diff --git a/include/sys/dbuf.h b/include/sys/dbuf.h index 32f036862cc1..ddce7deefcc0 100644 --- a/include/sys/dbuf.h +++ b/include/sys/dbuf.h @@ -1,408 +1,417 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2015 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. */ #ifndef _SYS_DBUF_H #define _SYS_DBUF_H #include #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif #define IN_DMU_SYNC 2 /* * define flags for dbuf_read */ #define DB_RF_MUST_SUCCEED (1 << 0) #define DB_RF_CANFAIL (1 << 1) #define DB_RF_HAVESTRUCT (1 << 2) #define DB_RF_NOPREFETCH (1 << 3) #define DB_RF_NEVERWAIT (1 << 4) #define DB_RF_CACHED (1 << 5) #define DB_RF_NO_DECRYPT (1 << 6) /* * The simplified state transition diagram for dbufs looks like: * * +----> READ ----+ * | | * | V * (alloc)-->UNCACHED CACHED-->EVICTING-->(free) * | ^ ^ * | | | * +----> FILL ----+ | * | | * | | * +--------> NOFILL -------+ * * DB_SEARCH is an invalid state for a dbuf. It is used by dbuf_free_range * to find all dbufs in a range of a dnode and must be less than any other * dbuf_states_t (see comment on dn_dbufs in dnode.h). */ typedef enum dbuf_states { DB_SEARCH = -1, DB_UNCACHED, DB_FILL, DB_NOFILL, DB_READ, DB_CACHED, DB_EVICTING } dbuf_states_t; struct dnode; struct dmu_tx; /* * level = 0 means the user data * level = 1 means the single indirect block * etc. */ struct dmu_buf_impl; typedef enum override_states { DR_NOT_OVERRIDDEN, DR_IN_DMU_SYNC, DR_OVERRIDDEN } override_states_t; typedef struct dbuf_dirty_record { /* link on our parents dirty list */ list_node_t dr_dirty_node; /* transaction group this data will sync in */ uint64_t dr_txg; /* zio of outstanding write IO */ zio_t *dr_zio; /* pointer back to our dbuf */ struct dmu_buf_impl *dr_dbuf; /* pointer to next dirty record */ struct dbuf_dirty_record *dr_next; /* pointer to parent dirty record */ struct dbuf_dirty_record *dr_parent; /* How much space was changed to dsl_pool_dirty_space() for this? */ unsigned int dr_accounted; /* A copy of the bp that points to us */ blkptr_t dr_bp_copy; union dirty_types { struct dirty_indirect { /* protect access to list */ kmutex_t dr_mtx; /* Our list of dirty children */ list_t dr_children; } di; struct dirty_leaf { /* * dr_data is set when we dirty the buffer * so that we can retain the pointer even if it * gets COW'd in a subsequent transaction group. */ arc_buf_t *dr_data; blkptr_t dr_overridden_by; override_states_t dr_override_state; uint8_t dr_copies; boolean_t dr_nopwrite; - boolean_t dr_raw; + boolean_t dr_has_raw_params; + + /* + * If dr_has_raw_params is set, the following crypt + * params will be set on the BP that's written. + */ + boolean_t dr_byteorder; + uint8_t dr_salt[ZIO_DATA_SALT_LEN]; + uint8_t dr_iv[ZIO_DATA_IV_LEN]; + uint8_t dr_mac[ZIO_DATA_MAC_LEN]; } dl; } dt; } dbuf_dirty_record_t; typedef struct dmu_buf_impl { /* * The following members are immutable, with the exception of * db.db_data, which is protected by db_mtx. */ /* the publicly visible structure */ dmu_buf_t db; /* the objset we belong to */ struct objset *db_objset; /* * handle to safely access the dnode we belong to (NULL when evicted) */ struct dnode_handle *db_dnode_handle; /* * our parent buffer; if the dnode points to us directly, * db_parent == db_dnode_handle->dnh_dnode->dn_dbuf * only accessed by sync thread ??? * (NULL when evicted) * May change from NULL to non-NULL under the protection of db_mtx * (see dbuf_check_blkptr()) */ struct dmu_buf_impl *db_parent; /* * link for hash table of all dmu_buf_impl_t's */ struct dmu_buf_impl *db_hash_next; /* our block number */ uint64_t db_blkid; /* * Pointer to the blkptr_t which points to us. May be NULL if we * don't have one yet. (NULL when evicted) */ blkptr_t *db_blkptr; /* * Our indirection level. Data buffers have db_level==0. * Indirect buffers which point to data buffers have * db_level==1. etc. Buffers which contain dnodes have * db_level==0, since the dnodes are stored in a file. */ uint8_t db_level; /* db_mtx protects the members below */ kmutex_t db_mtx; /* * Current state of the buffer */ dbuf_states_t db_state; /* * Refcount accessed by dmu_buf_{hold,rele}. * If nonzero, the buffer can't be destroyed. * Protected by db_mtx. */ refcount_t db_holds; /* buffer holding our data */ arc_buf_t *db_buf; kcondvar_t db_changed; dbuf_dirty_record_t *db_data_pending; /* pointer to most recent dirty record for this buffer */ dbuf_dirty_record_t *db_last_dirty; /* * Our link on the owner dnodes's dn_dbufs list. * Protected by its dn_dbufs_mtx. */ avl_node_t db_link; /* * Link in dbuf_cache. */ multilist_node_t db_cache_link; /* Data which is unique to data (leaf) blocks: */ /* User callback information. */ dmu_buf_user_t *db_user; /* * Evict user data as soon as the dirty and reference * counts are equal. */ uint8_t db_user_immediate_evict; /* * This block was freed while a read or write was * active. */ uint8_t db_freed_in_flight; /* * dnode_evict_dbufs() or dnode_evict_bonus() tried to * evict this dbuf, but couldn't due to outstanding * references. Evict once the refcount drops to 0. */ uint8_t db_pending_evict; uint8_t db_dirtycnt; } dmu_buf_impl_t; /* Note: the dbuf hash table is exposed only for the mdb module */ #define DBUF_MUTEXES 8192 #define DBUF_HASH_MUTEX(h, idx) (&(h)->hash_mutexes[(idx) & (DBUF_MUTEXES-1)]) typedef struct dbuf_hash_table { uint64_t hash_table_mask; dmu_buf_impl_t **hash_table; kmutex_t hash_mutexes[DBUF_MUTEXES]; } dbuf_hash_table_t; uint64_t dbuf_whichblock(const struct dnode *di, const int64_t level, const uint64_t offset); void dbuf_create_bonus(struct dnode *dn); int dbuf_spill_set_blksz(dmu_buf_t *db, uint64_t blksz, dmu_tx_t *tx); void dbuf_rm_spill(struct dnode *dn, dmu_tx_t *tx); dmu_buf_impl_t *dbuf_hold(struct dnode *dn, uint64_t blkid, void *tag); dmu_buf_impl_t *dbuf_hold_level(struct dnode *dn, int level, uint64_t blkid, void *tag); int dbuf_hold_impl(struct dnode *dn, uint8_t level, uint64_t blkid, boolean_t fail_sparse, boolean_t fail_uncached, void *tag, dmu_buf_impl_t **dbp); void dbuf_prefetch(struct dnode *dn, int64_t level, uint64_t blkid, zio_priority_t prio, arc_flags_t aflags); void dbuf_add_ref(dmu_buf_impl_t *db, void *tag); boolean_t dbuf_try_add_ref(dmu_buf_t *db, objset_t *os, uint64_t obj, uint64_t blkid, void *tag); uint64_t dbuf_refcount(dmu_buf_impl_t *db); void dbuf_rele(dmu_buf_impl_t *db, void *tag); void dbuf_rele_and_unlock(dmu_buf_impl_t *db, void *tag); dmu_buf_impl_t *dbuf_find(struct objset *os, uint64_t object, uint8_t level, uint64_t blkid); int dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags); void dmu_buf_will_not_fill(dmu_buf_t *db, dmu_tx_t *tx); void dmu_buf_will_fill(dmu_buf_t *db, dmu_tx_t *tx); void dmu_buf_fill_done(dmu_buf_t *db, dmu_tx_t *tx); void dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx); dbuf_dirty_record_t *dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx); arc_buf_t *dbuf_loan_arcbuf(dmu_buf_impl_t *db); void dmu_buf_write_embedded(dmu_buf_t *dbuf, void *data, bp_embedded_type_t etype, enum zio_compress comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx); void dbuf_destroy(dmu_buf_impl_t *db); void dbuf_unoverride(dbuf_dirty_record_t *dr); void dbuf_sync_list(list_t *list, int level, dmu_tx_t *tx); void dbuf_release_bp(dmu_buf_impl_t *db); boolean_t dbuf_can_remap(const dmu_buf_impl_t *buf); void dbuf_free_range(struct dnode *dn, uint64_t start, uint64_t end, struct dmu_tx *); void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx); void dbuf_stats_init(dbuf_hash_table_t *hash); void dbuf_stats_destroy(void); #define DB_DNODE(_db) ((_db)->db_dnode_handle->dnh_dnode) #define DB_DNODE_LOCK(_db) ((_db)->db_dnode_handle->dnh_zrlock) #define DB_DNODE_ENTER(_db) (zrl_add(&DB_DNODE_LOCK(_db))) #define DB_DNODE_EXIT(_db) (zrl_remove(&DB_DNODE_LOCK(_db))) #define DB_DNODE_HELD(_db) (!zrl_is_zero(&DB_DNODE_LOCK(_db))) void dbuf_init(void); void dbuf_fini(void); boolean_t dbuf_is_metadata(dmu_buf_impl_t *db); #define DBUF_GET_BUFC_TYPE(_db) \ (dbuf_is_metadata(_db) ? ARC_BUFC_METADATA : ARC_BUFC_DATA) #define DBUF_IS_CACHEABLE(_db) \ ((_db)->db_objset->os_primary_cache == ZFS_CACHE_ALL || \ (dbuf_is_metadata(_db) && \ ((_db)->db_objset->os_primary_cache == ZFS_CACHE_METADATA))) #define DBUF_IS_L2CACHEABLE(_db) \ ((_db)->db_objset->os_secondary_cache == ZFS_CACHE_ALL || \ (dbuf_is_metadata(_db) && \ ((_db)->db_objset->os_secondary_cache == ZFS_CACHE_METADATA))) #define DNODE_LEVEL_IS_L2CACHEABLE(_dn, _level) \ ((_dn)->dn_objset->os_secondary_cache == ZFS_CACHE_ALL || \ (((_level) > 0 || \ DMU_OT_IS_METADATA((_dn)->dn_handle->dnh_dnode->dn_type)) && \ ((_dn)->dn_objset->os_secondary_cache == ZFS_CACHE_METADATA))) #ifdef ZFS_DEBUG /* * There should be a ## between the string literal and fmt, to make it * clear that we're joining two strings together, but gcc does not * support that preprocessor token. */ #define dprintf_dbuf(dbuf, fmt, ...) do { \ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ char __db_buf[32]; \ uint64_t __db_obj = (dbuf)->db.db_object; \ if (__db_obj == DMU_META_DNODE_OBJECT) \ (void) strcpy(__db_buf, "mdn"); \ else \ (void) snprintf(__db_buf, sizeof (__db_buf), "%lld", \ (u_longlong_t)__db_obj); \ dprintf_ds((dbuf)->db_objset->os_dsl_dataset, \ "obj=%s lvl=%u blkid=%lld " fmt, \ __db_buf, (dbuf)->db_level, \ (u_longlong_t)(dbuf)->db_blkid, __VA_ARGS__); \ } \ _NOTE(CONSTCOND) } while (0) #define dprintf_dbuf_bp(db, bp, fmt, ...) do { \ if (zfs_flags & ZFS_DEBUG_DPRINTF) { \ char *__blkbuf = kmem_alloc(BP_SPRINTF_LEN, KM_SLEEP); \ snprintf_blkptr(__blkbuf, BP_SPRINTF_LEN, bp); \ dprintf_dbuf(db, fmt " %s\n", __VA_ARGS__, __blkbuf); \ kmem_free(__blkbuf, BP_SPRINTF_LEN); \ } \ _NOTE(CONSTCOND) } while (0) #define DBUF_VERIFY(db) dbuf_verify(db) #else #define dprintf_dbuf(db, fmt, ...) #define dprintf_dbuf_bp(db, bp, fmt, ...) #define DBUF_VERIFY(db) #endif #ifdef __cplusplus } #endif #endif /* _SYS_DBUF_H */ diff --git a/include/sys/dmu.h b/include/sys/dmu.h index ccc07006d30a..119584365a2c 100644 --- a/include/sys/dmu.h +++ b/include/sys/dmu.h @@ -1,1057 +1,1051 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2017 by Delphix. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, Joyent, Inc. All rights reserved. * Copyright 2014 HybridCluster. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright 2013 Saso Kiselkov. All rights reserved. */ /* Portions Copyright 2010 Robert Milkowski */ #ifndef _SYS_DMU_H #define _SYS_DMU_H /* * This file describes the interface that the DMU provides for its * consumers. * * The DMU also interacts with the SPA. That interface is described in * dmu_spa.h. */ #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif struct page; struct vnode; struct spa; struct zilog; struct zio; struct blkptr; struct zap_cursor; struct dsl_dataset; struct dsl_pool; struct dnode; struct drr_begin; struct drr_end; struct zbookmark_phys; struct spa; struct nvlist; struct arc_buf; struct zio_prop; struct sa_handle; struct dsl_crypto_params; typedef struct objset objset_t; typedef struct dmu_tx dmu_tx_t; typedef struct dsl_dir dsl_dir_t; typedef struct dnode dnode_t; typedef enum dmu_object_byteswap { DMU_BSWAP_UINT8, DMU_BSWAP_UINT16, DMU_BSWAP_UINT32, DMU_BSWAP_UINT64, DMU_BSWAP_ZAP, DMU_BSWAP_DNODE, DMU_BSWAP_OBJSET, DMU_BSWAP_ZNODE, DMU_BSWAP_OLDACL, DMU_BSWAP_ACL, /* * Allocating a new byteswap type number makes the on-disk format * incompatible with any other format that uses the same number. * * Data can usually be structured to work with one of the * DMU_BSWAP_UINT* or DMU_BSWAP_ZAP types. */ DMU_BSWAP_NUMFUNCS } dmu_object_byteswap_t; #define DMU_OT_NEWTYPE 0x80 #define DMU_OT_METADATA 0x40 #define DMU_OT_ENCRYPTED 0x20 #define DMU_OT_BYTESWAP_MASK 0x1f /* * Defines a uint8_t object type. Object types specify if the data * in the object is metadata (boolean) and how to byteswap the data * (dmu_object_byteswap_t). */ #define DMU_OT(byteswap, metadata, encrypted) \ (DMU_OT_NEWTYPE | \ ((metadata) ? DMU_OT_METADATA : 0) | \ ((encrypted) ? DMU_OT_ENCRYPTED : 0) | \ ((byteswap) & DMU_OT_BYTESWAP_MASK)) #define DMU_OT_IS_VALID(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_BYTESWAP_MASK) < DMU_BSWAP_NUMFUNCS : \ (ot) < DMU_OT_NUMTYPES) /* * MDB doesn't have dmu_ot; it defines these macros itself. */ #ifndef ZFS_MDB #define DMU_OT_IS_METADATA_IMPL(ot) (dmu_ot[ot].ot_metadata) #define DMU_OT_IS_ENCRYPTED_IMPL(ot) (dmu_ot[ot].ot_encrypt) #define DMU_OT_BYTESWAP_IMPL(ot) (dmu_ot[ot].ot_byteswap) #endif #define DMU_OT_IS_METADATA(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_METADATA) : \ DMU_OT_IS_METADATA_IMPL(ot)) #define DMU_OT_IS_ENCRYPTED(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_ENCRYPTED) : \ DMU_OT_IS_ENCRYPTED_IMPL(ot)) /* * These object types use bp_fill != 1 for their L0 bp's. Therefore they can't * have their data embedded (i.e. use a BP_IS_EMBEDDED() bp), because bp_fill * is repurposed for embedded BPs. */ #define DMU_OT_HAS_FILL(ot) \ ((ot) == DMU_OT_DNODE || (ot) == DMU_OT_OBJSET) #define DMU_OT_BYTESWAP(ot) (((ot) & DMU_OT_NEWTYPE) ? \ ((ot) & DMU_OT_BYTESWAP_MASK) : \ DMU_OT_BYTESWAP_IMPL(ot)) typedef enum dmu_object_type { DMU_OT_NONE, /* general: */ DMU_OT_OBJECT_DIRECTORY, /* ZAP */ DMU_OT_OBJECT_ARRAY, /* UINT64 */ DMU_OT_PACKED_NVLIST, /* UINT8 (XDR by nvlist_pack/unpack) */ DMU_OT_PACKED_NVLIST_SIZE, /* UINT64 */ DMU_OT_BPOBJ, /* UINT64 */ DMU_OT_BPOBJ_HDR, /* UINT64 */ /* spa: */ DMU_OT_SPACE_MAP_HEADER, /* UINT64 */ DMU_OT_SPACE_MAP, /* UINT64 */ /* zil: */ DMU_OT_INTENT_LOG, /* UINT64 */ /* dmu: */ DMU_OT_DNODE, /* DNODE */ DMU_OT_OBJSET, /* OBJSET */ /* dsl: */ DMU_OT_DSL_DIR, /* UINT64 */ DMU_OT_DSL_DIR_CHILD_MAP, /* ZAP */ DMU_OT_DSL_DS_SNAP_MAP, /* ZAP */ DMU_OT_DSL_PROPS, /* ZAP */ DMU_OT_DSL_DATASET, /* UINT64 */ /* zpl: */ DMU_OT_ZNODE, /* ZNODE */ DMU_OT_OLDACL, /* Old ACL */ DMU_OT_PLAIN_FILE_CONTENTS, /* UINT8 */ DMU_OT_DIRECTORY_CONTENTS, /* ZAP */ DMU_OT_MASTER_NODE, /* ZAP */ DMU_OT_UNLINKED_SET, /* ZAP */ /* zvol: */ DMU_OT_ZVOL, /* UINT8 */ DMU_OT_ZVOL_PROP, /* ZAP */ /* other; for testing only! */ DMU_OT_PLAIN_OTHER, /* UINT8 */ DMU_OT_UINT64_OTHER, /* UINT64 */ DMU_OT_ZAP_OTHER, /* ZAP */ /* new object types: */ DMU_OT_ERROR_LOG, /* ZAP */ DMU_OT_SPA_HISTORY, /* UINT8 */ DMU_OT_SPA_HISTORY_OFFSETS, /* spa_his_phys_t */ DMU_OT_POOL_PROPS, /* ZAP */ DMU_OT_DSL_PERMS, /* ZAP */ DMU_OT_ACL, /* ACL */ DMU_OT_SYSACL, /* SYSACL */ DMU_OT_FUID, /* FUID table (Packed NVLIST UINT8) */ DMU_OT_FUID_SIZE, /* FUID table size UINT64 */ DMU_OT_NEXT_CLONES, /* ZAP */ DMU_OT_SCAN_QUEUE, /* ZAP */ DMU_OT_USERGROUP_USED, /* ZAP */ DMU_OT_USERGROUP_QUOTA, /* ZAP */ DMU_OT_USERREFS, /* ZAP */ DMU_OT_DDT_ZAP, /* ZAP */ DMU_OT_DDT_STATS, /* ZAP */ DMU_OT_SA, /* System attr */ DMU_OT_SA_MASTER_NODE, /* ZAP */ DMU_OT_SA_ATTR_REGISTRATION, /* ZAP */ DMU_OT_SA_ATTR_LAYOUTS, /* ZAP */ DMU_OT_SCAN_XLATE, /* ZAP */ DMU_OT_DEDUP, /* fake dedup BP from ddt_bp_create() */ DMU_OT_DEADLIST, /* ZAP */ DMU_OT_DEADLIST_HDR, /* UINT64 */ DMU_OT_DSL_CLONES, /* ZAP */ DMU_OT_BPOBJ_SUBOBJ, /* UINT64 */ /* * Do not allocate new object types here. Doing so makes the on-disk * format incompatible with any other format that uses the same object * type number. * * When creating an object which does not have one of the above types * use the DMU_OTN_* type with the correct byteswap and metadata * values. * * The DMU_OTN_* types do not have entries in the dmu_ot table, * use the DMU_OT_IS_METDATA() and DMU_OT_BYTESWAP() macros instead * of indexing into dmu_ot directly (this works for both DMU_OT_* types * and DMU_OTN_* types). */ DMU_OT_NUMTYPES, /* * Names for valid types declared with DMU_OT(). */ DMU_OTN_UINT8_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_FALSE), DMU_OTN_UINT8_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_FALSE), DMU_OTN_UINT16_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_FALSE), DMU_OTN_UINT16_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_FALSE), DMU_OTN_UINT32_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_FALSE), DMU_OTN_UINT32_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_FALSE), DMU_OTN_UINT64_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_FALSE), DMU_OTN_UINT64_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_FALSE), DMU_OTN_ZAP_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_FALSE), DMU_OTN_ZAP_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_FALSE), DMU_OTN_UINT8_ENC_DATA = DMU_OT(DMU_BSWAP_UINT8, B_FALSE, B_TRUE), DMU_OTN_UINT8_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT8, B_TRUE, B_TRUE), DMU_OTN_UINT16_ENC_DATA = DMU_OT(DMU_BSWAP_UINT16, B_FALSE, B_TRUE), DMU_OTN_UINT16_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT16, B_TRUE, B_TRUE), DMU_OTN_UINT32_ENC_DATA = DMU_OT(DMU_BSWAP_UINT32, B_FALSE, B_TRUE), DMU_OTN_UINT32_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT32, B_TRUE, B_TRUE), DMU_OTN_UINT64_ENC_DATA = DMU_OT(DMU_BSWAP_UINT64, B_FALSE, B_TRUE), DMU_OTN_UINT64_ENC_METADATA = DMU_OT(DMU_BSWAP_UINT64, B_TRUE, B_TRUE), DMU_OTN_ZAP_ENC_DATA = DMU_OT(DMU_BSWAP_ZAP, B_FALSE, B_TRUE), DMU_OTN_ZAP_ENC_METADATA = DMU_OT(DMU_BSWAP_ZAP, B_TRUE, B_TRUE), } dmu_object_type_t; /* * These flags are intended to be used to specify the "txg_how" * parameter when calling the dmu_tx_assign() function. See the comment * above dmu_tx_assign() for more details on the meaning of these flags. */ #define TXG_NOWAIT (0ULL) #define TXG_WAIT (1ULL<<0) #define TXG_NOTHROTTLE (1ULL<<1) void byteswap_uint64_array(void *buf, size_t size); void byteswap_uint32_array(void *buf, size_t size); void byteswap_uint16_array(void *buf, size_t size); void byteswap_uint8_array(void *buf, size_t size); void zap_byteswap(void *buf, size_t size); void zfs_oldacl_byteswap(void *buf, size_t size); void zfs_acl_byteswap(void *buf, size_t size); void zfs_znode_byteswap(void *buf, size_t size); #define DS_FIND_SNAPSHOTS (1<<0) #define DS_FIND_CHILDREN (1<<1) #define DS_FIND_SERIALIZE (1<<2) /* * The maximum number of bytes that can be accessed as part of one * operation, including metadata. */ #define DMU_MAX_ACCESS (64 * 1024 * 1024) /* 64MB */ #define DMU_MAX_DELETEBLKCNT (20480) /* ~5MB of indirect blocks */ #define DMU_USERUSED_OBJECT (-1ULL) #define DMU_GROUPUSED_OBJECT (-2ULL) #define DMU_PROJECTUSED_OBJECT (-3ULL) /* * Zap prefix for object accounting in DMU_{USER,GROUP,PROJECT}USED_OBJECT. */ #define DMU_OBJACCT_PREFIX "obj-" #define DMU_OBJACCT_PREFIX_LEN 4 /* * artificial blkids for bonus buffer and spill blocks */ #define DMU_BONUS_BLKID (-1ULL) #define DMU_SPILL_BLKID (-2ULL) /* * Public routines to create, destroy, open, and close objsets. */ typedef void dmu_objset_create_sync_func_t(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); int dmu_objset_hold(const char *name, void *tag, objset_t **osp); int dmu_objset_own(const char *name, dmu_objset_type_t type, boolean_t readonly, boolean_t key_required, void *tag, objset_t **osp); void dmu_objset_rele(objset_t *os, void *tag); void dmu_objset_disown(objset_t *os, boolean_t key_required, void *tag); int dmu_objset_open_ds(struct dsl_dataset *ds, objset_t **osp); void dmu_objset_evict_dbufs(objset_t *os); int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, struct dsl_crypto_params *dcp, dmu_objset_create_sync_func_t func, void *arg); int dmu_objset_clone(const char *name, const char *origin); int dsl_destroy_snapshots_nvl(struct nvlist *snaps, boolean_t defer, struct nvlist *errlist); int dmu_objset_snapshot_one(const char *fsname, const char *snapname); int dmu_objset_snapshot_tmp(const char *, const char *, int); int dmu_objset_find(char *name, int func(const char *, void *), void *arg, int flags); void dmu_objset_byteswap(void *buf, size_t size); int dsl_dataset_rename_snapshot(const char *fsname, const char *oldsnapname, const char *newsnapname, boolean_t recursive); int dmu_objset_remap_indirects(const char *fsname); typedef struct dmu_buf { uint64_t db_object; /* object that this buffer is part of */ uint64_t db_offset; /* byte offset in this object */ uint64_t db_size; /* size of buffer in bytes */ void *db_data; /* data in buffer */ } dmu_buf_t; /* * The names of zap entries in the DIRECTORY_OBJECT of the MOS. */ #define DMU_POOL_DIRECTORY_OBJECT 1 #define DMU_POOL_CONFIG "config" #define DMU_POOL_FEATURES_FOR_WRITE "features_for_write" #define DMU_POOL_FEATURES_FOR_READ "features_for_read" #define DMU_POOL_FEATURE_DESCRIPTIONS "feature_descriptions" #define DMU_POOL_FEATURE_ENABLED_TXG "feature_enabled_txg" #define DMU_POOL_ROOT_DATASET "root_dataset" #define DMU_POOL_SYNC_BPOBJ "sync_bplist" #define DMU_POOL_ERRLOG_SCRUB "errlog_scrub" #define DMU_POOL_ERRLOG_LAST "errlog_last" #define DMU_POOL_SPARES "spares" #define DMU_POOL_DEFLATE "deflate" #define DMU_POOL_HISTORY "history" #define DMU_POOL_PROPS "pool_props" #define DMU_POOL_L2CACHE "l2cache" #define DMU_POOL_TMP_USERREFS "tmp_userrefs" #define DMU_POOL_DDT "DDT-%s-%s-%s" #define DMU_POOL_DDT_STATS "DDT-statistics" #define DMU_POOL_CREATION_VERSION "creation_version" #define DMU_POOL_SCAN "scan" #define DMU_POOL_FREE_BPOBJ "free_bpobj" #define DMU_POOL_BPTREE_OBJ "bptree_obj" #define DMU_POOL_EMPTY_BPOBJ "empty_bpobj" #define DMU_POOL_CHECKSUM_SALT "org.illumos:checksum_salt" #define DMU_POOL_VDEV_ZAP_MAP "com.delphix:vdev_zap_map" #define DMU_POOL_REMOVING "com.delphix:removing" #define DMU_POOL_OBSOLETE_BPOBJ "com.delphix:obsolete_bpobj" #define DMU_POOL_CONDENSING_INDIRECT "com.delphix:condensing_indirect" /* * Allocate an object from this objset. The range of object numbers * available is (0, DN_MAX_OBJECT). Object 0 is the meta-dnode. * * The transaction must be assigned to a txg. The newly allocated * object will be "held" in the transaction (ie. you can modify the * newly allocated object in this transaction). * * dmu_object_alloc() chooses an object and returns it in *objectp. * * dmu_object_claim() allocates a specific object number. If that * number is already allocated, it fails and returns EEXIST. * * Return 0 on success, or ENOSPC or EEXIST as specified above. */ uint64_t dmu_object_alloc(objset_t *os, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); uint64_t dmu_object_alloc_dnsize(objset_t *os, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonus_type, int bonus_len, int dnodesize, dmu_tx_t *tx); int dmu_object_claim(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonus_type, int bonus_len, dmu_tx_t *tx); int dmu_object_claim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonus_type, int bonus_len, int dnodesize, dmu_tx_t *tx); int dmu_object_reclaim(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, dmu_tx_t *txp); int dmu_object_reclaim_dnsize(objset_t *os, uint64_t object, dmu_object_type_t ot, int blocksize, dmu_object_type_t bonustype, int bonuslen, int dnodesize, dmu_tx_t *txp); /* * Free an object from this objset. * * The object's data will be freed as well (ie. you don't need to call * dmu_free(object, 0, -1, tx)). * * The object need not be held in the transaction. * * If there are any holds on this object's buffers (via dmu_buf_hold()), * or tx holds on the object (via dmu_tx_hold_object()), you can not * free it; it fails and returns EBUSY. * * If the object is not allocated, it fails and returns ENOENT. * * Return 0 on success, or EBUSY or ENOENT as specified above. */ int dmu_object_free(objset_t *os, uint64_t object, dmu_tx_t *tx); /* * Find the next allocated or free object. * * The objectp parameter is in-out. It will be updated to be the next * object which is allocated. Ignore objects which have not been * modified since txg. * * XXX Can only be called on a objset with no dirty data. * * Returns 0 on success, or ENOENT if there are no more objects. */ int dmu_object_next(objset_t *os, uint64_t *objectp, boolean_t hole, uint64_t txg); /* * Set the number of levels on a dnode. nlevels must be greater than the * current number of levels or an EINVAL will be returned. */ int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels, dmu_tx_t *tx); /* * Set the data blocksize for an object. * * The object cannot have any blocks allcated beyond the first. If * the first block is allocated already, the new size must be greater * than the current block size. If these conditions are not met, * ENOTSUP will be returned. * * Returns 0 on success, or EBUSY if there are any holds on the object * contents, or ENOTSUP as described above. */ int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs, dmu_tx_t *tx); /* * Manually set the maxblkid on a dnode. This will adjust nlevels accordingly * to accommodate the change. */ int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid, dmu_tx_t *tx); /* * Set the checksum property on a dnode. The new checksum algorithm will * apply to all newly written blocks; existing blocks will not be affected. */ void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, dmu_tx_t *tx); /* * Set the compress property on a dnode. The new compression algorithm will * apply to all newly written blocks; existing blocks will not be affected. */ void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, dmu_tx_t *tx); -int dmu_object_dirty_raw(objset_t *os, uint64_t object, dmu_tx_t *tx); int dmu_object_remap_indirects(objset_t *os, uint64_t object, uint64_t txg); void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, void *data, uint8_t etype, uint8_t comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx); /* * Decide how to write a block: checksum, compression, number of copies, etc. */ #define WP_NOFILL 0x1 #define WP_DMU_SYNC 0x2 #define WP_SPILL 0x4 void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, struct zio_prop *zp); /* * The bonus data is accessed more or less like a regular buffer. * You must dmu_bonus_hold() to get the buffer, which will give you a * dmu_buf_t with db_offset==-1ULL, and db_size = the size of the bonus * data. As with any normal buffer, you must call dmu_buf_will_dirty() * before modifying it, and the * object must be held in an assigned transaction before calling * dmu_buf_will_dirty. You may use dmu_buf_set_user() on the bonus * buffer as well. You must release what you hold with dmu_buf_rele(). * * Returns ENOENT, EIO, or 0. */ int dmu_bonus_hold_impl(objset_t *os, uint64_t object, void *tag, uint32_t flags, dmu_buf_t **dbp); int dmu_bonus_hold(objset_t *os, uint64_t object, void *tag, dmu_buf_t **); int dmu_bonus_max(void); int dmu_set_bonus(dmu_buf_t *, int, dmu_tx_t *); int dmu_set_bonustype(dmu_buf_t *, dmu_object_type_t, dmu_tx_t *); dmu_object_type_t dmu_get_bonustype(dmu_buf_t *); int dmu_rm_spill(objset_t *, uint64_t, dmu_tx_t *); /* * Special spill buffer support used by "SA" framework */ int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, void *tag, dmu_buf_t **dbp); int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp); /* * Obtain the DMU buffer from the specified object which contains the * specified offset. dmu_buf_hold() puts a "hold" on the buffer, so * that it will remain in memory. You must release the hold with * dmu_buf_rele(). You must not access the dmu_buf_t after releasing * what you hold. You must have a hold on any dmu_buf_t* you pass to the DMU. * * You must call dmu_buf_read, dmu_buf_will_dirty, or dmu_buf_will_fill * on the returned buffer before reading or writing the buffer's * db_data. The comments for those routines describe what particular * operations are valid after calling them. * * The object number must be a valid, allocated object number. */ int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, void *tag, dmu_buf_t **, int flags); int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, void *tag, dmu_buf_t **dbp, int flags); /* * Add a reference to a dmu buffer that has already been held via * dmu_buf_hold() in the current context. */ void dmu_buf_add_ref(dmu_buf_t *db, void* tag); /* * Attempt to add a reference to a dmu buffer that is in an unknown state, * using a pointer that may have been invalidated by eviction processing. * The request will succeed if the passed in dbuf still represents the * same os/object/blkid, is ineligible for eviction, and has at least * one hold by a user other than the syncer. */ boolean_t dmu_buf_try_add_ref(dmu_buf_t *, objset_t *os, uint64_t object, uint64_t blkid, void *tag); void dmu_buf_rele(dmu_buf_t *db, void *tag); uint64_t dmu_buf_refcount(dmu_buf_t *db); /* * dmu_buf_hold_array holds the DMU buffers which contain all bytes in a * range of an object. A pointer to an array of dmu_buf_t*'s is * returned (in *dbpp). * * dmu_buf_rele_array releases the hold on an array of dmu_buf_t*'s, and * frees the array. The hold on the array of buffers MUST be released * with dmu_buf_rele_array. You can NOT release the hold on each buffer * individually with dmu_buf_rele. */ int dmu_buf_hold_array_by_bonus(dmu_buf_t *db, uint64_t offset, uint64_t length, boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp); void dmu_buf_rele_array(dmu_buf_t **, int numbufs, void *tag); typedef void dmu_buf_evict_func_t(void *user_ptr); /* * A DMU buffer user object may be associated with a dbuf for the * duration of its lifetime. This allows the user of a dbuf (client) * to attach private data to a dbuf (e.g. in-core only data such as a * dnode_children_t, zap_t, or zap_leaf_t) and be optionally notified * when that dbuf has been evicted. Clients typically respond to the * eviction notification by freeing their private data, thus ensuring * the same lifetime for both dbuf and private data. * * The mapping from a dmu_buf_user_t to any client private data is the * client's responsibility. All current consumers of the API with private * data embed a dmu_buf_user_t as the first member of the structure for * their private data. This allows conversions between the two types * with a simple cast. Since the DMU buf user API never needs access * to the private data, other strategies can be employed if necessary * or convenient for the client (e.g. using container_of() to do the * conversion for private data that cannot have the dmu_buf_user_t as * its first member). * * Eviction callbacks are executed without the dbuf mutex held or any * other type of mechanism to guarantee that the dbuf is still available. * For this reason, users must assume the dbuf has already been freed * and not reference the dbuf from the callback context. * * Users requesting "immediate eviction" are notified as soon as the dbuf * is only referenced by dirty records (dirties == holds). Otherwise the * notification occurs after eviction processing for the dbuf begins. */ typedef struct dmu_buf_user { /* * Asynchronous user eviction callback state. */ taskq_ent_t dbu_tqent; /* * This instance's eviction function pointers. * * dbu_evict_func_sync is called synchronously and then * dbu_evict_func_async is executed asynchronously on a taskq. */ dmu_buf_evict_func_t *dbu_evict_func_sync; dmu_buf_evict_func_t *dbu_evict_func_async; #ifdef ZFS_DEBUG /* * Pointer to user's dbuf pointer. NULL for clients that do * not associate a dbuf with their user data. * * The dbuf pointer is cleared upon eviction so as to catch * use-after-evict bugs in clients. */ dmu_buf_t **dbu_clear_on_evict_dbufp; #endif } dmu_buf_user_t; /* * Initialize the given dmu_buf_user_t instance with the eviction function * evict_func, to be called when the user is evicted. * * NOTE: This function should only be called once on a given dmu_buf_user_t. * To allow enforcement of this, dbu must already be zeroed on entry. */ /*ARGSUSED*/ static inline void dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp) { ASSERT(dbu->dbu_evict_func_sync == NULL); ASSERT(dbu->dbu_evict_func_async == NULL); /* must have at least one evict func */ IMPLY(evict_func_sync == NULL, evict_func_async != NULL); dbu->dbu_evict_func_sync = evict_func_sync; dbu->dbu_evict_func_async = evict_func_async; taskq_init_ent(&dbu->dbu_tqent); #ifdef ZFS_DEBUG dbu->dbu_clear_on_evict_dbufp = clear_on_evict_dbufp; #endif } /* * Attach user data to a dbuf and mark it for normal (when the dbuf's * data is cleared or its reference count goes to zero) eviction processing. * * Returns NULL on success, or the existing user if another user currently * owns the buffer. */ void *dmu_buf_set_user(dmu_buf_t *db, dmu_buf_user_t *user); /* * Attach user data to a dbuf and mark it for immediate (its dirty and * reference counts are equal) eviction processing. * * Returns NULL on success, or the existing user if another user currently * owns the buffer. */ void *dmu_buf_set_user_ie(dmu_buf_t *db, dmu_buf_user_t *user); /* * Replace the current user of a dbuf. * * If given the current user of a dbuf, replaces the dbuf's user with * "new_user" and returns the user data pointer that was replaced. * Otherwise returns the current, and unmodified, dbuf user pointer. */ void *dmu_buf_replace_user(dmu_buf_t *db, dmu_buf_user_t *old_user, dmu_buf_user_t *new_user); /* * Remove the specified user data for a DMU buffer. * * Returns the user that was removed on success, or the current user if * another user currently owns the buffer. */ void *dmu_buf_remove_user(dmu_buf_t *db, dmu_buf_user_t *user); /* * Returns the user data (dmu_buf_user_t *) associated with this dbuf. */ void *dmu_buf_get_user(dmu_buf_t *db); objset_t *dmu_buf_get_objset(dmu_buf_t *db); dnode_t *dmu_buf_dnode_enter(dmu_buf_t *db); void dmu_buf_dnode_exit(dmu_buf_t *db); /* Block until any in-progress dmu buf user evictions complete. */ void dmu_buf_user_evict_wait(void); /* * Returns the blkptr associated with this dbuf, or NULL if not set. */ struct blkptr *dmu_buf_get_blkptr(dmu_buf_t *db); /* * Indicate that you are going to modify the buffer's data (db_data). * * The transaction (tx) must be assigned to a txg (ie. you've called * dmu_tx_assign()). The buffer's object must be held in the tx * (ie. you've called dmu_tx_hold_object(tx, db->db_object)). */ void dmu_buf_will_dirty(dmu_buf_t *db, dmu_tx_t *tx); -void dmu_buf_will_change_crypt_params(dmu_buf_t *db, dmu_tx_t *tx); +void dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder, + const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx); /* * You must create a transaction, then hold the objects which you will * (or might) modify as part of this transaction. Then you must assign * the transaction to a transaction group. Once the transaction has * been assigned, you can modify buffers which belong to held objects as * part of this transaction. You can't modify buffers before the * transaction has been assigned; you can't modify buffers which don't * belong to objects which this transaction holds; you can't hold * objects once the transaction has been assigned. You may hold an * object which you are going to free (with dmu_object_free()), but you * don't have to. * * You can abort the transaction before it has been assigned. * * Note that you may hold buffers (with dmu_buf_hold) at any time, * regardless of transaction state. */ #define DMU_NEW_OBJECT (-1ULL) #define DMU_OBJECT_END (-1ULL) dmu_tx_t *dmu_tx_create(objset_t *os); void dmu_tx_hold_write(dmu_tx_t *tx, uint64_t object, uint64_t off, int len); void dmu_tx_hold_write_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, int len); void dmu_tx_hold_free(dmu_tx_t *tx, uint64_t object, uint64_t off, uint64_t len); void dmu_tx_hold_free_by_dnode(dmu_tx_t *tx, dnode_t *dn, uint64_t off, uint64_t len); void dmu_tx_hold_remap_l1indirect(dmu_tx_t *tx, uint64_t object); void dmu_tx_hold_zap(dmu_tx_t *tx, uint64_t object, int add, const char *name); void dmu_tx_hold_zap_by_dnode(dmu_tx_t *tx, dnode_t *dn, int add, const char *name); void dmu_tx_hold_bonus(dmu_tx_t *tx, uint64_t object); void dmu_tx_hold_bonus_by_dnode(dmu_tx_t *tx, dnode_t *dn); void dmu_tx_hold_spill(dmu_tx_t *tx, uint64_t object); void dmu_tx_hold_sa(dmu_tx_t *tx, struct sa_handle *hdl, boolean_t may_grow); void dmu_tx_hold_sa_create(dmu_tx_t *tx, int total_size); void dmu_tx_abort(dmu_tx_t *tx); int dmu_tx_assign(dmu_tx_t *tx, uint64_t txg_how); void dmu_tx_wait(dmu_tx_t *tx); void dmu_tx_commit(dmu_tx_t *tx); void dmu_tx_mark_netfree(dmu_tx_t *tx); /* * To register a commit callback, dmu_tx_callback_register() must be called. * * dcb_data is a pointer to caller private data that is passed on as a * callback parameter. The caller is responsible for properly allocating and * freeing it. * * When registering a callback, the transaction must be already created, but * it cannot be committed or aborted. It can be assigned to a txg or not. * * The callback will be called after the transaction has been safely written * to stable storage and will also be called if the dmu_tx is aborted. * If there is any error which prevents the transaction from being committed to * disk, the callback will be called with a value of error != 0. * * When multiple callbacks are registered to the transaction, the callbacks * will be called in reverse order to let Lustre, the only user of commit * callback currently, take the fast path of its commit callback handling. */ typedef void dmu_tx_callback_func_t(void *dcb_data, int error); void dmu_tx_callback_register(dmu_tx_t *tx, dmu_tx_callback_func_t *dcb_func, void *dcb_data); void dmu_tx_do_callbacks(list_t *cb_list, int error); /* * Free up the data blocks for a defined range of a file. If size is * -1, the range from offset to end-of-file is freed. */ int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx); int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t size); -int dmu_free_long_range_raw(objset_t *os, uint64_t object, uint64_t offset, - uint64_t size); int dmu_free_long_object(objset_t *os, uint64_t object); -int dmu_free_long_object_raw(objset_t *os, uint64_t object); /* * Convenience functions. * * Canfail routines will return 0 on success, or an errno if there is a * nonrecoverable I/O error. */ #define DMU_READ_PREFETCH 0 /* prefetch */ #define DMU_READ_NO_PREFETCH 1 /* don't prefetch */ #define DMU_READ_NO_DECRYPT 2 /* don't decrypt */ int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, void *buf, uint32_t flags); int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, uint32_t flags); void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx); void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx); void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx); #ifdef _KERNEL #include int dmu_read_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size); int dmu_read_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size); int dmu_read_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size); int dmu_write_uio(objset_t *os, uint64_t object, struct uio *uio, uint64_t size, dmu_tx_t *tx); int dmu_write_uio_dbuf(dmu_buf_t *zdb, struct uio *uio, uint64_t size, dmu_tx_t *tx); int dmu_write_uio_dnode(dnode_t *dn, struct uio *uio, uint64_t size, dmu_tx_t *tx); #endif struct arc_buf *dmu_request_arcbuf(dmu_buf_t *handle, int size); void dmu_return_arcbuf(struct arc_buf *buf); void dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset, struct arc_buf *buf, dmu_tx_t *tx); void dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset, struct arc_buf *buf, dmu_tx_t *tx); #define dmu_assign_arcbuf dmu_assign_arcbuf_by_dbuf -int dmu_convert_mdn_block_to_raw(objset_t *os, uint64_t firstobj, - boolean_t byteorder, const uint8_t *salt, const uint8_t *iv, - const uint8_t *mac, dmu_tx_t *tx); void dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset, dmu_buf_t *handle, dmu_tx_t *tx); #ifdef HAVE_UIO_ZEROCOPY int dmu_xuio_init(struct xuio *uio, int niov); void dmu_xuio_fini(struct xuio *uio); int dmu_xuio_add(struct xuio *uio, struct arc_buf *abuf, offset_t off, size_t n); int dmu_xuio_cnt(struct xuio *uio); struct arc_buf *dmu_xuio_arcbuf(struct xuio *uio, int i); void dmu_xuio_clear(struct xuio *uio, int i); #endif /* HAVE_UIO_ZEROCOPY */ void xuio_stat_wbuf_copied(void); void xuio_stat_wbuf_nocopy(void); extern int zfs_prefetch_disable; extern int zfs_max_recordsize; /* * Asynchronously try to read in the data. */ void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, uint64_t len, enum zio_priority pri); typedef struct dmu_object_info { /* All sizes are in bytes unless otherwise indicated. */ uint32_t doi_data_block_size; uint32_t doi_metadata_block_size; dmu_object_type_t doi_type; dmu_object_type_t doi_bonus_type; uint64_t doi_bonus_size; uint8_t doi_indirection; /* 2 = dnode->indirect->data */ uint8_t doi_checksum; uint8_t doi_compress; uint8_t doi_nblkptr; uint8_t doi_pad[4]; uint64_t doi_dnodesize; uint64_t doi_physical_blocks_512; /* data + metadata, 512b blks */ uint64_t doi_max_offset; uint64_t doi_fill_count; /* number of non-empty blocks */ } dmu_object_info_t; typedef void (*const arc_byteswap_func_t)(void *buf, size_t size); typedef struct dmu_object_type_info { dmu_object_byteswap_t ot_byteswap; boolean_t ot_metadata; boolean_t ot_encrypt; char *ot_name; } dmu_object_type_info_t; typedef const struct dmu_object_byteswap_info { arc_byteswap_func_t ob_func; char *ob_name; } dmu_object_byteswap_info_t; extern const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES]; extern const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS]; /* * Get information on a DMU object. * * Return 0 on success or ENOENT if object is not allocated. * * If doi is NULL, just indicates whether the object exists. */ int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi); void __dmu_object_info_from_dnode(struct dnode *dn, dmu_object_info_t *doi); /* Like dmu_object_info, but faster if you have a held dnode in hand. */ void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi); /* Like dmu_object_info, but faster if you have a held dbuf in hand. */ void dmu_object_info_from_db(dmu_buf_t *db, dmu_object_info_t *doi); /* * Like dmu_object_info_from_db, but faster still when you only care about * the size. This is specifically optimized for zfs_getattr(). */ void dmu_object_size_from_db(dmu_buf_t *db, uint32_t *blksize, u_longlong_t *nblk512); void dmu_object_dnsize_from_db(dmu_buf_t *db, int *dnsize); typedef struct dmu_objset_stats { uint64_t dds_num_clones; /* number of clones of this */ uint64_t dds_creation_txg; uint64_t dds_guid; dmu_objset_type_t dds_type; uint8_t dds_is_snapshot; uint8_t dds_inconsistent; char dds_origin[ZFS_MAX_DATASET_NAME_LEN]; } dmu_objset_stats_t; /* * Get stats on a dataset. */ void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); /* * Add entries to the nvlist for all the objset's properties. See * zfs_prop_table[] and zfs(1m) for details on the properties. */ void dmu_objset_stats(objset_t *os, struct nvlist *nv); /* * Get the space usage statistics for statvfs(). * * refdbytes is the amount of space "referenced" by this objset. * availbytes is the amount of space available to this objset, taking * into account quotas & reservations, assuming that no other objsets * use the space first. These values correspond to the 'referenced' and * 'available' properties, described in the zfs(1m) manpage. * * usedobjs and availobjs are the number of objects currently allocated, * and available. */ void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, uint64_t *usedobjsp, uint64_t *availobjsp); /* * The fsid_guid is a 56-bit ID that can change to avoid collisions. * (Contrast with the ds_guid which is a 64-bit ID that will never * change, so there is a small probability that it will collide.) */ uint64_t dmu_objset_fsid_guid(objset_t *os); /* * Get the [cm]time for an objset's snapshot dir */ timestruc_t dmu_objset_snap_cmtime(objset_t *os); int dmu_objset_is_snapshot(objset_t *os); extern struct spa *dmu_objset_spa(objset_t *os); extern struct zilog *dmu_objset_zil(objset_t *os); extern struct dsl_pool *dmu_objset_pool(objset_t *os); extern struct dsl_dataset *dmu_objset_ds(objset_t *os); extern void dmu_objset_name(objset_t *os, char *buf); extern dmu_objset_type_t dmu_objset_type(objset_t *os); extern uint64_t dmu_objset_id(objset_t *os); extern uint64_t dmu_objset_dnodesize(objset_t *os); extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os); extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os); extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, uint64_t *id, uint64_t *offp, boolean_t *case_conflict); extern int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *val); extern int dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, boolean_t *conflict); extern int dmu_dir_list_next(objset_t *os, int namelen, char *name, uint64_t *idp, uint64_t *offp); typedef int objset_used_cb_t(dmu_object_type_t bonustype, void *bonus, uint64_t *userp, uint64_t *groupp, uint64_t *projectp); extern void dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb); extern void dmu_objset_set_user(objset_t *os, void *user_ptr); extern void *dmu_objset_get_user(objset_t *os); /* * Return the txg number for the given assigned transaction. */ uint64_t dmu_tx_get_txg(dmu_tx_t *tx); /* * Synchronous write. * If a parent zio is provided this function initiates a write on the * provided buffer as a child of the parent zio. * In the absence of a parent zio, the write is completed synchronously. * At write completion, blk is filled with the bp of the written block. * Note that while the data covered by this function will be on stable * storage when the write completes this new data does not become a * permanent part of the file until the associated transaction commits. */ /* * {zfs,zvol,ztest}_get_done() args */ typedef struct zgd { struct lwb *zgd_lwb; struct blkptr *zgd_bp; dmu_buf_t *zgd_db; struct rl *zgd_rl; void *zgd_private; } zgd_t; typedef void dmu_sync_cb_t(zgd_t *arg, int error); int dmu_sync(struct zio *zio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd); /* * Find the next hole or data block in file starting at *off * Return found offset in *off. Return ESRCH for end of file. */ int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off); /* * Initial setup and final teardown. */ extern void dmu_init(void); extern void dmu_fini(void); typedef void (*dmu_traverse_cb_t)(objset_t *os, void *arg, struct blkptr *bp, uint64_t object, uint64_t offset, int len); void dmu_traverse_objset(objset_t *os, uint64_t txg_start, dmu_traverse_cb_t cb, void *arg); int dmu_diff(const char *tosnap_name, const char *fromsnap_name, struct vnode *vp, offset_t *offp); /* CRC64 table */ #define ZFS_CRC64_POLY 0xC96C5795D7870F42ULL /* ECMA-182, reflected form */ extern uint64_t zfs_crc64_table[256]; #ifdef __cplusplus } #endif #endif /* _SYS_DMU_H */ diff --git a/include/sys/dmu_objset.h b/include/sys/dmu_objset.h index 54fe37e61d84..57872d8db8d3 100644 --- a/include/sys/dmu_objset.h +++ b/include/sys/dmu_objset.h @@ -1,252 +1,253 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2017 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. */ /* Portions Copyright 2010 Robert Milkowski */ #ifndef _SYS_DMU_OBJSET_H #define _SYS_DMU_OBJSET_H #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif extern krwlock_t os_lock; struct dsl_pool; struct dsl_dataset; struct dmu_tx; #define OBJSET_PHYS_SIZE_V1 1024 #define OBJSET_PHYS_SIZE_V2 2048 #define OBJSET_PHYS_SIZE_V3 4096 #define OBJSET_BUF_HAS_USERUSED(buf) \ (arc_buf_size(buf) >= OBJSET_PHYS_SIZE_V2) #define OBJSET_BUF_HAS_PROJECTUSED(buf) \ (arc_buf_size(buf) >= OBJSET_PHYS_SIZE_V3) #define OBJSET_FLAG_USERACCOUNTING_COMPLETE (1ULL << 0) #define OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE (1ULL << 1) #define OBJSET_FLAG_PROJECTQUOTA_COMPLETE (1ULL << 2) /* * This mask defines the set of flags which are "portable", meaning * that they can be preserved when doing a raw encrypted zfs send. * Flags included in this mask will be protected by os_portable_mac * when the block of dnodes is encrypted. No portable flags currently * exist. */ #define OBJSET_CRYPT_PORTABLE_FLAGS_MASK (0) typedef struct objset_phys { dnode_phys_t os_meta_dnode; zil_header_t os_zil_header; uint64_t os_type; uint64_t os_flags; uint8_t os_portable_mac[ZIO_OBJSET_MAC_LEN]; uint8_t os_local_mac[ZIO_OBJSET_MAC_LEN]; char os_pad0[OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)*3 - sizeof (zil_header_t) - sizeof (uint64_t)*2 - 2*ZIO_OBJSET_MAC_LEN]; dnode_phys_t os_userused_dnode; dnode_phys_t os_groupused_dnode; dnode_phys_t os_projectused_dnode; char os_pad1[OBJSET_PHYS_SIZE_V3 - OBJSET_PHYS_SIZE_V2 - sizeof (dnode_phys_t)]; } objset_phys_t; typedef int (*dmu_objset_upgrade_cb_t)(objset_t *); struct objset { /* Immutable: */ struct dsl_dataset *os_dsl_dataset; spa_t *os_spa; arc_buf_t *os_phys_buf; objset_phys_t *os_phys; boolean_t os_encrypted; /* * The following "special" dnodes have no parent, are exempt * from dnode_move(), and are not recorded in os_dnodes, but they * root their descendents in this objset using handles anyway, so * that all access to dnodes from dbufs consistently uses handles. */ dnode_handle_t os_meta_dnode; dnode_handle_t os_userused_dnode; dnode_handle_t os_groupused_dnode; dnode_handle_t os_projectused_dnode; zilog_t *os_zil; list_node_t os_evicting_node; /* can change, under dsl_dir's locks: */ uint64_t os_dnodesize; /* default dnode size for new objects */ enum zio_checksum os_checksum; enum zio_compress os_compress; uint8_t os_copies; enum zio_checksum os_dedup_checksum; boolean_t os_dedup_verify; zfs_logbias_op_t os_logbias; zfs_cache_type_t os_primary_cache; zfs_cache_type_t os_secondary_cache; zfs_sync_type_t os_sync; zfs_redundant_metadata_type_t os_redundant_metadata; int os_recordsize; /* * Pointer is constant; the blkptr it points to is protected by * os_dsl_dataset->ds_bp_rwlock */ blkptr_t *os_rootbp; /* no lock needed: */ struct dmu_tx *os_synctx; /* XXX sketchy */ zil_header_t os_zil_header; multilist_t *os_synced_dnodes; uint64_t os_flags; uint64_t os_freed_dnodes; boolean_t os_rescan_dnodes; + boolean_t os_raw_receive; /* os_phys_buf should be written raw next txg */ boolean_t os_next_write_raw[TXG_SIZE]; /* Protected by os_obj_lock */ kmutex_t os_obj_lock; uint64_t os_obj_next_chunk; /* Per-CPU next object to allocate, protected by atomic ops. */ uint64_t *os_obj_next_percpu; int os_obj_next_percpu_len; /* Protected by os_lock */ kmutex_t os_lock; multilist_t *os_dirty_dnodes[TXG_SIZE]; list_t os_dnodes; list_t os_downgraded_dbufs; /* Protects changes to DMU_{USER,GROUP,PROJECT}USED_OBJECT */ kmutex_t os_userused_lock; /* stuff we store for the user */ kmutex_t os_user_ptr_lock; void *os_user_ptr; sa_os_t *os_sa; /* kernel thread to upgrade this dataset */ kmutex_t os_upgrade_lock; taskqid_t os_upgrade_id; dmu_objset_upgrade_cb_t os_upgrade_cb; boolean_t os_upgrade_exit; int os_upgrade_status; }; #define DMU_META_OBJSET 0 #define DMU_META_DNODE_OBJECT 0 #define DMU_OBJECT_IS_SPECIAL(obj) ((int64_t)(obj) <= 0) #define DMU_META_DNODE(os) ((os)->os_meta_dnode.dnh_dnode) #define DMU_USERUSED_DNODE(os) ((os)->os_userused_dnode.dnh_dnode) #define DMU_GROUPUSED_DNODE(os) ((os)->os_groupused_dnode.dnh_dnode) #define DMU_PROJECTUSED_DNODE(os) ((os)->os_projectused_dnode.dnh_dnode) #define DMU_OS_IS_L2CACHEABLE(os) \ ((os)->os_secondary_cache == ZFS_CACHE_ALL || \ (os)->os_secondary_cache == ZFS_CACHE_METADATA) /* called from zpl */ int dmu_objset_hold(const char *name, void *tag, objset_t **osp); int dmu_objset_hold_flags(const char *name, boolean_t decrypt, void *tag, objset_t **osp); int dmu_objset_own(const char *name, dmu_objset_type_t type, boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp); int dmu_objset_own_obj(struct dsl_pool *dp, uint64_t obj, dmu_objset_type_t type, boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp); void dmu_objset_refresh_ownership(struct dsl_dataset *ds, struct dsl_dataset **newds, boolean_t decrypt, void *tag); void dmu_objset_rele(objset_t *os, void *tag); void dmu_objset_rele_flags(objset_t *os, boolean_t decrypt, void *tag); void dmu_objset_disown(objset_t *os, boolean_t decrypt, void *tag); int dmu_objset_from_ds(struct dsl_dataset *ds, objset_t **osp); void dmu_objset_stats(objset_t *os, nvlist_t *nv); void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat); void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, uint64_t *usedobjsp, uint64_t *availobjsp); uint64_t dmu_objset_fsid_guid(objset_t *os); int dmu_objset_find_dp(struct dsl_pool *dp, uint64_t ddobj, int func(struct dsl_pool *, struct dsl_dataset *, void *), void *arg, int flags); void dmu_objset_evict_dbufs(objset_t *os); timestruc_t dmu_objset_snap_cmtime(objset_t *os); /* called from dsl */ void dmu_objset_sync(objset_t *os, zio_t *zio, dmu_tx_t *tx); boolean_t dmu_objset_is_dirty(objset_t *os, uint64_t txg); objset_t *dmu_objset_create_impl_dnstats(spa_t *spa, struct dsl_dataset *ds, blkptr_t *bp, dmu_objset_type_t type, int levels, int blksz, int ibs, dmu_tx_t *tx); objset_t *dmu_objset_create_impl(spa_t *spa, struct dsl_dataset *ds, blkptr_t *bp, dmu_objset_type_t type, dmu_tx_t *tx); int dmu_objset_open_impl(spa_t *spa, struct dsl_dataset *ds, blkptr_t *bp, objset_t **osp); void dmu_objset_evict(objset_t *os); void dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx); void dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx); boolean_t dmu_objset_userused_enabled(objset_t *os); int dmu_objset_userspace_upgrade(objset_t *os); boolean_t dmu_objset_userspace_present(objset_t *os); boolean_t dmu_objset_userobjused_enabled(objset_t *os); boolean_t dmu_objset_userobjspace_upgradable(objset_t *os); boolean_t dmu_objset_userobjspace_present(objset_t *os); boolean_t dmu_objset_incompatible_encryption_version(objset_t *os); boolean_t dmu_objset_projectquota_enabled(objset_t *os); boolean_t dmu_objset_projectquota_present(objset_t *os); boolean_t dmu_objset_projectquota_upgradable(objset_t *os); void dmu_objset_id_quota_upgrade(objset_t *os); int dmu_fsname(const char *snapname, char *buf); void dmu_objset_evict_done(objset_t *os); void dmu_objset_willuse_space(objset_t *os, int64_t space, dmu_tx_t *tx); void dmu_objset_init(void); void dmu_objset_fini(void); #ifdef __cplusplus } #endif #endif /* _SYS_DMU_OBJSET_H */ diff --git a/module/zfs/dbuf.c b/module/zfs/dbuf.c index b0ae6cc7294d..bc03317a26c8 100644 --- a/module/zfs/dbuf.c +++ b/module/zfs/dbuf.c @@ -1,4382 +1,4399 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, 2017 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include kstat_t *dbuf_ksp; typedef struct dbuf_stats { /* * Various statistics about the size of the dbuf cache. */ kstat_named_t cache_count; kstat_named_t cache_size_bytes; kstat_named_t cache_size_bytes_max; /* * Statistics regarding the bounds on the dbuf cache size. */ kstat_named_t cache_target_bytes; kstat_named_t cache_lowater_bytes; kstat_named_t cache_hiwater_bytes; /* * Total number of dbuf cache evictions that have occurred. */ kstat_named_t cache_total_evicts; /* * The distribution of dbuf levels in the dbuf cache and * the total size of all dbufs at each level. */ kstat_named_t cache_levels[DN_MAX_LEVELS]; kstat_named_t cache_levels_bytes[DN_MAX_LEVELS]; /* * Statistics about the dbuf hash table. */ kstat_named_t hash_hits; kstat_named_t hash_misses; kstat_named_t hash_collisions; kstat_named_t hash_elements; kstat_named_t hash_elements_max; /* * Number of sublists containing more than one dbuf in the dbuf * hash table. Keep track of the longest hash chain. */ kstat_named_t hash_chains; kstat_named_t hash_chain_max; /* * Number of times a dbuf_create() discovers that a dbuf was * already created and in the dbuf hash table. */ kstat_named_t hash_insert_race; } dbuf_stats_t; dbuf_stats_t dbuf_stats = { { "cache_count", KSTAT_DATA_UINT64 }, { "cache_size_bytes", KSTAT_DATA_UINT64 }, { "cache_size_bytes_max", KSTAT_DATA_UINT64 }, { "cache_target_bytes", KSTAT_DATA_UINT64 }, { "cache_lowater_bytes", KSTAT_DATA_UINT64 }, { "cache_hiwater_bytes", KSTAT_DATA_UINT64 }, { "cache_total_evicts", KSTAT_DATA_UINT64 }, { { "cache_levels_N", KSTAT_DATA_UINT64 } }, { { "cache_levels_bytes_N", KSTAT_DATA_UINT64 } }, { "hash_hits", KSTAT_DATA_UINT64 }, { "hash_misses", KSTAT_DATA_UINT64 }, { "hash_collisions", KSTAT_DATA_UINT64 }, { "hash_elements", KSTAT_DATA_UINT64 }, { "hash_elements_max", KSTAT_DATA_UINT64 }, { "hash_chains", KSTAT_DATA_UINT64 }, { "hash_chain_max", KSTAT_DATA_UINT64 }, { "hash_insert_race", KSTAT_DATA_UINT64 } }; #define DBUF_STAT_INCR(stat, val) \ atomic_add_64(&dbuf_stats.stat.value.ui64, (val)); #define DBUF_STAT_DECR(stat, val) \ DBUF_STAT_INCR(stat, -(val)); #define DBUF_STAT_BUMP(stat) \ DBUF_STAT_INCR(stat, 1); #define DBUF_STAT_BUMPDOWN(stat) \ DBUF_STAT_INCR(stat, -1); #define DBUF_STAT_MAX(stat, v) { \ uint64_t _m; \ while ((v) > (_m = dbuf_stats.stat.value.ui64) && \ (_m != atomic_cas_64(&dbuf_stats.stat.value.ui64, _m, (v))))\ continue; \ } struct dbuf_hold_impl_data { /* Function arguments */ dnode_t *dh_dn; uint8_t dh_level; uint64_t dh_blkid; boolean_t dh_fail_sparse; boolean_t dh_fail_uncached; void *dh_tag; dmu_buf_impl_t **dh_dbp; /* Local variables */ dmu_buf_impl_t *dh_db; dmu_buf_impl_t *dh_parent; blkptr_t *dh_bp; int dh_err; dbuf_dirty_record_t *dh_dr; int dh_depth; }; static void __dbuf_hold_impl_init(struct dbuf_hold_impl_data *dh, dnode_t *dn, uint8_t level, uint64_t blkid, boolean_t fail_sparse, boolean_t fail_uncached, void *tag, dmu_buf_impl_t **dbp, int depth); static int __dbuf_hold_impl(struct dbuf_hold_impl_data *dh); uint_t zfs_dbuf_evict_key; static boolean_t dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx); static void dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx); extern inline void dmu_buf_init_user(dmu_buf_user_t *dbu, dmu_buf_evict_func_t *evict_func_sync, dmu_buf_evict_func_t *evict_func_async, dmu_buf_t **clear_on_evict_dbufp); /* * Global data structures and functions for the dbuf cache. */ static kmem_cache_t *dbuf_kmem_cache; static taskq_t *dbu_evict_taskq; static kthread_t *dbuf_cache_evict_thread; static kmutex_t dbuf_evict_lock; static kcondvar_t dbuf_evict_cv; static boolean_t dbuf_evict_thread_exit; /* * LRU cache of dbufs. The dbuf cache maintains a list of dbufs that * are not currently held but have been recently released. These dbufs * are not eligible for arc eviction until they are aged out of the cache. * Dbufs are added to the dbuf cache once the last hold is released. If a * dbuf is later accessed and still exists in the dbuf cache, then it will * be removed from the cache and later re-added to the head of the cache. * Dbufs that are aged out of the cache will be immediately destroyed and * become eligible for arc eviction. */ static multilist_t *dbuf_cache; static refcount_t dbuf_cache_size; unsigned long dbuf_cache_max_bytes = 0; /* Set the default size of the dbuf cache to log2 fraction of arc size. */ int dbuf_cache_shift = 5; /* * The dbuf cache uses a three-stage eviction policy: * - A low water marker designates when the dbuf eviction thread * should stop evicting from the dbuf cache. * - When we reach the maximum size (aka mid water mark), we * signal the eviction thread to run. * - The high water mark indicates when the eviction thread * is unable to keep up with the incoming load and eviction must * happen in the context of the calling thread. * * The dbuf cache: * (max size) * low water mid water hi water * +----------------------------------------+----------+----------+ * | | | | * | | | | * | | | | * | | | | * +----------------------------------------+----------+----------+ * stop signal evict * evicting eviction directly * thread * * The high and low water marks indicate the operating range for the eviction * thread. The low water mark is, by default, 90% of the total size of the * cache and the high water mark is at 110% (both of these percentages can be * changed by setting dbuf_cache_lowater_pct and dbuf_cache_hiwater_pct, * respectively). The eviction thread will try to ensure that the cache remains * within this range by waking up every second and checking if the cache is * above the low water mark. The thread can also be woken up by callers adding * elements into the cache if the cache is larger than the mid water (i.e max * cache size). Once the eviction thread is woken up and eviction is required, * it will continue evicting buffers until it's able to reduce the cache size * to the low water mark. If the cache size continues to grow and hits the high * water mark, then callers adding elements to the cache will begin to evict * directly from the cache until the cache is no longer above the high water * mark. */ /* * The percentage above and below the maximum cache size. */ uint_t dbuf_cache_hiwater_pct = 10; uint_t dbuf_cache_lowater_pct = 10; /* ARGSUSED */ static int dbuf_cons(void *vdb, void *unused, int kmflag) { dmu_buf_impl_t *db = vdb; bzero(db, sizeof (dmu_buf_impl_t)); mutex_init(&db->db_mtx, NULL, MUTEX_DEFAULT, NULL); cv_init(&db->db_changed, NULL, CV_DEFAULT, NULL); multilist_link_init(&db->db_cache_link); refcount_create(&db->db_holds); return (0); } /* ARGSUSED */ static void dbuf_dest(void *vdb, void *unused) { dmu_buf_impl_t *db = vdb; mutex_destroy(&db->db_mtx); cv_destroy(&db->db_changed); ASSERT(!multilist_link_active(&db->db_cache_link)); refcount_destroy(&db->db_holds); } /* * dbuf hash table routines */ static dbuf_hash_table_t dbuf_hash_table; static uint64_t dbuf_hash_count; static uint64_t dbuf_hash(void *os, uint64_t obj, uint8_t lvl, uint64_t blkid) { uintptr_t osv = (uintptr_t)os; uint64_t crc = -1ULL; ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY); crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (lvl)) & 0xFF]; crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 6)) & 0xFF]; crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF]; crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF]; crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (blkid >> 0)) & 0xFF]; crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (blkid >> 8)) & 0xFF]; crc ^= (osv>>14) ^ (obj>>16) ^ (blkid>>16); return (crc); } #define DBUF_EQUAL(dbuf, os, obj, level, blkid) \ ((dbuf)->db.db_object == (obj) && \ (dbuf)->db_objset == (os) && \ (dbuf)->db_level == (level) && \ (dbuf)->db_blkid == (blkid)) dmu_buf_impl_t * dbuf_find(objset_t *os, uint64_t obj, uint8_t level, uint64_t blkid) { dbuf_hash_table_t *h = &dbuf_hash_table; uint64_t hv; uint64_t idx; dmu_buf_impl_t *db; hv = dbuf_hash(os, obj, level, blkid); idx = hv & h->hash_table_mask; mutex_enter(DBUF_HASH_MUTEX(h, idx)); for (db = h->hash_table[idx]; db != NULL; db = db->db_hash_next) { if (DBUF_EQUAL(db, os, obj, level, blkid)) { mutex_enter(&db->db_mtx); if (db->db_state != DB_EVICTING) { mutex_exit(DBUF_HASH_MUTEX(h, idx)); return (db); } mutex_exit(&db->db_mtx); } } mutex_exit(DBUF_HASH_MUTEX(h, idx)); return (NULL); } static dmu_buf_impl_t * dbuf_find_bonus(objset_t *os, uint64_t object) { dnode_t *dn; dmu_buf_impl_t *db = NULL; if (dnode_hold(os, object, FTAG, &dn) == 0) { rw_enter(&dn->dn_struct_rwlock, RW_READER); if (dn->dn_bonus != NULL) { db = dn->dn_bonus; mutex_enter(&db->db_mtx); } rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); } return (db); } /* * Insert an entry into the hash table. If there is already an element * equal to elem in the hash table, then the already existing element * will be returned and the new element will not be inserted. * Otherwise returns NULL. */ static dmu_buf_impl_t * dbuf_hash_insert(dmu_buf_impl_t *db) { dbuf_hash_table_t *h = &dbuf_hash_table; objset_t *os = db->db_objset; uint64_t obj = db->db.db_object; int level = db->db_level; uint64_t blkid, hv, idx; dmu_buf_impl_t *dbf; uint32_t i; blkid = db->db_blkid; hv = dbuf_hash(os, obj, level, blkid); idx = hv & h->hash_table_mask; mutex_enter(DBUF_HASH_MUTEX(h, idx)); for (dbf = h->hash_table[idx], i = 0; dbf != NULL; dbf = dbf->db_hash_next, i++) { if (DBUF_EQUAL(dbf, os, obj, level, blkid)) { mutex_enter(&dbf->db_mtx); if (dbf->db_state != DB_EVICTING) { mutex_exit(DBUF_HASH_MUTEX(h, idx)); return (dbf); } mutex_exit(&dbf->db_mtx); } } if (i > 0) { DBUF_STAT_BUMP(hash_collisions); if (i == 1) DBUF_STAT_BUMP(hash_chains); DBUF_STAT_MAX(hash_chain_max, i); } mutex_enter(&db->db_mtx); db->db_hash_next = h->hash_table[idx]; h->hash_table[idx] = db; mutex_exit(DBUF_HASH_MUTEX(h, idx)); atomic_inc_64(&dbuf_hash_count); DBUF_STAT_MAX(hash_elements_max, dbuf_hash_count); return (NULL); } /* * Remove an entry from the hash table. It must be in the EVICTING state. */ static void dbuf_hash_remove(dmu_buf_impl_t *db) { dbuf_hash_table_t *h = &dbuf_hash_table; uint64_t hv, idx; dmu_buf_impl_t *dbf, **dbp; hv = dbuf_hash(db->db_objset, db->db.db_object, db->db_level, db->db_blkid); idx = hv & h->hash_table_mask; /* * We mustn't hold db_mtx to maintain lock ordering: * DBUF_HASH_MUTEX > db_mtx. */ ASSERT(refcount_is_zero(&db->db_holds)); ASSERT(db->db_state == DB_EVICTING); ASSERT(!MUTEX_HELD(&db->db_mtx)); mutex_enter(DBUF_HASH_MUTEX(h, idx)); dbp = &h->hash_table[idx]; while ((dbf = *dbp) != db) { dbp = &dbf->db_hash_next; ASSERT(dbf != NULL); } *dbp = db->db_hash_next; db->db_hash_next = NULL; if (h->hash_table[idx] && h->hash_table[idx]->db_hash_next == NULL) DBUF_STAT_BUMPDOWN(hash_chains); mutex_exit(DBUF_HASH_MUTEX(h, idx)); atomic_dec_64(&dbuf_hash_count); } typedef enum { DBVU_EVICTING, DBVU_NOT_EVICTING } dbvu_verify_type_t; static void dbuf_verify_user(dmu_buf_impl_t *db, dbvu_verify_type_t verify_type) { #ifdef ZFS_DEBUG int64_t holds; if (db->db_user == NULL) return; /* Only data blocks support the attachment of user data. */ ASSERT(db->db_level == 0); /* Clients must resolve a dbuf before attaching user data. */ ASSERT(db->db.db_data != NULL); ASSERT3U(db->db_state, ==, DB_CACHED); holds = refcount_count(&db->db_holds); if (verify_type == DBVU_EVICTING) { /* * Immediate eviction occurs when holds == dirtycnt. * For normal eviction buffers, holds is zero on * eviction, except when dbuf_fix_old_data() calls * dbuf_clear_data(). However, the hold count can grow * during eviction even though db_mtx is held (see * dmu_bonus_hold() for an example), so we can only * test the generic invariant that holds >= dirtycnt. */ ASSERT3U(holds, >=, db->db_dirtycnt); } else { if (db->db_user_immediate_evict == TRUE) ASSERT3U(holds, >=, db->db_dirtycnt); else ASSERT3U(holds, >, 0); } #endif } static void dbuf_evict_user(dmu_buf_impl_t *db) { dmu_buf_user_t *dbu = db->db_user; ASSERT(MUTEX_HELD(&db->db_mtx)); if (dbu == NULL) return; dbuf_verify_user(db, DBVU_EVICTING); db->db_user = NULL; #ifdef ZFS_DEBUG if (dbu->dbu_clear_on_evict_dbufp != NULL) *dbu->dbu_clear_on_evict_dbufp = NULL; #endif /* * There are two eviction callbacks - one that we call synchronously * and one that we invoke via a taskq. The async one is useful for * avoiding lock order reversals and limiting stack depth. * * Note that if we have a sync callback but no async callback, * it's likely that the sync callback will free the structure * containing the dbu. In that case we need to take care to not * dereference dbu after calling the sync evict func. */ boolean_t has_async = (dbu->dbu_evict_func_async != NULL); if (dbu->dbu_evict_func_sync != NULL) dbu->dbu_evict_func_sync(dbu); if (has_async) { taskq_dispatch_ent(dbu_evict_taskq, dbu->dbu_evict_func_async, dbu, 0, &dbu->dbu_tqent); } } boolean_t dbuf_is_metadata(dmu_buf_impl_t *db) { /* * Consider indirect blocks and spill blocks to be meta data. */ if (db->db_level > 0 || db->db_blkid == DMU_SPILL_BLKID) { return (B_TRUE); } else { boolean_t is_metadata; DB_DNODE_ENTER(db); is_metadata = DMU_OT_IS_METADATA(DB_DNODE(db)->dn_type); DB_DNODE_EXIT(db); return (is_metadata); } } /* * This function *must* return indices evenly distributed between all * sublists of the multilist. This is needed due to how the dbuf eviction * code is laid out; dbuf_evict_thread() assumes dbufs are evenly * distributed between all sublists and uses this assumption when * deciding which sublist to evict from and how much to evict from it. */ unsigned int dbuf_cache_multilist_index_func(multilist_t *ml, void *obj) { dmu_buf_impl_t *db = obj; /* * The assumption here, is the hash value for a given * dmu_buf_impl_t will remain constant throughout it's lifetime * (i.e. it's objset, object, level and blkid fields don't change). * Thus, we don't need to store the dbuf's sublist index * on insertion, as this index can be recalculated on removal. * * Also, the low order bits of the hash value are thought to be * distributed evenly. Otherwise, in the case that the multilist * has a power of two number of sublists, each sublists' usage * would not be evenly distributed. */ return (dbuf_hash(db->db_objset, db->db.db_object, db->db_level, db->db_blkid) % multilist_get_num_sublists(ml)); } static inline unsigned long dbuf_cache_target_bytes(void) { return MIN(dbuf_cache_max_bytes, arc_target_bytes() >> dbuf_cache_shift); } static inline uint64_t dbuf_cache_hiwater_bytes(void) { uint64_t dbuf_cache_target = dbuf_cache_target_bytes(); return (dbuf_cache_target + (dbuf_cache_target * dbuf_cache_hiwater_pct) / 100); } static inline uint64_t dbuf_cache_lowater_bytes(void) { uint64_t dbuf_cache_target = dbuf_cache_target_bytes(); return (dbuf_cache_target - (dbuf_cache_target * dbuf_cache_lowater_pct) / 100); } static inline boolean_t dbuf_cache_above_hiwater(void) { return (refcount_count(&dbuf_cache_size) > dbuf_cache_hiwater_bytes()); } static inline boolean_t dbuf_cache_above_lowater(void) { return (refcount_count(&dbuf_cache_size) > dbuf_cache_lowater_bytes()); } /* * Evict the oldest eligible dbuf from the dbuf cache. */ static void dbuf_evict_one(void) { int idx = multilist_get_random_index(dbuf_cache); multilist_sublist_t *mls = multilist_sublist_lock(dbuf_cache, idx); ASSERT(!MUTEX_HELD(&dbuf_evict_lock)); /* * Set the thread's tsd to indicate that it's processing evictions. * Once a thread stops evicting from the dbuf cache it will * reset its tsd to NULL. */ ASSERT3P(tsd_get(zfs_dbuf_evict_key), ==, NULL); (void) tsd_set(zfs_dbuf_evict_key, (void *)B_TRUE); dmu_buf_impl_t *db = multilist_sublist_tail(mls); while (db != NULL && mutex_tryenter(&db->db_mtx) == 0) { db = multilist_sublist_prev(mls, db); } DTRACE_PROBE2(dbuf__evict__one, dmu_buf_impl_t *, db, multilist_sublist_t *, mls); if (db != NULL) { multilist_sublist_remove(mls, db); multilist_sublist_unlock(mls); (void) refcount_remove_many(&dbuf_cache_size, db->db.db_size, db); DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[db->db_level], db->db.db_size); dbuf_destroy(db); DBUF_STAT_MAX(cache_size_bytes_max, refcount_count(&dbuf_cache_size)); DBUF_STAT_BUMP(cache_total_evicts); } else { multilist_sublist_unlock(mls); } (void) tsd_set(zfs_dbuf_evict_key, NULL); } /* * The dbuf evict thread is responsible for aging out dbufs from the * cache. Once the cache has reached it's maximum size, dbufs are removed * and destroyed. The eviction thread will continue running until the size * of the dbuf cache is at or below the maximum size. Once the dbuf is aged * out of the cache it is destroyed and becomes eligible for arc eviction. */ /* ARGSUSED */ static void dbuf_evict_thread(void *unused) { callb_cpr_t cpr; CALLB_CPR_INIT(&cpr, &dbuf_evict_lock, callb_generic_cpr, FTAG); mutex_enter(&dbuf_evict_lock); while (!dbuf_evict_thread_exit) { while (!dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) { CALLB_CPR_SAFE_BEGIN(&cpr); (void) cv_timedwait_sig_hires(&dbuf_evict_cv, &dbuf_evict_lock, SEC2NSEC(1), MSEC2NSEC(1), 0); CALLB_CPR_SAFE_END(&cpr, &dbuf_evict_lock); } mutex_exit(&dbuf_evict_lock); /* * Keep evicting as long as we're above the low water mark * for the cache. We do this without holding the locks to * minimize lock contention. */ while (dbuf_cache_above_lowater() && !dbuf_evict_thread_exit) { dbuf_evict_one(); } mutex_enter(&dbuf_evict_lock); } dbuf_evict_thread_exit = B_FALSE; cv_broadcast(&dbuf_evict_cv); CALLB_CPR_EXIT(&cpr); /* drops dbuf_evict_lock */ thread_exit(); } /* * Wake up the dbuf eviction thread if the dbuf cache is at its max size. * If the dbuf cache is at its high water mark, then evict a dbuf from the * dbuf cache using the callers context. */ static void dbuf_evict_notify(void) { /* * We use thread specific data to track when a thread has * started processing evictions. This allows us to avoid deeply * nested stacks that would have a call flow similar to this: * * dbuf_rele()-->dbuf_rele_and_unlock()-->dbuf_evict_notify() * ^ | * | | * +-----dbuf_destroy()<--dbuf_evict_one()<--------+ * * The dbuf_eviction_thread will always have its tsd set until * that thread exits. All other threads will only set their tsd * if they are participating in the eviction process. This only * happens if the eviction thread is unable to process evictions * fast enough. To keep the dbuf cache size in check, other threads * can evict from the dbuf cache directly. Those threads will set * their tsd values so that we ensure that they only evict one dbuf * from the dbuf cache. */ if (tsd_get(zfs_dbuf_evict_key) != NULL) return; /* * We check if we should evict without holding the dbuf_evict_lock, * because it's OK to occasionally make the wrong decision here, * and grabbing the lock results in massive lock contention. */ if (refcount_count(&dbuf_cache_size) > dbuf_cache_target_bytes()) { if (dbuf_cache_above_hiwater()) dbuf_evict_one(); cv_signal(&dbuf_evict_cv); } } static int dbuf_kstat_update(kstat_t *ksp, int rw) { dbuf_stats_t *ds = ksp->ks_data; if (rw == KSTAT_WRITE) { return (SET_ERROR(EACCES)); } else { ds->cache_size_bytes.value.ui64 = refcount_count(&dbuf_cache_size); ds->cache_target_bytes.value.ui64 = dbuf_cache_target_bytes(); ds->cache_hiwater_bytes.value.ui64 = dbuf_cache_hiwater_bytes(); ds->cache_lowater_bytes.value.ui64 = dbuf_cache_lowater_bytes(); ds->hash_elements.value.ui64 = dbuf_hash_count; } return (0); } void dbuf_init(void) { uint64_t hsize = 1ULL << 16; dbuf_hash_table_t *h = &dbuf_hash_table; int i; /* * The hash table is big enough to fill all of physical memory * with an average block size of zfs_arc_average_blocksize (default 8K). * By default, the table will take up * totalmem * sizeof(void*) / 8K (1MB per GB with 8-byte pointers). */ while (hsize * zfs_arc_average_blocksize < physmem * PAGESIZE) hsize <<= 1; retry: h->hash_table_mask = hsize - 1; #if defined(_KERNEL) && defined(HAVE_SPL) /* * Large allocations which do not require contiguous pages * should be using vmem_alloc() in the linux kernel */ h->hash_table = vmem_zalloc(hsize * sizeof (void *), KM_SLEEP); #else h->hash_table = kmem_zalloc(hsize * sizeof (void *), KM_NOSLEEP); #endif if (h->hash_table == NULL) { /* XXX - we should really return an error instead of assert */ ASSERT(hsize > (1ULL << 10)); hsize >>= 1; goto retry; } dbuf_kmem_cache = kmem_cache_create("dmu_buf_impl_t", sizeof (dmu_buf_impl_t), 0, dbuf_cons, dbuf_dest, NULL, NULL, NULL, 0); for (i = 0; i < DBUF_MUTEXES; i++) mutex_init(&h->hash_mutexes[i], NULL, MUTEX_DEFAULT, NULL); dbuf_stats_init(h); /* * Setup the parameters for the dbuf cache. We set the size of the * dbuf cache to 1/32nd (default) of the target size of the ARC. If * the value has been specified as a module option and it's not * greater than the target size of the ARC, then we honor that value. */ if (dbuf_cache_max_bytes == 0 || dbuf_cache_max_bytes >= arc_target_bytes()) { dbuf_cache_max_bytes = arc_target_bytes() >> dbuf_cache_shift; } /* * All entries are queued via taskq_dispatch_ent(), so min/maxalloc * configuration is not required. */ dbu_evict_taskq = taskq_create("dbu_evict", 1, defclsyspri, 0, 0, 0); dbuf_cache = multilist_create(sizeof (dmu_buf_impl_t), offsetof(dmu_buf_impl_t, db_cache_link), dbuf_cache_multilist_index_func); refcount_create(&dbuf_cache_size); tsd_create(&zfs_dbuf_evict_key, NULL); dbuf_evict_thread_exit = B_FALSE; mutex_init(&dbuf_evict_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&dbuf_evict_cv, NULL, CV_DEFAULT, NULL); dbuf_cache_evict_thread = thread_create(NULL, 0, dbuf_evict_thread, NULL, 0, &p0, TS_RUN, minclsyspri); dbuf_ksp = kstat_create("zfs", 0, "dbufstats", "misc", KSTAT_TYPE_NAMED, sizeof (dbuf_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (dbuf_ksp != NULL) { dbuf_ksp->ks_data = &dbuf_stats; dbuf_ksp->ks_update = dbuf_kstat_update; kstat_install(dbuf_ksp); for (i = 0; i < DN_MAX_LEVELS; i++) { snprintf(dbuf_stats.cache_levels[i].name, KSTAT_STRLEN, "cache_level_%d", i); dbuf_stats.cache_levels[i].data_type = KSTAT_DATA_UINT64; snprintf(dbuf_stats.cache_levels_bytes[i].name, KSTAT_STRLEN, "cache_level_%d_bytes", i); dbuf_stats.cache_levels_bytes[i].data_type = KSTAT_DATA_UINT64; } } } void dbuf_fini(void) { dbuf_hash_table_t *h = &dbuf_hash_table; int i; dbuf_stats_destroy(); for (i = 0; i < DBUF_MUTEXES; i++) mutex_destroy(&h->hash_mutexes[i]); #if defined(_KERNEL) && defined(HAVE_SPL) /* * Large allocations which do not require contiguous pages * should be using vmem_free() in the linux kernel */ vmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *)); #else kmem_free(h->hash_table, (h->hash_table_mask + 1) * sizeof (void *)); #endif kmem_cache_destroy(dbuf_kmem_cache); taskq_destroy(dbu_evict_taskq); mutex_enter(&dbuf_evict_lock); dbuf_evict_thread_exit = B_TRUE; while (dbuf_evict_thread_exit) { cv_signal(&dbuf_evict_cv); cv_wait(&dbuf_evict_cv, &dbuf_evict_lock); } mutex_exit(&dbuf_evict_lock); tsd_destroy(&zfs_dbuf_evict_key); mutex_destroy(&dbuf_evict_lock); cv_destroy(&dbuf_evict_cv); refcount_destroy(&dbuf_cache_size); multilist_destroy(dbuf_cache); if (dbuf_ksp != NULL) { kstat_delete(dbuf_ksp); dbuf_ksp = NULL; } } /* * Other stuff. */ #ifdef ZFS_DEBUG static void dbuf_verify(dmu_buf_impl_t *db) { dnode_t *dn; dbuf_dirty_record_t *dr; ASSERT(MUTEX_HELD(&db->db_mtx)); if (!(zfs_flags & ZFS_DEBUG_DBUF_VERIFY)) return; ASSERT(db->db_objset != NULL); DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dn == NULL) { ASSERT(db->db_parent == NULL); ASSERT(db->db_blkptr == NULL); } else { ASSERT3U(db->db.db_object, ==, dn->dn_object); ASSERT3P(db->db_objset, ==, dn->dn_objset); ASSERT3U(db->db_level, <, dn->dn_nlevels); ASSERT(db->db_blkid == DMU_BONUS_BLKID || db->db_blkid == DMU_SPILL_BLKID || !avl_is_empty(&dn->dn_dbufs)); } if (db->db_blkid == DMU_BONUS_BLKID) { ASSERT(dn != NULL); ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen); ASSERT3U(db->db.db_offset, ==, DMU_BONUS_BLKID); } else if (db->db_blkid == DMU_SPILL_BLKID) { ASSERT(dn != NULL); ASSERT0(db->db.db_offset); } else { ASSERT3U(db->db.db_offset, ==, db->db_blkid * db->db.db_size); } for (dr = db->db_data_pending; dr != NULL; dr = dr->dr_next) ASSERT(dr->dr_dbuf == db); for (dr = db->db_last_dirty; dr != NULL; dr = dr->dr_next) ASSERT(dr->dr_dbuf == db); /* * We can't assert that db_size matches dn_datablksz because it * can be momentarily different when another thread is doing * dnode_set_blksz(). */ if (db->db_level == 0 && db->db.db_object == DMU_META_DNODE_OBJECT) { dr = db->db_data_pending; /* * It should only be modified in syncing context, so * make sure we only have one copy of the data. */ ASSERT(dr == NULL || dr->dt.dl.dr_data == db->db_buf); } /* verify db->db_blkptr */ if (db->db_blkptr) { if (db->db_parent == dn->dn_dbuf) { /* db is pointed to by the dnode */ /* ASSERT3U(db->db_blkid, <, dn->dn_nblkptr); */ if (DMU_OBJECT_IS_SPECIAL(db->db.db_object)) ASSERT(db->db_parent == NULL); else ASSERT(db->db_parent != NULL); if (db->db_blkid != DMU_SPILL_BLKID) ASSERT3P(db->db_blkptr, ==, &dn->dn_phys->dn_blkptr[db->db_blkid]); } else { /* db is pointed to by an indirect block */ ASSERTV(int epb = db->db_parent->db.db_size >> SPA_BLKPTRSHIFT); ASSERT3U(db->db_parent->db_level, ==, db->db_level+1); ASSERT3U(db->db_parent->db.db_object, ==, db->db.db_object); /* * dnode_grow_indblksz() can make this fail if we don't * have the struct_rwlock. XXX indblksz no longer * grows. safe to do this now? */ if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) { ASSERT3P(db->db_blkptr, ==, ((blkptr_t *)db->db_parent->db.db_data + db->db_blkid % epb)); } } } if ((db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr)) && (db->db_buf == NULL || db->db_buf->b_data) && db->db.db_data && db->db_blkid != DMU_BONUS_BLKID && db->db_state != DB_FILL && !dn->dn_free_txg) { /* * If the blkptr isn't set but they have nonzero data, * it had better be dirty, otherwise we'll lose that * data when we evict this buffer. * * There is an exception to this rule for indirect blocks; in * this case, if the indirect block is a hole, we fill in a few * fields on each of the child blocks (importantly, birth time) * to prevent hole birth times from being lost when you * partially fill in a hole. */ if (db->db_dirtycnt == 0) { if (db->db_level == 0) { uint64_t *buf = db->db.db_data; int i; for (i = 0; i < db->db.db_size >> 3; i++) { ASSERT(buf[i] == 0); } } else { blkptr_t *bps = db->db.db_data; ASSERT3U(1 << DB_DNODE(db)->dn_indblkshift, ==, db->db.db_size); /* * We want to verify that all the blkptrs in the * indirect block are holes, but we may have * automatically set up a few fields for them. * We iterate through each blkptr and verify * they only have those fields set. */ for (int i = 0; i < db->db.db_size / sizeof (blkptr_t); i++) { blkptr_t *bp = &bps[i]; ASSERT(ZIO_CHECKSUM_IS_ZERO( &bp->blk_cksum)); ASSERT( DVA_IS_EMPTY(&bp->blk_dva[0]) && DVA_IS_EMPTY(&bp->blk_dva[1]) && DVA_IS_EMPTY(&bp->blk_dva[2])); ASSERT0(bp->blk_fill); ASSERT0(bp->blk_pad[0]); ASSERT0(bp->blk_pad[1]); ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT(BP_IS_HOLE(bp)); ASSERT0(bp->blk_phys_birth); } } } } DB_DNODE_EXIT(db); } #endif static void dbuf_clear_data(dmu_buf_impl_t *db) { ASSERT(MUTEX_HELD(&db->db_mtx)); dbuf_evict_user(db); ASSERT3P(db->db_buf, ==, NULL); db->db.db_data = NULL; if (db->db_state != DB_NOFILL) db->db_state = DB_UNCACHED; } static void dbuf_set_data(dmu_buf_impl_t *db, arc_buf_t *buf) { ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(buf != NULL); db->db_buf = buf; ASSERT(buf->b_data != NULL); db->db.db_data = buf->b_data; } /* * Loan out an arc_buf for read. Return the loaned arc_buf. */ arc_buf_t * dbuf_loan_arcbuf(dmu_buf_impl_t *db) { arc_buf_t *abuf; ASSERT(db->db_blkid != DMU_BONUS_BLKID); mutex_enter(&db->db_mtx); if (arc_released(db->db_buf) || refcount_count(&db->db_holds) > 1) { int blksz = db->db.db_size; spa_t *spa = db->db_objset->os_spa; mutex_exit(&db->db_mtx); abuf = arc_loan_buf(spa, B_FALSE, blksz); bcopy(db->db.db_data, abuf->b_data, blksz); } else { abuf = db->db_buf; arc_loan_inuse_buf(abuf, db); db->db_buf = NULL; dbuf_clear_data(db); mutex_exit(&db->db_mtx); } return (abuf); } /* * Calculate which level n block references the data at the level 0 offset * provided. */ uint64_t dbuf_whichblock(const dnode_t *dn, const int64_t level, const uint64_t offset) { if (dn->dn_datablkshift != 0 && dn->dn_indblkshift != 0) { /* * The level n blkid is equal to the level 0 blkid divided by * the number of level 0s in a level n block. * * The level 0 blkid is offset >> datablkshift = * offset / 2^datablkshift. * * The number of level 0s in a level n is the number of block * pointers in an indirect block, raised to the power of level. * This is 2^(indblkshift - SPA_BLKPTRSHIFT)^level = * 2^(level*(indblkshift - SPA_BLKPTRSHIFT)). * * Thus, the level n blkid is: offset / * ((2^datablkshift)*(2^(level*(indblkshift - SPA_BLKPTRSHIFT))) * = offset / 2^(datablkshift + level * * (indblkshift - SPA_BLKPTRSHIFT)) * = offset >> (datablkshift + level * * (indblkshift - SPA_BLKPTRSHIFT)) */ const unsigned exp = dn->dn_datablkshift + level * (dn->dn_indblkshift - SPA_BLKPTRSHIFT); if (exp >= 8 * sizeof (offset)) { /* This only happens on the highest indirection level */ ASSERT3U(level, ==, dn->dn_nlevels - 1); return (0); } ASSERT3U(exp, <, 8 * sizeof (offset)); return (offset >> exp); } else { ASSERT3U(offset, <, dn->dn_datablksz); return (0); } } static void dbuf_read_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *bp, arc_buf_t *buf, void *vdb) { dmu_buf_impl_t *db = vdb; mutex_enter(&db->db_mtx); ASSERT3U(db->db_state, ==, DB_READ); /* * All reads are synchronous, so we must have a hold on the dbuf */ ASSERT(refcount_count(&db->db_holds) > 0); ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); if (db->db_level == 0 && db->db_freed_in_flight) { /* we were freed in flight; disregard any error */ if (buf == NULL) { buf = arc_alloc_buf(db->db_objset->os_spa, db, DBUF_GET_BUFC_TYPE(db), db->db.db_size); } arc_release(buf, db); bzero(buf->b_data, db->db.db_size); arc_buf_freeze(buf); db->db_freed_in_flight = FALSE; dbuf_set_data(db, buf); db->db_state = DB_CACHED; } else if (buf != NULL) { dbuf_set_data(db, buf); db->db_state = DB_CACHED; } else { ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT3P(db->db_buf, ==, NULL); db->db_state = DB_UNCACHED; } cv_broadcast(&db->db_changed); dbuf_rele_and_unlock(db, NULL); } static int dbuf_read_impl(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags) { dnode_t *dn; zbookmark_phys_t zb; uint32_t aflags = ARC_FLAG_NOWAIT; int err, zio_flags = 0; DB_DNODE_ENTER(db); dn = DB_DNODE(db); ASSERT(!refcount_is_zero(&db->db_holds)); /* We need the struct_rwlock to prevent db_blkptr from changing. */ ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db_state == DB_UNCACHED); ASSERT(db->db_buf == NULL); if (db->db_blkid == DMU_BONUS_BLKID) { /* * The bonus length stored in the dnode may be less than * the maximum available space in the bonus buffer. */ int bonuslen = MIN(dn->dn_bonuslen, dn->dn_phys->dn_bonuslen); int max_bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots); arc_buf_t *dn_buf = (dn->dn_dbuf != NULL) ? dn->dn_dbuf->db_buf : NULL; /* if the underlying dnode block is encrypted, decrypt it */ if (dn_buf != NULL && dn->dn_objset->os_encrypted && DMU_OT_IS_ENCRYPTED(dn->dn_bonustype) && (flags & DB_RF_NO_DECRYPT) == 0 && arc_is_encrypted(dn_buf)) { SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), DMU_META_DNODE_OBJECT, 0, dn->dn_dbuf->db_blkid); err = arc_untransform(dn_buf, dn->dn_objset->os_spa, &zb, B_TRUE); if (err != 0) { DB_DNODE_EXIT(db); mutex_exit(&db->db_mtx); return (err); } } ASSERT3U(bonuslen, <=, db->db.db_size); db->db.db_data = kmem_alloc(max_bonuslen, KM_SLEEP); arc_space_consume(max_bonuslen, ARC_SPACE_BONUS); if (bonuslen < max_bonuslen) bzero(db->db.db_data, max_bonuslen); if (bonuslen) bcopy(DN_BONUS(dn->dn_phys), db->db.db_data, bonuslen); DB_DNODE_EXIT(db); db->db_state = DB_CACHED; mutex_exit(&db->db_mtx); return (0); } /* * Recheck BP_IS_HOLE() after dnode_block_freed() in case dnode_sync() * processes the delete record and clears the bp while we are waiting * for the dn_mtx (resulting in a "no" from block_freed). */ if (db->db_blkptr == NULL || BP_IS_HOLE(db->db_blkptr) || (db->db_level == 0 && (dnode_block_freed(dn, db->db_blkid) || BP_IS_HOLE(db->db_blkptr)))) { arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); dbuf_set_data(db, arc_alloc_buf(db->db_objset->os_spa, db, type, db->db.db_size)); bzero(db->db.db_data, db->db.db_size); if (db->db_blkptr != NULL && db->db_level > 0 && BP_IS_HOLE(db->db_blkptr) && db->db_blkptr->blk_birth != 0) { blkptr_t *bps = db->db.db_data; for (int i = 0; i < ((1 << DB_DNODE(db)->dn_indblkshift) / sizeof (blkptr_t)); i++) { blkptr_t *bp = &bps[i]; ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==, 1 << dn->dn_indblkshift); BP_SET_LSIZE(bp, BP_GET_LEVEL(db->db_blkptr) == 1 ? dn->dn_datablksz : BP_GET_LSIZE(db->db_blkptr)); BP_SET_TYPE(bp, BP_GET_TYPE(db->db_blkptr)); BP_SET_LEVEL(bp, BP_GET_LEVEL(db->db_blkptr) - 1); BP_SET_BIRTH(bp, db->db_blkptr->blk_birth, 0); } } DB_DNODE_EXIT(db); db->db_state = DB_CACHED; mutex_exit(&db->db_mtx); return (0); } DB_DNODE_EXIT(db); db->db_state = DB_READ; mutex_exit(&db->db_mtx); if (DBUF_IS_L2CACHEABLE(db)) aflags |= ARC_FLAG_L2CACHE; SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); /* * All bps of an encrypted os should have the encryption bit set. * If this is not true it indicates tampering and we report an error. */ if (db->db_objset->os_encrypted && !BP_USES_CRYPT(db->db_blkptr)) { spa_log_error(db->db_objset->os_spa, &zb); zfs_panic_recover("unencrypted block in encrypted " "object set %llu", dmu_objset_id(db->db_objset)); return (SET_ERROR(EIO)); } dbuf_add_ref(db, NULL); zio_flags = (flags & DB_RF_CANFAIL) ? ZIO_FLAG_CANFAIL : ZIO_FLAG_MUSTSUCCEED; if ((flags & DB_RF_NO_DECRYPT) && BP_IS_PROTECTED(db->db_blkptr)) zio_flags |= ZIO_FLAG_RAW; err = arc_read(zio, db->db_objset->os_spa, db->db_blkptr, dbuf_read_done, db, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb); return (err); } /* * This is our just-in-time copy function. It makes a copy of buffers that * have been modified in a previous transaction group before we access them in * the current active group. * * This function is used in three places: when we are dirtying a buffer for the * first time in a txg, when we are freeing a range in a dnode that includes * this buffer, and when we are accessing a buffer which was received compressed * and later referenced in a WRITE_BYREF record. * * Note that when we are called from dbuf_free_range() we do not put a hold on * the buffer, we just traverse the active dbuf list for the dnode. */ static void dbuf_fix_old_data(dmu_buf_impl_t *db, uint64_t txg) { dbuf_dirty_record_t *dr = db->db_last_dirty; ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(db->db.db_data != NULL); ASSERT(db->db_level == 0); ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT); if (dr == NULL || (dr->dt.dl.dr_data != ((db->db_blkid == DMU_BONUS_BLKID) ? db->db.db_data : db->db_buf))) return; /* * If the last dirty record for this dbuf has not yet synced * and its referencing the dbuf data, either: * reset the reference to point to a new copy, * or (if there a no active holders) * just null out the current db_data pointer. */ ASSERT3U(dr->dr_txg, >=, txg - 2); if (db->db_blkid == DMU_BONUS_BLKID) { dnode_t *dn = DB_DNODE(db); int bonuslen = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots); dr->dt.dl.dr_data = kmem_alloc(bonuslen, KM_SLEEP); arc_space_consume(bonuslen, ARC_SPACE_BONUS); bcopy(db->db.db_data, dr->dt.dl.dr_data, bonuslen); } else if (refcount_count(&db->db_holds) > db->db_dirtycnt) { dnode_t *dn = DB_DNODE(db); int size = arc_buf_size(db->db_buf); arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); spa_t *spa = db->db_objset->os_spa; enum zio_compress compress_type = arc_get_compression(db->db_buf); if (arc_is_encrypted(db->db_buf)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(db->db_buf, &byteorder, salt, iv, mac); dr->dt.dl.dr_data = arc_alloc_raw_buf(spa, db, dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac, dn->dn_type, size, arc_buf_lsize(db->db_buf), compress_type); } else if (compress_type != ZIO_COMPRESS_OFF) { ASSERT3U(type, ==, ARC_BUFC_DATA); dr->dt.dl.dr_data = arc_alloc_compressed_buf(spa, db, size, arc_buf_lsize(db->db_buf), compress_type); } else { dr->dt.dl.dr_data = arc_alloc_buf(spa, db, type, size); } bcopy(db->db.db_data, dr->dt.dl.dr_data->b_data, size); } else { db->db_buf = NULL; dbuf_clear_data(db); } } int dbuf_read(dmu_buf_impl_t *db, zio_t *zio, uint32_t flags) { int err = 0; boolean_t prefetch; dnode_t *dn; /* * We don't have to hold the mutex to check db_state because it * can't be freed while we have a hold on the buffer. */ ASSERT(!refcount_is_zero(&db->db_holds)); if (db->db_state == DB_NOFILL) return (SET_ERROR(EIO)); DB_DNODE_ENTER(db); dn = DB_DNODE(db); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_enter(&dn->dn_struct_rwlock, RW_READER); prefetch = db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID && (flags & DB_RF_NOPREFETCH) == 0 && dn != NULL && DBUF_IS_CACHEABLE(db); mutex_enter(&db->db_mtx); if (db->db_state == DB_CACHED) { spa_t *spa = dn->dn_objset->os_spa; /* * If the arc buf is compressed or encrypted, we need to * untransform it to read the data. This could happen during * the "zfs receive" of a stream which is deduplicated and * either raw or compressed. We do not need to do this if the * caller wants raw encrypted data. */ if (db->db_buf != NULL && (flags & DB_RF_NO_DECRYPT) == 0 && (arc_is_encrypted(db->db_buf) || arc_get_compression(db->db_buf) != ZIO_COMPRESS_OFF)) { zbookmark_phys_t zb; SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); dbuf_fix_old_data(db, spa_syncing_txg(spa)); err = arc_untransform(db->db_buf, spa, &zb, B_FALSE); dbuf_set_data(db, db->db_buf); } mutex_exit(&db->db_mtx); if (prefetch) dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(db); DBUF_STAT_BUMP(hash_hits); } else if (db->db_state == DB_UNCACHED) { spa_t *spa = dn->dn_objset->os_spa; boolean_t need_wait = B_FALSE; if (zio == NULL && db->db_blkptr != NULL && !BP_IS_HOLE(db->db_blkptr)) { zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL); need_wait = B_TRUE; } err = dbuf_read_impl(db, zio, flags); /* dbuf_read_impl has dropped db_mtx for us */ if (!err && prefetch) dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(db); DBUF_STAT_BUMP(hash_misses); if (!err && need_wait) err = zio_wait(zio); } else { /* * Another reader came in while the dbuf was in flight * between UNCACHED and CACHED. Either a writer will finish * writing the buffer (sending the dbuf to CACHED) or the * first reader's request will reach the read_done callback * and send the dbuf to CACHED. Otherwise, a failure * occurred and the dbuf went to UNCACHED. */ mutex_exit(&db->db_mtx); if (prefetch) dmu_zfetch(&dn->dn_zfetch, db->db_blkid, 1, B_TRUE); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(db); DBUF_STAT_BUMP(hash_misses); /* Skip the wait per the caller's request. */ mutex_enter(&db->db_mtx); if ((flags & DB_RF_NEVERWAIT) == 0) { while (db->db_state == DB_READ || db->db_state == DB_FILL) { ASSERT(db->db_state == DB_READ || (flags & DB_RF_HAVESTRUCT) == 0); DTRACE_PROBE2(blocked__read, dmu_buf_impl_t *, db, zio_t *, zio); cv_wait(&db->db_changed, &db->db_mtx); } if (db->db_state == DB_UNCACHED) err = SET_ERROR(EIO); } mutex_exit(&db->db_mtx); } return (err); } static void dbuf_noread(dmu_buf_impl_t *db) { ASSERT(!refcount_is_zero(&db->db_holds)); ASSERT(db->db_blkid != DMU_BONUS_BLKID); mutex_enter(&db->db_mtx); while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); if (db->db_state == DB_UNCACHED) { arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); spa_t *spa = db->db_objset->os_spa; ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); dbuf_set_data(db, arc_alloc_buf(spa, db, type, db->db.db_size)); db->db_state = DB_FILL; } else if (db->db_state == DB_NOFILL) { dbuf_clear_data(db); } else { ASSERT3U(db->db_state, ==, DB_CACHED); } mutex_exit(&db->db_mtx); } void dbuf_unoverride(dbuf_dirty_record_t *dr) { dmu_buf_impl_t *db = dr->dr_dbuf; blkptr_t *bp = &dr->dt.dl.dr_overridden_by; uint64_t txg = dr->dr_txg; ASSERT(MUTEX_HELD(&db->db_mtx)); /* * This assert is valid because dmu_sync() expects to be called by * a zilog's get_data while holding a range lock. This call only * comes from dbuf_dirty() callers who must also hold a range lock. */ ASSERT(dr->dt.dl.dr_override_state != DR_IN_DMU_SYNC); ASSERT(db->db_level == 0); if (db->db_blkid == DMU_BONUS_BLKID || dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN) return; ASSERT(db->db_data_pending != dr); /* free this block */ if (!BP_IS_HOLE(bp) && !dr->dt.dl.dr_nopwrite) zio_free(db->db_objset->os_spa, txg, bp); dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN; dr->dt.dl.dr_nopwrite = B_FALSE; - dr->dt.dl.dr_raw = B_FALSE; + dr->dt.dl.dr_has_raw_params = B_FALSE; /* * Release the already-written buffer, so we leave it in * a consistent dirty state. Note that all callers are * modifying the buffer, so they will immediately do * another (redundant) arc_release(). Therefore, leave * the buf thawed to save the effort of freezing & * immediately re-thawing it. */ arc_release(dr->dt.dl.dr_data, db); } /* * Evict (if its unreferenced) or clear (if its referenced) any level-0 * data blocks in the free range, so that any future readers will find * empty blocks. */ void dbuf_free_range(dnode_t *dn, uint64_t start_blkid, uint64_t end_blkid, dmu_tx_t *tx) { dmu_buf_impl_t *db_search; dmu_buf_impl_t *db, *db_next; uint64_t txg = tx->tx_txg; avl_index_t where; if (end_blkid > dn->dn_maxblkid && !(start_blkid == DMU_SPILL_BLKID || end_blkid == DMU_SPILL_BLKID)) end_blkid = dn->dn_maxblkid; dprintf_dnode(dn, "start=%llu end=%llu\n", start_blkid, end_blkid); db_search = kmem_alloc(sizeof (dmu_buf_impl_t), KM_SLEEP); db_search->db_level = 0; db_search->db_blkid = start_blkid; db_search->db_state = DB_SEARCH; mutex_enter(&dn->dn_dbufs_mtx); db = avl_find(&dn->dn_dbufs, db_search, &where); ASSERT3P(db, ==, NULL); db = avl_nearest(&dn->dn_dbufs, where, AVL_AFTER); for (; db != NULL; db = db_next) { db_next = AVL_NEXT(&dn->dn_dbufs, db); ASSERT(db->db_blkid != DMU_BONUS_BLKID); if (db->db_level != 0 || db->db_blkid > end_blkid) { break; } ASSERT3U(db->db_blkid, >=, start_blkid); /* found a level 0 buffer in the range */ mutex_enter(&db->db_mtx); if (dbuf_undirty(db, tx)) { /* mutex has been dropped and dbuf destroyed */ continue; } if (db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL || db->db_state == DB_EVICTING) { ASSERT(db->db.db_data == NULL); mutex_exit(&db->db_mtx); continue; } if (db->db_state == DB_READ || db->db_state == DB_FILL) { /* will be handled in dbuf_read_done or dbuf_rele */ db->db_freed_in_flight = TRUE; mutex_exit(&db->db_mtx); continue; } if (refcount_count(&db->db_holds) == 0) { ASSERT(db->db_buf); dbuf_destroy(db); continue; } /* The dbuf is referenced */ if (db->db_last_dirty != NULL) { dbuf_dirty_record_t *dr = db->db_last_dirty; if (dr->dr_txg == txg) { /* * This buffer is "in-use", re-adjust the file * size to reflect that this buffer may * contain new data when we sync. */ if (db->db_blkid != DMU_SPILL_BLKID && db->db_blkid > dn->dn_maxblkid) dn->dn_maxblkid = db->db_blkid; dbuf_unoverride(dr); } else { /* * This dbuf is not dirty in the open context. * Either uncache it (if its not referenced in * the open context) or reset its contents to * empty. */ dbuf_fix_old_data(db, txg); } } /* clear the contents if its cached */ if (db->db_state == DB_CACHED) { ASSERT(db->db.db_data != NULL); arc_release(db->db_buf, db); bzero(db->db.db_data, db->db.db_size); arc_buf_freeze(db->db_buf); } mutex_exit(&db->db_mtx); } kmem_free(db_search, sizeof (dmu_buf_impl_t)); mutex_exit(&dn->dn_dbufs_mtx); } void dbuf_new_size(dmu_buf_impl_t *db, int size, dmu_tx_t *tx) { arc_buf_t *buf, *obuf; int osize = db->db.db_size; arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); dnode_t *dn; ASSERT(db->db_blkid != DMU_BONUS_BLKID); DB_DNODE_ENTER(db); dn = DB_DNODE(db); /* XXX does *this* func really need the lock? */ ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock)); /* * This call to dmu_buf_will_dirty() with the dn_struct_rwlock held * is OK, because there can be no other references to the db * when we are changing its size, so no concurrent DB_FILL can * be happening. */ /* * XXX we should be doing a dbuf_read, checking the return * value and returning that up to our callers */ dmu_buf_will_dirty(&db->db, tx); /* create the data buffer for the new block */ buf = arc_alloc_buf(dn->dn_objset->os_spa, db, type, size); /* copy old block data to the new block */ obuf = db->db_buf; bcopy(obuf->b_data, buf->b_data, MIN(osize, size)); /* zero the remainder */ if (size > osize) bzero((uint8_t *)buf->b_data + osize, size - osize); mutex_enter(&db->db_mtx); dbuf_set_data(db, buf); arc_buf_destroy(obuf, db); db->db.db_size = size; if (db->db_level == 0) { ASSERT3U(db->db_last_dirty->dr_txg, ==, tx->tx_txg); db->db_last_dirty->dt.dl.dr_data = buf; } mutex_exit(&db->db_mtx); dmu_objset_willuse_space(dn->dn_objset, size - osize, tx); DB_DNODE_EXIT(db); } void dbuf_release_bp(dmu_buf_impl_t *db) { ASSERTV(objset_t *os = db->db_objset); ASSERT(dsl_pool_sync_context(dmu_objset_pool(os))); ASSERT(arc_released(os->os_phys_buf) || list_link_active(&os->os_dsl_dataset->ds_synced_link)); ASSERT(db->db_parent == NULL || arc_released(db->db_parent->db_buf)); (void) arc_release(db->db_buf, db); } /* * We already have a dirty record for this TXG, and we are being * dirtied again. */ static void dbuf_redirty(dbuf_dirty_record_t *dr) { dmu_buf_impl_t *db = dr->dr_dbuf; ASSERT(MUTEX_HELD(&db->db_mtx)); if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID) { /* * If this buffer has already been written out, * we now need to reset its state. */ dbuf_unoverride(dr); if (db->db.db_object != DMU_META_DNODE_OBJECT && db->db_state != DB_NOFILL) { /* Already released on initial dirty, so just thaw. */ ASSERT(arc_released(db->db_buf)); arc_buf_thaw(db->db_buf); } } } dbuf_dirty_record_t * dbuf_dirty(dmu_buf_impl_t *db, dmu_tx_t *tx) { dnode_t *dn; objset_t *os; dbuf_dirty_record_t **drp, *dr; int drop_struct_lock = FALSE; int txgoff = tx->tx_txg & TXG_MASK; ASSERT(tx->tx_txg != 0); ASSERT(!refcount_is_zero(&db->db_holds)); DMU_TX_DIRTY_BUF(tx, db); DB_DNODE_ENTER(db); dn = DB_DNODE(db); /* * Shouldn't dirty a regular buffer in syncing context. Private * objects may be dirtied in syncing context, but only if they * were already pre-dirtied in open context. */ #ifdef DEBUG if (dn->dn_objset->os_dsl_dataset != NULL) { rrw_enter(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG); } ASSERT(!dmu_tx_is_syncing(tx) || BP_IS_HOLE(dn->dn_objset->os_rootbp) || DMU_OBJECT_IS_SPECIAL(dn->dn_object) || dn->dn_objset->os_dsl_dataset == NULL); if (dn->dn_objset->os_dsl_dataset != NULL) rrw_exit(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, FTAG); #endif /* * We make this assert for private objects as well, but after we * check if we're already dirty. They are allowed to re-dirty * in syncing context. */ ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT || dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx == (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN)); mutex_enter(&db->db_mtx); /* * XXX make this true for indirects too? The problem is that * transactions created with dmu_tx_create_assigned() from * syncing context don't bother holding ahead. */ ASSERT(db->db_level != 0 || db->db_state == DB_CACHED || db->db_state == DB_FILL || db->db_state == DB_NOFILL); mutex_enter(&dn->dn_mtx); /* * Don't set dirtyctx to SYNC if we're just modifying this as we * initialize the objset. */ if (dn->dn_dirtyctx == DN_UNDIRTIED) { if (dn->dn_objset->os_dsl_dataset != NULL) { rrw_enter(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG); } if (!BP_IS_HOLE(dn->dn_objset->os_rootbp)) { dn->dn_dirtyctx = (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN); ASSERT(dn->dn_dirtyctx_firstset == NULL); dn->dn_dirtyctx_firstset = kmem_alloc(1, KM_SLEEP); } if (dn->dn_objset->os_dsl_dataset != NULL) { rrw_exit(&dn->dn_objset->os_dsl_dataset->ds_bp_rwlock, FTAG); } } if (tx->tx_txg > dn->dn_dirty_txg) dn->dn_dirty_txg = tx->tx_txg; mutex_exit(&dn->dn_mtx); if (db->db_blkid == DMU_SPILL_BLKID) dn->dn_have_spill = B_TRUE; /* * If this buffer is already dirty, we're done. */ drp = &db->db_last_dirty; ASSERT(*drp == NULL || (*drp)->dr_txg <= tx->tx_txg || db->db.db_object == DMU_META_DNODE_OBJECT); while ((dr = *drp) != NULL && dr->dr_txg > tx->tx_txg) drp = &dr->dr_next; if (dr && dr->dr_txg == tx->tx_txg) { DB_DNODE_EXIT(db); dbuf_redirty(dr); mutex_exit(&db->db_mtx); return (dr); } /* * Only valid if not already dirty. */ ASSERT(dn->dn_object == 0 || dn->dn_dirtyctx == DN_UNDIRTIED || dn->dn_dirtyctx == (dmu_tx_is_syncing(tx) ? DN_DIRTY_SYNC : DN_DIRTY_OPEN)); ASSERT3U(dn->dn_nlevels, >, db->db_level); /* * We should only be dirtying in syncing context if it's the * mos or we're initializing the os or it's a special object. * However, we are allowed to dirty in syncing context provided * we already dirtied it in open context. Hence we must make * this assertion only if we're not already dirty. */ os = dn->dn_objset; VERIFY3U(tx->tx_txg, <=, spa_final_dirty_txg(os->os_spa)); #ifdef DEBUG if (dn->dn_objset->os_dsl_dataset != NULL) rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_READER, FTAG); ASSERT(!dmu_tx_is_syncing(tx) || DMU_OBJECT_IS_SPECIAL(dn->dn_object) || os->os_dsl_dataset == NULL || BP_IS_HOLE(os->os_rootbp)); if (dn->dn_objset->os_dsl_dataset != NULL) rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG); #endif ASSERT(db->db.db_size != 0); dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size); if (db->db_blkid != DMU_BONUS_BLKID) { dmu_objset_willuse_space(os, db->db.db_size, tx); } /* * If this buffer is dirty in an old transaction group we need * to make a copy of it so that the changes we make in this * transaction group won't leak out when we sync the older txg. */ dr = kmem_zalloc(sizeof (dbuf_dirty_record_t), KM_SLEEP); list_link_init(&dr->dr_dirty_node); if (db->db_level == 0) { void *data_old = db->db_buf; if (db->db_state != DB_NOFILL) { if (db->db_blkid == DMU_BONUS_BLKID) { dbuf_fix_old_data(db, tx->tx_txg); data_old = db->db.db_data; } else if (db->db.db_object != DMU_META_DNODE_OBJECT) { /* * Release the data buffer from the cache so * that we can modify it without impacting * possible other users of this cached data * block. Note that indirect blocks and * private objects are not released until the * syncing state (since they are only modified * then). */ arc_release(db->db_buf, db); dbuf_fix_old_data(db, tx->tx_txg); data_old = db->db_buf; } ASSERT(data_old != NULL); } dr->dt.dl.dr_data = data_old; } else { mutex_init(&dr->dt.di.dr_mtx, NULL, MUTEX_NOLOCKDEP, NULL); list_create(&dr->dt.di.dr_children, sizeof (dbuf_dirty_record_t), offsetof(dbuf_dirty_record_t, dr_dirty_node)); } if (db->db_blkid != DMU_BONUS_BLKID && os->os_dsl_dataset != NULL) dr->dr_accounted = db->db.db_size; dr->dr_dbuf = db; dr->dr_txg = tx->tx_txg; dr->dr_next = *drp; *drp = dr; /* * We could have been freed_in_flight between the dbuf_noread * and dbuf_dirty. We win, as though the dbuf_noread() had * happened after the free. */ if (db->db_level == 0 && db->db_blkid != DMU_BONUS_BLKID && db->db_blkid != DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (dn->dn_free_ranges[txgoff] != NULL) { range_tree_clear(dn->dn_free_ranges[txgoff], db->db_blkid, 1); } mutex_exit(&dn->dn_mtx); db->db_freed_in_flight = FALSE; } /* * This buffer is now part of this txg */ dbuf_add_ref(db, (void *)(uintptr_t)tx->tx_txg); db->db_dirtycnt += 1; ASSERT3U(db->db_dirtycnt, <=, 3); mutex_exit(&db->db_mtx); if (db->db_blkid == DMU_BONUS_BLKID || db->db_blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&dn->dn_dirty_records[txgoff], dr); mutex_exit(&dn->dn_mtx); dnode_setdirty(dn, tx); DB_DNODE_EXIT(db); return (dr); } /* * The dn_struct_rwlock prevents db_blkptr from changing * due to a write from syncing context completing * while we are running, so we want to acquire it before * looking at db_blkptr. */ if (!RW_WRITE_HELD(&dn->dn_struct_rwlock)) { rw_enter(&dn->dn_struct_rwlock, RW_READER); drop_struct_lock = TRUE; } /* * We need to hold the dn_struct_rwlock to make this assertion, * because it protects dn_phys / dn_next_nlevels from changing. */ ASSERT((dn->dn_phys->dn_nlevels == 0 && db->db_level == 0) || dn->dn_phys->dn_nlevels > db->db_level || dn->dn_next_nlevels[txgoff] > db->db_level || dn->dn_next_nlevels[(tx->tx_txg-1) & TXG_MASK] > db->db_level || dn->dn_next_nlevels[(tx->tx_txg-2) & TXG_MASK] > db->db_level); /* * If we are overwriting a dedup BP, then unless it is snapshotted, * when we get to syncing context we will need to decrement its * refcount in the DDT. Prefetch the relevant DDT block so that * syncing context won't have to wait for the i/o. */ ddt_prefetch(os->os_spa, db->db_blkptr); if (db->db_level == 0) { dnode_new_blkid(dn, db->db_blkid, tx, drop_struct_lock); ASSERT(dn->dn_maxblkid >= db->db_blkid); } if (db->db_level+1 < dn->dn_nlevels) { dmu_buf_impl_t *parent = db->db_parent; dbuf_dirty_record_t *di; int parent_held = FALSE; if (db->db_parent == NULL || db->db_parent == dn->dn_dbuf) { int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; parent = dbuf_hold_level(dn, db->db_level+1, db->db_blkid >> epbs, FTAG); ASSERT(parent != NULL); parent_held = TRUE; } if (drop_struct_lock) rw_exit(&dn->dn_struct_rwlock); ASSERT3U(db->db_level+1, ==, parent->db_level); di = dbuf_dirty(parent, tx); if (parent_held) dbuf_rele(parent, FTAG); mutex_enter(&db->db_mtx); /* * Since we've dropped the mutex, it's possible that * dbuf_undirty() might have changed this out from under us. */ if (db->db_last_dirty == dr || dn->dn_object == DMU_META_DNODE_OBJECT) { mutex_enter(&di->dt.di.dr_mtx); ASSERT3U(di->dr_txg, ==, tx->tx_txg); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&di->dt.di.dr_children, dr); mutex_exit(&di->dt.di.dr_mtx); dr->dr_parent = di; } mutex_exit(&db->db_mtx); } else { ASSERT(db->db_level+1 == dn->dn_nlevels); ASSERT(db->db_blkid < dn->dn_nblkptr); ASSERT(db->db_parent == NULL || db->db_parent == dn->dn_dbuf); mutex_enter(&dn->dn_mtx); ASSERT(!list_link_active(&dr->dr_dirty_node)); list_insert_tail(&dn->dn_dirty_records[txgoff], dr); mutex_exit(&dn->dn_mtx); if (drop_struct_lock) rw_exit(&dn->dn_struct_rwlock); } dnode_setdirty(dn, tx); DB_DNODE_EXIT(db); return (dr); } /* * Undirty a buffer in the transaction group referenced by the given * transaction. Return whether this evicted the dbuf. */ static boolean_t dbuf_undirty(dmu_buf_impl_t *db, dmu_tx_t *tx) { dnode_t *dn; uint64_t txg = tx->tx_txg; dbuf_dirty_record_t *dr, **drp; ASSERT(txg != 0); /* * Due to our use of dn_nlevels below, this can only be called * in open context, unless we are operating on the MOS. * From syncing context, dn_nlevels may be different from the * dn_nlevels used when dbuf was dirtied. */ ASSERT(db->db_objset == dmu_objset_pool(db->db_objset)->dp_meta_objset || txg != spa_syncing_txg(dmu_objset_spa(db->db_objset))); ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT0(db->db_level); ASSERT(MUTEX_HELD(&db->db_mtx)); /* * If this buffer is not dirty, we're done. */ for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) if (dr->dr_txg <= txg) break; if (dr == NULL || dr->dr_txg < txg) return (B_FALSE); ASSERT(dr->dr_txg == txg); ASSERT(dr->dr_dbuf == db); DB_DNODE_ENTER(db); dn = DB_DNODE(db); dprintf_dbuf(db, "size=%llx\n", (u_longlong_t)db->db.db_size); ASSERT(db->db.db_size != 0); dsl_pool_undirty_space(dmu_objset_pool(dn->dn_objset), dr->dr_accounted, txg); *drp = dr->dr_next; /* * Note that there are three places in dbuf_dirty() * where this dirty record may be put on a list. * Make sure to do a list_remove corresponding to * every one of those list_insert calls. */ if (dr->dr_parent) { mutex_enter(&dr->dr_parent->dt.di.dr_mtx); list_remove(&dr->dr_parent->dt.di.dr_children, dr); mutex_exit(&dr->dr_parent->dt.di.dr_mtx); } else if (db->db_blkid == DMU_SPILL_BLKID || db->db_level + 1 == dn->dn_nlevels) { ASSERT(db->db_blkptr == NULL || db->db_parent == dn->dn_dbuf); mutex_enter(&dn->dn_mtx); list_remove(&dn->dn_dirty_records[txg & TXG_MASK], dr); mutex_exit(&dn->dn_mtx); } DB_DNODE_EXIT(db); if (db->db_state != DB_NOFILL) { dbuf_unoverride(dr); ASSERT(db->db_buf != NULL); ASSERT(dr->dt.dl.dr_data != NULL); if (dr->dt.dl.dr_data != db->db_buf) arc_buf_destroy(dr->dt.dl.dr_data, db); } kmem_free(dr, sizeof (dbuf_dirty_record_t)); ASSERT(db->db_dirtycnt > 0); db->db_dirtycnt -= 1; if (refcount_remove(&db->db_holds, (void *)(uintptr_t)txg) == 0) { ASSERT(db->db_state == DB_NOFILL || arc_released(db->db_buf)); dbuf_destroy(db); return (B_TRUE); } return (B_FALSE); } static void dmu_buf_will_dirty_impl(dmu_buf_t *db_fake, int flags, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; ASSERT(tx->tx_txg != 0); ASSERT(!refcount_is_zero(&db->db_holds)); /* * Quick check for dirtyness. For already dirty blocks, this * reduces runtime of this function by >90%, and overall performance * by 50% for some workloads (e.g. file deletion with indirect blocks * cached). */ mutex_enter(&db->db_mtx); dbuf_dirty_record_t *dr; for (dr = db->db_last_dirty; dr != NULL && dr->dr_txg >= tx->tx_txg; dr = dr->dr_next) { /* * It's possible that it is already dirty but not cached, * because there are some calls to dbuf_dirty() that don't * go through dmu_buf_will_dirty(). */ if (dr->dr_txg == tx->tx_txg && db->db_state == DB_CACHED) { /* This dbuf is already dirty and cached. */ dbuf_redirty(dr); mutex_exit(&db->db_mtx); return; } } mutex_exit(&db->db_mtx); DB_DNODE_ENTER(db); if (RW_WRITE_HELD(&DB_DNODE(db)->dn_struct_rwlock)) flags |= DB_RF_HAVESTRUCT; DB_DNODE_EXIT(db); (void) dbuf_read(db, NULL, flags); (void) dbuf_dirty(db, tx); } void dmu_buf_will_dirty(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_will_dirty_impl(db_fake, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH, tx); } void dmu_buf_will_not_fill(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; db->db_state = DB_NOFILL; dmu_buf_will_fill(db_fake, tx); } void dmu_buf_will_fill(dmu_buf_t *db_fake, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(tx->tx_txg != 0); ASSERT(db->db_level == 0); ASSERT(!refcount_is_zero(&db->db_holds)); ASSERT(db->db.db_object != DMU_META_DNODE_OBJECT || dmu_tx_private_ok(tx)); dbuf_noread(db); (void) dbuf_dirty(db, tx); } /* * This function is effectively the same as dmu_buf_will_dirty(), but - * indicates the caller expects raw encrypted data in the db. It will - * also set the raw flag on the created dirty record. + * indicates the caller expects raw encrypted data in the db, and provides + * the crypt params (byteorder, salt, iv, mac) which should be stored in the + * blkptr_t when this dbuf is written. This is only used for blocks of + * dnodes, during raw receive. */ void -dmu_buf_will_change_crypt_params(dmu_buf_t *db_fake, dmu_tx_t *tx) +dmu_buf_set_crypt_params(dmu_buf_t *db_fake, boolean_t byteorder, + const uint8_t *salt, const uint8_t *iv, const uint8_t *mac, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dbuf_dirty_record_t *dr; + /* + * dr_has_raw_params is only processed for blocks of dnodes + * (see dbuf_sync_dnode_leaf_crypt()). + */ + ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT); + ASSERT3U(db->db_level, ==, 0); + ASSERT(db->db_objset->os_raw_receive); + dmu_buf_will_dirty_impl(db_fake, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH | DB_RF_NO_DECRYPT, tx); dr = db->db_last_dirty; while (dr != NULL && dr->dr_txg > tx->tx_txg) dr = dr->dr_next; ASSERT3P(dr, !=, NULL); ASSERT3U(dr->dr_txg, ==, tx->tx_txg); - dr->dt.dl.dr_raw = B_TRUE; - db->db_objset->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_TRUE; + + dr->dt.dl.dr_has_raw_params = B_TRUE; + dr->dt.dl.dr_byteorder = byteorder; + bcopy(salt, dr->dt.dl.dr_salt, ZIO_DATA_SALT_LEN); + bcopy(iv, dr->dt.dl.dr_iv, ZIO_DATA_IV_LEN); + bcopy(mac, dr->dt.dl.dr_mac, ZIO_DATA_MAC_LEN); } #pragma weak dmu_buf_fill_done = dbuf_fill_done /* ARGSUSED */ void dbuf_fill_done(dmu_buf_impl_t *db, dmu_tx_t *tx) { mutex_enter(&db->db_mtx); DBUF_VERIFY(db); if (db->db_state == DB_FILL) { if (db->db_level == 0 && db->db_freed_in_flight) { ASSERT(db->db_blkid != DMU_BONUS_BLKID); /* we were freed while filling */ /* XXX dbuf_undirty? */ bzero(db->db.db_data, db->db.db_size); db->db_freed_in_flight = FALSE; } db->db_state = DB_CACHED; cv_broadcast(&db->db_changed); } mutex_exit(&db->db_mtx); } void dmu_buf_write_embedded(dmu_buf_t *dbuf, void *data, bp_embedded_type_t etype, enum zio_compress comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbuf; struct dirty_leaf *dl; dmu_object_type_t type; if (etype == BP_EMBEDDED_TYPE_DATA) { ASSERT(spa_feature_is_active(dmu_objset_spa(db->db_objset), SPA_FEATURE_EMBEDDED_DATA)); } DB_DNODE_ENTER(db); type = DB_DNODE(db)->dn_type; DB_DNODE_EXIT(db); ASSERT0(db->db_level); ASSERT(db->db_blkid != DMU_BONUS_BLKID); dmu_buf_will_not_fill(dbuf, tx); ASSERT3U(db->db_last_dirty->dr_txg, ==, tx->tx_txg); dl = &db->db_last_dirty->dt.dl; encode_embedded_bp_compressed(&dl->dr_overridden_by, data, comp, uncompressed_size, compressed_size); BPE_SET_ETYPE(&dl->dr_overridden_by, etype); BP_SET_TYPE(&dl->dr_overridden_by, type); BP_SET_LEVEL(&dl->dr_overridden_by, 0); BP_SET_BYTEORDER(&dl->dr_overridden_by, byteorder); dl->dr_override_state = DR_OVERRIDDEN; dl->dr_overridden_by.blk_birth = db->db_last_dirty->dr_txg; } /* * Directly assign a provided arc buf to a given dbuf if it's not referenced * by anybody except our caller. Otherwise copy arcbuf's contents to dbuf. */ void dbuf_assign_arcbuf(dmu_buf_impl_t *db, arc_buf_t *buf, dmu_tx_t *tx) { ASSERT(!refcount_is_zero(&db->db_holds)); ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(db->db_level == 0); ASSERT3U(dbuf_is_metadata(db), ==, arc_is_metadata(buf)); ASSERT(buf != NULL); ASSERT(arc_buf_lsize(buf) == db->db.db_size); ASSERT(tx->tx_txg != 0); arc_return_buf(buf, db); ASSERT(arc_released(buf)); mutex_enter(&db->db_mtx); while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); ASSERT(db->db_state == DB_CACHED || db->db_state == DB_UNCACHED); if (db->db_state == DB_CACHED && refcount_count(&db->db_holds) - 1 > db->db_dirtycnt) { /* * In practice, we will never have a case where we have an * encrypted arc buffer while additional holds exist on the * dbuf. We don't handle this here so we simply assert that * fact instead. */ ASSERT(!arc_is_encrypted(buf)); mutex_exit(&db->db_mtx); (void) dbuf_dirty(db, tx); bcopy(buf->b_data, db->db.db_data, db->db.db_size); arc_buf_destroy(buf, db); xuio_stat_wbuf_copied(); return; } xuio_stat_wbuf_nocopy(); if (db->db_state == DB_CACHED) { dbuf_dirty_record_t *dr = db->db_last_dirty; ASSERT(db->db_buf != NULL); if (dr != NULL && dr->dr_txg == tx->tx_txg) { ASSERT(dr->dt.dl.dr_data == db->db_buf); - IMPLY(arc_is_encrypted(buf), dr->dt.dl.dr_raw); if (!arc_released(db->db_buf)) { ASSERT(dr->dt.dl.dr_override_state == DR_OVERRIDDEN); arc_release(db->db_buf, db); } dr->dt.dl.dr_data = buf; arc_buf_destroy(db->db_buf, db); } else if (dr == NULL || dr->dt.dl.dr_data != db->db_buf) { arc_release(db->db_buf, db); arc_buf_destroy(db->db_buf, db); } db->db_buf = NULL; } ASSERT(db->db_buf == NULL); dbuf_set_data(db, buf); db->db_state = DB_FILL; mutex_exit(&db->db_mtx); (void) dbuf_dirty(db, tx); dmu_buf_fill_done(&db->db, tx); } void dbuf_destroy(dmu_buf_impl_t *db) { dnode_t *dn; dmu_buf_impl_t *parent = db->db_parent; dmu_buf_impl_t *dndb; ASSERT(MUTEX_HELD(&db->db_mtx)); ASSERT(refcount_is_zero(&db->db_holds)); if (db->db_buf != NULL) { arc_buf_destroy(db->db_buf, db); db->db_buf = NULL; } if (db->db_blkid == DMU_BONUS_BLKID) { int slots = DB_DNODE(db)->dn_num_slots; int bonuslen = DN_SLOTS_TO_BONUSLEN(slots); if (db->db.db_data != NULL) { kmem_free(db->db.db_data, bonuslen); arc_space_return(bonuslen, ARC_SPACE_BONUS); db->db_state = DB_UNCACHED; } } dbuf_clear_data(db); if (multilist_link_active(&db->db_cache_link)) { multilist_remove(dbuf_cache, db); (void) refcount_remove_many(&dbuf_cache_size, db->db.db_size, db); DBUF_STAT_BUMPDOWN(cache_levels[db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[db->db_level], db->db.db_size); } ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); ASSERT(db->db_data_pending == NULL); db->db_state = DB_EVICTING; db->db_blkptr = NULL; /* * Now that db_state is DB_EVICTING, nobody else can find this via * the hash table. We can now drop db_mtx, which allows us to * acquire the dn_dbufs_mtx. */ mutex_exit(&db->db_mtx); DB_DNODE_ENTER(db); dn = DB_DNODE(db); dndb = dn->dn_dbuf; if (db->db_blkid != DMU_BONUS_BLKID) { boolean_t needlock = !MUTEX_HELD(&dn->dn_dbufs_mtx); if (needlock) mutex_enter(&dn->dn_dbufs_mtx); avl_remove(&dn->dn_dbufs, db); atomic_dec_32(&dn->dn_dbufs_count); membar_producer(); DB_DNODE_EXIT(db); if (needlock) mutex_exit(&dn->dn_dbufs_mtx); /* * Decrementing the dbuf count means that the hold corresponding * to the removed dbuf is no longer discounted in dnode_move(), * so the dnode cannot be moved until after we release the hold. * The membar_producer() ensures visibility of the decremented * value in dnode_move(), since DB_DNODE_EXIT doesn't actually * release any lock. */ dnode_rele(dn, db); db->db_dnode_handle = NULL; dbuf_hash_remove(db); } else { DB_DNODE_EXIT(db); } ASSERT(refcount_is_zero(&db->db_holds)); db->db_parent = NULL; ASSERT(db->db_buf == NULL); ASSERT(db->db.db_data == NULL); ASSERT(db->db_hash_next == NULL); ASSERT(db->db_blkptr == NULL); ASSERT(db->db_data_pending == NULL); ASSERT(!multilist_link_active(&db->db_cache_link)); kmem_cache_free(dbuf_kmem_cache, db); arc_space_return(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); /* * If this dbuf is referenced from an indirect dbuf, * decrement the ref count on the indirect dbuf. */ if (parent && parent != dndb) dbuf_rele(parent, db); } /* * Note: While bpp will always be updated if the function returns success, * parentp will not be updated if the dnode does not have dn_dbuf filled in; * this happens when the dnode is the meta-dnode, or {user|group|project}used * object. */ __attribute__((always_inline)) static inline int dbuf_findbp(dnode_t *dn, int level, uint64_t blkid, int fail_sparse, dmu_buf_impl_t **parentp, blkptr_t **bpp, struct dbuf_hold_impl_data *dh) { *parentp = NULL; *bpp = NULL; ASSERT(blkid != DMU_BONUS_BLKID); if (blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (dn->dn_have_spill && (dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) *bpp = DN_SPILL_BLKPTR(dn->dn_phys); else *bpp = NULL; dbuf_add_ref(dn->dn_dbuf, NULL); *parentp = dn->dn_dbuf; mutex_exit(&dn->dn_mtx); return (0); } int nlevels = (dn->dn_phys->dn_nlevels == 0) ? 1 : dn->dn_phys->dn_nlevels; int epbs = dn->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT3U(level * epbs, <, 64); ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); /* * This assertion shouldn't trip as long as the max indirect block size * is less than 1M. The reason for this is that up to that point, * the number of levels required to address an entire object with blocks * of size SPA_MINBLOCKSIZE satisfies nlevels * epbs + 1 <= 64. In * other words, if N * epbs + 1 > 64, then if (N-1) * epbs + 1 > 55 * (i.e. we can address the entire object), objects will all use at most * N-1 levels and the assertion won't overflow. However, once epbs is * 13, 4 * 13 + 1 = 53, but 5 * 13 + 1 = 66. Then, 4 levels will not be * enough to address an entire object, so objects will have 5 levels, * but then this assertion will overflow. * * All this is to say that if we ever increase DN_MAX_INDBLKSHIFT, we * need to redo this logic to handle overflows. */ ASSERT(level >= nlevels || ((nlevels - level - 1) * epbs) + highbit64(dn->dn_phys->dn_nblkptr) <= 64); if (level >= nlevels || blkid >= ((uint64_t)dn->dn_phys->dn_nblkptr << ((nlevels - level - 1) * epbs)) || (fail_sparse && blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs)))) { /* the buffer has no parent yet */ return (SET_ERROR(ENOENT)); } else if (level < nlevels-1) { /* this block is referenced from an indirect block */ int err; if (dh == NULL) { err = dbuf_hold_impl(dn, level+1, blkid >> epbs, fail_sparse, FALSE, NULL, parentp); } else { __dbuf_hold_impl_init(dh + 1, dn, dh->dh_level + 1, blkid >> epbs, fail_sparse, FALSE, NULL, parentp, dh->dh_depth + 1); err = __dbuf_hold_impl(dh + 1); } if (err) return (err); err = dbuf_read(*parentp, NULL, (DB_RF_HAVESTRUCT | DB_RF_NOPREFETCH | DB_RF_CANFAIL)); if (err) { dbuf_rele(*parentp, NULL); *parentp = NULL; return (err); } *bpp = ((blkptr_t *)(*parentp)->db.db_data) + (blkid & ((1ULL << epbs) - 1)); if (blkid > (dn->dn_phys->dn_maxblkid >> (level * epbs))) ASSERT(BP_IS_HOLE(*bpp)); return (0); } else { /* the block is referenced from the dnode */ ASSERT3U(level, ==, nlevels-1); ASSERT(dn->dn_phys->dn_nblkptr == 0 || blkid < dn->dn_phys->dn_nblkptr); if (dn->dn_dbuf) { dbuf_add_ref(dn->dn_dbuf, NULL); *parentp = dn->dn_dbuf; } *bpp = &dn->dn_phys->dn_blkptr[blkid]; return (0); } } static dmu_buf_impl_t * dbuf_create(dnode_t *dn, uint8_t level, uint64_t blkid, dmu_buf_impl_t *parent, blkptr_t *blkptr) { objset_t *os = dn->dn_objset; dmu_buf_impl_t *db, *odb; ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); ASSERT(dn->dn_type != DMU_OT_NONE); db = kmem_cache_alloc(dbuf_kmem_cache, KM_SLEEP); db->db_objset = os; db->db.db_object = dn->dn_object; db->db_level = level; db->db_blkid = blkid; db->db_last_dirty = NULL; db->db_dirtycnt = 0; db->db_dnode_handle = dn->dn_handle; db->db_parent = parent; db->db_blkptr = blkptr; db->db_user = NULL; db->db_user_immediate_evict = FALSE; db->db_freed_in_flight = FALSE; db->db_pending_evict = FALSE; if (blkid == DMU_BONUS_BLKID) { ASSERT3P(parent, ==, dn->dn_dbuf); db->db.db_size = DN_SLOTS_TO_BONUSLEN(dn->dn_num_slots) - (dn->dn_nblkptr-1) * sizeof (blkptr_t); ASSERT3U(db->db.db_size, >=, dn->dn_bonuslen); db->db.db_offset = DMU_BONUS_BLKID; db->db_state = DB_UNCACHED; /* the bonus dbuf is not placed in the hash table */ arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); return (db); } else if (blkid == DMU_SPILL_BLKID) { db->db.db_size = (blkptr != NULL) ? BP_GET_LSIZE(blkptr) : SPA_MINBLOCKSIZE; db->db.db_offset = 0; } else { int blocksize = db->db_level ? 1 << dn->dn_indblkshift : dn->dn_datablksz; db->db.db_size = blocksize; db->db.db_offset = db->db_blkid * blocksize; } /* * Hold the dn_dbufs_mtx while we get the new dbuf * in the hash table *and* added to the dbufs list. * This prevents a possible deadlock with someone * trying to look up this dbuf before its added to the * dn_dbufs list. */ mutex_enter(&dn->dn_dbufs_mtx); db->db_state = DB_EVICTING; if ((odb = dbuf_hash_insert(db)) != NULL) { /* someone else inserted it first */ kmem_cache_free(dbuf_kmem_cache, db); mutex_exit(&dn->dn_dbufs_mtx); DBUF_STAT_BUMP(hash_insert_race); return (odb); } avl_add(&dn->dn_dbufs, db); db->db_state = DB_UNCACHED; mutex_exit(&dn->dn_dbufs_mtx); arc_space_consume(sizeof (dmu_buf_impl_t), ARC_SPACE_DBUF); if (parent && parent != dn->dn_dbuf) dbuf_add_ref(parent, db); ASSERT(dn->dn_object == DMU_META_DNODE_OBJECT || refcount_count(&dn->dn_holds) > 0); (void) refcount_add(&dn->dn_holds, db); atomic_inc_32(&dn->dn_dbufs_count); dprintf_dbuf(db, "db=%p\n", db); return (db); } typedef struct dbuf_prefetch_arg { spa_t *dpa_spa; /* The spa to issue the prefetch in. */ zbookmark_phys_t dpa_zb; /* The target block to prefetch. */ int dpa_epbs; /* Entries (blkptr_t's) Per Block Shift. */ int dpa_curlevel; /* The current level that we're reading */ dnode_t *dpa_dnode; /* The dnode associated with the prefetch */ zio_priority_t dpa_prio; /* The priority I/Os should be issued at. */ zio_t *dpa_zio; /* The parent zio_t for all prefetches. */ arc_flags_t dpa_aflags; /* Flags to pass to the final prefetch. */ } dbuf_prefetch_arg_t; /* * Actually issue the prefetch read for the block given. */ static void dbuf_issue_final_prefetch(dbuf_prefetch_arg_t *dpa, blkptr_t *bp) { if (BP_IS_HOLE(bp) || BP_IS_EMBEDDED(bp)) return; int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE; arc_flags_t aflags = dpa->dpa_aflags | ARC_FLAG_NOWAIT | ARC_FLAG_PREFETCH; /* dnodes are always read as raw and then converted later */ if (BP_GET_TYPE(bp) == DMU_OT_DNODE && BP_IS_PROTECTED(bp) && dpa->dpa_curlevel == 0) zio_flags |= ZIO_FLAG_RAW; ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp)); ASSERT3U(dpa->dpa_curlevel, ==, dpa->dpa_zb.zb_level); ASSERT(dpa->dpa_zio != NULL); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp, NULL, NULL, dpa->dpa_prio, zio_flags, &aflags, &dpa->dpa_zb); } /* * Called when an indirect block above our prefetch target is read in. This * will either read in the next indirect block down the tree or issue the actual * prefetch if the next block down is our target. */ static void dbuf_prefetch_indirect_done(zio_t *zio, const zbookmark_phys_t *zb, const blkptr_t *iobp, arc_buf_t *abuf, void *private) { dbuf_prefetch_arg_t *dpa = private; ASSERT3S(dpa->dpa_zb.zb_level, <, dpa->dpa_curlevel); ASSERT3S(dpa->dpa_curlevel, >, 0); /* * The dpa_dnode is only valid if we are called with a NULL * zio. This indicates that the arc_read() returned without * first calling zio_read() to issue a physical read. Once * a physical read is made the dpa_dnode must be invalidated * as the locks guarding it may have been dropped. If the * dpa_dnode is still valid, then we want to add it to the dbuf * cache. To do so, we must hold the dbuf associated with the block * we just prefetched, read its contents so that we associate it * with an arc_buf_t, and then release it. */ if (zio != NULL) { ASSERT3S(BP_GET_LEVEL(zio->io_bp), ==, dpa->dpa_curlevel); if (zio->io_flags & ZIO_FLAG_RAW_COMPRESS) { ASSERT3U(BP_GET_PSIZE(zio->io_bp), ==, zio->io_size); } else { ASSERT3U(BP_GET_LSIZE(zio->io_bp), ==, zio->io_size); } ASSERT3P(zio->io_spa, ==, dpa->dpa_spa); dpa->dpa_dnode = NULL; } else if (dpa->dpa_dnode != NULL) { uint64_t curblkid = dpa->dpa_zb.zb_blkid >> (dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level)); dmu_buf_impl_t *db = dbuf_hold_level(dpa->dpa_dnode, dpa->dpa_curlevel, curblkid, FTAG); (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED | DB_RF_NOPREFETCH | DB_RF_HAVESTRUCT); dbuf_rele(db, FTAG); } if (abuf == NULL) { kmem_free(dpa, sizeof (*dpa)); return; } dpa->dpa_curlevel--; uint64_t nextblkid = dpa->dpa_zb.zb_blkid >> (dpa->dpa_epbs * (dpa->dpa_curlevel - dpa->dpa_zb.zb_level)); blkptr_t *bp = ((blkptr_t *)abuf->b_data) + P2PHASE(nextblkid, 1ULL << dpa->dpa_epbs); if (BP_IS_HOLE(bp)) { kmem_free(dpa, sizeof (*dpa)); } else if (dpa->dpa_curlevel == dpa->dpa_zb.zb_level) { ASSERT3U(nextblkid, ==, dpa->dpa_zb.zb_blkid); dbuf_issue_final_prefetch(dpa, bp); kmem_free(dpa, sizeof (*dpa)); } else { arc_flags_t iter_aflags = ARC_FLAG_NOWAIT; zbookmark_phys_t zb; /* flag if L2ARC eligible, l2arc_noprefetch then decides */ if (dpa->dpa_aflags & ARC_FLAG_L2CACHE) iter_aflags |= ARC_FLAG_L2CACHE; ASSERT3U(dpa->dpa_curlevel, ==, BP_GET_LEVEL(bp)); SET_BOOKMARK(&zb, dpa->dpa_zb.zb_objset, dpa->dpa_zb.zb_object, dpa->dpa_curlevel, nextblkid); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, bp, dbuf_prefetch_indirect_done, dpa, dpa->dpa_prio, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &iter_aflags, &zb); } arc_buf_destroy(abuf, private); } /* * Issue prefetch reads for the given block on the given level. If the indirect * blocks above that block are not in memory, we will read them in * asynchronously. As a result, this call never blocks waiting for a read to * complete. Note that the prefetch might fail if the dataset is encrypted and * the encryption key is unmapped before the IO completes. */ void dbuf_prefetch(dnode_t *dn, int64_t level, uint64_t blkid, zio_priority_t prio, arc_flags_t aflags) { blkptr_t bp; int epbs, nlevels, curlevel; uint64_t curblkid; ASSERT(blkid != DMU_BONUS_BLKID); ASSERT(RW_LOCK_HELD(&dn->dn_struct_rwlock)); if (blkid > dn->dn_maxblkid) return; if (dnode_block_freed(dn, blkid)) return; /* * This dnode hasn't been written to disk yet, so there's nothing to * prefetch. */ nlevels = dn->dn_phys->dn_nlevels; if (level >= nlevels || dn->dn_phys->dn_nblkptr == 0) return; epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; if (dn->dn_phys->dn_maxblkid < blkid << (epbs * level)) return; dmu_buf_impl_t *db = dbuf_find(dn->dn_objset, dn->dn_object, level, blkid); if (db != NULL) { mutex_exit(&db->db_mtx); /* * This dbuf already exists. It is either CACHED, or * (we assume) about to be read or filled. */ return; } /* * Find the closest ancestor (indirect block) of the target block * that is present in the cache. In this indirect block, we will * find the bp that is at curlevel, curblkid. */ curlevel = level; curblkid = blkid; while (curlevel < nlevels - 1) { int parent_level = curlevel + 1; uint64_t parent_blkid = curblkid >> epbs; dmu_buf_impl_t *db; if (dbuf_hold_impl(dn, parent_level, parent_blkid, FALSE, TRUE, FTAG, &db) == 0) { blkptr_t *bpp = db->db_buf->b_data; bp = bpp[P2PHASE(curblkid, 1 << epbs)]; dbuf_rele(db, FTAG); break; } curlevel = parent_level; curblkid = parent_blkid; } if (curlevel == nlevels - 1) { /* No cached indirect blocks found. */ ASSERT3U(curblkid, <, dn->dn_phys->dn_nblkptr); bp = dn->dn_phys->dn_blkptr[curblkid]; } if (BP_IS_HOLE(&bp)) return; ASSERT3U(curlevel, ==, BP_GET_LEVEL(&bp)); zio_t *pio = zio_root(dmu_objset_spa(dn->dn_objset), NULL, NULL, ZIO_FLAG_CANFAIL); dbuf_prefetch_arg_t *dpa = kmem_zalloc(sizeof (*dpa), KM_SLEEP); dsl_dataset_t *ds = dn->dn_objset->os_dsl_dataset; SET_BOOKMARK(&dpa->dpa_zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET, dn->dn_object, level, blkid); dpa->dpa_curlevel = curlevel; dpa->dpa_prio = prio; dpa->dpa_aflags = aflags; dpa->dpa_spa = dn->dn_objset->os_spa; dpa->dpa_dnode = dn; dpa->dpa_epbs = epbs; dpa->dpa_zio = pio; /* flag if L2ARC eligible, l2arc_noprefetch then decides */ if (DNODE_LEVEL_IS_L2CACHEABLE(dn, level)) dpa->dpa_aflags |= ARC_FLAG_L2CACHE; /* * If we have the indirect just above us, no need to do the asynchronous * prefetch chain; we'll just run the last step ourselves. If we're at * a higher level, though, we want to issue the prefetches for all the * indirect blocks asynchronously, so we can go on with whatever we were * doing. */ if (curlevel == level) { ASSERT3U(curblkid, ==, blkid); dbuf_issue_final_prefetch(dpa, &bp); kmem_free(dpa, sizeof (*dpa)); } else { arc_flags_t iter_aflags = ARC_FLAG_NOWAIT; zbookmark_phys_t zb; /* flag if L2ARC eligible, l2arc_noprefetch then decides */ if (DNODE_LEVEL_IS_L2CACHEABLE(dn, level)) iter_aflags |= ARC_FLAG_L2CACHE; SET_BOOKMARK(&zb, ds != NULL ? ds->ds_object : DMU_META_OBJSET, dn->dn_object, curlevel, curblkid); (void) arc_read(dpa->dpa_zio, dpa->dpa_spa, &bp, dbuf_prefetch_indirect_done, dpa, prio, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE, &iter_aflags, &zb); } /* * We use pio here instead of dpa_zio since it's possible that * dpa may have already been freed. */ zio_nowait(pio); } #define DBUF_HOLD_IMPL_MAX_DEPTH 20 /* * Helper function for __dbuf_hold_impl() to copy a buffer. Handles * the case of encrypted, compressed and uncompressed buffers by * allocating the new buffer, respectively, with arc_alloc_raw_buf(), * arc_alloc_compressed_buf() or arc_alloc_buf().* * * NOTE: Declared noinline to avoid stack bloat in __dbuf_hold_impl(). */ noinline static void dbuf_hold_copy(struct dbuf_hold_impl_data *dh) { dnode_t *dn = dh->dh_dn; dmu_buf_impl_t *db = dh->dh_db; dbuf_dirty_record_t *dr = dh->dh_dr; arc_buf_t *data = dr->dt.dl.dr_data; enum zio_compress compress_type = arc_get_compression(data); if (arc_is_encrypted(data)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(data, &byteorder, salt, iv, mac); dbuf_set_data(db, arc_alloc_raw_buf(dn->dn_objset->os_spa, db, dmu_objset_id(dn->dn_objset), byteorder, salt, iv, mac, dn->dn_type, arc_buf_size(data), arc_buf_lsize(data), compress_type)); } else if (compress_type != ZIO_COMPRESS_OFF) { dbuf_set_data(db, arc_alloc_compressed_buf( dn->dn_objset->os_spa, db, arc_buf_size(data), arc_buf_lsize(data), compress_type)); } else { dbuf_set_data(db, arc_alloc_buf(dn->dn_objset->os_spa, db, DBUF_GET_BUFC_TYPE(db), db->db.db_size)); } bcopy(data->b_data, db->db.db_data, arc_buf_size(data)); } /* * Returns with db_holds incremented, and db_mtx not held. * Note: dn_struct_rwlock must be held. */ static int __dbuf_hold_impl(struct dbuf_hold_impl_data *dh) { ASSERT3S(dh->dh_depth, <, DBUF_HOLD_IMPL_MAX_DEPTH); dh->dh_parent = NULL; ASSERT(dh->dh_blkid != DMU_BONUS_BLKID); ASSERT(RW_LOCK_HELD(&dh->dh_dn->dn_struct_rwlock)); ASSERT3U(dh->dh_dn->dn_nlevels, >, dh->dh_level); *(dh->dh_dbp) = NULL; /* dbuf_find() returns with db_mtx held */ dh->dh_db = dbuf_find(dh->dh_dn->dn_objset, dh->dh_dn->dn_object, dh->dh_level, dh->dh_blkid); if (dh->dh_db == NULL) { dh->dh_bp = NULL; if (dh->dh_fail_uncached) return (SET_ERROR(ENOENT)); ASSERT3P(dh->dh_parent, ==, NULL); dh->dh_err = dbuf_findbp(dh->dh_dn, dh->dh_level, dh->dh_blkid, dh->dh_fail_sparse, &dh->dh_parent, &dh->dh_bp, dh); if (dh->dh_fail_sparse) { if (dh->dh_err == 0 && dh->dh_bp && BP_IS_HOLE(dh->dh_bp)) dh->dh_err = SET_ERROR(ENOENT); if (dh->dh_err) { if (dh->dh_parent) dbuf_rele(dh->dh_parent, NULL); return (dh->dh_err); } } if (dh->dh_err && dh->dh_err != ENOENT) return (dh->dh_err); dh->dh_db = dbuf_create(dh->dh_dn, dh->dh_level, dh->dh_blkid, dh->dh_parent, dh->dh_bp); } if (dh->dh_fail_uncached && dh->dh_db->db_state != DB_CACHED) { mutex_exit(&dh->dh_db->db_mtx); return (SET_ERROR(ENOENT)); } if (dh->dh_db->db_buf != NULL) { arc_buf_access(dh->dh_db->db_buf); ASSERT3P(dh->dh_db->db.db_data, ==, dh->dh_db->db_buf->b_data); } ASSERT(dh->dh_db->db_buf == NULL || arc_referenced(dh->dh_db->db_buf)); /* * If this buffer is currently syncing out, and we are are * still referencing it from db_data, we need to make a copy * of it in case we decide we want to dirty it again in this txg. */ if (dh->dh_db->db_level == 0 && dh->dh_db->db_blkid != DMU_BONUS_BLKID && dh->dh_dn->dn_object != DMU_META_DNODE_OBJECT && dh->dh_db->db_state == DB_CACHED && dh->dh_db->db_data_pending) { dh->dh_dr = dh->dh_db->db_data_pending; if (dh->dh_dr->dt.dl.dr_data == dh->dh_db->db_buf) dbuf_hold_copy(dh); } if (multilist_link_active(&dh->dh_db->db_cache_link)) { ASSERT(refcount_is_zero(&dh->dh_db->db_holds)); multilist_remove(dbuf_cache, dh->dh_db); (void) refcount_remove_many(&dbuf_cache_size, dh->dh_db->db.db_size, dh->dh_db); DBUF_STAT_BUMPDOWN(cache_levels[dh->dh_db->db_level]); DBUF_STAT_BUMPDOWN(cache_count); DBUF_STAT_DECR(cache_levels_bytes[dh->dh_db->db_level], dh->dh_db->db.db_size); } (void) refcount_add(&dh->dh_db->db_holds, dh->dh_tag); DBUF_VERIFY(dh->dh_db); mutex_exit(&dh->dh_db->db_mtx); /* NOTE: we can't rele the parent until after we drop the db_mtx */ if (dh->dh_parent) dbuf_rele(dh->dh_parent, NULL); ASSERT3P(DB_DNODE(dh->dh_db), ==, dh->dh_dn); ASSERT3U(dh->dh_db->db_blkid, ==, dh->dh_blkid); ASSERT3U(dh->dh_db->db_level, ==, dh->dh_level); *(dh->dh_dbp) = dh->dh_db; return (0); } /* * The following code preserves the recursive function dbuf_hold_impl() * but moves the local variables AND function arguments to the heap to * minimize the stack frame size. Enough space is initially allocated * on the stack for 20 levels of recursion. */ int dbuf_hold_impl(dnode_t *dn, uint8_t level, uint64_t blkid, boolean_t fail_sparse, boolean_t fail_uncached, void *tag, dmu_buf_impl_t **dbp) { struct dbuf_hold_impl_data *dh; int error; dh = kmem_alloc(sizeof (struct dbuf_hold_impl_data) * DBUF_HOLD_IMPL_MAX_DEPTH, KM_SLEEP); __dbuf_hold_impl_init(dh, dn, level, blkid, fail_sparse, fail_uncached, tag, dbp, 0); error = __dbuf_hold_impl(dh); kmem_free(dh, sizeof (struct dbuf_hold_impl_data) * DBUF_HOLD_IMPL_MAX_DEPTH); return (error); } static void __dbuf_hold_impl_init(struct dbuf_hold_impl_data *dh, dnode_t *dn, uint8_t level, uint64_t blkid, boolean_t fail_sparse, boolean_t fail_uncached, void *tag, dmu_buf_impl_t **dbp, int depth) { dh->dh_dn = dn; dh->dh_level = level; dh->dh_blkid = blkid; dh->dh_fail_sparse = fail_sparse; dh->dh_fail_uncached = fail_uncached; dh->dh_tag = tag; dh->dh_dbp = dbp; dh->dh_db = NULL; dh->dh_parent = NULL; dh->dh_bp = NULL; dh->dh_err = 0; dh->dh_dr = NULL; dh->dh_depth = depth; } dmu_buf_impl_t * dbuf_hold(dnode_t *dn, uint64_t blkid, void *tag) { return (dbuf_hold_level(dn, 0, blkid, tag)); } dmu_buf_impl_t * dbuf_hold_level(dnode_t *dn, int level, uint64_t blkid, void *tag) { dmu_buf_impl_t *db; int err = dbuf_hold_impl(dn, level, blkid, FALSE, FALSE, tag, &db); return (err ? NULL : db); } void dbuf_create_bonus(dnode_t *dn) { ASSERT(RW_WRITE_HELD(&dn->dn_struct_rwlock)); ASSERT(dn->dn_bonus == NULL); dn->dn_bonus = dbuf_create(dn, 0, DMU_BONUS_BLKID, dn->dn_dbuf, NULL); } int dbuf_spill_set_blksz(dmu_buf_t *db_fake, uint64_t blksz, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; if (db->db_blkid != DMU_SPILL_BLKID) return (SET_ERROR(ENOTSUP)); if (blksz == 0) blksz = SPA_MINBLOCKSIZE; ASSERT3U(blksz, <=, spa_maxblocksize(dmu_objset_spa(db->db_objset))); blksz = P2ROUNDUP(blksz, SPA_MINBLOCKSIZE); DB_DNODE_ENTER(db); dn = DB_DNODE(db); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); dbuf_new_size(db, blksz, tx); rw_exit(&dn->dn_struct_rwlock); DB_DNODE_EXIT(db); return (0); } void dbuf_rm_spill(dnode_t *dn, dmu_tx_t *tx) { dbuf_free_range(dn, DMU_SPILL_BLKID, DMU_SPILL_BLKID, tx); } #pragma weak dmu_buf_add_ref = dbuf_add_ref void dbuf_add_ref(dmu_buf_impl_t *db, void *tag) { int64_t holds = refcount_add(&db->db_holds, tag); VERIFY3S(holds, >, 1); } #pragma weak dmu_buf_try_add_ref = dbuf_try_add_ref boolean_t dbuf_try_add_ref(dmu_buf_t *db_fake, objset_t *os, uint64_t obj, uint64_t blkid, void *tag) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dmu_buf_impl_t *found_db; boolean_t result = B_FALSE; if (blkid == DMU_BONUS_BLKID) found_db = dbuf_find_bonus(os, obj); else found_db = dbuf_find(os, obj, 0, blkid); if (found_db != NULL) { if (db == found_db && dbuf_refcount(db) > db->db_dirtycnt) { (void) refcount_add(&db->db_holds, tag); result = B_TRUE; } mutex_exit(&found_db->db_mtx); } return (result); } /* * If you call dbuf_rele() you had better not be referencing the dnode handle * unless you have some other direct or indirect hold on the dnode. (An indirect * hold is a hold on one of the dnode's dbufs, including the bonus buffer.) * Without that, the dbuf_rele() could lead to a dnode_rele() followed by the * dnode's parent dbuf evicting its dnode handles. */ void dbuf_rele(dmu_buf_impl_t *db, void *tag) { mutex_enter(&db->db_mtx); dbuf_rele_and_unlock(db, tag); } void dmu_buf_rele(dmu_buf_t *db, void *tag) { dbuf_rele((dmu_buf_impl_t *)db, tag); } /* * dbuf_rele() for an already-locked dbuf. This is necessary to allow * db_dirtycnt and db_holds to be updated atomically. */ void dbuf_rele_and_unlock(dmu_buf_impl_t *db, void *tag) { int64_t holds; ASSERT(MUTEX_HELD(&db->db_mtx)); DBUF_VERIFY(db); /* * Remove the reference to the dbuf before removing its hold on the * dnode so we can guarantee in dnode_move() that a referenced bonus * buffer has a corresponding dnode hold. */ holds = refcount_remove(&db->db_holds, tag); ASSERT(holds >= 0); /* * We can't freeze indirects if there is a possibility that they * may be modified in the current syncing context. */ if (db->db_buf != NULL && holds == (db->db_level == 0 ? db->db_dirtycnt : 0)) { arc_buf_freeze(db->db_buf); } if (holds == db->db_dirtycnt && db->db_level == 0 && db->db_user_immediate_evict) dbuf_evict_user(db); if (holds == 0) { if (db->db_blkid == DMU_BONUS_BLKID) { dnode_t *dn; boolean_t evict_dbuf = db->db_pending_evict; /* * If the dnode moves here, we cannot cross this * barrier until the move completes. */ DB_DNODE_ENTER(db); dn = DB_DNODE(db); atomic_dec_32(&dn->dn_dbufs_count); /* * Decrementing the dbuf count means that the bonus * buffer's dnode hold is no longer discounted in * dnode_move(). The dnode cannot move until after * the dnode_rele() below. */ DB_DNODE_EXIT(db); /* * Do not reference db after its lock is dropped. * Another thread may evict it. */ mutex_exit(&db->db_mtx); if (evict_dbuf) dnode_evict_bonus(dn); dnode_rele(dn, db); } else if (db->db_buf == NULL) { /* * This is a special case: we never associated this * dbuf with any data allocated from the ARC. */ ASSERT(db->db_state == DB_UNCACHED || db->db_state == DB_NOFILL); dbuf_destroy(db); } else if (arc_released(db->db_buf)) { /* * This dbuf has anonymous data associated with it. */ dbuf_destroy(db); } else { boolean_t do_arc_evict = B_FALSE; blkptr_t bp; spa_t *spa = dmu_objset_spa(db->db_objset); if (!DBUF_IS_CACHEABLE(db) && db->db_blkptr != NULL && !BP_IS_HOLE(db->db_blkptr) && !BP_IS_EMBEDDED(db->db_blkptr)) { do_arc_evict = B_TRUE; bp = *db->db_blkptr; } if (!DBUF_IS_CACHEABLE(db) || db->db_pending_evict) { dbuf_destroy(db); } else if (!multilist_link_active(&db->db_cache_link)) { multilist_insert(dbuf_cache, db); (void) refcount_add_many(&dbuf_cache_size, db->db.db_size, db); DBUF_STAT_BUMP(cache_levels[db->db_level]); DBUF_STAT_BUMP(cache_count); DBUF_STAT_INCR(cache_levels_bytes[db->db_level], db->db.db_size); DBUF_STAT_MAX(cache_size_bytes_max, refcount_count(&dbuf_cache_size)); mutex_exit(&db->db_mtx); dbuf_evict_notify(); } if (do_arc_evict) arc_freed(spa, &bp); } } else { mutex_exit(&db->db_mtx); } } #pragma weak dmu_buf_refcount = dbuf_refcount uint64_t dbuf_refcount(dmu_buf_impl_t *db) { return (refcount_count(&db->db_holds)); } void * dmu_buf_replace_user(dmu_buf_t *db_fake, dmu_buf_user_t *old_user, dmu_buf_user_t *new_user) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; mutex_enter(&db->db_mtx); dbuf_verify_user(db, DBVU_NOT_EVICTING); if (db->db_user == old_user) db->db_user = new_user; else old_user = db->db_user; dbuf_verify_user(db, DBVU_NOT_EVICTING); mutex_exit(&db->db_mtx); return (old_user); } void * dmu_buf_set_user(dmu_buf_t *db_fake, dmu_buf_user_t *user) { return (dmu_buf_replace_user(db_fake, NULL, user)); } void * dmu_buf_set_user_ie(dmu_buf_t *db_fake, dmu_buf_user_t *user) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; db->db_user_immediate_evict = TRUE; return (dmu_buf_set_user(db_fake, user)); } void * dmu_buf_remove_user(dmu_buf_t *db_fake, dmu_buf_user_t *user) { return (dmu_buf_replace_user(db_fake, user, NULL)); } void * dmu_buf_get_user(dmu_buf_t *db_fake) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dbuf_verify_user(db, DBVU_NOT_EVICTING); return (db->db_user); } void dmu_buf_user_evict_wait() { taskq_wait(dbu_evict_taskq); } blkptr_t * dmu_buf_get_blkptr(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; return (dbi->db_blkptr); } objset_t * dmu_buf_get_objset(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; return (dbi->db_objset); } dnode_t * dmu_buf_dnode_enter(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; DB_DNODE_ENTER(dbi); return (DB_DNODE(dbi)); } void dmu_buf_dnode_exit(dmu_buf_t *db) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; DB_DNODE_EXIT(dbi); } static void dbuf_check_blkptr(dnode_t *dn, dmu_buf_impl_t *db) { /* ASSERT(dmu_tx_is_syncing(tx) */ ASSERT(MUTEX_HELD(&db->db_mtx)); if (db->db_blkptr != NULL) return; if (db->db_blkid == DMU_SPILL_BLKID) { db->db_blkptr = DN_SPILL_BLKPTR(dn->dn_phys); BP_ZERO(db->db_blkptr); return; } if (db->db_level == dn->dn_phys->dn_nlevels-1) { /* * This buffer was allocated at a time when there was * no available blkptrs from the dnode, or it was * inappropriate to hook it in (i.e., nlevels mis-match). */ ASSERT(db->db_blkid < dn->dn_phys->dn_nblkptr); ASSERT(db->db_parent == NULL); db->db_parent = dn->dn_dbuf; db->db_blkptr = &dn->dn_phys->dn_blkptr[db->db_blkid]; DBUF_VERIFY(db); } else { dmu_buf_impl_t *parent = db->db_parent; int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT(dn->dn_phys->dn_nlevels > 1); if (parent == NULL) { mutex_exit(&db->db_mtx); rw_enter(&dn->dn_struct_rwlock, RW_READER); parent = dbuf_hold_level(dn, db->db_level + 1, db->db_blkid >> epbs, db); rw_exit(&dn->dn_struct_rwlock); mutex_enter(&db->db_mtx); db->db_parent = parent; } db->db_blkptr = (blkptr_t *)parent->db.db_data + (db->db_blkid & ((1ULL << epbs) - 1)); DBUF_VERIFY(db); } } /* - * Ensure the dbuf's data is untransformed if the associated dirty - * record requires it. This is used by dbuf_sync_leaf() to ensure - * that a dnode block is decrypted before we write new data to it. - * For raw writes we assert that the buffer is already encrypted. + * When syncing out a blocks of dnodes, adjust the block to deal with + * encryption. Normally, we make sure the block is decrypted before writing + * it. If we have crypt params, then we are writing a raw (encrypted) block, + * from a raw receive. In this case, set the ARC buf's crypt params so + * that the BP will be filled with the correct byteorder, salt, iv, and mac. */ static void -dbuf_check_crypt(dbuf_dirty_record_t *dr) +dbuf_prepare_encrypted_dnode_leaf(dbuf_dirty_record_t *dr) { int err; dmu_buf_impl_t *db = dr->dr_dbuf; ASSERT(MUTEX_HELD(&db->db_mtx)); + ASSERT3U(db->db.db_object, ==, DMU_META_DNODE_OBJECT); + ASSERT3U(db->db_level, ==, 0); - if (!dr->dt.dl.dr_raw && arc_is_encrypted(db->db_buf)) { + if (!db->db_objset->os_raw_receive && arc_is_encrypted(db->db_buf)) { zbookmark_phys_t zb; /* * Unfortunately, there is currently no mechanism for * syncing context to handle decryption errors. An error * here is only possible if an attacker maliciously * changed a dnode block and updated the associated * checksums going up the block tree. */ SET_BOOKMARK(&zb, dmu_objset_id(db->db_objset), db->db.db_object, db->db_level, db->db_blkid); err = arc_untransform(db->db_buf, db->db_objset->os_spa, &zb, B_TRUE); if (err) panic("Invalid dnode block MAC"); - } else if (dr->dt.dl.dr_raw) { - /* - * Writing raw encrypted data requires the db's arc buffer - * to be converted to raw by the caller. - */ - ASSERT(arc_is_encrypted(db->db_buf)); + } else if (dr->dt.dl.dr_has_raw_params) { + (void) arc_release(dr->dt.dl.dr_data, db); + arc_convert_to_raw(dr->dt.dl.dr_data, + dmu_objset_id(db->db_objset), + dr->dt.dl.dr_byteorder, DMU_OT_DNODE, + dr->dt.dl.dr_salt, dr->dt.dl.dr_iv, dr->dt.dl.dr_mac); } } /* * dbuf_sync_indirect() is called recursively from dbuf_sync_list() so it * is critical the we not allow the compiler to inline this function in to * dbuf_sync_list() thereby drastically bloating the stack usage. */ noinline static void dbuf_sync_indirect(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn; zio_t *zio; ASSERT(dmu_tx_is_syncing(tx)); dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr); mutex_enter(&db->db_mtx); ASSERT(db->db_level > 0); DBUF_VERIFY(db); /* Read the block if it hasn't been read yet. */ if (db->db_buf == NULL) { mutex_exit(&db->db_mtx); (void) dbuf_read(db, NULL, DB_RF_MUST_SUCCEED); mutex_enter(&db->db_mtx); } ASSERT3U(db->db_state, ==, DB_CACHED); ASSERT(db->db_buf != NULL); DB_DNODE_ENTER(db); dn = DB_DNODE(db); /* Indirect block size must match what the dnode thinks it is. */ ASSERT3U(db->db.db_size, ==, 1<dn_phys->dn_indblkshift); dbuf_check_blkptr(dn, db); DB_DNODE_EXIT(db); /* Provide the pending dirty record to child dbufs */ db->db_data_pending = dr; mutex_exit(&db->db_mtx); dbuf_write(dr, db->db_buf, tx); zio = dr->dr_zio; mutex_enter(&dr->dt.di.dr_mtx); dbuf_sync_list(&dr->dt.di.dr_children, db->db_level - 1, tx); ASSERT(list_head(&dr->dt.di.dr_children) == NULL); mutex_exit(&dr->dt.di.dr_mtx); zio_nowait(zio); } /* * dbuf_sync_leaf() is called recursively from dbuf_sync_list() so it is * critical the we not allow the compiler to inline this function in to * dbuf_sync_list() thereby drastically bloating the stack usage. */ noinline static void dbuf_sync_leaf(dbuf_dirty_record_t *dr, dmu_tx_t *tx) { arc_buf_t **datap = &dr->dt.dl.dr_data; dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn; objset_t *os; uint64_t txg = tx->tx_txg; ASSERT(dmu_tx_is_syncing(tx)); dprintf_dbuf_bp(db, db->db_blkptr, "blkptr=%p", db->db_blkptr); mutex_enter(&db->db_mtx); /* * To be synced, we must be dirtied. But we * might have been freed after the dirty. */ if (db->db_state == DB_UNCACHED) { /* This buffer has been freed since it was dirtied */ ASSERT(db->db.db_data == NULL); } else if (db->db_state == DB_FILL) { /* This buffer was freed and is now being re-filled */ ASSERT(db->db.db_data != dr->dt.dl.dr_data); } else { ASSERT(db->db_state == DB_CACHED || db->db_state == DB_NOFILL); } DBUF_VERIFY(db); DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (db->db_blkid == DMU_SPILL_BLKID) { mutex_enter(&dn->dn_mtx); if (!(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR)) { /* * In the previous transaction group, the bonus buffer * was entirely used to store the attributes for the * dnode which overrode the dn_spill field. However, * when adding more attributes to the file a spill * block was required to hold the extra attributes. * * Make sure to clear the garbage left in the dn_spill * field from the previous attributes in the bonus * buffer. Otherwise, after writing out the spill * block to the new allocated dva, it will free * the old block pointed to by the invalid dn_spill. */ db->db_blkptr = NULL; } dn->dn_phys->dn_flags |= DNODE_FLAG_SPILL_BLKPTR; mutex_exit(&dn->dn_mtx); } /* * If this is a bonus buffer, simply copy the bonus data into the * dnode. It will be written out when the dnode is synced (and it * will be synced, since it must have been dirty for dbuf_sync to * be called). */ if (db->db_blkid == DMU_BONUS_BLKID) { dbuf_dirty_record_t **drp; ASSERT(*datap != NULL); ASSERT0(db->db_level); ASSERT3U(DN_MAX_BONUS_LEN(dn->dn_phys), <=, DN_SLOTS_TO_BONUSLEN(dn->dn_phys->dn_extra_slots + 1)); bcopy(*datap, DN_BONUS(dn->dn_phys), DN_MAX_BONUS_LEN(dn->dn_phys)); DB_DNODE_EXIT(db); if (*datap != db->db.db_data) { int slots = DB_DNODE(db)->dn_num_slots; int bonuslen = DN_SLOTS_TO_BONUSLEN(slots); kmem_free(*datap, bonuslen); arc_space_return(bonuslen, ARC_SPACE_BONUS); } db->db_data_pending = NULL; drp = &db->db_last_dirty; while (*drp != dr) drp = &(*drp)->dr_next; ASSERT(dr->dr_next == NULL); ASSERT(dr->dr_dbuf == db); *drp = dr->dr_next; if (dr->dr_dbuf->db_level != 0) { mutex_destroy(&dr->dt.di.dr_mtx); list_destroy(&dr->dt.di.dr_children); } kmem_free(dr, sizeof (dbuf_dirty_record_t)); ASSERT(db->db_dirtycnt > 0); db->db_dirtycnt -= 1; dbuf_rele_and_unlock(db, (void *)(uintptr_t)txg); return; } os = dn->dn_objset; /* * This function may have dropped the db_mtx lock allowing a dmu_sync * operation to sneak in. As a result, we need to ensure that we * don't check the dr_override_state until we have returned from * dbuf_check_blkptr. */ dbuf_check_blkptr(dn, db); /* * If this buffer is in the middle of an immediate write, * wait for the synchronous IO to complete. */ while (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC) { ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); cv_wait(&db->db_changed, &db->db_mtx); ASSERT(dr->dt.dl.dr_override_state != DR_NOT_OVERRIDDEN); } /* * If this is a dnode block, ensure it is appropriately encrypted * or decrypted, depending on what we are writing to it this txg. */ if (os->os_encrypted && dn->dn_object == DMU_META_DNODE_OBJECT) - dbuf_check_crypt(dr); + dbuf_prepare_encrypted_dnode_leaf(dr); if (db->db_state != DB_NOFILL && dn->dn_object != DMU_META_DNODE_OBJECT && refcount_count(&db->db_holds) > 1 && dr->dt.dl.dr_override_state != DR_OVERRIDDEN && *datap == db->db_buf) { /* * If this buffer is currently "in use" (i.e., there * are active holds and db_data still references it), * then make a copy before we start the write so that * any modifications from the open txg will not leak * into this write. * * NOTE: this copy does not need to be made for * objects only modified in the syncing context (e.g. * DNONE_DNODE blocks). */ int psize = arc_buf_size(*datap); int lsize = arc_buf_lsize(*datap); arc_buf_contents_t type = DBUF_GET_BUFC_TYPE(db); enum zio_compress compress_type = arc_get_compression(*datap); if (arc_is_encrypted(*datap)) { boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; arc_get_raw_params(*datap, &byteorder, salt, iv, mac); *datap = arc_alloc_raw_buf(os->os_spa, db, dmu_objset_id(os), byteorder, salt, iv, mac, dn->dn_type, psize, lsize, compress_type); } else if (compress_type != ZIO_COMPRESS_OFF) { ASSERT3U(type, ==, ARC_BUFC_DATA); *datap = arc_alloc_compressed_buf(os->os_spa, db, psize, lsize, compress_type); } else { *datap = arc_alloc_buf(os->os_spa, db, type, psize); } bcopy(db->db.db_data, (*datap)->b_data, psize); } db->db_data_pending = dr; mutex_exit(&db->db_mtx); dbuf_write(dr, *datap, tx); ASSERT(!list_link_active(&dr->dr_dirty_node)); if (dn->dn_object == DMU_META_DNODE_OBJECT) { list_insert_tail(&dn->dn_dirty_records[txg&TXG_MASK], dr); DB_DNODE_EXIT(db); } else { /* * Although zio_nowait() does not "wait for an IO", it does * initiate the IO. If this is an empty write it seems plausible * that the IO could actually be completed before the nowait * returns. We need to DB_DNODE_EXIT() first in case * zio_nowait() invalidates the dbuf. */ DB_DNODE_EXIT(db); zio_nowait(dr->dr_zio); } } void dbuf_sync_list(list_t *list, int level, dmu_tx_t *tx) { dbuf_dirty_record_t *dr; while ((dr = list_head(list))) { if (dr->dr_zio != NULL) { /* * If we find an already initialized zio then we * are processing the meta-dnode, and we have finished. * The dbufs for all dnodes are put back on the list * during processing, so that we can zio_wait() * these IOs after initiating all child IOs. */ ASSERT3U(dr->dr_dbuf->db.db_object, ==, DMU_META_DNODE_OBJECT); break; } if (dr->dr_dbuf->db_blkid != DMU_BONUS_BLKID && dr->dr_dbuf->db_blkid != DMU_SPILL_BLKID) { VERIFY3U(dr->dr_dbuf->db_level, ==, level); } list_remove(list, dr); if (dr->dr_dbuf->db_level > 0) dbuf_sync_indirect(dr, tx); else dbuf_sync_leaf(dr, tx); } } /* ARGSUSED */ static void dbuf_write_ready(zio_t *zio, arc_buf_t *buf, void *vdb) { dmu_buf_impl_t *db = vdb; dnode_t *dn; blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; spa_t *spa = zio->io_spa; int64_t delta; uint64_t fill = 0; int i; ASSERT3P(db->db_blkptr, !=, NULL); ASSERT3P(&db->db_data_pending->dr_bp_copy, ==, bp); DB_DNODE_ENTER(db); dn = DB_DNODE(db); delta = bp_get_dsize_sync(spa, bp) - bp_get_dsize_sync(spa, bp_orig); dnode_diduse_space(dn, delta - zio->io_prev_space_delta); zio->io_prev_space_delta = delta; if (bp->blk_birth != 0) { ASSERT((db->db_blkid != DMU_SPILL_BLKID && BP_GET_TYPE(bp) == dn->dn_type) || (db->db_blkid == DMU_SPILL_BLKID && BP_GET_TYPE(bp) == dn->dn_bonustype) || BP_IS_EMBEDDED(bp)); ASSERT(BP_GET_LEVEL(bp) == db->db_level); } mutex_enter(&db->db_mtx); #ifdef ZFS_DEBUG if (db->db_blkid == DMU_SPILL_BLKID) { ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR); ASSERT(!(BP_IS_HOLE(bp)) && db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys)); } #endif if (db->db_level == 0) { mutex_enter(&dn->dn_mtx); if (db->db_blkid > dn->dn_phys->dn_maxblkid && db->db_blkid != DMU_SPILL_BLKID) dn->dn_phys->dn_maxblkid = db->db_blkid; mutex_exit(&dn->dn_mtx); if (dn->dn_type == DMU_OT_DNODE) { i = 0; while (i < db->db.db_size) { dnode_phys_t *dnp = (void *)(((char *)db->db.db_data) + i); i += DNODE_MIN_SIZE; if (dnp->dn_type != DMU_OT_NONE) { fill++; i += dnp->dn_extra_slots * DNODE_MIN_SIZE; } } } else { if (BP_IS_HOLE(bp)) { fill = 0; } else { fill = 1; } } } else { blkptr_t *ibp = db->db.db_data; ASSERT3U(db->db.db_size, ==, 1<dn_phys->dn_indblkshift); for (i = db->db.db_size >> SPA_BLKPTRSHIFT; i > 0; i--, ibp++) { if (BP_IS_HOLE(ibp)) continue; fill += BP_GET_FILL(ibp); } } DB_DNODE_EXIT(db); if (!BP_IS_EMBEDDED(bp)) BP_SET_FILL(bp, fill); mutex_exit(&db->db_mtx); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); *db->db_blkptr = *bp; rw_exit(&dn->dn_struct_rwlock); } /* ARGSUSED */ /* * This function gets called just prior to running through the compression * stage of the zio pipeline. If we're an indirect block comprised of only * holes, then we want this indirect to be compressed away to a hole. In * order to do that we must zero out any information about the holes that * this indirect points to prior to before we try to compress it. */ static void dbuf_write_children_ready(zio_t *zio, arc_buf_t *buf, void *vdb) { dmu_buf_impl_t *db = vdb; dnode_t *dn; blkptr_t *bp; unsigned int epbs, i; ASSERT3U(db->db_level, >, 0); DB_DNODE_ENTER(db); dn = DB_DNODE(db); epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT; ASSERT3U(epbs, <, 31); /* Determine if all our children are holes */ for (i = 0, bp = db->db.db_data; i < 1ULL << epbs; i++, bp++) { if (!BP_IS_HOLE(bp)) break; } /* * If all the children are holes, then zero them all out so that * we may get compressed away. */ if (i == 1ULL << epbs) { /* * We only found holes. Grab the rwlock to prevent * anybody from reading the blocks we're about to * zero out. */ rw_enter(&dn->dn_struct_rwlock, RW_WRITER); bzero(db->db.db_data, db->db.db_size); rw_exit(&dn->dn_struct_rwlock); } DB_DNODE_EXIT(db); } /* * The SPA will call this callback several times for each zio - once * for every physical child i/o (zio->io_phys_children times). This * allows the DMU to monitor the progress of each logical i/o. For example, * there may be 2 copies of an indirect block, or many fragments of a RAID-Z * block. There may be a long delay before all copies/fragments are completed, * so this callback allows us to retire dirty space gradually, as the physical * i/os complete. */ /* ARGSUSED */ static void dbuf_write_physdone(zio_t *zio, arc_buf_t *buf, void *arg) { dmu_buf_impl_t *db = arg; objset_t *os = db->db_objset; dsl_pool_t *dp = dmu_objset_pool(os); dbuf_dirty_record_t *dr; int delta = 0; dr = db->db_data_pending; ASSERT3U(dr->dr_txg, ==, zio->io_txg); /* * The callback will be called io_phys_children times. Retire one * portion of our dirty space each time we are called. Any rounding * error will be cleaned up by dsl_pool_sync()'s call to * dsl_pool_undirty_space(). */ delta = dr->dr_accounted / zio->io_phys_children; dsl_pool_undirty_space(dp, delta, zio->io_txg); } /* ARGSUSED */ static void dbuf_write_done(zio_t *zio, arc_buf_t *buf, void *vdb) { dmu_buf_impl_t *db = vdb; blkptr_t *bp_orig = &zio->io_bp_orig; blkptr_t *bp = db->db_blkptr; objset_t *os = db->db_objset; dmu_tx_t *tx = os->os_synctx; dbuf_dirty_record_t **drp, *dr; ASSERT0(zio->io_error); ASSERT(db->db_blkptr == bp); /* * For nopwrites and rewrites we ensure that the bp matches our * original and bypass all the accounting. */ if (zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE)) { ASSERT(BP_EQUAL(bp, bp_orig)); } else { dsl_dataset_t *ds = os->os_dsl_dataset; (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); dsl_dataset_block_born(ds, bp, tx); } mutex_enter(&db->db_mtx); DBUF_VERIFY(db); drp = &db->db_last_dirty; while ((dr = *drp) != db->db_data_pending) drp = &dr->dr_next; ASSERT(!list_link_active(&dr->dr_dirty_node)); ASSERT(dr->dr_dbuf == db); ASSERT(dr->dr_next == NULL); *drp = dr->dr_next; #ifdef ZFS_DEBUG if (db->db_blkid == DMU_SPILL_BLKID) { dnode_t *dn; DB_DNODE_ENTER(db); dn = DB_DNODE(db); ASSERT(dn->dn_phys->dn_flags & DNODE_FLAG_SPILL_BLKPTR); ASSERT(!(BP_IS_HOLE(db->db_blkptr)) && db->db_blkptr == DN_SPILL_BLKPTR(dn->dn_phys)); DB_DNODE_EXIT(db); } #endif if (db->db_level == 0) { ASSERT(db->db_blkid != DMU_BONUS_BLKID); ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN); if (db->db_state != DB_NOFILL) { if (dr->dt.dl.dr_data != db->db_buf) arc_buf_destroy(dr->dt.dl.dr_data, db); } } else { dnode_t *dn; DB_DNODE_ENTER(db); dn = DB_DNODE(db); ASSERT(list_head(&dr->dt.di.dr_children) == NULL); ASSERT3U(db->db.db_size, ==, 1 << dn->dn_phys->dn_indblkshift); if (!BP_IS_HOLE(db->db_blkptr)) { ASSERTV(int epbs = dn->dn_phys->dn_indblkshift - SPA_BLKPTRSHIFT); ASSERT3U(db->db_blkid, <=, dn->dn_phys->dn_maxblkid >> (db->db_level * epbs)); ASSERT3U(BP_GET_LSIZE(db->db_blkptr), ==, db->db.db_size); } DB_DNODE_EXIT(db); mutex_destroy(&dr->dt.di.dr_mtx); list_destroy(&dr->dt.di.dr_children); } kmem_free(dr, sizeof (dbuf_dirty_record_t)); cv_broadcast(&db->db_changed); ASSERT(db->db_dirtycnt > 0); db->db_dirtycnt -= 1; db->db_data_pending = NULL; dbuf_rele_and_unlock(db, (void *)(uintptr_t)tx->tx_txg); } static void dbuf_write_nofill_ready(zio_t *zio) { dbuf_write_ready(zio, NULL, zio->io_private); } static void dbuf_write_nofill_done(zio_t *zio) { dbuf_write_done(zio, NULL, zio->io_private); } static void dbuf_write_override_ready(zio_t *zio) { dbuf_dirty_record_t *dr = zio->io_private; dmu_buf_impl_t *db = dr->dr_dbuf; dbuf_write_ready(zio, NULL, db); } static void dbuf_write_override_done(zio_t *zio) { dbuf_dirty_record_t *dr = zio->io_private; dmu_buf_impl_t *db = dr->dr_dbuf; blkptr_t *obp = &dr->dt.dl.dr_overridden_by; mutex_enter(&db->db_mtx); if (!BP_EQUAL(zio->io_bp, obp)) { if (!BP_IS_HOLE(obp)) dsl_free(spa_get_dsl(zio->io_spa), zio->io_txg, obp); arc_release(dr->dt.dl.dr_data, db); } mutex_exit(&db->db_mtx); dbuf_write_done(zio, NULL, db); if (zio->io_abd != NULL) abd_put(zio->io_abd); } typedef struct dbuf_remap_impl_callback_arg { objset_t *drica_os; uint64_t drica_blk_birth; dmu_tx_t *drica_tx; } dbuf_remap_impl_callback_arg_t; static void dbuf_remap_impl_callback(uint64_t vdev, uint64_t offset, uint64_t size, void *arg) { dbuf_remap_impl_callback_arg_t *drica = arg; objset_t *os = drica->drica_os; spa_t *spa = dmu_objset_spa(os); dmu_tx_t *tx = drica->drica_tx; ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); if (os == spa_meta_objset(spa)) { spa_vdev_indirect_mark_obsolete(spa, vdev, offset, size, tx); } else { dsl_dataset_block_remapped(dmu_objset_ds(os), vdev, offset, size, drica->drica_blk_birth, tx); } } static void dbuf_remap_impl(dnode_t *dn, blkptr_t *bp, dmu_tx_t *tx) { blkptr_t bp_copy = *bp; spa_t *spa = dmu_objset_spa(dn->dn_objset); dbuf_remap_impl_callback_arg_t drica; ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); drica.drica_os = dn->dn_objset; drica.drica_blk_birth = bp->blk_birth; drica.drica_tx = tx; if (spa_remap_blkptr(spa, &bp_copy, dbuf_remap_impl_callback, &drica)) { /* * The struct_rwlock prevents dbuf_read_impl() from * dereferencing the BP while we are changing it. To * avoid lock contention, only grab it when we are actually * changing the BP. */ rw_enter(&dn->dn_struct_rwlock, RW_WRITER); *bp = bp_copy; rw_exit(&dn->dn_struct_rwlock); } } /* * Returns true if a dbuf_remap would modify the dbuf. We do this by attempting * to remap a copy of every bp in the dbuf. */ boolean_t dbuf_can_remap(const dmu_buf_impl_t *db) { spa_t *spa = dmu_objset_spa(db->db_objset); blkptr_t *bp = db->db.db_data; boolean_t ret = B_FALSE; ASSERT3U(db->db_level, >, 0); ASSERT3S(db->db_state, ==, DB_CACHED); ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); for (int i = 0; i < db->db.db_size >> SPA_BLKPTRSHIFT; i++) { blkptr_t bp_copy = bp[i]; if (spa_remap_blkptr(spa, &bp_copy, NULL, NULL)) { ret = B_TRUE; break; } } spa_config_exit(spa, SCL_VDEV, FTAG); return (ret); } boolean_t dnode_needs_remap(const dnode_t *dn) { spa_t *spa = dmu_objset_spa(dn->dn_objset); boolean_t ret = B_FALSE; if (dn->dn_phys->dn_nlevels == 0) { return (B_FALSE); } ASSERT(spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)); spa_config_enter(spa, SCL_VDEV, FTAG, RW_READER); for (int j = 0; j < dn->dn_phys->dn_nblkptr; j++) { blkptr_t bp_copy = dn->dn_phys->dn_blkptr[j]; if (spa_remap_blkptr(spa, &bp_copy, NULL, NULL)) { ret = B_TRUE; break; } } spa_config_exit(spa, SCL_VDEV, FTAG); return (ret); } /* * Remap any existing BP's to concrete vdevs, if possible. */ static void dbuf_remap(dnode_t *dn, dmu_buf_impl_t *db, dmu_tx_t *tx) { spa_t *spa = dmu_objset_spa(db->db_objset); ASSERT(dsl_pool_sync_context(spa_get_dsl(spa))); if (!spa_feature_is_active(spa, SPA_FEATURE_DEVICE_REMOVAL)) return; if (db->db_level > 0) { blkptr_t *bp = db->db.db_data; for (int i = 0; i < db->db.db_size >> SPA_BLKPTRSHIFT; i++) { dbuf_remap_impl(dn, &bp[i], tx); } } else if (db->db.db_object == DMU_META_DNODE_OBJECT) { dnode_phys_t *dnp = db->db.db_data; ASSERT3U(db->db_dnode_handle->dnh_dnode->dn_type, ==, DMU_OT_DNODE); for (int i = 0; i < db->db.db_size >> DNODE_SHIFT; i += dnp[i].dn_extra_slots + 1) { for (int j = 0; j < dnp[i].dn_nblkptr; j++) { dbuf_remap_impl(dn, &dnp[i].dn_blkptr[j], tx); } } } } /* Issue I/O to commit a dirty buffer to disk. */ static void dbuf_write(dbuf_dirty_record_t *dr, arc_buf_t *data, dmu_tx_t *tx) { dmu_buf_impl_t *db = dr->dr_dbuf; dnode_t *dn; objset_t *os; dmu_buf_impl_t *parent = db->db_parent; uint64_t txg = tx->tx_txg; zbookmark_phys_t zb; zio_prop_t zp; zio_t *zio; int wp_flag = 0; ASSERT(dmu_tx_is_syncing(tx)); DB_DNODE_ENTER(db); dn = DB_DNODE(db); os = dn->dn_objset; if (db->db_state != DB_NOFILL) { if (db->db_level > 0 || dn->dn_type == DMU_OT_DNODE) { /* * Private object buffers are released here rather * than in dbuf_dirty() since they are only modified * in the syncing context and we don't want the * overhead of making multiple copies of the data. */ if (BP_IS_HOLE(db->db_blkptr)) { arc_buf_thaw(data); } else { dbuf_release_bp(db); } dbuf_remap(dn, db, tx); } } if (parent != dn->dn_dbuf) { /* Our parent is an indirect block. */ /* We have a dirty parent that has been scheduled for write. */ ASSERT(parent && parent->db_data_pending); /* Our parent's buffer is one level closer to the dnode. */ ASSERT(db->db_level == parent->db_level-1); /* * We're about to modify our parent's db_data by modifying * our block pointer, so the parent must be released. */ ASSERT(arc_released(parent->db_buf)); zio = parent->db_data_pending->dr_zio; } else { /* Our parent is the dnode itself. */ ASSERT((db->db_level == dn->dn_phys->dn_nlevels-1 && db->db_blkid != DMU_SPILL_BLKID) || (db->db_blkid == DMU_SPILL_BLKID && db->db_level == 0)); if (db->db_blkid != DMU_SPILL_BLKID) ASSERT3P(db->db_blkptr, ==, &dn->dn_phys->dn_blkptr[db->db_blkid]); zio = dn->dn_zio; } ASSERT(db->db_level == 0 || data == db->db_buf); ASSERT3U(db->db_blkptr->blk_birth, <=, txg); ASSERT(zio); SET_BOOKMARK(&zb, os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : DMU_META_OBJSET, db->db.db_object, db->db_level, db->db_blkid); if (db->db_blkid == DMU_SPILL_BLKID) wp_flag = WP_SPILL; wp_flag |= (db->db_state == DB_NOFILL) ? WP_NOFILL : 0; dmu_write_policy(os, dn, db->db_level, wp_flag, &zp); DB_DNODE_EXIT(db); /* * We copy the blkptr now (rather than when we instantiate the dirty * record), because its value can change between open context and * syncing context. We do not need to hold dn_struct_rwlock to read * db_blkptr because we are in syncing context. */ dr->dr_bp_copy = *db->db_blkptr; if (db->db_level == 0 && dr->dt.dl.dr_override_state == DR_OVERRIDDEN) { /* * The BP for this block has been provided by open context * (by dmu_sync() or dmu_buf_write_embedded()). */ abd_t *contents = (data != NULL) ? abd_get_from_buf(data->b_data, arc_buf_size(data)) : NULL; dr->dr_zio = zio_write(zio, os->os_spa, txg, &dr->dr_bp_copy, contents, db->db.db_size, db->db.db_size, &zp, dbuf_write_override_ready, NULL, NULL, dbuf_write_override_done, dr, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); mutex_enter(&db->db_mtx); dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN; zio_write_override(dr->dr_zio, &dr->dt.dl.dr_overridden_by, dr->dt.dl.dr_copies, dr->dt.dl.dr_nopwrite); mutex_exit(&db->db_mtx); } else if (db->db_state == DB_NOFILL) { ASSERT(zp.zp_checksum == ZIO_CHECKSUM_OFF || zp.zp_checksum == ZIO_CHECKSUM_NOPARITY); dr->dr_zio = zio_write(zio, os->os_spa, txg, &dr->dr_bp_copy, NULL, db->db.db_size, db->db.db_size, &zp, dbuf_write_nofill_ready, NULL, NULL, dbuf_write_nofill_done, db, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED | ZIO_FLAG_NODATA, &zb); } else { ASSERT(arc_released(data)); /* * For indirect blocks, we want to setup the children * ready callback so that we can properly handle an indirect * block that only contains holes. */ arc_write_done_func_t *children_ready_cb = NULL; if (db->db_level != 0) children_ready_cb = dbuf_write_children_ready; dr->dr_zio = arc_write(zio, os->os_spa, txg, &dr->dr_bp_copy, data, DBUF_IS_L2CACHEABLE(db), &zp, dbuf_write_ready, children_ready_cb, dbuf_write_physdone, dbuf_write_done, db, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); } } #if defined(_KERNEL) && defined(HAVE_SPL) EXPORT_SYMBOL(dbuf_find); EXPORT_SYMBOL(dbuf_is_metadata); EXPORT_SYMBOL(dbuf_destroy); EXPORT_SYMBOL(dbuf_loan_arcbuf); EXPORT_SYMBOL(dbuf_whichblock); EXPORT_SYMBOL(dbuf_read); EXPORT_SYMBOL(dbuf_unoverride); EXPORT_SYMBOL(dbuf_free_range); EXPORT_SYMBOL(dbuf_new_size); EXPORT_SYMBOL(dbuf_release_bp); EXPORT_SYMBOL(dbuf_dirty); -EXPORT_SYMBOL(dmu_buf_will_change_crypt_params); +EXPORT_SYMBOL(dmu_buf_set_crypt_params); EXPORT_SYMBOL(dmu_buf_will_dirty); EXPORT_SYMBOL(dmu_buf_will_not_fill); EXPORT_SYMBOL(dmu_buf_will_fill); EXPORT_SYMBOL(dmu_buf_fill_done); EXPORT_SYMBOL(dmu_buf_rele); EXPORT_SYMBOL(dbuf_assign_arcbuf); EXPORT_SYMBOL(dbuf_prefetch); EXPORT_SYMBOL(dbuf_hold_impl); EXPORT_SYMBOL(dbuf_hold); EXPORT_SYMBOL(dbuf_hold_level); EXPORT_SYMBOL(dbuf_create_bonus); EXPORT_SYMBOL(dbuf_spill_set_blksz); EXPORT_SYMBOL(dbuf_rm_spill); EXPORT_SYMBOL(dbuf_add_ref); EXPORT_SYMBOL(dbuf_rele); EXPORT_SYMBOL(dbuf_rele_and_unlock); EXPORT_SYMBOL(dbuf_refcount); EXPORT_SYMBOL(dbuf_sync_list); EXPORT_SYMBOL(dmu_buf_set_user); EXPORT_SYMBOL(dmu_buf_set_user_ie); EXPORT_SYMBOL(dmu_buf_get_user); EXPORT_SYMBOL(dmu_buf_get_blkptr); /* BEGIN CSTYLED */ module_param(dbuf_cache_max_bytes, ulong, 0644); MODULE_PARM_DESC(dbuf_cache_max_bytes, "Maximum size in bytes of the dbuf cache."); module_param(dbuf_cache_hiwater_pct, uint, 0644); MODULE_PARM_DESC(dbuf_cache_hiwater_pct, "Percentage over dbuf_cache_max_bytes when dbufs must be evicted " "directly."); module_param(dbuf_cache_lowater_pct, uint, 0644); MODULE_PARM_DESC(dbuf_cache_lowater_pct, "Percentage below dbuf_cache_max_bytes when the evict thread stops " "evicting dbufs."); module_param(dbuf_cache_shift, int, 0644); MODULE_PARM_DESC(dbuf_cache_shift, "Set the size of the dbuf cache to a log2 fraction of arc size."); /* END CSTYLED */ #endif diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c index 0352393dc276..2d2586a8c6b9 100644 --- a/module/zfs/dmu.c +++ b/module/zfs/dmu.c @@ -1,2667 +1,2554 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2017 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. * Copyright (c) 2016, Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2015 by Chunwei Chen. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #ifdef _KERNEL #include #include #endif /* * Enable/disable nopwrite feature. */ int zfs_nopwrite_enabled = 1; /* * Tunable to control percentage of dirtied blocks from frees in one TXG. * After this threshold is crossed, additional dirty blocks from frees * wait until the next TXG. * A value of zero will disable this throttle. */ unsigned long zfs_per_txg_dirty_frees_percent = 30; /* * Enable/disable forcing txg sync when dirty in dmu_offset_next. */ int zfs_dmu_offset_next_sync = 0; /* * This can be used for testing, to ensure that certain actions happen * while in the middle of a remap (which might otherwise complete too * quickly). */ int zfs_object_remap_one_indirect_delay_ticks = 0; const dmu_object_type_info_t dmu_ot[DMU_OT_NUMTYPES] = { { DMU_BSWAP_UINT8, TRUE, FALSE, "unallocated" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "object directory" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "object array" }, { DMU_BSWAP_UINT8, TRUE, FALSE, "packed nvlist" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "packed nvlist size" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "bpobj" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "bpobj header" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "SPA space map header" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "SPA space map" }, { DMU_BSWAP_UINT64, TRUE, TRUE, "ZIL intent log" }, { DMU_BSWAP_DNODE, TRUE, TRUE, "DMU dnode" }, { DMU_BSWAP_OBJSET, TRUE, FALSE, "DMU objset" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "DSL directory" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL directory child map"}, { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL dataset snap map" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL props" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "DSL dataset" }, { DMU_BSWAP_ZNODE, TRUE, FALSE, "ZFS znode" }, { DMU_BSWAP_OLDACL, TRUE, TRUE, "ZFS V0 ACL" }, { DMU_BSWAP_UINT8, FALSE, TRUE, "ZFS plain file" }, { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS directory" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "ZFS master node" }, { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS delete queue" }, { DMU_BSWAP_UINT8, FALSE, TRUE, "zvol object" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "zvol prop" }, { DMU_BSWAP_UINT8, FALSE, TRUE, "other uint8[]" }, { DMU_BSWAP_UINT64, FALSE, TRUE, "other uint64[]" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "other ZAP" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "persistent error log" }, { DMU_BSWAP_UINT8, TRUE, FALSE, "SPA history" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "SPA history offsets" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "Pool properties" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL permissions" }, { DMU_BSWAP_ACL, TRUE, TRUE, "ZFS ACL" }, { DMU_BSWAP_UINT8, TRUE, TRUE, "ZFS SYSACL" }, { DMU_BSWAP_UINT8, TRUE, TRUE, "FUID table" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "FUID table size" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL dataset next clones"}, { DMU_BSWAP_ZAP, TRUE, FALSE, "scan work queue" }, { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS user/group/project used" }, { DMU_BSWAP_ZAP, TRUE, TRUE, "ZFS user/group/project quota"}, { DMU_BSWAP_ZAP, TRUE, FALSE, "snapshot refcount tags"}, { DMU_BSWAP_ZAP, TRUE, FALSE, "DDT ZAP algorithm" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DDT statistics" }, { DMU_BSWAP_UINT8, TRUE, TRUE, "System attributes" }, { DMU_BSWAP_ZAP, TRUE, TRUE, "SA master node" }, { DMU_BSWAP_ZAP, TRUE, TRUE, "SA attr registration" }, { DMU_BSWAP_ZAP, TRUE, TRUE, "SA attr layouts" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "scan translations" }, { DMU_BSWAP_UINT8, FALSE, TRUE, "deduplicated block" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL deadlist map" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "DSL deadlist map hdr" }, { DMU_BSWAP_ZAP, TRUE, FALSE, "DSL dir clones" }, { DMU_BSWAP_UINT64, TRUE, FALSE, "bpobj subobj" } }; const dmu_object_byteswap_info_t dmu_ot_byteswap[DMU_BSWAP_NUMFUNCS] = { { byteswap_uint8_array, "uint8" }, { byteswap_uint16_array, "uint16" }, { byteswap_uint32_array, "uint32" }, { byteswap_uint64_array, "uint64" }, { zap_byteswap, "zap" }, { dnode_buf_byteswap, "dnode" }, { dmu_objset_byteswap, "objset" }, { zfs_znode_byteswap, "znode" }, { zfs_oldacl_byteswap, "oldacl" }, { zfs_acl_byteswap, "acl" } }; int dmu_buf_hold_noread_by_dnode(dnode_t *dn, uint64_t offset, void *tag, dmu_buf_t **dbp) { uint64_t blkid; dmu_buf_impl_t *db; blkid = dbuf_whichblock(dn, 0, offset); rw_enter(&dn->dn_struct_rwlock, RW_READER); db = dbuf_hold(dn, blkid, tag); rw_exit(&dn->dn_struct_rwlock); if (db == NULL) { *dbp = NULL; return (SET_ERROR(EIO)); } *dbp = &db->db; return (0); } int dmu_buf_hold_noread(objset_t *os, uint64_t object, uint64_t offset, void *tag, dmu_buf_t **dbp) { dnode_t *dn; uint64_t blkid; dmu_buf_impl_t *db; int err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); blkid = dbuf_whichblock(dn, 0, offset); rw_enter(&dn->dn_struct_rwlock, RW_READER); db = dbuf_hold(dn, blkid, tag); rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); if (db == NULL) { *dbp = NULL; return (SET_ERROR(EIO)); } *dbp = &db->db; return (err); } int dmu_buf_hold_by_dnode(dnode_t *dn, uint64_t offset, void *tag, dmu_buf_t **dbp, int flags) { int err; int db_flags = DB_RF_CANFAIL; if (flags & DMU_READ_NO_PREFETCH) db_flags |= DB_RF_NOPREFETCH; if (flags & DMU_READ_NO_DECRYPT) db_flags |= DB_RF_NO_DECRYPT; err = dmu_buf_hold_noread_by_dnode(dn, offset, tag, dbp); if (err == 0) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)(*dbp); err = dbuf_read(db, NULL, db_flags); if (err != 0) { dbuf_rele(db, tag); *dbp = NULL; } } return (err); } int dmu_buf_hold(objset_t *os, uint64_t object, uint64_t offset, void *tag, dmu_buf_t **dbp, int flags) { int err; int db_flags = DB_RF_CANFAIL; if (flags & DMU_READ_NO_PREFETCH) db_flags |= DB_RF_NOPREFETCH; if (flags & DMU_READ_NO_DECRYPT) db_flags |= DB_RF_NO_DECRYPT; err = dmu_buf_hold_noread(os, object, offset, tag, dbp); if (err == 0) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)(*dbp); err = dbuf_read(db, NULL, db_flags); if (err != 0) { dbuf_rele(db, tag); *dbp = NULL; } } return (err); } int dmu_bonus_max(void) { return (DN_OLD_MAX_BONUSLEN); } int dmu_set_bonus(dmu_buf_t *db_fake, int newsize, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; int error; DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dn->dn_bonus != db) { error = SET_ERROR(EINVAL); } else if (newsize < 0 || newsize > db_fake->db_size) { error = SET_ERROR(EINVAL); } else { dnode_setbonuslen(dn, newsize, tx); error = 0; } DB_DNODE_EXIT(db); return (error); } int dmu_set_bonustype(dmu_buf_t *db_fake, dmu_object_type_t type, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; int error; DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (!DMU_OT_IS_VALID(type)) { error = SET_ERROR(EINVAL); } else if (dn->dn_bonus != db) { error = SET_ERROR(EINVAL); } else { dnode_setbonus_type(dn, type, tx); error = 0; } DB_DNODE_EXIT(db); return (error); } dmu_object_type_t dmu_get_bonustype(dmu_buf_t *db_fake) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; dmu_object_type_t type; DB_DNODE_ENTER(db); dn = DB_DNODE(db); type = dn->dn_bonustype; DB_DNODE_EXIT(db); return (type); } int dmu_rm_spill(objset_t *os, uint64_t object, dmu_tx_t *tx) { dnode_t *dn; int error; error = dnode_hold(os, object, FTAG, &dn); dbuf_rm_spill(dn, tx); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); dnode_rm_spill(dn, tx); rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); return (error); } /* * returns ENOENT, EIO, or 0. */ int dmu_bonus_hold_impl(objset_t *os, uint64_t object, void *tag, uint32_t flags, dmu_buf_t **dbp) { dnode_t *dn; dmu_buf_impl_t *db; int error; uint32_t db_flags = DB_RF_MUST_SUCCEED; if (flags & DMU_READ_NO_PREFETCH) db_flags |= DB_RF_NOPREFETCH; if (flags & DMU_READ_NO_DECRYPT) db_flags |= DB_RF_NO_DECRYPT; error = dnode_hold(os, object, FTAG, &dn); if (error) return (error); rw_enter(&dn->dn_struct_rwlock, RW_READER); if (dn->dn_bonus == NULL) { rw_exit(&dn->dn_struct_rwlock); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); if (dn->dn_bonus == NULL) dbuf_create_bonus(dn); } db = dn->dn_bonus; /* as long as the bonus buf is held, the dnode will be held */ if (refcount_add(&db->db_holds, tag) == 1) { VERIFY(dnode_add_ref(dn, db)); atomic_inc_32(&dn->dn_dbufs_count); } /* * Wait to drop dn_struct_rwlock until after adding the bonus dbuf's * hold and incrementing the dbuf count to ensure that dnode_move() sees * a dnode hold for every dbuf. */ rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); error = dbuf_read(db, NULL, db_flags); if (error) { dnode_evict_bonus(dn); dbuf_rele(db, tag); *dbp = NULL; return (error); } *dbp = &db->db; return (0); } int dmu_bonus_hold(objset_t *os, uint64_t obj, void *tag, dmu_buf_t **dbp) { return (dmu_bonus_hold_impl(os, obj, tag, DMU_READ_NO_PREFETCH, dbp)); } /* * returns ENOENT, EIO, or 0. * * This interface will allocate a blank spill dbuf when a spill blk * doesn't already exist on the dnode. * * if you only want to find an already existing spill db, then * dmu_spill_hold_existing() should be used. */ int dmu_spill_hold_by_dnode(dnode_t *dn, uint32_t flags, void *tag, dmu_buf_t **dbp) { dmu_buf_impl_t *db = NULL; int err; if ((flags & DB_RF_HAVESTRUCT) == 0) rw_enter(&dn->dn_struct_rwlock, RW_READER); db = dbuf_hold(dn, DMU_SPILL_BLKID, tag); if ((flags & DB_RF_HAVESTRUCT) == 0) rw_exit(&dn->dn_struct_rwlock); if (db == NULL) { *dbp = NULL; return (SET_ERROR(EIO)); } err = dbuf_read(db, NULL, flags); if (err == 0) *dbp = &db->db; else { dbuf_rele(db, tag); *dbp = NULL; } return (err); } int dmu_spill_hold_existing(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)bonus; dnode_t *dn; int err; DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (spa_version(dn->dn_objset->os_spa) < SPA_VERSION_SA) { err = SET_ERROR(EINVAL); } else { rw_enter(&dn->dn_struct_rwlock, RW_READER); if (!dn->dn_have_spill) { err = SET_ERROR(ENOENT); } else { err = dmu_spill_hold_by_dnode(dn, DB_RF_HAVESTRUCT | DB_RF_CANFAIL, tag, dbp); } rw_exit(&dn->dn_struct_rwlock); } DB_DNODE_EXIT(db); return (err); } int dmu_spill_hold_by_bonus(dmu_buf_t *bonus, void *tag, dmu_buf_t **dbp) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)bonus; dnode_t *dn; int err; DB_DNODE_ENTER(db); dn = DB_DNODE(db); err = dmu_spill_hold_by_dnode(dn, DB_RF_CANFAIL, tag, dbp); DB_DNODE_EXIT(db); return (err); } /* * Note: longer-term, we should modify all of the dmu_buf_*() interfaces * to take a held dnode rather than -- the lookup is wasteful, * and can induce severe lock contention when writing to several files * whose dnodes are in the same block. */ static int dmu_buf_hold_array_by_dnode(dnode_t *dn, uint64_t offset, uint64_t length, boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp, uint32_t flags) { dmu_buf_t **dbp; uint64_t blkid, nblks, i; uint32_t dbuf_flags; int err; zio_t *zio; ASSERT(length <= DMU_MAX_ACCESS); /* * Note: We directly notify the prefetch code of this read, so that * we can tell it about the multi-block read. dbuf_read() only knows * about the one block it is accessing. */ dbuf_flags = DB_RF_CANFAIL | DB_RF_NEVERWAIT | DB_RF_HAVESTRUCT | DB_RF_NOPREFETCH; rw_enter(&dn->dn_struct_rwlock, RW_READER); if (dn->dn_datablkshift) { int blkshift = dn->dn_datablkshift; nblks = (P2ROUNDUP(offset + length, 1ULL << blkshift) - P2ALIGN(offset, 1ULL << blkshift)) >> blkshift; } else { if (offset + length > dn->dn_datablksz) { zfs_panic_recover("zfs: accessing past end of object " "%llx/%llx (size=%u access=%llu+%llu)", (longlong_t)dn->dn_objset-> os_dsl_dataset->ds_object, (longlong_t)dn->dn_object, dn->dn_datablksz, (longlong_t)offset, (longlong_t)length); rw_exit(&dn->dn_struct_rwlock); return (SET_ERROR(EIO)); } nblks = 1; } dbp = kmem_zalloc(sizeof (dmu_buf_t *) * nblks, KM_SLEEP); zio = zio_root(dn->dn_objset->os_spa, NULL, NULL, ZIO_FLAG_CANFAIL); blkid = dbuf_whichblock(dn, 0, offset); for (i = 0; i < nblks; i++) { dmu_buf_impl_t *db = dbuf_hold(dn, blkid + i, tag); if (db == NULL) { rw_exit(&dn->dn_struct_rwlock); dmu_buf_rele_array(dbp, nblks, tag); zio_nowait(zio); return (SET_ERROR(EIO)); } /* initiate async i/o */ if (read) (void) dbuf_read(db, zio, dbuf_flags); dbp[i] = &db->db; } if ((flags & DMU_READ_NO_PREFETCH) == 0 && DNODE_META_IS_CACHEABLE(dn) && length <= zfetch_array_rd_sz) { dmu_zfetch(&dn->dn_zfetch, blkid, nblks, read && DNODE_IS_CACHEABLE(dn)); } rw_exit(&dn->dn_struct_rwlock); /* wait for async i/o */ err = zio_wait(zio); if (err) { dmu_buf_rele_array(dbp, nblks, tag); return (err); } /* wait for other io to complete */ if (read) { for (i = 0; i < nblks; i++) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)dbp[i]; mutex_enter(&db->db_mtx); while (db->db_state == DB_READ || db->db_state == DB_FILL) cv_wait(&db->db_changed, &db->db_mtx); if (db->db_state == DB_UNCACHED) err = SET_ERROR(EIO); mutex_exit(&db->db_mtx); if (err) { dmu_buf_rele_array(dbp, nblks, tag); return (err); } } } *numbufsp = nblks; *dbpp = dbp; return (0); } static int dmu_buf_hold_array(objset_t *os, uint64_t object, uint64_t offset, uint64_t length, int read, void *tag, int *numbufsp, dmu_buf_t ***dbpp) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag, numbufsp, dbpp, DMU_READ_PREFETCH); dnode_rele(dn, FTAG); return (err); } int dmu_buf_hold_array_by_bonus(dmu_buf_t *db_fake, uint64_t offset, uint64_t length, boolean_t read, void *tag, int *numbufsp, dmu_buf_t ***dbpp) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; int err; DB_DNODE_ENTER(db); dn = DB_DNODE(db); err = dmu_buf_hold_array_by_dnode(dn, offset, length, read, tag, numbufsp, dbpp, DMU_READ_PREFETCH); DB_DNODE_EXIT(db); return (err); } void dmu_buf_rele_array(dmu_buf_t **dbp_fake, int numbufs, void *tag) { int i; dmu_buf_impl_t **dbp = (dmu_buf_impl_t **)dbp_fake; if (numbufs == 0) return; for (i = 0; i < numbufs; i++) { if (dbp[i]) dbuf_rele(dbp[i], tag); } kmem_free(dbp, sizeof (dmu_buf_t *) * numbufs); } /* * Issue prefetch i/os for the given blocks. If level is greater than 0, the * indirect blocks prefeteched will be those that point to the blocks containing * the data starting at offset, and continuing to offset + len. * * Note that if the indirect blocks above the blocks being prefetched are not * in cache, they will be asychronously read in. */ void dmu_prefetch(objset_t *os, uint64_t object, int64_t level, uint64_t offset, uint64_t len, zio_priority_t pri) { dnode_t *dn; uint64_t blkid; int nblks, err; if (len == 0) { /* they're interested in the bonus buffer */ dn = DMU_META_DNODE(os); if (object == 0 || object >= DN_MAX_OBJECT) return; rw_enter(&dn->dn_struct_rwlock, RW_READER); blkid = dbuf_whichblock(dn, level, object * sizeof (dnode_phys_t)); dbuf_prefetch(dn, level, blkid, pri, 0); rw_exit(&dn->dn_struct_rwlock); return; } /* * XXX - Note, if the dnode for the requested object is not * already cached, we will do a *synchronous* read in the * dnode_hold() call. The same is true for any indirects. */ err = dnode_hold(os, object, FTAG, &dn); if (err != 0) return; rw_enter(&dn->dn_struct_rwlock, RW_READER); /* * offset + len - 1 is the last byte we want to prefetch for, and offset * is the first. Then dbuf_whichblk(dn, level, off + len - 1) is the * last block we want to prefetch, and dbuf_whichblock(dn, level, * offset) is the first. Then the number we need to prefetch is the * last - first + 1. */ if (level > 0 || dn->dn_datablkshift != 0) { nblks = dbuf_whichblock(dn, level, offset + len - 1) - dbuf_whichblock(dn, level, offset) + 1; } else { nblks = (offset < dn->dn_datablksz); } if (nblks != 0) { blkid = dbuf_whichblock(dn, level, offset); for (int i = 0; i < nblks; i++) dbuf_prefetch(dn, level, blkid + i, pri, 0); } rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); } /* * Get the next "chunk" of file data to free. We traverse the file from * the end so that the file gets shorter over time (if we crashes in the * middle, this will leave us in a better state). We find allocated file * data by simply searching the allocated level 1 indirects. * * On input, *start should be the first offset that does not need to be * freed (e.g. "offset + length"). On return, *start will be the first * offset that should be freed. */ static int get_next_chunk(dnode_t *dn, uint64_t *start, uint64_t minimum) { uint64_t maxblks = DMU_MAX_ACCESS >> (dn->dn_indblkshift + 1); /* bytes of data covered by a level-1 indirect block */ uint64_t iblkrange = dn->dn_datablksz * EPB(dn->dn_indblkshift, SPA_BLKPTRSHIFT); ASSERT3U(minimum, <=, *start); if (*start - minimum <= iblkrange * maxblks) { *start = minimum; return (0); } ASSERT(ISP2(iblkrange)); for (uint64_t blks = 0; *start > minimum && blks < maxblks; blks++) { int err; /* * dnode_next_offset(BACKWARDS) will find an allocated L1 * indirect block at or before the input offset. We must * decrement *start so that it is at the end of the region * to search. */ (*start)--; err = dnode_next_offset(dn, DNODE_FIND_BACKWARDS, start, 2, 1, 0); /* if there are no indirect blocks before start, we are done */ if (err == ESRCH) { *start = minimum; break; } else if (err != 0) { return (err); } /* set start to the beginning of this L1 indirect */ *start = P2ALIGN(*start, iblkrange); } if (*start < minimum) *start = minimum; return (0); } /* * If this objset is of type OST_ZFS return true if vfs's unmounted flag is set, * otherwise return false. * Used below in dmu_free_long_range_impl() to enable abort when unmounting */ /*ARGSUSED*/ static boolean_t dmu_objset_zfs_unmounting(objset_t *os) { #ifdef _KERNEL if (dmu_objset_type(os) == DMU_OST_ZFS) return (zfs_get_vfs_flag_unmounted(os)); #endif return (B_FALSE); } static int dmu_free_long_range_impl(objset_t *os, dnode_t *dn, uint64_t offset, - uint64_t length, boolean_t raw) + uint64_t length) { uint64_t object_size; int err; uint64_t dirty_frees_threshold; dsl_pool_t *dp = dmu_objset_pool(os); if (dn == NULL) return (SET_ERROR(EINVAL)); object_size = (dn->dn_maxblkid + 1) * dn->dn_datablksz; if (offset >= object_size) return (0); if (zfs_per_txg_dirty_frees_percent <= 100) dirty_frees_threshold = zfs_per_txg_dirty_frees_percent * zfs_dirty_data_max / 100; else dirty_frees_threshold = zfs_dirty_data_max / 4; if (length == DMU_OBJECT_END || offset + length > object_size) length = object_size - offset; while (length != 0) { uint64_t chunk_end, chunk_begin, chunk_len; uint64_t long_free_dirty_all_txgs = 0; dmu_tx_t *tx; if (dmu_objset_zfs_unmounting(dn->dn_objset)) return (SET_ERROR(EINTR)); chunk_end = chunk_begin = offset + length; /* move chunk_begin backwards to the beginning of this chunk */ err = get_next_chunk(dn, &chunk_begin, offset); if (err) return (err); ASSERT3U(chunk_begin, >=, offset); ASSERT3U(chunk_begin, <=, chunk_end); chunk_len = chunk_end - chunk_begin; mutex_enter(&dp->dp_lock); for (int t = 0; t < TXG_SIZE; t++) { long_free_dirty_all_txgs += dp->dp_long_free_dirty_pertxg[t]; } mutex_exit(&dp->dp_lock); /* * To avoid filling up a TXG with just frees wait for * the next TXG to open before freeing more chunks if * we have reached the threshold of frees */ if (dirty_frees_threshold != 0 && long_free_dirty_all_txgs >= dirty_frees_threshold) { txg_wait_open(dp, 0); continue; } tx = dmu_tx_create(os); dmu_tx_hold_free(tx, dn->dn_object, chunk_begin, chunk_len); /* * Mark this transaction as typically resulting in a net * reduction in space used. */ dmu_tx_mark_netfree(tx); err = dmu_tx_assign(tx, TXG_WAIT); if (err) { dmu_tx_abort(tx); return (err); } mutex_enter(&dp->dp_lock); dp->dp_long_free_dirty_pertxg[dmu_tx_get_txg(tx) & TXG_MASK] += chunk_len; mutex_exit(&dp->dp_lock); DTRACE_PROBE3(free__long__range, uint64_t, long_free_dirty_all_txgs, uint64_t, chunk_len, uint64_t, dmu_tx_get_txg(tx)); dnode_free_range(dn, chunk_begin, chunk_len, tx); - /* if this is a raw free, mark the dirty record as such */ - if (raw) { - dbuf_dirty_record_t *dr = dn->dn_dbuf->db_last_dirty; - - while (dr != NULL && dr->dr_txg > tx->tx_txg) - dr = dr->dr_next; - if (dr != NULL && dr->dr_txg == tx->tx_txg) { - dr->dt.dl.dr_raw = B_TRUE; - dn->dn_objset->os_next_write_raw - [tx->tx_txg & TXG_MASK] = B_TRUE; - } - } - dmu_tx_commit(tx); length -= chunk_len; } return (0); } int dmu_free_long_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t length) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err != 0) return (err); - err = dmu_free_long_range_impl(os, dn, offset, length, B_FALSE); + err = dmu_free_long_range_impl(os, dn, offset, length); /* * It is important to zero out the maxblkid when freeing the entire * file, so that (a) subsequent calls to dmu_free_long_range_impl() * will take the fast path, and (b) dnode_reallocate() can verify * that the entire file has been freed. */ if (err == 0 && offset == 0 && length == DMU_OBJECT_END) dn->dn_maxblkid = 0; dnode_rele(dn, FTAG); return (err); } -/* - * This function is equivalent to dmu_free_long_range(), but also - * marks the new dirty record as a raw write. - */ int -dmu_free_long_range_raw(objset_t *os, uint64_t object, - uint64_t offset, uint64_t length) -{ - dnode_t *dn; - int err; - - err = dnode_hold(os, object, FTAG, &dn); - if (err != 0) - return (err); - err = dmu_free_long_range_impl(os, dn, offset, length, B_TRUE); - - /* - * It is important to zero out the maxblkid when freeing the entire - * file, so that (a) subsequent calls to dmu_free_long_range_impl() - * will take the fast path, and (b) dnode_reallocate() can verify - * that the entire file has been freed. - */ - if (err == 0 && offset == 0 && length == DMU_OBJECT_END) - dn->dn_maxblkid = 0; - - dnode_rele(dn, FTAG); - return (err); -} - -static int -dmu_free_long_object_impl(objset_t *os, uint64_t object, boolean_t raw) +dmu_free_long_object(objset_t *os, uint64_t object) { dmu_tx_t *tx; int err; err = dmu_free_long_range(os, object, 0, DMU_OBJECT_END); if (err != 0) return (err); tx = dmu_tx_create(os); dmu_tx_hold_bonus(tx, object); dmu_tx_hold_free(tx, object, 0, DMU_OBJECT_END); dmu_tx_mark_netfree(tx); err = dmu_tx_assign(tx, TXG_WAIT); if (err == 0) { - if (raw) - err = dmu_object_dirty_raw(os, object, tx); if (err == 0) err = dmu_object_free(os, object, tx); dmu_tx_commit(tx); } else { dmu_tx_abort(tx); } return (err); } -int -dmu_free_long_object(objset_t *os, uint64_t object) -{ - return (dmu_free_long_object_impl(os, object, B_FALSE)); -} - -int -dmu_free_long_object_raw(objset_t *os, uint64_t object) -{ - return (dmu_free_long_object_impl(os, object, B_TRUE)); -} - - int dmu_free_range(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx) { dnode_t *dn; int err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); ASSERT(offset < UINT64_MAX); ASSERT(size == DMU_OBJECT_END || size <= UINT64_MAX - offset); dnode_free_range(dn, offset, size, tx); dnode_rele(dn, FTAG); return (0); } static int dmu_read_impl(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, uint32_t flags) { dmu_buf_t **dbp; int numbufs, err = 0; /* * Deal with odd block sizes, where there can't be data past the first * block. If we ever do the tail block optimization, we will need to * handle that here as well. */ if (dn->dn_maxblkid == 0) { uint64_t newsz = offset > dn->dn_datablksz ? 0 : MIN(size, dn->dn_datablksz - offset); bzero((char *)buf + newsz, size - newsz); size = newsz; } while (size > 0) { uint64_t mylen = MIN(size, DMU_MAX_ACCESS / 2); int i; /* * NB: we could do this block-at-a-time, but it's nice * to be reading in parallel. */ err = dmu_buf_hold_array_by_dnode(dn, offset, mylen, TRUE, FTAG, &numbufs, &dbp, flags); if (err) break; for (i = 0; i < numbufs; i++) { uint64_t tocpy; int64_t bufoff; dmu_buf_t *db = dbp[i]; ASSERT(size > 0); bufoff = offset - db->db_offset; tocpy = MIN(db->db_size - bufoff, size); (void) memcpy(buf, (char *)db->db_data + bufoff, tocpy); offset += tocpy; size -= tocpy; buf = (char *)buf + tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); } return (err); } int dmu_read(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, void *buf, uint32_t flags) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err != 0) return (err); err = dmu_read_impl(dn, offset, size, buf, flags); dnode_rele(dn, FTAG); return (err); } int dmu_read_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, void *buf, uint32_t flags) { return (dmu_read_impl(dn, offset, size, buf, flags)); } static void dmu_write_impl(dmu_buf_t **dbp, int numbufs, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx) { int i; for (i = 0; i < numbufs; i++) { uint64_t tocpy; int64_t bufoff; dmu_buf_t *db = dbp[i]; ASSERT(size > 0); bufoff = offset - db->db_offset; tocpy = MIN(db->db_size - bufoff, size); ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); if (tocpy == db->db_size) dmu_buf_will_fill(db, tx); else dmu_buf_will_dirty(db, tx); (void) memcpy((char *)db->db_data + bufoff, buf, tocpy); if (tocpy == db->db_size) dmu_buf_fill_done(db, tx); offset += tocpy; size -= tocpy; buf = (char *)buf + tocpy; } } void dmu_write(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx) { dmu_buf_t **dbp; int numbufs; if (size == 0) return; VERIFY0(dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG, &numbufs, &dbp)); dmu_write_impl(dbp, numbufs, offset, size, buf, tx); dmu_buf_rele_array(dbp, numbufs, FTAG); } void dmu_write_by_dnode(dnode_t *dn, uint64_t offset, uint64_t size, const void *buf, dmu_tx_t *tx) { dmu_buf_t **dbp; int numbufs; if (size == 0) return; VERIFY0(dmu_buf_hold_array_by_dnode(dn, offset, size, FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH)); dmu_write_impl(dbp, numbufs, offset, size, buf, tx); dmu_buf_rele_array(dbp, numbufs, FTAG); } static int dmu_object_remap_one_indirect(objset_t *os, dnode_t *dn, uint64_t last_removal_txg, uint64_t offset) { uint64_t l1blkid = dbuf_whichblock(dn, 1, offset); int err = 0; rw_enter(&dn->dn_struct_rwlock, RW_READER); dmu_buf_impl_t *dbuf = dbuf_hold_level(dn, 1, l1blkid, FTAG); ASSERT3P(dbuf, !=, NULL); /* * If the block hasn't been written yet, this default will ensure * we don't try to remap it. */ uint64_t birth = UINT64_MAX; ASSERT3U(last_removal_txg, !=, UINT64_MAX); if (dbuf->db_blkptr != NULL) birth = dbuf->db_blkptr->blk_birth; rw_exit(&dn->dn_struct_rwlock); /* * If this L1 was already written after the last removal, then we've * already tried to remap it. */ if (birth <= last_removal_txg && dbuf_read(dbuf, NULL, DB_RF_MUST_SUCCEED) == 0 && dbuf_can_remap(dbuf)) { dmu_tx_t *tx = dmu_tx_create(os); dmu_tx_hold_remap_l1indirect(tx, dn->dn_object); err = dmu_tx_assign(tx, TXG_WAIT); if (err == 0) { (void) dbuf_dirty(dbuf, tx); dmu_tx_commit(tx); } else { dmu_tx_abort(tx); } } dbuf_rele(dbuf, FTAG); delay(zfs_object_remap_one_indirect_delay_ticks); return (err); } /* * Remap all blockpointers in the object, if possible, so that they reference * only concrete vdevs. * * To do this, iterate over the L0 blockpointers and remap any that reference * an indirect vdev. Note that we only examine L0 blockpointers; since we * cannot guarantee that we can remap all blockpointer anyways (due to split * blocks), we do not want to make the code unnecessarily complicated to * catch the unlikely case that there is an L1 block on an indirect vdev that * contains no indirect blockpointers. */ int dmu_object_remap_indirects(objset_t *os, uint64_t object, uint64_t last_removal_txg) { uint64_t offset, l1span; int err; dnode_t *dn; err = dnode_hold(os, object, FTAG, &dn); if (err != 0) { return (err); } if (dn->dn_nlevels <= 1) { if (issig(JUSTLOOKING) && issig(FORREAL)) { err = SET_ERROR(EINTR); } /* * If the dnode has no indirect blocks, we cannot dirty them. * We still want to remap the blkptr(s) in the dnode if * appropriate, so mark it as dirty. */ if (err == 0 && dnode_needs_remap(dn)) { dmu_tx_t *tx = dmu_tx_create(os); dmu_tx_hold_bonus(tx, dn->dn_object); if ((err = dmu_tx_assign(tx, TXG_WAIT)) == 0) { dnode_setdirty(dn, tx); dmu_tx_commit(tx); } else { dmu_tx_abort(tx); } } dnode_rele(dn, FTAG); return (err); } offset = 0; l1span = 1ULL << (dn->dn_indblkshift - SPA_BLKPTRSHIFT + dn->dn_datablkshift); /* * Find the next L1 indirect that is not a hole. */ while (dnode_next_offset(dn, 0, &offset, 2, 1, 0) == 0) { if (issig(JUSTLOOKING) && issig(FORREAL)) { err = SET_ERROR(EINTR); break; } if ((err = dmu_object_remap_one_indirect(os, dn, last_removal_txg, offset)) != 0) { break; } offset += l1span; } dnode_rele(dn, FTAG); return (err); } void dmu_prealloc(objset_t *os, uint64_t object, uint64_t offset, uint64_t size, dmu_tx_t *tx) { dmu_buf_t **dbp; int numbufs, i; if (size == 0) return; VERIFY(0 == dmu_buf_hold_array(os, object, offset, size, FALSE, FTAG, &numbufs, &dbp)); for (i = 0; i < numbufs; i++) { dmu_buf_t *db = dbp[i]; dmu_buf_will_not_fill(db, tx); } dmu_buf_rele_array(dbp, numbufs, FTAG); } void dmu_write_embedded(objset_t *os, uint64_t object, uint64_t offset, void *data, uint8_t etype, uint8_t comp, int uncompressed_size, int compressed_size, int byteorder, dmu_tx_t *tx) { dmu_buf_t *db; ASSERT3U(etype, <, NUM_BP_EMBEDDED_TYPES); ASSERT3U(comp, <, ZIO_COMPRESS_FUNCTIONS); VERIFY0(dmu_buf_hold_noread(os, object, offset, FTAG, &db)); dmu_buf_write_embedded(db, data, (bp_embedded_type_t)etype, (enum zio_compress)comp, uncompressed_size, compressed_size, byteorder, tx); dmu_buf_rele(db, FTAG); } /* * DMU support for xuio */ kstat_t *xuio_ksp = NULL; typedef struct xuio_stats { /* loaned yet not returned arc_buf */ kstat_named_t xuiostat_onloan_rbuf; kstat_named_t xuiostat_onloan_wbuf; /* whether a copy is made when loaning out a read buffer */ kstat_named_t xuiostat_rbuf_copied; kstat_named_t xuiostat_rbuf_nocopy; /* whether a copy is made when assigning a write buffer */ kstat_named_t xuiostat_wbuf_copied; kstat_named_t xuiostat_wbuf_nocopy; } xuio_stats_t; static xuio_stats_t xuio_stats = { { "onloan_read_buf", KSTAT_DATA_UINT64 }, { "onloan_write_buf", KSTAT_DATA_UINT64 }, { "read_buf_copied", KSTAT_DATA_UINT64 }, { "read_buf_nocopy", KSTAT_DATA_UINT64 }, { "write_buf_copied", KSTAT_DATA_UINT64 }, { "write_buf_nocopy", KSTAT_DATA_UINT64 } }; #define XUIOSTAT_INCR(stat, val) \ atomic_add_64(&xuio_stats.stat.value.ui64, (val)) #define XUIOSTAT_BUMP(stat) XUIOSTAT_INCR(stat, 1) #ifdef HAVE_UIO_ZEROCOPY int dmu_xuio_init(xuio_t *xuio, int nblk) { dmu_xuio_t *priv; uio_t *uio = &xuio->xu_uio; uio->uio_iovcnt = nblk; uio->uio_iov = kmem_zalloc(nblk * sizeof (iovec_t), KM_SLEEP); priv = kmem_zalloc(sizeof (dmu_xuio_t), KM_SLEEP); priv->cnt = nblk; priv->bufs = kmem_zalloc(nblk * sizeof (arc_buf_t *), KM_SLEEP); priv->iovp = (iovec_t *)uio->uio_iov; XUIO_XUZC_PRIV(xuio) = priv; if (XUIO_XUZC_RW(xuio) == UIO_READ) XUIOSTAT_INCR(xuiostat_onloan_rbuf, nblk); else XUIOSTAT_INCR(xuiostat_onloan_wbuf, nblk); return (0); } void dmu_xuio_fini(xuio_t *xuio) { dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); int nblk = priv->cnt; kmem_free(priv->iovp, nblk * sizeof (iovec_t)); kmem_free(priv->bufs, nblk * sizeof (arc_buf_t *)); kmem_free(priv, sizeof (dmu_xuio_t)); if (XUIO_XUZC_RW(xuio) == UIO_READ) XUIOSTAT_INCR(xuiostat_onloan_rbuf, -nblk); else XUIOSTAT_INCR(xuiostat_onloan_wbuf, -nblk); } /* * Initialize iov[priv->next] and priv->bufs[priv->next] with { off, n, abuf } * and increase priv->next by 1. */ int dmu_xuio_add(xuio_t *xuio, arc_buf_t *abuf, offset_t off, size_t n) { struct iovec *iov; uio_t *uio = &xuio->xu_uio; dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); int i = priv->next++; ASSERT(i < priv->cnt); ASSERT(off + n <= arc_buf_lsize(abuf)); iov = (iovec_t *)uio->uio_iov + i; iov->iov_base = (char *)abuf->b_data + off; iov->iov_len = n; priv->bufs[i] = abuf; return (0); } int dmu_xuio_cnt(xuio_t *xuio) { dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); return (priv->cnt); } arc_buf_t * dmu_xuio_arcbuf(xuio_t *xuio, int i) { dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); ASSERT(i < priv->cnt); return (priv->bufs[i]); } void dmu_xuio_clear(xuio_t *xuio, int i) { dmu_xuio_t *priv = XUIO_XUZC_PRIV(xuio); ASSERT(i < priv->cnt); priv->bufs[i] = NULL; } #endif /* HAVE_UIO_ZEROCOPY */ static void xuio_stat_init(void) { xuio_ksp = kstat_create("zfs", 0, "xuio_stats", "misc", KSTAT_TYPE_NAMED, sizeof (xuio_stats) / sizeof (kstat_named_t), KSTAT_FLAG_VIRTUAL); if (xuio_ksp != NULL) { xuio_ksp->ks_data = &xuio_stats; kstat_install(xuio_ksp); } } static void xuio_stat_fini(void) { if (xuio_ksp != NULL) { kstat_delete(xuio_ksp); xuio_ksp = NULL; } } void xuio_stat_wbuf_copied(void) { XUIOSTAT_BUMP(xuiostat_wbuf_copied); } void xuio_stat_wbuf_nocopy(void) { XUIOSTAT_BUMP(xuiostat_wbuf_nocopy); } #ifdef _KERNEL int dmu_read_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size) { dmu_buf_t **dbp; int numbufs, i, err; #ifdef HAVE_UIO_ZEROCOPY xuio_t *xuio = NULL; #endif /* * NB: we could do this block-at-a-time, but it's nice * to be reading in parallel. */ err = dmu_buf_hold_array_by_dnode(dn, uio->uio_loffset, size, TRUE, FTAG, &numbufs, &dbp, 0); if (err) return (err); for (i = 0; i < numbufs; i++) { uint64_t tocpy; int64_t bufoff; dmu_buf_t *db = dbp[i]; ASSERT(size > 0); bufoff = uio->uio_loffset - db->db_offset; tocpy = MIN(db->db_size - bufoff, size); #ifdef HAVE_UIO_ZEROCOPY if (xuio) { dmu_buf_impl_t *dbi = (dmu_buf_impl_t *)db; arc_buf_t *dbuf_abuf = dbi->db_buf; arc_buf_t *abuf = dbuf_loan_arcbuf(dbi); err = dmu_xuio_add(xuio, abuf, bufoff, tocpy); if (!err) { uio->uio_resid -= tocpy; uio->uio_loffset += tocpy; } if (abuf == dbuf_abuf) XUIOSTAT_BUMP(xuiostat_rbuf_nocopy); else XUIOSTAT_BUMP(xuiostat_rbuf_copied); } else #endif err = uiomove((char *)db->db_data + bufoff, tocpy, UIO_READ, uio); if (err) break; size -= tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); return (err); } /* * Read 'size' bytes into the uio buffer. * From object zdb->db_object. * Starting at offset uio->uio_loffset. * * If the caller already has a dbuf in the target object * (e.g. its bonus buffer), this routine is faster than dmu_read_uio(), * because we don't have to find the dnode_t for the object. */ int dmu_read_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)zdb; dnode_t *dn; int err; if (size == 0) return (0); DB_DNODE_ENTER(db); dn = DB_DNODE(db); err = dmu_read_uio_dnode(dn, uio, size); DB_DNODE_EXIT(db); return (err); } /* * Read 'size' bytes into the uio buffer. * From the specified object * Starting at offset uio->uio_loffset. */ int dmu_read_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size) { dnode_t *dn; int err; if (size == 0) return (0); err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dmu_read_uio_dnode(dn, uio, size); dnode_rele(dn, FTAG); return (err); } int dmu_write_uio_dnode(dnode_t *dn, uio_t *uio, uint64_t size, dmu_tx_t *tx) { dmu_buf_t **dbp; int numbufs; int err = 0; int i; err = dmu_buf_hold_array_by_dnode(dn, uio->uio_loffset, size, FALSE, FTAG, &numbufs, &dbp, DMU_READ_PREFETCH); if (err) return (err); for (i = 0; i < numbufs; i++) { uint64_t tocpy; int64_t bufoff; dmu_buf_t *db = dbp[i]; ASSERT(size > 0); bufoff = uio->uio_loffset - db->db_offset; tocpy = MIN(db->db_size - bufoff, size); ASSERT(i == 0 || i == numbufs-1 || tocpy == db->db_size); if (tocpy == db->db_size) dmu_buf_will_fill(db, tx); else dmu_buf_will_dirty(db, tx); /* * XXX uiomove could block forever (eg.nfs-backed * pages). There needs to be a uiolockdown() function * to lock the pages in memory, so that uiomove won't * block. */ err = uiomove((char *)db->db_data + bufoff, tocpy, UIO_WRITE, uio); if (tocpy == db->db_size) dmu_buf_fill_done(db, tx); if (err) break; size -= tocpy; } dmu_buf_rele_array(dbp, numbufs, FTAG); return (err); } /* * Write 'size' bytes from the uio buffer. * To object zdb->db_object. * Starting at offset uio->uio_loffset. * * If the caller already has a dbuf in the target object * (e.g. its bonus buffer), this routine is faster than dmu_write_uio(), * because we don't have to find the dnode_t for the object. */ int dmu_write_uio_dbuf(dmu_buf_t *zdb, uio_t *uio, uint64_t size, dmu_tx_t *tx) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)zdb; dnode_t *dn; int err; if (size == 0) return (0); DB_DNODE_ENTER(db); dn = DB_DNODE(db); err = dmu_write_uio_dnode(dn, uio, size, tx); DB_DNODE_EXIT(db); return (err); } /* * Write 'size' bytes from the uio buffer. * To the specified object. * Starting at offset uio->uio_loffset. */ int dmu_write_uio(objset_t *os, uint64_t object, uio_t *uio, uint64_t size, dmu_tx_t *tx) { dnode_t *dn; int err; if (size == 0) return (0); err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dmu_write_uio_dnode(dn, uio, size, tx); dnode_rele(dn, FTAG); return (err); } #endif /* _KERNEL */ /* * Allocate a loaned anonymous arc buffer. */ arc_buf_t * dmu_request_arcbuf(dmu_buf_t *handle, int size) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)handle; return (arc_loan_buf(db->db_objset->os_spa, B_FALSE, size)); } /* * Free a loaned arc buffer. */ void dmu_return_arcbuf(arc_buf_t *buf) { arc_return_buf(buf, FTAG); arc_buf_destroy(buf, FTAG); } -int -dmu_convert_mdn_block_to_raw(objset_t *os, uint64_t firstobj, - boolean_t byteorder, const uint8_t *salt, const uint8_t *iv, - const uint8_t *mac, dmu_tx_t *tx) -{ - int ret; - dmu_buf_t *handle = NULL; - dmu_buf_impl_t *db = NULL; - uint64_t offset = firstobj * DNODE_MIN_SIZE; - uint64_t dsobj = dmu_objset_id(os); - - ret = dmu_buf_hold_by_dnode(DMU_META_DNODE(os), offset, FTAG, &handle, - DMU_READ_PREFETCH | DMU_READ_NO_DECRYPT); - if (ret != 0) - return (ret); - - dmu_buf_will_change_crypt_params(handle, tx); - - db = (dmu_buf_impl_t *)handle; - ASSERT3P(db->db_buf, !=, NULL); - ASSERT3U(dsobj, !=, 0); - - /* - * This technically violates the assumption the dmu code makes - * that dnode blocks are only released in syncing context. - */ - (void) arc_release(db->db_buf, db); - arc_convert_to_raw(db->db_buf, dsobj, byteorder, DMU_OT_DNODE, - salt, iv, mac); - - dmu_buf_rele(handle, FTAG); - - return (0); -} - void dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset, dmu_buf_t *handle, dmu_tx_t *tx) { dmu_buf_t *dst_handle; dmu_buf_impl_t *dstdb; dmu_buf_impl_t *srcdb = (dmu_buf_impl_t *)handle; arc_buf_t *abuf; uint64_t datalen; boolean_t byteorder; uint8_t salt[ZIO_DATA_SALT_LEN]; uint8_t iv[ZIO_DATA_IV_LEN]; uint8_t mac[ZIO_DATA_MAC_LEN]; ASSERT3P(srcdb->db_buf, !=, NULL); /* hold the db that we want to write to */ VERIFY0(dmu_buf_hold(os, object, offset, FTAG, &dst_handle, DMU_READ_NO_DECRYPT)); dstdb = (dmu_buf_impl_t *)dst_handle; datalen = arc_buf_size(srcdb->db_buf); /* allocated an arc buffer that matches the type of srcdb->db_buf */ if (arc_is_encrypted(srcdb->db_buf)) { arc_get_raw_params(srcdb->db_buf, &byteorder, salt, iv, mac); abuf = arc_loan_raw_buf(os->os_spa, dmu_objset_id(os), byteorder, salt, iv, mac, DB_DNODE(dstdb)->dn_type, datalen, arc_buf_lsize(srcdb->db_buf), arc_get_compression(srcdb->db_buf)); } else { /* we won't get a compressed db back from dmu_buf_hold() */ ASSERT3U(arc_get_compression(srcdb->db_buf), ==, ZIO_COMPRESS_OFF); abuf = arc_loan_buf(os->os_spa, DMU_OT_IS_METADATA(DB_DNODE(dstdb)->dn_type), datalen); } ASSERT3U(datalen, ==, arc_buf_size(abuf)); /* copy the data to the new buffer and assign it to the dstdb */ bcopy(srcdb->db_buf->b_data, abuf->b_data, datalen); dbuf_assign_arcbuf(dstdb, abuf, tx); dmu_buf_rele(dst_handle, FTAG); } /* * When possible directly assign passed loaned arc buffer to a dbuf. * If this is not possible copy the contents of passed arc buf via * dmu_write(). */ void dmu_assign_arcbuf_by_dnode(dnode_t *dn, uint64_t offset, arc_buf_t *buf, dmu_tx_t *tx) { dmu_buf_impl_t *db; objset_t *os = dn->dn_objset; uint64_t object = dn->dn_object; uint32_t blksz = (uint32_t)arc_buf_lsize(buf); uint64_t blkid; rw_enter(&dn->dn_struct_rwlock, RW_READER); blkid = dbuf_whichblock(dn, 0, offset); VERIFY((db = dbuf_hold(dn, blkid, FTAG)) != NULL); rw_exit(&dn->dn_struct_rwlock); /* * We can only assign if the offset is aligned, the arc buf is the * same size as the dbuf, and the dbuf is not metadata. */ if (offset == db->db.db_offset && blksz == db->db.db_size) { dbuf_assign_arcbuf(db, buf, tx); dbuf_rele(db, FTAG); } else { /* compressed bufs must always be assignable to their dbuf */ ASSERT3U(arc_get_compression(buf), ==, ZIO_COMPRESS_OFF); ASSERT(!(buf->b_flags & ARC_BUF_FLAG_COMPRESSED)); dbuf_rele(db, FTAG); dmu_write(os, object, offset, blksz, buf->b_data, tx); dmu_return_arcbuf(buf); XUIOSTAT_BUMP(xuiostat_wbuf_copied); } } void dmu_assign_arcbuf_by_dbuf(dmu_buf_t *handle, uint64_t offset, arc_buf_t *buf, dmu_tx_t *tx) { dmu_buf_impl_t *dbuf = (dmu_buf_impl_t *)handle; DB_DNODE_ENTER(dbuf); dmu_assign_arcbuf_by_dnode(DB_DNODE(dbuf), offset, buf, tx); DB_DNODE_EXIT(dbuf); } typedef struct { dbuf_dirty_record_t *dsa_dr; dmu_sync_cb_t *dsa_done; zgd_t *dsa_zgd; dmu_tx_t *dsa_tx; } dmu_sync_arg_t; /* ARGSUSED */ static void dmu_sync_ready(zio_t *zio, arc_buf_t *buf, void *varg) { dmu_sync_arg_t *dsa = varg; dmu_buf_t *db = dsa->dsa_zgd->zgd_db; blkptr_t *bp = zio->io_bp; if (zio->io_error == 0) { if (BP_IS_HOLE(bp)) { /* * A block of zeros may compress to a hole, but the * block size still needs to be known for replay. */ BP_SET_LSIZE(bp, db->db_size); } else if (!BP_IS_EMBEDDED(bp)) { ASSERT(BP_GET_LEVEL(bp) == 0); BP_SET_FILL(bp, 1); } } } static void dmu_sync_late_arrival_ready(zio_t *zio) { dmu_sync_ready(zio, NULL, zio->io_private); } /* ARGSUSED */ static void dmu_sync_done(zio_t *zio, arc_buf_t *buf, void *varg) { dmu_sync_arg_t *dsa = varg; dbuf_dirty_record_t *dr = dsa->dsa_dr; dmu_buf_impl_t *db = dr->dr_dbuf; mutex_enter(&db->db_mtx); ASSERT(dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC); if (zio->io_error == 0) { dr->dt.dl.dr_nopwrite = !!(zio->io_flags & ZIO_FLAG_NOPWRITE); if (dr->dt.dl.dr_nopwrite) { blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; uint8_t chksum = BP_GET_CHECKSUM(bp_orig); ASSERT(BP_EQUAL(bp, bp_orig)); VERIFY(BP_EQUAL(bp, db->db_blkptr)); ASSERT(zio->io_prop.zp_compress != ZIO_COMPRESS_OFF); VERIFY(zio_checksum_table[chksum].ci_flags & ZCHECKSUM_FLAG_NOPWRITE); } dr->dt.dl.dr_overridden_by = *zio->io_bp; dr->dt.dl.dr_override_state = DR_OVERRIDDEN; dr->dt.dl.dr_copies = zio->io_prop.zp_copies; /* * Old style holes are filled with all zeros, whereas * new-style holes maintain their lsize, type, level, * and birth time (see zio_write_compress). While we * need to reset the BP_SET_LSIZE() call that happened * in dmu_sync_ready for old style holes, we do *not* * want to wipe out the information contained in new * style holes. Thus, only zero out the block pointer if * it's an old style hole. */ if (BP_IS_HOLE(&dr->dt.dl.dr_overridden_by) && dr->dt.dl.dr_overridden_by.blk_birth == 0) BP_ZERO(&dr->dt.dl.dr_overridden_by); } else { dr->dt.dl.dr_override_state = DR_NOT_OVERRIDDEN; } cv_broadcast(&db->db_changed); mutex_exit(&db->db_mtx); dsa->dsa_done(dsa->dsa_zgd, zio->io_error); kmem_free(dsa, sizeof (*dsa)); } static void dmu_sync_late_arrival_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; dmu_sync_arg_t *dsa = zio->io_private; ASSERTV(blkptr_t *bp_orig = &zio->io_bp_orig); if (zio->io_error == 0 && !BP_IS_HOLE(bp)) { ASSERT(!(zio->io_flags & ZIO_FLAG_NOPWRITE)); ASSERT(BP_IS_HOLE(bp_orig) || !BP_EQUAL(bp, bp_orig)); ASSERT(zio->io_bp->blk_birth == zio->io_txg); ASSERT(zio->io_txg > spa_syncing_txg(zio->io_spa)); zio_free(zio->io_spa, zio->io_txg, zio->io_bp); } dmu_tx_commit(dsa->dsa_tx); dsa->dsa_done(dsa->dsa_zgd, zio->io_error); abd_put(zio->io_abd); kmem_free(dsa, sizeof (*dsa)); } static int dmu_sync_late_arrival(zio_t *pio, objset_t *os, dmu_sync_cb_t *done, zgd_t *zgd, zio_prop_t *zp, zbookmark_phys_t *zb) { dmu_sync_arg_t *dsa; dmu_tx_t *tx; tx = dmu_tx_create(os); dmu_tx_hold_space(tx, zgd->zgd_db->db_size); if (dmu_tx_assign(tx, TXG_WAIT) != 0) { dmu_tx_abort(tx); /* Make zl_get_data do txg_waited_synced() */ return (SET_ERROR(EIO)); } /* * In order to prevent the zgd's lwb from being free'd prior to * dmu_sync_late_arrival_done() being called, we have to ensure * the lwb's "max txg" takes this tx's txg into account. */ zil_lwb_add_txg(zgd->zgd_lwb, dmu_tx_get_txg(tx)); dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP); dsa->dsa_dr = NULL; dsa->dsa_done = done; dsa->dsa_zgd = zgd; dsa->dsa_tx = tx; /* * Since we are currently syncing this txg, it's nontrivial to * determine what BP to nopwrite against, so we disable nopwrite. * * When syncing, the db_blkptr is initially the BP of the previous * txg. We can not nopwrite against it because it will be changed * (this is similar to the non-late-arrival case where the dbuf is * dirty in a future txg). * * Then dbuf_write_ready() sets bp_blkptr to the location we will write. * We can not nopwrite against it because although the BP will not * (typically) be changed, the data has not yet been persisted to this * location. * * Finally, when dbuf_write_done() is called, it is theoretically * possible to always nopwrite, because the data that was written in * this txg is the same data that we are trying to write. However we * would need to check that this dbuf is not dirty in any future * txg's (as we do in the normal dmu_sync() path). For simplicity, we * don't nopwrite in this case. */ zp->zp_nopwrite = B_FALSE; zio_nowait(zio_write(pio, os->os_spa, dmu_tx_get_txg(tx), zgd->zgd_bp, abd_get_from_buf(zgd->zgd_db->db_data, zgd->zgd_db->db_size), zgd->zgd_db->db_size, zgd->zgd_db->db_size, zp, dmu_sync_late_arrival_ready, NULL, NULL, dmu_sync_late_arrival_done, dsa, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, zb)); return (0); } /* * Intent log support: sync the block associated with db to disk. * N.B. and XXX: the caller is responsible for making sure that the * data isn't changing while dmu_sync() is writing it. * * Return values: * * EEXIST: this txg has already been synced, so there's nothing to do. * The caller should not log the write. * * ENOENT: the block was dbuf_free_range()'d, so there's nothing to do. * The caller should not log the write. * * EALREADY: this block is already in the process of being synced. * The caller should track its progress (somehow). * * EIO: could not do the I/O. * The caller should do a txg_wait_synced(). * * 0: the I/O has been initiated. * The caller should log this blkptr in the done callback. * It is possible that the I/O will fail, in which case * the error will be reported to the done callback and * propagated to pio from zio_done(). */ int dmu_sync(zio_t *pio, uint64_t txg, dmu_sync_cb_t *done, zgd_t *zgd) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)zgd->zgd_db; objset_t *os = db->db_objset; dsl_dataset_t *ds = os->os_dsl_dataset; dbuf_dirty_record_t *dr; dmu_sync_arg_t *dsa; zbookmark_phys_t zb; zio_prop_t zp; dnode_t *dn; ASSERT(pio != NULL); ASSERT(txg != 0); /* dbuf is within the locked range */ ASSERT3U(db->db.db_offset, >=, zgd->zgd_rl->r_off); ASSERT3U(db->db.db_offset + db->db.db_size, <=, zgd->zgd_rl->r_off + zgd->zgd_rl->r_len); SET_BOOKMARK(&zb, ds->ds_object, db->db.db_object, db->db_level, db->db_blkid); DB_DNODE_ENTER(db); dn = DB_DNODE(db); dmu_write_policy(os, dn, db->db_level, WP_DMU_SYNC, &zp); DB_DNODE_EXIT(db); /* * If we're frozen (running ziltest), we always need to generate a bp. */ if (txg > spa_freeze_txg(os->os_spa)) return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb)); /* * Grabbing db_mtx now provides a barrier between dbuf_sync_leaf() * and us. If we determine that this txg is not yet syncing, * but it begins to sync a moment later, that's OK because the * sync thread will block in dbuf_sync_leaf() until we drop db_mtx. */ mutex_enter(&db->db_mtx); if (txg <= spa_last_synced_txg(os->os_spa)) { /* * This txg has already synced. There's nothing to do. */ mutex_exit(&db->db_mtx); return (SET_ERROR(EEXIST)); } if (txg <= spa_syncing_txg(os->os_spa)) { /* * This txg is currently syncing, so we can't mess with * the dirty record anymore; just write a new log block. */ mutex_exit(&db->db_mtx); return (dmu_sync_late_arrival(pio, os, done, zgd, &zp, &zb)); } dr = db->db_last_dirty; while (dr && dr->dr_txg != txg) dr = dr->dr_next; if (dr == NULL) { /* * There's no dr for this dbuf, so it must have been freed. * There's no need to log writes to freed blocks, so we're done. */ mutex_exit(&db->db_mtx); return (SET_ERROR(ENOENT)); } ASSERT(dr->dr_next == NULL || dr->dr_next->dr_txg < txg); if (db->db_blkptr != NULL) { /* * We need to fill in zgd_bp with the current blkptr so that * the nopwrite code can check if we're writing the same * data that's already on disk. We can only nopwrite if we * are sure that after making the copy, db_blkptr will not * change until our i/o completes. We ensure this by * holding the db_mtx, and only allowing nopwrite if the * block is not already dirty (see below). This is verified * by dmu_sync_done(), which VERIFYs that the db_blkptr has * not changed. */ *zgd->zgd_bp = *db->db_blkptr; } /* * Assume the on-disk data is X, the current syncing data (in * txg - 1) is Y, and the current in-memory data is Z (currently * in dmu_sync). * * We usually want to perform a nopwrite if X and Z are the * same. However, if Y is different (i.e. the BP is going to * change before this write takes effect), then a nopwrite will * be incorrect - we would override with X, which could have * been freed when Y was written. * * (Note that this is not a concern when we are nop-writing from * syncing context, because X and Y must be identical, because * all previous txgs have been synced.) * * Therefore, we disable nopwrite if the current BP could change * before this TXG. There are two ways it could change: by * being dirty (dr_next is non-NULL), or by being freed * (dnode_block_freed()). This behavior is verified by * zio_done(), which VERIFYs that the override BP is identical * to the on-disk BP. */ DB_DNODE_ENTER(db); dn = DB_DNODE(db); if (dr->dr_next != NULL || dnode_block_freed(dn, db->db_blkid)) zp.zp_nopwrite = B_FALSE; DB_DNODE_EXIT(db); ASSERT(dr->dr_txg == txg); if (dr->dt.dl.dr_override_state == DR_IN_DMU_SYNC || dr->dt.dl.dr_override_state == DR_OVERRIDDEN) { /* * We have already issued a sync write for this buffer, * or this buffer has already been synced. It could not * have been dirtied since, or we would have cleared the state. */ mutex_exit(&db->db_mtx); return (SET_ERROR(EALREADY)); } ASSERT(dr->dt.dl.dr_override_state == DR_NOT_OVERRIDDEN); dr->dt.dl.dr_override_state = DR_IN_DMU_SYNC; mutex_exit(&db->db_mtx); dsa = kmem_alloc(sizeof (dmu_sync_arg_t), KM_SLEEP); dsa->dsa_dr = dr; dsa->dsa_done = done; dsa->dsa_zgd = zgd; dsa->dsa_tx = NULL; zio_nowait(arc_write(pio, os->os_spa, txg, zgd->zgd_bp, dr->dt.dl.dr_data, DBUF_IS_L2CACHEABLE(db), &zp, dmu_sync_ready, NULL, NULL, dmu_sync_done, dsa, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL, &zb)); return (0); } int dmu_object_set_nlevels(objset_t *os, uint64_t object, int nlevels, dmu_tx_t *tx) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dnode_set_nlevels(dn, nlevels, tx); dnode_rele(dn, FTAG); return (err); } int dmu_object_set_blocksize(objset_t *os, uint64_t object, uint64_t size, int ibs, dmu_tx_t *tx) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); err = dnode_set_blksz(dn, size, ibs, tx); dnode_rele(dn, FTAG); return (err); } int dmu_object_set_maxblkid(objset_t *os, uint64_t object, uint64_t maxblkid, dmu_tx_t *tx) { dnode_t *dn; int err; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); rw_enter(&dn->dn_struct_rwlock, RW_WRITER); dnode_new_blkid(dn, maxblkid, tx, B_FALSE); rw_exit(&dn->dn_struct_rwlock); dnode_rele(dn, FTAG); return (0); } void dmu_object_set_checksum(objset_t *os, uint64_t object, uint8_t checksum, dmu_tx_t *tx) { dnode_t *dn; /* * Send streams include each object's checksum function. This * check ensures that the receiving system can understand the * checksum function transmitted. */ ASSERT3U(checksum, <, ZIO_CHECKSUM_LEGACY_FUNCTIONS); VERIFY0(dnode_hold(os, object, FTAG, &dn)); ASSERT3U(checksum, <, ZIO_CHECKSUM_FUNCTIONS); dn->dn_checksum = checksum; dnode_setdirty(dn, tx); dnode_rele(dn, FTAG); } void dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress, dmu_tx_t *tx) { dnode_t *dn; /* * Send streams include each object's compression function. This * check ensures that the receiving system can understand the * compression function transmitted. */ ASSERT3U(compress, <, ZIO_COMPRESS_LEGACY_FUNCTIONS); VERIFY0(dnode_hold(os, object, FTAG, &dn)); dn->dn_compress = compress; dnode_setdirty(dn, tx); dnode_rele(dn, FTAG); } -/* - * Dirty an object and set the dirty record's raw flag. This is used - * when writing raw data to an object that will not effect the - * encryption parameters, specifically during raw receives. - */ -int -dmu_object_dirty_raw(objset_t *os, uint64_t object, dmu_tx_t *tx) -{ - dnode_t *dn; - int err; - - err = dnode_hold(os, object, FTAG, &dn); - if (err) - return (err); - dmu_buf_will_change_crypt_params((dmu_buf_t *)dn->dn_dbuf, tx); - dnode_rele(dn, FTAG); - return (err); -} - /* * When the "redundant_metadata" property is set to "most", only indirect * blocks of this level and higher will have an additional ditto block. */ int zfs_redundant_metadata_most_ditto_level = 2; void dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp) { dmu_object_type_t type = dn ? dn->dn_type : DMU_OT_OBJSET; boolean_t ismd = (level > 0 || DMU_OT_IS_METADATA(type) || (wp & WP_SPILL)); enum zio_checksum checksum = os->os_checksum; enum zio_compress compress = os->os_compress; enum zio_checksum dedup_checksum = os->os_dedup_checksum; boolean_t dedup = B_FALSE; boolean_t nopwrite = B_FALSE; boolean_t dedup_verify = os->os_dedup_verify; boolean_t encrypt = B_FALSE; int copies = os->os_copies; /* * We maintain different write policies for each of the following * types of data: * 1. metadata * 2. preallocated blocks (i.e. level-0 blocks of a dump device) * 3. all other level 0 blocks */ if (ismd) { /* * XXX -- we should design a compression algorithm * that specializes in arrays of bps. */ compress = zio_compress_select(os->os_spa, ZIO_COMPRESS_ON, ZIO_COMPRESS_ON); /* * Metadata always gets checksummed. If the data * checksum is multi-bit correctable, and it's not a * ZBT-style checksum, then it's suitable for metadata * as well. Otherwise, the metadata checksum defaults * to fletcher4. */ if (!(zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_METADATA) || (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED)) checksum = ZIO_CHECKSUM_FLETCHER_4; if (os->os_redundant_metadata == ZFS_REDUNDANT_METADATA_ALL || (os->os_redundant_metadata == ZFS_REDUNDANT_METADATA_MOST && (level >= zfs_redundant_metadata_most_ditto_level || DMU_OT_IS_METADATA(type) || (wp & WP_SPILL)))) copies++; } else if (wp & WP_NOFILL) { ASSERT(level == 0); /* * If we're writing preallocated blocks, we aren't actually * writing them so don't set any policy properties. These * blocks are currently only used by an external subsystem * outside of zfs (i.e. dump) and not written by the zio * pipeline. */ compress = ZIO_COMPRESS_OFF; checksum = ZIO_CHECKSUM_OFF; } else { compress = zio_compress_select(os->os_spa, dn->dn_compress, compress); checksum = (dedup_checksum == ZIO_CHECKSUM_OFF) ? zio_checksum_select(dn->dn_checksum, checksum) : dedup_checksum; /* * Determine dedup setting. If we are in dmu_sync(), * we won't actually dedup now because that's all * done in syncing context; but we do want to use the * dedup checkum. If the checksum is not strong * enough to ensure unique signatures, force * dedup_verify. */ if (dedup_checksum != ZIO_CHECKSUM_OFF) { dedup = (wp & WP_DMU_SYNC) ? B_FALSE : B_TRUE; if (!(zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_DEDUP)) dedup_verify = B_TRUE; } /* * Enable nopwrite if we have secure enough checksum * algorithm (see comment in zio_nop_write) and * compression is enabled. We don't enable nopwrite if * dedup is enabled as the two features are mutually * exclusive. */ nopwrite = (!dedup && (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_NOPWRITE) && compress != ZIO_COMPRESS_OFF && zfs_nopwrite_enabled); } /* * All objects in an encrypted objset are protected from modification * via a MAC. Encrypted objects store their IV and salt in the last DVA * in the bp, so we cannot use all copies. Encrypted objects are also * not subject to nopwrite since writing the same data will still * result in a new ciphertext. Only encrypted blocks can be dedup'd * to avoid ambiguity in the dedup code since the DDT does not store * object types. */ if (os->os_encrypted && (wp & WP_NOFILL) == 0) { encrypt = B_TRUE; if (DMU_OT_IS_ENCRYPTED(type)) { copies = MIN(copies, SPA_DVAS_PER_BP - 1); nopwrite = B_FALSE; } else { dedup = B_FALSE; } if (level <= 0 && (type == DMU_OT_DNODE || type == DMU_OT_OBJSET)) { compress = ZIO_COMPRESS_EMPTY; } } zp->zp_compress = compress; zp->zp_checksum = checksum; zp->zp_type = (wp & WP_SPILL) ? dn->dn_bonustype : type; zp->zp_level = level; zp->zp_copies = MIN(copies, spa_max_replication(os->os_spa)); zp->zp_dedup = dedup; zp->zp_dedup_verify = dedup && dedup_verify; zp->zp_nopwrite = nopwrite; zp->zp_encrypt = encrypt; zp->zp_byteorder = ZFS_HOST_BYTEORDER; bzero(zp->zp_salt, ZIO_DATA_SALT_LEN); bzero(zp->zp_iv, ZIO_DATA_IV_LEN); bzero(zp->zp_mac, ZIO_DATA_MAC_LEN); ASSERT3U(zp->zp_compress, !=, ZIO_COMPRESS_INHERIT); } /* * This function is only called from zfs_holey_common() for zpl_llseek() * in order to determine the location of holes. In order to accurately * report holes all dirty data must be synced to disk. This causes extremely * poor performance when seeking for holes in a dirty file. As a compromise, * only provide hole data when the dnode is clean. When a dnode is dirty * report the dnode as having no holes which is always a safe thing to do. */ int dmu_offset_next(objset_t *os, uint64_t object, boolean_t hole, uint64_t *off) { dnode_t *dn; int i, err; boolean_t clean = B_TRUE; err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); /* * Check if dnode is dirty */ for (i = 0; i < TXG_SIZE; i++) { if (multilist_link_active(&dn->dn_dirty_link[i])) { clean = B_FALSE; break; } } /* * If compatibility option is on, sync any current changes before * we go trundling through the block pointers. */ if (!clean && zfs_dmu_offset_next_sync) { clean = B_TRUE; dnode_rele(dn, FTAG); txg_wait_synced(dmu_objset_pool(os), 0); err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); } if (clean) err = dnode_next_offset(dn, (hole ? DNODE_FIND_HOLE : 0), off, 1, 1, 0); else err = SET_ERROR(EBUSY); dnode_rele(dn, FTAG); return (err); } void __dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi) { dnode_phys_t *dnp = dn->dn_phys; doi->doi_data_block_size = dn->dn_datablksz; doi->doi_metadata_block_size = dn->dn_indblkshift ? 1ULL << dn->dn_indblkshift : 0; doi->doi_type = dn->dn_type; doi->doi_bonus_type = dn->dn_bonustype; doi->doi_bonus_size = dn->dn_bonuslen; doi->doi_dnodesize = dn->dn_num_slots << DNODE_SHIFT; doi->doi_indirection = dn->dn_nlevels; doi->doi_checksum = dn->dn_checksum; doi->doi_compress = dn->dn_compress; doi->doi_nblkptr = dn->dn_nblkptr; doi->doi_physical_blocks_512 = (DN_USED_BYTES(dnp) + 256) >> 9; doi->doi_max_offset = (dn->dn_maxblkid + 1) * dn->dn_datablksz; doi->doi_fill_count = 0; for (int i = 0; i < dnp->dn_nblkptr; i++) doi->doi_fill_count += BP_GET_FILL(&dnp->dn_blkptr[i]); } void dmu_object_info_from_dnode(dnode_t *dn, dmu_object_info_t *doi) { rw_enter(&dn->dn_struct_rwlock, RW_READER); mutex_enter(&dn->dn_mtx); __dmu_object_info_from_dnode(dn, doi); mutex_exit(&dn->dn_mtx); rw_exit(&dn->dn_struct_rwlock); } /* * Get information on a DMU object. * If doi is NULL, just indicates whether the object exists. */ int dmu_object_info(objset_t *os, uint64_t object, dmu_object_info_t *doi) { dnode_t *dn; int err = dnode_hold(os, object, FTAG, &dn); if (err) return (err); if (doi != NULL) dmu_object_info_from_dnode(dn, doi); dnode_rele(dn, FTAG); return (0); } /* * As above, but faster; can be used when you have a held dbuf in hand. */ void dmu_object_info_from_db(dmu_buf_t *db_fake, dmu_object_info_t *doi) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; DB_DNODE_ENTER(db); dmu_object_info_from_dnode(DB_DNODE(db), doi); DB_DNODE_EXIT(db); } /* * Faster still when you only care about the size. * This is specifically optimized for zfs_getattr(). */ void dmu_object_size_from_db(dmu_buf_t *db_fake, uint32_t *blksize, u_longlong_t *nblk512) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; DB_DNODE_ENTER(db); dn = DB_DNODE(db); *blksize = dn->dn_datablksz; /* add in number of slots used for the dnode itself */ *nblk512 = ((DN_USED_BYTES(dn->dn_phys) + SPA_MINBLOCKSIZE/2) >> SPA_MINBLOCKSHIFT) + dn->dn_num_slots; DB_DNODE_EXIT(db); } void dmu_object_dnsize_from_db(dmu_buf_t *db_fake, int *dnsize) { dmu_buf_impl_t *db = (dmu_buf_impl_t *)db_fake; dnode_t *dn; DB_DNODE_ENTER(db); dn = DB_DNODE(db); *dnsize = dn->dn_num_slots << DNODE_SHIFT; DB_DNODE_EXIT(db); } void byteswap_uint64_array(void *vbuf, size_t size) { uint64_t *buf = vbuf; size_t count = size >> 3; int i; ASSERT((size & 7) == 0); for (i = 0; i < count; i++) buf[i] = BSWAP_64(buf[i]); } void byteswap_uint32_array(void *vbuf, size_t size) { uint32_t *buf = vbuf; size_t count = size >> 2; int i; ASSERT((size & 3) == 0); for (i = 0; i < count; i++) buf[i] = BSWAP_32(buf[i]); } void byteswap_uint16_array(void *vbuf, size_t size) { uint16_t *buf = vbuf; size_t count = size >> 1; int i; ASSERT((size & 1) == 0); for (i = 0; i < count; i++) buf[i] = BSWAP_16(buf[i]); } /* ARGSUSED */ void byteswap_uint8_array(void *vbuf, size_t size) { } void dmu_init(void) { abd_init(); zfs_dbgmsg_init(); sa_cache_init(); xuio_stat_init(); dmu_objset_init(); dnode_init(); zfetch_init(); dmu_tx_init(); l2arc_init(); arc_init(); dbuf_init(); } void dmu_fini(void) { arc_fini(); /* arc depends on l2arc, so arc must go first */ l2arc_fini(); dmu_tx_fini(); zfetch_fini(); dbuf_fini(); dnode_fini(); dmu_objset_fini(); xuio_stat_fini(); sa_cache_fini(); zfs_dbgmsg_fini(); abd_fini(); } #if defined(_KERNEL) && defined(HAVE_SPL) EXPORT_SYMBOL(dmu_bonus_hold); EXPORT_SYMBOL(dmu_buf_hold_array_by_bonus); EXPORT_SYMBOL(dmu_buf_rele_array); EXPORT_SYMBOL(dmu_prefetch); EXPORT_SYMBOL(dmu_free_range); EXPORT_SYMBOL(dmu_free_long_range); -EXPORT_SYMBOL(dmu_free_long_range_raw); EXPORT_SYMBOL(dmu_free_long_object); -EXPORT_SYMBOL(dmu_free_long_object_raw); EXPORT_SYMBOL(dmu_read); EXPORT_SYMBOL(dmu_read_by_dnode); EXPORT_SYMBOL(dmu_write); EXPORT_SYMBOL(dmu_write_by_dnode); EXPORT_SYMBOL(dmu_prealloc); EXPORT_SYMBOL(dmu_object_info); EXPORT_SYMBOL(dmu_object_info_from_dnode); EXPORT_SYMBOL(dmu_object_info_from_db); EXPORT_SYMBOL(dmu_object_size_from_db); EXPORT_SYMBOL(dmu_object_dnsize_from_db); EXPORT_SYMBOL(dmu_object_set_nlevels); EXPORT_SYMBOL(dmu_object_set_blocksize); EXPORT_SYMBOL(dmu_object_set_maxblkid); EXPORT_SYMBOL(dmu_object_set_checksum); EXPORT_SYMBOL(dmu_object_set_compress); EXPORT_SYMBOL(dmu_write_policy); EXPORT_SYMBOL(dmu_sync); EXPORT_SYMBOL(dmu_request_arcbuf); EXPORT_SYMBOL(dmu_return_arcbuf); EXPORT_SYMBOL(dmu_assign_arcbuf_by_dnode); EXPORT_SYMBOL(dmu_assign_arcbuf_by_dbuf); EXPORT_SYMBOL(dmu_buf_hold); EXPORT_SYMBOL(dmu_ot); /* BEGIN CSTYLED */ module_param(zfs_nopwrite_enabled, int, 0644); MODULE_PARM_DESC(zfs_nopwrite_enabled, "Enable NOP writes"); module_param(zfs_per_txg_dirty_frees_percent, ulong, 0644); MODULE_PARM_DESC(zfs_per_txg_dirty_frees_percent, "percentage of dirtied blocks from frees in one TXG"); module_param(zfs_dmu_offset_next_sync, int, 0644); MODULE_PARM_DESC(zfs_dmu_offset_next_sync, "Enable forcing txg sync to find holes"); /* END CSTYLED */ #endif diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c index a44f485b7b60..f1813015a965 100644 --- a/module/zfs/dmu_objset.c +++ b/module/zfs/dmu_objset.c @@ -1,2997 +1,2998 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2017 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. * Copyright (c) 2014 Spectra Logic Corporation, All rights reserved. * Copyright (c) 2015, STRATO AG, Inc. All rights reserved. * Copyright (c) 2016 Actifio, Inc. All rights reserved. * Copyright 2017 Nexenta Systems, Inc. * Copyright (c) 2017 Open-E, Inc. All Rights Reserved. */ /* Portions Copyright 2010 Robert Milkowski */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * Needed to close a window in dnode_move() that allows the objset to be freed * before it can be safely accessed. */ krwlock_t os_lock; /* * Tunable to overwrite the maximum number of threads for the parallelization * of dmu_objset_find_dp, needed to speed up the import of pools with many * datasets. * Default is 4 times the number of leaf vdevs. */ int dmu_find_threads = 0; /* * Backfill lower metadnode objects after this many have been freed. * Backfilling negatively impacts object creation rates, so only do it * if there are enough holes to fill. */ int dmu_rescan_dnode_threshold = 1 << DN_MAX_INDBLKSHIFT; static char *upgrade_tag = "upgrade_tag"; static void dmu_objset_find_dp_cb(void *arg); static void dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb); static void dmu_objset_upgrade_stop(objset_t *os); void dmu_objset_init(void) { rw_init(&os_lock, NULL, RW_DEFAULT, NULL); } void dmu_objset_fini(void) { rw_destroy(&os_lock); } spa_t * dmu_objset_spa(objset_t *os) { return (os->os_spa); } zilog_t * dmu_objset_zil(objset_t *os) { return (os->os_zil); } dsl_pool_t * dmu_objset_pool(objset_t *os) { dsl_dataset_t *ds; if ((ds = os->os_dsl_dataset) != NULL && ds->ds_dir) return (ds->ds_dir->dd_pool); else return (spa_get_dsl(os->os_spa)); } dsl_dataset_t * dmu_objset_ds(objset_t *os) { return (os->os_dsl_dataset); } dmu_objset_type_t dmu_objset_type(objset_t *os) { return (os->os_phys->os_type); } void dmu_objset_name(objset_t *os, char *buf) { dsl_dataset_name(os->os_dsl_dataset, buf); } uint64_t dmu_objset_id(objset_t *os) { dsl_dataset_t *ds = os->os_dsl_dataset; return (ds ? ds->ds_object : 0); } uint64_t dmu_objset_dnodesize(objset_t *os) { return (os->os_dnodesize); } zfs_sync_type_t dmu_objset_syncprop(objset_t *os) { return (os->os_sync); } zfs_logbias_op_t dmu_objset_logbias(objset_t *os) { return (os->os_logbias); } static void checksum_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; /* * Inheritance should have been done by now. */ ASSERT(newval != ZIO_CHECKSUM_INHERIT); os->os_checksum = zio_checksum_select(newval, ZIO_CHECKSUM_ON_VALUE); } static void compression_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; /* * Inheritance and range checking should have been done by now. */ ASSERT(newval != ZIO_COMPRESS_INHERIT); os->os_compress = zio_compress_select(os->os_spa, newval, ZIO_COMPRESS_ON); } static void copies_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; /* * Inheritance and range checking should have been done by now. */ ASSERT(newval > 0); ASSERT(newval <= spa_max_replication(os->os_spa)); os->os_copies = newval; } static void dedup_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; spa_t *spa = os->os_spa; enum zio_checksum checksum; /* * Inheritance should have been done by now. */ ASSERT(newval != ZIO_CHECKSUM_INHERIT); checksum = zio_checksum_dedup_select(spa, newval, ZIO_CHECKSUM_OFF); os->os_dedup_checksum = checksum & ZIO_CHECKSUM_MASK; os->os_dedup_verify = !!(checksum & ZIO_CHECKSUM_VERIFY); } static void primary_cache_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; /* * Inheritance and range checking should have been done by now. */ ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || newval == ZFS_CACHE_METADATA); os->os_primary_cache = newval; } static void secondary_cache_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; /* * Inheritance and range checking should have been done by now. */ ASSERT(newval == ZFS_CACHE_ALL || newval == ZFS_CACHE_NONE || newval == ZFS_CACHE_METADATA); os->os_secondary_cache = newval; } static void sync_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; /* * Inheritance and range checking should have been done by now. */ ASSERT(newval == ZFS_SYNC_STANDARD || newval == ZFS_SYNC_ALWAYS || newval == ZFS_SYNC_DISABLED); os->os_sync = newval; if (os->os_zil) zil_set_sync(os->os_zil, newval); } static void redundant_metadata_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; /* * Inheritance and range checking should have been done by now. */ ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL || newval == ZFS_REDUNDANT_METADATA_MOST); os->os_redundant_metadata = newval; } static void dnodesize_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; switch (newval) { case ZFS_DNSIZE_LEGACY: os->os_dnodesize = DNODE_MIN_SIZE; break; case ZFS_DNSIZE_AUTO: /* * Choose a dnode size that will work well for most * workloads if the user specified "auto". Future code * improvements could dynamically select a dnode size * based on observed workload patterns. */ os->os_dnodesize = DNODE_MIN_SIZE * 2; break; case ZFS_DNSIZE_1K: case ZFS_DNSIZE_2K: case ZFS_DNSIZE_4K: case ZFS_DNSIZE_8K: case ZFS_DNSIZE_16K: os->os_dnodesize = newval; break; } } static void logbias_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; ASSERT(newval == ZFS_LOGBIAS_LATENCY || newval == ZFS_LOGBIAS_THROUGHPUT); os->os_logbias = newval; if (os->os_zil) zil_set_logbias(os->os_zil, newval); } static void recordsize_changed_cb(void *arg, uint64_t newval) { objset_t *os = arg; os->os_recordsize = newval; } void dmu_objset_byteswap(void *buf, size_t size) { objset_phys_t *osp = buf; ASSERT(size == OBJSET_PHYS_SIZE_V1 || size == OBJSET_PHYS_SIZE_V2 || size == sizeof (objset_phys_t)); dnode_byteswap(&osp->os_meta_dnode); byteswap_uint64_array(&osp->os_zil_header, sizeof (zil_header_t)); osp->os_type = BSWAP_64(osp->os_type); osp->os_flags = BSWAP_64(osp->os_flags); if (size >= OBJSET_PHYS_SIZE_V2) { dnode_byteswap(&osp->os_userused_dnode); dnode_byteswap(&osp->os_groupused_dnode); if (size >= sizeof (objset_phys_t)) dnode_byteswap(&osp->os_projectused_dnode); } } /* * The hash is a CRC-based hash of the objset_t pointer and the object number. */ static uint64_t dnode_hash(const objset_t *os, uint64_t obj) { uintptr_t osv = (uintptr_t)os; uint64_t crc = -1ULL; ASSERT(zfs_crc64_table[128] == ZFS_CRC64_POLY); /* * The low 6 bits of the pointer don't have much entropy, because * the objset_t is larger than 2^6 bytes long. */ crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (osv >> 6)) & 0xFF]; crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 0)) & 0xFF]; crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 8)) & 0xFF]; crc = (crc >> 8) ^ zfs_crc64_table[(crc ^ (obj >> 16)) & 0xFF]; crc ^= (osv>>14) ^ (obj>>24); return (crc); } unsigned int dnode_multilist_index_func(multilist_t *ml, void *obj) { dnode_t *dn = obj; return (dnode_hash(dn->dn_objset, dn->dn_object) % multilist_get_num_sublists(ml)); } /* * Instantiates the objset_t in-memory structure corresponding to the * objset_phys_t that's pointed to by the specified blkptr_t. */ int dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, objset_t **osp) { objset_t *os; int i, err; ASSERT(ds == NULL || MUTEX_HELD(&ds->ds_opening_lock)); /* * The $ORIGIN dataset (if it exists) doesn't have an associated * objset, so there's no reason to open it. The $ORIGIN dataset * will not exist on pools older than SPA_VERSION_ORIGIN. */ if (ds != NULL && spa_get_dsl(spa) != NULL && spa_get_dsl(spa)->dp_origin_snap != NULL) { ASSERT3P(ds->ds_dir, !=, spa_get_dsl(spa)->dp_origin_snap->ds_dir); } os = kmem_zalloc(sizeof (objset_t), KM_SLEEP); os->os_dsl_dataset = ds; os->os_spa = spa; os->os_rootbp = bp; if (!BP_IS_HOLE(os->os_rootbp)) { arc_flags_t aflags = ARC_FLAG_WAIT; zbookmark_phys_t zb; int size; enum zio_flag zio_flags = ZIO_FLAG_CANFAIL; SET_BOOKMARK(&zb, ds ? ds->ds_object : DMU_META_OBJSET, ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); if (DMU_OS_IS_L2CACHEABLE(os)) aflags |= ARC_FLAG_L2CACHE; if (ds != NULL && ds->ds_dir->dd_crypto_obj != 0) { ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); ASSERT(BP_IS_AUTHENTICATED(bp)); zio_flags |= ZIO_FLAG_RAW; } dprintf_bp(os->os_rootbp, "reading %s", ""); err = arc_read(NULL, spa, os->os_rootbp, arc_getbuf_func, &os->os_phys_buf, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb); if (err != 0) { kmem_free(os, sizeof (objset_t)); /* convert checksum errors into IO errors */ if (err == ECKSUM) err = SET_ERROR(EIO); return (err); } if (spa_version(spa) < SPA_VERSION_USERSPACE) size = OBJSET_PHYS_SIZE_V1; else if (!spa_feature_is_enabled(spa, SPA_FEATURE_PROJECT_QUOTA)) size = OBJSET_PHYS_SIZE_V2; else size = sizeof (objset_phys_t); /* Increase the blocksize if we are permitted. */ if (arc_buf_size(os->os_phys_buf) < size) { arc_buf_t *buf = arc_alloc_buf(spa, &os->os_phys_buf, ARC_BUFC_METADATA, size); bzero(buf->b_data, size); bcopy(os->os_phys_buf->b_data, buf->b_data, arc_buf_size(os->os_phys_buf)); arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); os->os_phys_buf = buf; } os->os_phys = os->os_phys_buf->b_data; os->os_flags = os->os_phys->os_flags; } else { int size = spa_version(spa) >= SPA_VERSION_USERSPACE ? sizeof (objset_phys_t) : OBJSET_PHYS_SIZE_V1; os->os_phys_buf = arc_alloc_buf(spa, &os->os_phys_buf, ARC_BUFC_METADATA, size); os->os_phys = os->os_phys_buf->b_data; bzero(os->os_phys, size); } /* * Note: the changed_cb will be called once before the register * func returns, thus changing the checksum/compression from the * default (fletcher2/off). Snapshots don't need to know about * checksum/compression/copies. */ if (ds != NULL) { boolean_t needlock = B_FALSE; os->os_encrypted = (ds->ds_dir->dd_crypto_obj != 0); /* * Note: it's valid to open the objset if the dataset is * long-held, in which case the pool_config lock will not * be held. */ if (!dsl_pool_config_held(dmu_objset_pool(os))) { needlock = B_TRUE; dsl_pool_config_enter(dmu_objset_pool(os), FTAG); } err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE), primary_cache_changed_cb, os); if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SECONDARYCACHE), secondary_cache_changed_cb, os); } if (!ds->ds_is_snapshot) { if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_CHECKSUM), checksum_changed_cb, os); } if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_COMPRESSION), compression_changed_cb, os); } if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_COPIES), copies_changed_cb, os); } if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_DEDUP), dedup_changed_cb, os); } if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_LOGBIAS), logbias_changed_cb, os); } if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_SYNC), sync_changed_cb, os); } if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name( ZFS_PROP_REDUNDANT_METADATA), redundant_metadata_changed_cb, os); } if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), recordsize_changed_cb, os); } if (err == 0) { err = dsl_prop_register(ds, zfs_prop_to_name(ZFS_PROP_DNODESIZE), dnodesize_changed_cb, os); } } if (needlock) dsl_pool_config_exit(dmu_objset_pool(os), FTAG); if (err != 0) { arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); kmem_free(os, sizeof (objset_t)); return (err); } } else { /* It's the meta-objset. */ os->os_checksum = ZIO_CHECKSUM_FLETCHER_4; os->os_compress = ZIO_COMPRESS_ON; os->os_encrypted = B_FALSE; os->os_copies = spa_max_replication(spa); os->os_dedup_checksum = ZIO_CHECKSUM_OFF; os->os_dedup_verify = B_FALSE; os->os_logbias = ZFS_LOGBIAS_LATENCY; os->os_sync = ZFS_SYNC_STANDARD; os->os_primary_cache = ZFS_CACHE_ALL; os->os_secondary_cache = ZFS_CACHE_ALL; os->os_dnodesize = DNODE_MIN_SIZE; } if (ds == NULL || !ds->ds_is_snapshot) os->os_zil_header = os->os_phys->os_zil_header; os->os_zil = zil_alloc(os, &os->os_zil_header); for (i = 0; i < TXG_SIZE; i++) { os->os_dirty_dnodes[i] = multilist_create(sizeof (dnode_t), offsetof(dnode_t, dn_dirty_link[i]), dnode_multilist_index_func); } list_create(&os->os_dnodes, sizeof (dnode_t), offsetof(dnode_t, dn_link)); list_create(&os->os_downgraded_dbufs, sizeof (dmu_buf_impl_t), offsetof(dmu_buf_impl_t, db_link)); list_link_init(&os->os_evicting_node); mutex_init(&os->os_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&os->os_userused_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&os->os_obj_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&os->os_user_ptr_lock, NULL, MUTEX_DEFAULT, NULL); os->os_obj_next_percpu_len = boot_ncpus; os->os_obj_next_percpu = kmem_zalloc(os->os_obj_next_percpu_len * sizeof (os->os_obj_next_percpu[0]), KM_SLEEP); dnode_special_open(os, &os->os_phys->os_meta_dnode, DMU_META_DNODE_OBJECT, &os->os_meta_dnode); if (OBJSET_BUF_HAS_USERUSED(os->os_phys_buf)) { dnode_special_open(os, &os->os_phys->os_userused_dnode, DMU_USERUSED_OBJECT, &os->os_userused_dnode); dnode_special_open(os, &os->os_phys->os_groupused_dnode, DMU_GROUPUSED_OBJECT, &os->os_groupused_dnode); if (OBJSET_BUF_HAS_PROJECTUSED(os->os_phys_buf)) dnode_special_open(os, &os->os_phys->os_projectused_dnode, DMU_PROJECTUSED_OBJECT, &os->os_projectused_dnode); } mutex_init(&os->os_upgrade_lock, NULL, MUTEX_DEFAULT, NULL); *osp = os; return (0); } int dmu_objset_from_ds(dsl_dataset_t *ds, objset_t **osp) { int err = 0; /* * We shouldn't be doing anything with dsl_dataset_t's unless the * pool_config lock is held, or the dataset is long-held. */ ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool) || dsl_dataset_long_held(ds)); mutex_enter(&ds->ds_opening_lock); if (ds->ds_objset == NULL) { objset_t *os; rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); err = dmu_objset_open_impl(dsl_dataset_get_spa(ds), ds, dsl_dataset_get_blkptr(ds), &os); rrw_exit(&ds->ds_bp_rwlock, FTAG); if (err == 0) { mutex_enter(&ds->ds_lock); ASSERT(ds->ds_objset == NULL); ds->ds_objset = os; mutex_exit(&ds->ds_lock); } } *osp = ds->ds_objset; mutex_exit(&ds->ds_opening_lock); return (err); } /* * Holds the pool while the objset is held. Therefore only one objset * can be held at a time. */ int dmu_objset_hold_flags(const char *name, boolean_t decrypt, void *tag, objset_t **osp) { dsl_pool_t *dp; dsl_dataset_t *ds; int err; ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; err = dsl_pool_hold(name, tag, &dp); if (err != 0) return (err); err = dsl_dataset_hold_flags(dp, name, flags, tag, &ds); if (err != 0) { dsl_pool_rele(dp, tag); return (err); } err = dmu_objset_from_ds(ds, osp); if (err != 0) { dsl_dataset_rele(ds, tag); dsl_pool_rele(dp, tag); } return (err); } int dmu_objset_hold(const char *name, void *tag, objset_t **osp) { return (dmu_objset_hold_flags(name, B_FALSE, tag, osp)); } static int dmu_objset_own_impl(dsl_dataset_t *ds, dmu_objset_type_t type, boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) { int err; err = dmu_objset_from_ds(ds, osp); if (err != 0) { return (err); } else if (type != DMU_OST_ANY && type != (*osp)->os_phys->os_type) { return (SET_ERROR(EINVAL)); } else if (!readonly && dsl_dataset_is_snapshot(ds)) { return (SET_ERROR(EROFS)); } else if (!readonly && decrypt && dsl_dir_incompatible_encryption_version(ds->ds_dir)) { return (SET_ERROR(EROFS)); } /* if we are decrypting, we can now check MACs in os->os_phys_buf */ if (decrypt && arc_is_unauthenticated((*osp)->os_phys_buf)) { zbookmark_phys_t zb; SET_BOOKMARK(&zb, ds->ds_object, ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); err = arc_untransform((*osp)->os_phys_buf, (*osp)->os_spa, &zb, B_FALSE); if (err != 0) return (err); ASSERT0(arc_is_unauthenticated((*osp)->os_phys_buf)); } return (0); } /* * dsl_pool must not be held when this is called. * Upon successful return, there will be a longhold on the dataset, * and the dsl_pool will not be held. */ int dmu_objset_own(const char *name, dmu_objset_type_t type, boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) { dsl_pool_t *dp; dsl_dataset_t *ds; int err; ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; err = dsl_pool_hold(name, FTAG, &dp); if (err != 0) return (err); err = dsl_dataset_own(dp, name, flags, tag, &ds); if (err != 0) { dsl_pool_rele(dp, FTAG); return (err); } err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp); if (err != 0) { dsl_dataset_disown(ds, flags, tag); dsl_pool_rele(dp, FTAG); return (err); } /* * User accounting requires the dataset to be decrypted and rw. * We also don't begin user accounting during claiming to help * speed up pool import times and to keep this txg reserved * completely for recovery work. */ if ((dmu_objset_userobjspace_upgradable(*osp) || dmu_objset_projectquota_upgradable(*osp)) && !readonly && !dp->dp_spa->spa_claiming && (ds->ds_dir->dd_crypto_obj == 0 || decrypt)) dmu_objset_id_quota_upgrade(*osp); dsl_pool_rele(dp, FTAG); return (0); } int dmu_objset_own_obj(dsl_pool_t *dp, uint64_t obj, dmu_objset_type_t type, boolean_t readonly, boolean_t decrypt, void *tag, objset_t **osp) { dsl_dataset_t *ds; int err; ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; err = dsl_dataset_own_obj(dp, obj, flags, tag, &ds); if (err != 0) return (err); err = dmu_objset_own_impl(ds, type, readonly, decrypt, tag, osp); if (err != 0) { dsl_dataset_disown(ds, flags, tag); return (err); } return (0); } void dmu_objset_rele_flags(objset_t *os, boolean_t decrypt, void *tag) { ds_hold_flags_t flags = (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0; dsl_pool_t *dp = dmu_objset_pool(os); dsl_dataset_rele_flags(os->os_dsl_dataset, flags, tag); dsl_pool_rele(dp, tag); } void dmu_objset_rele(objset_t *os, void *tag) { dmu_objset_rele_flags(os, B_FALSE, tag); } /* * When we are called, os MUST refer to an objset associated with a dataset * that is owned by 'tag'; that is, is held and long held by 'tag' and ds_owner * == tag. We will then release and reacquire ownership of the dataset while * holding the pool config_rwlock to avoid intervening namespace or ownership * changes may occur. * * This exists solely to accommodate zfs_ioc_userspace_upgrade()'s desire to * release the hold on its dataset and acquire a new one on the dataset of the * same name so that it can be partially torn down and reconstructed. */ void dmu_objset_refresh_ownership(dsl_dataset_t *ds, dsl_dataset_t **newds, boolean_t decrypt, void *tag) { dsl_pool_t *dp; char name[ZFS_MAX_DATASET_NAME_LEN]; VERIFY3P(ds, !=, NULL); VERIFY3P(ds->ds_owner, ==, tag); VERIFY(dsl_dataset_long_held(ds)); dsl_dataset_name(ds, name); dp = ds->ds_dir->dd_pool; dsl_pool_config_enter(dp, FTAG); dsl_dataset_disown(ds, decrypt, tag); VERIFY0(dsl_dataset_own(dp, name, (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0, tag, newds)); dsl_pool_config_exit(dp, FTAG); } void dmu_objset_disown(objset_t *os, boolean_t decrypt, void *tag) { /* * Stop upgrading thread */ dmu_objset_upgrade_stop(os); dsl_dataset_disown(os->os_dsl_dataset, (decrypt) ? DS_HOLD_FLAG_DECRYPT : 0, tag); } void dmu_objset_evict_dbufs(objset_t *os) { dnode_t *dn_marker; dnode_t *dn; dn_marker = kmem_alloc(sizeof (dnode_t), KM_SLEEP); mutex_enter(&os->os_lock); dn = list_head(&os->os_dnodes); while (dn != NULL) { /* * Skip dnodes without holds. We have to do this dance * because dnode_add_ref() only works if there is already a * hold. If the dnode has no holds, then it has no dbufs. */ if (dnode_add_ref(dn, FTAG)) { list_insert_after(&os->os_dnodes, dn, dn_marker); mutex_exit(&os->os_lock); dnode_evict_dbufs(dn); dnode_rele(dn, FTAG); mutex_enter(&os->os_lock); dn = list_next(&os->os_dnodes, dn_marker); list_remove(&os->os_dnodes, dn_marker); } else { dn = list_next(&os->os_dnodes, dn); } } mutex_exit(&os->os_lock); kmem_free(dn_marker, sizeof (dnode_t)); if (DMU_USERUSED_DNODE(os) != NULL) { if (DMU_PROJECTUSED_DNODE(os) != NULL) dnode_evict_dbufs(DMU_PROJECTUSED_DNODE(os)); dnode_evict_dbufs(DMU_GROUPUSED_DNODE(os)); dnode_evict_dbufs(DMU_USERUSED_DNODE(os)); } dnode_evict_dbufs(DMU_META_DNODE(os)); } /* * Objset eviction processing is split into into two pieces. * The first marks the objset as evicting, evicts any dbufs that * have a refcount of zero, and then queues up the objset for the * second phase of eviction. Once os->os_dnodes has been cleared by * dnode_buf_pageout()->dnode_destroy(), the second phase is executed. * The second phase closes the special dnodes, dequeues the objset from * the list of those undergoing eviction, and finally frees the objset. * * NOTE: Due to asynchronous eviction processing (invocation of * dnode_buf_pageout()), it is possible for the meta dnode for the * objset to have no holds even though os->os_dnodes is not empty. */ void dmu_objset_evict(objset_t *os) { dsl_dataset_t *ds = os->os_dsl_dataset; for (int t = 0; t < TXG_SIZE; t++) ASSERT(!dmu_objset_is_dirty(os, t)); if (ds) dsl_prop_unregister_all(ds, os); if (os->os_sa) sa_tear_down(os); dmu_objset_evict_dbufs(os); mutex_enter(&os->os_lock); spa_evicting_os_register(os->os_spa, os); if (list_is_empty(&os->os_dnodes)) { mutex_exit(&os->os_lock); dmu_objset_evict_done(os); } else { mutex_exit(&os->os_lock); } } void dmu_objset_evict_done(objset_t *os) { ASSERT3P(list_head(&os->os_dnodes), ==, NULL); dnode_special_close(&os->os_meta_dnode); if (DMU_USERUSED_DNODE(os)) { if (DMU_PROJECTUSED_DNODE(os)) dnode_special_close(&os->os_projectused_dnode); dnode_special_close(&os->os_userused_dnode); dnode_special_close(&os->os_groupused_dnode); } zil_free(os->os_zil); arc_buf_destroy(os->os_phys_buf, &os->os_phys_buf); /* * This is a barrier to prevent the objset from going away in * dnode_move() until we can safely ensure that the objset is still in * use. We consider the objset valid before the barrier and invalid * after the barrier. */ rw_enter(&os_lock, RW_READER); rw_exit(&os_lock); kmem_free(os->os_obj_next_percpu, os->os_obj_next_percpu_len * sizeof (os->os_obj_next_percpu[0])); mutex_destroy(&os->os_lock); mutex_destroy(&os->os_userused_lock); mutex_destroy(&os->os_obj_lock); mutex_destroy(&os->os_user_ptr_lock); mutex_destroy(&os->os_upgrade_lock); for (int i = 0; i < TXG_SIZE; i++) { multilist_destroy(os->os_dirty_dnodes[i]); } spa_evicting_os_deregister(os->os_spa, os); kmem_free(os, sizeof (objset_t)); } timestruc_t dmu_objset_snap_cmtime(objset_t *os) { return (dsl_dir_snap_cmtime(os->os_dsl_dataset->ds_dir)); } objset_t * dmu_objset_create_impl_dnstats(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, dmu_objset_type_t type, int levels, int blksz, int ibs, dmu_tx_t *tx) { objset_t *os; dnode_t *mdn; ASSERT(dmu_tx_is_syncing(tx)); if (blksz == 0) blksz = DNODE_BLOCK_SIZE; if (ibs == 0) ibs = DN_MAX_INDBLKSHIFT; if (ds != NULL) VERIFY0(dmu_objset_from_ds(ds, &os)); else VERIFY0(dmu_objset_open_impl(spa, NULL, bp, &os)); mdn = DMU_META_DNODE(os); dnode_allocate(mdn, DMU_OT_DNODE, blksz, ibs, DMU_OT_NONE, 0, DNODE_MIN_SLOTS, tx); /* * We don't want to have to increase the meta-dnode's nlevels * later, because then we could do it in quescing context while * we are also accessing it in open context. * * This precaution is not necessary for the MOS (ds == NULL), * because the MOS is only updated in syncing context. * This is most fortunate: the MOS is the only objset that * needs to be synced multiple times as spa_sync() iterates * to convergence, so minimizing its dn_nlevels matters. */ if (ds != NULL) { if (levels == 0) { levels = 1; /* * Determine the number of levels necessary for the * meta-dnode to contain DN_MAX_OBJECT dnodes. Note * that in order to ensure that we do not overflow * 64 bits, there has to be a nlevels that gives us a * number of blocks > DN_MAX_OBJECT but < 2^64. * Therefore, (mdn->dn_indblkshift - SPA_BLKPTRSHIFT) * (10) must be less than (64 - log2(DN_MAX_OBJECT)) * (16). */ while ((uint64_t)mdn->dn_nblkptr << (mdn->dn_datablkshift - DNODE_SHIFT + (levels - 1) * (mdn->dn_indblkshift - SPA_BLKPTRSHIFT)) < DN_MAX_OBJECT) levels++; } mdn->dn_next_nlevels[tx->tx_txg & TXG_MASK] = mdn->dn_nlevels = levels; } ASSERT(type != DMU_OST_NONE); ASSERT(type != DMU_OST_ANY); ASSERT(type < DMU_OST_NUMTYPES); os->os_phys->os_type = type; /* * Enable user accounting if it is enabled and this is not an * encrypted receive. */ if (dmu_objset_userused_enabled(os) && (!os->os_encrypted || !dmu_objset_is_receiving(os))) { os->os_phys->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; if (dmu_objset_userobjused_enabled(os)) { ds->ds_feature_activation_needed[ SPA_FEATURE_USEROBJ_ACCOUNTING] = B_TRUE; os->os_phys->os_flags |= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; } if (dmu_objset_projectquota_enabled(os)) { ds->ds_feature_activation_needed[ SPA_FEATURE_PROJECT_QUOTA] = B_TRUE; os->os_phys->os_flags |= OBJSET_FLAG_PROJECTQUOTA_COMPLETE; } os->os_flags = os->os_phys->os_flags; } dsl_dataset_dirty(ds, tx); return (os); } /* called from dsl for meta-objset */ objset_t * dmu_objset_create_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp, dmu_objset_type_t type, dmu_tx_t *tx) { return (dmu_objset_create_impl_dnstats(spa, ds, bp, type, 0, 0, 0, tx)); } typedef struct dmu_objset_create_arg { const char *doca_name; cred_t *doca_cred; void (*doca_userfunc)(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx); void *doca_userarg; dmu_objset_type_t doca_type; uint64_t doca_flags; dsl_crypto_params_t *doca_dcp; } dmu_objset_create_arg_t; /*ARGSUSED*/ static int dmu_objset_create_check(void *arg, dmu_tx_t *tx) { dmu_objset_create_arg_t *doca = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dir_t *pdd; const char *tail; int error; if (strchr(doca->doca_name, '@') != NULL) return (SET_ERROR(EINVAL)); if (strlen(doca->doca_name) >= ZFS_MAX_DATASET_NAME_LEN) return (SET_ERROR(ENAMETOOLONG)); error = dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail); if (error != 0) return (error); if (tail == NULL) { dsl_dir_rele(pdd, FTAG); return (SET_ERROR(EEXIST)); } error = dmu_objset_create_crypt_check(pdd, doca->doca_dcp); if (error != 0) { dsl_dir_rele(pdd, FTAG); return (error); } error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, doca->doca_cred); dsl_dir_rele(pdd, FTAG); return (error); } static void dmu_objset_create_sync(void *arg, dmu_tx_t *tx) { dmu_objset_create_arg_t *doca = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dir_t *pdd; const char *tail; dsl_dataset_t *ds; uint64_t obj; blkptr_t *bp; objset_t *os; zio_t *rzio; VERIFY0(dsl_dir_hold(dp, doca->doca_name, FTAG, &pdd, &tail)); obj = dsl_dataset_create_sync(pdd, tail, NULL, doca->doca_flags, doca->doca_cred, doca->doca_dcp, tx); VERIFY0(dsl_dataset_hold_obj_flags(pdd->dd_pool, obj, DS_HOLD_FLAG_DECRYPT, FTAG, &ds)); rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); bp = dsl_dataset_get_blkptr(ds); os = dmu_objset_create_impl(pdd->dd_pool->dp_spa, ds, bp, doca->doca_type, tx); rrw_exit(&ds->ds_bp_rwlock, FTAG); if (doca->doca_userfunc != NULL) { doca->doca_userfunc(os, doca->doca_userarg, doca->doca_cred, tx); } /* * The doca_userfunc() may write out some data that needs to be * encrypted if the dataset is encrypted (specifically the root * directory). This data must be written out before the encryption * key mapping is removed by dsl_dataset_rele_flags(). Force the * I/O to occur immediately by invoking the relevant sections of * dsl_pool_sync(). */ if (os->os_encrypted) { dsl_dataset_t *tmpds = NULL; boolean_t need_sync_done = B_FALSE; mutex_enter(&ds->ds_lock); ds->ds_owner = FTAG; mutex_exit(&ds->ds_lock); rzio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds, tx->tx_txg); if (tmpds != NULL) { dsl_dataset_sync(ds, rzio, tx); need_sync_done = B_TRUE; } VERIFY0(zio_wait(rzio)); dmu_objset_do_userquota_updates(os, tx); taskq_wait(dp->dp_sync_taskq); rzio = zio_root(dp->dp_spa, NULL, NULL, ZIO_FLAG_MUSTSUCCEED); tmpds = txg_list_remove_this(&dp->dp_dirty_datasets, ds, tx->tx_txg); if (tmpds != NULL) { dmu_buf_rele(ds->ds_dbuf, ds); dsl_dataset_sync(ds, rzio, tx); } VERIFY0(zio_wait(rzio)); if (need_sync_done) dsl_dataset_sync_done(ds, tx); mutex_enter(&ds->ds_lock); ds->ds_owner = NULL; mutex_exit(&ds->ds_lock); } spa_history_log_internal_ds(ds, "create", tx, ""); zvol_create_minors(dp->dp_spa, doca->doca_name, B_TRUE); dsl_dataset_rele_flags(ds, DS_HOLD_FLAG_DECRYPT, FTAG); dsl_dir_rele(pdd, FTAG); } int dmu_objset_create(const char *name, dmu_objset_type_t type, uint64_t flags, dsl_crypto_params_t *dcp, dmu_objset_create_sync_func_t func, void *arg) { dmu_objset_create_arg_t doca; dsl_crypto_params_t tmp_dcp = { 0 }; doca.doca_name = name; doca.doca_cred = CRED(); doca.doca_flags = flags; doca.doca_userfunc = func; doca.doca_userarg = arg; doca.doca_type = type; /* * Some callers (mostly for testing) do not provide a dcp on their * own but various code inside the sync task will require it to be * allocated. Rather than adding NULL checks throughout this code * or adding dummy dcp's to all of the callers we simply create a * dummy one here and use that. This zero dcp will have the same * effect as asking for inheritence of all encryption params. */ doca.doca_dcp = (dcp != NULL) ? dcp : &tmp_dcp; return (dsl_sync_task(name, dmu_objset_create_check, dmu_objset_create_sync, &doca, 6, ZFS_SPACE_CHECK_NORMAL)); } typedef struct dmu_objset_clone_arg { const char *doca_clone; const char *doca_origin; cred_t *doca_cred; } dmu_objset_clone_arg_t; /*ARGSUSED*/ static int dmu_objset_clone_check(void *arg, dmu_tx_t *tx) { dmu_objset_clone_arg_t *doca = arg; dsl_dir_t *pdd; const char *tail; int error; dsl_dataset_t *origin; dsl_pool_t *dp = dmu_tx_pool(tx); if (strchr(doca->doca_clone, '@') != NULL) return (SET_ERROR(EINVAL)); if (strlen(doca->doca_clone) >= ZFS_MAX_DATASET_NAME_LEN) return (SET_ERROR(ENAMETOOLONG)); error = dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail); if (error != 0) return (error); if (tail == NULL) { dsl_dir_rele(pdd, FTAG); return (SET_ERROR(EEXIST)); } error = dsl_fs_ss_limit_check(pdd, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, doca->doca_cred); if (error != 0) { dsl_dir_rele(pdd, FTAG); return (SET_ERROR(EDQUOT)); } error = dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin); if (error != 0) { dsl_dir_rele(pdd, FTAG); return (error); } /* You can only clone snapshots, not the head datasets. */ if (!origin->ds_is_snapshot) { dsl_dataset_rele(origin, FTAG); dsl_dir_rele(pdd, FTAG); return (SET_ERROR(EINVAL)); } error = dmu_objset_clone_crypt_check(pdd, origin->ds_dir); if (error != 0) { dsl_dataset_rele(origin, FTAG); dsl_dir_rele(pdd, FTAG); return (error); } dsl_dataset_rele(origin, FTAG); dsl_dir_rele(pdd, FTAG); return (0); } static void dmu_objset_clone_sync(void *arg, dmu_tx_t *tx) { dmu_objset_clone_arg_t *doca = arg; dsl_pool_t *dp = dmu_tx_pool(tx); dsl_dir_t *pdd; const char *tail; dsl_dataset_t *origin, *ds; uint64_t obj; char namebuf[ZFS_MAX_DATASET_NAME_LEN]; VERIFY0(dsl_dir_hold(dp, doca->doca_clone, FTAG, &pdd, &tail)); VERIFY0(dsl_dataset_hold(dp, doca->doca_origin, FTAG, &origin)); obj = dsl_dataset_create_sync(pdd, tail, origin, 0, doca->doca_cred, NULL, tx); VERIFY0(dsl_dataset_hold_obj(pdd->dd_pool, obj, FTAG, &ds)); dsl_dataset_name(origin, namebuf); spa_history_log_internal_ds(ds, "clone", tx, "origin=%s (%llu)", namebuf, origin->ds_object); zvol_create_minors(dp->dp_spa, doca->doca_clone, B_TRUE); dsl_dataset_rele(ds, FTAG); dsl_dataset_rele(origin, FTAG); dsl_dir_rele(pdd, FTAG); } int dmu_objset_clone(const char *clone, const char *origin) { dmu_objset_clone_arg_t doca; doca.doca_clone = clone; doca.doca_origin = origin; doca.doca_cred = CRED(); return (dsl_sync_task(clone, dmu_objset_clone_check, dmu_objset_clone_sync, &doca, 6, ZFS_SPACE_CHECK_NORMAL)); } static int dmu_objset_remap_indirects_impl(objset_t *os, uint64_t last_removed_txg) { int error = 0; uint64_t object = 0; while ((error = dmu_object_next(os, &object, B_FALSE, 0)) == 0) { error = dmu_object_remap_indirects(os, object, last_removed_txg); /* * If the ZPL removed the object before we managed to dnode_hold * it, we would get an ENOENT. If the ZPL declares its intent * to remove the object (dnode_free) before we manage to * dnode_hold it, we would get an EEXIST. In either case, we * want to continue remapping the other objects in the objset; * in all other cases, we want to break early. */ if (error != 0 && error != ENOENT && error != EEXIST) { break; } } if (error == ESRCH) { error = 0; } return (error); } int dmu_objset_remap_indirects(const char *fsname) { int error = 0; objset_t *os = NULL; uint64_t last_removed_txg; uint64_t remap_start_txg; dsl_dir_t *dd; error = dmu_objset_hold(fsname, FTAG, &os); if (error != 0) { return (error); } dd = dmu_objset_ds(os)->ds_dir; if (!spa_feature_is_enabled(dmu_objset_spa(os), SPA_FEATURE_OBSOLETE_COUNTS)) { dmu_objset_rele(os, FTAG); return (SET_ERROR(ENOTSUP)); } if (dsl_dataset_is_snapshot(dmu_objset_ds(os))) { dmu_objset_rele(os, FTAG); return (SET_ERROR(EINVAL)); } /* * If there has not been a removal, we're done. */ last_removed_txg = spa_get_last_removal_txg(dmu_objset_spa(os)); if (last_removed_txg == -1ULL) { dmu_objset_rele(os, FTAG); return (0); } /* * If we have remapped since the last removal, we're done. */ if (dsl_dir_is_zapified(dd)) { uint64_t last_remap_txg; if (zap_lookup(spa_meta_objset(dmu_objset_spa(os)), dd->dd_object, DD_FIELD_LAST_REMAP_TXG, sizeof (last_remap_txg), 1, &last_remap_txg) == 0 && last_remap_txg > last_removed_txg) { dmu_objset_rele(os, FTAG); return (0); } } dsl_dataset_long_hold(dmu_objset_ds(os), FTAG); dsl_pool_rele(dmu_objset_pool(os), FTAG); remap_start_txg = spa_last_synced_txg(dmu_objset_spa(os)); error = dmu_objset_remap_indirects_impl(os, last_removed_txg); if (error == 0) { /* * We update the last_remap_txg to be the start txg so that * we can guarantee that every block older than last_remap_txg * that can be remapped has been remapped. */ error = dsl_dir_update_last_remap_txg(dd, remap_start_txg); } dsl_dataset_long_rele(dmu_objset_ds(os), FTAG); dsl_dataset_rele(dmu_objset_ds(os), FTAG); return (error); } int dmu_objset_snapshot_one(const char *fsname, const char *snapname) { int err; char *longsnap = kmem_asprintf("%s@%s", fsname, snapname); nvlist_t *snaps = fnvlist_alloc(); fnvlist_add_boolean(snaps, longsnap); strfree(longsnap); err = dsl_dataset_snapshot(snaps, NULL, NULL); fnvlist_free(snaps); return (err); } static void dmu_objset_upgrade_task_cb(void *data) { objset_t *os = data; mutex_enter(&os->os_upgrade_lock); os->os_upgrade_status = EINTR; if (!os->os_upgrade_exit) { mutex_exit(&os->os_upgrade_lock); os->os_upgrade_status = os->os_upgrade_cb(os); mutex_enter(&os->os_upgrade_lock); } os->os_upgrade_exit = B_TRUE; os->os_upgrade_id = 0; mutex_exit(&os->os_upgrade_lock); dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); } static void dmu_objset_upgrade(objset_t *os, dmu_objset_upgrade_cb_t cb) { if (os->os_upgrade_id != 0) return; ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); dsl_dataset_long_hold(dmu_objset_ds(os), upgrade_tag); mutex_enter(&os->os_upgrade_lock); if (os->os_upgrade_id == 0 && os->os_upgrade_status == 0) { os->os_upgrade_exit = B_FALSE; os->os_upgrade_cb = cb; os->os_upgrade_id = taskq_dispatch( os->os_spa->spa_upgrade_taskq, dmu_objset_upgrade_task_cb, os, TQ_SLEEP); if (os->os_upgrade_id == TASKQID_INVALID) { dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); os->os_upgrade_status = ENOMEM; } } mutex_exit(&os->os_upgrade_lock); } static void dmu_objset_upgrade_stop(objset_t *os) { mutex_enter(&os->os_upgrade_lock); os->os_upgrade_exit = B_TRUE; if (os->os_upgrade_id != 0) { taskqid_t id = os->os_upgrade_id; os->os_upgrade_id = 0; mutex_exit(&os->os_upgrade_lock); if ((taskq_cancel_id(os->os_spa->spa_upgrade_taskq, id)) == 0) { dsl_dataset_long_rele(dmu_objset_ds(os), upgrade_tag); } txg_wait_synced(os->os_spa->spa_dsl_pool, 0); } else { mutex_exit(&os->os_upgrade_lock); } } static void dmu_objset_sync_dnodes(multilist_sublist_t *list, dmu_tx_t *tx) { dnode_t *dn; while ((dn = multilist_sublist_head(list)) != NULL) { ASSERT(dn->dn_object != DMU_META_DNODE_OBJECT); ASSERT(dn->dn_dbuf->db_data_pending); /* * Initialize dn_zio outside dnode_sync() because the * meta-dnode needs to set it ouside dnode_sync(). */ dn->dn_zio = dn->dn_dbuf->db_data_pending->dr_zio; ASSERT(dn->dn_zio); ASSERT3U(dn->dn_nlevels, <=, DN_MAX_LEVELS); multilist_sublist_remove(list, dn); /* * If we are not doing useraccounting (os_synced_dnodes == NULL) * we are done with this dnode for this txg. Unset dn_dirty_txg * if later txgs aren't dirtying it so that future holders do * not get a stale value. Otherwise, we will do this in * userquota_updates_task() when processing has completely * finished for this txg. */ multilist_t *newlist = dn->dn_objset->os_synced_dnodes; if (newlist != NULL) { (void) dnode_add_ref(dn, newlist); multilist_insert(newlist, dn); } else { mutex_enter(&dn->dn_mtx); if (dn->dn_dirty_txg == tx->tx_txg) dn->dn_dirty_txg = 0; mutex_exit(&dn->dn_mtx); } dnode_sync(dn, tx); } } /* ARGSUSED */ static void dmu_objset_write_ready(zio_t *zio, arc_buf_t *abuf, void *arg) { blkptr_t *bp = zio->io_bp; objset_t *os = arg; dnode_phys_t *dnp = &os->os_phys->os_meta_dnode; uint64_t fill = 0; ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_OBJSET); ASSERT0(BP_GET_LEVEL(bp)); /* * Update rootbp fill count: it should be the number of objects * allocated in the object set (not counting the "special" * objects that are stored in the objset_phys_t -- the meta * dnode and user/group/project accounting objects). */ for (int i = 0; i < dnp->dn_nblkptr; i++) fill += BP_GET_FILL(&dnp->dn_blkptr[i]); BP_SET_FILL(bp, fill); if (os->os_dsl_dataset != NULL) rrw_enter(&os->os_dsl_dataset->ds_bp_rwlock, RW_WRITER, FTAG); *os->os_rootbp = *bp; if (os->os_dsl_dataset != NULL) rrw_exit(&os->os_dsl_dataset->ds_bp_rwlock, FTAG); } /* ARGSUSED */ static void dmu_objset_write_done(zio_t *zio, arc_buf_t *abuf, void *arg) { blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; objset_t *os = arg; if (zio->io_flags & ZIO_FLAG_IO_REWRITE) { ASSERT(BP_EQUAL(bp, bp_orig)); } else { dsl_dataset_t *ds = os->os_dsl_dataset; dmu_tx_t *tx = os->os_synctx; (void) dsl_dataset_block_kill(ds, bp_orig, tx, B_TRUE); dsl_dataset_block_born(ds, bp, tx); } kmem_free(bp, sizeof (*bp)); } typedef struct sync_dnodes_arg { multilist_t *sda_list; int sda_sublist_idx; multilist_t *sda_newlist; dmu_tx_t *sda_tx; } sync_dnodes_arg_t; static void sync_dnodes_task(void *arg) { sync_dnodes_arg_t *sda = arg; multilist_sublist_t *ms = multilist_sublist_lock(sda->sda_list, sda->sda_sublist_idx); dmu_objset_sync_dnodes(ms, sda->sda_tx); multilist_sublist_unlock(ms); kmem_free(sda, sizeof (*sda)); } /* called from dsl */ void dmu_objset_sync(objset_t *os, zio_t *pio, dmu_tx_t *tx) { int txgoff; zbookmark_phys_t zb; zio_prop_t zp; zio_t *zio; list_t *list; dbuf_dirty_record_t *dr; blkptr_t *blkptr_copy = kmem_alloc(sizeof (*os->os_rootbp), KM_SLEEP); *blkptr_copy = *os->os_rootbp; dprintf_ds(os->os_dsl_dataset, "txg=%llu\n", tx->tx_txg); ASSERT(dmu_tx_is_syncing(tx)); /* XXX the write_done callback should really give us the tx... */ os->os_synctx = tx; if (os->os_dsl_dataset == NULL) { /* * This is the MOS. If we have upgraded, * spa_max_replication() could change, so reset * os_copies here. */ os->os_copies = spa_max_replication(os->os_spa); } /* * Create the root block IO */ SET_BOOKMARK(&zb, os->os_dsl_dataset ? os->os_dsl_dataset->ds_object : DMU_META_OBJSET, ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); arc_release(os->os_phys_buf, &os->os_phys_buf); dmu_write_policy(os, NULL, 0, 0, &zp); /* - * If we are either claiming the ZIL or doing a raw receive write out - * the os_phys_buf raw. Neither of these actions will effect the MAC - * at this point. + * If we are either claiming the ZIL or doing a raw receive, write + * out the os_phys_buf raw. Neither of these actions will effect the + * MAC at this point. */ - if (os->os_next_write_raw[tx->tx_txg & TXG_MASK]) { + if (os->os_raw_receive || + os->os_next_write_raw[tx->tx_txg & TXG_MASK]) { ASSERT(os->os_encrypted); os->os_next_write_raw[tx->tx_txg & TXG_MASK] = B_FALSE; arc_convert_to_raw(os->os_phys_buf, os->os_dsl_dataset->ds_object, ZFS_HOST_BYTEORDER, DMU_OT_OBJSET, NULL, NULL, NULL); } zio = arc_write(pio, os->os_spa, tx->tx_txg, blkptr_copy, os->os_phys_buf, DMU_OS_IS_L2CACHEABLE(os), &zp, dmu_objset_write_ready, NULL, NULL, dmu_objset_write_done, os, ZIO_PRIORITY_ASYNC_WRITE, ZIO_FLAG_MUSTSUCCEED, &zb); /* * Sync special dnodes - the parent IO for the sync is the root block */ DMU_META_DNODE(os)->dn_zio = zio; dnode_sync(DMU_META_DNODE(os), tx); os->os_phys->os_flags = os->os_flags; if (DMU_USERUSED_DNODE(os) && DMU_USERUSED_DNODE(os)->dn_type != DMU_OT_NONE) { DMU_USERUSED_DNODE(os)->dn_zio = zio; dnode_sync(DMU_USERUSED_DNODE(os), tx); DMU_GROUPUSED_DNODE(os)->dn_zio = zio; dnode_sync(DMU_GROUPUSED_DNODE(os), tx); } if (DMU_PROJECTUSED_DNODE(os) && DMU_PROJECTUSED_DNODE(os)->dn_type != DMU_OT_NONE) { DMU_PROJECTUSED_DNODE(os)->dn_zio = zio; dnode_sync(DMU_PROJECTUSED_DNODE(os), tx); } txgoff = tx->tx_txg & TXG_MASK; if (dmu_objset_userused_enabled(os) && (!os->os_encrypted || !dmu_objset_is_receiving(os))) { /* * We must create the list here because it uses the * dn_dirty_link[] of this txg. But it may already * exist because we call dsl_dataset_sync() twice per txg. */ if (os->os_synced_dnodes == NULL) { os->os_synced_dnodes = multilist_create(sizeof (dnode_t), offsetof(dnode_t, dn_dirty_link[txgoff]), dnode_multilist_index_func); } else { ASSERT3U(os->os_synced_dnodes->ml_offset, ==, offsetof(dnode_t, dn_dirty_link[txgoff])); } } for (int i = 0; i < multilist_get_num_sublists(os->os_dirty_dnodes[txgoff]); i++) { sync_dnodes_arg_t *sda = kmem_alloc(sizeof (*sda), KM_SLEEP); sda->sda_list = os->os_dirty_dnodes[txgoff]; sda->sda_sublist_idx = i; sda->sda_tx = tx; (void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq, sync_dnodes_task, sda, 0); /* callback frees sda */ } taskq_wait(dmu_objset_pool(os)->dp_sync_taskq); list = &DMU_META_DNODE(os)->dn_dirty_records[txgoff]; while ((dr = list_head(list)) != NULL) { ASSERT0(dr->dr_dbuf->db_level); list_remove(list, dr); if (dr->dr_zio) zio_nowait(dr->dr_zio); } /* Enable dnode backfill if enough objects have been freed. */ if (os->os_freed_dnodes >= dmu_rescan_dnode_threshold) { os->os_rescan_dnodes = B_TRUE; os->os_freed_dnodes = 0; } /* * Free intent log blocks up to this tx. */ zil_sync(os->os_zil, tx); os->os_phys->os_zil_header = os->os_zil_header; zio_nowait(zio); } boolean_t dmu_objset_is_dirty(objset_t *os, uint64_t txg) { return (!multilist_is_empty(os->os_dirty_dnodes[txg & TXG_MASK])); } static objset_used_cb_t *used_cbs[DMU_OST_NUMTYPES]; void dmu_objset_register_type(dmu_objset_type_t ost, objset_used_cb_t *cb) { used_cbs[ost] = cb; } boolean_t dmu_objset_userused_enabled(objset_t *os) { return (spa_version(os->os_spa) >= SPA_VERSION_USERSPACE && used_cbs[os->os_phys->os_type] != NULL && DMU_USERUSED_DNODE(os) != NULL); } boolean_t dmu_objset_userobjused_enabled(objset_t *os) { return (dmu_objset_userused_enabled(os) && spa_feature_is_enabled(os->os_spa, SPA_FEATURE_USEROBJ_ACCOUNTING)); } boolean_t dmu_objset_projectquota_enabled(objset_t *os) { return (used_cbs[os->os_phys->os_type] != NULL && DMU_PROJECTUSED_DNODE(os) != NULL && spa_feature_is_enabled(os->os_spa, SPA_FEATURE_PROJECT_QUOTA)); } typedef struct userquota_node { /* must be in the first filed, see userquota_update_cache() */ char uqn_id[20 + DMU_OBJACCT_PREFIX_LEN]; int64_t uqn_delta; avl_node_t uqn_node; } userquota_node_t; typedef struct userquota_cache { avl_tree_t uqc_user_deltas; avl_tree_t uqc_group_deltas; avl_tree_t uqc_project_deltas; } userquota_cache_t; static int userquota_compare(const void *l, const void *r) { const userquota_node_t *luqn = l; const userquota_node_t *ruqn = r; int rv; /* * NB: can only access uqn_id because userquota_update_cache() doesn't * pass in an entire userquota_node_t. */ rv = strcmp(luqn->uqn_id, ruqn->uqn_id); return (AVL_ISIGN(rv)); } static void do_userquota_cacheflush(objset_t *os, userquota_cache_t *cache, dmu_tx_t *tx) { void *cookie; userquota_node_t *uqn; ASSERT(dmu_tx_is_syncing(tx)); cookie = NULL; while ((uqn = avl_destroy_nodes(&cache->uqc_user_deltas, &cookie)) != NULL) { /* * os_userused_lock protects against concurrent calls to * zap_increment_int(). It's needed because zap_increment_int() * is not thread-safe (i.e. not atomic). */ mutex_enter(&os->os_userused_lock); VERIFY0(zap_increment(os, DMU_USERUSED_OBJECT, uqn->uqn_id, uqn->uqn_delta, tx)); mutex_exit(&os->os_userused_lock); kmem_free(uqn, sizeof (*uqn)); } avl_destroy(&cache->uqc_user_deltas); cookie = NULL; while ((uqn = avl_destroy_nodes(&cache->uqc_group_deltas, &cookie)) != NULL) { mutex_enter(&os->os_userused_lock); VERIFY0(zap_increment(os, DMU_GROUPUSED_OBJECT, uqn->uqn_id, uqn->uqn_delta, tx)); mutex_exit(&os->os_userused_lock); kmem_free(uqn, sizeof (*uqn)); } avl_destroy(&cache->uqc_group_deltas); if (dmu_objset_projectquota_enabled(os)) { cookie = NULL; while ((uqn = avl_destroy_nodes(&cache->uqc_project_deltas, &cookie)) != NULL) { mutex_enter(&os->os_userused_lock); VERIFY0(zap_increment(os, DMU_PROJECTUSED_OBJECT, uqn->uqn_id, uqn->uqn_delta, tx)); mutex_exit(&os->os_userused_lock); kmem_free(uqn, sizeof (*uqn)); } avl_destroy(&cache->uqc_project_deltas); } } static void userquota_update_cache(avl_tree_t *avl, const char *id, int64_t delta) { userquota_node_t *uqn; avl_index_t idx; ASSERT(strlen(id) < sizeof (uqn->uqn_id)); /* * Use id directly for searching because uqn_id is the first field of * userquota_node_t and fields after uqn_id won't be accessed in * avl_find(). */ uqn = avl_find(avl, (const void *)id, &idx); if (uqn == NULL) { uqn = kmem_zalloc(sizeof (*uqn), KM_SLEEP); strlcpy(uqn->uqn_id, id, sizeof (uqn->uqn_id)); avl_insert(avl, uqn, idx); } uqn->uqn_delta += delta; } static void do_userquota_update(objset_t *os, userquota_cache_t *cache, uint64_t used, uint64_t flags, uint64_t user, uint64_t group, uint64_t project, boolean_t subtract) { if (flags & DNODE_FLAG_USERUSED_ACCOUNTED) { int64_t delta = DNODE_MIN_SIZE + used; char name[20]; if (subtract) delta = -delta; (void) sprintf(name, "%llx", (longlong_t)user); userquota_update_cache(&cache->uqc_user_deltas, name, delta); (void) sprintf(name, "%llx", (longlong_t)group); userquota_update_cache(&cache->uqc_group_deltas, name, delta); if (dmu_objset_projectquota_enabled(os)) { (void) sprintf(name, "%llx", (longlong_t)project); userquota_update_cache(&cache->uqc_project_deltas, name, delta); } } } static void do_userobjquota_update(objset_t *os, userquota_cache_t *cache, uint64_t flags, uint64_t user, uint64_t group, uint64_t project, boolean_t subtract) { if (flags & DNODE_FLAG_USEROBJUSED_ACCOUNTED) { char name[20 + DMU_OBJACCT_PREFIX_LEN]; int delta = subtract ? -1 : 1; (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", (longlong_t)user); userquota_update_cache(&cache->uqc_user_deltas, name, delta); (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", (longlong_t)group); userquota_update_cache(&cache->uqc_group_deltas, name, delta); if (dmu_objset_projectquota_enabled(os)) { (void) snprintf(name, sizeof (name), DMU_OBJACCT_PREFIX "%llx", (longlong_t)project); userquota_update_cache(&cache->uqc_project_deltas, name, delta); } } } typedef struct userquota_updates_arg { objset_t *uua_os; int uua_sublist_idx; dmu_tx_t *uua_tx; } userquota_updates_arg_t; static void userquota_updates_task(void *arg) { userquota_updates_arg_t *uua = arg; objset_t *os = uua->uua_os; dmu_tx_t *tx = uua->uua_tx; dnode_t *dn; userquota_cache_t cache = { { 0 } }; multilist_sublist_t *list = multilist_sublist_lock(os->os_synced_dnodes, uua->uua_sublist_idx); ASSERT(multilist_sublist_head(list) == NULL || dmu_objset_userused_enabled(os)); avl_create(&cache.uqc_user_deltas, userquota_compare, sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); avl_create(&cache.uqc_group_deltas, userquota_compare, sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); if (dmu_objset_projectquota_enabled(os)) avl_create(&cache.uqc_project_deltas, userquota_compare, sizeof (userquota_node_t), offsetof(userquota_node_t, uqn_node)); while ((dn = multilist_sublist_head(list)) != NULL) { int flags; ASSERT(!DMU_OBJECT_IS_SPECIAL(dn->dn_object)); ASSERT(dn->dn_phys->dn_type == DMU_OT_NONE || dn->dn_phys->dn_flags & DNODE_FLAG_USERUSED_ACCOUNTED); flags = dn->dn_id_flags; ASSERT(flags); if (flags & DN_ID_OLD_EXIST) { do_userquota_update(os, &cache, dn->dn_oldused, dn->dn_oldflags, dn->dn_olduid, dn->dn_oldgid, dn->dn_oldprojid, B_TRUE); do_userobjquota_update(os, &cache, dn->dn_oldflags, dn->dn_olduid, dn->dn_oldgid, dn->dn_oldprojid, B_TRUE); } if (flags & DN_ID_NEW_EXIST) { do_userquota_update(os, &cache, DN_USED_BYTES(dn->dn_phys), dn->dn_phys->dn_flags, dn->dn_newuid, dn->dn_newgid, dn->dn_newprojid, B_FALSE); do_userobjquota_update(os, &cache, dn->dn_phys->dn_flags, dn->dn_newuid, dn->dn_newgid, dn->dn_newprojid, B_FALSE); } mutex_enter(&dn->dn_mtx); dn->dn_oldused = 0; dn->dn_oldflags = 0; if (dn->dn_id_flags & DN_ID_NEW_EXIST) { dn->dn_olduid = dn->dn_newuid; dn->dn_oldgid = dn->dn_newgid; dn->dn_oldprojid = dn->dn_newprojid; dn->dn_id_flags |= DN_ID_OLD_EXIST; if (dn->dn_bonuslen == 0) dn->dn_id_flags |= DN_ID_CHKED_SPILL; else dn->dn_id_flags |= DN_ID_CHKED_BONUS; } dn->dn_id_flags &= ~(DN_ID_NEW_EXIST); if (dn->dn_dirty_txg == spa_syncing_txg(os->os_spa)) dn->dn_dirty_txg = 0; mutex_exit(&dn->dn_mtx); multilist_sublist_remove(list, dn); dnode_rele(dn, os->os_synced_dnodes); } do_userquota_cacheflush(os, &cache, tx); multilist_sublist_unlock(list); kmem_free(uua, sizeof (*uua)); } void dmu_objset_do_userquota_updates(objset_t *os, dmu_tx_t *tx) { if (!dmu_objset_userused_enabled(os)) return; /* * If this is a raw receive just return and handle accounting * later when we have the keys loaded. We also don't do user * accounting during claiming since the datasets are not owned * for the duration of claiming and this txg should only be * used for recovery. */ if (os->os_encrypted && dmu_objset_is_receiving(os)) return; if (tx->tx_txg <= os->os_spa->spa_claim_max_txg) return; /* Allocate the user/group/project used objects if necessary. */ if (DMU_USERUSED_DNODE(os)->dn_type == DMU_OT_NONE) { VERIFY0(zap_create_claim(os, DMU_USERUSED_OBJECT, DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); VERIFY0(zap_create_claim(os, DMU_GROUPUSED_OBJECT, DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); } if (dmu_objset_projectquota_enabled(os) && DMU_PROJECTUSED_DNODE(os)->dn_type == DMU_OT_NONE) { VERIFY0(zap_create_claim(os, DMU_PROJECTUSED_OBJECT, DMU_OT_USERGROUP_USED, DMU_OT_NONE, 0, tx)); } for (int i = 0; i < multilist_get_num_sublists(os->os_synced_dnodes); i++) { userquota_updates_arg_t *uua = kmem_alloc(sizeof (*uua), KM_SLEEP); uua->uua_os = os; uua->uua_sublist_idx = i; uua->uua_tx = tx; /* note: caller does taskq_wait() */ (void) taskq_dispatch(dmu_objset_pool(os)->dp_sync_taskq, userquota_updates_task, uua, 0); /* callback frees uua */ } } /* * Returns a pointer to data to find uid/gid from * * If a dirty record for transaction group that is syncing can't * be found then NULL is returned. In the NULL case it is assumed * the uid/gid aren't changing. */ static void * dmu_objset_userquota_find_data(dmu_buf_impl_t *db, dmu_tx_t *tx) { dbuf_dirty_record_t *dr, **drp; void *data; if (db->db_dirtycnt == 0) return (db->db.db_data); /* Nothing is changing */ for (drp = &db->db_last_dirty; (dr = *drp) != NULL; drp = &dr->dr_next) if (dr->dr_txg == tx->tx_txg) break; if (dr == NULL) { data = NULL; } else { dnode_t *dn; DB_DNODE_ENTER(dr->dr_dbuf); dn = DB_DNODE(dr->dr_dbuf); if (dn->dn_bonuslen == 0 && dr->dr_dbuf->db_blkid == DMU_SPILL_BLKID) data = dr->dt.dl.dr_data->b_data; else data = dr->dt.dl.dr_data; DB_DNODE_EXIT(dr->dr_dbuf); } return (data); } void dmu_objset_userquota_get_ids(dnode_t *dn, boolean_t before, dmu_tx_t *tx) { objset_t *os = dn->dn_objset; void *data = NULL; dmu_buf_impl_t *db = NULL; uint64_t *user = NULL; uint64_t *group = NULL; uint64_t *project = NULL; int flags = dn->dn_id_flags; int error; boolean_t have_spill = B_FALSE; if (!dmu_objset_userused_enabled(dn->dn_objset)) return; /* * Raw receives introduce a problem with user accounting. Raw * receives cannot update the user accounting info because the * user ids and the sizes are encrypted. To guarantee that we * never end up with bad user accounting, we simply disable it * during raw receives. We also disable this for normal receives * so that an incremental raw receive may be done on top of an * existing non-raw receive. */ if (os->os_encrypted && dmu_objset_is_receiving(os)) return; if (before && (flags & (DN_ID_CHKED_BONUS|DN_ID_OLD_EXIST| DN_ID_CHKED_SPILL))) return; if (before && dn->dn_bonuslen != 0) data = DN_BONUS(dn->dn_phys); else if (!before && dn->dn_bonuslen != 0) { if (dn->dn_bonus) { db = dn->dn_bonus; mutex_enter(&db->db_mtx); data = dmu_objset_userquota_find_data(db, tx); } else { data = DN_BONUS(dn->dn_phys); } } else if (dn->dn_bonuslen == 0 && dn->dn_bonustype == DMU_OT_SA) { int rf = 0; if (RW_WRITE_HELD(&dn->dn_struct_rwlock)) rf |= DB_RF_HAVESTRUCT; error = dmu_spill_hold_by_dnode(dn, rf | DB_RF_MUST_SUCCEED, FTAG, (dmu_buf_t **)&db); ASSERT(error == 0); mutex_enter(&db->db_mtx); data = (before) ? db->db.db_data : dmu_objset_userquota_find_data(db, tx); have_spill = B_TRUE; } else { mutex_enter(&dn->dn_mtx); dn->dn_id_flags |= DN_ID_CHKED_BONUS; mutex_exit(&dn->dn_mtx); return; } if (before) { ASSERT(data); user = &dn->dn_olduid; group = &dn->dn_oldgid; project = &dn->dn_oldprojid; } else if (data) { user = &dn->dn_newuid; group = &dn->dn_newgid; project = &dn->dn_newprojid; } /* * Must always call the callback in case the object * type has changed and that type isn't an object type to track */ error = used_cbs[os->os_phys->os_type](dn->dn_bonustype, data, user, group, project); /* * Preserve existing uid/gid when the callback can't determine * what the new uid/gid are and the callback returned EEXIST. * The EEXIST error tells us to just use the existing uid/gid. * If we don't know what the old values are then just assign * them to 0, since that is a new file being created. */ if (!before && data == NULL && error == EEXIST) { if (flags & DN_ID_OLD_EXIST) { dn->dn_newuid = dn->dn_olduid; dn->dn_newgid = dn->dn_oldgid; dn->dn_newprojid = dn->dn_oldprojid; } else { dn->dn_newuid = 0; dn->dn_newgid = 0; dn->dn_newprojid = ZFS_DEFAULT_PROJID; } error = 0; } if (db) mutex_exit(&db->db_mtx); mutex_enter(&dn->dn_mtx); if (error == 0 && before) dn->dn_id_flags |= DN_ID_OLD_EXIST; if (error == 0 && !before) dn->dn_id_flags |= DN_ID_NEW_EXIST; if (have_spill) { dn->dn_id_flags |= DN_ID_CHKED_SPILL; } else { dn->dn_id_flags |= DN_ID_CHKED_BONUS; } mutex_exit(&dn->dn_mtx); if (have_spill) dmu_buf_rele((dmu_buf_t *)db, FTAG); } boolean_t dmu_objset_userspace_present(objset_t *os) { return (os->os_phys->os_flags & OBJSET_FLAG_USERACCOUNTING_COMPLETE); } boolean_t dmu_objset_userobjspace_present(objset_t *os) { return (os->os_phys->os_flags & OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE); } boolean_t dmu_objset_projectquota_present(objset_t *os) { return (os->os_phys->os_flags & OBJSET_FLAG_PROJECTQUOTA_COMPLETE); } static int dmu_objset_space_upgrade(objset_t *os) { uint64_t obj; int err = 0; /* * We simply need to mark every object dirty, so that it will be * synced out and now accounted. If this is called * concurrently, or if we already did some work before crashing, * that's fine, since we track each object's accounted state * independently. */ for (obj = 0; err == 0; err = dmu_object_next(os, &obj, FALSE, 0)) { dmu_tx_t *tx; dmu_buf_t *db; int objerr; mutex_enter(&os->os_upgrade_lock); if (os->os_upgrade_exit) err = SET_ERROR(EINTR); mutex_exit(&os->os_upgrade_lock); if (err != 0) return (err); if (issig(JUSTLOOKING) && issig(FORREAL)) return (SET_ERROR(EINTR)); objerr = dmu_bonus_hold(os, obj, FTAG, &db); if (objerr != 0) continue; tx = dmu_tx_create(os); dmu_tx_hold_bonus(tx, obj); objerr = dmu_tx_assign(tx, TXG_WAIT); if (objerr != 0) { dmu_buf_rele(db, FTAG); dmu_tx_abort(tx); continue; } dmu_buf_will_dirty(db, tx); dmu_buf_rele(db, FTAG); dmu_tx_commit(tx); } return (0); } int dmu_objset_userspace_upgrade(objset_t *os) { int err = 0; if (dmu_objset_userspace_present(os)) return (0); if (dmu_objset_is_snapshot(os)) return (SET_ERROR(EINVAL)); if (!dmu_objset_userused_enabled(os)) return (SET_ERROR(ENOTSUP)); err = dmu_objset_space_upgrade(os); if (err) return (err); os->os_flags |= OBJSET_FLAG_USERACCOUNTING_COMPLETE; txg_wait_synced(dmu_objset_pool(os), 0); return (0); } static int dmu_objset_id_quota_upgrade_cb(objset_t *os) { int err = 0; if (dmu_objset_userobjspace_present(os) && dmu_objset_projectquota_present(os)) return (0); if (dmu_objset_is_snapshot(os)) return (SET_ERROR(EINVAL)); if (!dmu_objset_userobjused_enabled(os)) return (SET_ERROR(ENOTSUP)); if (!dmu_objset_projectquota_enabled(os) && dmu_objset_userobjspace_present(os)) return (SET_ERROR(ENOTSUP)); dmu_objset_ds(os)->ds_feature_activation_needed[ SPA_FEATURE_USEROBJ_ACCOUNTING] = B_TRUE; if (dmu_objset_projectquota_enabled(os)) dmu_objset_ds(os)->ds_feature_activation_needed[ SPA_FEATURE_PROJECT_QUOTA] = B_TRUE; err = dmu_objset_space_upgrade(os); if (err) return (err); os->os_flags |= OBJSET_FLAG_USEROBJACCOUNTING_COMPLETE; if (dmu_objset_projectquota_enabled(os)) os->os_flags |= OBJSET_FLAG_PROJECTQUOTA_COMPLETE; txg_wait_synced(dmu_objset_pool(os), 0); return (0); } void dmu_objset_id_quota_upgrade(objset_t *os) { dmu_objset_upgrade(os, dmu_objset_id_quota_upgrade_cb); } boolean_t dmu_objset_userobjspace_upgradable(objset_t *os) { return (dmu_objset_type(os) == DMU_OST_ZFS && !dmu_objset_is_snapshot(os) && dmu_objset_userobjused_enabled(os) && !dmu_objset_userobjspace_present(os)); } boolean_t dmu_objset_projectquota_upgradable(objset_t *os) { return (dmu_objset_type(os) == DMU_OST_ZFS && !dmu_objset_is_snapshot(os) && dmu_objset_projectquota_enabled(os) && !dmu_objset_projectquota_present(os)); } void dmu_objset_space(objset_t *os, uint64_t *refdbytesp, uint64_t *availbytesp, uint64_t *usedobjsp, uint64_t *availobjsp) { dsl_dataset_space(os->os_dsl_dataset, refdbytesp, availbytesp, usedobjsp, availobjsp); } uint64_t dmu_objset_fsid_guid(objset_t *os) { return (dsl_dataset_fsid_guid(os->os_dsl_dataset)); } void dmu_objset_fast_stat(objset_t *os, dmu_objset_stats_t *stat) { stat->dds_type = os->os_phys->os_type; if (os->os_dsl_dataset) dsl_dataset_fast_stat(os->os_dsl_dataset, stat); } void dmu_objset_stats(objset_t *os, nvlist_t *nv) { ASSERT(os->os_dsl_dataset || os->os_phys->os_type == DMU_OST_META); if (os->os_dsl_dataset != NULL) dsl_dataset_stats(os->os_dsl_dataset, nv); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_TYPE, os->os_phys->os_type); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_USERACCOUNTING, dmu_objset_userspace_present(os)); } int dmu_objset_is_snapshot(objset_t *os) { if (os->os_dsl_dataset != NULL) return (os->os_dsl_dataset->ds_is_snapshot); else return (B_FALSE); } int dmu_snapshot_realname(objset_t *os, char *name, char *real, int maxlen, boolean_t *conflict) { dsl_dataset_t *ds = os->os_dsl_dataset; uint64_t ignored; if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) return (SET_ERROR(ENOENT)); return (zap_lookup_norm(ds->ds_dir->dd_pool->dp_meta_objset, dsl_dataset_phys(ds)->ds_snapnames_zapobj, name, 8, 1, &ignored, MT_NORMALIZE, real, maxlen, conflict)); } int dmu_snapshot_list_next(objset_t *os, int namelen, char *name, uint64_t *idp, uint64_t *offp, boolean_t *case_conflict) { dsl_dataset_t *ds = os->os_dsl_dataset; zap_cursor_t cursor; zap_attribute_t attr; ASSERT(dsl_pool_config_held(dmu_objset_pool(os))); if (dsl_dataset_phys(ds)->ds_snapnames_zapobj == 0) return (SET_ERROR(ENOENT)); zap_cursor_init_serialized(&cursor, ds->ds_dir->dd_pool->dp_meta_objset, dsl_dataset_phys(ds)->ds_snapnames_zapobj, *offp); if (zap_cursor_retrieve(&cursor, &attr) != 0) { zap_cursor_fini(&cursor); return (SET_ERROR(ENOENT)); } if (strlen(attr.za_name) + 1 > namelen) { zap_cursor_fini(&cursor); return (SET_ERROR(ENAMETOOLONG)); } (void) strcpy(name, attr.za_name); if (idp) *idp = attr.za_first_integer; if (case_conflict) *case_conflict = attr.za_normalization_conflict; zap_cursor_advance(&cursor); *offp = zap_cursor_serialize(&cursor); zap_cursor_fini(&cursor); return (0); } int dmu_snapshot_lookup(objset_t *os, const char *name, uint64_t *value) { return (dsl_dataset_snap_lookup(os->os_dsl_dataset, name, value)); } int dmu_dir_list_next(objset_t *os, int namelen, char *name, uint64_t *idp, uint64_t *offp) { dsl_dir_t *dd = os->os_dsl_dataset->ds_dir; zap_cursor_t cursor; zap_attribute_t attr; /* there is no next dir on a snapshot! */ if (os->os_dsl_dataset->ds_object != dsl_dir_phys(dd)->dd_head_dataset_obj) return (SET_ERROR(ENOENT)); zap_cursor_init_serialized(&cursor, dd->dd_pool->dp_meta_objset, dsl_dir_phys(dd)->dd_child_dir_zapobj, *offp); if (zap_cursor_retrieve(&cursor, &attr) != 0) { zap_cursor_fini(&cursor); return (SET_ERROR(ENOENT)); } if (strlen(attr.za_name) + 1 > namelen) { zap_cursor_fini(&cursor); return (SET_ERROR(ENAMETOOLONG)); } (void) strcpy(name, attr.za_name); if (idp) *idp = attr.za_first_integer; zap_cursor_advance(&cursor); *offp = zap_cursor_serialize(&cursor); zap_cursor_fini(&cursor); return (0); } typedef struct dmu_objset_find_ctx { taskq_t *dc_tq; dsl_pool_t *dc_dp; uint64_t dc_ddobj; char *dc_ddname; /* last component of ddobj's name */ int (*dc_func)(dsl_pool_t *, dsl_dataset_t *, void *); void *dc_arg; int dc_flags; kmutex_t *dc_error_lock; int *dc_error; } dmu_objset_find_ctx_t; static void dmu_objset_find_dp_impl(dmu_objset_find_ctx_t *dcp) { dsl_pool_t *dp = dcp->dc_dp; dsl_dir_t *dd; dsl_dataset_t *ds; zap_cursor_t zc; zap_attribute_t *attr; uint64_t thisobj; int err = 0; /* don't process if there already was an error */ if (*dcp->dc_error != 0) goto out; /* * Note: passing the name (dc_ddname) here is optional, but it * improves performance because we don't need to call * zap_value_search() to determine the name. */ err = dsl_dir_hold_obj(dp, dcp->dc_ddobj, dcp->dc_ddname, FTAG, &dd); if (err != 0) goto out; /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ if (dd->dd_myname[0] == '$') { dsl_dir_rele(dd, FTAG); goto out; } thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); /* * Iterate over all children. */ if (dcp->dc_flags & DS_FIND_CHILDREN) { for (zap_cursor_init(&zc, dp->dp_meta_objset, dsl_dir_phys(dd)->dd_child_dir_zapobj); zap_cursor_retrieve(&zc, attr) == 0; (void) zap_cursor_advance(&zc)) { ASSERT3U(attr->za_integer_length, ==, sizeof (uint64_t)); ASSERT3U(attr->za_num_integers, ==, 1); dmu_objset_find_ctx_t *child_dcp = kmem_alloc(sizeof (*child_dcp), KM_SLEEP); *child_dcp = *dcp; child_dcp->dc_ddobj = attr->za_first_integer; child_dcp->dc_ddname = spa_strdup(attr->za_name); if (dcp->dc_tq != NULL) (void) taskq_dispatch(dcp->dc_tq, dmu_objset_find_dp_cb, child_dcp, TQ_SLEEP); else dmu_objset_find_dp_impl(child_dcp); } zap_cursor_fini(&zc); } /* * Iterate over all snapshots. */ if (dcp->dc_flags & DS_FIND_SNAPSHOTS) { dsl_dataset_t *ds; err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); if (err == 0) { uint64_t snapobj; snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; dsl_dataset_rele(ds, FTAG); for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); zap_cursor_retrieve(&zc, attr) == 0; (void) zap_cursor_advance(&zc)) { ASSERT3U(attr->za_integer_length, ==, sizeof (uint64_t)); ASSERT3U(attr->za_num_integers, ==, 1); err = dsl_dataset_hold_obj(dp, attr->za_first_integer, FTAG, &ds); if (err != 0) break; err = dcp->dc_func(dp, ds, dcp->dc_arg); dsl_dataset_rele(ds, FTAG); if (err != 0) break; } zap_cursor_fini(&zc); } } kmem_free(attr, sizeof (zap_attribute_t)); if (err != 0) { dsl_dir_rele(dd, FTAG); goto out; } /* * Apply to self. */ err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); /* * Note: we hold the dir while calling dsl_dataset_hold_obj() so * that the dir will remain cached, and we won't have to re-instantiate * it (which could be expensive due to finding its name via * zap_value_search()). */ dsl_dir_rele(dd, FTAG); if (err != 0) goto out; err = dcp->dc_func(dp, ds, dcp->dc_arg); dsl_dataset_rele(ds, FTAG); out: if (err != 0) { mutex_enter(dcp->dc_error_lock); /* only keep first error */ if (*dcp->dc_error == 0) *dcp->dc_error = err; mutex_exit(dcp->dc_error_lock); } if (dcp->dc_ddname != NULL) spa_strfree(dcp->dc_ddname); kmem_free(dcp, sizeof (*dcp)); } static void dmu_objset_find_dp_cb(void *arg) { dmu_objset_find_ctx_t *dcp = arg; dsl_pool_t *dp = dcp->dc_dp; /* * We need to get a pool_config_lock here, as there are several * asssert(pool_config_held) down the stack. Getting a lock via * dsl_pool_config_enter is risky, as it might be stalled by a * pending writer. This would deadlock, as the write lock can * only be granted when our parent thread gives up the lock. * The _prio interface gives us priority over a pending writer. */ dsl_pool_config_enter_prio(dp, FTAG); dmu_objset_find_dp_impl(dcp); dsl_pool_config_exit(dp, FTAG); } /* * Find objsets under and including ddobj, call func(ds) on each. * The order for the enumeration is completely undefined. * func is called with dsl_pool_config held. */ int dmu_objset_find_dp(dsl_pool_t *dp, uint64_t ddobj, int func(dsl_pool_t *, dsl_dataset_t *, void *), void *arg, int flags) { int error = 0; taskq_t *tq = NULL; int ntasks; dmu_objset_find_ctx_t *dcp; kmutex_t err_lock; mutex_init(&err_lock, NULL, MUTEX_DEFAULT, NULL); dcp = kmem_alloc(sizeof (*dcp), KM_SLEEP); dcp->dc_tq = NULL; dcp->dc_dp = dp; dcp->dc_ddobj = ddobj; dcp->dc_ddname = NULL; dcp->dc_func = func; dcp->dc_arg = arg; dcp->dc_flags = flags; dcp->dc_error_lock = &err_lock; dcp->dc_error = &error; if ((flags & DS_FIND_SERIALIZE) || dsl_pool_config_held_writer(dp)) { /* * In case a write lock is held we can't make use of * parallelism, as down the stack of the worker threads * the lock is asserted via dsl_pool_config_held. * In case of a read lock this is solved by getting a read * lock in each worker thread, which isn't possible in case * of a writer lock. So we fall back to the synchronous path * here. * In the future it might be possible to get some magic into * dsl_pool_config_held in a way that it returns true for * the worker threads so that a single lock held from this * thread suffices. For now, stay single threaded. */ dmu_objset_find_dp_impl(dcp); mutex_destroy(&err_lock); return (error); } ntasks = dmu_find_threads; if (ntasks == 0) ntasks = vdev_count_leaves(dp->dp_spa) * 4; tq = taskq_create("dmu_objset_find", ntasks, maxclsyspri, ntasks, INT_MAX, 0); if (tq == NULL) { kmem_free(dcp, sizeof (*dcp)); mutex_destroy(&err_lock); return (SET_ERROR(ENOMEM)); } dcp->dc_tq = tq; /* dcp will be freed by task */ (void) taskq_dispatch(tq, dmu_objset_find_dp_cb, dcp, TQ_SLEEP); /* * PORTING: this code relies on the property of taskq_wait to wait * until no more tasks are queued and no more tasks are active. As * we always queue new tasks from within other tasks, task_wait * reliably waits for the full recursion to finish, even though we * enqueue new tasks after taskq_wait has been called. * On platforms other than illumos, taskq_wait may not have this * property. */ taskq_wait(tq); taskq_destroy(tq); mutex_destroy(&err_lock); return (error); } /* * Find all objsets under name, and for each, call 'func(child_name, arg)'. * The dp_config_rwlock must not be held when this is called, and it * will not be held when the callback is called. * Therefore this function should only be used when the pool is not changing * (e.g. in syncing context), or the callback can deal with the possible races. */ static int dmu_objset_find_impl(spa_t *spa, const char *name, int func(const char *, void *), void *arg, int flags) { dsl_dir_t *dd; dsl_pool_t *dp = spa_get_dsl(spa); dsl_dataset_t *ds; zap_cursor_t zc; zap_attribute_t *attr; char *child; uint64_t thisobj; int err; dsl_pool_config_enter(dp, FTAG); err = dsl_dir_hold(dp, name, FTAG, &dd, NULL); if (err != 0) { dsl_pool_config_exit(dp, FTAG); return (err); } /* Don't visit hidden ($MOS & $ORIGIN) objsets. */ if (dd->dd_myname[0] == '$') { dsl_dir_rele(dd, FTAG); dsl_pool_config_exit(dp, FTAG); return (0); } thisobj = dsl_dir_phys(dd)->dd_head_dataset_obj; attr = kmem_alloc(sizeof (zap_attribute_t), KM_SLEEP); /* * Iterate over all children. */ if (flags & DS_FIND_CHILDREN) { for (zap_cursor_init(&zc, dp->dp_meta_objset, dsl_dir_phys(dd)->dd_child_dir_zapobj); zap_cursor_retrieve(&zc, attr) == 0; (void) zap_cursor_advance(&zc)) { ASSERT3U(attr->za_integer_length, ==, sizeof (uint64_t)); ASSERT3U(attr->za_num_integers, ==, 1); child = kmem_asprintf("%s/%s", name, attr->za_name); dsl_pool_config_exit(dp, FTAG); err = dmu_objset_find_impl(spa, child, func, arg, flags); dsl_pool_config_enter(dp, FTAG); strfree(child); if (err != 0) break; } zap_cursor_fini(&zc); if (err != 0) { dsl_dir_rele(dd, FTAG); dsl_pool_config_exit(dp, FTAG); kmem_free(attr, sizeof (zap_attribute_t)); return (err); } } /* * Iterate over all snapshots. */ if (flags & DS_FIND_SNAPSHOTS) { err = dsl_dataset_hold_obj(dp, thisobj, FTAG, &ds); if (err == 0) { uint64_t snapobj; snapobj = dsl_dataset_phys(ds)->ds_snapnames_zapobj; dsl_dataset_rele(ds, FTAG); for (zap_cursor_init(&zc, dp->dp_meta_objset, snapobj); zap_cursor_retrieve(&zc, attr) == 0; (void) zap_cursor_advance(&zc)) { ASSERT3U(attr->za_integer_length, ==, sizeof (uint64_t)); ASSERT3U(attr->za_num_integers, ==, 1); child = kmem_asprintf("%s@%s", name, attr->za_name); dsl_pool_config_exit(dp, FTAG); err = func(child, arg); dsl_pool_config_enter(dp, FTAG); strfree(child); if (err != 0) break; } zap_cursor_fini(&zc); } } dsl_dir_rele(dd, FTAG); kmem_free(attr, sizeof (zap_attribute_t)); dsl_pool_config_exit(dp, FTAG); if (err != 0) return (err); /* Apply to self. */ return (func(name, arg)); } /* * See comment above dmu_objset_find_impl(). */ int dmu_objset_find(char *name, int func(const char *, void *), void *arg, int flags) { spa_t *spa; int error; error = spa_open(name, &spa, FTAG); if (error != 0) return (error); error = dmu_objset_find_impl(spa, name, func, arg, flags); spa_close(spa, FTAG); return (error); } boolean_t dmu_objset_incompatible_encryption_version(objset_t *os) { return (dsl_dir_incompatible_encryption_version( os->os_dsl_dataset->ds_dir)); } void dmu_objset_set_user(objset_t *os, void *user_ptr) { ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); os->os_user_ptr = user_ptr; } void * dmu_objset_get_user(objset_t *os) { ASSERT(MUTEX_HELD(&os->os_user_ptr_lock)); return (os->os_user_ptr); } /* * Determine name of filesystem, given name of snapshot. * buf must be at least ZFS_MAX_DATASET_NAME_LEN bytes */ int dmu_fsname(const char *snapname, char *buf) { char *atp = strchr(snapname, '@'); if (atp == NULL) return (SET_ERROR(EINVAL)); if (atp - snapname >= ZFS_MAX_DATASET_NAME_LEN) return (SET_ERROR(ENAMETOOLONG)); (void) strlcpy(buf, snapname, atp - snapname + 1); return (0); } /* * Call when we think we're going to write/free space in open context to track * the amount of dirty data in the open txg, which is also the amount * of memory that can not be evicted until this txg syncs. */ void dmu_objset_willuse_space(objset_t *os, int64_t space, dmu_tx_t *tx) { dsl_dataset_t *ds = os->os_dsl_dataset; int64_t aspace = spa_get_worst_case_asize(os->os_spa, space); if (ds != NULL) { dsl_dir_willuse_space(ds->ds_dir, aspace, tx); dsl_pool_dirty_space(dmu_tx_pool(tx), space, tx); } } #if defined(_KERNEL) && defined(HAVE_SPL) EXPORT_SYMBOL(dmu_objset_zil); EXPORT_SYMBOL(dmu_objset_pool); EXPORT_SYMBOL(dmu_objset_ds); EXPORT_SYMBOL(dmu_objset_type); EXPORT_SYMBOL(dmu_objset_name); EXPORT_SYMBOL(dmu_objset_hold); EXPORT_SYMBOL(dmu_objset_hold_flags); EXPORT_SYMBOL(dmu_objset_own); EXPORT_SYMBOL(dmu_objset_rele); EXPORT_SYMBOL(dmu_objset_rele_flags); EXPORT_SYMBOL(dmu_objset_disown); EXPORT_SYMBOL(dmu_objset_from_ds); EXPORT_SYMBOL(dmu_objset_create); EXPORT_SYMBOL(dmu_objset_clone); EXPORT_SYMBOL(dmu_objset_stats); EXPORT_SYMBOL(dmu_objset_fast_stat); EXPORT_SYMBOL(dmu_objset_spa); EXPORT_SYMBOL(dmu_objset_space); EXPORT_SYMBOL(dmu_objset_fsid_guid); EXPORT_SYMBOL(dmu_objset_find); EXPORT_SYMBOL(dmu_objset_byteswap); EXPORT_SYMBOL(dmu_objset_evict_dbufs); EXPORT_SYMBOL(dmu_objset_snap_cmtime); EXPORT_SYMBOL(dmu_objset_dnodesize); EXPORT_SYMBOL(dmu_objset_sync); EXPORT_SYMBOL(dmu_objset_is_dirty); EXPORT_SYMBOL(dmu_objset_create_impl_dnstats); EXPORT_SYMBOL(dmu_objset_create_impl); EXPORT_SYMBOL(dmu_objset_open_impl); EXPORT_SYMBOL(dmu_objset_evict); EXPORT_SYMBOL(dmu_objset_register_type); EXPORT_SYMBOL(dmu_objset_do_userquota_updates); EXPORT_SYMBOL(dmu_objset_userquota_get_ids); EXPORT_SYMBOL(dmu_objset_userused_enabled); EXPORT_SYMBOL(dmu_objset_userspace_upgrade); EXPORT_SYMBOL(dmu_objset_userspace_present); EXPORT_SYMBOL(dmu_objset_userobjused_enabled); EXPORT_SYMBOL(dmu_objset_userobjspace_upgradable); EXPORT_SYMBOL(dmu_objset_userobjspace_present); EXPORT_SYMBOL(dmu_objset_projectquota_enabled); EXPORT_SYMBOL(dmu_objset_projectquota_present); EXPORT_SYMBOL(dmu_objset_projectquota_upgradable); EXPORT_SYMBOL(dmu_objset_id_quota_upgrade); #endif diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c index a007e96baa33..6c535e5411d8 100644 --- a/module/zfs/dmu_send.c +++ b/module/zfs/dmu_send.c @@ -1,4253 +1,4231 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2011, 2015 by Delphix. All rights reserved. * Copyright (c) 2014, Joyent, Inc. All rights reserved. * Copyright 2014 HybridCluster. All rights reserved. * Copyright 2016 RackTop Systems. * Copyright (c) 2016 Actifio, Inc. All rights reserved. */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* Set this tunable to TRUE to replace corrupt data with 0x2f5baddb10c */ int zfs_send_corrupt_data = B_FALSE; int zfs_send_queue_length = SPA_MAXBLOCKSIZE; int zfs_recv_queue_length = SPA_MAXBLOCKSIZE; /* Set this tunable to FALSE to disable setting of DRR_FLAG_FREERECORDS */ int zfs_send_set_freerecords_bit = B_TRUE; static char *dmu_recv_tag = "dmu_recv_tag"; const char *recv_clone_name = "%recv"; #define BP_SPAN(datablkszsec, indblkshift, level) \ (((uint64_t)datablkszsec) << (SPA_MINBLOCKSHIFT + \ (level) * (indblkshift - SPA_BLKPTRSHIFT))) static void byteswap_record(dmu_replay_record_t *drr); struct send_thread_arg { bqueue_t q; dsl_dataset_t *ds; /* Dataset to traverse */ uint64_t fromtxg; /* Traverse from this txg */ int flags; /* flags to pass to traverse_dataset */ int error_code; boolean_t cancel; zbookmark_phys_t resume; }; struct send_block_record { boolean_t eos_marker; /* Marks the end of the stream */ blkptr_t bp; zbookmark_phys_t zb; uint8_t indblkshift; uint16_t datablkszsec; bqueue_node_t ln; }; typedef struct dump_bytes_io { dmu_sendarg_t *dbi_dsp; void *dbi_buf; int dbi_len; } dump_bytes_io_t; static void dump_bytes_cb(void *arg) { dump_bytes_io_t *dbi = (dump_bytes_io_t *)arg; dmu_sendarg_t *dsp = dbi->dbi_dsp; dsl_dataset_t *ds = dmu_objset_ds(dsp->dsa_os); ssize_t resid; /* have to get resid to get detailed errno */ /* * The code does not rely on len being a multiple of 8. We keep * this assertion because of the corresponding assertion in * receive_read(). Keeping this assertion ensures that we do not * inadvertently break backwards compatibility (causing the assertion * in receive_read() to trigger on old software). Newer feature flags * (such as raw send) may break this assertion since they were * introduced after the requirement was made obsolete. */ ASSERT(dbi->dbi_len % 8 == 0 || (dsp->dsa_featureflags & DMU_BACKUP_FEATURE_RAW) != 0); dsp->dsa_err = vn_rdwr(UIO_WRITE, dsp->dsa_vp, (caddr_t)dbi->dbi_buf, dbi->dbi_len, 0, UIO_SYSSPACE, FAPPEND, RLIM64_INFINITY, CRED(), &resid); mutex_enter(&ds->ds_sendstream_lock); *dsp->dsa_off += dbi->dbi_len; mutex_exit(&ds->ds_sendstream_lock); } static int dump_bytes(dmu_sendarg_t *dsp, void *buf, int len) { dump_bytes_io_t dbi; dbi.dbi_dsp = dsp; dbi.dbi_buf = buf; dbi.dbi_len = len; #if defined(HAVE_LARGE_STACKS) dump_bytes_cb(&dbi); #else /* * The vn_rdwr() call is performed in a taskq to ensure that there is * always enough stack space to write safely to the target filesystem. * The ZIO_TYPE_FREE threads are used because there can be a lot of * them and they are used in vdev_file.c for a similar purpose. */ spa_taskq_dispatch_sync(dmu_objset_spa(dsp->dsa_os), ZIO_TYPE_FREE, ZIO_TASKQ_ISSUE, dump_bytes_cb, &dbi, TQ_SLEEP); #endif /* HAVE_LARGE_STACKS */ return (dsp->dsa_err); } /* * For all record types except BEGIN, fill in the checksum (overlaid in * drr_u.drr_checksum.drr_checksum). The checksum verifies everything * up to the start of the checksum itself. */ static int dump_record(dmu_sendarg_t *dsp, void *payload, int payload_len) { ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), ==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t)); (void) fletcher_4_incremental_native(dsp->dsa_drr, offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), &dsp->dsa_zc); if (dsp->dsa_drr->drr_type == DRR_BEGIN) { dsp->dsa_sent_begin = B_TRUE; } else { ASSERT(ZIO_CHECKSUM_IS_ZERO(&dsp->dsa_drr->drr_u. drr_checksum.drr_checksum)); dsp->dsa_drr->drr_u.drr_checksum.drr_checksum = dsp->dsa_zc; } if (dsp->dsa_drr->drr_type == DRR_END) { dsp->dsa_sent_end = B_TRUE; } (void) fletcher_4_incremental_native(&dsp->dsa_drr-> drr_u.drr_checksum.drr_checksum, sizeof (zio_cksum_t), &dsp->dsa_zc); if (dump_bytes(dsp, dsp->dsa_drr, sizeof (dmu_replay_record_t)) != 0) return (SET_ERROR(EINTR)); if (payload_len != 0) { (void) fletcher_4_incremental_native(payload, payload_len, &dsp->dsa_zc); if (dump_bytes(dsp, payload, payload_len) != 0) return (SET_ERROR(EINTR)); } return (0); } /* * Fill in the drr_free struct, or perform aggregation if the previous record is * also a free record, and the two are adjacent. * * Note that we send free records even for a full send, because we want to be * able to receive a full send as a clone, which requires a list of all the free * and freeobject records that were generated on the source. */ static int dump_free(dmu_sendarg_t *dsp, uint64_t object, uint64_t offset, uint64_t length) { struct drr_free *drrf = &(dsp->dsa_drr->drr_u.drr_free); /* * When we receive a free record, dbuf_free_range() assumes * that the receiving system doesn't have any dbufs in the range * being freed. This is always true because there is a one-record * constraint: we only send one WRITE record for any given * object,offset. We know that the one-record constraint is * true because we always send data in increasing order by * object,offset. * * If the increasing-order constraint ever changes, we should find * another way to assert that the one-record constraint is still * satisfied. */ ASSERT(object > dsp->dsa_last_data_object || (object == dsp->dsa_last_data_object && offset > dsp->dsa_last_data_offset)); /* * If there is a pending op, but it's not PENDING_FREE, push it out, * since free block aggregation can only be done for blocks of the * same type (i.e., DRR_FREE records can only be aggregated with * other DRR_FREE records. DRR_FREEOBJECTS records can only be * aggregated with other DRR_FREEOBJECTS records. */ if (dsp->dsa_pending_op != PENDING_NONE && dsp->dsa_pending_op != PENDING_FREE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } if (dsp->dsa_pending_op == PENDING_FREE) { /* * There should never be a PENDING_FREE if length is * DMU_OBJECT_END (because dump_dnode is the only place where * this function is called with a DMU_OBJECT_END, and only after * flushing any pending record). */ ASSERT(length != DMU_OBJECT_END); /* * Check to see whether this free block can be aggregated * with pending one. */ if (drrf->drr_object == object && drrf->drr_offset + drrf->drr_length == offset) { if (offset + length < offset) drrf->drr_length = DMU_OBJECT_END; else drrf->drr_length += length; return (0); } else { /* not a continuation. Push out pending record */ if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } } /* create a FREE record and make it pending */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_FREE; drrf->drr_object = object; drrf->drr_offset = offset; if (offset + length < offset) drrf->drr_length = DMU_OBJECT_END; else drrf->drr_length = length; drrf->drr_toguid = dsp->dsa_toguid; if (length == DMU_OBJECT_END) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); } else { dsp->dsa_pending_op = PENDING_FREE; } return (0); } static int dump_write(dmu_sendarg_t *dsp, dmu_object_type_t type, uint64_t object, uint64_t offset, int lsize, int psize, const blkptr_t *bp, void *data) { uint64_t payload_size; boolean_t raw = (dsp->dsa_featureflags & DMU_BACKUP_FEATURE_RAW); struct drr_write *drrw = &(dsp->dsa_drr->drr_u.drr_write); /* * We send data in increasing object, offset order. * See comment in dump_free() for details. */ ASSERT(object > dsp->dsa_last_data_object || (object == dsp->dsa_last_data_object && offset > dsp->dsa_last_data_offset)); dsp->dsa_last_data_object = object; dsp->dsa_last_data_offset = offset + lsize - 1; /* * If there is any kind of pending aggregation (currently either * a grouping of free objects or free blocks), push it out to * the stream, since aggregation can't be done across operations * of different types. */ if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } /* write a WRITE record */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_WRITE; drrw->drr_object = object; drrw->drr_type = type; drrw->drr_offset = offset; drrw->drr_toguid = dsp->dsa_toguid; drrw->drr_logical_size = lsize; /* only set the compression fields if the buf is compressed or raw */ if (raw || lsize != psize) { ASSERT(!BP_IS_EMBEDDED(bp)); ASSERT3S(psize, >, 0); if (raw) { ASSERT(BP_IS_PROTECTED(bp)); /* * This is a raw protected block so we need to pass * along everything the receiving side will need to * interpret this block, including the byteswap, salt, * IV, and MAC. */ if (BP_SHOULD_BYTESWAP(bp)) drrw->drr_flags |= DRR_RAW_BYTESWAP; zio_crypt_decode_params_bp(bp, drrw->drr_salt, drrw->drr_iv); zio_crypt_decode_mac_bp(bp, drrw->drr_mac); } else { /* this is a compressed block */ ASSERT(dsp->dsa_featureflags & DMU_BACKUP_FEATURE_COMPRESSED); ASSERT(!BP_SHOULD_BYTESWAP(bp)); ASSERT(!DMU_OT_IS_METADATA(BP_GET_TYPE(bp))); ASSERT3U(BP_GET_COMPRESS(bp), !=, ZIO_COMPRESS_OFF); ASSERT3S(lsize, >=, psize); } /* set fields common to compressed and raw sends */ drrw->drr_compressiontype = BP_GET_COMPRESS(bp); drrw->drr_compressed_size = psize; payload_size = drrw->drr_compressed_size; } else { payload_size = drrw->drr_logical_size; } if (bp == NULL || BP_IS_EMBEDDED(bp) || (BP_IS_PROTECTED(bp) && !raw)) { /* * There's no pre-computed checksum for partial-block writes, * embedded BP's, or encrypted BP's that are being sent as * plaintext, so (like fletcher4-checkummed blocks) userland * will have to compute a dedup-capable checksum itself. */ drrw->drr_checksumtype = ZIO_CHECKSUM_OFF; } else { drrw->drr_checksumtype = BP_GET_CHECKSUM(bp); if (zio_checksum_table[drrw->drr_checksumtype].ci_flags & ZCHECKSUM_FLAG_DEDUP) drrw->drr_flags |= DRR_CHECKSUM_DEDUP; DDK_SET_LSIZE(&drrw->drr_key, BP_GET_LSIZE(bp)); DDK_SET_PSIZE(&drrw->drr_key, BP_GET_PSIZE(bp)); DDK_SET_COMPRESS(&drrw->drr_key, BP_GET_COMPRESS(bp)); DDK_SET_CRYPT(&drrw->drr_key, BP_IS_PROTECTED(bp)); drrw->drr_key.ddk_cksum = bp->blk_cksum; } if (dump_record(dsp, data, payload_size) != 0) return (SET_ERROR(EINTR)); return (0); } static int dump_write_embedded(dmu_sendarg_t *dsp, uint64_t object, uint64_t offset, int blksz, const blkptr_t *bp) { char buf[BPE_PAYLOAD_SIZE]; struct drr_write_embedded *drrw = &(dsp->dsa_drr->drr_u.drr_write_embedded); if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } ASSERT(BP_IS_EMBEDDED(bp)); bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_WRITE_EMBEDDED; drrw->drr_object = object; drrw->drr_offset = offset; drrw->drr_length = blksz; drrw->drr_toguid = dsp->dsa_toguid; drrw->drr_compression = BP_GET_COMPRESS(bp); drrw->drr_etype = BPE_GET_ETYPE(bp); drrw->drr_lsize = BPE_GET_LSIZE(bp); drrw->drr_psize = BPE_GET_PSIZE(bp); decode_embedded_bp_compressed(bp, buf); if (dump_record(dsp, buf, P2ROUNDUP(drrw->drr_psize, 8)) != 0) return (SET_ERROR(EINTR)); return (0); } static int dump_spill(dmu_sendarg_t *dsp, const blkptr_t *bp, uint64_t object, void *data) { struct drr_spill *drrs = &(dsp->dsa_drr->drr_u.drr_spill); uint64_t blksz = BP_GET_LSIZE(bp); if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } /* write a SPILL record */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_SPILL; drrs->drr_object = object; drrs->drr_length = blksz; drrs->drr_toguid = dsp->dsa_toguid; /* handle raw send fields */ if (dsp->dsa_featureflags & DMU_BACKUP_FEATURE_RAW) { ASSERT(BP_IS_PROTECTED(bp)); if (BP_SHOULD_BYTESWAP(bp)) drrs->drr_flags |= DRR_RAW_BYTESWAP; drrs->drr_compressiontype = BP_GET_COMPRESS(bp); drrs->drr_compressed_size = BP_GET_PSIZE(bp); zio_crypt_decode_params_bp(bp, drrs->drr_salt, drrs->drr_iv); zio_crypt_decode_mac_bp(bp, drrs->drr_mac); } if (dump_record(dsp, data, blksz) != 0) return (SET_ERROR(EINTR)); return (0); } static int dump_freeobjects(dmu_sendarg_t *dsp, uint64_t firstobj, uint64_t numobjs) { struct drr_freeobjects *drrfo = &(dsp->dsa_drr->drr_u.drr_freeobjects); uint64_t maxobj = DNODES_PER_BLOCK * (DMU_META_DNODE(dsp->dsa_os)->dn_maxblkid + 1); /* * ZoL < 0.7 does not handle large FREEOBJECTS records correctly, * leading to zfs recv never completing. to avoid this issue, don't * send FREEOBJECTS records for object IDs which cannot exist on the * receiving side. */ if (maxobj > 0) { if (maxobj < firstobj) return (0); if (maxobj < firstobj + numobjs) numobjs = maxobj - firstobj; } /* * If there is a pending op, but it's not PENDING_FREEOBJECTS, * push it out, since free block aggregation can only be done for * blocks of the same type (i.e., DRR_FREE records can only be * aggregated with other DRR_FREE records. DRR_FREEOBJECTS records * can only be aggregated with other DRR_FREEOBJECTS records. */ if (dsp->dsa_pending_op != PENDING_NONE && dsp->dsa_pending_op != PENDING_FREEOBJECTS) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } if (dsp->dsa_pending_op == PENDING_FREEOBJECTS) { /* * See whether this free object array can be aggregated * with pending one */ if (drrfo->drr_firstobj + drrfo->drr_numobjs == firstobj) { drrfo->drr_numobjs += numobjs; return (0); } else { /* can't be aggregated. Push out pending record */ if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } } /* write a FREEOBJECTS record */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_FREEOBJECTS; drrfo->drr_firstobj = firstobj; drrfo->drr_numobjs = numobjs; drrfo->drr_toguid = dsp->dsa_toguid; dsp->dsa_pending_op = PENDING_FREEOBJECTS; return (0); } static int dump_dnode(dmu_sendarg_t *dsp, const blkptr_t *bp, uint64_t object, dnode_phys_t *dnp) { struct drr_object *drro = &(dsp->dsa_drr->drr_u.drr_object); int bonuslen; if (object < dsp->dsa_resume_object) { /* * Note: when resuming, we will visit all the dnodes in * the block of dnodes that we are resuming from. In * this case it's unnecessary to send the dnodes prior to * the one we are resuming from. We should be at most one * block's worth of dnodes behind the resume point. */ ASSERT3U(dsp->dsa_resume_object - object, <, 1 << (DNODE_BLOCK_SHIFT - DNODE_SHIFT)); return (0); } if (dnp == NULL || dnp->dn_type == DMU_OT_NONE) return (dump_freeobjects(dsp, object, 1)); if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } /* write an OBJECT record */ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_OBJECT; drro->drr_object = object; drro->drr_type = dnp->dn_type; drro->drr_bonustype = dnp->dn_bonustype; drro->drr_blksz = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT; drro->drr_bonuslen = dnp->dn_bonuslen; drro->drr_dn_slots = dnp->dn_extra_slots + 1; drro->drr_checksumtype = dnp->dn_checksum; drro->drr_compress = dnp->dn_compress; drro->drr_toguid = dsp->dsa_toguid; if (!(dsp->dsa_featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) && drro->drr_blksz > SPA_OLD_MAXBLOCKSIZE) drro->drr_blksz = SPA_OLD_MAXBLOCKSIZE; bonuslen = P2ROUNDUP(dnp->dn_bonuslen, 8); if ((dsp->dsa_featureflags & DMU_BACKUP_FEATURE_RAW)) { ASSERT(BP_IS_ENCRYPTED(bp)); if (BP_SHOULD_BYTESWAP(bp)) drro->drr_flags |= DRR_RAW_BYTESWAP; /* needed for reconstructing dnp on recv side */ drro->drr_maxblkid = dnp->dn_maxblkid; drro->drr_indblkshift = dnp->dn_indblkshift; drro->drr_nlevels = dnp->dn_nlevels; drro->drr_nblkptr = dnp->dn_nblkptr; /* * Since we encrypt the entire bonus area, the (raw) part * beyond the bonuslen is actually nonzero, so we need * to send it. */ if (bonuslen != 0) { drro->drr_raw_bonuslen = DN_MAX_BONUS_LEN(dnp); bonuslen = drro->drr_raw_bonuslen; } } if (dump_record(dsp, DN_BONUS(dnp), bonuslen) != 0) return (SET_ERROR(EINTR)); /* Free anything past the end of the file. */ if (dump_free(dsp, object, (dnp->dn_maxblkid + 1) * (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), DMU_OBJECT_END) != 0) return (SET_ERROR(EINTR)); if (dsp->dsa_err != 0) return (SET_ERROR(EINTR)); return (0); } static int dump_object_range(dmu_sendarg_t *dsp, const blkptr_t *bp, uint64_t firstobj, uint64_t numslots) { struct drr_object_range *drror = &(dsp->dsa_drr->drr_u.drr_object_range); /* we only use this record type for raw sends */ ASSERT(BP_IS_PROTECTED(bp)); ASSERT(dsp->dsa_featureflags & DMU_BACKUP_FEATURE_RAW); ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); ASSERT3U(BP_GET_TYPE(bp), ==, DMU_OT_DNODE); ASSERT0(BP_GET_LEVEL(bp)); if (dsp->dsa_pending_op != PENDING_NONE) { if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); dsp->dsa_pending_op = PENDING_NONE; } bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t)); dsp->dsa_drr->drr_type = DRR_OBJECT_RANGE; drror->drr_firstobj = firstobj; drror->drr_numslots = numslots; drror->drr_toguid = dsp->dsa_toguid; if (BP_SHOULD_BYTESWAP(bp)) drror->drr_flags |= DRR_RAW_BYTESWAP; zio_crypt_decode_params_bp(bp, drror->drr_salt, drror->drr_iv); zio_crypt_decode_mac_bp(bp, drror->drr_mac); if (dump_record(dsp, NULL, 0) != 0) return (SET_ERROR(EINTR)); return (0); } static boolean_t backup_do_embed(dmu_sendarg_t *dsp, const blkptr_t *bp) { if (!BP_IS_EMBEDDED(bp)) return (B_FALSE); /* * Compression function must be legacy, or explicitly enabled. */ if ((BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_LEGACY_FUNCTIONS && !(dsp->dsa_featureflags & DMU_BACKUP_FEATURE_LZ4))) return (B_FALSE); /* * Embed type must be explicitly enabled. */ switch (BPE_GET_ETYPE(bp)) { case BP_EMBEDDED_TYPE_DATA: if (dsp->dsa_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) return (B_TRUE); break; default: return (B_FALSE); } return (B_FALSE); } /* * This is the callback function to traverse_dataset that acts as the worker * thread for dmu_send_impl. */ /*ARGSUSED*/ static int send_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, const zbookmark_phys_t *zb, const struct dnode_phys *dnp, void *arg) { struct send_thread_arg *sta = arg; struct send_block_record *record; uint64_t record_size; int err = 0; ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT || zb->zb_object >= sta->resume.zb_object); ASSERT3P(sta->ds, !=, NULL); if (sta->cancel) return (SET_ERROR(EINTR)); if (bp == NULL) { ASSERT3U(zb->zb_level, ==, ZB_DNODE_LEVEL); return (0); } else if (zb->zb_level < 0) { return (0); } record = kmem_zalloc(sizeof (struct send_block_record), KM_SLEEP); record->eos_marker = B_FALSE; record->bp = *bp; record->zb = *zb; record->indblkshift = dnp->dn_indblkshift; record->datablkszsec = dnp->dn_datablkszsec; record_size = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT; bqueue_enqueue(&sta->q, record, record_size); return (err); } /* * This function kicks off the traverse_dataset. It also handles setting the * error code of the thread in case something goes wrong, and pushes the End of * Stream record when the traverse_dataset call has finished. If there is no * dataset to traverse, the thread immediately pushes End of Stream marker. */ static void send_traverse_thread(void *arg) { struct send_thread_arg *st_arg = arg; int err; struct send_block_record *data; fstrans_cookie_t cookie = spl_fstrans_mark(); if (st_arg->ds != NULL) { err = traverse_dataset_resume(st_arg->ds, st_arg->fromtxg, &st_arg->resume, st_arg->flags, send_cb, st_arg); if (err != EINTR) st_arg->error_code = err; } data = kmem_zalloc(sizeof (*data), KM_SLEEP); data->eos_marker = B_TRUE; bqueue_enqueue(&st_arg->q, data, 1); spl_fstrans_unmark(cookie); thread_exit(); } /* * This function actually handles figuring out what kind of record needs to be * dumped, reading the data (which has hopefully been prefetched), and calling * the appropriate helper function. */ static int do_dump(dmu_sendarg_t *dsa, struct send_block_record *data) { dsl_dataset_t *ds = dmu_objset_ds(dsa->dsa_os); const blkptr_t *bp = &data->bp; const zbookmark_phys_t *zb = &data->zb; uint8_t indblkshift = data->indblkshift; uint16_t dblkszsec = data->datablkszsec; spa_t *spa = ds->ds_dir->dd_pool->dp_spa; dmu_object_type_t type = bp ? BP_GET_TYPE(bp) : DMU_OT_NONE; int err = 0; ASSERT3U(zb->zb_level, >=, 0); ASSERT(zb->zb_object == DMU_META_DNODE_OBJECT || zb->zb_object >= dsa->dsa_resume_object); /* * All bps of an encrypted os should have the encryption bit set. * If this is not true it indicates tampering and we report an error. */ if (dsa->dsa_os->os_encrypted && !BP_IS_HOLE(bp) && !BP_USES_CRYPT(bp)) { spa_log_error(spa, zb); zfs_panic_recover("unencrypted block in encrypted " "object set %llu", ds->ds_object); return (SET_ERROR(EIO)); } if (zb->zb_object != DMU_META_DNODE_OBJECT && DMU_OBJECT_IS_SPECIAL(zb->zb_object)) { return (0); } else if (BP_IS_HOLE(bp) && zb->zb_object == DMU_META_DNODE_OBJECT) { uint64_t span = BP_SPAN(dblkszsec, indblkshift, zb->zb_level); uint64_t dnobj = (zb->zb_blkid * span) >> DNODE_SHIFT; err = dump_freeobjects(dsa, dnobj, span >> DNODE_SHIFT); } else if (BP_IS_HOLE(bp)) { uint64_t span = BP_SPAN(dblkszsec, indblkshift, zb->zb_level); uint64_t offset = zb->zb_blkid * span; /* Don't dump free records for offsets > DMU_OBJECT_END */ if (zb->zb_blkid == 0 || span <= DMU_OBJECT_END / zb->zb_blkid) err = dump_free(dsa, zb->zb_object, offset, span); } else if (zb->zb_level > 0 || type == DMU_OT_OBJSET) { return (0); } else if (type == DMU_OT_DNODE) { int epb = BP_GET_LSIZE(bp) >> DNODE_SHIFT; arc_flags_t aflags = ARC_FLAG_WAIT; arc_buf_t *abuf; enum zio_flag zioflags = ZIO_FLAG_CANFAIL; if (dsa->dsa_featureflags & DMU_BACKUP_FEATURE_RAW) { ASSERT(BP_IS_ENCRYPTED(bp)); ASSERT3U(BP_GET_COMPRESS(bp), ==, ZIO_COMPRESS_OFF); zioflags |= ZIO_FLAG_RAW; } ASSERT0(zb->zb_level); if (arc_read(NULL, spa, bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ, zioflags, &aflags, zb) != 0) return (SET_ERROR(EIO)); dnode_phys_t *blk = abuf->b_data; uint64_t dnobj = zb->zb_blkid * epb; /* * Raw sends require sending encryption parameters for the * block of dnodes. Regular sends do not need to send this * info. */ if (dsa->dsa_featureflags & DMU_BACKUP_FEATURE_RAW) { ASSERT(arc_is_encrypted(abuf)); err = dump_object_range(dsa, bp, dnobj, epb); } if (err == 0) { for (int i = 0; i < epb; i += blk[i].dn_extra_slots + 1) { err = dump_dnode(dsa, bp, dnobj + i, blk + i); if (err != 0) break; } } arc_buf_destroy(abuf, &abuf); } else if (type == DMU_OT_SA) { arc_flags_t aflags = ARC_FLAG_WAIT; arc_buf_t *abuf; enum zio_flag zioflags = ZIO_FLAG_CANFAIL; if (dsa->dsa_featureflags & DMU_BACKUP_FEATURE_RAW) { ASSERT(BP_IS_PROTECTED(bp)); zioflags |= ZIO_FLAG_RAW; } if (arc_read(NULL, spa, bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ, zioflags, &aflags, zb) != 0) return (SET_ERROR(EIO)); err = dump_spill(dsa, bp, zb->zb_object, abuf->b_data); arc_buf_destroy(abuf, &abuf); } else if (backup_do_embed(dsa, bp)) { /* it's an embedded level-0 block of a regular object */ int blksz = dblkszsec << SPA_MINBLOCKSHIFT; ASSERT0(zb->zb_level); err = dump_write_embedded(dsa, zb->zb_object, zb->zb_blkid * blksz, blksz, bp); } else { /* it's a level-0 block of a regular object */ arc_flags_t aflags = ARC_FLAG_WAIT; arc_buf_t *abuf; int blksz = dblkszsec << SPA_MINBLOCKSHIFT; uint64_t offset; /* * If we have large blocks stored on disk but the send flags * don't allow us to send large blocks, we split the data from * the arc buf into chunks. */ boolean_t split_large_blocks = blksz > SPA_OLD_MAXBLOCKSIZE && !(dsa->dsa_featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS); /* * Raw sends require that we always get raw data as it exists * on disk, so we assert that we are not splitting blocks here. */ boolean_t request_raw = (dsa->dsa_featureflags & DMU_BACKUP_FEATURE_RAW) != 0; /* * We should only request compressed data from the ARC if all * the following are true: * - stream compression was requested * - we aren't splitting large blocks into smaller chunks * - the data won't need to be byteswapped before sending * - this isn't an embedded block * - this isn't metadata (if receiving on a different endian * system it can be byteswapped more easily) */ boolean_t request_compressed = (dsa->dsa_featureflags & DMU_BACKUP_FEATURE_COMPRESSED) && !split_large_blocks && !BP_SHOULD_BYTESWAP(bp) && !BP_IS_EMBEDDED(bp) && !DMU_OT_IS_METADATA(BP_GET_TYPE(bp)); IMPLY(request_raw, !split_large_blocks); IMPLY(request_raw, BP_IS_PROTECTED(bp)); ASSERT0(zb->zb_level); ASSERT(zb->zb_object > dsa->dsa_resume_object || (zb->zb_object == dsa->dsa_resume_object && zb->zb_blkid * blksz >= dsa->dsa_resume_offset)); ASSERT3U(blksz, ==, BP_GET_LSIZE(bp)); enum zio_flag zioflags = ZIO_FLAG_CANFAIL; if (request_raw) zioflags |= ZIO_FLAG_RAW; else if (request_compressed) zioflags |= ZIO_FLAG_RAW_COMPRESS; if (arc_read(NULL, spa, bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_ASYNC_READ, zioflags, &aflags, zb) != 0) { if (zfs_send_corrupt_data) { /* Send a block filled with 0x"zfs badd bloc" */ abuf = arc_alloc_buf(spa, &abuf, ARC_BUFC_DATA, blksz); uint64_t *ptr; for (ptr = abuf->b_data; (char *)ptr < (char *)abuf->b_data + blksz; ptr++) *ptr = 0x2f5baddb10cULL; } else { return (SET_ERROR(EIO)); } } offset = zb->zb_blkid * blksz; if (split_large_blocks) { ASSERT0(arc_is_encrypted(abuf)); ASSERT3U(arc_get_compression(abuf), ==, ZIO_COMPRESS_OFF); char *buf = abuf->b_data; while (blksz > 0 && err == 0) { int n = MIN(blksz, SPA_OLD_MAXBLOCKSIZE); err = dump_write(dsa, type, zb->zb_object, offset, n, n, NULL, buf); offset += n; buf += n; blksz -= n; } } else { err = dump_write(dsa, type, zb->zb_object, offset, blksz, arc_buf_size(abuf), bp, abuf->b_data); } arc_buf_destroy(abuf, &abuf); } ASSERT(err == 0 || err == EINTR); return (err); } /* * Pop the new data off the queue, and free the old data. */ static struct send_block_record * get_next_record(bqueue_t *bq, struct send_block_record *data) { struct send_block_record *tmp = bqueue_dequeue(bq); kmem_free(data, sizeof (*data)); return (tmp); } /* * Actually do the bulk of the work in a zfs send. * * Note: Releases dp using the specified tag. */ static int dmu_send_impl(void *tag, dsl_pool_t *dp, dsl_dataset_t *to_ds, zfs_bookmark_phys_t *ancestor_zb, boolean_t is_clone, boolean_t embedok, boolean_t large_block_ok, boolean_t compressok, boolean_t rawok, int outfd, uint64_t resumeobj, uint64_t resumeoff, vnode_t *vp, offset_t *off) { objset_t *os; dmu_replay_record_t *drr; dmu_sendarg_t *dsp; int err; uint64_t fromtxg = 0; uint64_t featureflags = 0; struct send_thread_arg to_arg; void *payload = NULL; size_t payload_len = 0; struct send_block_record *to_data; err = dmu_objset_from_ds(to_ds, &os); if (err != 0) { dsl_pool_rele(dp, tag); return (err); } /* * If this is a non-raw send of an encrypted ds, we can ensure that * the objset_phys_t is authenticated. This is safe because this is * either a snapshot or we have owned the dataset, ensuring that * it can't be modified. */ if (!rawok && os->os_encrypted && arc_is_unauthenticated(os->os_phys_buf)) { zbookmark_phys_t zb; SET_BOOKMARK(&zb, to_ds->ds_object, ZB_ROOT_OBJECT, ZB_ROOT_LEVEL, ZB_ROOT_BLKID); err = arc_untransform(os->os_phys_buf, os->os_spa, &zb, B_FALSE); if (err != 0) { dsl_pool_rele(dp, tag); return (err); } ASSERT0(arc_is_unauthenticated(os->os_phys_buf)); } drr = kmem_zalloc(sizeof (dmu_replay_record_t), KM_SLEEP); drr->drr_type = DRR_BEGIN; drr->drr_u.drr_begin.drr_magic = DMU_BACKUP_MAGIC; DMU_SET_STREAM_HDRTYPE(drr->drr_u.drr_begin.drr_versioninfo, DMU_SUBSTREAM); bzero(&to_arg, sizeof (to_arg)); #ifdef _KERNEL if (dmu_objset_type(os) == DMU_OST_ZFS) { uint64_t version; if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &version) != 0) { kmem_free(drr, sizeof (dmu_replay_record_t)); dsl_pool_rele(dp, tag); return (SET_ERROR(EINVAL)); } if (version >= ZPL_VERSION_SA) { featureflags |= DMU_BACKUP_FEATURE_SA_SPILL; } } #endif /* raw sends imply large_block_ok */ if ((large_block_ok || rawok) && to_ds->ds_feature_inuse[SPA_FEATURE_LARGE_BLOCKS]) featureflags |= DMU_BACKUP_FEATURE_LARGE_BLOCKS; if (to_ds->ds_feature_inuse[SPA_FEATURE_LARGE_DNODE]) featureflags |= DMU_BACKUP_FEATURE_LARGE_DNODE; /* encrypted datasets will not have embedded blocks */ if ((embedok || rawok) && !os->os_encrypted && spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) { featureflags |= DMU_BACKUP_FEATURE_EMBED_DATA; } /* raw send implies compressok */ if (compressok || rawok) featureflags |= DMU_BACKUP_FEATURE_COMPRESSED; if (rawok && os->os_encrypted) featureflags |= DMU_BACKUP_FEATURE_RAW; if ((featureflags & (DMU_BACKUP_FEATURE_EMBED_DATA | DMU_BACKUP_FEATURE_COMPRESSED | DMU_BACKUP_FEATURE_RAW)) != 0 && spa_feature_is_active(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) { featureflags |= DMU_BACKUP_FEATURE_LZ4; } if (resumeobj != 0 || resumeoff != 0) { featureflags |= DMU_BACKUP_FEATURE_RESUMING; } DMU_SET_FEATUREFLAGS(drr->drr_u.drr_begin.drr_versioninfo, featureflags); drr->drr_u.drr_begin.drr_creation_time = dsl_dataset_phys(to_ds)->ds_creation_time; drr->drr_u.drr_begin.drr_type = dmu_objset_type(os); if (is_clone) drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_CLONE; drr->drr_u.drr_begin.drr_toguid = dsl_dataset_phys(to_ds)->ds_guid; if (dsl_dataset_phys(to_ds)->ds_flags & DS_FLAG_CI_DATASET) drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_CI_DATA; if (zfs_send_set_freerecords_bit) drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_FREERECORDS; if (ancestor_zb != NULL) { drr->drr_u.drr_begin.drr_fromguid = ancestor_zb->zbm_guid; fromtxg = ancestor_zb->zbm_creation_txg; } dsl_dataset_name(to_ds, drr->drr_u.drr_begin.drr_toname); if (!to_ds->ds_is_snapshot) { (void) strlcat(drr->drr_u.drr_begin.drr_toname, "@--head--", sizeof (drr->drr_u.drr_begin.drr_toname)); } dsp = kmem_zalloc(sizeof (dmu_sendarg_t), KM_SLEEP); dsp->dsa_drr = drr; dsp->dsa_vp = vp; dsp->dsa_outfd = outfd; dsp->dsa_proc = curproc; dsp->dsa_os = os; dsp->dsa_off = off; dsp->dsa_toguid = dsl_dataset_phys(to_ds)->ds_guid; dsp->dsa_pending_op = PENDING_NONE; dsp->dsa_featureflags = featureflags; dsp->dsa_resume_object = resumeobj; dsp->dsa_resume_offset = resumeoff; mutex_enter(&to_ds->ds_sendstream_lock); list_insert_head(&to_ds->ds_sendstreams, dsp); mutex_exit(&to_ds->ds_sendstream_lock); dsl_dataset_long_hold(to_ds, FTAG); dsl_pool_rele(dp, tag); /* handle features that require a DRR_BEGIN payload */ if (featureflags & (DMU_BACKUP_FEATURE_RESUMING | DMU_BACKUP_FEATURE_RAW)) { nvlist_t *keynvl = NULL; nvlist_t *nvl = fnvlist_alloc(); if (featureflags & DMU_BACKUP_FEATURE_RESUMING) { dmu_object_info_t to_doi; err = dmu_object_info(os, resumeobj, &to_doi); if (err != 0) { fnvlist_free(nvl); goto out; } SET_BOOKMARK(&to_arg.resume, to_ds->ds_object, resumeobj, 0, resumeoff / to_doi.doi_data_block_size); fnvlist_add_uint64(nvl, "resume_object", resumeobj); fnvlist_add_uint64(nvl, "resume_offset", resumeoff); } if (featureflags & DMU_BACKUP_FEATURE_RAW) { ASSERT(os->os_encrypted); err = dsl_crypto_populate_key_nvlist(to_ds, &keynvl); if (err != 0) { fnvlist_free(nvl); goto out; } fnvlist_add_nvlist(nvl, "crypt_keydata", keynvl); } payload = fnvlist_pack(nvl, &payload_len); drr->drr_payloadlen = payload_len; fnvlist_free(keynvl); fnvlist_free(nvl); } err = dump_record(dsp, payload, payload_len); fnvlist_pack_free(payload, payload_len); if (err != 0) { err = dsp->dsa_err; goto out; } err = bqueue_init(&to_arg.q, MAX(zfs_send_queue_length, 2 * zfs_max_recordsize), offsetof(struct send_block_record, ln)); to_arg.error_code = 0; to_arg.cancel = B_FALSE; to_arg.ds = to_ds; to_arg.fromtxg = fromtxg; to_arg.flags = TRAVERSE_PRE | TRAVERSE_PREFETCH; if (rawok) to_arg.flags |= TRAVERSE_NO_DECRYPT; (void) thread_create(NULL, 0, send_traverse_thread, &to_arg, 0, curproc, TS_RUN, minclsyspri); to_data = bqueue_dequeue(&to_arg.q); while (!to_data->eos_marker && err == 0) { err = do_dump(dsp, to_data); to_data = get_next_record(&to_arg.q, to_data); if (issig(JUSTLOOKING) && issig(FORREAL)) err = EINTR; } if (err != 0) { to_arg.cancel = B_TRUE; while (!to_data->eos_marker) { to_data = get_next_record(&to_arg.q, to_data); } } kmem_free(to_data, sizeof (*to_data)); bqueue_destroy(&to_arg.q); if (err == 0 && to_arg.error_code != 0) err = to_arg.error_code; if (err != 0) goto out; if (dsp->dsa_pending_op != PENDING_NONE) if (dump_record(dsp, NULL, 0) != 0) err = SET_ERROR(EINTR); if (err != 0) { if (err == EINTR && dsp->dsa_err != 0) err = dsp->dsa_err; goto out; } bzero(drr, sizeof (dmu_replay_record_t)); drr->drr_type = DRR_END; drr->drr_u.drr_end.drr_checksum = dsp->dsa_zc; drr->drr_u.drr_end.drr_toguid = dsp->dsa_toguid; if (dump_record(dsp, NULL, 0) != 0) err = dsp->dsa_err; out: mutex_enter(&to_ds->ds_sendstream_lock); list_remove(&to_ds->ds_sendstreams, dsp); mutex_exit(&to_ds->ds_sendstream_lock); VERIFY(err != 0 || (dsp->dsa_sent_begin && dsp->dsa_sent_end)); kmem_free(drr, sizeof (dmu_replay_record_t)); kmem_free(dsp, sizeof (dmu_sendarg_t)); dsl_dataset_long_rele(to_ds, FTAG); return (err); } int dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap, boolean_t embedok, boolean_t large_block_ok, boolean_t compressok, boolean_t rawok, int outfd, vnode_t *vp, offset_t *off) { dsl_pool_t *dp; dsl_dataset_t *ds; dsl_dataset_t *fromds = NULL; ds_hold_flags_t dsflags = (rawok) ? 0 : DS_HOLD_FLAG_DECRYPT; int err; err = dsl_pool_hold(pool, FTAG, &dp); if (err != 0) return (err); err = dsl_dataset_hold_obj_flags(dp, tosnap, dsflags, FTAG, &ds); if (err != 0) { dsl_pool_rele(dp, FTAG); return (err); } if (fromsnap != 0) { zfs_bookmark_phys_t zb; boolean_t is_clone; err = dsl_dataset_hold_obj(dp, fromsnap, FTAG, &fromds); if (err != 0) { dsl_dataset_rele_flags(ds, dsflags, FTAG); dsl_pool_rele(dp, FTAG); return (err); } if (!dsl_dataset_is_before(ds, fromds, 0)) err = SET_ERROR(EXDEV); zb.zbm_creation_time = dsl_dataset_phys(fromds)->ds_creation_time; zb.zbm_creation_txg = dsl_dataset_phys(fromds)->ds_creation_txg; zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid; is_clone = (fromds->ds_dir != ds->ds_dir); dsl_dataset_rele(fromds, FTAG); err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone, embedok, large_block_ok, compressok, rawok, outfd, 0, 0, vp, off); } else { err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE, embedok, large_block_ok, compressok, rawok, outfd, 0, 0, vp, off); } dsl_dataset_rele_flags(ds, dsflags, FTAG); return (err); } int dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok, boolean_t large_block_ok, boolean_t compressok, boolean_t rawok, int outfd, uint64_t resumeobj, uint64_t resumeoff, vnode_t *vp, offset_t *off) { dsl_pool_t *dp; dsl_dataset_t *ds; int err; ds_hold_flags_t dsflags = (rawok) ? 0 : DS_HOLD_FLAG_DECRYPT; boolean_t owned = B_FALSE; if (fromsnap != NULL && strpbrk(fromsnap, "@#") == NULL) return (SET_ERROR(EINVAL)); err = dsl_pool_hold(tosnap, FTAG, &dp); if (err != 0) return (err); if (strchr(tosnap, '@') == NULL && spa_writeable(dp->dp_spa)) { /* * We are sending a filesystem or volume. Ensure * that it doesn't change by owning the dataset. */ err = dsl_dataset_own(dp, tosnap, dsflags, FTAG, &ds); owned = B_TRUE; } else { err = dsl_dataset_hold_flags(dp, tosnap, dsflags, FTAG, &ds); } if (err != 0) { dsl_pool_rele(dp, FTAG); return (err); } if (fromsnap != NULL) { zfs_bookmark_phys_t zb; boolean_t is_clone = B_FALSE; int fsnamelen = strchr(tosnap, '@') - tosnap; /* * If the fromsnap is in a different filesystem, then * mark the send stream as a clone. */ if (strncmp(tosnap, fromsnap, fsnamelen) != 0 || (fromsnap[fsnamelen] != '@' && fromsnap[fsnamelen] != '#')) { is_clone = B_TRUE; } if (strchr(fromsnap, '@')) { dsl_dataset_t *fromds; err = dsl_dataset_hold(dp, fromsnap, FTAG, &fromds); if (err == 0) { if (!dsl_dataset_is_before(ds, fromds, 0)) err = SET_ERROR(EXDEV); zb.zbm_creation_time = dsl_dataset_phys(fromds)->ds_creation_time; zb.zbm_creation_txg = dsl_dataset_phys(fromds)->ds_creation_txg; zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid; is_clone = (ds->ds_dir != fromds->ds_dir); dsl_dataset_rele(fromds, FTAG); } } else { err = dsl_bookmark_lookup(dp, fromsnap, ds, &zb); } if (err != 0) { if (owned) dsl_dataset_disown(ds, dsflags, FTAG); else dsl_dataset_rele_flags(ds, dsflags, FTAG); dsl_pool_rele(dp, FTAG); return (err); } err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone, embedok, large_block_ok, compressok, rawok, outfd, resumeobj, resumeoff, vp, off); } else { err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE, embedok, large_block_ok, compressok, rawok, outfd, resumeobj, resumeoff, vp, off); } if (owned) dsl_dataset_disown(ds, dsflags, FTAG); else dsl_dataset_rele_flags(ds, dsflags, FTAG); return (err); } static int dmu_adjust_send_estimate_for_indirects(dsl_dataset_t *ds, uint64_t uncompressed, uint64_t compressed, boolean_t stream_compressed, uint64_t *sizep) { int err; uint64_t size; /* * Assume that space (both on-disk and in-stream) is dominated by * data. We will adjust for indirect blocks and the copies property, * but ignore per-object space used (eg, dnodes and DRR_OBJECT records). */ uint64_t recordsize; uint64_t record_count; objset_t *os; VERIFY0(dmu_objset_from_ds(ds, &os)); /* Assume all (uncompressed) blocks are recordsize. */ if (os->os_phys->os_type == DMU_OST_ZVOL) { err = dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &recordsize); } else { err = dsl_prop_get_int_ds(ds, zfs_prop_to_name(ZFS_PROP_RECORDSIZE), &recordsize); } if (err != 0) return (err); record_count = uncompressed / recordsize; /* * If we're estimating a send size for a compressed stream, use the * compressed data size to estimate the stream size. Otherwise, use the * uncompressed data size. */ size = stream_compressed ? compressed : uncompressed; /* * Subtract out approximate space used by indirect blocks. * Assume most space is used by data blocks (non-indirect, non-dnode). * Assume no ditto blocks or internal fragmentation. * * Therefore, space used by indirect blocks is sizeof(blkptr_t) per * block. */ size -= record_count * sizeof (blkptr_t); /* Add in the space for the record associated with each block. */ size += record_count * sizeof (dmu_replay_record_t); *sizep = size; return (0); } int dmu_send_estimate(dsl_dataset_t *ds, dsl_dataset_t *fromds, boolean_t stream_compressed, uint64_t *sizep) { int err; uint64_t uncomp, comp; ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); /* tosnap must be a snapshot */ if (!ds->ds_is_snapshot) return (SET_ERROR(EINVAL)); /* fromsnap, if provided, must be a snapshot */ if (fromds != NULL && !fromds->ds_is_snapshot) return (SET_ERROR(EINVAL)); /* * fromsnap must be an earlier snapshot from the same fs as tosnap, * or the origin's fs. */ if (fromds != NULL && !dsl_dataset_is_before(ds, fromds, 0)) return (SET_ERROR(EXDEV)); /* Get compressed and uncompressed size estimates of changed data. */ if (fromds == NULL) { uncomp = dsl_dataset_phys(ds)->ds_uncompressed_bytes; comp = dsl_dataset_phys(ds)->ds_compressed_bytes; } else { uint64_t used; err = dsl_dataset_space_written(fromds, ds, &used, &comp, &uncomp); if (err != 0) return (err); } err = dmu_adjust_send_estimate_for_indirects(ds, uncomp, comp, stream_compressed, sizep); /* * Add the size of the BEGIN and END records to the estimate. */ *sizep += 2 * sizeof (dmu_replay_record_t); return (err); } struct calculate_send_arg { uint64_t uncompressed; uint64_t compressed; }; /* * Simple callback used to traverse the blocks of a snapshot and sum their * uncompressed and compressed sizes. */ /* ARGSUSED */ static int dmu_calculate_send_traversal(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg) { struct calculate_send_arg *space = arg; if (bp != NULL && !BP_IS_HOLE(bp)) { space->uncompressed += BP_GET_UCSIZE(bp); space->compressed += BP_GET_PSIZE(bp); } return (0); } /* * Given a desination snapshot and a TXG, calculate the approximate size of a * send stream sent from that TXG. from_txg may be zero, indicating that the * whole snapshot will be sent. */ int dmu_send_estimate_from_txg(dsl_dataset_t *ds, uint64_t from_txg, boolean_t stream_compressed, uint64_t *sizep) { int err; struct calculate_send_arg size = { 0 }; ASSERT(dsl_pool_config_held(ds->ds_dir->dd_pool)); /* tosnap must be a snapshot */ if (!dsl_dataset_is_snapshot(ds)) return (SET_ERROR(EINVAL)); /* verify that from_txg is before the provided snapshot was taken */ if (from_txg >= dsl_dataset_phys(ds)->ds_creation_txg) { return (SET_ERROR(EXDEV)); } /* * traverse the blocks of the snapshot with birth times after * from_txg, summing their uncompressed size */ err = traverse_dataset(ds, from_txg, TRAVERSE_POST | TRAVERSE_NO_DECRYPT, dmu_calculate_send_traversal, &size); if (err) return (err); err = dmu_adjust_send_estimate_for_indirects(ds, size.uncompressed, size.compressed, stream_compressed, sizep); return (err); } typedef struct dmu_recv_begin_arg { const char *drba_origin; dmu_recv_cookie_t *drba_cookie; cred_t *drba_cred; uint64_t drba_snapobj; } dmu_recv_begin_arg_t; static int recv_begin_check_existing_impl(dmu_recv_begin_arg_t *drba, dsl_dataset_t *ds, uint64_t fromguid) { uint64_t val; int error; dsl_pool_t *dp = ds->ds_dir->dd_pool; struct drr_begin *drrb = drba->drba_cookie->drc_drrb; uint64_t featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); boolean_t encrypted = ds->ds_dir->dd_crypto_obj != 0; boolean_t raw = (featureflags & DMU_BACKUP_FEATURE_RAW) != 0; /* temporary clone name must not exist */ error = zap_lookup(dp->dp_meta_objset, dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, recv_clone_name, 8, 1, &val); if (error != ENOENT) return (error == 0 ? EBUSY : error); /* new snapshot name must not exist */ error = zap_lookup(dp->dp_meta_objset, dsl_dataset_phys(ds)->ds_snapnames_zapobj, drba->drba_cookie->drc_tosnap, 8, 1, &val); if (error != ENOENT) return (error == 0 ? EEXIST : error); /* * Check snapshot limit before receiving. We'll recheck again at the * end, but might as well abort before receiving if we're already over * the limit. * * Note that we do not check the file system limit with * dsl_dir_fscount_check because the temporary %clones don't count * against that limit. */ error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT, NULL, drba->drba_cred); if (error != 0) return (error); if (fromguid != 0) { dsl_dataset_t *snap; uint64_t obj = dsl_dataset_phys(ds)->ds_prev_snap_obj; /* Can't perform a raw receive on top of a non-raw receive */ if (!encrypted && raw) return (SET_ERROR(EINVAL)); /* Find snapshot in this dir that matches fromguid. */ while (obj != 0) { error = dsl_dataset_hold_obj(dp, obj, FTAG, &snap); if (error != 0) return (SET_ERROR(ENODEV)); if (snap->ds_dir != ds->ds_dir) { dsl_dataset_rele(snap, FTAG); return (SET_ERROR(ENODEV)); } if (dsl_dataset_phys(snap)->ds_guid == fromguid) break; obj = dsl_dataset_phys(snap)->ds_prev_snap_obj; dsl_dataset_rele(snap, FTAG); } if (obj == 0) return (SET_ERROR(ENODEV)); if (drba->drba_cookie->drc_force) { drba->drba_snapobj = obj; } else { /* * If we are not forcing, there must be no * changes since fromsnap. */ if (dsl_dataset_modified_since_snap(ds, snap)) { dsl_dataset_rele(snap, FTAG); return (SET_ERROR(ETXTBSY)); } drba->drba_snapobj = ds->ds_prev->ds_object; } dsl_dataset_rele(snap, FTAG); } else { /* if full, then must be forced */ if (!drba->drba_cookie->drc_force) return (SET_ERROR(EEXIST)); /* * We don't support using zfs recv -F to blow away * encrypted filesystems. This would require the * dsl dir to point to the old encryption key and * the new one at the same time during the receive. */ if ((!encrypted && raw) || encrypted) return (SET_ERROR(EINVAL)); drba->drba_snapobj = 0; } return (0); } static int dmu_recv_begin_check(void *arg, dmu_tx_t *tx) { dmu_recv_begin_arg_t *drba = arg; dsl_pool_t *dp = dmu_tx_pool(tx); struct drr_begin *drrb = drba->drba_cookie->drc_drrb; uint64_t fromguid = drrb->drr_fromguid; int flags = drrb->drr_flags; ds_hold_flags_t dsflags = 0; int error; uint64_t featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); dsl_dataset_t *ds; const char *tofs = drba->drba_cookie->drc_tofs; /* already checked */ ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC); ASSERT(!(featureflags & DMU_BACKUP_FEATURE_RESUMING)); if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_COMPOUNDSTREAM || drrb->drr_type >= DMU_OST_NUMTYPES || ((flags & DRR_FLAG_CLONE) && drba->drba_origin == NULL)) return (SET_ERROR(EINVAL)); /* Verify pool version supports SA if SA_SPILL feature set */ if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) && spa_version(dp->dp_spa) < SPA_VERSION_SA) return (SET_ERROR(ENOTSUP)); if (drba->drba_cookie->drc_resumable && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EXTENSIBLE_DATASET)) return (SET_ERROR(ENOTSUP)); /* * The receiving code doesn't know how to translate a WRITE_EMBEDDED * record to a plain WRITE record, so the pool must have the * EMBEDDED_DATA feature enabled if the stream has WRITE_EMBEDDED * records. Same with WRITE_EMBEDDED records that use LZ4 compression. */ if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) return (SET_ERROR(ENOTSUP)); if ((featureflags & DMU_BACKUP_FEATURE_LZ4) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) return (SET_ERROR(ENOTSUP)); /* * The receiving code doesn't know how to translate large blocks * to smaller ones, so the pool must have the LARGE_BLOCKS * feature enabled if the stream has LARGE_BLOCKS. Same with * large dnodes. */ if ((featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LARGE_BLOCKS)) return (SET_ERROR(ENOTSUP)); if ((featureflags & DMU_BACKUP_FEATURE_LARGE_DNODE) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LARGE_DNODE)) return (SET_ERROR(ENOTSUP)); if ((featureflags & DMU_BACKUP_FEATURE_RAW)) { /* raw receives require the encryption feature */ if (!spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_ENCRYPTION)) return (SET_ERROR(ENOTSUP)); } else { dsflags |= DS_HOLD_FLAG_DECRYPT; } error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds); if (error == 0) { /* target fs already exists; recv into temp clone */ /* Can't recv a clone into an existing fs */ if (flags & DRR_FLAG_CLONE || drba->drba_origin) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(EINVAL)); } error = recv_begin_check_existing_impl(drba, ds, fromguid); dsl_dataset_rele_flags(ds, dsflags, FTAG); } else if (error == ENOENT) { /* target fs does not exist; must be a full backup or clone */ char buf[ZFS_MAX_DATASET_NAME_LEN]; /* * If it's a non-clone incremental, we are missing the * target fs, so fail the recv. */ if (fromguid != 0 && !(flags & DRR_FLAG_CLONE || drba->drba_origin)) return (SET_ERROR(ENOENT)); /* * If we're receiving a full send as a clone, and it doesn't * contain all the necessary free records and freeobject * records, reject it. */ if (fromguid == 0 && drba->drba_origin && !(flags & DRR_FLAG_FREERECORDS)) return (SET_ERROR(EINVAL)); /* Open the parent of tofs */ ASSERT3U(strlen(tofs), <, sizeof (buf)); (void) strlcpy(buf, tofs, strrchr(tofs, '/') - tofs + 1); error = dsl_dataset_hold_flags(dp, buf, dsflags, FTAG, &ds); if (error != 0) return (error); /* * Check filesystem and snapshot limits before receiving. We'll * recheck snapshot limits again at the end (we create the * filesystems and increment those counts during begin_sync). */ error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_FILESYSTEM_LIMIT, NULL, drba->drba_cred); if (error != 0) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (error); } error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT, NULL, drba->drba_cred); if (error != 0) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (error); } if (drba->drba_origin != NULL) { dsl_dataset_t *origin; error = dsl_dataset_hold_flags(dp, drba->drba_origin, dsflags, FTAG, &origin); if (error != 0) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (error); } if (!origin->ds_is_snapshot) { dsl_dataset_rele_flags(origin, dsflags, FTAG); dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(EINVAL)); } if (dsl_dataset_phys(origin)->ds_guid != fromguid && fromguid != 0) { dsl_dataset_rele_flags(origin, dsflags, FTAG); dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(ENODEV)); } dsl_dataset_rele_flags(origin, dsflags, FTAG); } dsl_dataset_rele_flags(ds, dsflags, FTAG); error = 0; } return (error); } static void dmu_recv_begin_sync(void *arg, dmu_tx_t *tx) { dmu_recv_begin_arg_t *drba = arg; dsl_pool_t *dp = dmu_tx_pool(tx); objset_t *mos = dp->dp_meta_objset; struct drr_begin *drrb = drba->drba_cookie->drc_drrb; const char *tofs = drba->drba_cookie->drc_tofs; uint64_t featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); dsl_dataset_t *ds, *newds; objset_t *os; uint64_t dsobj; ds_hold_flags_t dsflags = 0; int error; uint64_t crflags = 0; dsl_crypto_params_t *dcpp = NULL; dsl_crypto_params_t dcp = { 0 }; if (drrb->drr_flags & DRR_FLAG_CI_DATA) crflags |= DS_FLAG_CI_DATASET; if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0) { dsflags |= DS_HOLD_FLAG_DECRYPT; } else { dcp.cp_cmd = DCP_CMD_RAW_RECV; } error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds); if (error == 0) { /* create temporary clone */ dsl_dataset_t *snap = NULL; if (drba->drba_snapobj != 0) { VERIFY0(dsl_dataset_hold_obj(dp, drba->drba_snapobj, FTAG, &snap)); } else { /* we use the dcp whenever we are not making a clone */ dcpp = &dcp; } dsobj = dsl_dataset_create_sync(ds->ds_dir, recv_clone_name, snap, crflags, drba->drba_cred, dcpp, tx); if (drba->drba_snapobj != 0) dsl_dataset_rele(snap, FTAG); dsl_dataset_rele_flags(ds, dsflags, FTAG); } else { dsl_dir_t *dd; const char *tail; dsl_dataset_t *origin = NULL; VERIFY0(dsl_dir_hold(dp, tofs, FTAG, &dd, &tail)); if (drba->drba_origin != NULL) { VERIFY0(dsl_dataset_hold(dp, drba->drba_origin, FTAG, &origin)); } else { /* we use the dcp whenever we are not making a clone */ dcpp = &dcp; } /* Create new dataset. */ dsobj = dsl_dataset_create_sync(dd, strrchr(tofs, '/') + 1, origin, crflags, drba->drba_cred, dcpp, tx); if (origin != NULL) dsl_dataset_rele(origin, FTAG); dsl_dir_rele(dd, FTAG); drba->drba_cookie->drc_newfs = B_TRUE; } VERIFY0(dsl_dataset_own_obj(dp, dsobj, dsflags, dmu_recv_tag, &newds)); VERIFY0(dmu_objset_from_ds(newds, &os)); if (drba->drba_cookie->drc_resumable) { dsl_dataset_zapify(newds, tx); if (drrb->drr_fromguid != 0) { VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_FROMGUID, 8, 1, &drrb->drr_fromguid, tx)); } VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TOGUID, 8, 1, &drrb->drr_toguid, tx)); VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_TONAME, 1, strlen(drrb->drr_toname) + 1, drrb->drr_toname, tx)); uint64_t one = 1; uint64_t zero = 0; VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OBJECT, 8, 1, &one, tx)); VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_OFFSET, 8, 1, &zero, tx)); VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_BYTES, 8, 1, &zero, tx)); if (featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) { VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_LARGEBLOCK, 8, 1, &one, tx)); } if (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) { VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_EMBEDOK, 8, 1, &one, tx)); } if (featureflags & DMU_BACKUP_FEATURE_COMPRESSED) { VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_COMPRESSOK, 8, 1, &one, tx)); } if (featureflags & DMU_BACKUP_FEATURE_RAW) { VERIFY0(zap_add(mos, dsobj, DS_FIELD_RESUME_RAWOK, 8, 1, &one, tx)); } } /* * Usually the os->os_encrypted value is tied to the presence of a * DSL Crypto Key object in the dd. However, that will not be received * until dmu_recv_stream(), so we set the value manually for now. */ if (featureflags & DMU_BACKUP_FEATURE_RAW) { os->os_encrypted = B_TRUE; drba->drba_cookie->drc_raw = B_TRUE; } dmu_buf_will_dirty(newds->ds_dbuf, tx); dsl_dataset_phys(newds)->ds_flags |= DS_FLAG_INCONSISTENT; /* * If we actually created a non-clone, we need to create the objset * in our new dataset. If this is a raw send we postpone this until * dmu_recv_stream() so that we can allocate the metadnode with the * properties from the DRR_BEGIN payload. */ rrw_enter(&newds->ds_bp_rwlock, RW_READER, FTAG); if (BP_IS_HOLE(dsl_dataset_get_blkptr(newds)) && (featureflags & DMU_BACKUP_FEATURE_RAW) == 0) { (void) dmu_objset_create_impl(dp->dp_spa, newds, dsl_dataset_get_blkptr(newds), drrb->drr_type, tx); } rrw_exit(&newds->ds_bp_rwlock, FTAG); drba->drba_cookie->drc_ds = newds; spa_history_log_internal_ds(newds, "receive", tx, ""); } static int dmu_recv_resume_begin_check(void *arg, dmu_tx_t *tx) { dmu_recv_begin_arg_t *drba = arg; dsl_pool_t *dp = dmu_tx_pool(tx); struct drr_begin *drrb = drba->drba_cookie->drc_drrb; int error; ds_hold_flags_t dsflags = 0; uint64_t featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); dsl_dataset_t *ds; const char *tofs = drba->drba_cookie->drc_tofs; /* already checked */ ASSERT3U(drrb->drr_magic, ==, DMU_BACKUP_MAGIC); ASSERT(featureflags & DMU_BACKUP_FEATURE_RESUMING); if (DMU_GET_STREAM_HDRTYPE(drrb->drr_versioninfo) == DMU_COMPOUNDSTREAM || drrb->drr_type >= DMU_OST_NUMTYPES) return (SET_ERROR(EINVAL)); /* Verify pool version supports SA if SA_SPILL feature set */ if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) && spa_version(dp->dp_spa) < SPA_VERSION_SA) return (SET_ERROR(ENOTSUP)); /* * The receiving code doesn't know how to translate a WRITE_EMBEDDED * record to a plain WRITE record, so the pool must have the * EMBEDDED_DATA feature enabled if the stream has WRITE_EMBEDDED * records. Same with WRITE_EMBEDDED records that use LZ4 compression. */ if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) return (SET_ERROR(ENOTSUP)); if ((featureflags & DMU_BACKUP_FEATURE_LZ4) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS)) return (SET_ERROR(ENOTSUP)); /* * The receiving code doesn't know how to translate large blocks * to smaller ones, so the pool must have the LARGE_BLOCKS * feature enabled if the stream has LARGE_BLOCKS. Same with * large dnodes. */ if ((featureflags & DMU_BACKUP_FEATURE_LARGE_BLOCKS) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LARGE_BLOCKS)) return (SET_ERROR(ENOTSUP)); if ((featureflags & DMU_BACKUP_FEATURE_LARGE_DNODE) && !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LARGE_DNODE)) return (SET_ERROR(ENOTSUP)); /* 6 extra bytes for /%recv */ char recvname[ZFS_MAX_DATASET_NAME_LEN + 6]; (void) snprintf(recvname, sizeof (recvname), "%s/%s", tofs, recv_clone_name); if ((featureflags & DMU_BACKUP_FEATURE_RAW) == 0) dsflags |= DS_HOLD_FLAG_DECRYPT; if (dsl_dataset_hold_flags(dp, recvname, dsflags, FTAG, &ds) != 0) { /* %recv does not exist; continue in tofs */ error = dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds); if (error != 0) return (error); } /* check that ds is marked inconsistent */ if (!DS_IS_INCONSISTENT(ds)) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(EINVAL)); } /* check that there is resuming data, and that the toguid matches */ if (!dsl_dataset_is_zapified(ds)) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(EINVAL)); } uint64_t val; error = zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TOGUID, sizeof (val), 1, &val); if (error != 0 || drrb->drr_toguid != val) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(EINVAL)); } /* * Check if the receive is still running. If so, it will be owned. * Note that nothing else can own the dataset (e.g. after the receive * fails) because it will be marked inconsistent. */ if (dsl_dataset_has_owner(ds)) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(EBUSY)); } /* There should not be any snapshots of this fs yet. */ if (ds->ds_prev != NULL && ds->ds_prev->ds_dir == ds->ds_dir) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(EINVAL)); } /* * Note: resume point will be checked when we process the first WRITE * record. */ /* check that the origin matches */ val = 0; (void) zap_lookup(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_FROMGUID, sizeof (val), 1, &val); if (drrb->drr_fromguid != val) { dsl_dataset_rele_flags(ds, dsflags, FTAG); return (SET_ERROR(EINVAL)); } dsl_dataset_rele_flags(ds, dsflags, FTAG); return (0); } static void dmu_recv_resume_begin_sync(void *arg, dmu_tx_t *tx) { dmu_recv_begin_arg_t *drba = arg; dsl_pool_t *dp = dmu_tx_pool(tx); const char *tofs = drba->drba_cookie->drc_tofs; struct drr_begin *drrb = drba->drba_cookie->drc_drrb; uint64_t featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo); dsl_dataset_t *ds; objset_t *os; ds_hold_flags_t dsflags = 0; uint64_t dsobj; /* 6 extra bytes for /%recv */ char recvname[ZFS_MAX_DATASET_NAME_LEN + 6]; (void) snprintf(recvname, sizeof (recvname), "%s/%s", tofs, recv_clone_name); if (featureflags & DMU_BACKUP_FEATURE_RAW) { drba->drba_cookie->drc_raw = B_TRUE; } else { dsflags |= DS_HOLD_FLAG_DECRYPT; } if (dsl_dataset_hold_flags(dp, recvname, dsflags, FTAG, &ds) != 0) { /* %recv does not exist; continue in tofs */ VERIFY0(dsl_dataset_hold_flags(dp, tofs, dsflags, FTAG, &ds)); drba->drba_cookie->drc_newfs = B_TRUE; } /* clear the inconsistent flag so that we can own it */ ASSERT(DS_IS_INCONSISTENT(ds)); dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT; dsobj = ds->ds_object; dsl_dataset_rele_flags(ds, dsflags, FTAG); VERIFY0(dsl_dataset_own_obj(dp, dsobj, dsflags, dmu_recv_tag, &ds)); VERIFY0(dmu_objset_from_ds(ds, &os)); dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_flags |= DS_FLAG_INCONSISTENT; rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); ASSERT(!BP_IS_HOLE(dsl_dataset_get_blkptr(ds)) || drba->drba_cookie->drc_raw); rrw_exit(&ds->ds_bp_rwlock, FTAG); drba->drba_cookie->drc_ds = ds; spa_history_log_internal_ds(ds, "resume receive", tx, ""); } /* * NB: callers *MUST* call dmu_recv_stream() if dmu_recv_begin() * succeeds; otherwise we will leak the holds on the datasets. */ int dmu_recv_begin(char *tofs, char *tosnap, dmu_replay_record_t *drr_begin, boolean_t force, boolean_t resumable, char *origin, dmu_recv_cookie_t *drc) { dmu_recv_begin_arg_t drba = { 0 }; bzero(drc, sizeof (dmu_recv_cookie_t)); drc->drc_drr_begin = drr_begin; drc->drc_drrb = &drr_begin->drr_u.drr_begin; drc->drc_tosnap = tosnap; drc->drc_tofs = tofs; drc->drc_force = force; drc->drc_resumable = resumable; drc->drc_cred = CRED(); drc->drc_clone = (origin != NULL); if (drc->drc_drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC)) { drc->drc_byteswap = B_TRUE; (void) fletcher_4_incremental_byteswap(drr_begin, sizeof (dmu_replay_record_t), &drc->drc_cksum); byteswap_record(drr_begin); } else if (drc->drc_drrb->drr_magic == DMU_BACKUP_MAGIC) { (void) fletcher_4_incremental_native(drr_begin, sizeof (dmu_replay_record_t), &drc->drc_cksum); } else { return (SET_ERROR(EINVAL)); } drba.drba_origin = origin; drba.drba_cookie = drc; drba.drba_cred = CRED(); if (DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo) & DMU_BACKUP_FEATURE_RESUMING) { return (dsl_sync_task(tofs, dmu_recv_resume_begin_check, dmu_recv_resume_begin_sync, &drba, 5, ZFS_SPACE_CHECK_NORMAL)); } else { return (dsl_sync_task(tofs, dmu_recv_begin_check, dmu_recv_begin_sync, &drba, 5, ZFS_SPACE_CHECK_NORMAL)); } } struct receive_record_arg { dmu_replay_record_t header; void *payload; /* Pointer to a buffer containing the payload */ /* * If the record is a write, pointer to the arc_buf_t containing the * payload. */ arc_buf_t *arc_buf; int payload_size; uint64_t bytes_read; /* bytes read from stream when record created */ boolean_t eos_marker; /* Marks the end of the stream */ bqueue_node_t node; }; struct receive_writer_arg { objset_t *os; boolean_t byteswap; bqueue_t q; /* * These three args are used to signal to the main thread that we're * done. */ kmutex_t mutex; kcondvar_t cv; boolean_t done; int err; /* A map from guid to dataset to help handle dedup'd streams. */ avl_tree_t *guid_to_ds_map; boolean_t resumable; boolean_t raw; uint64_t last_object; uint64_t last_offset; uint64_t max_object; /* highest object ID referenced in stream */ uint64_t bytes_read; /* bytes read when current record created */ /* Encryption parameters for the last received DRR_OBJECT_RANGE */ + boolean_t or_crypt_params_present; uint64_t or_firstobj; uint64_t or_numslots; uint8_t or_salt[ZIO_DATA_SALT_LEN]; uint8_t or_iv[ZIO_DATA_IV_LEN]; uint8_t or_mac[ZIO_DATA_MAC_LEN]; boolean_t or_byteorder; }; struct objlist { list_t list; /* List of struct receive_objnode. */ /* * Last object looked up. Used to assert that objects are being looked * up in ascending order. */ uint64_t last_lookup; }; struct receive_objnode { list_node_t node; uint64_t object; }; struct receive_arg { objset_t *os; vnode_t *vp; /* The vnode to read the stream from */ uint64_t voff; /* The current offset in the stream */ uint64_t bytes_read; /* * A record that has had its payload read in, but hasn't yet been handed * off to the worker thread. */ struct receive_record_arg *rrd; /* A record that has had its header read in, but not its payload. */ struct receive_record_arg *next_rrd; zio_cksum_t cksum; zio_cksum_t prev_cksum; int err; boolean_t byteswap; boolean_t raw; uint64_t featureflags; /* Sorted list of objects not to issue prefetches for. */ struct objlist ignore_objlist; }; typedef struct guid_map_entry { uint64_t guid; boolean_t raw; dsl_dataset_t *gme_ds; avl_node_t avlnode; } guid_map_entry_t; static int guid_compare(const void *arg1, const void *arg2) { const guid_map_entry_t *gmep1 = (const guid_map_entry_t *)arg1; const guid_map_entry_t *gmep2 = (const guid_map_entry_t *)arg2; return (AVL_CMP(gmep1->guid, gmep2->guid)); } static void free_guid_map_onexit(void *arg) { avl_tree_t *ca = arg; void *cookie = NULL; guid_map_entry_t *gmep; while ((gmep = avl_destroy_nodes(ca, &cookie)) != NULL) { dsl_dataset_long_rele(gmep->gme_ds, gmep); dsl_dataset_rele_flags(gmep->gme_ds, (gmep->raw) ? 0 : DS_HOLD_FLAG_DECRYPT, gmep); kmem_free(gmep, sizeof (guid_map_entry_t)); } avl_destroy(ca); kmem_free(ca, sizeof (avl_tree_t)); } static int receive_read(struct receive_arg *ra, int len, void *buf) { int done = 0; /* * The code doesn't rely on this (lengths being multiples of 8). See * comment in dump_bytes. */ ASSERT(len % 8 == 0 || (ra->featureflags & DMU_BACKUP_FEATURE_RAW) != 0); while (done < len) { ssize_t resid; ra->err = vn_rdwr(UIO_READ, ra->vp, (char *)buf + done, len - done, ra->voff, UIO_SYSSPACE, FAPPEND, RLIM64_INFINITY, CRED(), &resid); if (resid == len - done) { /* * Note: ECKSUM indicates that the receive * was interrupted and can potentially be resumed. */ ra->err = SET_ERROR(ECKSUM); } ra->voff += len - done - resid; done = len - resid; if (ra->err != 0) return (ra->err); } ra->bytes_read += len; ASSERT3U(done, ==, len); return (0); } noinline static void byteswap_record(dmu_replay_record_t *drr) { #define DO64(X) (drr->drr_u.X = BSWAP_64(drr->drr_u.X)) #define DO32(X) (drr->drr_u.X = BSWAP_32(drr->drr_u.X)) drr->drr_type = BSWAP_32(drr->drr_type); drr->drr_payloadlen = BSWAP_32(drr->drr_payloadlen); switch (drr->drr_type) { case DRR_BEGIN: DO64(drr_begin.drr_magic); DO64(drr_begin.drr_versioninfo); DO64(drr_begin.drr_creation_time); DO32(drr_begin.drr_type); DO32(drr_begin.drr_flags); DO64(drr_begin.drr_toguid); DO64(drr_begin.drr_fromguid); break; case DRR_OBJECT: DO64(drr_object.drr_object); DO32(drr_object.drr_type); DO32(drr_object.drr_bonustype); DO32(drr_object.drr_blksz); DO32(drr_object.drr_bonuslen); DO32(drr_object.drr_raw_bonuslen); DO64(drr_object.drr_toguid); DO64(drr_object.drr_maxblkid); break; case DRR_FREEOBJECTS: DO64(drr_freeobjects.drr_firstobj); DO64(drr_freeobjects.drr_numobjs); DO64(drr_freeobjects.drr_toguid); break; case DRR_WRITE: DO64(drr_write.drr_object); DO32(drr_write.drr_type); DO64(drr_write.drr_offset); DO64(drr_write.drr_logical_size); DO64(drr_write.drr_toguid); ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_write.drr_key.ddk_cksum); DO64(drr_write.drr_key.ddk_prop); DO64(drr_write.drr_compressed_size); break; case DRR_WRITE_BYREF: DO64(drr_write_byref.drr_object); DO64(drr_write_byref.drr_offset); DO64(drr_write_byref.drr_length); DO64(drr_write_byref.drr_toguid); DO64(drr_write_byref.drr_refguid); DO64(drr_write_byref.drr_refobject); DO64(drr_write_byref.drr_refoffset); ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_write_byref. drr_key.ddk_cksum); DO64(drr_write_byref.drr_key.ddk_prop); break; case DRR_WRITE_EMBEDDED: DO64(drr_write_embedded.drr_object); DO64(drr_write_embedded.drr_offset); DO64(drr_write_embedded.drr_length); DO64(drr_write_embedded.drr_toguid); DO32(drr_write_embedded.drr_lsize); DO32(drr_write_embedded.drr_psize); break; case DRR_FREE: DO64(drr_free.drr_object); DO64(drr_free.drr_offset); DO64(drr_free.drr_length); DO64(drr_free.drr_toguid); break; case DRR_SPILL: DO64(drr_spill.drr_object); DO64(drr_spill.drr_length); DO64(drr_spill.drr_toguid); DO64(drr_spill.drr_compressed_size); DO32(drr_spill.drr_type); break; case DRR_OBJECT_RANGE: DO64(drr_object_range.drr_firstobj); DO64(drr_object_range.drr_numslots); DO64(drr_object_range.drr_toguid); break; case DRR_END: DO64(drr_end.drr_toguid); ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_end.drr_checksum); break; default: break; } if (drr->drr_type != DRR_BEGIN) { ZIO_CHECKSUM_BSWAP(&drr->drr_u.drr_checksum.drr_checksum); } #undef DO64 #undef DO32 } static inline uint8_t deduce_nblkptr(dmu_object_type_t bonus_type, uint64_t bonus_size) { if (bonus_type == DMU_OT_SA) { return (1); } else { return (1 + ((DN_OLD_MAX_BONUSLEN - MIN(DN_OLD_MAX_BONUSLEN, bonus_size)) >> SPA_BLKPTRSHIFT)); } } static void save_resume_state(struct receive_writer_arg *rwa, uint64_t object, uint64_t offset, dmu_tx_t *tx) { int txgoff = dmu_tx_get_txg(tx) & TXG_MASK; if (!rwa->resumable) return; /* * We use ds_resume_bytes[] != 0 to indicate that we need to * update this on disk, so it must not be 0. */ ASSERT(rwa->bytes_read != 0); /* * We only resume from write records, which have a valid * (non-meta-dnode) object number. */ ASSERT(object != 0); /* * For resuming to work correctly, we must receive records in order, * sorted by object,offset. This is checked by the callers, but * assert it here for good measure. */ ASSERT3U(object, >=, rwa->os->os_dsl_dataset->ds_resume_object[txgoff]); ASSERT(object != rwa->os->os_dsl_dataset->ds_resume_object[txgoff] || offset >= rwa->os->os_dsl_dataset->ds_resume_offset[txgoff]); ASSERT3U(rwa->bytes_read, >=, rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff]); rwa->os->os_dsl_dataset->ds_resume_object[txgoff] = object; rwa->os->os_dsl_dataset->ds_resume_offset[txgoff] = offset; rwa->os->os_dsl_dataset->ds_resume_bytes[txgoff] = rwa->bytes_read; } noinline static int receive_object(struct receive_writer_arg *rwa, struct drr_object *drro, void *data) { dmu_object_info_t doi; dmu_tx_t *tx; uint64_t object; int err; if (drro->drr_type == DMU_OT_NONE || !DMU_OT_IS_VALID(drro->drr_type) || !DMU_OT_IS_VALID(drro->drr_bonustype) || drro->drr_checksumtype >= ZIO_CHECKSUM_FUNCTIONS || drro->drr_compress >= ZIO_COMPRESS_FUNCTIONS || P2PHASE(drro->drr_blksz, SPA_MINBLOCKSIZE) || drro->drr_blksz < SPA_MINBLOCKSIZE || drro->drr_blksz > spa_maxblocksize(dmu_objset_spa(rwa->os)) || drro->drr_bonuslen > DN_BONUS_SIZE(spa_maxdnodesize(dmu_objset_spa(rwa->os))) || drro->drr_dn_slots > (spa_maxdnodesize(dmu_objset_spa(rwa->os)) >> DNODE_SHIFT)) { return (SET_ERROR(EINVAL)); } if (rwa->raw) { /* * We should have received a DRR_OBJECT_RANGE record * containing this block and stored it in rwa. */ if (drro->drr_object < rwa->or_firstobj || drro->drr_object >= rwa->or_firstobj + rwa->or_numslots || drro->drr_raw_bonuslen < drro->drr_bonuslen || drro->drr_indblkshift > SPA_MAXBLOCKSHIFT || drro->drr_nlevels > DN_MAX_LEVELS || drro->drr_nblkptr > DN_MAX_NBLKPTR || DN_SLOTS_TO_BONUSLEN(drro->drr_dn_slots) < drro->drr_raw_bonuslen) return (SET_ERROR(EINVAL)); } else { if (drro->drr_flags != 0 || drro->drr_raw_bonuslen != 0 || drro->drr_indblkshift != 0 || drro->drr_nlevels != 0 || drro->drr_nblkptr != 0) return (SET_ERROR(EINVAL)); } err = dmu_object_info(rwa->os, drro->drr_object, &doi); if (err != 0 && err != ENOENT && err != EEXIST) return (SET_ERROR(EINVAL)); if (drro->drr_object > rwa->max_object) rwa->max_object = drro->drr_object; /* * If we are losing blkptrs or changing the block size this must * be a new file instance. We must clear out the previous file * contents before we can change this type of metadata in the dnode. * Raw receives will also check that the indirect structure of the * dnode hasn't changed. */ if (err == 0) { uint32_t indblksz = drro->drr_indblkshift ? 1ULL << drro->drr_indblkshift : 0; int nblkptr = deduce_nblkptr(drro->drr_bonustype, drro->drr_bonuslen); object = drro->drr_object; /* nblkptr will be bounded by the bonus size and type */ if (rwa->raw && nblkptr != drro->drr_nblkptr) return (SET_ERROR(EINVAL)); - if (rwa->raw && - (drro->drr_blksz != doi.doi_data_block_size || + if (drro->drr_blksz != doi.doi_data_block_size || nblkptr < doi.doi_nblkptr || - indblksz != doi.doi_metadata_block_size || - drro->drr_nlevels < doi.doi_indirection || - drro->drr_dn_slots != doi.doi_dnodesize >> DNODE_SHIFT)) { - err = dmu_free_long_range_raw(rwa->os, + drro->drr_dn_slots != doi.doi_dnodesize >> DNODE_SHIFT || + (rwa->raw && + (indblksz != doi.doi_metadata_block_size || + drro->drr_nlevels < doi.doi_indirection))) { + err = dmu_free_long_range(rwa->os, drro->drr_object, 0, DMU_OBJECT_END); if (err != 0) return (SET_ERROR(EINVAL)); - } else if (drro->drr_blksz != doi.doi_data_block_size || - nblkptr < doi.doi_nblkptr || - drro->drr_dn_slots != doi.doi_dnodesize >> DNODE_SHIFT) { - err = dmu_free_long_range(rwa->os, drro->drr_object, - 0, DMU_OBJECT_END); - if (err != 0) - return (SET_ERROR(EINVAL)); } /* * The dmu does not currently support decreasing nlevels * on an object. For non-raw sends, this does not matter * and the new object can just use the previous one's nlevels. * For raw sends, however, the structure of the received dnode * (including nlevels) must match that of the send side. * Therefore, instead of using dmu_object_reclaim(), we must * free the object completely and call dmu_object_claim_dnsize() * instead. */ if ((rwa->raw && drro->drr_nlevels < doi.doi_indirection) || drro->drr_dn_slots != doi.doi_dnodesize >> DNODE_SHIFT) { - if (rwa->raw) { - err = dmu_free_long_object_raw(rwa->os, - drro->drr_object); - } else { - err = dmu_free_long_object(rwa->os, - drro->drr_object); - } + err = dmu_free_long_object(rwa->os, drro->drr_object); if (err != 0) return (SET_ERROR(EINVAL)); txg_wait_synced(dmu_objset_pool(rwa->os), 0); object = DMU_NEW_OBJECT; } } else if (err == EEXIST) { /* * The object requested is currently an interior slot of a * multi-slot dnode. This will be resolved when the next txg * is synced out, since the send stream will have told us * to free this slot when we freed the associated dnode * earlier in the stream. */ txg_wait_synced(dmu_objset_pool(rwa->os), 0); object = drro->drr_object; } else { /* object is free and we are about to allocate a new one */ object = DMU_NEW_OBJECT; } /* * If this is a multi-slot dnode there is a chance that this * object will expand into a slot that is already used by * another object from the previous snapshot. We must free * these objects before we attempt to allocate the new dnode. */ if (drro->drr_dn_slots > 1) { boolean_t need_sync = B_FALSE; for (uint64_t slot = drro->drr_object + 1; slot < drro->drr_object + drro->drr_dn_slots; slot++) { dmu_object_info_t slot_doi; err = dmu_object_info(rwa->os, slot, &slot_doi); if (err == ENOENT || err == EEXIST) continue; else if (err != 0) return (err); - if (rwa->raw) - err = dmu_free_long_object_raw(rwa->os, slot); - else - err = dmu_free_long_object(rwa->os, slot); + err = dmu_free_long_object(rwa->os, slot); if (err != 0) return (err); need_sync = B_TRUE; } if (need_sync) txg_wait_synced(dmu_objset_pool(rwa->os), 0); } tx = dmu_tx_create(rwa->os); dmu_tx_hold_bonus(tx, object); dmu_tx_hold_write(tx, object, 0, 0); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); return (err); } if (object == DMU_NEW_OBJECT) { /* currently free, want to be allocated */ err = dmu_object_claim_dnsize(rwa->os, drro->drr_object, drro->drr_type, drro->drr_blksz, drro->drr_bonustype, drro->drr_bonuslen, drro->drr_dn_slots << DNODE_SHIFT, tx); } else if (drro->drr_type != doi.doi_type || drro->drr_blksz != doi.doi_data_block_size || drro->drr_bonustype != doi.doi_bonus_type || drro->drr_bonuslen != doi.doi_bonus_size) { /* currently allocated, but with different properties */ err = dmu_object_reclaim(rwa->os, drro->drr_object, drro->drr_type, drro->drr_blksz, drro->drr_bonustype, drro->drr_bonuslen, tx); } if (err != 0) { dmu_tx_commit(tx); return (SET_ERROR(EINVAL)); } - if (rwa->raw) { + if (rwa->or_crypt_params_present) { /* - * Convert the buffer associated with this range of dnodes - * to a raw buffer. This ensures that it will be written out - * as a raw buffer when we fill in the dnode object. Since we - * are committing this tx now, it is possible for the dnode - * block to end up on-disk with the incorrect MAC. Despite - * this, the dataset is marked as inconsistent so no other - * code paths (apart from scrubs) will attempt to read this - * data. Scrubs will not be effected by this either since - * scrubs only read raw data and do not attempt to check - * the MAC. + * Set the crypt params for the buffer associated with this + * range of dnodes. This causes the blkptr_t to have the + * same crypt params (byteorder, salt, iv, mac) as on the + * sending side. + * + * Since we are committing this tx now, it is possible for + * the dnode block to end up on-disk with the incorrect MAC, + * if subsequent objects in this block are received in a + * different txg. However, since the dataset is marked as + * inconsistent, no code paths will do a non-raw read (or + * decrypt the block / verify the MAC). The receive code and + * scrub code can safely do raw reads and verify the + * checksum. They don't need to verify the MAC. */ - err = dmu_convert_mdn_block_to_raw(rwa->os, rwa->or_firstobj, - rwa->or_byteorder, rwa->or_salt, rwa->or_iv, rwa->or_mac, - tx); + dmu_buf_t *db = NULL; + uint64_t offset = rwa->or_firstobj * DNODE_MIN_SIZE; + + err = dmu_buf_hold_by_dnode(DMU_META_DNODE(rwa->os), + offset, FTAG, &db, DMU_READ_PREFETCH | DMU_READ_NO_DECRYPT); if (err != 0) { dmu_tx_commit(tx); return (SET_ERROR(EINVAL)); } + + dmu_buf_set_crypt_params(db, rwa->or_byteorder, + rwa->or_salt, rwa->or_iv, rwa->or_mac, tx); + + dmu_buf_rele(db, FTAG); + + rwa->or_crypt_params_present = B_FALSE; } dmu_object_set_checksum(rwa->os, drro->drr_object, drro->drr_checksumtype, tx); dmu_object_set_compress(rwa->os, drro->drr_object, drro->drr_compress, tx); /* handle more restrictive dnode structuring for raw recvs */ if (rwa->raw) { /* * Set the indirect block shift and nlevels. This will not fail * because we ensured all of the blocks were free earlier if * this is a new object. */ VERIFY0(dmu_object_set_blocksize(rwa->os, drro->drr_object, drro->drr_blksz, drro->drr_indblkshift, tx)); VERIFY0(dmu_object_set_nlevels(rwa->os, drro->drr_object, drro->drr_nlevels, tx)); VERIFY0(dmu_object_set_maxblkid(rwa->os, drro->drr_object, drro->drr_maxblkid, tx)); } if (data != NULL) { dmu_buf_t *db; uint32_t flags = DMU_READ_NO_PREFETCH; if (rwa->raw) flags |= DMU_READ_NO_DECRYPT; VERIFY0(dmu_bonus_hold_impl(rwa->os, drro->drr_object, FTAG, flags, &db)); dmu_buf_will_dirty(db, tx); ASSERT3U(db->db_size, >=, drro->drr_bonuslen); bcopy(data, db->db_data, DRR_OBJECT_PAYLOAD_SIZE(drro)); /* * Raw bonus buffers have their byteorder determined by the * DRR_OBJECT_RANGE record. */ if (rwa->byteswap && !rwa->raw) { dmu_object_byteswap_t byteswap = DMU_OT_BYTESWAP(drro->drr_bonustype); dmu_ot_byteswap[byteswap].ob_func(db->db_data, DRR_OBJECT_PAYLOAD_SIZE(drro)); } dmu_buf_rele(db, FTAG); } dmu_tx_commit(tx); return (0); } /* ARGSUSED */ noinline static int receive_freeobjects(struct receive_writer_arg *rwa, struct drr_freeobjects *drrfo) { uint64_t obj; int next_err = 0; if (drrfo->drr_firstobj + drrfo->drr_numobjs < drrfo->drr_firstobj) return (SET_ERROR(EINVAL)); for (obj = drrfo->drr_firstobj == 0 ? 1 : drrfo->drr_firstobj; obj < drrfo->drr_firstobj + drrfo->drr_numobjs && next_err == 0; next_err = dmu_object_next(rwa->os, &obj, FALSE, 0)) { dmu_object_info_t doi; int err; err = dmu_object_info(rwa->os, obj, &doi); if (err == ENOENT) continue; else if (err != 0) return (err); - if (rwa->raw) - err = dmu_free_long_object_raw(rwa->os, obj); - else - err = dmu_free_long_object(rwa->os, obj); + err = dmu_free_long_object(rwa->os, obj); if (err != 0) return (err); if (obj > rwa->max_object) rwa->max_object = obj; } if (next_err != ESRCH) return (next_err); return (0); } noinline static int receive_write(struct receive_writer_arg *rwa, struct drr_write *drrw, arc_buf_t *abuf) { int err; dmu_tx_t *tx; dnode_t *dn; if (drrw->drr_offset + drrw->drr_logical_size < drrw->drr_offset || !DMU_OT_IS_VALID(drrw->drr_type)) return (SET_ERROR(EINVAL)); /* * For resuming to work, records must be in increasing order * by (object, offset). */ if (drrw->drr_object < rwa->last_object || (drrw->drr_object == rwa->last_object && drrw->drr_offset < rwa->last_offset)) { return (SET_ERROR(EINVAL)); } rwa->last_object = drrw->drr_object; rwa->last_offset = drrw->drr_offset; if (rwa->last_object > rwa->max_object) rwa->max_object = rwa->last_object; if (dmu_object_info(rwa->os, drrw->drr_object, NULL) != 0) return (SET_ERROR(EINVAL)); tx = dmu_tx_create(rwa->os); dmu_tx_hold_write(tx, drrw->drr_object, drrw->drr_offset, drrw->drr_logical_size); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); return (err); } - if (rwa->raw) - VERIFY0(dmu_object_dirty_raw(rwa->os, drrw->drr_object, tx)); - if (rwa->byteswap && !arc_is_encrypted(abuf) && arc_get_compression(abuf) == ZIO_COMPRESS_OFF) { dmu_object_byteswap_t byteswap = DMU_OT_BYTESWAP(drrw->drr_type); dmu_ot_byteswap[byteswap].ob_func(abuf->b_data, DRR_WRITE_PAYLOAD_SIZE(drrw)); } VERIFY0(dnode_hold(rwa->os, drrw->drr_object, FTAG, &dn)); dmu_assign_arcbuf_by_dnode(dn, drrw->drr_offset, abuf, tx); dnode_rele(dn, FTAG); /* * Note: If the receive fails, we want the resume stream to start * with the same record that we last successfully received (as opposed * to the next record), so that we can verify that we are * resuming from the correct location. */ save_resume_state(rwa, drrw->drr_object, drrw->drr_offset, tx); dmu_tx_commit(tx); return (0); } /* * Handle a DRR_WRITE_BYREF record. This record is used in dedup'ed * streams to refer to a copy of the data that is already on the * system because it came in earlier in the stream. This function * finds the earlier copy of the data, and uses that copy instead of * data from the stream to fulfill this write. */ static int receive_write_byref(struct receive_writer_arg *rwa, struct drr_write_byref *drrwbr) { dmu_tx_t *tx; int err; guid_map_entry_t gmesrch; guid_map_entry_t *gmep; avl_index_t where; objset_t *ref_os = NULL; int flags = DMU_READ_PREFETCH; dmu_buf_t *dbp; if (drrwbr->drr_offset + drrwbr->drr_length < drrwbr->drr_offset) return (SET_ERROR(EINVAL)); /* * If the GUID of the referenced dataset is different from the * GUID of the target dataset, find the referenced dataset. */ if (drrwbr->drr_toguid != drrwbr->drr_refguid) { gmesrch.guid = drrwbr->drr_refguid; if ((gmep = avl_find(rwa->guid_to_ds_map, &gmesrch, &where)) == NULL) { return (SET_ERROR(EINVAL)); } if (dmu_objset_from_ds(gmep->gme_ds, &ref_os)) return (SET_ERROR(EINVAL)); } else { ref_os = rwa->os; } if (drrwbr->drr_object > rwa->max_object) rwa->max_object = drrwbr->drr_object; if (rwa->raw) flags |= DMU_READ_NO_DECRYPT; /* may return either a regular db or an encrypted one */ err = dmu_buf_hold(ref_os, drrwbr->drr_refobject, drrwbr->drr_refoffset, FTAG, &dbp, flags); if (err != 0) return (err); tx = dmu_tx_create(rwa->os); dmu_tx_hold_write(tx, drrwbr->drr_object, drrwbr->drr_offset, drrwbr->drr_length); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); return (err); } if (rwa->raw) { - VERIFY0(dmu_object_dirty_raw(rwa->os, drrwbr->drr_object, tx)); dmu_copy_from_buf(rwa->os, drrwbr->drr_object, drrwbr->drr_offset, dbp, tx); } else { dmu_write(rwa->os, drrwbr->drr_object, drrwbr->drr_offset, drrwbr->drr_length, dbp->db_data, tx); } dmu_buf_rele(dbp, FTAG); /* See comment in restore_write. */ save_resume_state(rwa, drrwbr->drr_object, drrwbr->drr_offset, tx); dmu_tx_commit(tx); return (0); } static int receive_write_embedded(struct receive_writer_arg *rwa, struct drr_write_embedded *drrwe, void *data) { dmu_tx_t *tx; int err; if (drrwe->drr_offset + drrwe->drr_length < drrwe->drr_offset) return (SET_ERROR(EINVAL)); if (drrwe->drr_psize > BPE_PAYLOAD_SIZE) return (SET_ERROR(EINVAL)); if (drrwe->drr_etype >= NUM_BP_EMBEDDED_TYPES) return (SET_ERROR(EINVAL)); if (drrwe->drr_compression >= ZIO_COMPRESS_FUNCTIONS) return (SET_ERROR(EINVAL)); if (rwa->raw) return (SET_ERROR(EINVAL)); if (drrwe->drr_object > rwa->max_object) rwa->max_object = drrwe->drr_object; tx = dmu_tx_create(rwa->os); dmu_tx_hold_write(tx, drrwe->drr_object, drrwe->drr_offset, drrwe->drr_length); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_tx_abort(tx); return (err); } dmu_write_embedded(rwa->os, drrwe->drr_object, drrwe->drr_offset, data, drrwe->drr_etype, drrwe->drr_compression, drrwe->drr_lsize, drrwe->drr_psize, rwa->byteswap ^ ZFS_HOST_BYTEORDER, tx); /* See comment in restore_write. */ save_resume_state(rwa, drrwe->drr_object, drrwe->drr_offset, tx); dmu_tx_commit(tx); return (0); } static int receive_spill(struct receive_writer_arg *rwa, struct drr_spill *drrs, arc_buf_t *abuf) { dmu_tx_t *tx; dmu_buf_t *db, *db_spill; int err; if (drrs->drr_length < SPA_MINBLOCKSIZE || drrs->drr_length > spa_maxblocksize(dmu_objset_spa(rwa->os))) return (SET_ERROR(EINVAL)); if (rwa->raw) { if (!DMU_OT_IS_VALID(drrs->drr_type) || drrs->drr_compressiontype >= ZIO_COMPRESS_FUNCTIONS || drrs->drr_compressed_size == 0) return (SET_ERROR(EINVAL)); } if (dmu_object_info(rwa->os, drrs->drr_object, NULL) != 0) return (SET_ERROR(EINVAL)); if (drrs->drr_object > rwa->max_object) rwa->max_object = drrs->drr_object; VERIFY0(dmu_bonus_hold(rwa->os, drrs->drr_object, FTAG, &db)); if ((err = dmu_spill_hold_by_bonus(db, FTAG, &db_spill)) != 0) { dmu_buf_rele(db, FTAG); return (err); } tx = dmu_tx_create(rwa->os); dmu_tx_hold_spill(tx, db->db_object); err = dmu_tx_assign(tx, TXG_WAIT); if (err != 0) { dmu_buf_rele(db, FTAG); dmu_buf_rele(db_spill, FTAG); dmu_tx_abort(tx); return (err); } - - if (rwa->raw) { - VERIFY0(dmu_object_dirty_raw(rwa->os, drrs->drr_object, tx)); - dmu_buf_will_change_crypt_params(db_spill, tx); - } else { - dmu_buf_will_dirty(db_spill, tx); - } + dmu_buf_will_dirty(db_spill, tx); if (db_spill->db_size < drrs->drr_length) VERIFY(0 == dbuf_spill_set_blksz(db_spill, drrs->drr_length, tx)); if (rwa->byteswap && !arc_is_encrypted(abuf) && arc_get_compression(abuf) == ZIO_COMPRESS_OFF) { dmu_object_byteswap_t byteswap = DMU_OT_BYTESWAP(drrs->drr_type); dmu_ot_byteswap[byteswap].ob_func(abuf->b_data, DRR_SPILL_PAYLOAD_SIZE(drrs)); } dbuf_assign_arcbuf((dmu_buf_impl_t *)db_spill, abuf, tx); dmu_buf_rele(db, FTAG); dmu_buf_rele(db_spill, FTAG); dmu_tx_commit(tx); return (0); } /* ARGSUSED */ noinline static int receive_free(struct receive_writer_arg *rwa, struct drr_free *drrf) { int err; if (drrf->drr_length != DMU_OBJECT_END && drrf->drr_offset + drrf->drr_length < drrf->drr_offset) return (SET_ERROR(EINVAL)); if (dmu_object_info(rwa->os, drrf->drr_object, NULL) != 0) return (SET_ERROR(EINVAL)); if (drrf->drr_object > rwa->max_object) rwa->max_object = drrf->drr_object; - if (rwa->raw) { - err = dmu_free_long_range_raw(rwa->os, drrf->drr_object, - drrf->drr_offset, drrf->drr_length); - } else { - err = dmu_free_long_range(rwa->os, drrf->drr_object, - drrf->drr_offset, drrf->drr_length); - } + err = dmu_free_long_range(rwa->os, drrf->drr_object, + drrf->drr_offset, drrf->drr_length); return (err); } static int receive_object_range(struct receive_writer_arg *rwa, struct drr_object_range *drror) { /* * By default, we assume this block is in our native format * (ZFS_HOST_BYTEORDER). We then take into account whether * the send stream is byteswapped (rwa->byteswap). Finally, * we need to byteswap again if this particular block was * in non-native format on the send side. */ boolean_t byteorder = ZFS_HOST_BYTEORDER ^ rwa->byteswap ^ !!DRR_IS_RAW_BYTESWAPPED(drror->drr_flags); /* * Since dnode block sizes are constant, we should not need to worry * about making sure that the dnode block size is the same on the * sending and receiving sides for the time being. For non-raw sends, * this does not matter (and in fact we do not send a DRR_OBJECT_RANGE * record at all). Raw sends require this record type because the * encryption parameters are used to protect an entire block of bonus * buffers. If the size of dnode blocks ever becomes variable, * handling will need to be added to ensure that dnode block sizes * match on the sending and receiving side. */ if (drror->drr_numslots != DNODES_PER_BLOCK || P2PHASE(drror->drr_firstobj, DNODES_PER_BLOCK) != 0 || !rwa->raw) return (SET_ERROR(EINVAL)); if (drror->drr_firstobj > rwa->max_object) rwa->max_object = drror->drr_firstobj; /* * The DRR_OBJECT_RANGE handling must be deferred to receive_object() - * so that the encryption parameters are set with each object that is - * written into that block. + * so that the block of dnodes is not written out when it's empty, + * and converted to a HOLE BP. */ + rwa->or_crypt_params_present = B_TRUE; rwa->or_firstobj = drror->drr_firstobj; rwa->or_numslots = drror->drr_numslots; bcopy(drror->drr_salt, rwa->or_salt, ZIO_DATA_SALT_LEN); bcopy(drror->drr_iv, rwa->or_iv, ZIO_DATA_IV_LEN); bcopy(drror->drr_mac, rwa->or_mac, ZIO_DATA_MAC_LEN); rwa->or_byteorder = byteorder; return (0); } /* used to destroy the drc_ds on error */ static void dmu_recv_cleanup_ds(dmu_recv_cookie_t *drc) { dsl_dataset_t *ds = drc->drc_ds; ds_hold_flags_t dsflags = (drc->drc_raw) ? 0 : DS_HOLD_FLAG_DECRYPT; /* * Wait for the txg sync before cleaning up the receive. For * resumable receives, this ensures that our resume state has * been written out to disk. For raw receives, this ensures * that the user accounting code will not attempt to do anything * after we stopped receiving the dataset. */ txg_wait_synced(ds->ds_dir->dd_pool, 0); + ds->ds_objset->os_raw_receive = B_FALSE; rrw_enter(&ds->ds_bp_rwlock, RW_READER, FTAG); if (drc->drc_resumable && !BP_IS_HOLE(dsl_dataset_get_blkptr(ds))) { rrw_exit(&ds->ds_bp_rwlock, FTAG); dsl_dataset_disown(ds, dsflags, dmu_recv_tag); } else { char name[ZFS_MAX_DATASET_NAME_LEN]; rrw_exit(&ds->ds_bp_rwlock, FTAG); dsl_dataset_name(ds, name); dsl_dataset_disown(ds, dsflags, dmu_recv_tag); (void) dsl_destroy_head(name); } } static void receive_cksum(struct receive_arg *ra, int len, void *buf) { if (ra->byteswap) { (void) fletcher_4_incremental_byteswap(buf, len, &ra->cksum); } else { (void) fletcher_4_incremental_native(buf, len, &ra->cksum); } } /* * Read the payload into a buffer of size len, and update the current record's * payload field. * Allocate ra->next_rrd and read the next record's header into * ra->next_rrd->header. * Verify checksum of payload and next record. */ static int receive_read_payload_and_next_header(struct receive_arg *ra, int len, void *buf) { int err; zio_cksum_t cksum_orig; zio_cksum_t *cksump; if (len != 0) { ASSERT3U(len, <=, SPA_MAXBLOCKSIZE); err = receive_read(ra, len, buf); if (err != 0) return (err); receive_cksum(ra, len, buf); /* note: rrd is NULL when reading the begin record's payload */ if (ra->rrd != NULL) { ra->rrd->payload = buf; ra->rrd->payload_size = len; ra->rrd->bytes_read = ra->bytes_read; } } ra->prev_cksum = ra->cksum; ra->next_rrd = kmem_zalloc(sizeof (*ra->next_rrd), KM_SLEEP); err = receive_read(ra, sizeof (ra->next_rrd->header), &ra->next_rrd->header); ra->next_rrd->bytes_read = ra->bytes_read; if (err != 0) { kmem_free(ra->next_rrd, sizeof (*ra->next_rrd)); ra->next_rrd = NULL; return (err); } if (ra->next_rrd->header.drr_type == DRR_BEGIN) { kmem_free(ra->next_rrd, sizeof (*ra->next_rrd)); ra->next_rrd = NULL; return (SET_ERROR(EINVAL)); } /* * Note: checksum is of everything up to but not including the * checksum itself. */ ASSERT3U(offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), ==, sizeof (dmu_replay_record_t) - sizeof (zio_cksum_t)); receive_cksum(ra, offsetof(dmu_replay_record_t, drr_u.drr_checksum.drr_checksum), &ra->next_rrd->header); cksum_orig = ra->next_rrd->header.drr_u.drr_checksum.drr_checksum; cksump = &ra->next_rrd->header.drr_u.drr_checksum.drr_checksum; if (ra->byteswap) byteswap_record(&ra->next_rrd->header); if ((!ZIO_CHECKSUM_IS_ZERO(cksump)) && !ZIO_CHECKSUM_EQUAL(ra->cksum, *cksump)) { kmem_free(ra->next_rrd, sizeof (*ra->next_rrd)); ra->next_rrd = NULL; return (SET_ERROR(ECKSUM)); } receive_cksum(ra, sizeof (cksum_orig), &cksum_orig); return (0); } static void objlist_create(struct objlist *list) { list_create(&list->list, sizeof (struct receive_objnode), offsetof(struct receive_objnode, node)); list->last_lookup = 0; } static void objlist_destroy(struct objlist *list) { for (struct receive_objnode *n = list_remove_head(&list->list); n != NULL; n = list_remove_head(&list->list)) { kmem_free(n, sizeof (*n)); } list_destroy(&list->list); } /* * This function looks through the objlist to see if the specified object number * is contained in the objlist. In the process, it will remove all object * numbers in the list that are smaller than the specified object number. Thus, * any lookup of an object number smaller than a previously looked up object * number will always return false; therefore, all lookups should be done in * ascending order. */ static boolean_t objlist_exists(struct objlist *list, uint64_t object) { struct receive_objnode *node = list_head(&list->list); ASSERT3U(object, >=, list->last_lookup); list->last_lookup = object; while (node != NULL && node->object < object) { VERIFY3P(node, ==, list_remove_head(&list->list)); kmem_free(node, sizeof (*node)); node = list_head(&list->list); } return (node != NULL && node->object == object); } /* * The objlist is a list of object numbers stored in ascending order. However, * the insertion of new object numbers does not seek out the correct location to * store a new object number; instead, it appends it to the list for simplicity. * Thus, any users must take care to only insert new object numbers in ascending * order. */ static void objlist_insert(struct objlist *list, uint64_t object) { struct receive_objnode *node = kmem_zalloc(sizeof (*node), KM_SLEEP); node->object = object; #ifdef ZFS_DEBUG { struct receive_objnode *last_object = list_tail(&list->list); uint64_t last_objnum = (last_object != NULL ? last_object->object : 0); ASSERT3U(node->object, >, last_objnum); } #endif list_insert_tail(&list->list, node); } /* * Issue the prefetch reads for any necessary indirect blocks. * * We use the object ignore list to tell us whether or not to issue prefetches * for a given object. We do this for both correctness (in case the blocksize * of an object has changed) and performance (if the object doesn't exist, don't * needlessly try to issue prefetches). We also trim the list as we go through * the stream to prevent it from growing to an unbounded size. * * The object numbers within will always be in sorted order, and any write * records we see will also be in sorted order, but they're not sorted with * respect to each other (i.e. we can get several object records before * receiving each object's write records). As a result, once we've reached a * given object number, we can safely remove any reference to lower object * numbers in the ignore list. In practice, we receive up to 32 object records * before receiving write records, so the list can have up to 32 nodes in it. */ /* ARGSUSED */ static void receive_read_prefetch(struct receive_arg *ra, uint64_t object, uint64_t offset, uint64_t length) { if (!objlist_exists(&ra->ignore_objlist, object)) { dmu_prefetch(ra->os, object, 1, offset, length, ZIO_PRIORITY_SYNC_READ); } } /* * Read records off the stream, issuing any necessary prefetches. */ static int receive_read_record(struct receive_arg *ra) { int err; switch (ra->rrd->header.drr_type) { case DRR_OBJECT: { struct drr_object *drro = &ra->rrd->header.drr_u.drr_object; uint32_t size = DRR_OBJECT_PAYLOAD_SIZE(drro); void *buf = kmem_zalloc(size, KM_SLEEP); dmu_object_info_t doi; err = receive_read_payload_and_next_header(ra, size, buf); if (err != 0) { kmem_free(buf, size); return (err); } err = dmu_object_info(ra->os, drro->drr_object, &doi); /* * See receive_read_prefetch for an explanation why we're * storing this object in the ignore_obj_list. */ if (err == ENOENT || err == EEXIST || (err == 0 && doi.doi_data_block_size != drro->drr_blksz)) { objlist_insert(&ra->ignore_objlist, drro->drr_object); err = 0; } return (err); } case DRR_FREEOBJECTS: { err = receive_read_payload_and_next_header(ra, 0, NULL); return (err); } case DRR_WRITE: { struct drr_write *drrw = &ra->rrd->header.drr_u.drr_write; arc_buf_t *abuf; boolean_t is_meta = DMU_OT_IS_METADATA(drrw->drr_type); if (ra->raw) { boolean_t byteorder = ZFS_HOST_BYTEORDER ^ !!DRR_IS_RAW_BYTESWAPPED(drrw->drr_flags) ^ ra->byteswap; abuf = arc_loan_raw_buf(dmu_objset_spa(ra->os), drrw->drr_object, byteorder, drrw->drr_salt, drrw->drr_iv, drrw->drr_mac, drrw->drr_type, drrw->drr_compressed_size, drrw->drr_logical_size, drrw->drr_compressiontype); } else if (DRR_WRITE_COMPRESSED(drrw)) { ASSERT3U(drrw->drr_compressed_size, >, 0); ASSERT3U(drrw->drr_logical_size, >=, drrw->drr_compressed_size); ASSERT(!is_meta); abuf = arc_loan_compressed_buf( dmu_objset_spa(ra->os), drrw->drr_compressed_size, drrw->drr_logical_size, drrw->drr_compressiontype); } else { abuf = arc_loan_buf(dmu_objset_spa(ra->os), is_meta, drrw->drr_logical_size); } err = receive_read_payload_and_next_header(ra, DRR_WRITE_PAYLOAD_SIZE(drrw), abuf->b_data); if (err != 0) { dmu_return_arcbuf(abuf); return (err); } ra->rrd->arc_buf = abuf; receive_read_prefetch(ra, drrw->drr_object, drrw->drr_offset, drrw->drr_logical_size); return (err); } case DRR_WRITE_BYREF: { struct drr_write_byref *drrwb = &ra->rrd->header.drr_u.drr_write_byref; err = receive_read_payload_and_next_header(ra, 0, NULL); receive_read_prefetch(ra, drrwb->drr_object, drrwb->drr_offset, drrwb->drr_length); return (err); } case DRR_WRITE_EMBEDDED: { struct drr_write_embedded *drrwe = &ra->rrd->header.drr_u.drr_write_embedded; uint32_t size = P2ROUNDUP(drrwe->drr_psize, 8); void *buf = kmem_zalloc(size, KM_SLEEP); err = receive_read_payload_and_next_header(ra, size, buf); if (err != 0) { kmem_free(buf, size); return (err); } receive_read_prefetch(ra, drrwe->drr_object, drrwe->drr_offset, drrwe->drr_length); return (err); } case DRR_FREE: { /* * It might be beneficial to prefetch indirect blocks here, but * we don't really have the data to decide for sure. */ err = receive_read_payload_and_next_header(ra, 0, NULL); return (err); } case DRR_END: { struct drr_end *drre = &ra->rrd->header.drr_u.drr_end; if (!ZIO_CHECKSUM_EQUAL(ra->prev_cksum, drre->drr_checksum)) return (SET_ERROR(ECKSUM)); return (0); } case DRR_SPILL: { struct drr_spill *drrs = &ra->rrd->header.drr_u.drr_spill; arc_buf_t *abuf; int len = DRR_SPILL_PAYLOAD_SIZE(drrs); /* DRR_SPILL records are either raw or uncompressed */ if (ra->raw) { boolean_t byteorder = ZFS_HOST_BYTEORDER ^ !!DRR_IS_RAW_BYTESWAPPED(drrs->drr_flags) ^ ra->byteswap; abuf = arc_loan_raw_buf(dmu_objset_spa(ra->os), drrs->drr_object, byteorder, drrs->drr_salt, drrs->drr_iv, drrs->drr_mac, drrs->drr_type, drrs->drr_compressed_size, drrs->drr_length, drrs->drr_compressiontype); } else { abuf = arc_loan_buf(dmu_objset_spa(ra->os), DMU_OT_IS_METADATA(drrs->drr_type), drrs->drr_length); } err = receive_read_payload_and_next_header(ra, len, abuf->b_data); if (err != 0) { dmu_return_arcbuf(abuf); return (err); } ra->rrd->arc_buf = abuf; return (err); } case DRR_OBJECT_RANGE: { err = receive_read_payload_and_next_header(ra, 0, NULL); return (err); } default: return (SET_ERROR(EINVAL)); } } static void dprintf_drr(struct receive_record_arg *rrd, int err) { #ifdef ZFS_DEBUG switch (rrd->header.drr_type) { case DRR_OBJECT: { struct drr_object *drro = &rrd->header.drr_u.drr_object; dprintf("drr_type = OBJECT obj = %llu type = %u " "bonustype = %u blksz = %u bonuslen = %u cksumtype = %u " "compress = %u dn_slots = %u err = %d\n", drro->drr_object, drro->drr_type, drro->drr_bonustype, drro->drr_blksz, drro->drr_bonuslen, drro->drr_checksumtype, drro->drr_compress, drro->drr_dn_slots, err); break; } case DRR_FREEOBJECTS: { struct drr_freeobjects *drrfo = &rrd->header.drr_u.drr_freeobjects; dprintf("drr_type = FREEOBJECTS firstobj = %llu " "numobjs = %llu err = %d\n", drrfo->drr_firstobj, drrfo->drr_numobjs, err); break; } case DRR_WRITE: { struct drr_write *drrw = &rrd->header.drr_u.drr_write; dprintf("drr_type = WRITE obj = %llu type = %u offset = %llu " "lsize = %llu cksumtype = %u cksumflags = %u " "compress = %u psize = %llu err = %d\n", drrw->drr_object, drrw->drr_type, drrw->drr_offset, drrw->drr_logical_size, drrw->drr_checksumtype, drrw->drr_flags, drrw->drr_compressiontype, drrw->drr_compressed_size, err); break; } case DRR_WRITE_BYREF: { struct drr_write_byref *drrwbr = &rrd->header.drr_u.drr_write_byref; dprintf("drr_type = WRITE_BYREF obj = %llu offset = %llu " "length = %llu toguid = %llx refguid = %llx " "refobject = %llu refoffset = %llu cksumtype = %u " "cksumflags = %u err = %d\n", drrwbr->drr_object, drrwbr->drr_offset, drrwbr->drr_length, drrwbr->drr_toguid, drrwbr->drr_refguid, drrwbr->drr_refobject, drrwbr->drr_refoffset, drrwbr->drr_checksumtype, drrwbr->drr_flags, err); break; } case DRR_WRITE_EMBEDDED: { struct drr_write_embedded *drrwe = &rrd->header.drr_u.drr_write_embedded; dprintf("drr_type = WRITE_EMBEDDED obj = %llu offset = %llu " "length = %llu compress = %u etype = %u lsize = %u " "psize = %u err = %d\n", drrwe->drr_object, drrwe->drr_offset, drrwe->drr_length, drrwe->drr_compression, drrwe->drr_etype, drrwe->drr_lsize, drrwe->drr_psize, err); break; } case DRR_FREE: { struct drr_free *drrf = &rrd->header.drr_u.drr_free; dprintf("drr_type = FREE obj = %llu offset = %llu " "length = %lld err = %d\n", drrf->drr_object, drrf->drr_offset, drrf->drr_length, err); break; } case DRR_SPILL: { struct drr_spill *drrs = &rrd->header.drr_u.drr_spill; dprintf("drr_type = SPILL obj = %llu length = %llu " "err = %d\n", drrs->drr_object, drrs->drr_length, err); break; } default: return; } #endif } /* * Commit the records to the pool. */ static int receive_process_record(struct receive_writer_arg *rwa, struct receive_record_arg *rrd) { int err; /* Processing in order, therefore bytes_read should be increasing. */ ASSERT3U(rrd->bytes_read, >=, rwa->bytes_read); rwa->bytes_read = rrd->bytes_read; switch (rrd->header.drr_type) { case DRR_OBJECT: { struct drr_object *drro = &rrd->header.drr_u.drr_object; err = receive_object(rwa, drro, rrd->payload); kmem_free(rrd->payload, rrd->payload_size); rrd->payload = NULL; break; } case DRR_FREEOBJECTS: { struct drr_freeobjects *drrfo = &rrd->header.drr_u.drr_freeobjects; err = receive_freeobjects(rwa, drrfo); break; } case DRR_WRITE: { struct drr_write *drrw = &rrd->header.drr_u.drr_write; err = receive_write(rwa, drrw, rrd->arc_buf); /* if receive_write() is successful, it consumes the arc_buf */ if (err != 0) dmu_return_arcbuf(rrd->arc_buf); rrd->arc_buf = NULL; rrd->payload = NULL; break; } case DRR_WRITE_BYREF: { struct drr_write_byref *drrwbr = &rrd->header.drr_u.drr_write_byref; err = receive_write_byref(rwa, drrwbr); break; } case DRR_WRITE_EMBEDDED: { struct drr_write_embedded *drrwe = &rrd->header.drr_u.drr_write_embedded; err = receive_write_embedded(rwa, drrwe, rrd->payload); kmem_free(rrd->payload, rrd->payload_size); rrd->payload = NULL; break; } case DRR_FREE: { struct drr_free *drrf = &rrd->header.drr_u.drr_free; err = receive_free(rwa, drrf); break; } case DRR_SPILL: { struct drr_spill *drrs = &rrd->header.drr_u.drr_spill; err = receive_spill(rwa, drrs, rrd->arc_buf); /* if receive_spill() is successful, it consumes the arc_buf */ if (err != 0) dmu_return_arcbuf(rrd->arc_buf); rrd->arc_buf = NULL; rrd->payload = NULL; break; } case DRR_OBJECT_RANGE: { struct drr_object_range *drror = &rrd->header.drr_u.drr_object_range; return (receive_object_range(rwa, drror)); } default: return (SET_ERROR(EINVAL)); } if (err != 0) dprintf_drr(rrd, err); return (err); } /* * dmu_recv_stream's worker thread; pull records off the queue, and then call * receive_process_record When we're done, signal the main thread and exit. */ static void receive_writer_thread(void *arg) { struct receive_writer_arg *rwa = arg; struct receive_record_arg *rrd; fstrans_cookie_t cookie = spl_fstrans_mark(); for (rrd = bqueue_dequeue(&rwa->q); !rrd->eos_marker; rrd = bqueue_dequeue(&rwa->q)) { /* * If there's an error, the main thread will stop putting things * on the queue, but we need to clear everything in it before we * can exit. */ if (rwa->err == 0) { rwa->err = receive_process_record(rwa, rrd); } else if (rrd->arc_buf != NULL) { dmu_return_arcbuf(rrd->arc_buf); rrd->arc_buf = NULL; rrd->payload = NULL; } else if (rrd->payload != NULL) { kmem_free(rrd->payload, rrd->payload_size); rrd->payload = NULL; } kmem_free(rrd, sizeof (*rrd)); } kmem_free(rrd, sizeof (*rrd)); mutex_enter(&rwa->mutex); rwa->done = B_TRUE; cv_signal(&rwa->cv); mutex_exit(&rwa->mutex); spl_fstrans_unmark(cookie); thread_exit(); } static int resume_check(struct receive_arg *ra, nvlist_t *begin_nvl) { uint64_t val; objset_t *mos = dmu_objset_pool(ra->os)->dp_meta_objset; uint64_t dsobj = dmu_objset_id(ra->os); uint64_t resume_obj, resume_off; if (nvlist_lookup_uint64(begin_nvl, "resume_object", &resume_obj) != 0 || nvlist_lookup_uint64(begin_nvl, "resume_offset", &resume_off) != 0) { return (SET_ERROR(EINVAL)); } VERIFY0(zap_lookup(mos, dsobj, DS_FIELD_RESUME_OBJECT, sizeof (val), 1, &val)); if (resume_obj != val) return (SET_ERROR(EINVAL)); VERIFY0(zap_lookup(mos, dsobj, DS_FIELD_RESUME_OFFSET, sizeof (val), 1, &val)); if (resume_off != val) return (SET_ERROR(EINVAL)); return (0); } /* * Read in the stream's records, one by one, and apply them to the pool. There * are two threads involved; the thread that calls this function will spin up a * worker thread, read the records off the stream one by one, and issue * prefetches for any necessary indirect blocks. It will then push the records * onto an internal blocking queue. The worker thread will pull the records off * the queue, and actually write the data into the DMU. This way, the worker * thread doesn't have to wait for reads to complete, since everything it needs * (the indirect blocks) will be prefetched. * * NB: callers *must* call dmu_recv_end() if this succeeds. */ int dmu_recv_stream(dmu_recv_cookie_t *drc, vnode_t *vp, offset_t *voffp, int cleanup_fd, uint64_t *action_handlep) { int err = 0; struct receive_arg *ra; struct receive_writer_arg *rwa; int featureflags; uint32_t payloadlen; void *payload; nvlist_t *begin_nvl = NULL; ra = kmem_zalloc(sizeof (*ra), KM_SLEEP); rwa = kmem_zalloc(sizeof (*rwa), KM_SLEEP); ra->byteswap = drc->drc_byteswap; ra->raw = drc->drc_raw; ra->cksum = drc->drc_cksum; ra->vp = vp; ra->voff = *voffp; if (dsl_dataset_is_zapified(drc->drc_ds)) { (void) zap_lookup(drc->drc_ds->ds_dir->dd_pool->dp_meta_objset, drc->drc_ds->ds_object, DS_FIELD_RESUME_BYTES, sizeof (ra->bytes_read), 1, &ra->bytes_read); } objlist_create(&ra->ignore_objlist); /* these were verified in dmu_recv_begin */ ASSERT3U(DMU_GET_STREAM_HDRTYPE(drc->drc_drrb->drr_versioninfo), ==, DMU_SUBSTREAM); ASSERT3U(drc->drc_drrb->drr_type, <, DMU_OST_NUMTYPES); /* * Open the objset we are modifying. */ VERIFY0(dmu_objset_from_ds(drc->drc_ds, &ra->os)); ASSERT(dsl_dataset_phys(drc->drc_ds)->ds_flags & DS_FLAG_INCONSISTENT); featureflags = DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo); ra->featureflags = featureflags; /* embedded data is incompatible with encrypted datasets */ if (ra->os->os_encrypted && (featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)) { err = SET_ERROR(EINVAL); goto out; } /* if this stream is dedup'ed, set up the avl tree for guid mapping */ if (featureflags & DMU_BACKUP_FEATURE_DEDUP) { minor_t minor; if (cleanup_fd == -1) { err = SET_ERROR(EBADF); goto out; } err = zfs_onexit_fd_hold(cleanup_fd, &minor); if (err != 0) { cleanup_fd = -1; goto out; } if (*action_handlep == 0) { rwa->guid_to_ds_map = kmem_alloc(sizeof (avl_tree_t), KM_SLEEP); avl_create(rwa->guid_to_ds_map, guid_compare, sizeof (guid_map_entry_t), offsetof(guid_map_entry_t, avlnode)); err = zfs_onexit_add_cb(minor, free_guid_map_onexit, rwa->guid_to_ds_map, action_handlep); if (err != 0) goto out; } else { err = zfs_onexit_cb_data(minor, *action_handlep, (void **)&rwa->guid_to_ds_map); if (err != 0) goto out; } drc->drc_guid_to_ds_map = rwa->guid_to_ds_map; } payloadlen = drc->drc_drr_begin->drr_payloadlen; payload = NULL; if (payloadlen != 0) payload = kmem_alloc(payloadlen, KM_SLEEP); err = receive_read_payload_and_next_header(ra, payloadlen, payload); if (err != 0) { if (payloadlen != 0) kmem_free(payload, payloadlen); goto out; } if (payloadlen != 0) { err = nvlist_unpack(payload, payloadlen, &begin_nvl, KM_SLEEP); kmem_free(payload, payloadlen); if (err != 0) goto out; } /* handle DSL encryption key payload */ if (featureflags & DMU_BACKUP_FEATURE_RAW) { nvlist_t *keynvl = NULL; ASSERT(ra->os->os_encrypted); ASSERT(drc->drc_raw); err = nvlist_lookup_nvlist(begin_nvl, "crypt_keydata", &keynvl); if (err != 0) goto out; /* * If this is a new dataset we set the key immediately. * Otherwise we don't want to change the key until we * are sure the rest of the receive succeeded so we stash * the keynvl away until then. */ err = dsl_crypto_recv_raw(spa_name(ra->os->os_spa), drc->drc_ds->ds_object, drc->drc_drrb->drr_type, keynvl, drc->drc_newfs); if (err != 0) goto out; if (!drc->drc_newfs) drc->drc_keynvl = fnvlist_dup(keynvl); } if (featureflags & DMU_BACKUP_FEATURE_RESUMING) { err = resume_check(ra, begin_nvl); if (err != 0) goto out; } (void) bqueue_init(&rwa->q, MAX(zfs_recv_queue_length, 2 * zfs_max_recordsize), offsetof(struct receive_record_arg, node)); cv_init(&rwa->cv, NULL, CV_DEFAULT, NULL); mutex_init(&rwa->mutex, NULL, MUTEX_DEFAULT, NULL); rwa->os = ra->os; rwa->byteswap = drc->drc_byteswap; rwa->resumable = drc->drc_resumable; rwa->raw = drc->drc_raw; + rwa->os->os_raw_receive = drc->drc_raw; (void) thread_create(NULL, 0, receive_writer_thread, rwa, 0, curproc, TS_RUN, minclsyspri); /* * We're reading rwa->err without locks, which is safe since we are the * only reader, and the worker thread is the only writer. It's ok if we * miss a write for an iteration or two of the loop, since the writer * thread will keep freeing records we send it until we send it an eos * marker. * * We can leave this loop in 3 ways: First, if rwa->err is * non-zero. In that case, the writer thread will free the rrd we just * pushed. Second, if we're interrupted; in that case, either it's the * first loop and ra->rrd was never allocated, or it's later and ra->rrd * has been handed off to the writer thread who will free it. Finally, * if receive_read_record fails or we're at the end of the stream, then * we free ra->rrd and exit. */ while (rwa->err == 0) { if (issig(JUSTLOOKING) && issig(FORREAL)) { err = SET_ERROR(EINTR); break; } ASSERT3P(ra->rrd, ==, NULL); ra->rrd = ra->next_rrd; ra->next_rrd = NULL; /* Allocates and loads header into ra->next_rrd */ err = receive_read_record(ra); if (ra->rrd->header.drr_type == DRR_END || err != 0) { kmem_free(ra->rrd, sizeof (*ra->rrd)); ra->rrd = NULL; break; } bqueue_enqueue(&rwa->q, ra->rrd, sizeof (struct receive_record_arg) + ra->rrd->payload_size); ra->rrd = NULL; } if (ra->next_rrd == NULL) ra->next_rrd = kmem_zalloc(sizeof (*ra->next_rrd), KM_SLEEP); ra->next_rrd->eos_marker = B_TRUE; bqueue_enqueue(&rwa->q, ra->next_rrd, 1); mutex_enter(&rwa->mutex); while (!rwa->done) { cv_wait(&rwa->cv, &rwa->mutex); } mutex_exit(&rwa->mutex); /* * If we are receiving a full stream as a clone, all object IDs which * are greater than the maximum ID referenced in the stream are * by definition unused and must be freed. */ if (drc->drc_clone && drc->drc_drrb->drr_fromguid == 0) { uint64_t obj = rwa->max_object + 1; int free_err = 0; int next_err = 0; while (next_err == 0) { - if (drc->drc_raw) { - free_err = dmu_free_long_object_raw(rwa->os, - obj); - } else { - free_err = dmu_free_long_object(rwa->os, obj); - } + free_err = dmu_free_long_object(rwa->os, obj); if (free_err != 0 && free_err != ENOENT) break; next_err = dmu_object_next(rwa->os, &obj, FALSE, 0); } if (err == 0) { if (free_err != 0 && free_err != ENOENT) err = free_err; else if (next_err != ESRCH) err = next_err; } } cv_destroy(&rwa->cv); mutex_destroy(&rwa->mutex); bqueue_destroy(&rwa->q); if (err == 0) err = rwa->err; out: nvlist_free(begin_nvl); if ((featureflags & DMU_BACKUP_FEATURE_DEDUP) && (cleanup_fd != -1)) zfs_onexit_fd_rele(cleanup_fd); if (err != 0) { /* * Clean up references. If receive is not resumable, * destroy what we created, so we don't leave it in * the inconsistent state. */ dmu_recv_cleanup_ds(drc); nvlist_free(drc->drc_keynvl); } *voffp = ra->voff; objlist_destroy(&ra->ignore_objlist); kmem_free(ra, sizeof (*ra)); kmem_free(rwa, sizeof (*rwa)); return (err); } static int dmu_recv_end_check(void *arg, dmu_tx_t *tx) { dmu_recv_cookie_t *drc = arg; dsl_pool_t *dp = dmu_tx_pool(tx); int error; ASSERT3P(drc->drc_ds->ds_owner, ==, dmu_recv_tag); if (!drc->drc_newfs) { dsl_dataset_t *origin_head; error = dsl_dataset_hold(dp, drc->drc_tofs, FTAG, &origin_head); if (error != 0) return (error); if (drc->drc_force) { /* * We will destroy any snapshots in tofs (i.e. before * origin_head) that are after the origin (which is * the snap before drc_ds, because drc_ds can not * have any snaps of its own). */ uint64_t obj; obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj; while (obj != dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) { dsl_dataset_t *snap; error = dsl_dataset_hold_obj(dp, obj, FTAG, &snap); if (error != 0) break; if (snap->ds_dir != origin_head->ds_dir) error = SET_ERROR(EINVAL); if (error == 0) { error = dsl_destroy_snapshot_check_impl( snap, B_FALSE); } obj = dsl_dataset_phys(snap)->ds_prev_snap_obj; dsl_dataset_rele(snap, FTAG); if (error != 0) break; } if (error != 0) { dsl_dataset_rele(origin_head, FTAG); return (error); } } if (drc->drc_keynvl != NULL) { error = dsl_crypto_recv_raw_key_check(drc->drc_ds, drc->drc_keynvl, tx); if (error != 0) { dsl_dataset_rele(origin_head, FTAG); return (error); } } error = dsl_dataset_clone_swap_check_impl(drc->drc_ds, origin_head, drc->drc_force, drc->drc_owner, tx); if (error != 0) { dsl_dataset_rele(origin_head, FTAG); return (error); } error = dsl_dataset_snapshot_check_impl(origin_head, drc->drc_tosnap, tx, B_TRUE, 1, drc->drc_cred); dsl_dataset_rele(origin_head, FTAG); if (error != 0) return (error); error = dsl_destroy_head_check_impl(drc->drc_ds, 1); } else { error = dsl_dataset_snapshot_check_impl(drc->drc_ds, drc->drc_tosnap, tx, B_TRUE, 1, drc->drc_cred); } return (error); } static void dmu_recv_end_sync(void *arg, dmu_tx_t *tx) { dmu_recv_cookie_t *drc = arg; dsl_pool_t *dp = dmu_tx_pool(tx); boolean_t encrypted = drc->drc_ds->ds_dir->dd_crypto_obj != 0; spa_history_log_internal_ds(drc->drc_ds, "finish receiving", tx, "snap=%s", drc->drc_tosnap); + drc->drc_ds->ds_objset->os_raw_receive = B_FALSE; if (!drc->drc_newfs) { dsl_dataset_t *origin_head; VERIFY0(dsl_dataset_hold(dp, drc->drc_tofs, FTAG, &origin_head)); if (drc->drc_force) { /* * Destroy any snapshots of drc_tofs (origin_head) * after the origin (the snap before drc_ds). */ uint64_t obj; obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj; while (obj != dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) { dsl_dataset_t *snap; VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG, &snap)); ASSERT3P(snap->ds_dir, ==, origin_head->ds_dir); obj = dsl_dataset_phys(snap)->ds_prev_snap_obj; dsl_destroy_snapshot_sync_impl(snap, B_FALSE, tx); dsl_dataset_rele(snap, FTAG); } } if (drc->drc_keynvl != NULL) { dsl_crypto_recv_raw_key_sync(drc->drc_ds, drc->drc_keynvl, tx); nvlist_free(drc->drc_keynvl); drc->drc_keynvl = NULL; } VERIFY3P(drc->drc_ds->ds_prev, ==, origin_head->ds_prev); dsl_dataset_clone_swap_sync_impl(drc->drc_ds, origin_head, tx); dsl_dataset_snapshot_sync_impl(origin_head, drc->drc_tosnap, tx); /* set snapshot's creation time and guid */ dmu_buf_will_dirty(origin_head->ds_prev->ds_dbuf, tx); dsl_dataset_phys(origin_head->ds_prev)->ds_creation_time = drc->drc_drrb->drr_creation_time; dsl_dataset_phys(origin_head->ds_prev)->ds_guid = drc->drc_drrb->drr_toguid; dsl_dataset_phys(origin_head->ds_prev)->ds_flags &= ~DS_FLAG_INCONSISTENT; dmu_buf_will_dirty(origin_head->ds_dbuf, tx); dsl_dataset_phys(origin_head)->ds_flags &= ~DS_FLAG_INCONSISTENT; drc->drc_newsnapobj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj; dsl_dataset_rele(origin_head, FTAG); dsl_destroy_head_sync_impl(drc->drc_ds, tx); if (drc->drc_owner != NULL) VERIFY3P(origin_head->ds_owner, ==, drc->drc_owner); } else { dsl_dataset_t *ds = drc->drc_ds; dsl_dataset_snapshot_sync_impl(ds, drc->drc_tosnap, tx); /* set snapshot's creation time and guid */ dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx); dsl_dataset_phys(ds->ds_prev)->ds_creation_time = drc->drc_drrb->drr_creation_time; dsl_dataset_phys(ds->ds_prev)->ds_guid = drc->drc_drrb->drr_toguid; dsl_dataset_phys(ds->ds_prev)->ds_flags &= ~DS_FLAG_INCONSISTENT; dmu_buf_will_dirty(ds->ds_dbuf, tx); dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT; if (dsl_dataset_has_resume_receive_state(ds)) { (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_FROMGUID, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_OBJECT, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_OFFSET, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_BYTES, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TOGUID, tx); (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_TONAME, tx); } drc->drc_newsnapobj = dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj; } zvol_create_minors(dp->dp_spa, drc->drc_tofs, B_TRUE); /* * Release the hold from dmu_recv_begin. This must be done before * we return to open context, so that when we free the dataset's dnode * we can evict its bonus buffer. Since the dataset may be destroyed * at this point (and therefore won't have a valid pointer to the spa) * we release the key mapping manually here while we do have a valid * pointer, if it exists. */ if (!drc->drc_raw && encrypted) { (void) spa_keystore_remove_mapping(dmu_tx_pool(tx)->dp_spa, drc->drc_ds->ds_object, drc->drc_ds); } dsl_dataset_disown(drc->drc_ds, 0, dmu_recv_tag); drc->drc_ds = NULL; } static int add_ds_to_guidmap(const char *name, avl_tree_t *guid_map, uint64_t snapobj, boolean_t raw) { dsl_pool_t *dp; dsl_dataset_t *snapds; guid_map_entry_t *gmep; ds_hold_flags_t dsflags = (raw) ? 0 : DS_HOLD_FLAG_DECRYPT; int err; ASSERT(guid_map != NULL); err = dsl_pool_hold(name, FTAG, &dp); if (err != 0) return (err); gmep = kmem_alloc(sizeof (*gmep), KM_SLEEP); err = dsl_dataset_hold_obj_flags(dp, snapobj, dsflags, gmep, &snapds); if (err == 0) { gmep->guid = dsl_dataset_phys(snapds)->ds_guid; gmep->raw = raw; gmep->gme_ds = snapds; avl_add(guid_map, gmep); dsl_dataset_long_hold(snapds, gmep); } else { kmem_free(gmep, sizeof (*gmep)); } dsl_pool_rele(dp, FTAG); return (err); } static int dmu_recv_end_modified_blocks = 3; static int dmu_recv_existing_end(dmu_recv_cookie_t *drc) { #ifdef _KERNEL /* * We will be destroying the ds; make sure its origin is unmounted if * necessary. */ char name[ZFS_MAX_DATASET_NAME_LEN]; dsl_dataset_name(drc->drc_ds, name); zfs_destroy_unmount_origin(name); #endif return (dsl_sync_task(drc->drc_tofs, dmu_recv_end_check, dmu_recv_end_sync, drc, dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL)); } static int dmu_recv_new_end(dmu_recv_cookie_t *drc) { return (dsl_sync_task(drc->drc_tofs, dmu_recv_end_check, dmu_recv_end_sync, drc, dmu_recv_end_modified_blocks, ZFS_SPACE_CHECK_NORMAL)); } int dmu_recv_end(dmu_recv_cookie_t *drc, void *owner) { int error; drc->drc_owner = owner; if (drc->drc_newfs) error = dmu_recv_new_end(drc); else error = dmu_recv_existing_end(drc); if (error != 0) { dmu_recv_cleanup_ds(drc); nvlist_free(drc->drc_keynvl); } else if (drc->drc_guid_to_ds_map != NULL) { (void) add_ds_to_guidmap(drc->drc_tofs, drc->drc_guid_to_ds_map, drc->drc_newsnapobj, drc->drc_raw); } return (error); } /* * Return TRUE if this objset is currently being received into. */ boolean_t dmu_objset_is_receiving(objset_t *os) { return (os->os_dsl_dataset != NULL && os->os_dsl_dataset->ds_owner == dmu_recv_tag); } #if defined(_KERNEL) module_param(zfs_send_corrupt_data, int, 0644); MODULE_PARM_DESC(zfs_send_corrupt_data, "Allow sending corrupt data"); module_param(zfs_send_queue_length, int, 0644); MODULE_PARM_DESC(zfs_send_queue_length, "Maximum send queue length"); module_param(zfs_recv_queue_length, int, 0644); MODULE_PARM_DESC(zfs_recv_queue_length, "Maximum receive queue length"); #endif