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. 24663207adSDon Brady * Copyright (c) 2011, 2018 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. 31663207adSDon Brady * Copyright (c) 2017, Intel Corporation. 32c3d26abcSMatthew Ahrens * Copyright (c) 2014 Integros [integros.com] 33ec5cf9d5SAlexander Stetsenko */ 34fa9e4066Sahrens 3555da60b9SMark J Musante /* Portions Copyright 2010 Robert Milkowski */ 3655da60b9SMark J Musante 37fa9e4066Sahrens #ifndef _SYS_DMU_H 38fa9e4066Sahrens #define _SYS_DMU_H 39fa9e4066Sahrens 40fa9e4066Sahrens /* 41fa9e4066Sahrens * This file describes the interface that the DMU provides for its 42fa9e4066Sahrens * consumers. 43fa9e4066Sahrens * 44fa9e4066Sahrens * The DMU also interacts with the SPA. That interface is described in 45fa9e4066Sahrens * dmu_spa.h. 46fa9e4066Sahrens */ 47fa9e4066Sahrens 48bc9014e6SJustin Gibbs #include <sys/zfs_context.h> 49fa9e4066Sahrens #include <sys/inttypes.h> 50ecd6cf80Smarks #include <sys/cred.h> 514445fffbSMatthew Ahrens #include <sys/fs/zfs.h> 525602294fSDan Kimmel #include <sys/zio_compress.h> 53a2cdcdd2SPaul Dagnelie #include <sys/zio_priority.h> 54fa9e4066Sahrens 55fa9e4066Sahrens #ifdef __cplusplus 56fa9e4066Sahrens extern "C" { 57fa9e4066Sahrens #endif 58fa9e4066Sahrens 59fa9e4066Sahrens struct uio; 60c242f9a0Schunli zhang - Sun Microsystems - Irvine United States struct xuio; 6144eda4d7Smaybee struct page; 62fa9e4066Sahrens struct vnode; 63fa9e4066Sahrens struct spa; 64fa9e4066Sahrens struct zilog; 65fa9e4066Sahrens struct zio; 66fa9e4066Sahrens struct blkptr; 67fa9e4066Sahrens struct zap_cursor; 68fa9e4066Sahrens struct dsl_dataset; 69fa9e4066Sahrens struct dsl_pool; 70fa9e4066Sahrens struct dnode; 71fa9e4066Sahrens struct drr_begin; 72fa9e4066Sahrens struct drr_end; 737802d7bfSMatthew Ahrens struct zbookmark_phys; 7444cd46caSbillm struct spa; 75a2eea2e1Sahrens struct nvlist; 762fdbea25SAleksandr Guzovskiy struct arc_buf; 77b24ab676SJeff Bonwick struct zio_prop; 780a586ceaSMark Shellenbaum struct sa_handle; 7979315247SMatthew Ahrens struct locked_range; 80*eb633035STom Caputi struct dsl_crypto_params; 81fa9e4066Sahrens 82fa9e4066Sahrens typedef struct objset objset_t; 83fa9e4066Sahrens typedef struct dmu_tx dmu_tx_t; 84fa9e4066Sahrens typedef struct dsl_dir dsl_dir_t; 8579d72832SMatthew Ahrens typedef struct dnode dnode_t; 86fa9e4066Sahrens 87ad135b5dSChristopher Siden typedef enum dmu_object_byteswap { 88ad135b5dSChristopher Siden DMU_BSWAP_UINT8, 89ad135b5dSChristopher Siden DMU_BSWAP_UINT16, 90ad135b5dSChristopher Siden DMU_BSWAP_UINT32, 91ad135b5dSChristopher Siden DMU_BSWAP_UINT64, 92ad135b5dSChristopher Siden DMU_BSWAP_ZAP, 93ad135b5dSChristopher Siden DMU_BSWAP_DNODE, 94ad135b5dSChristopher Siden DMU_BSWAP_OBJSET, 95ad135b5dSChristopher Siden DMU_BSWAP_ZNODE, 96ad135b5dSChristopher Siden DMU_BSWAP_OLDACL, 97ad135b5dSChristopher Siden DMU_BSWAP_ACL, 98ad135b5dSChristopher Siden /* 99ad135b5dSChristopher Siden * Allocating a new byteswap type number makes the on-disk format 100ad135b5dSChristopher Siden * incompatible with any other format that uses the same number. 101ad135b5dSChristopher Siden * 102ad135b5dSChristopher Siden * Data can usually be structured to work with one of the 103ad135b5dSChristopher Siden * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. 104ad135b5dSChristopher Siden */ 105ad135b5dSChristopher Siden DMU_BSWAP_NUMFUNCS 106ad135b5dSChristopher Siden } dmu_object_byteswap_t; 107ad135b5dSChristopher Siden 108ad135b5dSChristopher Siden #define DMU_OT_NEWTYPE 0x80 109ad135b5dSChristopher Siden #define DMU_OT_METADATA 0x40 110*eb633035STom Caputi #define DMU_OT_ENCRYPTED 0x20 111*eb633035STom Caputi #define DMU_OT_BYTESWAP_MASK 0x1f 112ad135b5dSChristopher Siden 113ad135b5dSChristopher Siden /* 114ad135b5dSChristopher Siden * Defines a uint8_t object type. Object types specify if the data 115ad135b5dSChristopher Siden * in the object is metadata (boolean) and how to byteswap the data 116adb52d92SMatthew Ahrens * (dmu_object_byteswap_t). All of the types created by this method 117adb52d92SMatthew Ahrens * are cached in the dbuf metadata cache. 118ad135b5dSChristopher Siden */ 119*eb633035STom Caputi #define DMU_OT(byteswap, metadata, encrypted) \ 120ad135b5dSChristopher Siden (DMU_OT_NEWTYPE | \ 121ad135b5dSChristopher Siden ((metadata) ? DMU_OT_METADATA : 0) | \ 122*eb633035STom Caputi ((encrypted) ? DMU_OT_ENCRYPTED : 0) | \ 123ad135b5dSChristopher Siden ((byteswap) & DMU_OT_BYTESWAP_MASK)) 124ad135b5dSChristopher Siden 125ad135b5dSChristopher Siden #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 126ad135b5dSChristopher Siden ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ 127ad135b5dSChristopher Siden (ot) < DMU_OT_NUMTYPES) 128ad135b5dSChristopher Siden 129*eb633035STom Caputi /* 130*eb633035STom Caputi * MDB doesn't have dmu_ot; it defines these macros itself. 131*eb633035STom Caputi */ 132*eb633035STom Caputi #ifndef ZFS_MDB 133*eb633035STom Caputi #define DMU_OT_IS_METADATA_IMPL(ot) (dmu_ot[ot].ot_metadata) 134*eb633035STom Caputi #define DMU_OT_IS_ENCRYPTED_IMPL(ot) (dmu_ot[ot].ot_encrypt) 135*eb633035STom Caputi #define DMU_OT_BYTESWAP_IMPL(ot) (dmu_ot[ot].ot_byteswap) 136*eb633035STom Caputi #endif 137*eb633035STom Caputi 138ad135b5dSChristopher Siden #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 139ad135b5dSChristopher Siden ((ot) & DMU_OT_METADATA) : \ 140*eb633035STom Caputi DMU_OT_IS_METADATA_IMPL(ot)) 141ad135b5dSChristopher Siden 142663207adSDon Brady #define DMU_OT_IS_DDT(ot) \ 143663207adSDon Brady ((ot) == DMU_OT_DDT_ZAP) 144663207adSDon Brady 145663207adSDon Brady #define DMU_OT_IS_ZIL(ot) \ 146663207adSDon Brady ((ot) == DMU_OT_INTENT_LOG) 147663207adSDon Brady 148663207adSDon Brady /* Note: ztest uses DMU_OT_UINT64_OTHER as a proxy for file blocks */ 149663207adSDon Brady #define DMU_OT_IS_FILE(ot) \ 150663207adSDon Brady ((ot) == DMU_OT_PLAIN_FILE_CONTENTS || (ot) == DMU_OT_UINT64_OTHER) 151663207adSDon Brady 152adb52d92SMatthew Ahrens #define DMU_OT_IS_METADATA_CACHED(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 153adb52d92SMatthew Ahrens B_TRUE : dmu_ot[(ot)].ot_dbuf_metadata_cache) 154adb52d92SMatthew Ahrens 155*eb633035STom Caputi #define DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 156*eb633035STom Caputi ((ot) & DMU_OT_ENCRYPTED) : \ 157*eb633035STom Caputi DMU_OT_IS_ENCRYPTED_IMPL(ot)) 158*eb633035STom Caputi 1595d7b4d43SMatthew Ahrens /* 1605d7b4d43SMatthew Ahrens * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't 1615d7b4d43SMatthew Ahrens * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill 1625d7b4d43SMatthew Ahrens * is repurposed for embedded BPs. 1635d7b4d43SMatthew Ahrens */ 1645d7b4d43SMatthew Ahrens #define DMU_OT_HAS_FILL(ot) \ 1655d7b4d43SMatthew Ahrens ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) 1665d7b4d43SMatthew Ahrens 167ad135b5dSChristopher Siden #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ 168ad135b5dSChristopher Siden ((ot) & DMU_OT_BYTESWAP_MASK) : \ 169*eb633035STom Caputi DMU_OT_BYTESWAP_IMPL(ot)) 170ad135b5dSChristopher Siden 171fa9e4066Sahrens typedef enum dmu_object_type { 172fa9e4066Sahrens DMU_OT_NONE, 173fa9e4066Sahrens /* general: */ 174fa9e4066Sahrens DMU_OT_OBJECT_DIRECTORY, /* ZAP */ 175fa9e4066Sahrens DMU_OT_OBJECT_ARRAY, /* UINT64 */ 176fa9e4066Sahrens DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ 177fa9e4066Sahrens DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ 178cde58dbcSMatthew Ahrens DMU_OT_BPOBJ, /* UINT64 */ 179cde58dbcSMatthew Ahrens DMU_OT_BPOBJ_HDR, /* UINT64 */ 180fa9e4066Sahrens /* spa: */ 181fa9e4066Sahrens DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ 182fa9e4066Sahrens DMU_OT_SPACE_MAP, /* UINT64 */ 183fa9e4066Sahrens /* zil: */ 184fa9e4066Sahrens DMU_OT_INTENT_LOG, /* UINT64 */ 185fa9e4066Sahrens /* dmu: */ 186fa9e4066Sahrens DMU_OT_DNODE, /* DNODE */ 187fa9e4066Sahrens DMU_OT_OBJSET, /* OBJSET */ 188fa9e4066Sahrens /* dsl: */ 1891649cd4bStabriz DMU_OT_DSL_DIR, /* UINT64 */ 19087e5029aSahrens DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ 19187e5029aSahrens DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ 192fa9e4066Sahrens DMU_OT_DSL_PROPS, /* ZAP */ 1931649cd4bStabriz DMU_OT_DSL_DATASET, /* UINT64 */ 194fa9e4066Sahrens /* zpl: */ 195fa9e4066Sahrens DMU_OT_ZNODE, /* ZNODE */ 196da6c28aaSamw DMU_OT_OLDACL, /* Old ACL */ 197fa9e4066Sahrens DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ 198fa9e4066Sahrens DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ 199fa9e4066Sahrens DMU_OT_MASTER_NODE, /* ZAP */ 200893a6d32Sahrens DMU_OT_UNLINKED_SET, /* ZAP */ 201fa9e4066Sahrens /* zvol: */ 202fa9e4066Sahrens DMU_OT_ZVOL, /* UINT8 */ 203fa9e4066Sahrens DMU_OT_ZVOL_PROP, /* ZAP */ 204fa9e4066Sahrens /* other; for testing only! */ 205fa9e4066Sahrens DMU_OT_PLAIN_OTHER, /* UINT8 */ 206fa9e4066Sahrens DMU_OT_UINT64_OTHER, /* UINT64 */ 207fa9e4066Sahrens DMU_OT_ZAP_OTHER, /* ZAP */ 208ea8dc4b6Seschrock /* new object types: */ 209ea8dc4b6Seschrock DMU_OT_ERROR_LOG, /* ZAP */ 21006eeb2adSek DMU_OT_SPA_HISTORY, /* UINT8 */ 21106eeb2adSek DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ 212b1b8ab34Slling DMU_OT_POOL_PROPS, /* ZAP */ 213ecd6cf80Smarks DMU_OT_DSL_PERMS, /* ZAP */ 214da6c28aaSamw DMU_OT_ACL, /* ACL */ 215da6c28aaSamw DMU_OT_SYSACL, /* SYSACL */ 216da6c28aaSamw DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ 217da6c28aaSamw DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ 218088f3894Sahrens DMU_OT_NEXT_CLONES, /* ZAP */ 2193f9d6ad7SLin Ling DMU_OT_SCAN_QUEUE, /* ZAP */ 22014843421SMatthew Ahrens DMU_OT_USERGROUP_USED, /* ZAP */ 22114843421SMatthew Ahrens DMU_OT_USERGROUP_QUOTA, /* ZAP */ 222842727c2SChris Kirby DMU_OT_USERREFS, /* ZAP */ 223b24ab676SJeff Bonwick DMU_OT_DDT_ZAP, /* ZAP */ 224b24ab676SJeff Bonwick DMU_OT_DDT_STATS, /* ZAP */ 2250a586ceaSMark Shellenbaum DMU_OT_SA, /* System attr */ 2260a586ceaSMark Shellenbaum DMU_OT_SA_MASTER_NODE, /* ZAP */ 2270a586ceaSMark Shellenbaum DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ 2280a586ceaSMark Shellenbaum DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ 2293f9d6ad7SLin Ling DMU_OT_SCAN_XLATE, /* ZAP */ 2303f9d6ad7SLin Ling DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ 231cde58dbcSMatthew Ahrens DMU_OT_DEADLIST, /* ZAP */ 232cde58dbcSMatthew Ahrens DMU_OT_DEADLIST_HDR, /* UINT64 */ 233cde58dbcSMatthew Ahrens DMU_OT_DSL_CLONES, /* ZAP */ 234cde58dbcSMatthew Ahrens DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ 235ad135b5dSChristopher Siden /* 236ad135b5dSChristopher Siden * Do not allocate new object types here. Doing so makes the on-disk 237ad135b5dSChristopher Siden * format incompatible with any other format that uses the same object 238ad135b5dSChristopher Siden * type number. 239ad135b5dSChristopher Siden * 240ad135b5dSChristopher Siden * When creating an object which does not have one of the above types 241ad135b5dSChristopher Siden * use the DMU_OTN_* type with the correct byteswap and metadata 242ad135b5dSChristopher Siden * values. 243ad135b5dSChristopher Siden * 244ad135b5dSChristopher Siden * The DMU_OTN_* types do not have entries in the dmu_ot table, 245ad135b5dSChristopher Siden * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead 246663207adSDon Brady * use the DMU_OT_IS_METADATA() and DMU_OT_BYTESWAP() macros instead 247ad135b5dSChristopher Siden * of indexing into dmu_ot directly (this works for both DMU_OT_* types 248ad135b5dSChristopher Siden * and DMU_OTN_* types). 249ad135b5dSChristopher Siden */ 250ad135b5dSChristopher Siden DMU_OT_NUMTYPES, 251ad135b5dSChristopher Siden 252ad135b5dSChristopher Siden /* 253ad135b5dSChristopher Siden * Names for valid types declared with DMU_OT(). 254ad135b5dSChristopher Siden */ 255*eb633035STom Caputi DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_FALSE), 256*eb633035STom Caputi DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_FALSE), 257*eb633035STom Caputi DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_FALSE), 258*eb633035STom Caputi DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_FALSE), 259*eb633035STom Caputi DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_FALSE), 260*eb633035STom Caputi DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_FALSE), 261*eb633035STom Caputi DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_FALSE), 262*eb633035STom Caputi DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_FALSE), 263*eb633035STom Caputi DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_FALSE), 264*eb633035STom Caputi DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_FALSE), 265*eb633035STom Caputi 266*eb633035STom Caputi DMU_OTN_UINT8_ENC_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_TRUE), 267*eb633035STom Caputi DMU_OTN_UINT8_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_TRUE), 268*eb633035STom Caputi DMU_OTN_UINT16_ENC_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_TRUE), 269*eb633035STom Caputi DMU_OTN_UINT16_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_TRUE), 270*eb633035STom Caputi DMU_OTN_UINT32_ENC_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_TRUE), 271*eb633035STom Caputi DMU_OTN_UINT32_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_TRUE), 272*eb633035STom Caputi DMU_OTN_UINT64_ENC_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_TRUE), 273*eb633035STom Caputi DMU_OTN_UINT64_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_TRUE), 274*eb633035STom Caputi DMU_OTN_ZAP_ENC_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_TRUE), 275*eb633035STom Caputi DMU_OTN_ZAP_ENC_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_TRUE), 276fa9e4066Sahrens } dmu_object_type_t; 277fa9e4066Sahrens 278f864f99eSPrakash Surya /* 279f864f99eSPrakash Surya * These flags are intended to be used to specify the "txg_how" 280f864f99eSPrakash Surya * parameter when calling the dmu_tx_assign() function. See the comment 281f864f99eSPrakash Surya * above dmu_tx_assign() for more details on the meaning of these flags. 282f864f99eSPrakash Surya */ 283f864f99eSPrakash Surya #define TXG_NOWAIT (0ULL) 284f864f99eSPrakash Surya #define TXG_WAIT (1ULL<<0) 285f864f99eSPrakash Surya #define TXG_NOTHROTTLE (1ULL<<1) 2863b2aab18SMatthew Ahrens 287fa9e4066Sahrens void byteswap_uint64_array(void *buf, size_t size); 288fa9e4066Sahrens void byteswap_uint32_array(void *buf, size_t size); 289fa9e4066Sahrens void byteswap_uint16_array(void *buf, size_t size); 290fa9e4066Sahrens void byteswap_uint8_array(void *buf, size_t size); 291fa9e4066Sahrens void zap_byteswap(void *buf, size_t size); 292da6c28aaSamw void zfs_oldacl_byteswap(void *buf, size_t size); 293fa9e4066Sahrens void zfs_acl_byteswap(void *buf, size_t size); 294fa9e4066Sahrens void zfs_znode_byteswap(void *buf, size_t size); 295fa9e4066Sahrens 2960b69c2f0Sahrens #define DS_FIND_SNAPSHOTS (1<<0) 2970b69c2f0Sahrens #define DS_FIND_CHILDREN (1<<1) 29812380e1eSArne Jansen #define DS_FIND_SERIALIZE (1<<2) 299fa9e4066Sahrens 300fa9e4066Sahrens /* 301fa9e4066Sahrens * The maximum number of bytes that can be accessed as part of one 302fa9e4066Sahrens * operation, including metadata. 303fa9e4066Sahrens */ 304b5152584SMatthew Ahrens #define DMU_MAX_ACCESS (32 * 1024 * 1024) /* 32MB */ 305cdb0ab79Smaybee #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ 306fa9e4066Sahrens 30714843421SMatthew Ahrens #define DMU_USERUSED_OBJECT (-1ULL) 30814843421SMatthew Ahrens #define DMU_GROUPUSED_OBJECT (-2ULL) 30914843421SMatthew Ahrens 3100a586ceaSMark Shellenbaum /* 3110a586ceaSMark Shellenbaum * artificial blkids for bonus buffer and spill blocks 3120a586ceaSMark Shellenbaum */ 3130a586ceaSMark Shellenbaum #define DMU_BONUS_BLKID (-1ULL) 3140a586ceaSMark Shellenbaum #define DMU_SPILL_BLKID (-2ULL) 315*eb633035STom Caputi 316fa9e4066Sahrens /* 317fa9e4066Sahrens * Public routines to create, destroy, open, and close objsets. 318fa9e4066Sahrens */ 319*eb633035STom Caputi typedef void dmu_objset_create_sync_func_t(objset_t *os, void *arg, 320*eb633035STom Caputi cred_t *cr, dmu_tx_t *tx); 321*eb633035STom Caputi 322503ad85cSMatthew Ahrens int dmu_objset_hold(const char *name, void *tag, objset_t **osp); 323503ad85cSMatthew Ahrens int dmu_objset_own(const char *name, dmu_objset_type_t type, 324*eb633035STom Caputi boolean_t readonly, boolean_t key_required, void *tag, objset_t **osp); 325503ad85cSMatthew Ahrens void dmu_objset_rele(objset_t *os, void *tag); 326*eb633035STom Caputi void dmu_objset_disown(objset_t *os, boolean_t key_required, void *tag); 327503ad85cSMatthew Ahrens int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); 328503ad85cSMatthew Ahrens 3293b2aab18SMatthew Ahrens void dmu_objset_evict_dbufs(objset_t *os); 330ae46e4c7SMatthew Ahrens int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, 331*eb633035STom Caputi struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func, 332*eb633035STom Caputi void *arg); 3333b2aab18SMatthew Ahrens int dmu_objset_clone(const char *name, const char *origin); 3343b2aab18SMatthew Ahrens int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, 3354445fffbSMatthew Ahrens struct nvlist *errlist); 3364445fffbSMatthew Ahrens int dmu_objset_snapshot_one(const char *fsname, const char *snapname); 3374445fffbSMatthew Ahrens int dmu_objset_snapshot_tmp(const char *, const char *, int); 338fd136879SMatthew Ahrens int dmu_objset_find(char *name, int func(const char *, void *), void *arg, 339fa9e4066Sahrens int flags); 340fa9e4066Sahrens void dmu_objset_byteswap(void *buf, size_t size); 3413b2aab18SMatthew Ahrens int dsl_dataset_rename_snapshot(const char *fsname, 3423b2aab18SMatthew Ahrens const char *oldsnapname, const char *newsnapname, boolean_t recursive); 3435cabbc6bSPrashanth Sreenivasa int dmu_objset_remap_indirects(const char *fsname); 344fa9e4066Sahrens 345fa9e4066Sahrens typedef struct dmu_buf { 346fa9e4066Sahrens uint64_t db_object; /* object that this buffer is part of */ 347fa9e4066Sahrens uint64_t db_offset; /* byte offset in this object */ 348fa9e4066Sahrens uint64_t db_size; /* size of buffer in bytes */ 349fa9e4066Sahrens void *db_data; /* data in buffer */ 350fa9e4066Sahrens } dmu_buf_t; 351fa9e4066Sahrens 35299653d4eSeschrock /* 35399653d4eSeschrock * The names of zap entries in the DIRECTORY_OBJECT of the MOS. 35499653d4eSeschrock */ 355fa9e4066Sahrens #define DMU_POOL_DIRECTORY_OBJECT 1 356fa9e4066Sahrens #define DMU_POOL_CONFIG "config" 357ad135b5dSChristopher Siden #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" 358ad135b5dSChristopher Siden #define DMU_POOL_FEATURES_FOR_READ "features_for_read" 359ad135b5dSChristopher Siden #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" 36043466aaeSMax Grossman #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" 361fa9e4066Sahrens #define DMU_POOL_ROOT_DATASET "root_dataset" 362cde58dbcSMatthew Ahrens #define DMU_POOL_SYNC_BPOBJ "sync_bplist" 363ea8dc4b6Seschrock #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" 364ea8dc4b6Seschrock #define DMU_POOL_ERRLOG_LAST "errlog_last" 36599653d4eSeschrock #define DMU_POOL_SPARES "spares" 36699653d4eSeschrock #define DMU_POOL_DEFLATE "deflate" 36706eeb2adSek #define DMU_POOL_HISTORY "history" 368b1b8ab34Slling #define DMU_POOL_PROPS "pool_props" 369fa94a07fSbrendan #define DMU_POOL_L2CACHE "l2cache" 370ca45db41SChris Kirby #define DMU_POOL_TMP_USERREFS "tmp_userrefs" 371b24ab676SJeff Bonwick #define DMU_POOL_DDT "DDT-%s-%s-%s" 372b24ab676SJeff Bonwick #define DMU_POOL_DDT_STATS "DDT-statistics" 3733f9d6ad7SLin Ling #define DMU_POOL_CREATION_VERSION "creation_version" 3743f9d6ad7SLin Ling #define DMU_POOL_SCAN "scan" 375cde58dbcSMatthew Ahrens #define DMU_POOL_FREE_BPOBJ "free_bpobj" 376ad135b5dSChristopher Siden #define DMU_POOL_BPTREE_OBJ "bptree_obj" 377f1745736SMatthew Ahrens #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" 37845818ee1SMatthew Ahrens #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" 379215198a6SJoe Stein #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map" 3805cabbc6bSPrashanth Sreenivasa #define DMU_POOL_REMOVING "com.delphix:removing" 3815cabbc6bSPrashanth Sreenivasa #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" 3825cabbc6bSPrashanth Sreenivasa #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" 38386714001SSerapheim Dimitropoulos #define DMU_POOL_ZPOOL_CHECKPOINT "com.delphix:zpool_checkpoint" 384088f3894Sahrens 385fa9e4066Sahrens /* 386fa9e4066Sahrens * Allocate an object from this objset. The range of object numbers 387fa9e4066Sahrens * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. 388fa9e4066Sahrens * 389fa9e4066Sahrens * The transaction must be assigned to a txg. The newly allocated 390fa9e4066Sahrens * object will be "held" in the transaction (ie. you can modify the 391fa9e4066Sahrens * newly allocated object in this transaction). 392fa9e4066Sahrens * 393fa9e4066Sahrens * dmu_object_alloc() chooses an object and returns it in *objectp. 394fa9e4066Sahrens * 395fa9e4066Sahrens * dmu_object_claim() allocates a specific object number. If that 396fa9e4066Sahrens * number is already allocated, it fails and returns EEXIST. 397fa9e4066Sahrens * 398fa9e4066Sahrens * Return 0 on success, or ENOSPC or EEXIST as specified above. 399fa9e4066Sahrens */ 400fa9e4066Sahrens uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, 401fa9e4066Sahrens int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 402221813c1SMatthew Ahrens uint64_t dmu_object_alloc_ibs(objset_t *os, dmu_object_type_t ot, int blocksize, 403221813c1SMatthew Ahrens int indirect_blockshift, 404221813c1SMatthew Ahrens dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *tx); 40554811da5SToomas Soome uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, 40654811da5SToomas Soome int blocksize, dmu_object_type_t bonus_type, int bonus_len, 40754811da5SToomas Soome int dnodesize, dmu_tx_t *tx); 40854811da5SToomas Soome int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, 40954811da5SToomas Soome int blocksize, dmu_object_type_t bonus_type, int bonus_len, 41054811da5SToomas Soome int dnodesize, dmu_tx_t *tx); 41154811da5SToomas Soome int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, 41254811da5SToomas Soome dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, 413*eb633035STom Caputi int bonuslen, int dnodesize, boolean_t keep_spill, dmu_tx_t *txp); 414fa9e4066Sahrens int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, 415fa9e4066Sahrens int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); 416fa9e4066Sahrens int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, 417e77d42eaSMatthew Ahrens int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); 418*eb633035STom Caputi int dmu_object_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx); 419fa9e4066Sahrens 420fa9e4066Sahrens /* 421fa9e4066Sahrens * Free an object from this objset. 422fa9e4066Sahrens * 423fa9e4066Sahrens * The object's data will be freed as well (ie. you don't need to call 424fa9e4066Sahrens * dmu_free(object, 0, -1, tx)). 425fa9e4066Sahrens * 426fa9e4066Sahrens * The object need not be held in the transaction. 427fa9e4066Sahrens * 428fa9e4066Sahrens * If there are any holds on this object's buffers (via dmu_buf_hold()), 429fa9e4066Sahrens * or tx holds on the object (via dmu_tx_hold_object()), you can not 430fa9e4066Sahrens * free it; it fails and returns EBUSY. 431fa9e4066Sahrens * 432fa9e4066Sahrens * If the object is not allocated, it fails and returns ENOENT. 433fa9e4066Sahrens * 434fa9e4066Sahrens * Return 0 on success, or EBUSY or ENOENT as specified above. 435fa9e4066Sahrens */ 436fa9e4066Sahrens int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); 437fa9e4066Sahrens 438fa9e4066Sahrens /* 439fa9e4066Sahrens * Find the next allocated or free object. 440fa9e4066Sahrens * 441fa9e4066Sahrens * The objectp parameter is in-out. It will be updated to be the next 4426754306eSahrens * object which is allocated. Ignore objects which have not been 4436754306eSahrens * modified since txg. 444fa9e4066Sahrens * 445fa9e4066Sahrens * XXX Can only be called on a objset with no dirty data. 446fa9e4066Sahrens * 447fa9e4066Sahrens * Returns 0 on success, or ENOENT if there are no more objects. 448fa9e4066Sahrens */ 4496754306eSahrens int dmu_object_next(objset_t *os, uint64_t *objectp, 4506754306eSahrens boolean_t hole, uint64_t txg); 451fa9e4066Sahrens 452*eb633035STom Caputi /* 453*eb633035STom Caputi * Set the number of levels on a dnode. nlevels must be greater than the 454*eb633035STom Caputi * current number of levels or an EINVAL will be returned. 455*eb633035STom Caputi */ 456*eb633035STom Caputi int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels, 457*eb633035STom Caputi dmu_tx_t *tx); 458*eb633035STom Caputi 459fa9e4066Sahrens /* 460fa9e4066Sahrens * Set the data blocksize for an object. 461fa9e4066Sahrens * 462fa9e4066Sahrens * The object cannot have any blocks allcated beyond the first. If 463fa9e4066Sahrens * the first block is allocated already, the new size must be greater 464fa9e4066Sahrens * than the current block size. If these conditions are not met, 465fa9e4066Sahrens * ENOTSUP will be returned. 466fa9e4066Sahrens * 467fa9e4066Sahrens * Returns 0 on success, or EBUSY if there are any holds on the object 468fa9e4066Sahrens * contents, or ENOTSUP as described above. 469fa9e4066Sahrens */ 470fa9e4066Sahrens int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, 471fa9e4066Sahrens int ibs, dmu_tx_t *tx); 472fa9e4066Sahrens 473*eb633035STom Caputi /* 474*eb633035STom Caputi * Manually set the maxblkid on a dnode. This will adjust nlevels accordingly 475*eb633035STom Caputi * to accommodate the change. When calling this function, the caller must 476*eb633035STom Caputi * ensure that the object's nlevels can sufficiently support the new maxblkid. 477*eb633035STom Caputi */ 478*eb633035STom Caputi int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid, 479*eb633035STom Caputi dmu_tx_t *tx); 480*eb633035STom Caputi 481fa9e4066Sahrens /* 482fa9e4066Sahrens * Set the checksum property on a dnode. The new checksum algorithm will 483fa9e4066Sahrens * apply to all newly written blocks; existing blocks will not be affected. 484fa9e4066Sahrens */ 485fa9e4066Sahrens void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, 486fa9e4066Sahrens dmu_tx_t *tx); 487fa9e4066Sahrens 488fa9e4066Sahrens /* 489fa9e4066Sahrens * Set the compress property on a dnode. The new compression algorithm will 490fa9e4066Sahrens * apply to all newly written blocks; existing blocks will not be affected. 491fa9e4066Sahrens */ 492fa9e4066Sahrens void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, 493fa9e4066Sahrens dmu_tx_t *tx); 494fa9e4066Sahrens 4955cabbc6bSPrashanth Sreenivasa int dmu_object_remap_indirects(objset_t *os, uint64_t object, uint64_t txg); 4965cabbc6bSPrashanth Sreenivasa 4975d7b4d43SMatthew Ahrens void 4985d7b4d43SMatthew Ahrens dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, 4995d7b4d43SMatthew Ahrens void *data, uint8_t etype, uint8_t comp, int uncompressed_size, 5005d7b4d43SMatthew Ahrens int compressed_size, int byteorder, dmu_tx_t *tx); 5015d7b4d43SMatthew Ahrens 50244cd46caSbillm /* 503b24ab676SJeff Bonwick * Decide how to write a block: checksum, compression, number of copies, etc. 50444cd46caSbillm */ 505b24ab676SJeff Bonwick #define WP_NOFILL 0x1 506b24ab676SJeff Bonwick #define WP_DMU_SYNC 0x2 5070a586ceaSMark Shellenbaum #define WP_SPILL 0x4 508b24ab676SJeff Bonwick 50979d72832SMatthew Ahrens void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, 510adaec86aSMatthew Ahrens struct zio_prop *zp); 511*eb633035STom Caputi void dmu_write_policy_override_compress(struct zio_prop *zp, 512*eb633035STom Caputi enum zio_compress compress); 513*eb633035STom Caputi void dmu_write_policy_override_encrypt(struct zio_prop *zp, 514*eb633035STom Caputi dmu_object_type_t ot, boolean_t byteorder, enum zio_compress compress, 515*eb633035STom Caputi const uint8_t *salt, const uint8_t *iv, const uint8_t *mac); 516fa9e4066Sahrens /* 517fa9e4066Sahrens * The bonus data is accessed more or less like a regular buffer. 518fa9e4066Sahrens * You must dmu_bonus_hold() to get the buffer, which will give you a 519fa9e4066Sahrens * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus 5205cabbc6bSPrashanth Sreenivasa * data. As with any normal buffer, you must call dmu_buf_will_dirty() 5215cabbc6bSPrashanth Sreenivasa * before modifying it, and the 522fa9e4066Sahrens * object must be held in an assigned transaction before calling 523fa9e4066Sahrens * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus 524fa9e4066Sahrens * buffer as well. You must release your hold with dmu_buf_rele(). 5253e30c24aSWill Andrews * 5263e30c24aSWill Andrews * Returns ENOENT, EIO, or 0. 527fa9e4066Sahrens */ 528*eb633035STom Caputi int dmu_bonus_hold_impl(objset_t *os, uint64_t object, void *tag, 529*eb633035STom Caputi uint32_t flags, dmu_buf_t **dbp); 530ea8dc4b6Seschrock int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); 531*eb633035STom Caputi int dmu_bonus_hold_by_dnode(dnode_t *dn, void *tag, dmu_buf_t **dbp, 532*eb633035STom Caputi uint32_t flags); 533fa9e4066Sahrens int dmu_bonus_max(void); 5341934e92fSmaybee int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); 5350a586ceaSMark Shellenbaum int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); 536744947dcSTom Erickson dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); 5370a586ceaSMark Shellenbaum int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); 5380a586ceaSMark Shellenbaum 5390a586ceaSMark Shellenbaum /* 5400a586ceaSMark Shellenbaum * Special spill buffer support used by "SA" framework 5410a586ceaSMark Shellenbaum */ 5420a586ceaSMark Shellenbaum 543*eb633035STom Caputi int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, uint32_t flags, void *tag, 544*eb633035STom Caputi dmu_buf_t **dbp); 54579d72832SMatthew Ahrens int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, 5460a586ceaSMark Shellenbaum void *tag, dmu_buf_t **dbp); 5470a586ceaSMark Shellenbaum int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); 548fa9e4066Sahrens 549fa9e4066Sahrens /* 550fa9e4066Sahrens * Obtain the DMU buffer from the specified object which contains the 551fa9e4066Sahrens * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so 552fa9e4066Sahrens * that it will remain in memory. You must release the hold with 553fa9e4066Sahrens * dmu_buf_rele(). You musn't access the dmu_buf_t after releasing your 554fa9e4066Sahrens * hold. You must have a hold on any dmu_buf_t* you pass to the DMU. 555fa9e4066Sahrens * 556fa9e4066Sahrens * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill 557fa9e4066Sahrens * on the returned buffer before reading or writing the buffer's 558fa9e4066Sahrens * db_data. The comments for those routines describe what particular 559fa9e4066Sahrens * operations are valid after calling them. 560fa9e4066Sahrens * 561fa9e4066Sahrens * The object number must be a valid, allocated object number. 562fa9e4066Sahrens */ 563ea8dc4b6Seschrock int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, 56447cb52daSJeff Bonwick void *tag, dmu_buf_t **, int flags); 56579d72832SMatthew Ahrens int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, 56679d72832SMatthew Ahrens void *tag, dmu_buf_t **dbp, int flags); 567e57a022bSJustin T. Gibbs 568e57a022bSJustin T. Gibbs /* 569e57a022bSJustin T. Gibbs * Add a reference to a dmu buffer that has already been held via 570e57a022bSJustin T. Gibbs * dmu_buf_hold() in the current context. 571e57a022bSJustin T. Gibbs */ 572fa9e4066Sahrens void dmu_buf_add_ref(dmu_buf_t *db, void* tag); 573e57a022bSJustin T. Gibbs 574e57a022bSJustin T. Gibbs /* 575e57a022bSJustin T. Gibbs * Attempt to add a reference to a dmu buffer that is in an unknown state, 576e57a022bSJustin T. Gibbs * using a pointer that may have been invalidated by eviction processing. 577e57a022bSJustin T. Gibbs * The request will succeed if the passed in dbuf still represents the 578e57a022bSJustin T. Gibbs * same os/object/blkid, is ineligible for eviction, and has at least 579e57a022bSJustin T. Gibbs * one hold by a user other than the syncer. 580e57a022bSJustin T. Gibbs */ 581e57a022bSJustin T. Gibbs boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, 582e57a022bSJustin T. Gibbs uint64_t blkid, void *tag); 583e57a022bSJustin T. Gibbs 584ea8dc4b6Seschrock void dmu_buf_rele(dmu_buf_t *db, void *tag); 585fa9e4066Sahrens uint64_t dmu_buf_refcount(dmu_buf_t *db); 586*eb633035STom Caputi uint64_t dmu_buf_user_refcount(dmu_buf_t *db); 587fa9e4066Sahrens 588fa9e4066Sahrens /* 589fa9e4066Sahrens * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a 590fa9e4066Sahrens * range of an object. A pointer to an array of dmu_buf_t*'s is 591fa9e4066Sahrens * returned (in *dbpp). 592fa9e4066Sahrens * 593fa9e4066Sahrens * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and 594fa9e4066Sahrens * frees the array. The hold on the array of buffers MUST be released 595fa9e4066Sahrens * with dmu_buf_rele_array. You can NOT release the hold on each buffer 596fa9e4066Sahrens * individually with dmu_buf_rele. 597fa9e4066Sahrens */ 59813506d1eSmaybee int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, 599cf6106c8SMatthew Ahrens uint64_t length, boolean_t read, void *tag, 600cf6106c8SMatthew Ahrens int *numbufsp, dmu_buf_t ***dbpp); 6018dfe5547SRichard Yao int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length, 6028dfe5547SRichard Yao boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp, 6038dfe5547SRichard Yao uint32_t flags); 604ea8dc4b6Seschrock void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); 605fa9e4066Sahrens 606bc9014e6SJustin Gibbs typedef void dmu_buf_evict_func_t(void *user_ptr); 607bc9014e6SJustin Gibbs 608fa9e4066Sahrens /* 609bc9014e6SJustin Gibbs * A DMU buffer user object may be associated with a dbuf for the 610bc9014e6SJustin Gibbs * duration of its lifetime. This allows the user of a dbuf (client) 611bc9014e6SJustin Gibbs * to attach private data to a dbuf (e.g. in-core only data such as a 612bc9014e6SJustin Gibbs * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified 613bc9014e6SJustin Gibbs * when that dbuf has been evicted. Clients typically respond to the 614bc9014e6SJustin Gibbs * eviction notification by freeing their private data, thus ensuring 615bc9014e6SJustin Gibbs * the same lifetime for both dbuf and private data. 616fa9e4066Sahrens * 617bc9014e6SJustin Gibbs * The mapping from a dmu_buf_user_t to any client private data is the 618bc9014e6SJustin Gibbs * client's responsibility. All current consumers of the API with private 619bc9014e6SJustin Gibbs * data embed a dmu_buf_user_t as the first member of the structure for 620bc9014e6SJustin Gibbs * their private data. This allows conversions between the two types 621bc9014e6SJustin Gibbs * with a simple cast. Since the DMU buf user API never needs access 622bc9014e6SJustin Gibbs * to the private data, other strategies can be employed if necessary 62394c3dad2SToomas Soome * or convenient for the client (e.g. using __containerof() to do the 624bc9014e6SJustin Gibbs * conversion for private data that cannot have the dmu_buf_user_t as 625bc9014e6SJustin Gibbs * its first member). 626fa9e4066Sahrens * 627bc9014e6SJustin Gibbs * Eviction callbacks are executed without the dbuf mutex held or any 628bc9014e6SJustin Gibbs * other type of mechanism to guarantee that the dbuf is still available. 629bc9014e6SJustin Gibbs * For this reason, users must assume the dbuf has already been freed 630bc9014e6SJustin Gibbs * and not reference the dbuf from the callback context. 631fa9e4066Sahrens * 632bc9014e6SJustin Gibbs * Users requesting "immediate eviction" are notified as soon as the dbuf 633bc9014e6SJustin Gibbs * is only referenced by dirty records (dirties == holds). Otherwise the 634bc9014e6SJustin Gibbs * notification occurs after eviction processing for the dbuf begins. 635fa9e4066Sahrens */ 636bc9014e6SJustin Gibbs typedef struct dmu_buf_user { 637bc9014e6SJustin Gibbs /* 638bc9014e6SJustin Gibbs * Asynchronous user eviction callback state. 639bc9014e6SJustin Gibbs */ 640bc9014e6SJustin Gibbs taskq_ent_t dbu_tqent; 641bc9014e6SJustin Gibbs 64240510e8eSJosef 'Jeff' Sipek /* 64340510e8eSJosef 'Jeff' Sipek * This instance's eviction function pointers. 64440510e8eSJosef 'Jeff' Sipek * 64540510e8eSJosef 'Jeff' Sipek * dbu_evict_func_sync is called synchronously and then 64640510e8eSJosef 'Jeff' Sipek * dbu_evict_func_async is executed asynchronously on a taskq. 64740510e8eSJosef 'Jeff' Sipek */ 64840510e8eSJosef 'Jeff' Sipek dmu_buf_evict_func_t *dbu_evict_func_sync; 64940510e8eSJosef 'Jeff' Sipek dmu_buf_evict_func_t *dbu_evict_func_async; 650bc9014e6SJustin Gibbs #ifdef ZFS_DEBUG 651bc9014e6SJustin Gibbs /* 652bc9014e6SJustin Gibbs * Pointer to user's dbuf pointer. NULL for clients that do 653bc9014e6SJustin Gibbs * not associate a dbuf with their user data. 654bc9014e6SJustin Gibbs * 655bc9014e6SJustin Gibbs * The dbuf pointer is cleared upon eviction so as to catch 656bc9014e6SJustin Gibbs * use-after-evict bugs in clients. 657bc9014e6SJustin Gibbs */ 658bc9014e6SJustin Gibbs dmu_buf_t **dbu_clear_on_evict_dbufp; 659bc9014e6SJustin Gibbs #endif 660bc9014e6SJustin Gibbs } dmu_buf_user_t; 661bc9014e6SJustin Gibbs 662fa9e4066Sahrens /* 663bc9014e6SJustin Gibbs * Initialize the given dmu_buf_user_t instance with the eviction function 664bc9014e6SJustin Gibbs * evict_func, to be called when the user is evicted. 665bc9014e6SJustin Gibbs * 666bc9014e6SJustin Gibbs * NOTE: This function should only be called once on a given dmu_buf_user_t. 667bc9014e6SJustin Gibbs * To allow enforcement of this, dbu must already be zeroed on entry. 668fa9e4066Sahrens */ 6694ba5b961SDan Kimmel /*ARGSUSED*/ 670bc9014e6SJustin Gibbs inline void 67140510e8eSJosef 'Jeff' Sipek dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, 67240510e8eSJosef 'Jeff' Sipek dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp) 673bc9014e6SJustin Gibbs { 67440510e8eSJosef 'Jeff' Sipek ASSERT(dbu->dbu_evict_func_sync == NULL); 67540510e8eSJosef 'Jeff' Sipek ASSERT(dbu->dbu_evict_func_async == NULL); 67640510e8eSJosef 'Jeff' Sipek 67740510e8eSJosef 'Jeff' Sipek /* must have at least one evict func */ 67840510e8eSJosef 'Jeff' Sipek IMPLY(evict_func_sync == NULL, evict_func_async != NULL); 67940510e8eSJosef 'Jeff' Sipek dbu->dbu_evict_func_sync = evict_func_sync; 68040510e8eSJosef 'Jeff' Sipek dbu->dbu_evict_func_async = evict_func_async; 681bc9014e6SJustin Gibbs #ifdef ZFS_DEBUG 682bc9014e6SJustin Gibbs dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; 683bc9014e6SJustin Gibbs #endif 684bc9014e6SJustin Gibbs } 685fa9e4066Sahrens 686fa9e4066Sahrens /* 687bc9014e6SJustin Gibbs * Attach user data to a dbuf and mark it for normal (when the dbuf's 688bc9014e6SJustin Gibbs * data is cleared or its reference count goes to zero) eviction processing. 689bc9014e6SJustin Gibbs * 690bc9014e6SJustin Gibbs * Returns NULL on success, or the existing user if another user currently 691bc9014e6SJustin Gibbs * owns the buffer. 692bc9014e6SJustin Gibbs */ 693bc9014e6SJustin Gibbs void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); 694bc9014e6SJustin Gibbs 695bc9014e6SJustin Gibbs /* 696bc9014e6SJustin Gibbs * Attach user data to a dbuf and mark it for immediate (its dirty and 697bc9014e6SJustin Gibbs * reference counts are equal) eviction processing. 698bc9014e6SJustin Gibbs * 699bc9014e6SJustin Gibbs * Returns NULL on success, or the existing user if another user currently 700bc9014e6SJustin Gibbs * owns the buffer. 701bc9014e6SJustin Gibbs */ 702bc9014e6SJustin Gibbs void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); 703bc9014e6SJustin Gibbs 704bc9014e6SJustin Gibbs /* 705bc9014e6SJustin Gibbs * Replace the current user of a dbuf. 706bc9014e6SJustin Gibbs * 707bc9014e6SJustin Gibbs * If given the current user of a dbuf, replaces the dbuf's user with 708bc9014e6SJustin Gibbs * "new_user" and returns the user data pointer that was replaced. 709bc9014e6SJustin Gibbs * Otherwise returns the current, and unmodified, dbuf user pointer. 710bc9014e6SJustin Gibbs */ 711bc9014e6SJustin Gibbs void *dmu_buf_replace_user(dmu_buf_t *db, 712bc9014e6SJustin Gibbs dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); 713bc9014e6SJustin Gibbs 714bc9014e6SJustin Gibbs /* 715bc9014e6SJustin Gibbs * Remove the specified user data for a DMU buffer. 716bc9014e6SJustin Gibbs * 717bc9014e6SJustin Gibbs * Returns the user that was removed on success, or the current user if 718bc9014e6SJustin Gibbs * another user currently owns the buffer. 719bc9014e6SJustin Gibbs */ 720bc9014e6SJustin Gibbs void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); 721bc9014e6SJustin Gibbs 722bc9014e6SJustin Gibbs /* 723bc9014e6SJustin Gibbs * Returns the user data (dmu_buf_user_t *) associated with this dbuf. 724fa9e4066Sahrens */ 725fa9e4066Sahrens void *dmu_buf_get_user(dmu_buf_t *db); 726fa9e4066Sahrens 727ae972795SMatthew Ahrens objset_t *dmu_buf_get_objset(dmu_buf_t *db); 72879d72832SMatthew Ahrens dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db); 72979d72832SMatthew Ahrens void dmu_buf_dnode_exit(dmu_buf_t *db); 730ae972795SMatthew Ahrens 731bc9014e6SJustin Gibbs /* Block until any in-progress dmu buf user evictions complete. */ 732bc9014e6SJustin Gibbs void dmu_buf_user_evict_wait(void); 733bc9014e6SJustin Gibbs 73480901aeaSGeorge Wilson /* 73580901aeaSGeorge Wilson * Returns the blkptr associated with this dbuf, or NULL if not set. 73680901aeaSGeorge Wilson */ 73780901aeaSGeorge Wilson struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); 73880901aeaSGeorge Wilson 739fa9e4066Sahrens /* 740fa9e4066Sahrens * Indicate that you are going to modify the buffer's data (db_data). 741fa9e4066Sahrens * 742fa9e4066Sahrens * The transaction (tx) must be assigned to a txg (ie. you've called 743fa9e4066Sahrens * dmu_tx_assign()). The buffer's object must be held in the tx 744fa9e4066Sahrens * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). 745fa9e4066Sahrens */ 746fa9e4066Sahrens void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); 747*eb633035STom Caputi void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder, 748*eb633035STom Caputi const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx); 749fa9e4066Sahrens 750fa9e4066Sahrens /* 751fa9e4066Sahrens * You must create a transaction, then hold the objects which you will 752fa9e4066Sahrens * (or might) modify as part of this transaction. Then you must assign 753fa9e4066Sahrens * the transaction to a transaction group. Once the transaction has 754fa9e4066Sahrens * been assigned, you can modify buffers which belong to held objects as 755fa9e4066Sahrens * part of this transaction. You can't modify buffers before the 756fa9e4066Sahrens * transaction has been assigned; you can't modify buffers which don't 757fa9e4066Sahrens * belong to objects which this transaction holds; you can't hold 758fa9e4066Sahrens * objects once the transaction has been assigned. You may hold an 759fa9e4066Sahrens * object which you are going to free (with dmu_object_free()), but you 760fa9e4066Sahrens * don't have to. 761fa9e4066Sahrens * 762fa9e4066Sahrens * You can abort the transaction before it has been assigned. 763fa9e4066Sahrens * 764fa9e4066Sahrens * Note that you may hold buffers (with dmu_buf_hold) at any time, 765fa9e4066Sahrens * regardless of transaction state. 766fa9e4066Sahrens */ 767fa9e4066Sahrens 768fa9e4066Sahrens #define DMU_NEW_OBJECT (-1ULL) 769fa9e4066Sahrens #define DMU_OBJECT_END (-1ULL) 770fa9e4066Sahrens 771fa9e4066Sahrens dmu_tx_t *dmu_tx_create(objset_t *os); 772fa9e4066Sahrens void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); 773b0c42cd4Sbzzz void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, 774b0c42cd4Sbzzz int len); 775fa9e4066Sahrens void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, 776fa9e4066Sahrens uint64_t len); 777b0c42cd4Sbzzz void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, 778b0c42cd4Sbzzz uint64_t len); 7795cabbc6bSPrashanth Sreenivasa void dmu_tx_hold_remap_l1indirect(dmu_tx_t *tx, uint64_t object); 78014843421SMatthew Ahrens void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); 781b0c42cd4Sbzzz void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, 782b0c42cd4Sbzzz const char *name); 783fa9e4066Sahrens void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); 784b0c42cd4Sbzzz void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn); 7850a586ceaSMark Shellenbaum void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); 7860a586ceaSMark Shellenbaum void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); 7870a586ceaSMark Shellenbaum void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); 788fa9e4066Sahrens void dmu_tx_abort(dmu_tx_t *tx); 789f864f99eSPrakash Surya int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); 7908a2f1b91Sahrens void dmu_tx_wait(dmu_tx_t *tx); 791fa9e4066Sahrens void dmu_tx_commit(dmu_tx_t *tx); 7924bb73804SMatthew Ahrens void dmu_tx_mark_netfree(dmu_tx_t *tx); 793fa9e4066Sahrens 794d20e665cSRicardo M. Correia /* 795d20e665cSRicardo M. Correia * To register a commit callback, dmu_tx_callback_register() must be called. 796d20e665cSRicardo M. Correia * 797d20e665cSRicardo M. Correia * dcb_data is a pointer to caller private data that is passed on as a 798d20e665cSRicardo M. Correia * callback parameter. The caller is responsible for properly allocating and 799d20e665cSRicardo M. Correia * freeing it. 800d20e665cSRicardo M. Correia * 801d20e665cSRicardo M. Correia * When registering a callback, the transaction must be already created, but 802d20e665cSRicardo M. Correia * it cannot be committed or aborted. It can be assigned to a txg or not. 803d20e665cSRicardo M. Correia * 804d20e665cSRicardo M. Correia * The callback will be called after the transaction has been safely written 805d20e665cSRicardo M. Correia * to stable storage and will also be called if the dmu_tx is aborted. 806d20e665cSRicardo M. Correia * If there is any error which prevents the transaction from being committed to 807d20e665cSRicardo M. Correia * disk, the callback will be called with a value of error != 0. 808d20e665cSRicardo M. Correia */ 809d20e665cSRicardo M. Correia typedef void dmu_tx_callback_func_t(void *dcb_data, int error); 810d20e665cSRicardo M. Correia 811d20e665cSRicardo M. Correia void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, 812d20e665cSRicardo M. Correia void *dcb_data); 813d20e665cSRicardo M. Correia 814fa9e4066Sahrens /* 815fa9e4066Sahrens * Free up the data blocks for a defined range of a file. If size is 816ad135b5dSChristopher Siden * -1, the range from offset to end-of-file is freed. 817fa9e4066Sahrens */ 818ea8dc4b6Seschrock int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, 819*eb633035STom Caputi uint64_t size, dmu_tx_t *tx); 820cdb0ab79Smaybee int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, 821*eb633035STom Caputi uint64_t size); 822713d6c20SMatthew Ahrens int dmu_free_long_object(objset_t *os, uint64_t object); 823fa9e4066Sahrens 824fa9e4066Sahrens /* 825fa9e4066Sahrens * Convenience functions. 826fa9e4066Sahrens * 827fa9e4066Sahrens * Canfail routines will return 0 on success, or an errno if there is a 828fa9e4066Sahrens * nonrecoverable I/O error. 829fa9e4066Sahrens */ 8307bfdf011SNeil Perrin #define DMU_READ_PREFETCH 0 /* prefetch */ 8317bfdf011SNeil Perrin #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ 832*eb633035STom Caputi #define DMU_READ_NO_DECRYPT 2 /* don't decrypt */ 833ea8dc4b6Seschrock int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 8347bfdf011SNeil Perrin void *buf, uint32_t flags); 835b0c42cd4Sbzzz int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, 836b0c42cd4Sbzzz uint32_t flags); 837fa9e4066Sahrens void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 838fa9e4066Sahrens const void *buf, dmu_tx_t *tx); 839b0c42cd4Sbzzz void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, 840b0c42cd4Sbzzz const void *buf, dmu_tx_t *tx); 84182c9918fSTim Haley void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, 84282c9918fSTim Haley dmu_tx_t *tx); 843feb08c6bSbillm int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); 844f8554bb9SMatthew Ahrens int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size); 8458dfe5547SRichard Yao int dmu_read_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size); 846feb08c6bSbillm int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, 847feb08c6bSbillm dmu_tx_t *tx); 84894d1a210STim Haley int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, 84994d1a210STim Haley dmu_tx_t *tx); 8508dfe5547SRichard Yao int dmu_write_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size, 8518dfe5547SRichard Yao dmu_tx_t *tx); 85244eda4d7Smaybee int dmu_write_pages(objset_t *os, uint64_t object, uint64_t offset, 85344eda4d7Smaybee uint64_t size, struct page *pp, dmu_tx_t *tx); 8542fdbea25SAleksandr Guzovskiy struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); 8552fdbea25SAleksandr Guzovskiy void dmu_return_arcbuf(struct arc_buf *buf); 856*eb633035STom Caputi int dmu_assign_arcbuf_by_dnode(dnode_t *handle, uint64_t offset, 8578dfe5547SRichard Yao struct arc_buf *buf, dmu_tx_t *tx); 858*eb633035STom Caputi int dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset, 859*eb633035STom Caputi struct arc_buf *buf, dmu_tx_t *tx); 860*eb633035STom Caputi void dmu_convert_to_raw(dmu_buf_t *handle, boolean_t byteorder, 861*eb633035STom Caputi const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx); 862*eb633035STom Caputi #define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf 863*eb633035STom Caputi void dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset, 864*eb633035STom Caputi dmu_buf_t *handle, dmu_tx_t *tx); 865c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_init(struct xuio *uio, int niov); 866c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void dmu_xuio_fini(struct xuio *uio); 867c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, 868c242f9a0Schunli zhang - Sun Microsystems - Irvine United States size_t n); 869c242f9a0Schunli zhang - Sun Microsystems - Irvine United States int dmu_xuio_cnt(struct xuio *uio); 870c242f9a0Schunli zhang - Sun Microsystems - Irvine United States struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); 871c242f9a0Schunli zhang - Sun Microsystems - Irvine United States void dmu_xuio_clear(struct xuio *uio, int i); 87299aa8b55SPrashanth Sreenivasa void xuio_stat_wbuf_copied(void); 87399aa8b55SPrashanth Sreenivasa void xuio_stat_wbuf_nocopy(void); 874fa9e4066Sahrens 875cf6106c8SMatthew Ahrens extern boolean_t zfs_prefetch_disable; 876b5152584SMatthew Ahrens extern int zfs_max_recordsize; 877416e0cd8Sek 878fa9e4066Sahrens /* 879fa9e4066Sahrens * Asynchronously try to read in the data. 880fa9e4066Sahrens */ 881a2cdcdd2SPaul Dagnelie void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, 882a2cdcdd2SPaul Dagnelie uint64_t len, enum zio_priority pri); 883fa9e4066Sahrens 884fa9e4066Sahrens typedef struct dmu_object_info { 885b24ab676SJeff Bonwick /* All sizes are in bytes unless otherwise indicated. */ 886fa9e4066Sahrens uint32_t doi_data_block_size; 887fa9e4066Sahrens uint32_t doi_metadata_block_size; 888fa9e4066Sahrens dmu_object_type_t doi_type; 889fa9e4066Sahrens dmu_object_type_t doi_bonus_type; 890b24ab676SJeff Bonwick uint64_t doi_bonus_size; 891fa9e4066Sahrens uint8_t doi_indirection; /* 2 = dnode->indirect->data */ 892fa9e4066Sahrens uint8_t doi_checksum; 893fa9e4066Sahrens uint8_t doi_compress; 894e77d42eaSMatthew Ahrens uint8_t doi_nblkptr; 89554811da5SToomas Soome int8_t doi_pad[4]; 89654811da5SToomas Soome uint64_t doi_dnodesize; 897b24ab676SJeff Bonwick uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ 898b24ab676SJeff Bonwick uint64_t doi_max_offset; 899b24ab676SJeff Bonwick uint64_t doi_fill_count; /* number of non-empty blocks */ 900fa9e4066Sahrens } dmu_object_info_t; 901fa9e4066Sahrens 902e45ce728Sahrens typedef void arc_byteswap_func_t(void *buf, size_t size); 903e45ce728Sahrens 904fa9e4066Sahrens typedef struct dmu_object_type_info { 905ad135b5dSChristopher Siden dmu_object_byteswap_t ot_byteswap; 906fa9e4066Sahrens boolean_t ot_metadata; 907adb52d92SMatthew Ahrens boolean_t ot_dbuf_metadata_cache; 908*eb633035STom Caputi boolean_t ot_encrypt; 909fa9e4066Sahrens char *ot_name; 910fa9e4066Sahrens } dmu_object_type_info_t; 911fa9e4066Sahrens 912ad135b5dSChristopher Siden typedef struct dmu_object_byteswap_info { 913ad135b5dSChristopher Siden arc_byteswap_func_t *ob_func; 914ad135b5dSChristopher Siden char *ob_name; 915ad135b5dSChristopher Siden } dmu_object_byteswap_info_t; 916ad135b5dSChristopher Siden 917fa9e4066Sahrens extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; 918ad135b5dSChristopher Siden extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; 919fa9e4066Sahrens 920fa9e4066Sahrens /* 921fa9e4066Sahrens * Get information on a DMU object. 922fa9e4066Sahrens * 923fa9e4066Sahrens * Return 0 on success or ENOENT if object is not allocated. 924fa9e4066Sahrens * 925fa9e4066Sahrens * If doi is NULL, just indicates whether the object exists. 926fa9e4066Sahrens */ 927fa9e4066Sahrens int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); 9283e30c24aSWill Andrews /* Like dmu_object_info, but faster if you have a held dnode in hand. */ 92979d72832SMatthew Ahrens void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi); 9303e30c24aSWill Andrews /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ 931fa9e4066Sahrens void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); 9323e30c24aSWill Andrews /* 9333e30c24aSWill Andrews * Like dmu_object_info_from_db, but faster still when you only care about 9343e30c24aSWill Andrews * the size. This is specifically optimized for zfs_getattr(). 9353e30c24aSWill Andrews */ 936fa9e4066Sahrens void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, 937fa9e4066Sahrens u_longlong_t *nblk512); 938fa9e4066Sahrens 93954811da5SToomas Soome void dmu_object_dnsize_from_db(dmu_buf_t *db, int *dnsize); 94054811da5SToomas Soome 941fa9e4066Sahrens typedef struct dmu_objset_stats { 942a2eea2e1Sahrens uint64_t dds_num_clones; /* number of clones of this */ 943a2eea2e1Sahrens uint64_t dds_creation_txg; 9443cb34c60Sahrens uint64_t dds_guid; 945fa9e4066Sahrens dmu_objset_type_t dds_type; 946fa9e4066Sahrens uint8_t dds_is_snapshot; 94731fd60d3Sahrens uint8_t dds_inconsistent; 9489adfa60dSMatthew Ahrens char dds_origin[ZFS_MAX_DATASET_NAME_LEN]; 949a2eea2e1Sahrens } dmu_objset_stats_t; 950fa9e4066Sahrens 951a2eea2e1Sahrens /* 952a2eea2e1Sahrens * Get stats on a dataset. 953a2eea2e1Sahrens */ 954a2eea2e1Sahrens void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); 955fa9e4066Sahrens 956a2eea2e1Sahrens /* 957a2eea2e1Sahrens * Add entries to the nvlist for all the objset's properties. See 958a2eea2e1Sahrens * zfs_prop_table[] and zfs(1m) for details on the properties. 959a2eea2e1Sahrens */ 960a2eea2e1Sahrens void dmu_objset_stats(objset_t *os, struct nvlist *nv); 961fa9e4066Sahrens 962a2eea2e1Sahrens /* 963a2eea2e1Sahrens * Get the space usage statistics for statvfs(). 964a2eea2e1Sahrens * 965a2eea2e1Sahrens * refdbytes is the amount of space "referenced" by this objset. 966a2eea2e1Sahrens * availbytes is the amount of space available to this objset, taking 967a2eea2e1Sahrens * into account quotas & reservations, assuming that no other objsets 968a2eea2e1Sahrens * use the space first. These values correspond to the 'referenced' and 969a2eea2e1Sahrens * 'available' properties, described in the zfs(1m) manpage. 970a2eea2e1Sahrens * 971a2eea2e1Sahrens * usedobjs and availobjs are the number of objects currently allocated, 972a2eea2e1Sahrens * and available. 973a2eea2e1Sahrens */ 974a2eea2e1Sahrens void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, 975a2eea2e1Sahrens uint64_t *usedobjsp, uint64_t *availobjsp); 976fa9e4066Sahrens 977fa9e4066Sahrens /* 978a2eea2e1Sahrens * The fsid_guid is a 56-bit ID that can change to avoid collisions. 979a2eea2e1Sahrens * (Contrast with the ds_guid which is a 64-bit ID that will never 980a2eea2e1Sahrens * change, so there is a small probability that it will collide.) 981fa9e4066Sahrens */ 982a2eea2e1Sahrens uint64_t dmu_objset_fsid_guid(objset_t *os); 983fa9e4066Sahrens 98471eb0538SChris Kirby /* 98571eb0538SChris Kirby * Get the [cm]time for an objset's snapshot dir 98671eb0538SChris Kirby */ 98771eb0538SChris Kirby timestruc_t dmu_objset_snap_cmtime(objset_t *os); 98871eb0538SChris Kirby 989fa9e4066Sahrens int dmu_objset_is_snapshot(objset_t *os); 990fa9e4066Sahrens 991fa9e4066Sahrens extern struct spa *dmu_objset_spa(objset_t *os); 992fa9e4066Sahrens extern struct zilog *dmu_objset_zil(objset_t *os); 993fa9e4066Sahrens extern struct dsl_pool *dmu_objset_pool(objset_t *os); 994fa9e4066Sahrens extern struct dsl_dataset *dmu_objset_ds(objset_t *os); 995fa9e4066Sahrens extern void dmu_objset_name(objset_t *os, char *buf); 996fa9e4066Sahrens extern dmu_objset_type_t dmu_objset_type(objset_t *os); 997fa9e4066Sahrens extern uint64_t dmu_objset_id(objset_t *os); 99854811da5SToomas Soome extern uint64_t dmu_objset_dnodesize(objset_t *os); 999edf345e6SMatthew Ahrens extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); 1000edf345e6SMatthew Ahrens extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); 1001fa9e4066Sahrens extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, 1002b38f0970Sck uint64_t *id, uint64_t *offp, boolean_t *case_conflict); 1003ab04eb8eStimh extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, 1004ab04eb8eStimh int maxlen, boolean_t *conflict); 100587e5029aSahrens extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, 100687e5029aSahrens uint64_t *idp, uint64_t *offp); 100714843421SMatthew Ahrens 10089966ca11SMatthew Ahrens typedef int objset_used_cb_t(dmu_object_type_t bonustype, 10099966ca11SMatthew Ahrens void *bonus, uint64_t *userp, uint64_t *groupp); 101014843421SMatthew Ahrens extern void dmu_objset_register_type(dmu_objset_type_t ost, 101114843421SMatthew Ahrens objset_used_cb_t *cb); 1012f18faf3fSek extern void dmu_objset_set_user(objset_t *os, void *user_ptr); 1013f18faf3fSek extern void *dmu_objset_get_user(objset_t *os); 1014fa9e4066Sahrens 1015fa9e4066Sahrens /* 1016fa9e4066Sahrens * Return the txg number for the given assigned transaction. 1017fa9e4066Sahrens */ 101887e5029aSahrens uint64_t dmu_tx_get_txg(dmu_tx_t *tx); 1019fa9e4066Sahrens 1020fa9e4066Sahrens /* 1021fa9e4066Sahrens * Synchronous write. 1022c5c6ffa0Smaybee * If a parent zio is provided this function initiates a write on the 1023c5c6ffa0Smaybee * provided buffer as a child of the parent zio. 1024da6c28aaSamw * In the absence of a parent zio, the write is completed synchronously. 1025c5c6ffa0Smaybee * At write completion, blk is filled with the bp of the written block. 1026fa9e4066Sahrens * Note that while the data covered by this function will be on stable 1027c5c6ffa0Smaybee * storage when the write completes this new data does not become a 1028fa9e4066Sahrens * permanent part of the file until the associated transaction commits. 1029fa9e4066Sahrens */ 1030b24ab676SJeff Bonwick 1031b24ab676SJeff Bonwick /* 1032b24ab676SJeff Bonwick * {zfs,zvol,ztest}_get_done() args 1033b24ab676SJeff Bonwick */ 1034b24ab676SJeff Bonwick typedef struct zgd { 10351271e4b1SPrakash Surya struct lwb *zgd_lwb; 1036b24ab676SJeff Bonwick struct blkptr *zgd_bp; 1037b24ab676SJeff Bonwick dmu_buf_t *zgd_db; 103879315247SMatthew Ahrens struct locked_range *zgd_lr; 1039b24ab676SJeff Bonwick void *zgd_private; 1040b24ab676SJeff Bonwick } zgd_t; 1041b24ab676SJeff Bonwick 1042b24ab676SJeff Bonwick typedef void dmu_sync_cb_t(zgd_t *arg, int error); 1043b24ab676SJeff Bonwick int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); 1044fa9e4066Sahrens 1045fa9e4066Sahrens /* 1046fa9e4066Sahrens * Find the next hole or data block in file starting at *off 1047fa9e4066Sahrens * Return found offset in *off. Return ESRCH for end of file. 1048fa9e4066Sahrens */ 1049fa9e4066Sahrens int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, 1050fa9e4066Sahrens uint64_t *off); 1051fa9e4066Sahrens 10522bcf0248SMax Grossman /* 10532bcf0248SMax Grossman * Check if a DMU object has any dirty blocks. If so, sync out 10542bcf0248SMax Grossman * all pending transaction groups. Otherwise, this function 10552bcf0248SMax Grossman * does not alter DMU state. This could be improved to only sync 10562bcf0248SMax Grossman * out the necessary transaction groups for this particular 10572bcf0248SMax Grossman * object. 10582bcf0248SMax Grossman */ 10592bcf0248SMax Grossman int dmu_object_wait_synced(objset_t *os, uint64_t object); 10602bcf0248SMax Grossman 1061fa9e4066Sahrens /* 1062fa9e4066Sahrens * Initial setup and final teardown. 1063fa9e4066Sahrens */ 1064fa9e4066Sahrens extern void dmu_init(void); 1065fa9e4066Sahrens extern void dmu_fini(void); 1066fa9e4066Sahrens 1067fa9e4066Sahrens typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, 1068fa9e4066Sahrens uint64_t object, uint64_t offset, int len); 1069fa9e4066Sahrens void dmu_traverse_objset(objset_t *os, uint64_t txg_start, 1070fa9e4066Sahrens dmu_traverse_cb_t cb, void *arg); 1071fa9e4066Sahrens 10723b2aab18SMatthew Ahrens int dmu_diff(const char *tosnap_name, const char *fromsnap_name, 10733b2aab18SMatthew Ahrens struct vnode *vp, offset_t *offp); 107499d5e173STim Haley 1075fa9e4066Sahrens /* CRC64 table */ 1076fa9e4066Sahrens #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ 1077fa9e4066Sahrens extern uint64_t zfs_crc64_table[256]; 1078fa9e4066Sahrens 1079fa9e4066Sahrens #ifdef __cplusplus 1080fa9e4066Sahrens } 1081fa9e4066Sahrens #endif 1082fa9e4066Sahrens 1083fa9e4066Sahrens #endif /* _SYS_DMU_H */ 1084