; 298fa9e4066Sahrens 29999653d4eSeschrock /* 30099653d4eSeschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 30199653d4eSeschrock */ 302fa9e4066Sahrens #define DMU_POOL_DIRECTORY_OBJECT 1 303fa9e4066Sahrens #define DMU_POOL_CONFIG "config" 304ad135b5dSChristopher Siden #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 305ad135b5dSChristopher Siden #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 306ad135b5dSChristopher Siden #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 30743466aaeSMax Grossman #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" 308fa9e4066Sahrens #define DMU_POOL_ROOT_DATASET "root_dataset" 309cde58dbcSMatthew Ahrens #define DMU_POOL_SYNC_BPOBJ "sync_bplist" 310ea8dc4b6Seschrock #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 311ea8dc4b6Seschrock #define DMU_POOL_ERRLOG_LAST "errlog_last" 31299653d4eSeschrock #define DMU_POOL_SPARES "spares" 31399653d4eSeschrock #define DMU_POOL_DEFLATE "deflate" 31406eeb2adSek #define DMU_POOL_HISTORY "history" 315b1b8ab34Slling #define DMU_POOL_PROPS "pool_props" 316fa94a07fSbrendan #define DMU_POOL_L2CACHE "l2cache" 317ca45db41SChris Kirby #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 318b24ab676SJeff Bonwick #define DMU_POOL_DDT "DDT-%s-%s-%s" 319b24ab676SJeff Bonwick #define DMU_POOL_DDT_STATS "DDT-statistics" 3203f9d6ad7SLin Ling #define DMU_POOL_CREATION_VERSION "creation_version" 3213f9d6ad7SLin Ling #define DMU_POOL_SCAN "scan" 322cde58dbcSMatthew Ahrens #define DMU_POOL_FREE_BPOBJ "free_bpobj" 323ad135b5dSChristopher Siden #define DMU_POOL_BPTREE_OBJ "bptree_obj" 324f1745736SMatthew Ahrens #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 32545818ee1SMatthew Ahrens #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 326215198a6SJoe Stein #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map" 327088f3894Sahrens 328fa9e4066Sahrens /* 329fa9e4066Sahrens * Allocate an object from this objset. The range of object numbers 330fa9e4066Sahrens * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 331fa9e4066Sahrens * 332fa9e4066Sahrens * The transaction must be assigned to a txg. The newly allocated 333fa9e4066Sahrens * object will be "held" in the transaction (ie. you can modify the 334fa9e4066Sahrens * newly allocated object in this transaction). 335fa9e4066Sahrens * 336fa9e4066Sahrens * dmu_object_alloc() chooses an object and returns it in *objectp. 337fa9e4066Sahrens * 338fa9e4066Sahrens * dmu_object_claim() allocates a specific object number. If that 339fa9e4066Sahrens * number is already allocated, it fails and returns EEXIST. 340fa9e4066Sahrens * 341fa9e4066Sahrens * Return 0 on success, or ENOSPC or EEXIST as specified above. 342fa9e4066Sahrens */ 343fa9e4066Sahrens uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 344fa9e4066Sahrens int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 345fa9e4066Sahrens int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 346fa9e4066Sahrens int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 347fa9e4066Sahrens int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 348e77d42eaSMatthew Ahrens int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); 349fa9e4066Sahrens 350fa9e4066Sahrens /* 351fa9e4066Sahrens * Free an object from this objset. 352fa9e4066Sahrens * 353fa9e4066Sahrens * The object's data will be freed as well (ie. you don't need to call 354fa9e4066Sahrens * dmu_free(object, 0, -1, tx)). 355fa9e4066Sahrens * 356fa9e4066Sahrens * The object need not be held in the transaction. 357fa9e4066Sahrens * 358fa9e4066Sahrens * If there are any holds on this object's buffers (via dmu_buf_hold()), 359fa9e4066Sahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not 360fa9e4066Sahrens * free it; it fails and returns EBUSY. 361fa9e4066Sahrens * 362fa9e4066Sahrens * If the object is not allocated, it fails and returns ENOENT. 363fa9e4066Sahrens * 364fa9e4066Sahrens * Return 0 on success, or EBUSY or ENOENT as specified above. 365fa9e4066Sahrens */ 366fa9e4066Sahrens int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 367fa9e4066Sahrens 368fa9e4066Sahrens /* 369fa9e4066Sahrens * Find the next allocated or free object. 370fa9e4066Sahrens * 371fa9e4066Sahrens * The objectp parameter is in-out. It will be updated to be the next 3726754306eSahrens * object which is allocated. Ignore objects which have not been 3736754306eSahrens * modified since txg. 374fa9e4066Sahrens * 375fa9e4066Sahrens * XXX Can only be called on a objset with no dirty data. 376fa9e4066Sahrens * 377fa9e4066Sahrens * Returns 0 on success, or ENOENT if there are no more objects. 378fa9e4066Sahrens */ 3796754306eSahrens int dmu_object_next(objset_t *os, uint64_t *objectp, 3806754306eSahrens boolean_t hole, uint64_t txg); 381fa9e4066Sahrens 382fa9e4066Sahrens /* 383fa9e4066Sahrens * Set the data blocksize for an object. 384fa9e4066Sahrens * 385fa9e4066Sahrens * The object cannot have any blocks allcated beyond the first. If 386fa9e4066Sahrens * the first block is allocated already, the new size must be greater 387fa9e4066Sahrens * than the current block size. If these conditions are not met, 388fa9e4066Sahrens * ENOTSUP will be returned. 389fa9e4066Sahrens * 390fa9e4066Sahrens * Returns 0 on success, or EBUSY if there are any holds on the object 391fa9e4066Sahrens * contents, or ENOTSUP as described above. 392fa9e4066Sahrens */ 393fa9e4066Sahrens int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 394fa9e4066Sahrens int ibs, dmu_tx_t *tx); 395fa9e4066Sahrens 396fa9e4066Sahrens /* 397fa9e4066Sahrens * Set the checksum property on a dnode. The new checksum algorithm will 398fa9e4066Sahrens * apply to all newly written blocks; existing blocks will not be affected. 399fa9e4066Sahrens */ 400fa9e4066Sahrens void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 401fa9e4066Sahrens dmu_tx_t *tx); 402fa9e4066Sahrens 403fa9e4066Sahrens /* 404fa9e4066Sahrens * Set the compress property on a dnode. The new compression algorithm will 405fa9e4066Sahrens * apply to all newly written blocks; existing blocks will not be affected. 406fa9e4066Sahrens */ 407fa9e4066Sahrens void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 408fa9e4066Sahrens dmu_tx_t *tx); 409fa9e4066Sahrens 4105d7b4d43SMatthew Ahrens void 4115d7b4d43SMatthew Ahrens dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, 4125d7b4d43SMatthew Ahrens void *data, uint8_t etype, uint8_t comp, int uncompressed_size, 4135d7b4d43SMatthew Ahrens int compressed_size, int byteorder, dmu_tx_t *tx); 4145d7b4d43SMatthew Ahrens 41544cd46caSbillm /* 416b24ab676SJeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc. 41744cd46caSbillm */ 418b24ab676SJeff Bonwick #define WP_NOFILL 0x1 419b24ab676SJeff Bonwick #define WP_DMU_SYNC 0x2 4200a586ceaSMark Shellenbaum #define WP_SPILL 0x4 421b24ab676SJeff Bonwick 42279d72832SMatthew Ahrens void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, 4235602294fSDan Kimmel enum zio_compress compress_override, struct zio_prop *zp); 424fa9e4066Sahrens /* 425fa9e4066Sahrens * The bonus data is accessed more or less like a regular buffer. 426fa9e4066Sahrens * You must dmu_bonus_hold() to get the buffer, which will give you a 427fa9e4066Sahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 428fa9e4066Sahrens * data. As with any normal buffer, you must call dmu_buf_read() to 429fa9e4066Sahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the 430fa9e4066Sahrens * object must be held in an assigned transaction before calling 431fa9e4066Sahrens * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 432fa9e4066Sahrens * buffer as well. You must release your hold with dmu_buf_rele(). 4333e30c24aSWill Andrews * 4343e30c24aSWill Andrews * Returns ENOENT, EIO, or 0. 435fa9e4066Sahrens */ 436ea8dc4b6Seschrock int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); 437fa9e4066Sahrens int dmu_bonus_max(void); 4381934e92fSmaybee int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 4390a586ceaSMark Shellenbaum int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 440744947dcSTom Erickson dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); 4410a586ceaSMark Shellenbaum int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 4420a586ceaSMark Shellenbaum 4430a586ceaSMark Shellenbaum /* 4440a586ceaSMark Shellenbaum * Special spill buffer support used by "SA" framework 4450a586ceaSMark Shellenbaum */ 4460a586ceaSMark Shellenbaum 4470a586ceaSMark Shellenbaum int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 44879d72832SMatthew Ahrens int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, 4490a586ceaSMark Shellenbaum void *tag, dmu_buf_t **dbp); 4500a586ceaSMark Shellenbaum int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 451fa9e4066Sahrens 452fa9e4066Sahrens /* 453fa9e4066Sahrens * Obtain the DMU buffer from the specified object which contains the 454fa9e4066Sahrens * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 455fa9e4066Sahrens * that it will remain in memory. You must release the hold with 456fa9e4066Sahrens * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your 457fa9e4066Sahrens * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 458fa9e4066Sahrens * 459fa9e4066Sahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 460fa9e4066Sahrens * on the returned buffer before reading or writing the buffer's 461fa9e4066Sahrens * db_data. The comments for those routines describe what particular 462fa9e4066Sahrens * operations are valid after calling them. 463fa9e4066Sahrens * 464fa9e4066Sahrens * The object number must be a valid, allocated object number. 465fa9e4066Sahrens */ 466ea8dc4b6Seschrock int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 46747cb52daSJeff Bonwick void *tag, dmu_buf_t **, int flags); 46879d72832SMatthew Ahrens int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, 46979d72832SMatthew Ahrens void *tag, dmu_buf_t **dbp, int flags); 470e57a022bSJustin T. Gibbs 471e57a022bSJustin T. Gibbs /* 472e57a022bSJustin T. Gibbs * Add a reference to a dmu buffer that has already been held via 473e57a022bSJustin T. Gibbs * dmu_buf_hold() in the current context. 474e57a022bSJustin T. Gibbs */ 475fa9e4066Sahrens void dmu_buf_add_ref(dmu_buf_t *db, void* tag); 476e57a022bSJustin T. Gibbs 477e57a022bSJustin T. Gibbs /* 478e57a022bSJustin T. Gibbs * Attempt to add a reference to a dmu buffer that is in an unknown state, 479e57a022bSJustin T. Gibbs * using a pointer that may have been invalidated by eviction processing. 480e57a022bSJustin T. Gibbs * The request will succeed if the passed in dbuf still represents the 481e57a022bSJustin T. Gibbs * same os/object/blkid, is ineligible for eviction, and has at least 482e57a022bSJustin T. Gibbs * one hold by a user other than the syncer. 483e57a022bSJustin T. Gibbs */ 484e57a022bSJustin T. Gibbs boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, 485e57a022bSJustin T. Gibbs uint64_t blkid, void *tag); 486e57a022bSJustin T. Gibbs 487ea8dc4b6Seschrock void dmu_buf_rele(dmu_buf_t *db, void *tag); 488fa9e4066Sahrens uint64_t dmu_buf_refcount(dmu_buf_t *db); 489fa9e4066Sahrens 490fa9e4066Sahrens /* 491fa9e4066Sahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 492fa9e4066Sahrens * range of an object. A pointer to an array of dmu_buf_t*'s is 493fa9e4066Sahrens * returned (in *dbpp). 494fa9e4066Sahrens * 495fa9e4066Sahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 496fa9e4066Sahrens * frees the array. The hold on the array of buffers MUST be released 497fa9e4066Sahrens * with dmu_buf_rele_array. You can NOT release the hold on each buffer 498fa9e4066Sahrens * individually with dmu_buf_rele. 499fa9e4066Sahrens */ 50013506d1eSmaybee int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 501cf6106c8SMatthew Ahrens uint64_t length, boolean_t read, void *tag, 502cf6106c8SMatthew Ahrens int *numbufsp, dmu_buf_t ***dbpp); 503ea8dc4b6Seschrock void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); 504fa9e4066Sahrens 505bc9014e6SJustin Gibbs typedef void dmu_buf_evict_func_t(void *user_ptr); 506bc9014e6SJustin Gibbs 507fa9e4066Sahrens /* 508bc9014e6SJustin Gibbs * A DMU buffer user object may be associated with a dbuf for the 509bc9014e6SJustin Gibbs * duration of its lifetime. This allows the user of a dbuf (client) 510bc9014e6SJustin Gibbs * to attach private data to a dbuf (e.g. in-core only data such as a 511bc9014e6SJustin Gibbs * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified 512bc9014e6SJustin Gibbs * when that dbuf has been evicted. Clients typically respond to the 513bc9014e6SJustin Gibbs * eviction notification by freeing their private data, thus ensuring 514bc9014e6SJustin Gibbs * the same lifetime for both dbuf and private data. 515fa9e4066Sahrens * 516bc9014e6SJustin Gibbs * The mapping from a dmu_buf_user_t to any client private data is the 517bc9014e6SJustin Gibbs * client's responsibility. All current consumers of the API with private 518bc9014e6SJustin Gibbs * data embed a dmu_buf_user_t as the first member of the structure for 519bc9014e6SJustin Gibbs * their private data. This allows conversions between the two types 520bc9014e6SJustin Gibbs * with a simple cast. Since the DMU buf user API never needs access 521bc9014e6SJustin Gibbs * to the private data, other strategies can be employed if necessary 522bc9014e6SJustin Gibbs * or convenient for the client (e.g. using container_of() to do the 523bc9014e6SJustin Gibbs * conversion for private data that cannot have the dmu_buf_user_t as 524bc9014e6SJustin Gibbs * its first member). 525fa9e4066Sahrens * 526bc9014e6SJustin Gibbs * Eviction callbacks are executed without the dbuf mutex held or any 527bc9014e6SJustin Gibbs * other type of mechanism to guarantee that the dbuf is still available. 528bc9014e6SJustin Gibbs * For this reason, users must assume the dbuf has already been freed 529bc9014e6SJustin Gibbs * and not reference the dbuf from the callback context. 530fa9e4066Sahrens * 531bc9014e6SJustin Gibbs * Users requesting "immediate eviction" are notified as soon as the dbuf 532bc9014e6SJustin Gibbs * is only referenced by dirty records (dirties == holds). Otherwise the 533bc9014e6SJustin Gibbs * notification occurs after eviction processing for the dbuf begins. 534fa9e4066Sahrens */ 535bc9014e6SJustin Gibbs typedef struct dmu_buf_user { 536bc9014e6SJustin Gibbs /* 537bc9014e6SJustin Gibbs * Asynchronous user eviction callback state. 538bc9014e6SJustin Gibbs */ 539bc9014e6SJustin Gibbs taskq_ent_t dbu_tqent; 540bc9014e6SJustin Gibbs 54140510e8eSJosef 'Jeff' Sipek /* 54240510e8eSJosef 'Jeff' Sipek * This instance's eviction function pointers. 54340510e8eSJosef 'Jeff' Sipek * 54440510e8eSJosef 'Jeff' Sipek * dbu_evict_func_sync is called synchronously and then 54540510e8eSJosef 'Jeff' Sipek * dbu_evict_func_async is executed asynchronously on a taskq. 54640510e8eSJosef 'Jeff' Sipek */ 54740510e8eSJosef 'Jeff' Sipek dmu_buf_evict_func_t *dbu_evict_func_sync; 54840510e8eSJosef 'Jeff' Sipek dmu_buf_evict_func_t *dbu_evict_func_async; 549bc9014e6SJustin Gibbs #ifdef ZFS_DEBUG 550bc9014e6SJustin Gibbs /* 551bc9014e6SJustin Gibbs * Pointer to user's dbuf pointer. NULL for clients that do 552bc9014e6SJustin Gibbs * not associate a dbuf with their user data. 553bc9014e6SJustin Gibbs * 554bc9014e6SJustin Gibbs * The dbuf pointer is cleared upon eviction so as to catch 555bc9014e6SJustin Gibbs * use-after-evict bugs in clients. 556bc9014e6SJustin Gibbs */ 557bc9014e6SJustin Gibbs dmu_buf_t **dbu_clear_on_evict_dbufp; 558bc9014e6SJustin Gibbs #endif 559bc9014e6SJustin Gibbs } dmu_buf_user_t; 560bc9014e6SJustin Gibbs 561fa9e4066Sahrens /* 562bc9014e6SJustin Gibbs * Initialize the given dmu_buf_user_t instance with the eviction function 563bc9014e6SJustin Gibbs * evict_func, to be called when the user is evicted. 564bc9014e6SJustin Gibbs * 565bc9014e6SJustin Gibbs * NOTE: This function should only be called once on a given dmu_buf_user_t. 566bc9014e6SJustin Gibbs * To allow enforcement of this, dbu must already be zeroed on entry. 567fa9e4066Sahrens */ 5684ba5b961SDan Kimmel /*ARGSUSED*/ 569bc9014e6SJustin Gibbs inline void 57040510e8eSJosef 'Jeff' Sipek dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, 57140510e8eSJosef 'Jeff' Sipek dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp) 572bc9014e6SJustin Gibbs { 57340510e8eSJosef 'Jeff' Sipek ASSERT(dbu->dbu_evict_func_sync == NULL); 57440510e8eSJosef 'Jeff' Sipek ASSERT(dbu->dbu_evict_func_async == NULL); 57540510e8eSJosef 'Jeff' Sipek 57640510e8eSJosef 'Jeff' Sipek /* must have at least one evict func */ 57740510e8eSJosef 'Jeff' Sipek IMPLY(evict_func_sync == NULL, evict_func_async != NULL); 57840510e8eSJosef 'Jeff' Sipek dbu->dbu_evict_func_sync = evict_func_sync; 57940510e8eSJosef 'Jeff' Sipek dbu->dbu_evict_func_async = evict_func_async; 580bc9014e6SJustin Gibbs #ifdef ZFS_DEBUG 581bc9014e6SJustin Gibbs dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; 582bc9014e6SJustin Gibbs #endif 583bc9014e6SJustin Gibbs } 584fa9e4066Sahrens 585fa9e4066Sahrens /* 586bc9014e6SJustin Gibbs * Attach user data to a dbuf and mark it for normal (when the dbuf's 587bc9014e6SJustin Gibbs * data is cleared or its reference count goes to zero) eviction processing. 588bc9014e6SJustin Gibbs * 589bc9014e6SJustin Gibbs * Returns NULL on success, or the existing user if another user currently 590bc9014e6SJustin Gibbs * owns the buffer. 591bc9014e6SJustin Gibbs */ 592bc9014e6SJustin Gibbs void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); 593bc9014e6SJustin Gibbs 594bc9014e6SJustin Gibbs /* 595bc9014e6SJustin Gibbs * Attach user data to a dbuf and mark it for immediate (its dirty and 596bc9014e6SJustin Gibbs * reference counts are equal) eviction processing. 597bc9014e6SJustin Gibbs * 598bc9014e6SJustin Gibbs * Returns NULL on success, or the existing user if another user currently 599bc9014e6SJustin Gibbs * owns the buffer. 600bc9014e6SJustin Gibbs */ 601bc9014e6SJustin Gibbs void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); 602bc9014e6SJustin Gibbs 603bc9014e6SJustin Gibbs /* 604bc9014e6SJustin Gibbs * Replace the current user of a dbuf. 605bc9014e6SJustin Gibbs * 606bc9014e6SJustin Gibbs * If given the current user of a dbuf, replaces the dbuf's user with 607bc9014e6SJustin Gibbs * "new_user" and returns the user data pointer that was replaced. 608bc9014e6SJustin Gibbs * Otherwise returns the current, and unmodified, dbuf user pointer. 609bc9014e6SJustin Gibbs */ 610bc9014e6SJustin Gibbs void *dmu_buf_replace_user(dmu_buf_t *db, 611bc9014e6SJustin Gibbs dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); 612bc9014e6SJustin Gibbs 613bc9014e6SJustin Gibbs /* 614bc9014e6SJustin Gibbs * Remove the specified user data for a DMU buffer. 615bc9014e6SJustin Gibbs * 616bc9014e6SJustin Gibbs * Returns the user that was removed on success, or the current user if 617bc9014e6SJustin Gibbs * another user currently owns the buffer. 618bc9014e6SJustin Gibbs */ 619bc9014e6SJustin Gibbs void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); 620bc9014e6SJustin Gibbs 621bc9014e6SJustin Gibbs /* 622bc9014e6SJustin Gibbs * Returns the user data (dmu_buf_user_t *) associated with this dbuf. 623fa9e4066Sahrens */ 624fa9e4066Sahrens void *dmu_buf_get_user(dmu_buf_t *db); 625fa9e4066Sahrens 626ae972795SMatthew Ahrens objset_t *dmu_buf_get_objset(dmu_buf_t *db); 62779d72832SMatthew Ahrens dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db); 62879d72832SMatthew Ahrens void dmu_buf_dnode_exit(dmu_buf_t *db); 629ae972795SMatthew Ahrens 630bc9014e6SJustin Gibbs /* Block until any in-progress dmu buf user evictions complete. */ 631bc9014e6SJustin Gibbs void dmu_buf_user_evict_wait(void); 632bc9014e6SJustin Gibbs 63380901aeaSGeorge Wilson /* 63480901aeaSGeorge Wilson * Returns the blkptr associated with this dbuf, or NULL if not set. 63580901aeaSGeorge Wilson */ 63680901aeaSGeorge Wilson struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); 63780901aeaSGeorge Wilson 638fa9e4066Sahrens /* 639fa9e4066Sahrens * Indicate that you are going to modify the buffer's data (db_data). 640fa9e4066Sahrens * 641fa9e4066Sahrens * The transaction (tx) must be assigned to a txg (ie. you've called 642fa9e4066Sahrens * dmu_tx_assign()). The buffer's object must be held in the tx 643fa9e4066Sahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 644fa9e4066Sahrens */ 645fa9e4066Sahrens void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 646fa9e4066Sahrens 647fa9e4066Sahrens /* 648fa9e4066Sahrens * You must create a transaction, then hold the objects which you will 649fa9e4066Sahrens * (or might) modify as part of this transaction. Then you must assign 650fa9e4066Sahrens * the transaction to a transaction group. Once the transaction has 651fa9e4066Sahrens * been assigned, you can modify buffers which belong to held objects as 652fa9e4066Sahrens * part of this transaction. You can't modify buffers before the 653fa9e4066Sahrens * transaction has been assigned; you can't modify buffers which don't 654fa9e4066Sahrens * belong to objects which this transaction holds; you can't hold 655fa9e4066Sahrens * objects once the transaction has been assigned. You may hold an 656fa9e4066Sahrens * object which you are going to free (with dmu_object_free()), but you 657fa9e4066Sahrens * don't have to. 658fa9e4066Sahrens * 659fa9e4066Sahrens * You can abort the transaction before it has been assigned. 660fa9e4066Sahrens * 661fa9e4066Sahrens * Note that you may hold buffers (with dmu_buf_hold) at any time, 662fa9e4066Sahrens * regardless of transaction state. 663fa9e4066Sahrens */ 664fa9e4066Sahrens 665fa9e4066Sahrens #define DMU_NEW_OBJECT (-1ULL) 666fa9e4066Sahrens #define DMU_OBJECT_END (-1ULL) 667fa9e4066Sahrens 668fa9e4066Sahrens dmu_tx_t *dmu_tx_create(objset_t *os); 669fa9e4066Sahrens void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); 670*b0c42cd4Sbzzz void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, 671*b0c42cd4Sbzzz int len); 672fa9e4066Sahrens void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 673fa9e4066Sahrens uint64_t len); 674*b0c42cd4Sbzzz void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, 675*b0c42cd4Sbzzz uint64_t len); 67614843421SMatthew Ahrens void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 677*b0c42cd4Sbzzz void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, 678*b0c42cd4Sbzzz const char *name); 679fa9e4066Sahrens void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 680*b0c42cd4Sbzzz void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn); 6810a586ceaSMark Shellenbaum void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 6820a586ceaSMark Shellenbaum void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 6830a586ceaSMark Shellenbaum void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 684fa9e4066Sahrens void dmu_tx_abort(dmu_tx_t *tx); 6853b2aab18SMatthew Ahrens int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how); 6868a2f1b91Sahrens void dmu_tx_wait(dmu_tx_t *tx); 687fa9e4066Sahrens void dmu_tx_commit(dmu_tx_t *tx); 6884bb73804SMatthew Ahrens void dmu_tx_mark_netfree(dmu_tx_t *tx); 689fa9e4066Sahrens 690d20e665cSRicardo M. Correia /* 691d20e665cSRicardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called. 692d20e665cSRicardo M. Correia * 693d20e665cSRicardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a 694d20e665cSRicardo M. Correia * callback parameter. The caller is responsible for properly allocating and 695d20e665cSRicardo M. Correia * freeing it. 696d20e665cSRicardo M. Correia * 697d20e665cSRicardo M. Correia * When registering a callback, the transaction must be already created, but 698d20e665cSRicardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not. 699d20e665cSRicardo M. Correia * 700d20e665cSRicardo M. Correia * The callback will be called after the transaction has been safely written 701d20e665cSRicardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted. 702d20e665cSRicardo M. Correia * If there is any error which prevents the transaction from being committed to 703d20e665cSRicardo M. Correia * disk, the callback will be called with a value of error != 0. 704d20e665cSRicardo M. Correia */ 705d20e665cSRicardo M. Correia typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 706d20e665cSRicardo M. Correia 707d20e665cSRicardo M. Correia void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 708d20e665cSRicardo M. Correia void *dcb_data); 709d20e665cSRicardo M. Correia 710fa9e4066Sahrens /* 711fa9e4066Sahrens * Free up the data blocks for a defined range of a file. If size is 712ad135b5dSChristopher Siden * -1, the range from offset to end-of-file is freed. 713fa9e4066Sahrens */ 714ea8dc4b6Seschrock int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 715fa9e4066Sahrens uint64_t size, dmu_tx_t *tx); 716cdb0ab79Smaybee int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 717cdb0ab79Smaybee uint64_t size); 718713d6c20SMatthew Ahrens int dmu_free_long_object(objset_t *os, uint64_t object); 719fa9e4066Sahrens 720fa9e4066Sahrens /* 721fa9e4066Sahrens * Convenience functions. 722fa9e4066Sahrens * 723fa9e4066Sahrens * Canfail routines will return 0 on success, or an errno if there is a 724fa9e4066Sahrens * nonrecoverable I/O error. 725fa9e4066Sahrens */ 7267bfdf011SNeil Perrin #define DMU_READ_PREFETCH 0 /* prefetch */ 7277bfdf011SNeil Perrin #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 728ea8dc4b6Seschrock int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 7297bfdf011SNeil Perrin void *buf, uint32_t flags); 730*b0c42cd4Sbzzz int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, 731*b0c42cd4Sbzzz uint32_t flags); 732fa9e4066Sahrens void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 733fa9e4066Sahrens const void *buf, dmu_tx_t *tx); 734*b0c42cd4Sbzzz void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, 735*b0c42cd4Sbzzz const void *buf, dmu_tx_t *tx); 73682c9918fSTim Haley void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 73782c9918fSTim Haley dmu_tx_t *tx); 738feb08c6bSbillm int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); 739f8554bb9SMatthew Ahrens int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size); 740feb08c6bSbillm int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, 741feb08c6bSbillm dmu_tx_t *tx); 74294d1a210STim Haley int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, 74394d1a210STim Haley dmu_tx_t *tx); 74444eda4d7Smaybee int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, 74544eda4d7Smaybee uint64_t size, struct page *pp, dmu_tx_t *tx); 7462fdbea25SAleksandr Guzovskiy struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 7472fdbea25SAleksandr Guzovskiy void dmu_return_arcbuf(struct arc_buf *buf); 7482fdbea25SAleksandr Guzovskiy void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, 7492fdbea25SAleksandr Guzovskiy dmu_tx_t *tx); 750c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_init(struct xuio *uio, int niov); 751c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void dmu_xuio_fini(struct xuio *uio); 752c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, 753c242f9a0Schunli zhang - Sun Microsystems - Irvine United States size_t n); 754c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_cnt(struct xuio *uio); 755c242f9a0Schunli zhang - Sun Microsystems - Irvine United States struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); 756c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void dmu_xuio_clear(struct xuio *uio, int i); 75799aa8b55SPrashanth Sreenivasa void xuio_stat_wbuf_copied(void); 75899aa8b55SPrashanth Sreenivasa void xuio_stat_wbuf_nocopy(void); 759fa9e4066Sahrens 760cf6106c8SMatthew Ahrens extern boolean_t zfs_prefetch_disable; 761b5152584SMatthew Ahrens extern int zfs_max_recordsize; 762416e0cd8Sek 763fa9e4066Sahrens /* 764fa9e4066Sahrens * Asynchronously try to read in the data. 765fa9e4066Sahrens */ 766a2cdcdd2SPaul Dagnelie void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, 767a2cdcdd2SPaul Dagnelie uint64_t len, enum zio_priority pri); 768fa9e4066Sahrens 769fa9e4066Sahrens typedef struct dmu_object_info { 770b24ab676SJeff Bonwick /* All sizes are in bytes unless otherwise indicated. */ 771fa9e4066Sahrens uint32_t doi_data_block_size; 772fa9e4066Sahrens uint32_t doi_metadata_block_size; 773fa9e4066Sahrens dmu_object_type_t doi_type; 774fa9e4066Sahrens dmu_object_type_t doi_bonus_type; 775b24ab676SJeff Bonwick uint64_t doi_bonus_size; 776fa9e4066Sahrens uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 777fa9e4066Sahrens uint8_t doi_checksum; 778fa9e4066Sahrens uint8_t doi_compress; 779e77d42eaSMatthew Ahrens uint8_t doi_nblkptr; 780e77d42eaSMatthew Ahrens uint8_t doi_pad[4]; 781b24ab676SJeff Bonwick uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 782b24ab676SJeff Bonwick uint64_t doi_max_offset; 783b24ab676SJeff Bonwick uint64_t doi_fill_count; /* number of non-empty blocks */ 784fa9e4066Sahrens } dmu_object_info_t; 785fa9e4066Sahrens 786e45ce728Sahrens typedef void arc_byteswap_func_t(void *buf, size_t size); 787e45ce728Sahrens 788fa9e4066Sahrens typedef struct dmu_object_type_info { 789ad135b5dSChristopher Siden dmu_object_byteswap_t ot_byteswap; 790fa9e4066Sahrens boolean_t ot_metadata; 791fa9e4066Sahrens char *ot_name; 792fa9e4066Sahrens } dmu_object_type_info_t; 793fa9e4066Sahrens 794ad135b5dSChristopher Siden typedef struct dmu_object_byteswap_info { 795ad135b5dSChristopher Siden arc_byteswap_func_t *ob_func; 796ad135b5dSChristopher Siden char *ob_name; 797ad135b5dSChristopher Siden } dmu_object_byteswap_info_t; 798ad135b5dSChristopher Siden 799fa9e4066Sahrens extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 800ad135b5dSChristopher Siden extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; 801fa9e4066Sahrens 802fa9e4066Sahrens /* 803fa9e4066Sahrens * Get information on a DMU object. 804fa9e4066Sahrens * 805fa9e4066Sahrens * Return 0 on success or ENOENT if object is not allocated. 806fa9e4066Sahrens * 807fa9e4066Sahrens * If doi is NULL, just indicates whether the object exists. 808fa9e4066Sahrens */ 809fa9e4066Sahrens int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 8103e30c24aSWill Andrews /* Like dmu_object_info, but faster if you have a held dnode in hand. */ 81179d72832SMatthew Ahrens void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi); 8123e30c24aSWill Andrews /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ 813fa9e4066Sahrens void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 8143e30c24aSWill Andrews /* 8153e30c24aSWill Andrews * Like dmu_object_info_from_db, but faster still when you only care about 8163e30c24aSWill Andrews * the size. This is specifically optimized for zfs_getattr(). 8173e30c24aSWill Andrews */ 818fa9e4066Sahrens void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 819fa9e4066Sahrens u_longlong_t *nblk512); 820fa9e4066Sahrens 821fa9e4066Sahrens typedef struct dmu_objset_stats { 822a2eea2e1Sahrens uint64_t dds_num_clones; /* number of clones of this */ 823a2eea2e1Sahrens uint64_t dds_creation_txg; 8243cb34c60Sahrens uint64_t dds_guid; 825fa9e4066Sahrens dmu_objset_type_t dds_type; 826fa9e4066Sahrens uint8_t dds_is_snapshot; 82731fd60d3Sahrens uint8_t dds_inconsistent; 8289adfa60dSMatthew Ahrens char dds_origin[ZFS_MAX_DATASET_NAME_LEN]; 829a2eea2e1Sahrens } dmu_objset_stats_t; 830fa9e4066Sahrens 831a2eea2e1Sahrens /* 832a2eea2e1Sahrens * Get stats on a dataset. 833a2eea2e1Sahrens */ 834a2eea2e1Sahrens void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 835fa9e4066Sahrens 836a2eea2e1Sahrens /* 837a2eea2e1Sahrens * Add entries to the nvlist for all the objset's properties. See 838a2eea2e1Sahrens * zfs_prop_table[] and zfs(1m) for details on the properties. 839a2eea2e1Sahrens */ 840a2eea2e1Sahrens void dmu_objset_stats(objset_t *os, struct nvlist *nv); 841fa9e4066Sahrens 842a2eea2e1Sahrens /* 843a2eea2e1Sahrens * Get the space usage statistics for statvfs(). 844a2eea2e1Sahrens * 845a2eea2e1Sahrens * refdbytes is the amount of space "referenced" by this objset. 846a2eea2e1Sahrens * availbytes is the amount of space available to this objset, taking 847a2eea2e1Sahrens * into account quotas & reservations, assuming that no other objsets 848a2eea2e1Sahrens * use the space first. These values correspond to the 'referenced' and 849a2eea2e1Sahrens * 'available' properties, described in the zfs(1m) manpage. 850a2eea2e1Sahrens * 851a2eea2e1Sahrens * usedobjs and availobjs are the number of objects currently allocated, 852a2eea2e1Sahrens * and available. 853a2eea2e1Sahrens */ 854a2eea2e1Sahrens void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 855a2eea2e1Sahrens uint64_t *usedobjsp, uint64_t *availobjsp); 856fa9e4066Sahrens 857fa9e4066Sahrens /* 858a2eea2e1Sahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions. 859a2eea2e1Sahrens * (Contrast with the ds_guid which is a 64-bit ID that will never 860a2eea2e1Sahrens * change, so there is a small probability that it will collide.) 861fa9e4066Sahrens */ 862a2eea2e1Sahrens uint64_t dmu_objset_fsid_guid(objset_t *os); 863fa9e4066Sahrens 86471eb0538SChris Kirby /* 86571eb0538SChris Kirby * Get the [cm]time for an objset's snapshot dir 86671eb0538SChris Kirby */ 86771eb0538SChris Kirby timestruc_t dmu_objset_snap_cmtime(objset_t *os); 86871eb0538SChris Kirby 869fa9e4066Sahrens int dmu_objset_is_snapshot(objset_t *os); 870fa9e4066Sahrens 871fa9e4066Sahrens extern struct spa *dmu_objset_spa(objset_t *os); 872fa9e4066Sahrens extern struct zilog *dmu_objset_zil(objset_t *os); 873fa9e4066Sahrens extern struct dsl_pool *dmu_objset_pool(objset_t *os); 874fa9e4066Sahrens extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 875fa9e4066Sahrens extern void dmu_objset_name(objset_t *os, char *buf); 876fa9e4066Sahrens extern dmu_objset_type_t dmu_objset_type(objset_t *os); 877fa9e4066Sahrens extern uint64_t dmu_objset_id(objset_t *os); 878edf345e6SMatthew Ahrens extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); 879edf345e6SMatthew Ahrens extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); 880fa9e4066Sahrens extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, 881b38f0970Sck uint64_t *id, uint64_t *offp, boolean_t *case_conflict); 882ab04eb8eStimh extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, 883ab04eb8eStimh int maxlen, boolean_t *conflict); 88487e5029aSahrens extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, 88587e5029aSahrens uint64_t *idp, uint64_t *offp); 88614843421SMatthew Ahrens 8879966ca11SMatthew Ahrens typedef int objset_used_cb_t(dmu_object_type_t bonustype, 8889966ca11SMatthew Ahrens void *bonus, uint64_t *userp, uint64_t *groupp); 88914843421SMatthew Ahrens extern void dmu_objset_register_type(dmu_objset_type_t ost, 89014843421SMatthew Ahrens objset_used_cb_t *cb); 891f18faf3fSek extern void dmu_objset_set_user(objset_t *os, void *user_ptr); 892f18faf3fSek extern void *dmu_objset_get_user(objset_t *os); 893fa9e4066Sahrens 894fa9e4066Sahrens /* 895fa9e4066Sahrens * Return the txg number for the given assigned transaction. 896fa9e4066Sahrens */ 89787e5029aSahrens uint64_t dmu_tx_get_txg(dmu_tx_t *tx); 898fa9e4066Sahrens 899fa9e4066Sahrens /* 900fa9e4066Sahrens * Synchronous write. 901c5c6ffa0Smaybee * If a parent zio is provided this function initiates a write on the 902c5c6ffa0Smaybee * provided buffer as a child of the parent zio. 903da6c28aaSamw * In the absence of a parent zio, the write is completed synchronously. 904c5c6ffa0Smaybee * At write completion, blk is filled with the bp of the written block. 905fa9e4066Sahrens * Note that while the data covered by this function will be on stable 906c5c6ffa0Smaybee * storage when the write completes this new data does not become a 907fa9e4066Sahrens * permanent part of the file until the associated transaction commits. 908fa9e4066Sahrens */ 909b24ab676SJeff Bonwick 910b24ab676SJeff Bonwick /* 911b24ab676SJeff Bonwick * {zfs,zvol,ztest}_get_done() args 912b24ab676SJeff Bonwick */ 913b24ab676SJeff Bonwick typedef struct zgd { 914b24ab676SJeff Bonwick struct zilog *zgd_zilog; 915b24ab676SJeff Bonwick struct blkptr *zgd_bp; 916b24ab676SJeff Bonwick dmu_buf_t *zgd_db; 917b24ab676SJeff Bonwick struct rl *zgd_rl; 918b24ab676SJeff Bonwick void *zgd_private; 919b24ab676SJeff Bonwick } zgd_t; 920b24ab676SJeff Bonwick 921b24ab676SJeff Bonwick typedef void dmu_sync_cb_t(zgd_t *arg, int error); 922b24ab676SJeff Bonwick int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); 923fa9e4066Sahrens 924fa9e4066Sahrens /* 925fa9e4066Sahrens * Find the next hole or data block in file starting at *off 926fa9e4066Sahrens * Return found offset in *off. Return ESRCH for end of file. 927fa9e4066Sahrens */ 928fa9e4066Sahrens int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, 929fa9e4066Sahrens uint64_t *off); 930fa9e4066Sahrens 9312bcf0248SMax Grossman /* 9322bcf0248SMax Grossman * Check if a DMU object has any dirty blocks. If so, sync out 9332bcf0248SMax Grossman * all pending transaction groups. Otherwise, this function 9342bcf0248SMax Grossman * does not alter DMU state. This could be improved to only sync 9352bcf0248SMax Grossman * out the necessary transaction groups for this particular 9362bcf0248SMax Grossman * object. 9372bcf0248SMax Grossman */ 9382bcf0248SMax Grossman int dmu_object_wait_synced(objset_t *os, uint64_t object); 9392bcf0248SMax Grossman 940fa9e4066Sahrens /* 941fa9e4066Sahrens * Initial setup and final teardown. 942fa9e4066Sahrens */ 943fa9e4066Sahrens extern void dmu_init(void); 944fa9e4066Sahrens extern void dmu_fini(void); 945fa9e4066Sahrens 946fa9e4066Sahrens typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, 947fa9e4066Sahrens uint64_t object, uint64_t offset, int len); 948fa9e4066Sahrens void dmu_traverse_objset(objset_t *os, uint64_t txg_start, 949fa9e4066Sahrens dmu_traverse_cb_t cb, void *arg); 950fa9e4066Sahrens 9513b2aab18SMatthew Ahrens int dmu_diff(const char *tosnap_name, const char *fromsnap_name, 9523b2aab18SMatthew Ahrens struct vnode *vp, offset_t *offp); 95399d5e173STim Haley 954fa9e4066Sahrens /* CRC64 table */ 955fa9e4066Sahrens #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 956fa9e4066Sahrens extern uint64_t zfs_crc64_table[256]; 957fa9e4066Sahrens 958e4be62a2SSaso Kiselkov extern int zfs_mdcomp_disable; 959e4be62a2SSaso Kiselkov 960fa9e4066Sahrens #ifdef __cplusplus 961fa9e4066Sahrens } 962fa9e4066Sahrens #endif 963fa9e4066Sahrens 964fa9e4066Sahrens #endif /* _SYS_DMU_H */ 965