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> 51a2cdcdd2SPaul Dagnelie #include <sys/zio_priority.h> 52fa9e4066Sahrens 53fa9e4066Sahrens #ifdef __cplusplus 54fa9e4066Sahrens extern "C" { 55fa9e4066Sahrens #endif 56fa9e4066Sahrens 57fa9e4066Sahrens struct uio; 58c242f9a0Schunli zhang - Sun Microsystems - Irvine United States struct xuio; 5944eda4d7Smaybee struct page; 60fa9e4066Sahrens struct vnode; 61fa9e4066Sahrens struct spa; 62fa9e4066Sahrens struct zilog; 63fa9e4066Sahrens struct zio; 64fa9e4066Sahrens struct blkptr; 65fa9e4066Sahrens struct zap_cursor; 66fa9e4066Sahrens struct dsl_dataset; 67fa9e4066Sahrens struct dsl_pool; 68fa9e4066Sahrens struct dnode; 69fa9e4066Sahrens struct drr_begin; 70fa9e4066Sahrens struct drr_end; 717802d7bfSMatthew Ahrens struct zbookmark_phys; 7244cd46caSbillm struct spa; 73a2eea2e1Sahrens struct nvlist; 742fdbea25SAleksandr Guzovskiy struct arc_buf; 75b24ab676SJeff Bonwick struct zio_prop; 760a586ceaSMark Shellenbaum struct sa_handle; 77fa9e4066Sahrens 78fa9e4066Sahrens typedef struct objset objset_t; 79fa9e4066Sahrens typedef struct dmu_tx dmu_tx_t; 80fa9e4066Sahrens typedef struct dsl_dir dsl_dir_t; 81*79d72832SMatthew Ahrens typedef struct dnode dnode_t; 82fa9e4066Sahrens 83ad135b5dSChristopher Siden typedef enum dmu_object_byteswap { 84ad135b5dSChristopher Siden DMU_BSWAP_UINT8, 85ad135b5dSChristopher Siden DMU_BSWAP_UINT16, 86ad135b5dSChristopher Siden DMU_BSWAP_UINT32, 87ad135b5dSChristopher Siden DMU_BSWAP_UINT64, 88ad135b5dSChristopher Siden DMU_BSWAP_ZAP, 89ad135b5dSChristopher Siden DMU_BSWAP_DNODE, 90ad135b5dSChristopher Siden DMU_BSWAP_OBJSET, 91ad135b5dSChristopher Siden DMU_BSWAP_ZNODE, 92ad135b5dSChristopher Siden DMU_BSWAP_OLDACL, 93ad135b5dSChristopher Siden DMU_BSWAP_ACL, 94ad135b5dSChristopher Siden /* 95ad135b5dSChristopher Siden * Allocating a new byteswap type number makes the on-disk format 96ad135b5dSChristopher Siden * incompatible with any other format that uses the same number. 97ad135b5dSChristopher Siden * 98ad135b5dSChristopher Siden * Data can usually be structured to work with one of the 99ad135b5dSChristopher Siden * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 100ad135b5dSChristopher Siden */ 101ad135b5dSChristopher Siden DMU_BSWAP_NUMFUNCS 102ad135b5dSChristopher Siden } dmu_object_byteswap_t; 103ad135b5dSChristopher Siden 104ad135b5dSChristopher Siden #define DMU_OT_NEWTYPE 0x80 105ad135b5dSChristopher Siden #define DMU_OT_METADATA 0x40 106ad135b5dSChristopher Siden #define DMU_OT_BYTESWAP_MASK 0x3f 107ad135b5dSChristopher Siden 108ad135b5dSChristopher Siden /* 109ad135b5dSChristopher Siden * Defines a uint8_t object type. Object types specify if the data 110ad135b5dSChristopher Siden * in the object is metadata (boolean) and how to byteswap the data 111ad135b5dSChristopher Siden * (dmu_object_byteswap_t). 112ad135b5dSChristopher Siden */ 113ad135b5dSChristopher Siden #define DMU_OT(byteswap, metadata) \ 114ad135b5dSChristopher Siden (DMU_OT_NEWTYPE | \ 115ad135b5dSChristopher Siden ((metadata) ? DMU_OT_METADATA : 0) | \ 116ad135b5dSChristopher Siden ((byteswap) & DMU_OT_BYTESWAP_MASK)) 117ad135b5dSChristopher Siden 118ad135b5dSChristopher Siden #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 119ad135b5dSChristopher Siden ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ 120ad135b5dSChristopher Siden (ot) < DMU_OT_NUMTYPES) 121ad135b5dSChristopher Siden 122ad135b5dSChristopher Siden #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 123ad135b5dSChristopher Siden ((ot) & DMU_OT_METADATA) : \ 124ad135b5dSChristopher Siden dmu_ot[(ot)].ot_metadata) 125ad135b5dSChristopher Siden 1265d7b4d43SMatthew Ahrens /* 1275d7b4d43SMatthew Ahrens * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't 1285d7b4d43SMatthew Ahrens * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill 1295d7b4d43SMatthew Ahrens * is repurposed for embedded BPs. 1305d7b4d43SMatthew Ahrens */ 1315d7b4d43SMatthew Ahrens #define DMU_OT_HAS_FILL(ot) \ 1325d7b4d43SMatthew Ahrens ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) 1335d7b4d43SMatthew Ahrens 134ad135b5dSChristopher Siden #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 135ad135b5dSChristopher Siden ((ot) & DMU_OT_BYTESWAP_MASK) : \ 136ad135b5dSChristopher Siden dmu_ot[(ot)].ot_byteswap) 137ad135b5dSChristopher Siden 138fa9e4066Sahrens typedef enum dmu_object_type { 139fa9e4066Sahrens DMU_OT_NONE, 140fa9e4066Sahrens /* general: */ 141fa9e4066Sahrens DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 142fa9e4066Sahrens DMU_OT_OBJECT_ARRAY, /* UINT64 */ 143fa9e4066Sahrens DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 144fa9e4066Sahrens DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 145cde58dbcSMatthew Ahrens DMU_OT_BPOBJ, /* UINT64 */ 146cde58dbcSMatthew Ahrens DMU_OT_BPOBJ_HDR, /* UINT64 */ 147fa9e4066Sahrens /* spa: */ 148fa9e4066Sahrens DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 149fa9e4066Sahrens DMU_OT_SPACE_MAP, /* UINT64 */ 150fa9e4066Sahrens /* zil: */ 151fa9e4066Sahrens DMU_OT_INTENT_LOG, /* UINT64 */ 152fa9e4066Sahrens /* dmu: */ 153fa9e4066Sahrens DMU_OT_DNODE, /* DNODE */ 154fa9e4066Sahrens DMU_OT_OBJSET, /* OBJSET */ 155fa9e4066Sahrens /* dsl: */ 1561649cd4bStabriz DMU_OT_DSL_DIR, /* UINT64 */ 15787e5029aSahrens DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 15887e5029aSahrens DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 159fa9e4066Sahrens DMU_OT_DSL_PROPS, /* ZAP */ 1601649cd4bStabriz DMU_OT_DSL_DATASET, /* UINT64 */ 161fa9e4066Sahrens /* zpl: */ 162fa9e4066Sahrens DMU_OT_ZNODE, /* ZNODE */ 163da6c28aaSamw DMU_OT_OLDACL, /* Old ACL */ 164fa9e4066Sahrens DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 165fa9e4066Sahrens DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 166fa9e4066Sahrens DMU_OT_MASTER_NODE, /* ZAP */ 167893a6d32Sahrens DMU_OT_UNLINKED_SET, /* ZAP */ 168fa9e4066Sahrens /* zvol: */ 169fa9e4066Sahrens DMU_OT_ZVOL, /* UINT8 */ 170fa9e4066Sahrens DMU_OT_ZVOL_PROP, /* ZAP */ 171fa9e4066Sahrens /* other; for testing only! */ 172fa9e4066Sahrens DMU_OT_PLAIN_OTHER, /* UINT8 */ 173fa9e4066Sahrens DMU_OT_UINT64_OTHER, /* UINT64 */ 174fa9e4066Sahrens DMU_OT_ZAP_OTHER, /* ZAP */ 175ea8dc4b6Seschrock /* new object types: */ 176ea8dc4b6Seschrock DMU_OT_ERROR_LOG, /* ZAP */ 17706eeb2adSek DMU_OT_SPA_HISTORY, /* UINT8 */ 17806eeb2adSek DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 179b1b8ab34Slling DMU_OT_POOL_PROPS, /* ZAP */ 180ecd6cf80Smarks DMU_OT_DSL_PERMS, /* ZAP */ 181da6c28aaSamw DMU_OT_ACL, /* ACL */ 182da6c28aaSamw DMU_OT_SYSACL, /* SYSACL */ 183da6c28aaSamw DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 184da6c28aaSamw DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 185088f3894Sahrens DMU_OT_NEXT_CLONES, /* ZAP */ 1863f9d6ad7SLin Ling DMU_OT_SCAN_QUEUE, /* ZAP */ 18714843421SMatthew Ahrens DMU_OT_USERGROUP_USED, /* ZAP */ 18814843421SMatthew Ahrens DMU_OT_USERGROUP_QUOTA, /* ZAP */ 189842727c2SChris Kirby DMU_OT_USERREFS, /* ZAP */ 190b24ab676SJeff Bonwick DMU_OT_DDT_ZAP, /* ZAP */ 191b24ab676SJeff Bonwick DMU_OT_DDT_STATS, /* ZAP */ 1920a586ceaSMark Shellenbaum DMU_OT_SA, /* System attr */ 1930a586ceaSMark Shellenbaum DMU_OT_SA_MASTER_NODE, /* ZAP */ 1940a586ceaSMark Shellenbaum DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 1950a586ceaSMark Shellenbaum DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 1963f9d6ad7SLin Ling DMU_OT_SCAN_XLATE, /* ZAP */ 1973f9d6ad7SLin Ling DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 198cde58dbcSMatthew Ahrens DMU_OT_DEADLIST, /* ZAP */ 199cde58dbcSMatthew Ahrens DMU_OT_DEADLIST_HDR, /* UINT64 */ 200cde58dbcSMatthew Ahrens DMU_OT_DSL_CLONES, /* ZAP */ 201cde58dbcSMatthew Ahrens DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 202ad135b5dSChristopher Siden /* 203ad135b5dSChristopher Siden * Do not allocate new object types here. Doing so makes the on-disk 204ad135b5dSChristopher Siden * format incompatible with any other format that uses the same object 205ad135b5dSChristopher Siden * type number. 206ad135b5dSChristopher Siden * 207ad135b5dSChristopher Siden * When creating an object which does not have one of the above types 208ad135b5dSChristopher Siden * use the DMU_OTN_* type with the correct byteswap and metadata 209ad135b5dSChristopher Siden * values. 210ad135b5dSChristopher Siden * 211ad135b5dSChristopher Siden * The DMU_OTN_* types do not have entries in the dmu_ot table, 212ad135b5dSChristopher Siden * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead 213ad135b5dSChristopher Siden * of indexing into dmu_ot directly (this works for both DMU_OT_* types 214ad135b5dSChristopher Siden * and DMU_OTN_* types). 215ad135b5dSChristopher Siden */ 216ad135b5dSChristopher Siden DMU_OT_NUMTYPES, 217ad135b5dSChristopher Siden 218ad135b5dSChristopher Siden /* 219ad135b5dSChristopher Siden * Names for valid types declared with DMU_OT(). 220ad135b5dSChristopher Siden */ 221ad135b5dSChristopher Siden DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE), 222ad135b5dSChristopher Siden DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE), 223ad135b5dSChristopher Siden DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE), 224ad135b5dSChristopher Siden DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE), 225ad135b5dSChristopher Siden DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE), 226ad135b5dSChristopher Siden DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE), 227ad135b5dSChristopher Siden DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE), 228ad135b5dSChristopher Siden DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE), 229ad135b5dSChristopher Siden DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE), 230ad135b5dSChristopher Siden DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE), 231fa9e4066Sahrens } dmu_object_type_t; 232fa9e4066Sahrens 2333b2aab18SMatthew Ahrens typedef enum txg_how { 2343b2aab18SMatthew Ahrens TXG_WAIT = 1, 2353b2aab18SMatthew Ahrens TXG_NOWAIT, 23669962b56SMatthew Ahrens TXG_WAITED, 2373b2aab18SMatthew Ahrens } txg_how_t; 2383b2aab18SMatthew Ahrens 239fa9e4066Sahrens void byteswap_uint64_array(void *buf, size_t size); 240fa9e4066Sahrens void byteswap_uint32_array(void *buf, size_t size); 241fa9e4066Sahrens void byteswap_uint16_array(void *buf, size_t size); 242fa9e4066Sahrens void byteswap_uint8_array(void *buf, size_t size); 243fa9e4066Sahrens void zap_byteswap(void *buf, size_t size); 244da6c28aaSamw void zfs_oldacl_byteswap(void *buf, size_t size); 245fa9e4066Sahrens void zfs_acl_byteswap(void *buf, size_t size); 246fa9e4066Sahrens void zfs_znode_byteswap(void *buf, size_t size); 247fa9e4066Sahrens 2480b69c2f0Sahrens #define DS_FIND_SNAPSHOTS (1<<0) 2490b69c2f0Sahrens #define DS_FIND_CHILDREN (1<<1) 25012380e1eSArne Jansen #define DS_FIND_SERIALIZE (1<<2) 251fa9e4066Sahrens 252fa9e4066Sahrens /* 253fa9e4066Sahrens * The maximum number of bytes that can be accessed as part of one 254fa9e4066Sahrens * operation, including metadata. 255fa9e4066Sahrens */ 256b5152584SMatthew Ahrens #define DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */ 257cdb0ab79Smaybee #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ 258fa9e4066Sahrens 25914843421SMatthew Ahrens #define DMU_USERUSED_OBJECT (-1ULL) 26014843421SMatthew Ahrens #define DMU_GROUPUSED_OBJECT (-2ULL) 26114843421SMatthew Ahrens 2620a586ceaSMark Shellenbaum /* 2630a586ceaSMark Shellenbaum * artificial blkids for bonus buffer and spill blocks 2640a586ceaSMark Shellenbaum */ 2650a586ceaSMark Shellenbaum #define DMU_BONUS_BLKID (-1ULL) 2660a586ceaSMark Shellenbaum #define DMU_SPILL_BLKID (-2ULL) 267fa9e4066Sahrens /* 268fa9e4066Sahrens * Public routines to create, destroy, open, and close objsets. 269fa9e4066Sahrens */ 270503ad85cSMatthew Ahrens int dmu_objset_hold(const char *name, void *tag, objset_t **osp); 271503ad85cSMatthew Ahrens int dmu_objset_own(const char *name, dmu_objset_type_t type, 272503ad85cSMatthew Ahrens boolean_t readonly, void *tag, objset_t **osp); 273503ad85cSMatthew Ahrens void dmu_objset_rele(objset_t *os, void *tag); 274503ad85cSMatthew Ahrens void dmu_objset_disown(objset_t *os, void *tag); 275503ad85cSMatthew Ahrens int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 276503ad85cSMatthew Ahrens 2773b2aab18SMatthew Ahrens void dmu_objset_evict_dbufs(objset_t *os); 278ae46e4c7SMatthew Ahrens int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 279ecd6cf80Smarks void (*func)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx), void *arg); 2803b2aab18SMatthew Ahrens int dmu_objset_clone(const char *name, const char *origin); 2813b2aab18SMatthew Ahrens int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, 2824445fffbSMatthew Ahrens struct nvlist *errlist); 2834445fffbSMatthew Ahrens int dmu_objset_snapshot_one(const char *fsname, const char *snapname); 2844445fffbSMatthew Ahrens int dmu_objset_snapshot_tmp(const char *, const char *, int); 285fd136879SMatthew Ahrens int dmu_objset_find(char *name, int func(const char *, void *), void *arg, 286fa9e4066Sahrens int flags); 287fa9e4066Sahrens void dmu_objset_byteswap(void *buf, size_t size); 2883b2aab18SMatthew Ahrens int dsl_dataset_rename_snapshot(const char *fsname, 2893b2aab18SMatthew Ahrens const char *oldsnapname, const char *newsnapname, boolean_t recursive); 290fa9e4066Sahrens 291fa9e4066Sahrens typedef struct dmu_buf { 292fa9e4066Sahrens uint64_t db_object; /* object that this buffer is part of */ 293fa9e4066Sahrens uint64_t db_offset; /* byte offset in this object */ 294fa9e4066Sahrens uint64_t db_size; /* size of buffer in bytes */ 295fa9e4066Sahrens void *db_data; /* data in buffer */ 296fa9e4066Sahrens } dmu_buf_t; 297fa9e4066Sahrens 29899653d4eSeschrock /* 29999653d4eSeschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 30099653d4eSeschrock */ 301fa9e4066Sahrens #define DMU_POOL_DIRECTORY_OBJECT 1 302fa9e4066Sahrens #define DMU_POOL_CONFIG "config" 303ad135b5dSChristopher Siden #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 304ad135b5dSChristopher Siden #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 305ad135b5dSChristopher Siden #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 30643466aaeSMax Grossman #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" 307fa9e4066Sahrens #define DMU_POOL_ROOT_DATASET "root_dataset" 308cde58dbcSMatthew Ahrens #define DMU_POOL_SYNC_BPOBJ "sync_bplist" 309ea8dc4b6Seschrock #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 310ea8dc4b6Seschrock #define DMU_POOL_ERRLOG_LAST "errlog_last" 31199653d4eSeschrock #define DMU_POOL_SPARES "spares" 31299653d4eSeschrock #define DMU_POOL_DEFLATE "deflate" 31306eeb2adSek #define DMU_POOL_HISTORY "history" 314b1b8ab34Slling #define DMU_POOL_PROPS "pool_props" 315fa94a07fSbrendan #define DMU_POOL_L2CACHE "l2cache" 316ca45db41SChris Kirby #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 317b24ab676SJeff Bonwick #define DMU_POOL_DDT "DDT-%s-%s-%s" 318b24ab676SJeff Bonwick #define DMU_POOL_DDT_STATS "DDT-statistics" 3193f9d6ad7SLin Ling #define DMU_POOL_CREATION_VERSION "creation_version" 3203f9d6ad7SLin Ling #define DMU_POOL_SCAN "scan" 321cde58dbcSMatthew Ahrens #define DMU_POOL_FREE_BPOBJ "free_bpobj" 322ad135b5dSChristopher Siden #define DMU_POOL_BPTREE_OBJ "bptree_obj" 323f1745736SMatthew Ahrens #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 32445818ee1SMatthew Ahrens #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 325215198a6SJoe Stein #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map" 326088f3894Sahrens 327fa9e4066Sahrens /* 328fa9e4066Sahrens * Allocate an object from this objset. The range of object numbers 329fa9e4066Sahrens * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 330fa9e4066Sahrens * 331fa9e4066Sahrens * The transaction must be assigned to a txg. The newly allocated 332fa9e4066Sahrens * object will be "held" in the transaction (ie. you can modify the 333fa9e4066Sahrens * newly allocated object in this transaction). 334fa9e4066Sahrens * 335fa9e4066Sahrens * dmu_object_alloc() chooses an object and returns it in *objectp. 336fa9e4066Sahrens * 337fa9e4066Sahrens * dmu_object_claim() allocates a specific object number. If that 338fa9e4066Sahrens * number is already allocated, it fails and returns EEXIST. 339fa9e4066Sahrens * 340fa9e4066Sahrens * Return 0 on success, or ENOSPC or EEXIST as specified above. 341fa9e4066Sahrens */ 342fa9e4066Sahrens uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 343fa9e4066Sahrens int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 344fa9e4066Sahrens int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 345fa9e4066Sahrens int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 346fa9e4066Sahrens int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 347e77d42eaSMatthew Ahrens int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); 348fa9e4066Sahrens 349fa9e4066Sahrens /* 350fa9e4066Sahrens * Free an object from this objset. 351fa9e4066Sahrens * 352fa9e4066Sahrens * The object's data will be freed as well (ie. you don't need to call 353fa9e4066Sahrens * dmu_free(object, 0, -1, tx)). 354fa9e4066Sahrens * 355fa9e4066Sahrens * The object need not be held in the transaction. 356fa9e4066Sahrens * 357fa9e4066Sahrens * If there are any holds on this object's buffers (via dmu_buf_hold()), 358fa9e4066Sahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not 359fa9e4066Sahrens * free it; it fails and returns EBUSY. 360fa9e4066Sahrens * 361fa9e4066Sahrens * If the object is not allocated, it fails and returns ENOENT. 362fa9e4066Sahrens * 363fa9e4066Sahrens * Return 0 on success, or EBUSY or ENOENT as specified above. 364fa9e4066Sahrens */ 365fa9e4066Sahrens int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 366fa9e4066Sahrens 367fa9e4066Sahrens /* 368fa9e4066Sahrens * Find the next allocated or free object. 369fa9e4066Sahrens * 370fa9e4066Sahrens * The objectp parameter is in-out. It will be updated to be the next 3716754306eSahrens * object which is allocated. Ignore objects which have not been 3726754306eSahrens * modified since txg. 373fa9e4066Sahrens * 374fa9e4066Sahrens * XXX Can only be called on a objset with no dirty data. 375fa9e4066Sahrens * 376fa9e4066Sahrens * Returns 0 on success, or ENOENT if there are no more objects. 377fa9e4066Sahrens */ 3786754306eSahrens int dmu_object_next(objset_t *os, uint64_t *objectp, 3796754306eSahrens boolean_t hole, uint64_t txg); 380fa9e4066Sahrens 381fa9e4066Sahrens /* 382fa9e4066Sahrens * Set the data blocksize for an object. 383fa9e4066Sahrens * 384fa9e4066Sahrens * The object cannot have any blocks allcated beyond the first. If 385fa9e4066Sahrens * the first block is allocated already, the new size must be greater 386fa9e4066Sahrens * than the current block size. If these conditions are not met, 387fa9e4066Sahrens * ENOTSUP will be returned. 388fa9e4066Sahrens * 389fa9e4066Sahrens * Returns 0 on success, or EBUSY if there are any holds on the object 390fa9e4066Sahrens * contents, or ENOTSUP as described above. 391fa9e4066Sahrens */ 392fa9e4066Sahrens int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 393fa9e4066Sahrens int ibs, dmu_tx_t *tx); 394fa9e4066Sahrens 395fa9e4066Sahrens /* 396fa9e4066Sahrens * Set the checksum property on a dnode. The new checksum algorithm will 397fa9e4066Sahrens * apply to all newly written blocks; existing blocks will not be affected. 398fa9e4066Sahrens */ 399fa9e4066Sahrens void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 400fa9e4066Sahrens dmu_tx_t *tx); 401fa9e4066Sahrens 402fa9e4066Sahrens /* 403fa9e4066Sahrens * Set the compress property on a dnode. The new compression algorithm will 404fa9e4066Sahrens * apply to all newly written blocks; existing blocks will not be affected. 405fa9e4066Sahrens */ 406fa9e4066Sahrens void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 407fa9e4066Sahrens dmu_tx_t *tx); 408fa9e4066Sahrens 4095d7b4d43SMatthew Ahrens void 4105d7b4d43SMatthew Ahrens dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, 4115d7b4d43SMatthew Ahrens void *data, uint8_t etype, uint8_t comp, int uncompressed_size, 4125d7b4d43SMatthew Ahrens int compressed_size, int byteorder, dmu_tx_t *tx); 4135d7b4d43SMatthew Ahrens 41444cd46caSbillm /* 415b24ab676SJeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc. 41644cd46caSbillm */ 417b24ab676SJeff Bonwick #define WP_NOFILL 0x1 418b24ab676SJeff Bonwick #define WP_DMU_SYNC 0x2 4190a586ceaSMark Shellenbaum #define WP_SPILL 0x4 420b24ab676SJeff Bonwick 421*79d72832SMatthew Ahrens void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, 422b24ab676SJeff Bonwick struct zio_prop *zp); 423fa9e4066Sahrens /* 424fa9e4066Sahrens * The bonus data is accessed more or less like a regular buffer. 425fa9e4066Sahrens * You must dmu_bonus_hold() to get the buffer, which will give you a 426fa9e4066Sahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 427fa9e4066Sahrens * data. As with any normal buffer, you must call dmu_buf_read() to 428fa9e4066Sahrens * read db_data, dmu_buf_will_dirty() before modifying it, and the 429fa9e4066Sahrens * object must be held in an assigned transaction before calling 430fa9e4066Sahrens * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 431fa9e4066Sahrens * buffer as well. You must release your hold with dmu_buf_rele(). 4323e30c24aSWill Andrews * 4333e30c24aSWill Andrews * Returns ENOENT, EIO, or 0. 434fa9e4066Sahrens */ 435ea8dc4b6Seschrock int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); 436fa9e4066Sahrens int dmu_bonus_max(void); 4371934e92fSmaybee int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 4380a586ceaSMark Shellenbaum int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 439744947dcSTom Erickson dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); 4400a586ceaSMark Shellenbaum int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 4410a586ceaSMark Shellenbaum 4420a586ceaSMark Shellenbaum /* 4430a586ceaSMark Shellenbaum * Special spill buffer support used by "SA" framework 4440a586ceaSMark Shellenbaum */ 4450a586ceaSMark Shellenbaum 4460a586ceaSMark Shellenbaum int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 447*79d72832SMatthew Ahrens int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, 4480a586ceaSMark Shellenbaum void *tag, dmu_buf_t **dbp); 4490a586ceaSMark Shellenbaum int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 450fa9e4066Sahrens 451fa9e4066Sahrens /* 452fa9e4066Sahrens * Obtain the DMU buffer from the specified object which contains the 453fa9e4066Sahrens * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 454fa9e4066Sahrens * that it will remain in memory. You must release the hold with 455fa9e4066Sahrens * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your 456fa9e4066Sahrens * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 457fa9e4066Sahrens * 458fa9e4066Sahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 459fa9e4066Sahrens * on the returned buffer before reading or writing the buffer's 460fa9e4066Sahrens * db_data. The comments for those routines describe what particular 461fa9e4066Sahrens * operations are valid after calling them. 462fa9e4066Sahrens * 463fa9e4066Sahrens * The object number must be a valid, allocated object number. 464fa9e4066Sahrens */ 465ea8dc4b6Seschrock int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 46647cb52daSJeff Bonwick void *tag, dmu_buf_t **, int flags); 467*79d72832SMatthew Ahrens int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, 468*79d72832SMatthew Ahrens void *tag, dmu_buf_t **dbp, int flags); 469e57a022bSJustin T. Gibbs 470e57a022bSJustin T. Gibbs /* 471e57a022bSJustin T. Gibbs * Add a reference to a dmu buffer that has already been held via 472e57a022bSJustin T. Gibbs * dmu_buf_hold() in the current context. 473e57a022bSJustin T. Gibbs */ 474fa9e4066Sahrens void dmu_buf_add_ref(dmu_buf_t *db, void* tag); 475e57a022bSJustin T. Gibbs 476e57a022bSJustin T. Gibbs /* 477e57a022bSJustin T. Gibbs * Attempt to add a reference to a dmu buffer that is in an unknown state, 478e57a022bSJustin T. Gibbs * using a pointer that may have been invalidated by eviction processing. 479e57a022bSJustin T. Gibbs * The request will succeed if the passed in dbuf still represents the 480e57a022bSJustin T. Gibbs * same os/object/blkid, is ineligible for eviction, and has at least 481e57a022bSJustin T. Gibbs * one hold by a user other than the syncer. 482e57a022bSJustin T. Gibbs */ 483e57a022bSJustin T. Gibbs boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, 484e57a022bSJustin T. Gibbs uint64_t blkid, void *tag); 485e57a022bSJustin T. Gibbs 486ea8dc4b6Seschrock void dmu_buf_rele(dmu_buf_t *db, void *tag); 487fa9e4066Sahrens uint64_t dmu_buf_refcount(dmu_buf_t *db); 488fa9e4066Sahrens 489fa9e4066Sahrens /* 490fa9e4066Sahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 491fa9e4066Sahrens * range of an object. A pointer to an array of dmu_buf_t*'s is 492fa9e4066Sahrens * returned (in *dbpp). 493fa9e4066Sahrens * 494fa9e4066Sahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 495fa9e4066Sahrens * frees the array. The hold on the array of buffers MUST be released 496fa9e4066Sahrens * with dmu_buf_rele_array. You can NOT release the hold on each buffer 497fa9e4066Sahrens * individually with dmu_buf_rele. 498fa9e4066Sahrens */ 49913506d1eSmaybee int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 500cf6106c8SMatthew Ahrens uint64_t length, boolean_t read, void *tag, 501cf6106c8SMatthew Ahrens int *numbufsp, dmu_buf_t ***dbpp); 502ea8dc4b6Seschrock void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); 503fa9e4066Sahrens 504bc9014e6SJustin Gibbs typedef void dmu_buf_evict_func_t(void *user_ptr); 505bc9014e6SJustin Gibbs 506fa9e4066Sahrens /* 507bc9014e6SJustin Gibbs * A DMU buffer user object may be associated with a dbuf for the 508bc9014e6SJustin Gibbs * duration of its lifetime. This allows the user of a dbuf (client) 509bc9014e6SJustin Gibbs * to attach private data to a dbuf (e.g. in-core only data such as a 510bc9014e6SJustin Gibbs * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified 511bc9014e6SJustin Gibbs * when that dbuf has been evicted. Clients typically respond to the 512bc9014e6SJustin Gibbs * eviction notification by freeing their private data, thus ensuring 513bc9014e6SJustin Gibbs * the same lifetime for both dbuf and private data. 514fa9e4066Sahrens * 515bc9014e6SJustin Gibbs * The mapping from a dmu_buf_user_t to any client private data is the 516bc9014e6SJustin Gibbs * client's responsibility. All current consumers of the API with private 517bc9014e6SJustin Gibbs * data embed a dmu_buf_user_t as the first member of the structure for 518bc9014e6SJustin Gibbs * their private data. This allows conversions between the two types 519bc9014e6SJustin Gibbs * with a simple cast. Since the DMU buf user API never needs access 520bc9014e6SJustin Gibbs * to the private data, other strategies can be employed if necessary 521bc9014e6SJustin Gibbs * or convenient for the client (e.g. using container_of() to do the 522bc9014e6SJustin Gibbs * conversion for private data that cannot have the dmu_buf_user_t as 523bc9014e6SJustin Gibbs * its first member). 524fa9e4066Sahrens * 525bc9014e6SJustin Gibbs * Eviction callbacks are executed without the dbuf mutex held or any 526bc9014e6SJustin Gibbs * other type of mechanism to guarantee that the dbuf is still available. 527bc9014e6SJustin Gibbs * For this reason, users must assume the dbuf has already been freed 528bc9014e6SJustin Gibbs * and not reference the dbuf from the callback context. 529fa9e4066Sahrens * 530bc9014e6SJustin Gibbs * Users requesting "immediate eviction" are notified as soon as the dbuf 531bc9014e6SJustin Gibbs * is only referenced by dirty records (dirties == holds). Otherwise the 532bc9014e6SJustin Gibbs * notification occurs after eviction processing for the dbuf begins. 533fa9e4066Sahrens */ 534bc9014e6SJustin Gibbs typedef struct dmu_buf_user { 535bc9014e6SJustin Gibbs /* 536bc9014e6SJustin Gibbs * Asynchronous user eviction callback state. 537bc9014e6SJustin Gibbs */ 538bc9014e6SJustin Gibbs taskq_ent_t dbu_tqent; 539bc9014e6SJustin Gibbs 540bc9014e6SJustin Gibbs /* This instance's eviction function pointer. */ 541bc9014e6SJustin Gibbs dmu_buf_evict_func_t *dbu_evict_func; 542bc9014e6SJustin Gibbs #ifdef ZFS_DEBUG 543bc9014e6SJustin Gibbs /* 544bc9014e6SJustin Gibbs * Pointer to user's dbuf pointer. NULL for clients that do 545bc9014e6SJustin Gibbs * not associate a dbuf with their user data. 546bc9014e6SJustin Gibbs * 547bc9014e6SJustin Gibbs * The dbuf pointer is cleared upon eviction so as to catch 548bc9014e6SJustin Gibbs * use-after-evict bugs in clients. 549bc9014e6SJustin Gibbs */ 550bc9014e6SJustin Gibbs dmu_buf_t **dbu_clear_on_evict_dbufp; 551bc9014e6SJustin Gibbs #endif 552bc9014e6SJustin Gibbs } dmu_buf_user_t; 553bc9014e6SJustin Gibbs 554fa9e4066Sahrens /* 555bc9014e6SJustin Gibbs * Initialize the given dmu_buf_user_t instance with the eviction function 556bc9014e6SJustin Gibbs * evict_func, to be called when the user is evicted. 557bc9014e6SJustin Gibbs * 558bc9014e6SJustin Gibbs * NOTE: This function should only be called once on a given dmu_buf_user_t. 559bc9014e6SJustin Gibbs * To allow enforcement of this, dbu must already be zeroed on entry. 560fa9e4066Sahrens */ 561bc9014e6SJustin Gibbs #ifdef __lint 562bc9014e6SJustin Gibbs /* Very ugly, but it beats issuing suppression directives in many Makefiles. */ 563bc9014e6SJustin Gibbs extern void 564bc9014e6SJustin Gibbs dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func, 565bc9014e6SJustin Gibbs dmu_buf_t **clear_on_evict_dbufp); 566bc9014e6SJustin Gibbs #else /* __lint */ 567bc9014e6SJustin Gibbs inline void 568bc9014e6SJustin Gibbs dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func, 569bc9014e6SJustin Gibbs dmu_buf_t **clear_on_evict_dbufp) 570bc9014e6SJustin Gibbs { 571bc9014e6SJustin Gibbs ASSERT(dbu->dbu_evict_func == NULL); 572bc9014e6SJustin Gibbs ASSERT(evict_func != NULL); 573bc9014e6SJustin Gibbs dbu->dbu_evict_func = evict_func; 574bc9014e6SJustin Gibbs #ifdef ZFS_DEBUG 575bc9014e6SJustin Gibbs dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; 576bc9014e6SJustin Gibbs #endif 577bc9014e6SJustin Gibbs } 578bc9014e6SJustin Gibbs #endif /* __lint */ 579fa9e4066Sahrens 580fa9e4066Sahrens /* 581bc9014e6SJustin Gibbs * Attach user data to a dbuf and mark it for normal (when the dbuf's 582bc9014e6SJustin Gibbs * data is cleared or its reference count goes to zero) 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(dmu_buf_t *db, dmu_buf_user_t *user); 588bc9014e6SJustin Gibbs 589bc9014e6SJustin Gibbs /* 590bc9014e6SJustin Gibbs * Attach user data to a dbuf and mark it for immediate (its dirty and 591bc9014e6SJustin Gibbs * reference counts are equal) eviction processing. 592bc9014e6SJustin Gibbs * 593bc9014e6SJustin Gibbs * Returns NULL on success, or the existing user if another user currently 594bc9014e6SJustin Gibbs * owns the buffer. 595bc9014e6SJustin Gibbs */ 596bc9014e6SJustin Gibbs void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); 597bc9014e6SJustin Gibbs 598bc9014e6SJustin Gibbs /* 599bc9014e6SJustin Gibbs * Replace the current user of a dbuf. 600bc9014e6SJustin Gibbs * 601bc9014e6SJustin Gibbs * If given the current user of a dbuf, replaces the dbuf's user with 602bc9014e6SJustin Gibbs * "new_user" and returns the user data pointer that was replaced. 603bc9014e6SJustin Gibbs * Otherwise returns the current, and unmodified, dbuf user pointer. 604bc9014e6SJustin Gibbs */ 605bc9014e6SJustin Gibbs void *dmu_buf_replace_user(dmu_buf_t *db, 606bc9014e6SJustin Gibbs dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); 607bc9014e6SJustin Gibbs 608bc9014e6SJustin Gibbs /* 609bc9014e6SJustin Gibbs * Remove the specified user data for a DMU buffer. 610bc9014e6SJustin Gibbs * 611bc9014e6SJustin Gibbs * Returns the user that was removed on success, or the current user if 612bc9014e6SJustin Gibbs * another user currently owns the buffer. 613bc9014e6SJustin Gibbs */ 614bc9014e6SJustin Gibbs void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); 615bc9014e6SJustin Gibbs 616bc9014e6SJustin Gibbs /* 617bc9014e6SJustin Gibbs * Returns the user data (dmu_buf_user_t *) associated with this dbuf. 618fa9e4066Sahrens */ 619fa9e4066Sahrens void *dmu_buf_get_user(dmu_buf_t *db); 620fa9e4066Sahrens 621ae972795SMatthew Ahrens objset_t *dmu_buf_get_objset(dmu_buf_t *db); 622*79d72832SMatthew Ahrens dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db); 623*79d72832SMatthew Ahrens void dmu_buf_dnode_exit(dmu_buf_t *db); 624ae972795SMatthew Ahrens 625bc9014e6SJustin Gibbs /* Block until any in-progress dmu buf user evictions complete. */ 626bc9014e6SJustin Gibbs void dmu_buf_user_evict_wait(void); 627bc9014e6SJustin Gibbs 62880901aeaSGeorge Wilson /* 62980901aeaSGeorge Wilson * Returns the blkptr associated with this dbuf, or NULL if not set. 63080901aeaSGeorge Wilson */ 63180901aeaSGeorge Wilson struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); 63280901aeaSGeorge Wilson 633fa9e4066Sahrens /* 634fa9e4066Sahrens * Indicate that you are going to modify the buffer's data (db_data). 635fa9e4066Sahrens * 636fa9e4066Sahrens * The transaction (tx) must be assigned to a txg (ie. you've called 637fa9e4066Sahrens * dmu_tx_assign()). The buffer's object must be held in the tx 638fa9e4066Sahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 639fa9e4066Sahrens */ 640fa9e4066Sahrens void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 641fa9e4066Sahrens 6423d692628SSanjeev Bagewadi /* 6433d692628SSanjeev Bagewadi * Tells if the given dbuf is freeable. 6443d692628SSanjeev Bagewadi */ 6453d692628SSanjeev Bagewadi boolean_t dmu_buf_freeable(dmu_buf_t *); 6463d692628SSanjeev Bagewadi 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); 670fa9e4066Sahrens void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 671fa9e4066Sahrens uint64_t len); 67214843421SMatthew Ahrens void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 673fa9e4066Sahrens void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 6740a586ceaSMark Shellenbaum void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 6750a586ceaSMark Shellenbaum void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 6760a586ceaSMark Shellenbaum void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 677fa9e4066Sahrens void dmu_tx_abort(dmu_tx_t *tx); 6783b2aab18SMatthew Ahrens int dmu_tx_assign(dmu_tx_t *tx, enum txg_how txg_how); 6798a2f1b91Sahrens void dmu_tx_wait(dmu_tx_t *tx); 680fa9e4066Sahrens void dmu_tx_commit(dmu_tx_t *tx); 6814bb73804SMatthew Ahrens void dmu_tx_mark_netfree(dmu_tx_t *tx); 682fa9e4066Sahrens 683d20e665cSRicardo M. Correia /* 684d20e665cSRicardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called. 685d20e665cSRicardo M. Correia * 686d20e665cSRicardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a 687d20e665cSRicardo M. Correia * callback parameter. The caller is responsible for properly allocating and 688d20e665cSRicardo M. Correia * freeing it. 689d20e665cSRicardo M. Correia * 690d20e665cSRicardo M. Correia * When registering a callback, the transaction must be already created, but 691d20e665cSRicardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not. 692d20e665cSRicardo M. Correia * 693d20e665cSRicardo M. Correia * The callback will be called after the transaction has been safely written 694d20e665cSRicardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted. 695d20e665cSRicardo M. Correia * If there is any error which prevents the transaction from being committed to 696d20e665cSRicardo M. Correia * disk, the callback will be called with a value of error != 0. 697d20e665cSRicardo M. Correia */ 698d20e665cSRicardo M. Correia typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 699d20e665cSRicardo M. Correia 700d20e665cSRicardo M. Correia void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 701d20e665cSRicardo M. Correia void *dcb_data); 702d20e665cSRicardo M. Correia 703fa9e4066Sahrens /* 704fa9e4066Sahrens * Free up the data blocks for a defined range of a file. If size is 705ad135b5dSChristopher Siden * -1, the range from offset to end-of-file is freed. 706fa9e4066Sahrens */ 707ea8dc4b6Seschrock int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 708fa9e4066Sahrens uint64_t size, dmu_tx_t *tx); 709cdb0ab79Smaybee int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 710cdb0ab79Smaybee uint64_t size); 711713d6c20SMatthew Ahrens int dmu_free_long_object(objset_t *os, uint64_t object); 712fa9e4066Sahrens 713fa9e4066Sahrens /* 714fa9e4066Sahrens * Convenience functions. 715fa9e4066Sahrens * 716fa9e4066Sahrens * Canfail routines will return 0 on success, or an errno if there is a 717fa9e4066Sahrens * nonrecoverable I/O error. 718fa9e4066Sahrens */ 7197bfdf011SNeil Perrin #define DMU_READ_PREFETCH 0 /* prefetch */ 7207bfdf011SNeil Perrin #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 721ea8dc4b6Seschrock int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 7227bfdf011SNeil Perrin void *buf, uint32_t flags); 723fa9e4066Sahrens void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 724fa9e4066Sahrens const void *buf, dmu_tx_t *tx); 72582c9918fSTim Haley void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 72682c9918fSTim Haley dmu_tx_t *tx); 727feb08c6bSbillm int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); 728f8554bb9SMatthew Ahrens int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size); 729feb08c6bSbillm int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, 730feb08c6bSbillm dmu_tx_t *tx); 73194d1a210STim Haley int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, 73294d1a210STim Haley dmu_tx_t *tx); 73344eda4d7Smaybee int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, 73444eda4d7Smaybee uint64_t size, struct page *pp, dmu_tx_t *tx); 7352fdbea25SAleksandr Guzovskiy struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 7362fdbea25SAleksandr Guzovskiy void dmu_return_arcbuf(struct arc_buf *buf); 7372fdbea25SAleksandr Guzovskiy void dmu_assign_arcbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, 7382fdbea25SAleksandr Guzovskiy dmu_tx_t *tx); 739c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_init(struct xuio *uio, int niov); 740c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void dmu_xuio_fini(struct xuio *uio); 741c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, 742c242f9a0Schunli zhang - Sun Microsystems - Irvine United States size_t n); 743c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_cnt(struct xuio *uio); 744c242f9a0Schunli zhang - Sun Microsystems - Irvine United States struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); 745c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void dmu_xuio_clear(struct xuio *uio, int i); 746c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void xuio_stat_wbuf_copied(); 747c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void xuio_stat_wbuf_nocopy(); 748fa9e4066Sahrens 749cf6106c8SMatthew Ahrens extern boolean_t zfs_prefetch_disable; 750b5152584SMatthew Ahrens extern int zfs_max_recordsize; 751416e0cd8Sek 752fa9e4066Sahrens /* 753fa9e4066Sahrens * Asynchronously try to read in the data. 754fa9e4066Sahrens */ 755a2cdcdd2SPaul Dagnelie void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, 756a2cdcdd2SPaul Dagnelie uint64_t len, enum zio_priority pri); 757fa9e4066Sahrens 758fa9e4066Sahrens typedef struct dmu_object_info { 759b24ab676SJeff Bonwick /* All sizes are in bytes unless otherwise indicated. */ 760fa9e4066Sahrens uint32_t doi_data_block_size; 761fa9e4066Sahrens uint32_t doi_metadata_block_size; 762fa9e4066Sahrens dmu_object_type_t doi_type; 763fa9e4066Sahrens dmu_object_type_t doi_bonus_type; 764b24ab676SJeff Bonwick uint64_t doi_bonus_size; 765fa9e4066Sahrens uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 766fa9e4066Sahrens uint8_t doi_checksum; 767fa9e4066Sahrens uint8_t doi_compress; 768e77d42eaSMatthew Ahrens uint8_t doi_nblkptr; 769e77d42eaSMatthew Ahrens uint8_t doi_pad[4]; 770b24ab676SJeff Bonwick uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 771b24ab676SJeff Bonwick uint64_t doi_max_offset; 772b24ab676SJeff Bonwick uint64_t doi_fill_count; /* number of non-empty blocks */ 773fa9e4066Sahrens } dmu_object_info_t; 774fa9e4066Sahrens 775e45ce728Sahrens typedef void arc_byteswap_func_t(void *buf, size_t size); 776e45ce728Sahrens 777fa9e4066Sahrens typedef struct dmu_object_type_info { 778ad135b5dSChristopher Siden dmu_object_byteswap_t ot_byteswap; 779fa9e4066Sahrens boolean_t ot_metadata; 780fa9e4066Sahrens char *ot_name; 781fa9e4066Sahrens } dmu_object_type_info_t; 782fa9e4066Sahrens 783ad135b5dSChristopher Siden typedef struct dmu_object_byteswap_info { 784ad135b5dSChristopher Siden arc_byteswap_func_t *ob_func; 785ad135b5dSChristopher Siden char *ob_name; 786ad135b5dSChristopher Siden } dmu_object_byteswap_info_t; 787ad135b5dSChristopher Siden 788fa9e4066Sahrens extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 789ad135b5dSChristopher Siden extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; 790fa9e4066Sahrens 791fa9e4066Sahrens /* 792fa9e4066Sahrens * Get information on a DMU object. 793fa9e4066Sahrens * 794fa9e4066Sahrens * Return 0 on success or ENOENT if object is not allocated. 795fa9e4066Sahrens * 796fa9e4066Sahrens * If doi is NULL, just indicates whether the object exists. 797fa9e4066Sahrens */ 798fa9e4066Sahrens int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 7993e30c24aSWill Andrews /* Like dmu_object_info, but faster if you have a held dnode in hand. */ 800*79d72832SMatthew Ahrens void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi); 8013e30c24aSWill Andrews /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ 802fa9e4066Sahrens void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 8033e30c24aSWill Andrews /* 8043e30c24aSWill Andrews * Like dmu_object_info_from_db, but faster still when you only care about 8053e30c24aSWill Andrews * the size. This is specifically optimized for zfs_getattr(). 8063e30c24aSWill Andrews */ 807fa9e4066Sahrens void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 808fa9e4066Sahrens u_longlong_t *nblk512); 809fa9e4066Sahrens 810fa9e4066Sahrens typedef struct dmu_objset_stats { 811a2eea2e1Sahrens uint64_t dds_num_clones; /* number of clones of this */ 812a2eea2e1Sahrens uint64_t dds_creation_txg; 8133cb34c60Sahrens uint64_t dds_guid; 814fa9e4066Sahrens dmu_objset_type_t dds_type; 815fa9e4066Sahrens uint8_t dds_is_snapshot; 81631fd60d3Sahrens uint8_t dds_inconsistent; 8179adfa60dSMatthew Ahrens char dds_origin[ZFS_MAX_DATASET_NAME_LEN]; 818a2eea2e1Sahrens } dmu_objset_stats_t; 819fa9e4066Sahrens 820a2eea2e1Sahrens /* 821a2eea2e1Sahrens * Get stats on a dataset. 822a2eea2e1Sahrens */ 823a2eea2e1Sahrens void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 824fa9e4066Sahrens 825a2eea2e1Sahrens /* 826a2eea2e1Sahrens * Add entries to the nvlist for all the objset's properties. See 827a2eea2e1Sahrens * zfs_prop_table[] and zfs(1m) for details on the properties. 828a2eea2e1Sahrens */ 829a2eea2e1Sahrens void dmu_objset_stats(objset_t *os, struct nvlist *nv); 830fa9e4066Sahrens 831a2eea2e1Sahrens /* 832a2eea2e1Sahrens * Get the space usage statistics for statvfs(). 833a2eea2e1Sahrens * 834a2eea2e1Sahrens * refdbytes is the amount of space "referenced" by this objset. 835a2eea2e1Sahrens * availbytes is the amount of space available to this objset, taking 836a2eea2e1Sahrens * into account quotas & reservations, assuming that no other objsets 837a2eea2e1Sahrens * use the space first. These values correspond to the 'referenced' and 838a2eea2e1Sahrens * 'available' properties, described in the zfs(1m) manpage. 839a2eea2e1Sahrens * 840a2eea2e1Sahrens * usedobjs and availobjs are the number of objects currently allocated, 841a2eea2e1Sahrens * and available. 842a2eea2e1Sahrens */ 843a2eea2e1Sahrens void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 844a2eea2e1Sahrens uint64_t *usedobjsp, uint64_t *availobjsp); 845fa9e4066Sahrens 846fa9e4066Sahrens /* 847a2eea2e1Sahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions. 848a2eea2e1Sahrens * (Contrast with the ds_guid which is a 64-bit ID that will never 849a2eea2e1Sahrens * change, so there is a small probability that it will collide.) 850fa9e4066Sahrens */ 851a2eea2e1Sahrens uint64_t dmu_objset_fsid_guid(objset_t *os); 852fa9e4066Sahrens 85371eb0538SChris Kirby /* 85471eb0538SChris Kirby * Get the [cm]time for an objset's snapshot dir 85571eb0538SChris Kirby */ 85671eb0538SChris Kirby timestruc_t dmu_objset_snap_cmtime(objset_t *os); 85771eb0538SChris Kirby 858fa9e4066Sahrens int dmu_objset_is_snapshot(objset_t *os); 859fa9e4066Sahrens 860fa9e4066Sahrens extern struct spa *dmu_objset_spa(objset_t *os); 861fa9e4066Sahrens extern struct zilog *dmu_objset_zil(objset_t *os); 862fa9e4066Sahrens extern struct dsl_pool *dmu_objset_pool(objset_t *os); 863fa9e4066Sahrens extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 864fa9e4066Sahrens extern void dmu_objset_name(objset_t *os, char *buf); 865fa9e4066Sahrens extern dmu_objset_type_t dmu_objset_type(objset_t *os); 866fa9e4066Sahrens extern uint64_t dmu_objset_id(objset_t *os); 867edf345e6SMatthew Ahrens extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); 868edf345e6SMatthew Ahrens extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); 869fa9e4066Sahrens extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, 870b38f0970Sck uint64_t *id, uint64_t *offp, boolean_t *case_conflict); 871ab04eb8eStimh extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, 872ab04eb8eStimh int maxlen, boolean_t *conflict); 87387e5029aSahrens extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, 87487e5029aSahrens uint64_t *idp, uint64_t *offp); 87514843421SMatthew Ahrens 8769966ca11SMatthew Ahrens typedef int objset_used_cb_t(dmu_object_type_t bonustype, 8779966ca11SMatthew Ahrens void *bonus, uint64_t *userp, uint64_t *groupp); 87814843421SMatthew Ahrens extern void dmu_objset_register_type(dmu_objset_type_t ost, 87914843421SMatthew Ahrens objset_used_cb_t *cb); 880f18faf3fSek extern void dmu_objset_set_user(objset_t *os, void *user_ptr); 881f18faf3fSek extern void *dmu_objset_get_user(objset_t *os); 882fa9e4066Sahrens 883fa9e4066Sahrens /* 884fa9e4066Sahrens * Return the txg number for the given assigned transaction. 885fa9e4066Sahrens */ 88687e5029aSahrens uint64_t dmu_tx_get_txg(dmu_tx_t *tx); 887fa9e4066Sahrens 888fa9e4066Sahrens /* 889fa9e4066Sahrens * Synchronous write. 890c5c6ffa0Smaybee * If a parent zio is provided this function initiates a write on the 891c5c6ffa0Smaybee * provided buffer as a child of the parent zio. 892da6c28aaSamw * In the absence of a parent zio, the write is completed synchronously. 893c5c6ffa0Smaybee * At write completion, blk is filled with the bp of the written block. 894fa9e4066Sahrens * Note that while the data covered by this function will be on stable 895c5c6ffa0Smaybee * storage when the write completes this new data does not become a 896fa9e4066Sahrens * permanent part of the file until the associated transaction commits. 897fa9e4066Sahrens */ 898b24ab676SJeff Bonwick 899b24ab676SJeff Bonwick /* 900b24ab676SJeff Bonwick * {zfs,zvol,ztest}_get_done() args 901b24ab676SJeff Bonwick */ 902b24ab676SJeff Bonwick typedef struct zgd { 903b24ab676SJeff Bonwick struct zilog *zgd_zilog; 904b24ab676SJeff Bonwick struct blkptr *zgd_bp; 905b24ab676SJeff Bonwick dmu_buf_t *zgd_db; 906b24ab676SJeff Bonwick struct rl *zgd_rl; 907b24ab676SJeff Bonwick void *zgd_private; 908b24ab676SJeff Bonwick } zgd_t; 909b24ab676SJeff Bonwick 910b24ab676SJeff Bonwick typedef void dmu_sync_cb_t(zgd_t *arg, int error); 911b24ab676SJeff Bonwick int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); 912fa9e4066Sahrens 913fa9e4066Sahrens /* 914fa9e4066Sahrens * Find the next hole or data block in file starting at *off 915fa9e4066Sahrens * Return found offset in *off. Return ESRCH for end of file. 916fa9e4066Sahrens */ 917fa9e4066Sahrens int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, 918fa9e4066Sahrens uint64_t *off); 919fa9e4066Sahrens 9202bcf0248SMax Grossman /* 9212bcf0248SMax Grossman * Check if a DMU object has any dirty blocks. If so, sync out 9222bcf0248SMax Grossman * all pending transaction groups. Otherwise, this function 9232bcf0248SMax Grossman * does not alter DMU state. This could be improved to only sync 9242bcf0248SMax Grossman * out the necessary transaction groups for this particular 9252bcf0248SMax Grossman * object. 9262bcf0248SMax Grossman */ 9272bcf0248SMax Grossman int dmu_object_wait_synced(objset_t *os, uint64_t object); 9282bcf0248SMax Grossman 929fa9e4066Sahrens /* 930fa9e4066Sahrens * Initial setup and final teardown. 931fa9e4066Sahrens */ 932fa9e4066Sahrens extern void dmu_init(void); 933fa9e4066Sahrens extern void dmu_fini(void); 934fa9e4066Sahrens 935fa9e4066Sahrens typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, 936fa9e4066Sahrens uint64_t object, uint64_t offset, int len); 937fa9e4066Sahrens void dmu_traverse_objset(objset_t *os, uint64_t txg_start, 938fa9e4066Sahrens dmu_traverse_cb_t cb, void *arg); 939fa9e4066Sahrens 9403b2aab18SMatthew Ahrens int dmu_diff(const char *tosnap_name, const char *fromsnap_name, 9413b2aab18SMatthew Ahrens struct vnode *vp, offset_t *offp); 94299d5e173STim Haley 943fa9e4066Sahrens /* CRC64 table */ 944fa9e4066Sahrens #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 945fa9e4066Sahrens extern uint64_t zfs_crc64_table[256]; 946fa9e4066Sahrens 947e4be62a2SSaso Kiselkov extern int zfs_mdcomp_disable; 948e4be62a2SSaso Kiselkov 949fa9e4066Sahrens #ifdef __cplusplus 950fa9e4066Sahrens } 951fa9e4066Sahrens #endif 952fa9e4066Sahrens 953fa9e4066Sahrens #endif /* _SYS_DMU_H */ 954