Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil.h =================================================================== --- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil.h (revision 308172) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil.h (revision 308173) @@ -1,434 +1,435 @@ /* * 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 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ /* Portions Copyright 2010 Robert Milkowski */ #ifndef _SYS_ZIL_H #define _SYS_ZIL_H #include #include #include #include #ifdef __cplusplus extern "C" { #endif struct dsl_pool; struct dsl_dataset; /* * Intent log format: * * Each objset has its own intent log. The log header (zil_header_t) * for objset N's intent log is kept in the Nth object of the SPA's * intent_log objset. The log header points to a chain of log blocks, * each of which contains log records (i.e., transactions) followed by * a log block trailer (zil_trailer_t). The format of a log record * depends on the record (or transaction) type, but all records begin * with a common structure that defines the type, length, and txg. */ /* * Intent log header - this on disk structure holds fields to manage * the log. All fields are 64 bit to easily handle cross architectures. */ typedef struct zil_header { uint64_t zh_claim_txg; /* txg in which log blocks were claimed */ uint64_t zh_replay_seq; /* highest replayed sequence number */ blkptr_t zh_log; /* log chain */ uint64_t zh_claim_blk_seq; /* highest claimed block sequence number */ uint64_t zh_flags; /* header flags */ uint64_t zh_claim_lr_seq; /* highest claimed lr sequence number */ uint64_t zh_pad[3]; } zil_header_t; /* * zh_flags bit settings */ #define ZIL_REPLAY_NEEDED 0x1 /* replay needed - internal only */ #define ZIL_CLAIM_LR_SEQ_VALID 0x2 /* zh_claim_lr_seq field is valid */ /* * Log block chaining. * * Log blocks are chained together. Originally they were chained at the * end of the block. For performance reasons the chain was moved to the * beginning of the block which allows writes for only the data being used. * The older position is supported for backwards compatability. * * The zio_eck_t contains a zec_cksum which for the intent log is * the sequence number of this log block. A seq of 0 is invalid. * The zec_cksum is checked by the SPA against the sequence * number passed in the blk_cksum field of the blkptr_t */ typedef struct zil_chain { uint64_t zc_pad; blkptr_t zc_next_blk; /* next block in chain */ uint64_t zc_nused; /* bytes in log block used */ zio_eck_t zc_eck; /* block trailer */ } zil_chain_t; #define ZIL_MIN_BLKSZ 4096ULL /* * The words of a log block checksum. */ #define ZIL_ZC_GUID_0 0 #define ZIL_ZC_GUID_1 1 #define ZIL_ZC_OBJSET 2 #define ZIL_ZC_SEQ 3 typedef enum zil_create { Z_FILE, Z_DIR, Z_XATTRDIR, } zil_create_t; /* * size of xvattr log section. * its composed of lr_attr_t + xvattr bitmap + 2 64 bit timestamps * for create time and a single 64 bit integer for all of the attributes, * and 4 64 bit integers (32 bytes) for the scanstamp. * */ #define ZIL_XVAT_SIZE(mapsize) \ sizeof (lr_attr_t) + (sizeof (uint32_t) * (mapsize - 1)) + \ (sizeof (uint64_t) * 7) /* * Size of ACL in log. The ACE data is padded out to properly align * on 8 byte boundary. */ #define ZIL_ACE_LENGTH(x) (roundup(x, sizeof (uint64_t))) /* * Intent log transaction types and record structures */ #define TX_CREATE 1 /* Create file */ #define TX_MKDIR 2 /* Make directory */ #define TX_MKXATTR 3 /* Make XATTR directory */ #define TX_SYMLINK 4 /* Create symbolic link to a file */ #define TX_REMOVE 5 /* Remove file */ #define TX_RMDIR 6 /* Remove directory */ #define TX_LINK 7 /* Create hard link to a file */ #define TX_RENAME 8 /* Rename a file */ #define TX_WRITE 9 /* File write */ #define TX_TRUNCATE 10 /* Truncate a file */ #define TX_SETATTR 11 /* Set file attributes */ #define TX_ACL_V0 12 /* Set old formatted ACL */ #define TX_ACL 13 /* Set ACL */ #define TX_CREATE_ACL 14 /* create with ACL */ #define TX_CREATE_ATTR 15 /* create + attrs */ #define TX_CREATE_ACL_ATTR 16 /* create with ACL + attrs */ #define TX_MKDIR_ACL 17 /* mkdir with ACL */ #define TX_MKDIR_ATTR 18 /* mkdir with attr */ #define TX_MKDIR_ACL_ATTR 19 /* mkdir with ACL + attrs */ #define TX_WRITE2 20 /* dmu_sync EALREADY write */ #define TX_MAX_TYPE 21 /* Max transaction type */ /* * The transactions for mkdir, symlink, remove, rmdir, link, and rename * may have the following bit set, indicating the original request * specified case-insensitive handling of names. */ #define TX_CI ((uint64_t)0x1 << 63) /* case-insensitive behavior requested */ /* * Transactions for write, truncate, setattr, acl_v0, and acl can be logged * out of order. For convenience in the code, all such records must have * lr_foid at the same offset. */ #define TX_OOO(txtype) \ ((txtype) == TX_WRITE || \ (txtype) == TX_TRUNCATE || \ (txtype) == TX_SETATTR || \ (txtype) == TX_ACL_V0 || \ (txtype) == TX_ACL || \ (txtype) == TX_WRITE2) /* * Format of log records. * The fields are carefully defined to allow them to be aligned * and sized the same on sparc & intel architectures. * Each log record has a common structure at the beginning. * * The log record on disk (lrc_seq) holds the sequence number of all log * records which is used to ensure we don't replay the same record. */ typedef struct { /* common log record header */ uint64_t lrc_txtype; /* intent log transaction type */ uint64_t lrc_reclen; /* transaction record length */ uint64_t lrc_txg; /* dmu transaction group number */ uint64_t lrc_seq; /* see comment above */ } lr_t; /* * Common start of all out-of-order record types (TX_OOO() above). */ typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_foid; /* object id */ } lr_ooo_t; /* * Handle option extended vattr attributes. * * Whenever new attributes are added the version number * will need to be updated as will code in * zfs_log.c and zfs_replay.c */ typedef struct { uint32_t lr_attr_masksize; /* number of elements in array */ uint32_t lr_attr_bitmap; /* First entry of array */ /* remainder of array and any additional fields */ } lr_attr_t; /* * log record for creates without optional ACL. * This log record does support optional xvattr_t attributes. */ typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_doid; /* object id of directory */ uint64_t lr_foid; /* object id of created file object */ uint64_t lr_mode; /* mode of object */ uint64_t lr_uid; /* uid of object */ uint64_t lr_gid; /* gid of object */ uint64_t lr_gen; /* generation (txg of creation) */ uint64_t lr_crtime[2]; /* creation time */ uint64_t lr_rdev; /* rdev of object to create */ /* name of object to create follows this */ /* for symlinks, link content follows name */ /* for creates with xvattr data, the name follows the xvattr info */ } lr_create_t; /* * FUID ACL record will be an array of ACEs from the original ACL. * If this array includes ephemeral IDs, the record will also include * an array of log-specific FUIDs to replace the ephemeral IDs. * Only one copy of each unique domain will be present, so the log-specific * FUIDs will use an index into a compressed domain table. On replay this * information will be used to construct real FUIDs (and bypass idmap, * since it may not be available). */ /* * Log record for creates with optional ACL * This log record is also used for recording any FUID * information needed for replaying the create. If the * file doesn't have any actual ACEs then the lr_aclcnt * would be zero. * * After lr_acl_flags, there are a lr_acl_bytes number of variable sized ace's. * If create is also setting xvattr's, then acl data follows xvattr. * If ACE FUIDs are needed then they will follow the xvattr_t. Following * the FUIDs will be the domain table information. The FUIDs for the owner * and group will be in lr_create. Name follows ACL data. */ typedef struct { lr_create_t lr_create; /* common create portion */ uint64_t lr_aclcnt; /* number of ACEs in ACL */ uint64_t lr_domcnt; /* number of unique domains */ uint64_t lr_fuidcnt; /* number of real fuids */ uint64_t lr_acl_bytes; /* number of bytes in ACL */ uint64_t lr_acl_flags; /* ACL flags */ } lr_acl_create_t; typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_doid; /* obj id of directory */ /* name of object to remove follows this */ } lr_remove_t; typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_doid; /* obj id of directory */ uint64_t lr_link_obj; /* obj id of link */ /* name of object to link follows this */ } lr_link_t; typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_sdoid; /* obj id of source directory */ uint64_t lr_tdoid; /* obj id of target directory */ /* 2 strings: names of source and destination follow this */ } lr_rename_t; typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_foid; /* file object to write */ uint64_t lr_offset; /* offset to write to */ uint64_t lr_length; /* user data length to write */ uint64_t lr_blkoff; /* no longer used */ blkptr_t lr_blkptr; /* spa block pointer for replay */ /* write data will follow for small writes */ } lr_write_t; typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_foid; /* object id of file to truncate */ uint64_t lr_offset; /* offset to truncate from */ uint64_t lr_length; /* length to truncate */ } lr_truncate_t; typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_foid; /* file object to change attributes */ uint64_t lr_mask; /* mask of attributes to set */ uint64_t lr_mode; /* mode to set */ uint64_t lr_uid; /* uid to set */ uint64_t lr_gid; /* gid to set */ uint64_t lr_size; /* size to set */ uint64_t lr_atime[2]; /* access time */ uint64_t lr_mtime[2]; /* modification time */ /* optional attribute lr_attr_t may be here */ } lr_setattr_t; typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_foid; /* obj id of file */ uint64_t lr_aclcnt; /* number of acl entries */ /* lr_aclcnt number of ace_t entries follow this */ } lr_acl_v0_t; typedef struct { lr_t lr_common; /* common portion of log record */ uint64_t lr_foid; /* obj id of file */ uint64_t lr_aclcnt; /* number of ACEs in ACL */ uint64_t lr_domcnt; /* number of unique domains */ uint64_t lr_fuidcnt; /* number of real fuids */ uint64_t lr_acl_bytes; /* number of bytes in ACL */ uint64_t lr_acl_flags; /* ACL flags */ /* lr_acl_bytes number of variable sized ace's follows */ } lr_acl_t; /* * ZIL structure definitions, interface function prototype and globals. */ /* * Writes are handled in three different ways: * * WR_INDIRECT: * In this mode, if we need to commit the write later, then the block * is immediately written into the file system (using dmu_sync), * and a pointer to the block is put into the log record. * When the txg commits the block is linked in. * This saves additionally writing the data into the log record. * There are a few requirements for this to occur: * - write is greater than zfs/zvol_immediate_write_sz * - not using slogs (as slogs are assumed to always be faster * than writing into the main pool) * - the write occupies only one block * WR_COPIED: * If we know we'll immediately be committing the * transaction (FSYNC or FDSYNC), the we allocate a larger * log record here for the data and copy the data in. * WR_NEED_COPY: * Otherwise we don't allocate a buffer, and *if* we need to * flush the write later then a buffer is allocated and * we retrieve the data using the dmu. */ typedef enum { WR_INDIRECT, /* indirect - a large write (dmu_sync() data */ /* and put blkptr in log, rather than actual data) */ WR_COPIED, /* immediate - data is copied into lr_write_t */ WR_NEED_COPY, /* immediate - data needs to be copied if pushed */ WR_NUM_STATES /* number of states */ } itx_wr_state_t; typedef struct itx { list_node_t itx_node; /* linkage on zl_itx_list */ void *itx_private; /* type-specific opaque data */ itx_wr_state_t itx_wr_state; /* write state */ uint8_t itx_sync; /* synchronous transaction */ uint64_t itx_sod; /* record size on disk */ uint64_t itx_oid; /* object id */ lr_t itx_lr; /* common part of log record */ /* followed by type-specific part of lr_xx_t and its immediate data */ } itx_t; typedef int zil_parse_blk_func_t(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t txg); typedef int zil_parse_lr_func_t(zilog_t *zilog, lr_t *lr, void *arg, uint64_t txg); typedef int zil_replay_func_t(); typedef int zil_get_data_t(void *arg, lr_write_t *lr, char *dbuf, zio_t *zio); extern int zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func, zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg); extern void zil_init(void); extern void zil_fini(void); extern zilog_t *zil_alloc(objset_t *os, zil_header_t *zh_phys); extern void zil_free(zilog_t *zilog); extern zilog_t *zil_open(objset_t *os, zil_get_data_t *get_data); extern void zil_close(zilog_t *zilog); extern void zil_replay(objset_t *os, void *arg, zil_replay_func_t *replay_func[TX_MAX_TYPE]); extern boolean_t zil_replaying(zilog_t *zilog, dmu_tx_t *tx); extern void zil_destroy(zilog_t *zilog, boolean_t keep_first); extern void zil_destroy_sync(zilog_t *zilog, dmu_tx_t *tx); extern void zil_rollback_destroy(zilog_t *zilog, dmu_tx_t *tx); extern itx_t *zil_itx_create(uint64_t txtype, size_t lrsize); extern void zil_itx_destroy(itx_t *itx); extern void zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx); +extern void zil_async_to_sync(zilog_t *zilog, uint64_t oid); extern void zil_commit(zilog_t *zilog, uint64_t oid); extern int zil_vdev_offline(const char *osname, void *txarg); extern int zil_claim(struct dsl_pool *dp, struct dsl_dataset *ds, void *txarg); extern int zil_check_log_chain(struct dsl_pool *dp, struct dsl_dataset *ds, void *tx); extern void zil_sync(zilog_t *zilog, dmu_tx_t *tx); extern void zil_clean(zilog_t *zilog, uint64_t synced_txg); extern int zil_suspend(const char *osname, void **cookiep); extern void zil_resume(void *cookie); extern void zil_add_block(zilog_t *zilog, const blkptr_t *bp); extern int zil_bp_tree_add(zilog_t *zilog, const blkptr_t *bp); extern void zil_set_sync(zilog_t *zilog, uint64_t syncval); extern void zil_set_logbias(zilog_t *zilog, uint64_t slogval); extern int zil_replay_disable; #ifdef __cplusplus } #endif #endif /* _SYS_ZIL_H */ Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c =================================================================== --- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c (revision 308172) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c (revision 308173) @@ -1,2151 +1,2149 @@ /* * 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, 2015 by Delphix. All rights reserved. * Copyright (c) 2011, 2014 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ /* Portions Copyright 2010 Robert Milkowski */ #include #include #include #include #include #include #include #include #include #include #include #include #include /* * The zfs intent log (ZIL) saves transaction records of system calls * that change the file system in memory with enough information * to be able to replay them. These are stored in memory until * either the DMU transaction group (txg) commits them to the stable pool * and they can be discarded, or they are flushed to the stable log * (also in the pool) due to a fsync, O_DSYNC or other synchronous * requirement. In the event of a panic or power fail then those log * records (transactions) are replayed. * * There is one ZIL per file system. Its on-disk (pool) format consists * of 3 parts: * * - ZIL header * - ZIL blocks * - ZIL records * * A log record holds a system call transaction. Log blocks can * hold many log records and the blocks are chained together. * Each ZIL block contains a block pointer (blkptr_t) to the next * ZIL block in the chain. The ZIL header points to the first * block in the chain. Note there is not a fixed place in the pool * to hold blocks. They are dynamically allocated and freed as * needed from the blocks available. Figure X shows the ZIL structure: */ /* * Disable intent logging replay. This global ZIL switch affects all pools. */ int zil_replay_disable = 0; SYSCTL_DECL(_vfs_zfs); SYSCTL_INT(_vfs_zfs, OID_AUTO, zil_replay_disable, CTLFLAG_RWTUN, &zil_replay_disable, 0, "Disable intent logging replay"); /* * Tunable parameter for debugging or performance analysis. Setting * zfs_nocacheflush will cause corruption on power loss if a volatile * out-of-order write cache is enabled. */ boolean_t zfs_nocacheflush = B_FALSE; SYSCTL_INT(_vfs_zfs, OID_AUTO, cache_flush_disable, CTLFLAG_RDTUN, &zfs_nocacheflush, 0, "Disable cache flush"); boolean_t zfs_trim_enabled = B_TRUE; SYSCTL_DECL(_vfs_zfs_trim); SYSCTL_INT(_vfs_zfs_trim, OID_AUTO, enabled, CTLFLAG_RDTUN, &zfs_trim_enabled, 0, "Enable ZFS TRIM"); static kmem_cache_t *zil_lwb_cache; -static void zil_async_to_sync(zilog_t *zilog, uint64_t foid); - #define LWB_EMPTY(lwb) ((BP_GET_LSIZE(&lwb->lwb_blk) - \ sizeof (zil_chain_t)) == (lwb->lwb_sz - lwb->lwb_nused)) /* * ziltest is by and large an ugly hack, but very useful in * checking replay without tedious work. * When running ziltest we want to keep all itx's and so maintain * a single list in the zl_itxg[] that uses a high txg: ZILTEST_TXG * We subtract TXG_CONCURRENT_STATES to allow for common code. */ #define ZILTEST_TXG (UINT64_MAX - TXG_CONCURRENT_STATES) static int zil_bp_compare(const void *x1, const void *x2) { const dva_t *dva1 = &((zil_bp_node_t *)x1)->zn_dva; const dva_t *dva2 = &((zil_bp_node_t *)x2)->zn_dva; if (DVA_GET_VDEV(dva1) < DVA_GET_VDEV(dva2)) return (-1); if (DVA_GET_VDEV(dva1) > DVA_GET_VDEV(dva2)) return (1); if (DVA_GET_OFFSET(dva1) < DVA_GET_OFFSET(dva2)) return (-1); if (DVA_GET_OFFSET(dva1) > DVA_GET_OFFSET(dva2)) return (1); return (0); } static void zil_bp_tree_init(zilog_t *zilog) { avl_create(&zilog->zl_bp_tree, zil_bp_compare, sizeof (zil_bp_node_t), offsetof(zil_bp_node_t, zn_node)); } static void zil_bp_tree_fini(zilog_t *zilog) { avl_tree_t *t = &zilog->zl_bp_tree; zil_bp_node_t *zn; void *cookie = NULL; while ((zn = avl_destroy_nodes(t, &cookie)) != NULL) kmem_free(zn, sizeof (zil_bp_node_t)); avl_destroy(t); } int zil_bp_tree_add(zilog_t *zilog, const blkptr_t *bp) { avl_tree_t *t = &zilog->zl_bp_tree; const dva_t *dva; zil_bp_node_t *zn; avl_index_t where; if (BP_IS_EMBEDDED(bp)) return (0); dva = BP_IDENTITY(bp); if (avl_find(t, dva, &where) != NULL) return (SET_ERROR(EEXIST)); zn = kmem_alloc(sizeof (zil_bp_node_t), KM_SLEEP); zn->zn_dva = *dva; avl_insert(t, zn, where); return (0); } static zil_header_t * zil_header_in_syncing_context(zilog_t *zilog) { return ((zil_header_t *)zilog->zl_header); } static void zil_init_log_chain(zilog_t *zilog, blkptr_t *bp) { zio_cksum_t *zc = &bp->blk_cksum; zc->zc_word[ZIL_ZC_GUID_0] = spa_get_random(-1ULL); zc->zc_word[ZIL_ZC_GUID_1] = spa_get_random(-1ULL); zc->zc_word[ZIL_ZC_OBJSET] = dmu_objset_id(zilog->zl_os); zc->zc_word[ZIL_ZC_SEQ] = 1ULL; } /* * Read a log block and make sure it's valid. */ static int zil_read_log_block(zilog_t *zilog, const blkptr_t *bp, blkptr_t *nbp, void *dst, char **end) { enum zio_flag zio_flags = ZIO_FLAG_CANFAIL; arc_flags_t aflags = ARC_FLAG_WAIT; arc_buf_t *abuf = NULL; zbookmark_phys_t zb; int error; if (zilog->zl_header->zh_claim_txg == 0) zio_flags |= ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SCRUB; if (!(zilog->zl_header->zh_flags & ZIL_CLAIM_LR_SEQ_VALID)) zio_flags |= ZIO_FLAG_SPECULATIVE; SET_BOOKMARK(&zb, bp->blk_cksum.zc_word[ZIL_ZC_OBJSET], ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, bp->blk_cksum.zc_word[ZIL_ZC_SEQ]); error = arc_read(NULL, zilog->zl_spa, bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb); if (error == 0) { zio_cksum_t cksum = bp->blk_cksum; /* * Validate the checksummed log block. * * Sequence numbers should be... sequential. The checksum * verifier for the next block should be bp's checksum plus 1. * * Also check the log chain linkage and size used. */ cksum.zc_word[ZIL_ZC_SEQ]++; if (BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_ZILOG2) { zil_chain_t *zilc = abuf->b_data; char *lr = (char *)(zilc + 1); uint64_t len = zilc->zc_nused - sizeof (zil_chain_t); if (bcmp(&cksum, &zilc->zc_next_blk.blk_cksum, sizeof (cksum)) || BP_IS_HOLE(&zilc->zc_next_blk)) { error = SET_ERROR(ECKSUM); } else { ASSERT3U(len, <=, SPA_OLD_MAXBLOCKSIZE); bcopy(lr, dst, len); *end = (char *)dst + len; *nbp = zilc->zc_next_blk; } } else { char *lr = abuf->b_data; uint64_t size = BP_GET_LSIZE(bp); zil_chain_t *zilc = (zil_chain_t *)(lr + size) - 1; if (bcmp(&cksum, &zilc->zc_next_blk.blk_cksum, sizeof (cksum)) || BP_IS_HOLE(&zilc->zc_next_blk) || (zilc->zc_nused > (size - sizeof (*zilc)))) { error = SET_ERROR(ECKSUM); } else { ASSERT3U(zilc->zc_nused, <=, SPA_OLD_MAXBLOCKSIZE); bcopy(lr, dst, zilc->zc_nused); *end = (char *)dst + zilc->zc_nused; *nbp = zilc->zc_next_blk; } } arc_buf_destroy(abuf, &abuf); } return (error); } /* * Read a TX_WRITE log data block. */ static int zil_read_log_data(zilog_t *zilog, const lr_write_t *lr, void *wbuf) { enum zio_flag zio_flags = ZIO_FLAG_CANFAIL; const blkptr_t *bp = &lr->lr_blkptr; arc_flags_t aflags = ARC_FLAG_WAIT; arc_buf_t *abuf = NULL; zbookmark_phys_t zb; int error; if (BP_IS_HOLE(bp)) { if (wbuf != NULL) bzero(wbuf, MAX(BP_GET_LSIZE(bp), lr->lr_length)); return (0); } if (zilog->zl_header->zh_claim_txg == 0) zio_flags |= ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SCRUB; SET_BOOKMARK(&zb, dmu_objset_id(zilog->zl_os), lr->lr_foid, ZB_ZIL_LEVEL, lr->lr_offset / BP_GET_LSIZE(bp)); error = arc_read(NULL, zilog->zl_spa, bp, arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zb); if (error == 0) { if (wbuf != NULL) bcopy(abuf->b_data, wbuf, arc_buf_size(abuf)); arc_buf_destroy(abuf, &abuf); } return (error); } /* * Parse the intent log, and call parse_func for each valid record within. */ int zil_parse(zilog_t *zilog, zil_parse_blk_func_t *parse_blk_func, zil_parse_lr_func_t *parse_lr_func, void *arg, uint64_t txg) { const zil_header_t *zh = zilog->zl_header; boolean_t claimed = !!zh->zh_claim_txg; uint64_t claim_blk_seq = claimed ? zh->zh_claim_blk_seq : UINT64_MAX; uint64_t claim_lr_seq = claimed ? zh->zh_claim_lr_seq : UINT64_MAX; uint64_t max_blk_seq = 0; uint64_t max_lr_seq = 0; uint64_t blk_count = 0; uint64_t lr_count = 0; blkptr_t blk, next_blk; char *lrbuf, *lrp; int error = 0; /* * Old logs didn't record the maximum zh_claim_lr_seq. */ if (!(zh->zh_flags & ZIL_CLAIM_LR_SEQ_VALID)) claim_lr_seq = UINT64_MAX; /* * Starting at the block pointed to by zh_log we read the log chain. * For each block in the chain we strongly check that block to * ensure its validity. We stop when an invalid block is found. * For each block pointer in the chain we call parse_blk_func(). * For each record in each valid block we call parse_lr_func(). * If the log has been claimed, stop if we encounter a sequence * number greater than the highest claimed sequence number. */ lrbuf = zio_buf_alloc(SPA_OLD_MAXBLOCKSIZE); zil_bp_tree_init(zilog); for (blk = zh->zh_log; !BP_IS_HOLE(&blk); blk = next_blk) { uint64_t blk_seq = blk.blk_cksum.zc_word[ZIL_ZC_SEQ]; int reclen; char *end; if (blk_seq > claim_blk_seq) break; if ((error = parse_blk_func(zilog, &blk, arg, txg)) != 0) break; ASSERT3U(max_blk_seq, <, blk_seq); max_blk_seq = blk_seq; blk_count++; if (max_lr_seq == claim_lr_seq && max_blk_seq == claim_blk_seq) break; error = zil_read_log_block(zilog, &blk, &next_blk, lrbuf, &end); if (error != 0) break; for (lrp = lrbuf; lrp < end; lrp += reclen) { lr_t *lr = (lr_t *)lrp; reclen = lr->lrc_reclen; ASSERT3U(reclen, >=, sizeof (lr_t)); if (lr->lrc_seq > claim_lr_seq) goto done; if ((error = parse_lr_func(zilog, lr, arg, txg)) != 0) goto done; ASSERT3U(max_lr_seq, <, lr->lrc_seq); max_lr_seq = lr->lrc_seq; lr_count++; } } done: zilog->zl_parse_error = error; zilog->zl_parse_blk_seq = max_blk_seq; zilog->zl_parse_lr_seq = max_lr_seq; zilog->zl_parse_blk_count = blk_count; zilog->zl_parse_lr_count = lr_count; ASSERT(!claimed || !(zh->zh_flags & ZIL_CLAIM_LR_SEQ_VALID) || (max_blk_seq == claim_blk_seq && max_lr_seq == claim_lr_seq)); zil_bp_tree_fini(zilog); zio_buf_free(lrbuf, SPA_OLD_MAXBLOCKSIZE); return (error); } static int zil_claim_log_block(zilog_t *zilog, blkptr_t *bp, void *tx, uint64_t first_txg) { /* * Claim log block if not already committed and not already claimed. * If tx == NULL, just verify that the block is claimable. */ if (BP_IS_HOLE(bp) || bp->blk_birth < first_txg || zil_bp_tree_add(zilog, bp) != 0) return (0); return (zio_wait(zio_claim(NULL, zilog->zl_spa, tx == NULL ? 0 : first_txg, bp, spa_claim_notify, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_SCRUB))); } static int zil_claim_log_record(zilog_t *zilog, lr_t *lrc, void *tx, uint64_t first_txg) { lr_write_t *lr = (lr_write_t *)lrc; int error; if (lrc->lrc_txtype != TX_WRITE) return (0); /* * If the block is not readable, don't claim it. This can happen * in normal operation when a log block is written to disk before * some of the dmu_sync() blocks it points to. In this case, the * transaction cannot have been committed to anyone (we would have * waited for all writes to be stable first), so it is semantically * correct to declare this the end of the log. */ if (lr->lr_blkptr.blk_birth >= first_txg && (error = zil_read_log_data(zilog, lr, NULL)) != 0) return (error); return (zil_claim_log_block(zilog, &lr->lr_blkptr, tx, first_txg)); } /* ARGSUSED */ static int zil_free_log_block(zilog_t *zilog, blkptr_t *bp, void *tx, uint64_t claim_txg) { zio_free_zil(zilog->zl_spa, dmu_tx_get_txg(tx), bp); return (0); } static int zil_free_log_record(zilog_t *zilog, lr_t *lrc, void *tx, uint64_t claim_txg) { lr_write_t *lr = (lr_write_t *)lrc; blkptr_t *bp = &lr->lr_blkptr; /* * If we previously claimed it, we need to free it. */ if (claim_txg != 0 && lrc->lrc_txtype == TX_WRITE && bp->blk_birth >= claim_txg && zil_bp_tree_add(zilog, bp) == 0 && !BP_IS_HOLE(bp)) zio_free(zilog->zl_spa, dmu_tx_get_txg(tx), bp); return (0); } static lwb_t * zil_alloc_lwb(zilog_t *zilog, blkptr_t *bp, uint64_t txg) { lwb_t *lwb; lwb = kmem_cache_alloc(zil_lwb_cache, KM_SLEEP); lwb->lwb_zilog = zilog; lwb->lwb_blk = *bp; lwb->lwb_buf = zio_buf_alloc(BP_GET_LSIZE(bp)); lwb->lwb_max_txg = txg; lwb->lwb_zio = NULL; lwb->lwb_tx = NULL; if (BP_GET_CHECKSUM(bp) == ZIO_CHECKSUM_ZILOG2) { lwb->lwb_nused = sizeof (zil_chain_t); lwb->lwb_sz = BP_GET_LSIZE(bp); } else { lwb->lwb_nused = 0; lwb->lwb_sz = BP_GET_LSIZE(bp) - sizeof (zil_chain_t); } mutex_enter(&zilog->zl_lock); list_insert_tail(&zilog->zl_lwb_list, lwb); mutex_exit(&zilog->zl_lock); return (lwb); } /* * Called when we create in-memory log transactions so that we know * to cleanup the itxs at the end of spa_sync(). */ void zilog_dirty(zilog_t *zilog, uint64_t txg) { dsl_pool_t *dp = zilog->zl_dmu_pool; dsl_dataset_t *ds = dmu_objset_ds(zilog->zl_os); if (ds->ds_is_snapshot) panic("dirtying snapshot!"); if (txg_list_add(&dp->dp_dirty_zilogs, zilog, txg)) { /* up the hold count until we can be written out */ dmu_buf_add_ref(ds->ds_dbuf, zilog); } } boolean_t zilog_is_dirty(zilog_t *zilog) { dsl_pool_t *dp = zilog->zl_dmu_pool; for (int t = 0; t < TXG_SIZE; t++) { if (txg_list_member(&dp->dp_dirty_zilogs, zilog, t)) return (B_TRUE); } return (B_FALSE); } /* * Create an on-disk intent log. */ static lwb_t * zil_create(zilog_t *zilog) { const zil_header_t *zh = zilog->zl_header; lwb_t *lwb = NULL; uint64_t txg = 0; dmu_tx_t *tx = NULL; blkptr_t blk; int error = 0; /* * Wait for any previous destroy to complete. */ txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg); ASSERT(zh->zh_claim_txg == 0); ASSERT(zh->zh_replay_seq == 0); blk = zh->zh_log; /* * Allocate an initial log block if: * - there isn't one already * - the existing block is the wrong endianess */ if (BP_IS_HOLE(&blk) || BP_SHOULD_BYTESWAP(&blk)) { tx = dmu_tx_create(zilog->zl_os); VERIFY(dmu_tx_assign(tx, TXG_WAIT) == 0); dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); txg = dmu_tx_get_txg(tx); if (!BP_IS_HOLE(&blk)) { zio_free_zil(zilog->zl_spa, txg, &blk); BP_ZERO(&blk); } error = zio_alloc_zil(zilog->zl_spa, txg, &blk, NULL, ZIL_MIN_BLKSZ, zilog->zl_logbias == ZFS_LOGBIAS_LATENCY); if (error == 0) zil_init_log_chain(zilog, &blk); } /* * Allocate a log write buffer (lwb) for the first log block. */ if (error == 0) lwb = zil_alloc_lwb(zilog, &blk, txg); /* * If we just allocated the first log block, commit our transaction * and wait for zil_sync() to stuff the block poiner into zh_log. * (zh is part of the MOS, so we cannot modify it in open context.) */ if (tx != NULL) { dmu_tx_commit(tx); txg_wait_synced(zilog->zl_dmu_pool, txg); } ASSERT(bcmp(&blk, &zh->zh_log, sizeof (blk)) == 0); return (lwb); } /* * In one tx, free all log blocks and clear the log header. * If keep_first is set, then we're replaying a log with no content. * We want to keep the first block, however, so that the first * synchronous transaction doesn't require a txg_wait_synced() * in zil_create(). We don't need to txg_wait_synced() here either * when keep_first is set, because both zil_create() and zil_destroy() * will wait for any in-progress destroys to complete. */ void zil_destroy(zilog_t *zilog, boolean_t keep_first) { const zil_header_t *zh = zilog->zl_header; lwb_t *lwb; dmu_tx_t *tx; uint64_t txg; /* * Wait for any previous destroy to complete. */ txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg); zilog->zl_old_header = *zh; /* debugging aid */ if (BP_IS_HOLE(&zh->zh_log)) return; tx = dmu_tx_create(zilog->zl_os); VERIFY(dmu_tx_assign(tx, TXG_WAIT) == 0); dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); txg = dmu_tx_get_txg(tx); mutex_enter(&zilog->zl_lock); ASSERT3U(zilog->zl_destroy_txg, <, txg); zilog->zl_destroy_txg = txg; zilog->zl_keep_first = keep_first; if (!list_is_empty(&zilog->zl_lwb_list)) { ASSERT(zh->zh_claim_txg == 0); VERIFY(!keep_first); while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) { list_remove(&zilog->zl_lwb_list, lwb); if (lwb->lwb_buf != NULL) zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); zio_free_zil(zilog->zl_spa, txg, &lwb->lwb_blk); kmem_cache_free(zil_lwb_cache, lwb); } } else if (!keep_first) { zil_destroy_sync(zilog, tx); } mutex_exit(&zilog->zl_lock); dmu_tx_commit(tx); } void zil_destroy_sync(zilog_t *zilog, dmu_tx_t *tx) { ASSERT(list_is_empty(&zilog->zl_lwb_list)); (void) zil_parse(zilog, zil_free_log_block, zil_free_log_record, tx, zilog->zl_header->zh_claim_txg); } int zil_claim(dsl_pool_t *dp, dsl_dataset_t *ds, void *txarg) { dmu_tx_t *tx = txarg; uint64_t first_txg = dmu_tx_get_txg(tx); zilog_t *zilog; zil_header_t *zh; objset_t *os; int error; error = dmu_objset_own_obj(dp, ds->ds_object, DMU_OST_ANY, B_FALSE, FTAG, &os); if (error != 0) { /* * EBUSY indicates that the objset is inconsistent, in which * case it can not have a ZIL. */ if (error != EBUSY) { cmn_err(CE_WARN, "can't open objset for %llu, error %u", (unsigned long long)ds->ds_object, error); } return (0); } zilog = dmu_objset_zil(os); zh = zil_header_in_syncing_context(zilog); if (spa_get_log_state(zilog->zl_spa) == SPA_LOG_CLEAR) { if (!BP_IS_HOLE(&zh->zh_log)) zio_free_zil(zilog->zl_spa, first_txg, &zh->zh_log); BP_ZERO(&zh->zh_log); dsl_dataset_dirty(dmu_objset_ds(os), tx); dmu_objset_disown(os, FTAG); return (0); } /* * Claim all log blocks if we haven't already done so, and remember * the highest claimed sequence number. This ensures that if we can * read only part of the log now (e.g. due to a missing device), * but we can read the entire log later, we will not try to replay * or destroy beyond the last block we successfully claimed. */ ASSERT3U(zh->zh_claim_txg, <=, first_txg); if (zh->zh_claim_txg == 0 && !BP_IS_HOLE(&zh->zh_log)) { (void) zil_parse(zilog, zil_claim_log_block, zil_claim_log_record, tx, first_txg); zh->zh_claim_txg = first_txg; zh->zh_claim_blk_seq = zilog->zl_parse_blk_seq; zh->zh_claim_lr_seq = zilog->zl_parse_lr_seq; if (zilog->zl_parse_lr_count || zilog->zl_parse_blk_count > 1) zh->zh_flags |= ZIL_REPLAY_NEEDED; zh->zh_flags |= ZIL_CLAIM_LR_SEQ_VALID; dsl_dataset_dirty(dmu_objset_ds(os), tx); } ASSERT3U(first_txg, ==, (spa_last_synced_txg(zilog->zl_spa) + 1)); dmu_objset_disown(os, FTAG); return (0); } /* * Check the log by walking the log chain. * Checksum errors are ok as they indicate the end of the chain. * Any other error (no device or read failure) returns an error. */ /* ARGSUSED */ int zil_check_log_chain(dsl_pool_t *dp, dsl_dataset_t *ds, void *tx) { zilog_t *zilog; objset_t *os; blkptr_t *bp; int error; ASSERT(tx == NULL); error = dmu_objset_from_ds(ds, &os); if (error != 0) { cmn_err(CE_WARN, "can't open objset %llu, error %d", (unsigned long long)ds->ds_object, error); return (0); } zilog = dmu_objset_zil(os); bp = (blkptr_t *)&zilog->zl_header->zh_log; /* * Check the first block and determine if it's on a log device * which may have been removed or faulted prior to loading this * pool. If so, there's no point in checking the rest of the log * as its content should have already been synced to the pool. */ if (!BP_IS_HOLE(bp)) { vdev_t *vd; boolean_t valid = B_TRUE; spa_config_enter(os->os_spa, SCL_STATE, FTAG, RW_READER); vd = vdev_lookup_top(os->os_spa, DVA_GET_VDEV(&bp->blk_dva[0])); if (vd->vdev_islog && vdev_is_dead(vd)) valid = vdev_log_state_valid(vd); spa_config_exit(os->os_spa, SCL_STATE, FTAG); if (!valid) return (0); } /* * Because tx == NULL, zil_claim_log_block() will not actually claim * any blocks, but just determine whether it is possible to do so. * In addition to checking the log chain, zil_claim_log_block() * will invoke zio_claim() with a done func of spa_claim_notify(), * which will update spa_max_claim_txg. See spa_load() for details. */ error = zil_parse(zilog, zil_claim_log_block, zil_claim_log_record, tx, zilog->zl_header->zh_claim_txg ? -1ULL : spa_first_txg(os->os_spa)); return ((error == ECKSUM || error == ENOENT) ? 0 : error); } static int zil_vdev_compare(const void *x1, const void *x2) { const uint64_t v1 = ((zil_vdev_node_t *)x1)->zv_vdev; const uint64_t v2 = ((zil_vdev_node_t *)x2)->zv_vdev; if (v1 < v2) return (-1); if (v1 > v2) return (1); return (0); } void zil_add_block(zilog_t *zilog, const blkptr_t *bp) { avl_tree_t *t = &zilog->zl_vdev_tree; avl_index_t where; zil_vdev_node_t *zv, zvsearch; int ndvas = BP_GET_NDVAS(bp); int i; if (zfs_nocacheflush) return; ASSERT(zilog->zl_writer); /* * Even though we're zl_writer, we still need a lock because the * zl_get_data() callbacks may have dmu_sync() done callbacks * that will run concurrently. */ mutex_enter(&zilog->zl_vdev_lock); for (i = 0; i < ndvas; i++) { zvsearch.zv_vdev = DVA_GET_VDEV(&bp->blk_dva[i]); if (avl_find(t, &zvsearch, &where) == NULL) { zv = kmem_alloc(sizeof (*zv), KM_SLEEP); zv->zv_vdev = zvsearch.zv_vdev; avl_insert(t, zv, where); } } mutex_exit(&zilog->zl_vdev_lock); } static void zil_flush_vdevs(zilog_t *zilog) { spa_t *spa = zilog->zl_spa; avl_tree_t *t = &zilog->zl_vdev_tree; void *cookie = NULL; zil_vdev_node_t *zv; zio_t *zio; ASSERT(zilog->zl_writer); /* * We don't need zl_vdev_lock here because we're the zl_writer, * and all zl_get_data() callbacks are done. */ if (avl_numnodes(t) == 0) return; spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); zio = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL); while ((zv = avl_destroy_nodes(t, &cookie)) != NULL) { vdev_t *vd = vdev_lookup_top(spa, zv->zv_vdev); if (vd != NULL) zio_flush(zio, vd); kmem_free(zv, sizeof (*zv)); } /* * Wait for all the flushes to complete. Not all devices actually * support the DKIOCFLUSHWRITECACHE ioctl, so it's OK if it fails. */ (void) zio_wait(zio); spa_config_exit(spa, SCL_STATE, FTAG); } /* * Function called when a log block write completes */ static void zil_lwb_write_done(zio_t *zio) { lwb_t *lwb = zio->io_private; zilog_t *zilog = lwb->lwb_zilog; dmu_tx_t *tx = lwb->lwb_tx; ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF); ASSERT(BP_GET_TYPE(zio->io_bp) == DMU_OT_INTENT_LOG); ASSERT(BP_GET_LEVEL(zio->io_bp) == 0); ASSERT(BP_GET_BYTEORDER(zio->io_bp) == ZFS_HOST_BYTEORDER); ASSERT(!BP_IS_GANG(zio->io_bp)); ASSERT(!BP_IS_HOLE(zio->io_bp)); ASSERT(BP_GET_FILL(zio->io_bp) == 0); /* * Ensure the lwb buffer pointer is cleared before releasing * the txg. If we have had an allocation failure and * the txg is waiting to sync then we want want zil_sync() * to remove the lwb so that it's not picked up as the next new * one in zil_commit_writer(). zil_sync() will only remove * the lwb if lwb_buf is null. */ zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); mutex_enter(&zilog->zl_lock); lwb->lwb_buf = NULL; lwb->lwb_tx = NULL; mutex_exit(&zilog->zl_lock); /* * Now that we've written this log block, we have a stable pointer * to the next block in the chain, so it's OK to let the txg in * which we allocated the next block sync. */ dmu_tx_commit(tx); } /* * Initialize the io for a log block. */ static void zil_lwb_write_init(zilog_t *zilog, lwb_t *lwb) { zbookmark_phys_t zb; SET_BOOKMARK(&zb, lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_OBJSET], ZB_ZIL_OBJECT, ZB_ZIL_LEVEL, lwb->lwb_blk.blk_cksum.zc_word[ZIL_ZC_SEQ]); if (zilog->zl_root_zio == NULL) { zilog->zl_root_zio = zio_root(zilog->zl_spa, NULL, NULL, ZIO_FLAG_CANFAIL); } if (lwb->lwb_zio == NULL) { lwb->lwb_zio = zio_rewrite(zilog->zl_root_zio, zilog->zl_spa, 0, &lwb->lwb_blk, lwb->lwb_buf, BP_GET_LSIZE(&lwb->lwb_blk), zil_lwb_write_done, lwb, ZIO_PRIORITY_SYNC_WRITE, ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE, &zb); } } /* * Define a limited set of intent log block sizes. * * These must be a multiple of 4KB. Note only the amount used (again * aligned to 4KB) actually gets written. However, we can't always just * allocate SPA_OLD_MAXBLOCKSIZE as the slog space could be exhausted. */ uint64_t zil_block_buckets[] = { 4096, /* non TX_WRITE */ 8192+4096, /* data base */ 32*1024 + 4096, /* NFS writes */ UINT64_MAX }; /* * Use the slog as long as the logbias is 'latency' and the current commit size * is less than the limit or the total list size is less than 2X the limit. * Limit checking is disabled by setting zil_slog_limit to UINT64_MAX. */ uint64_t zil_slog_limit = 1024 * 1024; SYSCTL_QUAD(_vfs_zfs, OID_AUTO, zil_slog_limit, CTLFLAG_RWTUN, &zil_slog_limit, 0, "Maximal commit size to use SLOG"); #define USE_SLOG(zilog) (((zilog)->zl_logbias == ZFS_LOGBIAS_LATENCY) && \ (((zilog)->zl_cur_used < zil_slog_limit) || \ ((zilog)->zl_itx_list_sz < (zil_slog_limit << 1)))) /* * Start a log block write and advance to the next log block. * Calls are serialized. */ static lwb_t * zil_lwb_write_start(zilog_t *zilog, lwb_t *lwb) { lwb_t *nlwb = NULL; zil_chain_t *zilc; spa_t *spa = zilog->zl_spa; blkptr_t *bp; dmu_tx_t *tx; uint64_t txg; uint64_t zil_blksz, wsz; int i, error; if (BP_GET_CHECKSUM(&lwb->lwb_blk) == ZIO_CHECKSUM_ZILOG2) { zilc = (zil_chain_t *)lwb->lwb_buf; bp = &zilc->zc_next_blk; } else { zilc = (zil_chain_t *)(lwb->lwb_buf + lwb->lwb_sz); bp = &zilc->zc_next_blk; } ASSERT(lwb->lwb_nused <= lwb->lwb_sz); /* * Allocate the next block and save its address in this block * before writing it in order to establish the log chain. * Note that if the allocation of nlwb synced before we wrote * the block that points at it (lwb), we'd leak it if we crashed. * Therefore, we don't do dmu_tx_commit() until zil_lwb_write_done(). * We dirty the dataset to ensure that zil_sync() will be called * to clean up in the event of allocation failure or I/O failure. */ tx = dmu_tx_create(zilog->zl_os); VERIFY(dmu_tx_assign(tx, TXG_WAIT) == 0); dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); txg = dmu_tx_get_txg(tx); lwb->lwb_tx = tx; /* * Log blocks are pre-allocated. Here we select the size of the next * block, based on size used in the last block. * - first find the smallest bucket that will fit the block from a * limited set of block sizes. This is because it's faster to write * blocks allocated from the same metaslab as they are adjacent or * close. * - next find the maximum from the new suggested size and an array of * previous sizes. This lessens a picket fence effect of wrongly * guesssing the size if we have a stream of say 2k, 64k, 2k, 64k * requests. * * Note we only write what is used, but we can't just allocate * the maximum block size because we can exhaust the available * pool log space. */ zil_blksz = zilog->zl_cur_used + sizeof (zil_chain_t); for (i = 0; zil_blksz > zil_block_buckets[i]; i++) continue; zil_blksz = zil_block_buckets[i]; if (zil_blksz == UINT64_MAX) zil_blksz = SPA_OLD_MAXBLOCKSIZE; zilog->zl_prev_blks[zilog->zl_prev_rotor] = zil_blksz; for (i = 0; i < ZIL_PREV_BLKS; i++) zil_blksz = MAX(zil_blksz, zilog->zl_prev_blks[i]); zilog->zl_prev_rotor = (zilog->zl_prev_rotor + 1) & (ZIL_PREV_BLKS - 1); BP_ZERO(bp); /* pass the old blkptr in order to spread log blocks across devs */ error = zio_alloc_zil(spa, txg, bp, &lwb->lwb_blk, zil_blksz, USE_SLOG(zilog)); if (error == 0) { ASSERT3U(bp->blk_birth, ==, txg); bp->blk_cksum = lwb->lwb_blk.blk_cksum; bp->blk_cksum.zc_word[ZIL_ZC_SEQ]++; /* * Allocate a new log write buffer (lwb). */ nlwb = zil_alloc_lwb(zilog, bp, txg); /* Record the block for later vdev flushing */ zil_add_block(zilog, &lwb->lwb_blk); } if (BP_GET_CHECKSUM(&lwb->lwb_blk) == ZIO_CHECKSUM_ZILOG2) { /* For Slim ZIL only write what is used. */ wsz = P2ROUNDUP_TYPED(lwb->lwb_nused, ZIL_MIN_BLKSZ, uint64_t); ASSERT3U(wsz, <=, lwb->lwb_sz); zio_shrink(lwb->lwb_zio, wsz); } else { wsz = lwb->lwb_sz; } zilc->zc_pad = 0; zilc->zc_nused = lwb->lwb_nused; zilc->zc_eck.zec_cksum = lwb->lwb_blk.blk_cksum; /* * clear unused data for security */ bzero(lwb->lwb_buf + lwb->lwb_nused, wsz - lwb->lwb_nused); zio_nowait(lwb->lwb_zio); /* Kick off the write for the old log block */ /* * If there was an allocation failure then nlwb will be null which * forces a txg_wait_synced(). */ return (nlwb); } static lwb_t * zil_lwb_commit(zilog_t *zilog, itx_t *itx, lwb_t *lwb) { lr_t *lrc = &itx->itx_lr; /* common log record */ lr_write_t *lrw = (lr_write_t *)lrc; char *lr_buf; uint64_t txg = lrc->lrc_txg; uint64_t reclen = lrc->lrc_reclen; uint64_t dlen = 0; if (lwb == NULL) return (NULL); ASSERT(lwb->lwb_buf != NULL); ASSERT(zilog_is_dirty(zilog) || spa_freeze_txg(zilog->zl_spa) != UINT64_MAX); if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY) dlen = P2ROUNDUP_TYPED( lrw->lr_length, sizeof (uint64_t), uint64_t); zilog->zl_cur_used += (reclen + dlen); zil_lwb_write_init(zilog, lwb); /* * If this record won't fit in the current log block, start a new one. */ if (lwb->lwb_nused + reclen + dlen > lwb->lwb_sz) { lwb = zil_lwb_write_start(zilog, lwb); if (lwb == NULL) return (NULL); zil_lwb_write_init(zilog, lwb); ASSERT(LWB_EMPTY(lwb)); if (lwb->lwb_nused + reclen + dlen > lwb->lwb_sz) { txg_wait_synced(zilog->zl_dmu_pool, txg); return (lwb); } } lr_buf = lwb->lwb_buf + lwb->lwb_nused; bcopy(lrc, lr_buf, reclen); lrc = (lr_t *)lr_buf; lrw = (lr_write_t *)lrc; /* * If it's a write, fetch the data or get its blkptr as appropriate. */ if (lrc->lrc_txtype == TX_WRITE) { if (txg > spa_freeze_txg(zilog->zl_spa)) txg_wait_synced(zilog->zl_dmu_pool, txg); if (itx->itx_wr_state != WR_COPIED) { char *dbuf; int error; if (dlen) { ASSERT(itx->itx_wr_state == WR_NEED_COPY); dbuf = lr_buf + reclen; lrw->lr_common.lrc_reclen += dlen; } else { ASSERT(itx->itx_wr_state == WR_INDIRECT); dbuf = NULL; } error = zilog->zl_get_data( itx->itx_private, lrw, dbuf, lwb->lwb_zio); if (error == EIO) { txg_wait_synced(zilog->zl_dmu_pool, txg); return (lwb); } if (error != 0) { ASSERT(error == ENOENT || error == EEXIST || error == EALREADY); return (lwb); } } } /* * We're actually making an entry, so update lrc_seq to be the * log record sequence number. Note that this is generally not * equal to the itx sequence number because not all transactions * are synchronous, and sometimes spa_sync() gets there first. */ lrc->lrc_seq = ++zilog->zl_lr_seq; /* we are single threaded */ lwb->lwb_nused += reclen + dlen; lwb->lwb_max_txg = MAX(lwb->lwb_max_txg, txg); ASSERT3U(lwb->lwb_nused, <=, lwb->lwb_sz); ASSERT0(P2PHASE(lwb->lwb_nused, sizeof (uint64_t))); return (lwb); } itx_t * zil_itx_create(uint64_t txtype, size_t lrsize) { itx_t *itx; lrsize = P2ROUNDUP_TYPED(lrsize, sizeof (uint64_t), size_t); #ifdef __FreeBSD__ if (offsetof(itx_t, itx_lr) + lrsize > PAGE_SIZE) itx = zio_buf_alloc(offsetof(itx_t, itx_lr) + lrsize); else #endif itx = kmem_alloc(offsetof(itx_t, itx_lr) + lrsize, KM_SLEEP); itx->itx_lr.lrc_txtype = txtype; itx->itx_lr.lrc_reclen = lrsize; itx->itx_sod = lrsize; /* if write & WR_NEED_COPY will be increased */ itx->itx_lr.lrc_seq = 0; /* defensive */ itx->itx_sync = B_TRUE; /* default is synchronous */ return (itx); } void zil_itx_destroy(itx_t *itx) { #ifdef __FreeBSD__ if (offsetof(itx_t, itx_lr) + itx->itx_lr.lrc_reclen > PAGE_SIZE) zio_buf_free(itx, offsetof(itx_t, itx_lr) + itx->itx_lr.lrc_reclen); else #endif kmem_free(itx, offsetof(itx_t, itx_lr) + itx->itx_lr.lrc_reclen); } /* * Free up the sync and async itxs. The itxs_t has already been detached * so no locks are needed. */ static void zil_itxg_clean(itxs_t *itxs) { itx_t *itx; list_t *list; avl_tree_t *t; void *cookie; itx_async_node_t *ian; list = &itxs->i_sync_list; while ((itx = list_head(list)) != NULL) { list_remove(list, itx); zil_itx_destroy(itx); } cookie = NULL; t = &itxs->i_async_tree; while ((ian = avl_destroy_nodes(t, &cookie)) != NULL) { list = &ian->ia_list; while ((itx = list_head(list)) != NULL) { list_remove(list, itx); zil_itx_destroy(itx); } list_destroy(list); kmem_free(ian, sizeof (itx_async_node_t)); } avl_destroy(t); kmem_free(itxs, sizeof (itxs_t)); } static int zil_aitx_compare(const void *x1, const void *x2) { const uint64_t o1 = ((itx_async_node_t *)x1)->ia_foid; const uint64_t o2 = ((itx_async_node_t *)x2)->ia_foid; if (o1 < o2) return (-1); if (o1 > o2) return (1); return (0); } /* * Remove all async itx with the given oid. */ static void zil_remove_async(zilog_t *zilog, uint64_t oid) { uint64_t otxg, txg; itx_async_node_t *ian; avl_tree_t *t; avl_index_t where; list_t clean_list; itx_t *itx; ASSERT(oid != 0); list_create(&clean_list, sizeof (itx_t), offsetof(itx_t, itx_node)); if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */ otxg = ZILTEST_TXG; else otxg = spa_last_synced_txg(zilog->zl_spa) + 1; for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) { itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK]; mutex_enter(&itxg->itxg_lock); if (itxg->itxg_txg != txg) { mutex_exit(&itxg->itxg_lock); continue; } /* * Locate the object node and append its list. */ t = &itxg->itxg_itxs->i_async_tree; ian = avl_find(t, &oid, &where); if (ian != NULL) list_move_tail(&clean_list, &ian->ia_list); mutex_exit(&itxg->itxg_lock); } while ((itx = list_head(&clean_list)) != NULL) { list_remove(&clean_list, itx); zil_itx_destroy(itx); } list_destroy(&clean_list); } void zil_itx_assign(zilog_t *zilog, itx_t *itx, dmu_tx_t *tx) { uint64_t txg; itxg_t *itxg; itxs_t *itxs, *clean = NULL; /* * Object ids can be re-instantiated in the next txg so * remove any async transactions to avoid future leaks. * This can happen if a fsync occurs on the re-instantiated * object for a WR_INDIRECT or WR_NEED_COPY write, which gets * the new file data and flushes a write record for the old object. */ if ((itx->itx_lr.lrc_txtype & ~TX_CI) == TX_REMOVE) zil_remove_async(zilog, itx->itx_oid); /* * Ensure the data of a renamed file is committed before the rename. */ if ((itx->itx_lr.lrc_txtype & ~TX_CI) == TX_RENAME) zil_async_to_sync(zilog, itx->itx_oid); if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) txg = ZILTEST_TXG; else txg = dmu_tx_get_txg(tx); itxg = &zilog->zl_itxg[txg & TXG_MASK]; mutex_enter(&itxg->itxg_lock); itxs = itxg->itxg_itxs; if (itxg->itxg_txg != txg) { if (itxs != NULL) { /* * The zil_clean callback hasn't got around to cleaning * this itxg. Save the itxs for release below. * This should be rare. */ atomic_add_64(&zilog->zl_itx_list_sz, -itxg->itxg_sod); itxg->itxg_sod = 0; clean = itxg->itxg_itxs; } ASSERT(itxg->itxg_sod == 0); itxg->itxg_txg = txg; itxs = itxg->itxg_itxs = kmem_zalloc(sizeof (itxs_t), KM_SLEEP); list_create(&itxs->i_sync_list, sizeof (itx_t), offsetof(itx_t, itx_node)); avl_create(&itxs->i_async_tree, zil_aitx_compare, sizeof (itx_async_node_t), offsetof(itx_async_node_t, ia_node)); } if (itx->itx_sync) { list_insert_tail(&itxs->i_sync_list, itx); atomic_add_64(&zilog->zl_itx_list_sz, itx->itx_sod); itxg->itxg_sod += itx->itx_sod; } else { avl_tree_t *t = &itxs->i_async_tree; uint64_t foid = ((lr_ooo_t *)&itx->itx_lr)->lr_foid; itx_async_node_t *ian; avl_index_t where; ian = avl_find(t, &foid, &where); if (ian == NULL) { ian = kmem_alloc(sizeof (itx_async_node_t), KM_SLEEP); list_create(&ian->ia_list, sizeof (itx_t), offsetof(itx_t, itx_node)); ian->ia_foid = foid; avl_insert(t, ian, where); } list_insert_tail(&ian->ia_list, itx); } itx->itx_lr.lrc_txg = dmu_tx_get_txg(tx); zilog_dirty(zilog, txg); mutex_exit(&itxg->itxg_lock); /* Release the old itxs now we've dropped the lock */ if (clean != NULL) zil_itxg_clean(clean); } /* * If there are any in-memory intent log transactions which have now been * synced then start up a taskq to free them. We should only do this after we * have written out the uberblocks (i.e. txg has been comitted) so that * don't inadvertently clean out in-memory log records that would be required * by zil_commit(). */ void zil_clean(zilog_t *zilog, uint64_t synced_txg) { itxg_t *itxg = &zilog->zl_itxg[synced_txg & TXG_MASK]; itxs_t *clean_me; mutex_enter(&itxg->itxg_lock); if (itxg->itxg_itxs == NULL || itxg->itxg_txg == ZILTEST_TXG) { mutex_exit(&itxg->itxg_lock); return; } ASSERT3U(itxg->itxg_txg, <=, synced_txg); ASSERT(itxg->itxg_txg != 0); ASSERT(zilog->zl_clean_taskq != NULL); atomic_add_64(&zilog->zl_itx_list_sz, -itxg->itxg_sod); itxg->itxg_sod = 0; clean_me = itxg->itxg_itxs; itxg->itxg_itxs = NULL; itxg->itxg_txg = 0; mutex_exit(&itxg->itxg_lock); /* * Preferably start a task queue to free up the old itxs but * if taskq_dispatch can't allocate resources to do that then * free it in-line. This should be rare. Note, using TQ_SLEEP * created a bad performance problem. */ if (taskq_dispatch(zilog->zl_clean_taskq, (void (*)(void *))zil_itxg_clean, clean_me, TQ_NOSLEEP) == 0) zil_itxg_clean(clean_me); } /* * Get the list of itxs to commit into zl_itx_commit_list. */ static void zil_get_commit_list(zilog_t *zilog) { uint64_t otxg, txg; list_t *commit_list = &zilog->zl_itx_commit_list; uint64_t push_sod = 0; if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */ otxg = ZILTEST_TXG; else otxg = spa_last_synced_txg(zilog->zl_spa) + 1; for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) { itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK]; mutex_enter(&itxg->itxg_lock); if (itxg->itxg_txg != txg) { mutex_exit(&itxg->itxg_lock); continue; } list_move_tail(commit_list, &itxg->itxg_itxs->i_sync_list); push_sod += itxg->itxg_sod; itxg->itxg_sod = 0; mutex_exit(&itxg->itxg_lock); } atomic_add_64(&zilog->zl_itx_list_sz, -push_sod); } /* * Move the async itxs for a specified object to commit into sync lists. */ -static void +void zil_async_to_sync(zilog_t *zilog, uint64_t foid) { uint64_t otxg, txg; itx_async_node_t *ian; avl_tree_t *t; avl_index_t where; if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */ otxg = ZILTEST_TXG; else otxg = spa_last_synced_txg(zilog->zl_spa) + 1; for (txg = otxg; txg < (otxg + TXG_CONCURRENT_STATES); txg++) { itxg_t *itxg = &zilog->zl_itxg[txg & TXG_MASK]; mutex_enter(&itxg->itxg_lock); if (itxg->itxg_txg != txg) { mutex_exit(&itxg->itxg_lock); continue; } /* * If a foid is specified then find that node and append its * list. Otherwise walk the tree appending all the lists * to the sync list. We add to the end rather than the * beginning to ensure the create has happened. */ t = &itxg->itxg_itxs->i_async_tree; if (foid != 0) { ian = avl_find(t, &foid, &where); if (ian != NULL) { list_move_tail(&itxg->itxg_itxs->i_sync_list, &ian->ia_list); } } else { void *cookie = NULL; while ((ian = avl_destroy_nodes(t, &cookie)) != NULL) { list_move_tail(&itxg->itxg_itxs->i_sync_list, &ian->ia_list); list_destroy(&ian->ia_list); kmem_free(ian, sizeof (itx_async_node_t)); } } mutex_exit(&itxg->itxg_lock); } } static void zil_commit_writer(zilog_t *zilog) { uint64_t txg; itx_t *itx; lwb_t *lwb; spa_t *spa = zilog->zl_spa; int error = 0; ASSERT(zilog->zl_root_zio == NULL); mutex_exit(&zilog->zl_lock); zil_get_commit_list(zilog); /* * Return if there's nothing to commit before we dirty the fs by * calling zil_create(). */ if (list_head(&zilog->zl_itx_commit_list) == NULL) { mutex_enter(&zilog->zl_lock); return; } if (zilog->zl_suspend) { lwb = NULL; } else { lwb = list_tail(&zilog->zl_lwb_list); if (lwb == NULL) lwb = zil_create(zilog); } DTRACE_PROBE1(zil__cw1, zilog_t *, zilog); while (itx = list_head(&zilog->zl_itx_commit_list)) { txg = itx->itx_lr.lrc_txg; ASSERT(txg); if (txg > spa_last_synced_txg(spa) || txg > spa_freeze_txg(spa)) lwb = zil_lwb_commit(zilog, itx, lwb); list_remove(&zilog->zl_itx_commit_list, itx); zil_itx_destroy(itx); } DTRACE_PROBE1(zil__cw2, zilog_t *, zilog); /* write the last block out */ if (lwb != NULL && lwb->lwb_zio != NULL) lwb = zil_lwb_write_start(zilog, lwb); zilog->zl_cur_used = 0; /* * Wait if necessary for the log blocks to be on stable storage. */ if (zilog->zl_root_zio) { error = zio_wait(zilog->zl_root_zio); zilog->zl_root_zio = NULL; zil_flush_vdevs(zilog); } if (error || lwb == NULL) txg_wait_synced(zilog->zl_dmu_pool, 0); mutex_enter(&zilog->zl_lock); /* * Remember the highest committed log sequence number for ztest. * We only update this value when all the log writes succeeded, * because ztest wants to ASSERT that it got the whole log chain. */ if (error == 0 && lwb != NULL) zilog->zl_commit_lr_seq = zilog->zl_lr_seq; } /* * Commit zfs transactions to stable storage. * If foid is 0 push out all transactions, otherwise push only those * for that object or might reference that object. * * itxs are committed in batches. In a heavily stressed zil there will be * a commit writer thread who is writing out a bunch of itxs to the log * for a set of committing threads (cthreads) in the same batch as the writer. * Those cthreads are all waiting on the same cv for that batch. * * There will also be a different and growing batch of threads that are * waiting to commit (qthreads). When the committing batch completes * a transition occurs such that the cthreads exit and the qthreads become * cthreads. One of the new cthreads becomes the writer thread for the * batch. Any new threads arriving become new qthreads. * * Only 2 condition variables are needed and there's no transition * between the two cvs needed. They just flip-flop between qthreads * and cthreads. * * Using this scheme we can efficiently wakeup up only those threads * that have been committed. */ void zil_commit(zilog_t *zilog, uint64_t foid) { uint64_t mybatch; if (zilog->zl_sync == ZFS_SYNC_DISABLED) return; /* move the async itxs for the foid to the sync queues */ zil_async_to_sync(zilog, foid); mutex_enter(&zilog->zl_lock); mybatch = zilog->zl_next_batch; while (zilog->zl_writer) { cv_wait(&zilog->zl_cv_batch[mybatch & 1], &zilog->zl_lock); if (mybatch <= zilog->zl_com_batch) { mutex_exit(&zilog->zl_lock); return; } } zilog->zl_next_batch++; zilog->zl_writer = B_TRUE; zil_commit_writer(zilog); zilog->zl_com_batch = mybatch; zilog->zl_writer = B_FALSE; mutex_exit(&zilog->zl_lock); /* wake up one thread to become the next writer */ cv_signal(&zilog->zl_cv_batch[(mybatch+1) & 1]); /* wake up all threads waiting for this batch to be committed */ cv_broadcast(&zilog->zl_cv_batch[mybatch & 1]); } /* * Called in syncing context to free committed log blocks and update log header. */ void zil_sync(zilog_t *zilog, dmu_tx_t *tx) { zil_header_t *zh = zil_header_in_syncing_context(zilog); uint64_t txg = dmu_tx_get_txg(tx); spa_t *spa = zilog->zl_spa; uint64_t *replayed_seq = &zilog->zl_replayed_seq[txg & TXG_MASK]; lwb_t *lwb; /* * We don't zero out zl_destroy_txg, so make sure we don't try * to destroy it twice. */ if (spa_sync_pass(spa) != 1) return; mutex_enter(&zilog->zl_lock); ASSERT(zilog->zl_stop_sync == 0); if (*replayed_seq != 0) { ASSERT(zh->zh_replay_seq < *replayed_seq); zh->zh_replay_seq = *replayed_seq; *replayed_seq = 0; } if (zilog->zl_destroy_txg == txg) { blkptr_t blk = zh->zh_log; ASSERT(list_head(&zilog->zl_lwb_list) == NULL); bzero(zh, sizeof (zil_header_t)); bzero(zilog->zl_replayed_seq, sizeof (zilog->zl_replayed_seq)); if (zilog->zl_keep_first) { /* * If this block was part of log chain that couldn't * be claimed because a device was missing during * zil_claim(), but that device later returns, * then this block could erroneously appear valid. * To guard against this, assign a new GUID to the new * log chain so it doesn't matter what blk points to. */ zil_init_log_chain(zilog, &blk); zh->zh_log = blk; } } while ((lwb = list_head(&zilog->zl_lwb_list)) != NULL) { zh->zh_log = lwb->lwb_blk; if (lwb->lwb_buf != NULL || lwb->lwb_max_txg > txg) break; list_remove(&zilog->zl_lwb_list, lwb); zio_free_zil(spa, txg, &lwb->lwb_blk); kmem_cache_free(zil_lwb_cache, lwb); /* * If we don't have anything left in the lwb list then * we've had an allocation failure and we need to zero * out the zil_header blkptr so that we don't end * up freeing the same block twice. */ if (list_head(&zilog->zl_lwb_list) == NULL) BP_ZERO(&zh->zh_log); } mutex_exit(&zilog->zl_lock); } void zil_init(void) { zil_lwb_cache = kmem_cache_create("zil_lwb_cache", sizeof (struct lwb), 0, NULL, NULL, NULL, NULL, NULL, 0); } void zil_fini(void) { kmem_cache_destroy(zil_lwb_cache); } void zil_set_sync(zilog_t *zilog, uint64_t sync) { zilog->zl_sync = sync; } void zil_set_logbias(zilog_t *zilog, uint64_t logbias) { zilog->zl_logbias = logbias; } zilog_t * zil_alloc(objset_t *os, zil_header_t *zh_phys) { zilog_t *zilog; zilog = kmem_zalloc(sizeof (zilog_t), KM_SLEEP); zilog->zl_header = zh_phys; zilog->zl_os = os; zilog->zl_spa = dmu_objset_spa(os); zilog->zl_dmu_pool = dmu_objset_pool(os); zilog->zl_destroy_txg = TXG_INITIAL - 1; zilog->zl_logbias = dmu_objset_logbias(os); zilog->zl_sync = dmu_objset_syncprop(os); zilog->zl_next_batch = 1; mutex_init(&zilog->zl_lock, NULL, MUTEX_DEFAULT, NULL); for (int i = 0; i < TXG_SIZE; i++) { mutex_init(&zilog->zl_itxg[i].itxg_lock, NULL, MUTEX_DEFAULT, NULL); } list_create(&zilog->zl_lwb_list, sizeof (lwb_t), offsetof(lwb_t, lwb_node)); list_create(&zilog->zl_itx_commit_list, sizeof (itx_t), offsetof(itx_t, itx_node)); mutex_init(&zilog->zl_vdev_lock, NULL, MUTEX_DEFAULT, NULL); avl_create(&zilog->zl_vdev_tree, zil_vdev_compare, sizeof (zil_vdev_node_t), offsetof(zil_vdev_node_t, zv_node)); cv_init(&zilog->zl_cv_writer, NULL, CV_DEFAULT, NULL); cv_init(&zilog->zl_cv_suspend, NULL, CV_DEFAULT, NULL); cv_init(&zilog->zl_cv_batch[0], NULL, CV_DEFAULT, NULL); cv_init(&zilog->zl_cv_batch[1], NULL, CV_DEFAULT, NULL); return (zilog); } void zil_free(zilog_t *zilog) { zilog->zl_stop_sync = 1; ASSERT0(zilog->zl_suspend); ASSERT0(zilog->zl_suspending); ASSERT(list_is_empty(&zilog->zl_lwb_list)); list_destroy(&zilog->zl_lwb_list); avl_destroy(&zilog->zl_vdev_tree); mutex_destroy(&zilog->zl_vdev_lock); ASSERT(list_is_empty(&zilog->zl_itx_commit_list)); list_destroy(&zilog->zl_itx_commit_list); for (int i = 0; i < TXG_SIZE; i++) { /* * It's possible for an itx to be generated that doesn't dirty * a txg (e.g. ztest TX_TRUNCATE). So there's no zil_clean() * callback to remove the entry. We remove those here. * * Also free up the ziltest itxs. */ if (zilog->zl_itxg[i].itxg_itxs) zil_itxg_clean(zilog->zl_itxg[i].itxg_itxs); mutex_destroy(&zilog->zl_itxg[i].itxg_lock); } mutex_destroy(&zilog->zl_lock); cv_destroy(&zilog->zl_cv_writer); cv_destroy(&zilog->zl_cv_suspend); cv_destroy(&zilog->zl_cv_batch[0]); cv_destroy(&zilog->zl_cv_batch[1]); kmem_free(zilog, sizeof (zilog_t)); } /* * Open an intent log. */ zilog_t * zil_open(objset_t *os, zil_get_data_t *get_data) { zilog_t *zilog = dmu_objset_zil(os); ASSERT(zilog->zl_clean_taskq == NULL); ASSERT(zilog->zl_get_data == NULL); ASSERT(list_is_empty(&zilog->zl_lwb_list)); zilog->zl_get_data = get_data; zilog->zl_clean_taskq = taskq_create("zil_clean", 1, minclsyspri, 2, 2, TASKQ_PREPOPULATE); return (zilog); } /* * Close an intent log. */ void zil_close(zilog_t *zilog) { lwb_t *lwb; uint64_t txg = 0; zil_commit(zilog, 0); /* commit all itx */ /* * The lwb_max_txg for the stubby lwb will reflect the last activity * for the zil. After a txg_wait_synced() on the txg we know all the * callbacks have occurred that may clean the zil. Only then can we * destroy the zl_clean_taskq. */ mutex_enter(&zilog->zl_lock); lwb = list_tail(&zilog->zl_lwb_list); if (lwb != NULL) txg = lwb->lwb_max_txg; mutex_exit(&zilog->zl_lock); if (txg) txg_wait_synced(zilog->zl_dmu_pool, txg); ASSERT(!zilog_is_dirty(zilog)); taskq_destroy(zilog->zl_clean_taskq); zilog->zl_clean_taskq = NULL; zilog->zl_get_data = NULL; /* * We should have only one LWB left on the list; remove it now. */ mutex_enter(&zilog->zl_lock); lwb = list_head(&zilog->zl_lwb_list); if (lwb != NULL) { ASSERT(lwb == list_tail(&zilog->zl_lwb_list)); list_remove(&zilog->zl_lwb_list, lwb); zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); kmem_cache_free(zil_lwb_cache, lwb); } mutex_exit(&zilog->zl_lock); } static char *suspend_tag = "zil suspending"; /* * Suspend an intent log. While in suspended mode, we still honor * synchronous semantics, but we rely on txg_wait_synced() to do it. * On old version pools, we suspend the log briefly when taking a * snapshot so that it will have an empty intent log. * * Long holds are not really intended to be used the way we do here -- * held for such a short time. A concurrent caller of dsl_dataset_long_held() * could fail. Therefore we take pains to only put a long hold if it is * actually necessary. Fortunately, it will only be necessary if the * objset is currently mounted (or the ZVOL equivalent). In that case it * will already have a long hold, so we are not really making things any worse. * * Ideally, we would locate the existing long-holder (i.e. the zfsvfs_t or * zvol_state_t), and use their mechanism to prevent their hold from being * dropped (e.g. VFS_HOLD()). However, that would be even more pain for * very little gain. * * if cookiep == NULL, this does both the suspend & resume. * Otherwise, it returns with the dataset "long held", and the cookie * should be passed into zil_resume(). */ int zil_suspend(const char *osname, void **cookiep) { objset_t *os; zilog_t *zilog; const zil_header_t *zh; int error; error = dmu_objset_hold(osname, suspend_tag, &os); if (error != 0) return (error); zilog = dmu_objset_zil(os); mutex_enter(&zilog->zl_lock); zh = zilog->zl_header; if (zh->zh_flags & ZIL_REPLAY_NEEDED) { /* unplayed log */ mutex_exit(&zilog->zl_lock); dmu_objset_rele(os, suspend_tag); return (SET_ERROR(EBUSY)); } /* * Don't put a long hold in the cases where we can avoid it. This * is when there is no cookie so we are doing a suspend & resume * (i.e. called from zil_vdev_offline()), and there's nothing to do * for the suspend because it's already suspended, or there's no ZIL. */ if (cookiep == NULL && !zilog->zl_suspending && (zilog->zl_suspend > 0 || BP_IS_HOLE(&zh->zh_log))) { mutex_exit(&zilog->zl_lock); dmu_objset_rele(os, suspend_tag); return (0); } dsl_dataset_long_hold(dmu_objset_ds(os), suspend_tag); dsl_pool_rele(dmu_objset_pool(os), suspend_tag); zilog->zl_suspend++; if (zilog->zl_suspend > 1) { /* * Someone else is already suspending it. * Just wait for them to finish. */ while (zilog->zl_suspending) cv_wait(&zilog->zl_cv_suspend, &zilog->zl_lock); mutex_exit(&zilog->zl_lock); if (cookiep == NULL) zil_resume(os); else *cookiep = os; return (0); } /* * If there is no pointer to an on-disk block, this ZIL must not * be active (e.g. filesystem not mounted), so there's nothing * to clean up. */ if (BP_IS_HOLE(&zh->zh_log)) { ASSERT(cookiep != NULL); /* fast path already handled */ *cookiep = os; mutex_exit(&zilog->zl_lock); return (0); } zilog->zl_suspending = B_TRUE; mutex_exit(&zilog->zl_lock); zil_commit(zilog, 0); zil_destroy(zilog, B_FALSE); mutex_enter(&zilog->zl_lock); zilog->zl_suspending = B_FALSE; cv_broadcast(&zilog->zl_cv_suspend); mutex_exit(&zilog->zl_lock); if (cookiep == NULL) zil_resume(os); else *cookiep = os; return (0); } void zil_resume(void *cookie) { objset_t *os = cookie; zilog_t *zilog = dmu_objset_zil(os); mutex_enter(&zilog->zl_lock); ASSERT(zilog->zl_suspend != 0); zilog->zl_suspend--; mutex_exit(&zilog->zl_lock); dsl_dataset_long_rele(dmu_objset_ds(os), suspend_tag); dsl_dataset_rele(dmu_objset_ds(os), suspend_tag); } typedef struct zil_replay_arg { zil_replay_func_t **zr_replay; void *zr_arg; boolean_t zr_byteswap; char *zr_lr; } zil_replay_arg_t; static int zil_replay_error(zilog_t *zilog, lr_t *lr, int error) { char name[ZFS_MAX_DATASET_NAME_LEN]; zilog->zl_replaying_seq--; /* didn't actually replay this one */ dmu_objset_name(zilog->zl_os, name); cmn_err(CE_WARN, "ZFS replay transaction error %d, " "dataset %s, seq 0x%llx, txtype %llu %s\n", error, name, (u_longlong_t)lr->lrc_seq, (u_longlong_t)(lr->lrc_txtype & ~TX_CI), (lr->lrc_txtype & TX_CI) ? "CI" : ""); return (error); } static int zil_replay_log_record(zilog_t *zilog, lr_t *lr, void *zra, uint64_t claim_txg) { zil_replay_arg_t *zr = zra; const zil_header_t *zh = zilog->zl_header; uint64_t reclen = lr->lrc_reclen; uint64_t txtype = lr->lrc_txtype; int error = 0; zilog->zl_replaying_seq = lr->lrc_seq; if (lr->lrc_seq <= zh->zh_replay_seq) /* already replayed */ return (0); if (lr->lrc_txg < claim_txg) /* already committed */ return (0); /* Strip case-insensitive bit, still present in log record */ txtype &= ~TX_CI; if (txtype == 0 || txtype >= TX_MAX_TYPE) return (zil_replay_error(zilog, lr, EINVAL)); /* * If this record type can be logged out of order, the object * (lr_foid) may no longer exist. That's legitimate, not an error. */ if (TX_OOO(txtype)) { error = dmu_object_info(zilog->zl_os, ((lr_ooo_t *)lr)->lr_foid, NULL); if (error == ENOENT || error == EEXIST) return (0); } /* * Make a copy of the data so we can revise and extend it. */ bcopy(lr, zr->zr_lr, reclen); /* * If this is a TX_WRITE with a blkptr, suck in the data. */ if (txtype == TX_WRITE && reclen == sizeof (lr_write_t)) { error = zil_read_log_data(zilog, (lr_write_t *)lr, zr->zr_lr + reclen); if (error != 0) return (zil_replay_error(zilog, lr, error)); } /* * The log block containing this lr may have been byteswapped * so that we can easily examine common fields like lrc_txtype. * However, the log is a mix of different record types, and only the * replay vectors know how to byteswap their records. Therefore, if * the lr was byteswapped, undo it before invoking the replay vector. */ if (zr->zr_byteswap) byteswap_uint64_array(zr->zr_lr, reclen); /* * We must now do two things atomically: replay this log record, * and update the log header sequence number to reflect the fact that * we did so. At the end of each replay function the sequence number * is updated if we are in replay mode. */ error = zr->zr_replay[txtype](zr->zr_arg, zr->zr_lr, zr->zr_byteswap); if (error != 0) { /* * The DMU's dnode layer doesn't see removes until the txg * commits, so a subsequent claim can spuriously fail with * EEXIST. So if we receive any error we try syncing out * any removes then retry the transaction. Note that we * specify B_FALSE for byteswap now, so we don't do it twice. */ txg_wait_synced(spa_get_dsl(zilog->zl_spa), 0); error = zr->zr_replay[txtype](zr->zr_arg, zr->zr_lr, B_FALSE); if (error != 0) return (zil_replay_error(zilog, lr, error)); } return (0); } /* ARGSUSED */ static int zil_incr_blks(zilog_t *zilog, blkptr_t *bp, void *arg, uint64_t claim_txg) { zilog->zl_replay_blks++; return (0); } /* * If this dataset has a non-empty intent log, replay it and destroy it. */ void zil_replay(objset_t *os, void *arg, zil_replay_func_t *replay_func[TX_MAX_TYPE]) { zilog_t *zilog = dmu_objset_zil(os); const zil_header_t *zh = zilog->zl_header; zil_replay_arg_t zr; if ((zh->zh_flags & ZIL_REPLAY_NEEDED) == 0) { zil_destroy(zilog, B_TRUE); return; } zr.zr_replay = replay_func; zr.zr_arg = arg; zr.zr_byteswap = BP_SHOULD_BYTESWAP(&zh->zh_log); zr.zr_lr = kmem_alloc(2 * SPA_MAXBLOCKSIZE, KM_SLEEP); /* * Wait for in-progress removes to sync before starting replay. */ txg_wait_synced(zilog->zl_dmu_pool, 0); zilog->zl_replay = B_TRUE; zilog->zl_replay_time = ddi_get_lbolt(); ASSERT(zilog->zl_replay_blks == 0); (void) zil_parse(zilog, zil_incr_blks, zil_replay_log_record, &zr, zh->zh_claim_txg); kmem_free(zr.zr_lr, 2 * SPA_MAXBLOCKSIZE); zil_destroy(zilog, B_FALSE); txg_wait_synced(zilog->zl_dmu_pool, zilog->zl_destroy_txg); zilog->zl_replay = B_FALSE; } boolean_t zil_replaying(zilog_t *zilog, dmu_tx_t *tx) { if (zilog->zl_sync == ZFS_SYNC_DISABLED) return (B_TRUE); if (zilog->zl_replay) { dsl_dataset_dirty(dmu_objset_ds(zilog->zl_os), tx); zilog->zl_replayed_seq[dmu_tx_get_txg(tx) & TXG_MASK] = zilog->zl_replaying_seq; return (B_TRUE); } return (B_FALSE); } /* ARGSUSED */ int zil_vdev_offline(const char *osname, void *arg) { int error; error = zil_suspend(osname, NULL); if (error != 0) return (SET_ERROR(EEXIST)); return (0); } Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zvol.c =================================================================== --- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zvol.c (revision 308172) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zvol.c (revision 308173) @@ -1,3224 +1,3234 @@ /* * 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) 2006-2010 Pawel Jakub Dawidek * All rights reserved. * * Portions Copyright 2010 Robert Milkowski * * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, 2014 by Delphix. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ /* Portions Copyright 2011 Martin Matuska */ /* * ZFS volume emulation driver. * * Makes a DMU object look like a volume of arbitrary size, up to 2^64 bytes. * Volumes are accessed through the symbolic links named: * * /dev/zvol/dsk// * /dev/zvol/rdsk// * * These links are created by the /dev filesystem (sdev_zvolops.c). * Volumes are persistent through reboot. No user command needs to be * run before opening and using a device. * * FreeBSD notes. * On FreeBSD ZVOLs are simply GEOM providers like any other storage device * in the system. */ #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 #include #include #include #include #include #include #include #include #include #include #include #include #include "zfs_namecheck.h" #ifndef illumos struct g_class zfs_zvol_class = { .name = "ZFS::ZVOL", .version = G_VERSION, }; DECLARE_GEOM_CLASS(zfs_zvol_class, zfs_zvol); #endif void *zfsdev_state; static char *zvol_tag = "zvol_tag"; #define ZVOL_DUMPSIZE "dumpsize" /* * This lock protects the zfsdev_state structure from being modified * while it's being used, e.g. an open that comes in before a create * finishes. It also protects temporary opens of the dataset so that, * e.g., an open doesn't get a spurious EBUSY. */ #ifdef illumos kmutex_t zfsdev_state_lock; #else /* * In FreeBSD we've replaced the upstream zfsdev_state_lock with the * spa_namespace_lock in the ZVOL code. */ #define zfsdev_state_lock spa_namespace_lock #endif static uint32_t zvol_minors; #ifndef illumos SYSCTL_DECL(_vfs_zfs); SYSCTL_NODE(_vfs_zfs, OID_AUTO, vol, CTLFLAG_RW, 0, "ZFS VOLUME"); static int volmode = ZFS_VOLMODE_GEOM; SYSCTL_INT(_vfs_zfs_vol, OID_AUTO, mode, CTLFLAG_RWTUN, &volmode, 0, "Expose as GEOM providers (1), device files (2) or neither"); static boolean_t zpool_on_zvol = B_FALSE; SYSCTL_INT(_vfs_zfs_vol, OID_AUTO, recursive, CTLFLAG_RWTUN, &zpool_on_zvol, 0, "Allow zpools to use zvols as vdevs (DANGEROUS)"); #endif typedef struct zvol_extent { list_node_t ze_node; dva_t ze_dva; /* dva associated with this extent */ uint64_t ze_nblks; /* number of blocks in extent */ } zvol_extent_t; /* * The in-core state of each volume. */ typedef struct zvol_state { #ifndef illumos LIST_ENTRY(zvol_state) zv_links; #endif char zv_name[MAXPATHLEN]; /* pool/dd name */ uint64_t zv_volsize; /* amount of space we advertise */ uint64_t zv_volblocksize; /* volume block size */ #ifdef illumos minor_t zv_minor; /* minor number */ #else struct cdev *zv_dev; /* non-GEOM device */ struct g_provider *zv_provider; /* GEOM provider */ #endif uint8_t zv_min_bs; /* minimum addressable block shift */ uint8_t zv_flags; /* readonly, dumpified, etc. */ objset_t *zv_objset; /* objset handle */ #ifdef illumos uint32_t zv_open_count[OTYPCNT]; /* open counts */ #endif uint32_t zv_total_opens; /* total open count */ + uint32_t zv_sync_cnt; /* synchronous open count */ zilog_t *zv_zilog; /* ZIL handle */ list_t zv_extents; /* List of extents for dump */ znode_t zv_znode; /* for range locking */ dmu_buf_t *zv_dbuf; /* bonus handle */ #ifndef illumos int zv_state; int zv_volmode; /* Provide GEOM or cdev */ struct bio_queue_head zv_queue; struct mtx zv_queue_mtx; /* zv_queue mutex */ #endif } zvol_state_t; #ifndef illumos static LIST_HEAD(, zvol_state) all_zvols; #endif /* * zvol specific flags */ #define ZVOL_RDONLY 0x1 #define ZVOL_DUMPIFIED 0x2 #define ZVOL_EXCL 0x4 #define ZVOL_WCE 0x8 /* * zvol maximum transfer in one DMU tx. */ int zvol_maxphys = DMU_MAX_ACCESS/2; /* * Toggle unmap functionality. */ boolean_t zvol_unmap_enabled = B_TRUE; #ifndef illumos SYSCTL_INT(_vfs_zfs_vol, OID_AUTO, unmap_enabled, CTLFLAG_RWTUN, &zvol_unmap_enabled, 0, "Enable UNMAP functionality"); static d_open_t zvol_d_open; static d_close_t zvol_d_close; static d_read_t zvol_read; static d_write_t zvol_write; static d_ioctl_t zvol_d_ioctl; static d_strategy_t zvol_strategy; static struct cdevsw zvol_cdevsw = { .d_version = D_VERSION, .d_open = zvol_d_open, .d_close = zvol_d_close, .d_read = zvol_read, .d_write = zvol_write, .d_ioctl = zvol_d_ioctl, .d_strategy = zvol_strategy, .d_name = "zvol", .d_flags = D_DISK | D_TRACKCLOSE, }; static void zvol_geom_run(zvol_state_t *zv); static void zvol_geom_destroy(zvol_state_t *zv); static int zvol_geom_access(struct g_provider *pp, int acr, int acw, int ace); static void zvol_geom_start(struct bio *bp); static void zvol_geom_worker(void *arg); static void zvol_log_truncate(zvol_state_t *zv, dmu_tx_t *tx, uint64_t off, uint64_t len, boolean_t sync); #endif /* !illumos */ extern int zfs_set_prop_nvlist(const char *, zprop_source_t, nvlist_t *, nvlist_t *); static int zvol_remove_zv(zvol_state_t *); static int zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio); static int zvol_dumpify(zvol_state_t *zv); static int zvol_dump_fini(zvol_state_t *zv); static int zvol_dump_init(zvol_state_t *zv, boolean_t resize); static void zvol_size_changed(zvol_state_t *zv, uint64_t volsize) { #ifdef illumos dev_t dev = makedevice(ddi_driver_major(zfs_dip), zv->zv_minor); zv->zv_volsize = volsize; VERIFY(ddi_prop_update_int64(dev, zfs_dip, "Size", volsize) == DDI_SUCCESS); VERIFY(ddi_prop_update_int64(dev, zfs_dip, "Nblocks", lbtodb(volsize)) == DDI_SUCCESS); /* Notify specfs to invalidate the cached size */ spec_size_invalidate(dev, VBLK); spec_size_invalidate(dev, VCHR); #else /* !illumos */ zv->zv_volsize = volsize; if (zv->zv_volmode == ZFS_VOLMODE_GEOM) { struct g_provider *pp; pp = zv->zv_provider; if (pp == NULL) return; g_topology_lock(); g_resize_provider(pp, zv->zv_volsize); g_topology_unlock(); } #endif /* illumos */ } int zvol_check_volsize(uint64_t volsize, uint64_t blocksize) { if (volsize == 0) return (SET_ERROR(EINVAL)); if (volsize % blocksize != 0) return (SET_ERROR(EINVAL)); #ifdef _ILP32 if (volsize - 1 > SPEC_MAXOFFSET_T) return (SET_ERROR(EOVERFLOW)); #endif return (0); } int zvol_check_volblocksize(uint64_t volblocksize) { if (volblocksize < SPA_MINBLOCKSIZE || volblocksize > SPA_OLD_MAXBLOCKSIZE || !ISP2(volblocksize)) return (SET_ERROR(EDOM)); return (0); } int zvol_get_stats(objset_t *os, nvlist_t *nv) { int error; dmu_object_info_t doi; uint64_t val; error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &val); if (error) return (error); dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLSIZE, val); error = dmu_object_info(os, ZVOL_OBJ, &doi); if (error == 0) { dsl_prop_nvlist_add_uint64(nv, ZFS_PROP_VOLBLOCKSIZE, doi.doi_data_block_size); } return (error); } static zvol_state_t * zvol_minor_lookup(const char *name) { #ifdef illumos minor_t minor; #endif zvol_state_t *zv; ASSERT(MUTEX_HELD(&zfsdev_state_lock)); #ifdef illumos for (minor = 1; minor <= ZFSDEV_MAX_MINOR; minor++) { zv = zfsdev_get_soft_state(minor, ZSST_ZVOL); if (zv == NULL) continue; #else LIST_FOREACH(zv, &all_zvols, zv_links) { #endif if (strcmp(zv->zv_name, name) == 0) return (zv); } return (NULL); } /* extent mapping arg */ struct maparg { zvol_state_t *ma_zv; uint64_t ma_blks; }; /*ARGSUSED*/ static int zvol_map_block(spa_t *spa, zilog_t *zilog, const blkptr_t *bp, const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg) { struct maparg *ma = arg; zvol_extent_t *ze; int bs = ma->ma_zv->zv_volblocksize; if (bp == NULL || BP_IS_HOLE(bp) || zb->zb_object != ZVOL_OBJ || zb->zb_level != 0) return (0); VERIFY(!BP_IS_EMBEDDED(bp)); VERIFY3U(ma->ma_blks, ==, zb->zb_blkid); ma->ma_blks++; /* Abort immediately if we have encountered gang blocks */ if (BP_IS_GANG(bp)) return (SET_ERROR(EFRAGS)); /* * See if the block is at the end of the previous extent. */ ze = list_tail(&ma->ma_zv->zv_extents); if (ze && DVA_GET_VDEV(BP_IDENTITY(bp)) == DVA_GET_VDEV(&ze->ze_dva) && DVA_GET_OFFSET(BP_IDENTITY(bp)) == DVA_GET_OFFSET(&ze->ze_dva) + ze->ze_nblks * bs) { ze->ze_nblks++; return (0); } dprintf_bp(bp, "%s", "next blkptr:"); /* start a new extent */ ze = kmem_zalloc(sizeof (zvol_extent_t), KM_SLEEP); ze->ze_dva = bp->blk_dva[0]; /* structure assignment */ ze->ze_nblks = 1; list_insert_tail(&ma->ma_zv->zv_extents, ze); return (0); } static void zvol_free_extents(zvol_state_t *zv) { zvol_extent_t *ze; while (ze = list_head(&zv->zv_extents)) { list_remove(&zv->zv_extents, ze); kmem_free(ze, sizeof (zvol_extent_t)); } } static int zvol_get_lbas(zvol_state_t *zv) { objset_t *os = zv->zv_objset; struct maparg ma; int err; ma.ma_zv = zv; ma.ma_blks = 0; zvol_free_extents(zv); /* commit any in-flight changes before traversing the dataset */ txg_wait_synced(dmu_objset_pool(os), 0); err = traverse_dataset(dmu_objset_ds(os), 0, TRAVERSE_PRE | TRAVERSE_PREFETCH_METADATA, zvol_map_block, &ma); if (err || ma.ma_blks != (zv->zv_volsize / zv->zv_volblocksize)) { zvol_free_extents(zv); return (err ? err : EIO); } return (0); } /* ARGSUSED */ void zvol_create_cb(objset_t *os, void *arg, cred_t *cr, dmu_tx_t *tx) { zfs_creat_t *zct = arg; nvlist_t *nvprops = zct->zct_props; int error; uint64_t volblocksize, volsize; VERIFY(nvlist_lookup_uint64(nvprops, zfs_prop_to_name(ZFS_PROP_VOLSIZE), &volsize) == 0); if (nvlist_lookup_uint64(nvprops, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &volblocksize) != 0) volblocksize = zfs_prop_default_numeric(ZFS_PROP_VOLBLOCKSIZE); /* * These properties must be removed from the list so the generic * property setting step won't apply to them. */ VERIFY(nvlist_remove_all(nvprops, zfs_prop_to_name(ZFS_PROP_VOLSIZE)) == 0); (void) nvlist_remove_all(nvprops, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE)); error = dmu_object_claim(os, ZVOL_OBJ, DMU_OT_ZVOL, volblocksize, DMU_OT_NONE, 0, tx); ASSERT(error == 0); error = zap_create_claim(os, ZVOL_ZAP_OBJ, DMU_OT_ZVOL_PROP, DMU_OT_NONE, 0, tx); ASSERT(error == 0); error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx); ASSERT(error == 0); } /* * Replay a TX_TRUNCATE ZIL transaction if asked. TX_TRUNCATE is how we * implement DKIOCFREE/free-long-range. */ static int zvol_replay_truncate(zvol_state_t *zv, lr_truncate_t *lr, boolean_t byteswap) { uint64_t offset, length; if (byteswap) byteswap_uint64_array(lr, sizeof (*lr)); offset = lr->lr_offset; length = lr->lr_length; return (dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, offset, length)); } /* * Replay a TX_WRITE ZIL transaction that didn't get committed * after a system failure */ static int zvol_replay_write(zvol_state_t *zv, lr_write_t *lr, boolean_t byteswap) { objset_t *os = zv->zv_objset; char *data = (char *)(lr + 1); /* data follows lr_write_t */ uint64_t offset, length; dmu_tx_t *tx; int error; if (byteswap) byteswap_uint64_array(lr, sizeof (*lr)); offset = lr->lr_offset; length = lr->lr_length; /* If it's a dmu_sync() block, write the whole block */ if (lr->lr_common.lrc_reclen == sizeof (lr_write_t)) { uint64_t blocksize = BP_GET_LSIZE(&lr->lr_blkptr); if (length < blocksize) { offset -= offset % blocksize; length = blocksize; } } tx = dmu_tx_create(os); dmu_tx_hold_write(tx, ZVOL_OBJ, offset, length); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); } else { dmu_write(os, ZVOL_OBJ, offset, length, data, tx); dmu_tx_commit(tx); } return (error); } /* ARGSUSED */ static int zvol_replay_err(zvol_state_t *zv, lr_t *lr, boolean_t byteswap) { return (SET_ERROR(ENOTSUP)); } /* * Callback vectors for replaying records. * Only TX_WRITE and TX_TRUNCATE are needed for zvol. */ zil_replay_func_t *zvol_replay_vector[TX_MAX_TYPE] = { zvol_replay_err, /* 0 no such transaction type */ zvol_replay_err, /* TX_CREATE */ zvol_replay_err, /* TX_MKDIR */ zvol_replay_err, /* TX_MKXATTR */ zvol_replay_err, /* TX_SYMLINK */ zvol_replay_err, /* TX_REMOVE */ zvol_replay_err, /* TX_RMDIR */ zvol_replay_err, /* TX_LINK */ zvol_replay_err, /* TX_RENAME */ zvol_replay_write, /* TX_WRITE */ zvol_replay_truncate, /* TX_TRUNCATE */ zvol_replay_err, /* TX_SETATTR */ zvol_replay_err, /* TX_ACL */ zvol_replay_err, /* TX_CREATE_ACL */ zvol_replay_err, /* TX_CREATE_ATTR */ zvol_replay_err, /* TX_CREATE_ACL_ATTR */ zvol_replay_err, /* TX_MKDIR_ACL */ zvol_replay_err, /* TX_MKDIR_ATTR */ zvol_replay_err, /* TX_MKDIR_ACL_ATTR */ zvol_replay_err, /* TX_WRITE2 */ }; #ifdef illumos int zvol_name2minor(const char *name, minor_t *minor) { zvol_state_t *zv; mutex_enter(&zfsdev_state_lock); zv = zvol_minor_lookup(name); if (minor && zv) *minor = zv->zv_minor; mutex_exit(&zfsdev_state_lock); return (zv ? 0 : -1); } #endif /* illumos */ /* * Create a minor node (plus a whole lot more) for the specified volume. */ int zvol_create_minor(const char *name) { zfs_soft_state_t *zs; zvol_state_t *zv; objset_t *os; dmu_object_info_t doi; #ifdef illumos minor_t minor = 0; char chrbuf[30], blkbuf[30]; #else struct g_provider *pp; struct g_geom *gp; uint64_t volsize, mode; #endif int error; #ifndef illumos ZFS_LOG(1, "Creating ZVOL %s...", name); #endif mutex_enter(&zfsdev_state_lock); if (zvol_minor_lookup(name) != NULL) { mutex_exit(&zfsdev_state_lock); return (SET_ERROR(EEXIST)); } /* lie and say we're read-only */ error = dmu_objset_own(name, DMU_OST_ZVOL, B_TRUE, FTAG, &os); if (error) { mutex_exit(&zfsdev_state_lock); return (error); } #ifdef illumos if ((minor = zfsdev_minor_alloc()) == 0) { dmu_objset_disown(os, FTAG); mutex_exit(&zfsdev_state_lock); return (SET_ERROR(ENXIO)); } if (ddi_soft_state_zalloc(zfsdev_state, minor) != DDI_SUCCESS) { dmu_objset_disown(os, FTAG); mutex_exit(&zfsdev_state_lock); return (SET_ERROR(EAGAIN)); } (void) ddi_prop_update_string(minor, zfs_dip, ZVOL_PROP_NAME, (char *)name); (void) snprintf(chrbuf, sizeof (chrbuf), "%u,raw", minor); if (ddi_create_minor_node(zfs_dip, chrbuf, S_IFCHR, minor, DDI_PSEUDO, 0) == DDI_FAILURE) { ddi_soft_state_free(zfsdev_state, minor); dmu_objset_disown(os, FTAG); mutex_exit(&zfsdev_state_lock); return (SET_ERROR(EAGAIN)); } (void) snprintf(blkbuf, sizeof (blkbuf), "%u", minor); if (ddi_create_minor_node(zfs_dip, blkbuf, S_IFBLK, minor, DDI_PSEUDO, 0) == DDI_FAILURE) { ddi_remove_minor_node(zfs_dip, chrbuf); ddi_soft_state_free(zfsdev_state, minor); dmu_objset_disown(os, FTAG); mutex_exit(&zfsdev_state_lock); return (SET_ERROR(EAGAIN)); } zs = ddi_get_soft_state(zfsdev_state, minor); zs->zss_type = ZSST_ZVOL; zv = zs->zss_data = kmem_zalloc(sizeof (zvol_state_t), KM_SLEEP); #else /* !illumos */ zv = kmem_zalloc(sizeof(*zv), KM_SLEEP); zv->zv_state = 0; error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); if (error) { kmem_free(zv, sizeof(*zv)); dmu_objset_disown(os, zvol_tag); mutex_exit(&zfsdev_state_lock); return (error); } error = dsl_prop_get_integer(name, zfs_prop_to_name(ZFS_PROP_VOLMODE), &mode, NULL); if (error != 0 || mode == ZFS_VOLMODE_DEFAULT) mode = volmode; DROP_GIANT(); zv->zv_volsize = volsize; zv->zv_volmode = mode; if (zv->zv_volmode == ZFS_VOLMODE_GEOM) { g_topology_lock(); gp = g_new_geomf(&zfs_zvol_class, "zfs::zvol::%s", name); gp->start = zvol_geom_start; gp->access = zvol_geom_access; pp = g_new_providerf(gp, "%s/%s", ZVOL_DRIVER, name); pp->flags |= G_PF_DIRECT_RECEIVE | G_PF_DIRECT_SEND; pp->sectorsize = DEV_BSIZE; pp->mediasize = zv->zv_volsize; pp->private = zv; zv->zv_provider = pp; bioq_init(&zv->zv_queue); mtx_init(&zv->zv_queue_mtx, "zvol", NULL, MTX_DEF); } else if (zv->zv_volmode == ZFS_VOLMODE_DEV) { struct make_dev_args args; make_dev_args_init(&args); args.mda_flags = MAKEDEV_CHECKNAME | MAKEDEV_WAITOK; args.mda_devsw = &zvol_cdevsw; args.mda_cr = NULL; args.mda_uid = UID_ROOT; args.mda_gid = GID_OPERATOR; args.mda_mode = 0640; args.mda_si_drv2 = zv; error = make_dev_s(&args, &zv->zv_dev, "%s/%s", ZVOL_DRIVER, name); if (error != 0) { kmem_free(zv, sizeof(*zv)); dmu_objset_disown(os, FTAG); mutex_exit(&zfsdev_state_lock); return (error); } zv->zv_dev->si_iosize_max = MAXPHYS; } LIST_INSERT_HEAD(&all_zvols, zv, zv_links); #endif /* illumos */ (void) strlcpy(zv->zv_name, name, MAXPATHLEN); zv->zv_min_bs = DEV_BSHIFT; #ifdef illumos zv->zv_minor = minor; #endif zv->zv_objset = os; if (dmu_objset_is_snapshot(os) || !spa_writeable(dmu_objset_spa(os))) zv->zv_flags |= ZVOL_RDONLY; mutex_init(&zv->zv_znode.z_range_lock, NULL, MUTEX_DEFAULT, NULL); avl_create(&zv->zv_znode.z_range_avl, zfs_range_compare, sizeof (rl_t), offsetof(rl_t, r_node)); list_create(&zv->zv_extents, sizeof (zvol_extent_t), offsetof(zvol_extent_t, ze_node)); /* get and cache the blocksize */ error = dmu_object_info(os, ZVOL_OBJ, &doi); ASSERT(error == 0); zv->zv_volblocksize = doi.doi_data_block_size; if (spa_writeable(dmu_objset_spa(os))) { if (zil_replay_disable) zil_destroy(dmu_objset_zil(os), B_FALSE); else zil_replay(os, zv, zvol_replay_vector); } dmu_objset_disown(os, FTAG); zv->zv_objset = NULL; zvol_minors++; mutex_exit(&zfsdev_state_lock); #ifndef illumos if (zv->zv_volmode == ZFS_VOLMODE_GEOM) { zvol_geom_run(zv); g_topology_unlock(); } PICKUP_GIANT(); ZFS_LOG(1, "ZVOL %s created.", name); #endif return (0); } /* * Remove minor node for the specified volume. */ static int zvol_remove_zv(zvol_state_t *zv) { #ifdef illumos char nmbuf[20]; minor_t minor = zv->zv_minor; #endif ASSERT(MUTEX_HELD(&zfsdev_state_lock)); if (zv->zv_total_opens != 0) return (SET_ERROR(EBUSY)); #ifdef illumos (void) snprintf(nmbuf, sizeof (nmbuf), "%u,raw", minor); ddi_remove_minor_node(zfs_dip, nmbuf); (void) snprintf(nmbuf, sizeof (nmbuf), "%u", minor); ddi_remove_minor_node(zfs_dip, nmbuf); #else ZFS_LOG(1, "ZVOL %s destroyed.", zv->zv_name); LIST_REMOVE(zv, zv_links); if (zv->zv_volmode == ZFS_VOLMODE_GEOM) { g_topology_lock(); zvol_geom_destroy(zv); g_topology_unlock(); } else if (zv->zv_volmode == ZFS_VOLMODE_DEV) { if (zv->zv_dev != NULL) destroy_dev(zv->zv_dev); } #endif avl_destroy(&zv->zv_znode.z_range_avl); mutex_destroy(&zv->zv_znode.z_range_lock); kmem_free(zv, sizeof (zvol_state_t)); #ifdef illumos ddi_soft_state_free(zfsdev_state, minor); #endif zvol_minors--; return (0); } int zvol_remove_minor(const char *name) { zvol_state_t *zv; int rc; mutex_enter(&zfsdev_state_lock); if ((zv = zvol_minor_lookup(name)) == NULL) { mutex_exit(&zfsdev_state_lock); return (SET_ERROR(ENXIO)); } rc = zvol_remove_zv(zv); mutex_exit(&zfsdev_state_lock); return (rc); } int zvol_first_open(zvol_state_t *zv) { objset_t *os; uint64_t volsize; int error; uint64_t readonly; /* lie and say we're read-only */ error = dmu_objset_own(zv->zv_name, DMU_OST_ZVOL, B_TRUE, zvol_tag, &os); if (error) return (error); zv->zv_objset = os; error = zap_lookup(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize); if (error) { ASSERT(error == 0); dmu_objset_disown(os, zvol_tag); return (error); } error = dmu_bonus_hold(os, ZVOL_OBJ, zvol_tag, &zv->zv_dbuf); if (error) { dmu_objset_disown(os, zvol_tag); return (error); } zvol_size_changed(zv, volsize); zv->zv_zilog = zil_open(os, zvol_get_data); VERIFY(dsl_prop_get_integer(zv->zv_name, "readonly", &readonly, NULL) == 0); if (readonly || dmu_objset_is_snapshot(os) || !spa_writeable(dmu_objset_spa(os))) zv->zv_flags |= ZVOL_RDONLY; else zv->zv_flags &= ~ZVOL_RDONLY; return (error); } void zvol_last_close(zvol_state_t *zv) { zil_close(zv->zv_zilog); zv->zv_zilog = NULL; dmu_buf_rele(zv->zv_dbuf, zvol_tag); zv->zv_dbuf = NULL; /* * Evict cached data */ if (dsl_dataset_is_dirty(dmu_objset_ds(zv->zv_objset)) && !(zv->zv_flags & ZVOL_RDONLY)) txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0); dmu_objset_evict_dbufs(zv->zv_objset); dmu_objset_disown(zv->zv_objset, zvol_tag); zv->zv_objset = NULL; } #ifdef illumos int zvol_prealloc(zvol_state_t *zv) { objset_t *os = zv->zv_objset; dmu_tx_t *tx; uint64_t refd, avail, usedobjs, availobjs; uint64_t resid = zv->zv_volsize; uint64_t off = 0; /* Check the space usage before attempting to allocate the space */ dmu_objset_space(os, &refd, &avail, &usedobjs, &availobjs); if (avail < zv->zv_volsize) return (SET_ERROR(ENOSPC)); /* Free old extents if they exist */ zvol_free_extents(zv); while (resid != 0) { int error; uint64_t bytes = MIN(resid, SPA_OLD_MAXBLOCKSIZE); tx = dmu_tx_create(os); dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); (void) dmu_free_long_range(os, ZVOL_OBJ, 0, off); return (error); } dmu_prealloc(os, ZVOL_OBJ, off, bytes, tx); dmu_tx_commit(tx); off += bytes; resid -= bytes; } txg_wait_synced(dmu_objset_pool(os), 0); return (0); } #endif /* illumos */ static int zvol_update_volsize(objset_t *os, uint64_t volsize) { dmu_tx_t *tx; int error; ASSERT(MUTEX_HELD(&zfsdev_state_lock)); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); dmu_tx_mark_netfree(tx); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); return (error); } error = zap_update(os, ZVOL_ZAP_OBJ, "size", 8, 1, &volsize, tx); dmu_tx_commit(tx); if (error == 0) error = dmu_free_long_range(os, ZVOL_OBJ, volsize, DMU_OBJECT_END); return (error); } void zvol_remove_minors(const char *name) { #ifdef illumos zvol_state_t *zv; char *namebuf; minor_t minor; namebuf = kmem_zalloc(strlen(name) + 2, KM_SLEEP); (void) strncpy(namebuf, name, strlen(name)); (void) strcat(namebuf, "/"); mutex_enter(&zfsdev_state_lock); for (minor = 1; minor <= ZFSDEV_MAX_MINOR; minor++) { zv = zfsdev_get_soft_state(minor, ZSST_ZVOL); if (zv == NULL) continue; if (strncmp(namebuf, zv->zv_name, strlen(namebuf)) == 0) (void) zvol_remove_zv(zv); } kmem_free(namebuf, strlen(name) + 2); mutex_exit(&zfsdev_state_lock); #else /* !illumos */ zvol_state_t *zv, *tzv; size_t namelen; namelen = strlen(name); DROP_GIANT(); mutex_enter(&zfsdev_state_lock); LIST_FOREACH_SAFE(zv, &all_zvols, zv_links, tzv) { if (strcmp(zv->zv_name, name) == 0 || (strncmp(zv->zv_name, name, namelen) == 0 && strlen(zv->zv_name) > namelen && (zv->zv_name[namelen] == '/' || zv->zv_name[namelen] == '@'))) { (void) zvol_remove_zv(zv); } } mutex_exit(&zfsdev_state_lock); PICKUP_GIANT(); #endif /* illumos */ } static int zvol_update_live_volsize(zvol_state_t *zv, uint64_t volsize) { uint64_t old_volsize = 0ULL; int error = 0; ASSERT(MUTEX_HELD(&zfsdev_state_lock)); /* * Reinitialize the dump area to the new size. If we * failed to resize the dump area then restore it back to * its original size. We must set the new volsize prior * to calling dumpvp_resize() to ensure that the devices' * size(9P) is not visible by the dump subsystem. */ old_volsize = zv->zv_volsize; zvol_size_changed(zv, volsize); #ifdef ZVOL_DUMP if (zv->zv_flags & ZVOL_DUMPIFIED) { if ((error = zvol_dumpify(zv)) != 0 || (error = dumpvp_resize()) != 0) { int dumpify_error; (void) zvol_update_volsize(zv->zv_objset, old_volsize); zvol_size_changed(zv, old_volsize); dumpify_error = zvol_dumpify(zv); error = dumpify_error ? dumpify_error : error; } } #endif /* ZVOL_DUMP */ #ifdef illumos /* * Generate a LUN expansion event. */ if (error == 0) { sysevent_id_t eid; nvlist_t *attr; char *physpath = kmem_zalloc(MAXPATHLEN, KM_SLEEP); (void) snprintf(physpath, MAXPATHLEN, "%s%u", ZVOL_PSEUDO_DEV, zv->zv_minor); VERIFY(nvlist_alloc(&attr, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_string(attr, DEV_PHYS_PATH, physpath) == 0); (void) ddi_log_sysevent(zfs_dip, SUNW_VENDOR, EC_DEV_STATUS, ESC_DEV_DLE, attr, &eid, DDI_SLEEP); nvlist_free(attr); kmem_free(physpath, MAXPATHLEN); } #endif /* illumos */ return (error); } int zvol_set_volsize(const char *name, uint64_t volsize) { zvol_state_t *zv = NULL; objset_t *os; int error; dmu_object_info_t doi; uint64_t readonly; boolean_t owned = B_FALSE; error = dsl_prop_get_integer(name, zfs_prop_to_name(ZFS_PROP_READONLY), &readonly, NULL); if (error != 0) return (error); if (readonly) return (SET_ERROR(EROFS)); mutex_enter(&zfsdev_state_lock); zv = zvol_minor_lookup(name); if (zv == NULL || zv->zv_objset == NULL) { if ((error = dmu_objset_own(name, DMU_OST_ZVOL, B_FALSE, FTAG, &os)) != 0) { mutex_exit(&zfsdev_state_lock); return (error); } owned = B_TRUE; if (zv != NULL) zv->zv_objset = os; } else { os = zv->zv_objset; } if ((error = dmu_object_info(os, ZVOL_OBJ, &doi)) != 0 || (error = zvol_check_volsize(volsize, doi.doi_data_block_size)) != 0) goto out; error = zvol_update_volsize(os, volsize); if (error == 0 && zv != NULL) error = zvol_update_live_volsize(zv, volsize); out: if (owned) { dmu_objset_disown(os, FTAG); if (zv != NULL) zv->zv_objset = NULL; } mutex_exit(&zfsdev_state_lock); return (error); } /*ARGSUSED*/ #ifdef illumos int zvol_open(dev_t *devp, int flag, int otyp, cred_t *cr) #else static int zvol_open(struct g_provider *pp, int flag, int count) #endif { zvol_state_t *zv; int err = 0; #ifdef illumos mutex_enter(&zfsdev_state_lock); zv = zfsdev_get_soft_state(getminor(*devp), ZSST_ZVOL); if (zv == NULL) { mutex_exit(&zfsdev_state_lock); return (SET_ERROR(ENXIO)); } if (zv->zv_total_opens == 0) err = zvol_first_open(zv); if (err) { mutex_exit(&zfsdev_state_lock); return (err); } #else /* !illumos */ boolean_t locked = B_FALSE; if (!zpool_on_zvol && tsd_get(zfs_geom_probe_vdev_key) != NULL) { /* * if zfs_geom_probe_vdev_key is set, that means that zfs is * attempting to probe geom providers while looking for a * replacement for a missing VDEV. In this case, the * spa_namespace_lock will not be held, but it is still illegal * to use a zvol as a vdev. Deadlocks can result if another * thread has spa_namespace_lock */ return (EOPNOTSUPP); } /* * Protect against recursively entering spa_namespace_lock * when spa_open() is used for a pool on a (local) ZVOL(s). * This is needed since we replaced upstream zfsdev_state_lock * with spa_namespace_lock in the ZVOL code. * We are using the same trick as spa_open(). * Note that calls in zvol_first_open which need to resolve * pool name to a spa object will enter spa_open() * recursively, but that function already has all the * necessary protection. */ if (!MUTEX_HELD(&zfsdev_state_lock)) { mutex_enter(&zfsdev_state_lock); locked = B_TRUE; } zv = pp->private; if (zv == NULL) { if (locked) mutex_exit(&zfsdev_state_lock); return (SET_ERROR(ENXIO)); } if (zv->zv_total_opens == 0) { err = zvol_first_open(zv); if (err) { if (locked) mutex_exit(&zfsdev_state_lock); return (err); } pp->mediasize = zv->zv_volsize; pp->stripeoffset = 0; pp->stripesize = zv->zv_volblocksize; } #endif /* illumos */ if ((flag & FWRITE) && (zv->zv_flags & ZVOL_RDONLY)) { err = SET_ERROR(EROFS); goto out; } if (zv->zv_flags & ZVOL_EXCL) { err = SET_ERROR(EBUSY); goto out; } #ifdef FEXCL if (flag & FEXCL) { if (zv->zv_total_opens != 0) { err = SET_ERROR(EBUSY); goto out; } zv->zv_flags |= ZVOL_EXCL; } #endif #ifdef illumos if (zv->zv_open_count[otyp] == 0 || otyp == OTYP_LYR) { zv->zv_open_count[otyp]++; zv->zv_total_opens++; } mutex_exit(&zfsdev_state_lock); #else zv->zv_total_opens += count; if (locked) mutex_exit(&zfsdev_state_lock); #endif return (err); out: if (zv->zv_total_opens == 0) zvol_last_close(zv); #ifdef illumos mutex_exit(&zfsdev_state_lock); #else if (locked) mutex_exit(&zfsdev_state_lock); #endif return (err); } /*ARGSUSED*/ #ifdef illumos int zvol_close(dev_t dev, int flag, int otyp, cred_t *cr) { minor_t minor = getminor(dev); zvol_state_t *zv; int error = 0; mutex_enter(&zfsdev_state_lock); zv = zfsdev_get_soft_state(minor, ZSST_ZVOL); if (zv == NULL) { mutex_exit(&zfsdev_state_lock); #else /* !illumos */ static int zvol_close(struct g_provider *pp, int flag, int count) { zvol_state_t *zv; int error = 0; boolean_t locked = B_FALSE; /* See comment in zvol_open(). */ if (!MUTEX_HELD(&zfsdev_state_lock)) { mutex_enter(&zfsdev_state_lock); locked = B_TRUE; } zv = pp->private; if (zv == NULL) { if (locked) mutex_exit(&zfsdev_state_lock); #endif /* illumos */ return (SET_ERROR(ENXIO)); } if (zv->zv_flags & ZVOL_EXCL) { ASSERT(zv->zv_total_opens == 1); zv->zv_flags &= ~ZVOL_EXCL; } /* * If the open count is zero, this is a spurious close. * That indicates a bug in the kernel / DDI framework. */ #ifdef illumos ASSERT(zv->zv_open_count[otyp] != 0); #endif ASSERT(zv->zv_total_opens != 0); /* * You may get multiple opens, but only one close. */ #ifdef illumos zv->zv_open_count[otyp]--; zv->zv_total_opens--; #else zv->zv_total_opens -= count; #endif if (zv->zv_total_opens == 0) zvol_last_close(zv); #ifdef illumos mutex_exit(&zfsdev_state_lock); #else if (locked) mutex_exit(&zfsdev_state_lock); #endif return (error); } static void zvol_get_done(zgd_t *zgd, int error) { if (zgd->zgd_db) dmu_buf_rele(zgd->zgd_db, zgd); zfs_range_unlock(zgd->zgd_rl); if (error == 0 && zgd->zgd_bp) zil_add_block(zgd->zgd_zilog, zgd->zgd_bp); kmem_free(zgd, sizeof (zgd_t)); } /* * Get data to generate a TX_WRITE intent log record. */ static int zvol_get_data(void *arg, lr_write_t *lr, char *buf, zio_t *zio) { zvol_state_t *zv = arg; objset_t *os = zv->zv_objset; uint64_t object = ZVOL_OBJ; uint64_t offset = lr->lr_offset; uint64_t size = lr->lr_length; /* length of user data */ blkptr_t *bp = &lr->lr_blkptr; dmu_buf_t *db; zgd_t *zgd; int error; ASSERT(zio != NULL); ASSERT(size != 0); zgd = kmem_zalloc(sizeof (zgd_t), KM_SLEEP); zgd->zgd_zilog = zv->zv_zilog; zgd->zgd_rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_READER); /* * Write records come in two flavors: immediate and indirect. * For small writes it's cheaper to store the data with the * log record (immediate); for large writes it's cheaper to * sync the data and get a pointer to it (indirect) so that * we don't have to write the data twice. */ if (buf != NULL) { /* immediate write */ error = dmu_read(os, object, offset, size, buf, DMU_READ_NO_PREFETCH); } else { size = zv->zv_volblocksize; offset = P2ALIGN(offset, size); error = dmu_buf_hold(os, object, offset, zgd, &db, DMU_READ_NO_PREFETCH); if (error == 0) { blkptr_t *obp = dmu_buf_get_blkptr(db); if (obp) { ASSERT(BP_IS_HOLE(bp)); *bp = *obp; } zgd->zgd_db = db; zgd->zgd_bp = bp; ASSERT(db->db_offset == offset); ASSERT(db->db_size == size); error = dmu_sync(zio, lr->lr_common.lrc_txg, zvol_get_done, zgd); if (error == 0) return (0); } } zvol_get_done(zgd, error); return (error); } /* * zvol_log_write() handles synchronous writes using TX_WRITE ZIL transactions. * * We store data in the log buffers if it's small enough. * Otherwise we will later flush the data out via dmu_sync(). */ ssize_t zvol_immediate_write_sz = 32768; #ifdef _KERNEL SYSCTL_LONG(_vfs_zfs_vol, OID_AUTO, immediate_write_sz, CTLFLAG_RWTUN, &zvol_immediate_write_sz, 0, "Minimal size for indirect log write"); #endif static void zvol_log_write(zvol_state_t *zv, dmu_tx_t *tx, offset_t off, ssize_t resid, boolean_t sync) { uint32_t blocksize = zv->zv_volblocksize; zilog_t *zilog = zv->zv_zilog; boolean_t slogging; ssize_t immediate_write_sz; if (zil_replaying(zilog, tx)) return; immediate_write_sz = (zilog->zl_logbias == ZFS_LOGBIAS_THROUGHPUT) ? 0 : zvol_immediate_write_sz; slogging = spa_has_slogs(zilog->zl_spa) && (zilog->zl_logbias == ZFS_LOGBIAS_LATENCY); while (resid) { itx_t *itx; lr_write_t *lr; ssize_t len; itx_wr_state_t write_state; /* * Unlike zfs_log_write() we can be called with * upto DMU_MAX_ACCESS/2 (5MB) writes. */ if (blocksize > immediate_write_sz && !slogging && resid >= blocksize && off % blocksize == 0) { write_state = WR_INDIRECT; /* uses dmu_sync */ len = blocksize; } else if (sync) { write_state = WR_COPIED; len = MIN(ZIL_MAX_LOG_DATA, resid); } else { write_state = WR_NEED_COPY; len = MIN(ZIL_MAX_LOG_DATA, resid); } itx = zil_itx_create(TX_WRITE, sizeof (*lr) + (write_state == WR_COPIED ? len : 0)); lr = (lr_write_t *)&itx->itx_lr; if (write_state == WR_COPIED && dmu_read(zv->zv_objset, ZVOL_OBJ, off, len, lr + 1, DMU_READ_NO_PREFETCH) != 0) { zil_itx_destroy(itx); itx = zil_itx_create(TX_WRITE, sizeof (*lr)); lr = (lr_write_t *)&itx->itx_lr; write_state = WR_NEED_COPY; } itx->itx_wr_state = write_state; if (write_state == WR_NEED_COPY) itx->itx_sod += len; lr->lr_foid = ZVOL_OBJ; lr->lr_offset = off; lr->lr_length = len; lr->lr_blkoff = 0; BP_ZERO(&lr->lr_blkptr); itx->itx_private = zv; - itx->itx_sync = sync; + if (!sync && (zv->zv_sync_cnt == 0)) + itx->itx_sync = B_FALSE; + zil_itx_assign(zilog, itx, tx); off += len; resid -= len; } } #ifdef illumos static int zvol_dumpio_vdev(vdev_t *vd, void *addr, uint64_t offset, uint64_t origoffset, uint64_t size, boolean_t doread, boolean_t isdump) { vdev_disk_t *dvd; int c; int numerrors = 0; if (vd->vdev_ops == &vdev_mirror_ops || vd->vdev_ops == &vdev_replacing_ops || vd->vdev_ops == &vdev_spare_ops) { for (c = 0; c < vd->vdev_children; c++) { int err = zvol_dumpio_vdev(vd->vdev_child[c], addr, offset, origoffset, size, doread, isdump); if (err != 0) { numerrors++; } else if (doread) { break; } } } if (!vd->vdev_ops->vdev_op_leaf && vd->vdev_ops != &vdev_raidz_ops) return (numerrors < vd->vdev_children ? 0 : EIO); if (doread && !vdev_readable(vd)) return (SET_ERROR(EIO)); else if (!doread && !vdev_writeable(vd)) return (SET_ERROR(EIO)); if (vd->vdev_ops == &vdev_raidz_ops) { return (vdev_raidz_physio(vd, addr, size, offset, origoffset, doread, isdump)); } offset += VDEV_LABEL_START_SIZE; if (ddi_in_panic() || isdump) { ASSERT(!doread); if (doread) return (SET_ERROR(EIO)); dvd = vd->vdev_tsd; ASSERT3P(dvd, !=, NULL); return (ldi_dump(dvd->vd_lh, addr, lbtodb(offset), lbtodb(size))); } else { dvd = vd->vdev_tsd; ASSERT3P(dvd, !=, NULL); return (vdev_disk_ldi_physio(dvd->vd_lh, addr, size, offset, doread ? B_READ : B_WRITE)); } } static int zvol_dumpio(zvol_state_t *zv, void *addr, uint64_t offset, uint64_t size, boolean_t doread, boolean_t isdump) { vdev_t *vd; int error; zvol_extent_t *ze; spa_t *spa = dmu_objset_spa(zv->zv_objset); /* Must be sector aligned, and not stradle a block boundary. */ if (P2PHASE(offset, DEV_BSIZE) || P2PHASE(size, DEV_BSIZE) || P2BOUNDARY(offset, size, zv->zv_volblocksize)) { return (SET_ERROR(EINVAL)); } ASSERT(size <= zv->zv_volblocksize); /* Locate the extent this belongs to */ ze = list_head(&zv->zv_extents); while (offset >= ze->ze_nblks * zv->zv_volblocksize) { offset -= ze->ze_nblks * zv->zv_volblocksize; ze = list_next(&zv->zv_extents, ze); } if (ze == NULL) return (SET_ERROR(EINVAL)); if (!ddi_in_panic()) spa_config_enter(spa, SCL_STATE, FTAG, RW_READER); vd = vdev_lookup_top(spa, DVA_GET_VDEV(&ze->ze_dva)); offset += DVA_GET_OFFSET(&ze->ze_dva); error = zvol_dumpio_vdev(vd, addr, offset, DVA_GET_OFFSET(&ze->ze_dva), size, doread, isdump); if (!ddi_in_panic()) spa_config_exit(spa, SCL_STATE, FTAG); return (error); } int zvol_strategy(buf_t *bp) { zfs_soft_state_t *zs = NULL; #else /* !illumos */ void zvol_strategy(struct bio *bp) { #endif /* illumos */ zvol_state_t *zv; uint64_t off, volsize; size_t resid; char *addr; objset_t *os; rl_t *rl; int error = 0; #ifdef illumos boolean_t doread = bp->b_flags & B_READ; #else boolean_t doread = 0; #endif boolean_t is_dumpified; boolean_t sync; #ifdef illumos if (getminor(bp->b_edev) == 0) { error = SET_ERROR(EINVAL); } else { zs = ddi_get_soft_state(zfsdev_state, getminor(bp->b_edev)); if (zs == NULL) error = SET_ERROR(ENXIO); else if (zs->zss_type != ZSST_ZVOL) error = SET_ERROR(EINVAL); } if (error) { bioerror(bp, error); biodone(bp); return (0); } zv = zs->zss_data; if (!(bp->b_flags & B_READ) && (zv->zv_flags & ZVOL_RDONLY)) { bioerror(bp, EROFS); biodone(bp); return (0); } off = ldbtob(bp->b_blkno); #else /* !illumos */ if (bp->bio_to) zv = bp->bio_to->private; else zv = bp->bio_dev->si_drv2; if (zv == NULL) { error = SET_ERROR(ENXIO); goto out; } if (bp->bio_cmd != BIO_READ && (zv->zv_flags & ZVOL_RDONLY)) { error = SET_ERROR(EROFS); goto out; } switch (bp->bio_cmd) { case BIO_FLUSH: goto sync; case BIO_READ: doread = 1; case BIO_WRITE: case BIO_DELETE: break; default: error = EOPNOTSUPP; goto out; } off = bp->bio_offset; #endif /* illumos */ volsize = zv->zv_volsize; os = zv->zv_objset; ASSERT(os != NULL); #ifdef illumos bp_mapin(bp); addr = bp->b_un.b_addr; resid = bp->b_bcount; if (resid > 0 && (off < 0 || off >= volsize)) { bioerror(bp, EIO); biodone(bp); return (0); } is_dumpified = zv->zv_flags & ZVOL_DUMPIFIED; sync = ((!(bp->b_flags & B_ASYNC) && !(zv->zv_flags & ZVOL_WCE)) || (zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS)) && !doread && !is_dumpified; #else /* !illumos */ addr = bp->bio_data; resid = bp->bio_length; if (resid > 0 && (off < 0 || off >= volsize)) { error = SET_ERROR(EIO); goto out; } is_dumpified = B_FALSE; sync = !doread && !is_dumpified && zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS; #endif /* illumos */ /* * There must be no buffer changes when doing a dmu_sync() because * we can't change the data whilst calculating the checksum. */ rl = zfs_range_lock(&zv->zv_znode, off, resid, doread ? RL_READER : RL_WRITER); #ifndef illumos if (bp->bio_cmd == BIO_DELETE) { dmu_tx_t *tx = dmu_tx_create(zv->zv_objset); error = dmu_tx_assign(tx, TXG_WAIT); if (error != 0) { dmu_tx_abort(tx); } else { zvol_log_truncate(zv, tx, off, resid, sync); dmu_tx_commit(tx); error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, off, resid); resid = 0; } goto unlock; } #endif while (resid != 0 && off < volsize) { size_t size = MIN(resid, zvol_maxphys); #ifdef illumos if (is_dumpified) { size = MIN(size, P2END(off, zv->zv_volblocksize) - off); error = zvol_dumpio(zv, addr, off, size, doread, B_FALSE); } else if (doread) { #else if (doread) { #endif error = dmu_read(os, ZVOL_OBJ, off, size, addr, DMU_READ_PREFETCH); } else { dmu_tx_t *tx = dmu_tx_create(os); dmu_tx_hold_write(tx, ZVOL_OBJ, off, size); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); } else { dmu_write(os, ZVOL_OBJ, off, size, addr, tx); zvol_log_write(zv, tx, off, size, sync); dmu_tx_commit(tx); } } if (error) { /* convert checksum errors into IO errors */ if (error == ECKSUM) error = SET_ERROR(EIO); break; } off += size; addr += size; resid -= size; } #ifndef illumos unlock: #endif zfs_range_unlock(rl); #ifdef illumos if ((bp->b_resid = resid) == bp->b_bcount) bioerror(bp, off > volsize ? EINVAL : error); if (sync) zil_commit(zv->zv_zilog, ZVOL_OBJ); biodone(bp); return (0); #else /* !illumos */ bp->bio_completed = bp->bio_length - resid; if (bp->bio_completed < bp->bio_length && off > volsize) error = EINVAL; if (sync) { sync: zil_commit(zv->zv_zilog, ZVOL_OBJ); } out: if (bp->bio_to) g_io_deliver(bp, error); else biofinish(bp, NULL, error); #endif /* illumos */ } #ifdef illumos /* * Set the buffer count to the zvol maximum transfer. * Using our own routine instead of the default minphys() * means that for larger writes we write bigger buffers on X86 * (128K instead of 56K) and flush the disk write cache less often * (every zvol_maxphys - currently 1MB) instead of minphys (currently * 56K on X86 and 128K on sparc). */ void zvol_minphys(struct buf *bp) { if (bp->b_bcount > zvol_maxphys) bp->b_bcount = zvol_maxphys; } int zvol_dump(dev_t dev, caddr_t addr, daddr_t blkno, int nblocks) { minor_t minor = getminor(dev); zvol_state_t *zv; int error = 0; uint64_t size; uint64_t boff; uint64_t resid; zv = zfsdev_get_soft_state(minor, ZSST_ZVOL); if (zv == NULL) return (SET_ERROR(ENXIO)); if ((zv->zv_flags & ZVOL_DUMPIFIED) == 0) return (SET_ERROR(EINVAL)); boff = ldbtob(blkno); resid = ldbtob(nblocks); VERIFY3U(boff + resid, <=, zv->zv_volsize); while (resid) { size = MIN(resid, P2END(boff, zv->zv_volblocksize) - boff); error = zvol_dumpio(zv, addr, boff, size, B_FALSE, B_TRUE); if (error) break; boff += size; addr += size; resid -= size; } return (error); } /*ARGSUSED*/ int zvol_read(dev_t dev, uio_t *uio, cred_t *cr) { minor_t minor = getminor(dev); #else /* !illumos */ int zvol_read(struct cdev *dev, struct uio *uio, int ioflag) { #endif /* illumos */ zvol_state_t *zv; uint64_t volsize; rl_t *rl; int error = 0; #ifdef illumos zv = zfsdev_get_soft_state(minor, ZSST_ZVOL); if (zv == NULL) return (SET_ERROR(ENXIO)); #else zv = dev->si_drv2; #endif volsize = zv->zv_volsize; /* uio_loffset == volsize isn't an error as its required for EOF processing. */ if (uio->uio_resid > 0 && (uio->uio_loffset < 0 || uio->uio_loffset > volsize)) return (SET_ERROR(EIO)); #ifdef illumos if (zv->zv_flags & ZVOL_DUMPIFIED) { error = physio(zvol_strategy, NULL, dev, B_READ, zvol_minphys, uio); return (error); } #endif rl = zfs_range_lock(&zv->zv_znode, uio->uio_loffset, uio->uio_resid, RL_READER); while (uio->uio_resid > 0 && uio->uio_loffset < volsize) { uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1); /* don't read past the end */ if (bytes > volsize - uio->uio_loffset) bytes = volsize - uio->uio_loffset; error = dmu_read_uio_dbuf(zv->zv_dbuf, uio, bytes); if (error) { /* convert checksum errors into IO errors */ if (error == ECKSUM) error = SET_ERROR(EIO); break; } } zfs_range_unlock(rl); return (error); } #ifdef illumos /*ARGSUSED*/ int zvol_write(dev_t dev, uio_t *uio, cred_t *cr) { minor_t minor = getminor(dev); #else /* !illumos */ int zvol_write(struct cdev *dev, struct uio *uio, int ioflag) { #endif /* illumos */ zvol_state_t *zv; uint64_t volsize; rl_t *rl; int error = 0; boolean_t sync; #ifdef illumos zv = zfsdev_get_soft_state(minor, ZSST_ZVOL); if (zv == NULL) return (SET_ERROR(ENXIO)); #else zv = dev->si_drv2; #endif volsize = zv->zv_volsize; /* uio_loffset == volsize isn't an error as its required for EOF processing. */ if (uio->uio_resid > 0 && (uio->uio_loffset < 0 || uio->uio_loffset > volsize)) return (SET_ERROR(EIO)); #ifdef illumos if (zv->zv_flags & ZVOL_DUMPIFIED) { error = physio(zvol_strategy, NULL, dev, B_WRITE, zvol_minphys, uio); return (error); } sync = !(zv->zv_flags & ZVOL_WCE) || #else sync = (ioflag & IO_SYNC) || #endif (zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS); rl = zfs_range_lock(&zv->zv_znode, uio->uio_loffset, uio->uio_resid, RL_WRITER); while (uio->uio_resid > 0 && uio->uio_loffset < volsize) { uint64_t bytes = MIN(uio->uio_resid, DMU_MAX_ACCESS >> 1); uint64_t off = uio->uio_loffset; dmu_tx_t *tx = dmu_tx_create(zv->zv_objset); if (bytes > volsize - off) /* don't write past the end */ bytes = volsize - off; dmu_tx_hold_write(tx, ZVOL_OBJ, off, bytes); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); break; } error = dmu_write_uio_dbuf(zv->zv_dbuf, uio, bytes, tx); if (error == 0) zvol_log_write(zv, tx, off, bytes, sync); dmu_tx_commit(tx); if (error) break; } zfs_range_unlock(rl); if (sync) zil_commit(zv->zv_zilog, ZVOL_OBJ); return (error); } #ifdef illumos int zvol_getefi(void *arg, int flag, uint64_t vs, uint8_t bs) { struct uuid uuid = EFI_RESERVED; efi_gpe_t gpe = { 0 }; uint32_t crc; dk_efi_t efi; int length; char *ptr; if (ddi_copyin(arg, &efi, sizeof (dk_efi_t), flag)) return (SET_ERROR(EFAULT)); ptr = (char *)(uintptr_t)efi.dki_data_64; length = efi.dki_length; /* * Some clients may attempt to request a PMBR for the * zvol. Currently this interface will return EINVAL to * such requests. These requests could be supported by * adding a check for lba == 0 and consing up an appropriate * PMBR. */ if (efi.dki_lba < 1 || efi.dki_lba > 2 || length <= 0) return (SET_ERROR(EINVAL)); gpe.efi_gpe_StartingLBA = LE_64(34ULL); gpe.efi_gpe_EndingLBA = LE_64((vs >> bs) - 1); UUID_LE_CONVERT(gpe.efi_gpe_PartitionTypeGUID, uuid); if (efi.dki_lba == 1) { efi_gpt_t gpt = { 0 }; gpt.efi_gpt_Signature = LE_64(EFI_SIGNATURE); gpt.efi_gpt_Revision = LE_32(EFI_VERSION_CURRENT); gpt.efi_gpt_HeaderSize = LE_32(sizeof (gpt)); gpt.efi_gpt_MyLBA = LE_64(1ULL); gpt.efi_gpt_FirstUsableLBA = LE_64(34ULL); gpt.efi_gpt_LastUsableLBA = LE_64((vs >> bs) - 1); gpt.efi_gpt_PartitionEntryLBA = LE_64(2ULL); gpt.efi_gpt_NumberOfPartitionEntries = LE_32(1); gpt.efi_gpt_SizeOfPartitionEntry = LE_32(sizeof (efi_gpe_t)); CRC32(crc, &gpe, sizeof (gpe), -1U, crc32_table); gpt.efi_gpt_PartitionEntryArrayCRC32 = LE_32(~crc); CRC32(crc, &gpt, sizeof (gpt), -1U, crc32_table); gpt.efi_gpt_HeaderCRC32 = LE_32(~crc); if (ddi_copyout(&gpt, ptr, MIN(sizeof (gpt), length), flag)) return (SET_ERROR(EFAULT)); ptr += sizeof (gpt); length -= sizeof (gpt); } if (length > 0 && ddi_copyout(&gpe, ptr, MIN(sizeof (gpe), length), flag)) return (SET_ERROR(EFAULT)); return (0); } /* * BEGIN entry points to allow external callers access to the volume. */ /* * Return the volume parameters needed for access from an external caller. * These values are invariant as long as the volume is held open. */ int zvol_get_volume_params(minor_t minor, uint64_t *blksize, uint64_t *max_xfer_len, void **minor_hdl, void **objset_hdl, void **zil_hdl, void **rl_hdl, void **bonus_hdl) { zvol_state_t *zv; zv = zfsdev_get_soft_state(minor, ZSST_ZVOL); if (zv == NULL) return (SET_ERROR(ENXIO)); if (zv->zv_flags & ZVOL_DUMPIFIED) return (SET_ERROR(ENXIO)); ASSERT(blksize && max_xfer_len && minor_hdl && objset_hdl && zil_hdl && rl_hdl && bonus_hdl); *blksize = zv->zv_volblocksize; *max_xfer_len = (uint64_t)zvol_maxphys; *minor_hdl = zv; *objset_hdl = zv->zv_objset; *zil_hdl = zv->zv_zilog; *rl_hdl = &zv->zv_znode; *bonus_hdl = zv->zv_dbuf; return (0); } /* * Return the current volume size to an external caller. * The size can change while the volume is open. */ uint64_t zvol_get_volume_size(void *minor_hdl) { zvol_state_t *zv = minor_hdl; return (zv->zv_volsize); } /* * Return the current WCE setting to an external caller. * The WCE setting can change while the volume is open. */ int zvol_get_volume_wce(void *minor_hdl) { zvol_state_t *zv = minor_hdl; return ((zv->zv_flags & ZVOL_WCE) ? 1 : 0); } /* * Entry point for external callers to zvol_log_write */ void zvol_log_write_minor(void *minor_hdl, dmu_tx_t *tx, offset_t off, ssize_t resid, boolean_t sync) { zvol_state_t *zv = minor_hdl; zvol_log_write(zv, tx, off, resid, sync); } /* * END entry points to allow external callers access to the volume. */ #endif /* illumos */ /* * Log a DKIOCFREE/free-long-range to the ZIL with TX_TRUNCATE. */ static void zvol_log_truncate(zvol_state_t *zv, dmu_tx_t *tx, uint64_t off, uint64_t len, boolean_t sync) { itx_t *itx; lr_truncate_t *lr; zilog_t *zilog = zv->zv_zilog; if (zil_replaying(zilog, tx)) return; itx = zil_itx_create(TX_TRUNCATE, sizeof (*lr)); lr = (lr_truncate_t *)&itx->itx_lr; lr->lr_foid = ZVOL_OBJ; lr->lr_offset = off; lr->lr_length = len; - itx->itx_sync = sync; + itx->itx_sync = (sync || zv->zv_sync_cnt != 0); zil_itx_assign(zilog, itx, tx); } #ifdef illumos /* * Dirtbag ioctls to support mkfs(1M) for UFS filesystems. See dkio(7I). * Also a dirtbag dkio ioctl for unmap/free-block functionality. */ /*ARGSUSED*/ int zvol_ioctl(dev_t dev, int cmd, intptr_t arg, int flag, cred_t *cr, int *rvalp) { zvol_state_t *zv; struct dk_callback *dkc; int error = 0; rl_t *rl; mutex_enter(&zfsdev_state_lock); zv = zfsdev_get_soft_state(getminor(dev), ZSST_ZVOL); if (zv == NULL) { mutex_exit(&zfsdev_state_lock); return (SET_ERROR(ENXIO)); } ASSERT(zv->zv_total_opens > 0); switch (cmd) { case DKIOCINFO: { struct dk_cinfo dki; bzero(&dki, sizeof (dki)); (void) strcpy(dki.dki_cname, "zvol"); (void) strcpy(dki.dki_dname, "zvol"); dki.dki_ctype = DKC_UNKNOWN; dki.dki_unit = getminor(dev); dki.dki_maxtransfer = 1 << (SPA_OLD_MAXBLOCKSHIFT - zv->zv_min_bs); mutex_exit(&zfsdev_state_lock); if (ddi_copyout(&dki, (void *)arg, sizeof (dki), flag)) error = SET_ERROR(EFAULT); return (error); } case DKIOCGMEDIAINFO: { struct dk_minfo dkm; bzero(&dkm, sizeof (dkm)); dkm.dki_lbsize = 1U << zv->zv_min_bs; dkm.dki_capacity = zv->zv_volsize >> zv->zv_min_bs; dkm.dki_media_type = DK_UNKNOWN; mutex_exit(&zfsdev_state_lock); if (ddi_copyout(&dkm, (void *)arg, sizeof (dkm), flag)) error = SET_ERROR(EFAULT); return (error); } case DKIOCGMEDIAINFOEXT: { struct dk_minfo_ext dkmext; bzero(&dkmext, sizeof (dkmext)); dkmext.dki_lbsize = 1U << zv->zv_min_bs; dkmext.dki_pbsize = zv->zv_volblocksize; dkmext.dki_capacity = zv->zv_volsize >> zv->zv_min_bs; dkmext.dki_media_type = DK_UNKNOWN; mutex_exit(&zfsdev_state_lock); if (ddi_copyout(&dkmext, (void *)arg, sizeof (dkmext), flag)) error = SET_ERROR(EFAULT); return (error); } case DKIOCGETEFI: { uint64_t vs = zv->zv_volsize; uint8_t bs = zv->zv_min_bs; mutex_exit(&zfsdev_state_lock); error = zvol_getefi((void *)arg, flag, vs, bs); return (error); } case DKIOCFLUSHWRITECACHE: dkc = (struct dk_callback *)arg; mutex_exit(&zfsdev_state_lock); zil_commit(zv->zv_zilog, ZVOL_OBJ); if ((flag & FKIOCTL) && dkc != NULL && dkc->dkc_callback) { (*dkc->dkc_callback)(dkc->dkc_cookie, error); error = 0; } return (error); case DKIOCGETWCE: { int wce = (zv->zv_flags & ZVOL_WCE) ? 1 : 0; if (ddi_copyout(&wce, (void *)arg, sizeof (int), flag)) error = SET_ERROR(EFAULT); break; } case DKIOCSETWCE: { int wce; if (ddi_copyin((void *)arg, &wce, sizeof (int), flag)) { error = SET_ERROR(EFAULT); break; } if (wce) { zv->zv_flags |= ZVOL_WCE; mutex_exit(&zfsdev_state_lock); } else { zv->zv_flags &= ~ZVOL_WCE; mutex_exit(&zfsdev_state_lock); zil_commit(zv->zv_zilog, ZVOL_OBJ); } return (0); } case DKIOCGGEOM: case DKIOCGVTOC: /* * commands using these (like prtvtoc) expect ENOTSUP * since we're emulating an EFI label */ error = SET_ERROR(ENOTSUP); break; case DKIOCDUMPINIT: rl = zfs_range_lock(&zv->zv_znode, 0, zv->zv_volsize, RL_WRITER); error = zvol_dumpify(zv); zfs_range_unlock(rl); break; case DKIOCDUMPFINI: if (!(zv->zv_flags & ZVOL_DUMPIFIED)) break; rl = zfs_range_lock(&zv->zv_znode, 0, zv->zv_volsize, RL_WRITER); error = zvol_dump_fini(zv); zfs_range_unlock(rl); break; case DKIOCFREE: { dkioc_free_t df; dmu_tx_t *tx; if (!zvol_unmap_enabled) break; if (ddi_copyin((void *)arg, &df, sizeof (df), flag)) { error = SET_ERROR(EFAULT); break; } /* * Apply Postel's Law to length-checking. If they overshoot, * just blank out until the end, if there's a need to blank * out anything. */ if (df.df_start >= zv->zv_volsize) break; /* No need to do anything... */ mutex_exit(&zfsdev_state_lock); rl = zfs_range_lock(&zv->zv_znode, df.df_start, df.df_length, RL_WRITER); tx = dmu_tx_create(zv->zv_objset); dmu_tx_mark_netfree(tx); error = dmu_tx_assign(tx, TXG_WAIT); if (error != 0) { dmu_tx_abort(tx); } else { zvol_log_truncate(zv, tx, df.df_start, df.df_length, B_TRUE); dmu_tx_commit(tx); error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, df.df_start, df.df_length); } zfs_range_unlock(rl); if (error == 0) { /* * If the write-cache is disabled or 'sync' property * is set to 'always' then treat this as a synchronous * operation (i.e. commit to zil). */ if (!(zv->zv_flags & ZVOL_WCE) || (zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS)) zil_commit(zv->zv_zilog, ZVOL_OBJ); /* * If the caller really wants synchronous writes, and * can't wait for them, don't return until the write * is done. */ if (df.df_flags & DF_WAIT_SYNC) { txg_wait_synced( dmu_objset_pool(zv->zv_objset), 0); } } return (error); } default: error = SET_ERROR(ENOTTY); break; } mutex_exit(&zfsdev_state_lock); return (error); } #endif /* illumos */ int zvol_busy(void) { return (zvol_minors != 0); } void zvol_init(void) { VERIFY(ddi_soft_state_init(&zfsdev_state, sizeof (zfs_soft_state_t), 1) == 0); #ifdef illumos mutex_init(&zfsdev_state_lock, NULL, MUTEX_DEFAULT, NULL); #else ZFS_LOG(1, "ZVOL Initialized."); #endif } void zvol_fini(void) { #ifdef illumos mutex_destroy(&zfsdev_state_lock); #endif ddi_soft_state_fini(&zfsdev_state); ZFS_LOG(1, "ZVOL Deinitialized."); } #ifdef illumos /*ARGSUSED*/ static int zfs_mvdev_dump_feature_check(void *arg, dmu_tx_t *tx) { spa_t *spa = dmu_tx_pool(tx)->dp_spa; if (spa_feature_is_active(spa, SPA_FEATURE_MULTI_VDEV_CRASH_DUMP)) return (1); return (0); } /*ARGSUSED*/ static void zfs_mvdev_dump_activate_feature_sync(void *arg, dmu_tx_t *tx) { spa_t *spa = dmu_tx_pool(tx)->dp_spa; spa_feature_incr(spa, SPA_FEATURE_MULTI_VDEV_CRASH_DUMP, tx); } static int zvol_dump_init(zvol_state_t *zv, boolean_t resize) { dmu_tx_t *tx; int error; objset_t *os = zv->zv_objset; spa_t *spa = dmu_objset_spa(os); vdev_t *vd = spa->spa_root_vdev; nvlist_t *nv = NULL; uint64_t version = spa_version(spa); uint64_t checksum, compress, refresrv, vbs, dedup; ASSERT(MUTEX_HELD(&zfsdev_state_lock)); ASSERT(vd->vdev_ops == &vdev_root_ops); error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, 0, DMU_OBJECT_END); if (error != 0) return (error); /* wait for dmu_free_long_range to actually free the blocks */ txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0); /* * If the pool on which the dump device is being initialized has more * than one child vdev, check that the MULTI_VDEV_CRASH_DUMP feature is * enabled. If so, bump that feature's counter to indicate that the * feature is active. We also check the vdev type to handle the * following case: * # zpool create test raidz disk1 disk2 disk3 * Now have spa_root_vdev->vdev_children == 1 (the raidz vdev), * the raidz vdev itself has 3 children. */ if (vd->vdev_children > 1 || vd->vdev_ops == &vdev_raidz_ops) { if (!spa_feature_is_enabled(spa, SPA_FEATURE_MULTI_VDEV_CRASH_DUMP)) return (SET_ERROR(ENOTSUP)); (void) dsl_sync_task(spa_name(spa), zfs_mvdev_dump_feature_check, zfs_mvdev_dump_activate_feature_sync, NULL, 2, ZFS_SPACE_CHECK_RESERVED); } if (!resize) { error = dsl_prop_get_integer(zv->zv_name, zfs_prop_to_name(ZFS_PROP_COMPRESSION), &compress, NULL); if (error == 0) { error = dsl_prop_get_integer(zv->zv_name, zfs_prop_to_name(ZFS_PROP_CHECKSUM), &checksum, NULL); } if (error == 0) { error = dsl_prop_get_integer(zv->zv_name, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), &refresrv, NULL); } if (error == 0) { error = dsl_prop_get_integer(zv->zv_name, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), &vbs, NULL); } if (version >= SPA_VERSION_DEDUP && error == 0) { error = dsl_prop_get_integer(zv->zv_name, zfs_prop_to_name(ZFS_PROP_DEDUP), &dedup, NULL); } } if (error != 0) return (error); tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); dmu_tx_hold_bonus(tx, ZVOL_OBJ); error = dmu_tx_assign(tx, TXG_WAIT); if (error != 0) { dmu_tx_abort(tx); return (error); } /* * If we are resizing the dump device then we only need to * update the refreservation to match the newly updated * zvolsize. Otherwise, we save off the original state of the * zvol so that we can restore them if the zvol is ever undumpified. */ if (resize) { error = zap_update(os, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, &zv->zv_volsize, tx); } else { error = zap_update(os, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_COMPRESSION), 8, 1, &compress, tx); if (error == 0) { error = zap_update(os, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_CHECKSUM), 8, 1, &checksum, tx); } if (error == 0) { error = zap_update(os, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, &refresrv, tx); } if (error == 0) { error = zap_update(os, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 8, 1, &vbs, tx); } if (error == 0) { error = dmu_object_set_blocksize( os, ZVOL_OBJ, SPA_OLD_MAXBLOCKSIZE, 0, tx); } if (version >= SPA_VERSION_DEDUP && error == 0) { error = zap_update(os, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_DEDUP), 8, 1, &dedup, tx); } if (error == 0) zv->zv_volblocksize = SPA_OLD_MAXBLOCKSIZE; } dmu_tx_commit(tx); /* * We only need update the zvol's property if we are initializing * the dump area for the first time. */ if (error == 0 && !resize) { /* * If MULTI_VDEV_CRASH_DUMP is active, use the NOPARITY checksum * function. Otherwise, use the old default -- OFF. */ checksum = spa_feature_is_active(spa, SPA_FEATURE_MULTI_VDEV_CRASH_DUMP) ? ZIO_CHECKSUM_NOPARITY : ZIO_CHECKSUM_OFF; VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); VERIFY(nvlist_add_uint64(nv, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 0) == 0); VERIFY(nvlist_add_uint64(nv, zfs_prop_to_name(ZFS_PROP_COMPRESSION), ZIO_COMPRESS_OFF) == 0); VERIFY(nvlist_add_uint64(nv, zfs_prop_to_name(ZFS_PROP_CHECKSUM), checksum) == 0); if (version >= SPA_VERSION_DEDUP) { VERIFY(nvlist_add_uint64(nv, zfs_prop_to_name(ZFS_PROP_DEDUP), ZIO_CHECKSUM_OFF) == 0); } error = zfs_set_prop_nvlist(zv->zv_name, ZPROP_SRC_LOCAL, nv, NULL); nvlist_free(nv); } /* Allocate the space for the dump */ if (error == 0) error = zvol_prealloc(zv); return (error); } static int zvol_dumpify(zvol_state_t *zv) { int error = 0; uint64_t dumpsize = 0; dmu_tx_t *tx; objset_t *os = zv->zv_objset; if (zv->zv_flags & ZVOL_RDONLY) return (SET_ERROR(EROFS)); if (zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, 8, 1, &dumpsize) != 0 || dumpsize != zv->zv_volsize) { boolean_t resize = (dumpsize > 0); if ((error = zvol_dump_init(zv, resize)) != 0) { (void) zvol_dump_fini(zv); return (error); } } /* * Build up our lba mapping. */ error = zvol_get_lbas(zv); if (error) { (void) zvol_dump_fini(zv); return (error); } tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); (void) zvol_dump_fini(zv); return (error); } zv->zv_flags |= ZVOL_DUMPIFIED; error = zap_update(os, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, 8, 1, &zv->zv_volsize, tx); dmu_tx_commit(tx); if (error) { (void) zvol_dump_fini(zv); return (error); } txg_wait_synced(dmu_objset_pool(os), 0); return (0); } static int zvol_dump_fini(zvol_state_t *zv) { dmu_tx_t *tx; objset_t *os = zv->zv_objset; nvlist_t *nv; int error = 0; uint64_t checksum, compress, refresrv, vbs, dedup; uint64_t version = spa_version(dmu_objset_spa(zv->zv_objset)); /* * Attempt to restore the zvol back to its pre-dumpified state. * This is a best-effort attempt as it's possible that not all * of these properties were initialized during the dumpify process * (i.e. error during zvol_dump_init). */ tx = dmu_tx_create(os); dmu_tx_hold_zap(tx, ZVOL_ZAP_OBJ, TRUE, NULL); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); return (error); } (void) zap_remove(os, ZVOL_ZAP_OBJ, ZVOL_DUMPSIZE, tx); dmu_tx_commit(tx); (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_CHECKSUM), 8, 1, &checksum); (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_COMPRESSION), 8, 1, &compress); (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), 8, 1, &refresrv); (void) zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_VOLBLOCKSIZE), 8, 1, &vbs); VERIFY(nvlist_alloc(&nv, NV_UNIQUE_NAME, KM_SLEEP) == 0); (void) nvlist_add_uint64(nv, zfs_prop_to_name(ZFS_PROP_CHECKSUM), checksum); (void) nvlist_add_uint64(nv, zfs_prop_to_name(ZFS_PROP_COMPRESSION), compress); (void) nvlist_add_uint64(nv, zfs_prop_to_name(ZFS_PROP_REFRESERVATION), refresrv); if (version >= SPA_VERSION_DEDUP && zap_lookup(zv->zv_objset, ZVOL_ZAP_OBJ, zfs_prop_to_name(ZFS_PROP_DEDUP), 8, 1, &dedup) == 0) { (void) nvlist_add_uint64(nv, zfs_prop_to_name(ZFS_PROP_DEDUP), dedup); } (void) zfs_set_prop_nvlist(zv->zv_name, ZPROP_SRC_LOCAL, nv, NULL); nvlist_free(nv); zvol_free_extents(zv); zv->zv_flags &= ~ZVOL_DUMPIFIED; (void) dmu_free_long_range(os, ZVOL_OBJ, 0, DMU_OBJECT_END); /* wait for dmu_free_long_range to actually free the blocks */ txg_wait_synced(dmu_objset_pool(zv->zv_objset), 0); tx = dmu_tx_create(os); dmu_tx_hold_bonus(tx, ZVOL_OBJ); error = dmu_tx_assign(tx, TXG_WAIT); if (error) { dmu_tx_abort(tx); return (error); } if (dmu_object_set_blocksize(os, ZVOL_OBJ, vbs, 0, tx) == 0) zv->zv_volblocksize = vbs; dmu_tx_commit(tx); return (0); } #else /* !illumos */ static void zvol_geom_run(zvol_state_t *zv) { struct g_provider *pp; pp = zv->zv_provider; g_error_provider(pp, 0); kproc_kthread_add(zvol_geom_worker, zv, &zfsproc, NULL, 0, 0, "zfskern", "zvol %s", pp->name + sizeof(ZVOL_DRIVER)); } static void zvol_geom_destroy(zvol_state_t *zv) { struct g_provider *pp; g_topology_assert(); mtx_lock(&zv->zv_queue_mtx); zv->zv_state = 1; wakeup_one(&zv->zv_queue); while (zv->zv_state != 2) msleep(&zv->zv_state, &zv->zv_queue_mtx, 0, "zvol:w", 0); mtx_destroy(&zv->zv_queue_mtx); pp = zv->zv_provider; zv->zv_provider = NULL; pp->private = NULL; g_wither_geom(pp->geom, ENXIO); } static int zvol_geom_access(struct g_provider *pp, int acr, int acw, int ace) { int count, error, flags; g_topology_assert(); /* * To make it easier we expect either open or close, but not both * at the same time. */ KASSERT((acr >= 0 && acw >= 0 && ace >= 0) || (acr <= 0 && acw <= 0 && ace <= 0), ("Unsupported access request to %s (acr=%d, acw=%d, ace=%d).", pp->name, acr, acw, ace)); if (pp->private == NULL) { if (acr <= 0 && acw <= 0 && ace <= 0) return (0); return (pp->error); } /* * We don't pass FEXCL flag to zvol_open()/zvol_close() if ace != 0, * because GEOM already handles that and handles it a bit differently. * GEOM allows for multiple read/exclusive consumers and ZFS allows * only one exclusive consumer, no matter if it is reader or writer. * I like better the way GEOM works so I'll leave it for GEOM to * decide what to do. */ count = acr + acw + ace; if (count == 0) return (0); flags = 0; if (acr != 0 || ace != 0) flags |= FREAD; if (acw != 0) flags |= FWRITE; g_topology_unlock(); if (count > 0) error = zvol_open(pp, flags, count); else error = zvol_close(pp, flags, -count); g_topology_lock(); return (error); } static void zvol_geom_start(struct bio *bp) { zvol_state_t *zv; boolean_t first; zv = bp->bio_to->private; ASSERT(zv != NULL); switch (bp->bio_cmd) { case BIO_FLUSH: if (!THREAD_CAN_SLEEP()) goto enqueue; zil_commit(zv->zv_zilog, ZVOL_OBJ); g_io_deliver(bp, 0); break; case BIO_READ: case BIO_WRITE: case BIO_DELETE: if (!THREAD_CAN_SLEEP()) goto enqueue; zvol_strategy(bp); break; case BIO_GETATTR: { spa_t *spa = dmu_objset_spa(zv->zv_objset); uint64_t refd, avail, usedobjs, availobjs, val; if (g_handleattr_int(bp, "GEOM::candelete", 1)) return; if (strcmp(bp->bio_attribute, "blocksavail") == 0) { dmu_objset_space(zv->zv_objset, &refd, &avail, &usedobjs, &availobjs); if (g_handleattr_off_t(bp, "blocksavail", avail / DEV_BSIZE)) return; } else if (strcmp(bp->bio_attribute, "blocksused") == 0) { dmu_objset_space(zv->zv_objset, &refd, &avail, &usedobjs, &availobjs); if (g_handleattr_off_t(bp, "blocksused", refd / DEV_BSIZE)) return; } else if (strcmp(bp->bio_attribute, "poolblocksavail") == 0) { avail = metaslab_class_get_space(spa_normal_class(spa)); avail -= metaslab_class_get_alloc(spa_normal_class(spa)); if (g_handleattr_off_t(bp, "poolblocksavail", avail / DEV_BSIZE)) return; } else if (strcmp(bp->bio_attribute, "poolblocksused") == 0) { refd = metaslab_class_get_alloc(spa_normal_class(spa)); if (g_handleattr_off_t(bp, "poolblocksused", refd / DEV_BSIZE)) return; } /* FALLTHROUGH */ } default: g_io_deliver(bp, EOPNOTSUPP); break; } return; enqueue: mtx_lock(&zv->zv_queue_mtx); first = (bioq_first(&zv->zv_queue) == NULL); bioq_insert_tail(&zv->zv_queue, bp); mtx_unlock(&zv->zv_queue_mtx); if (first) wakeup_one(&zv->zv_queue); } static void zvol_geom_worker(void *arg) { zvol_state_t *zv; struct bio *bp; thread_lock(curthread); sched_prio(curthread, PRIBIO); thread_unlock(curthread); zv = arg; for (;;) { mtx_lock(&zv->zv_queue_mtx); bp = bioq_takefirst(&zv->zv_queue); if (bp == NULL) { if (zv->zv_state == 1) { zv->zv_state = 2; wakeup(&zv->zv_state); mtx_unlock(&zv->zv_queue_mtx); kthread_exit(); } msleep(&zv->zv_queue, &zv->zv_queue_mtx, PRIBIO | PDROP, "zvol:io", 0); continue; } mtx_unlock(&zv->zv_queue_mtx); switch (bp->bio_cmd) { case BIO_FLUSH: zil_commit(zv->zv_zilog, ZVOL_OBJ); g_io_deliver(bp, 0); break; case BIO_READ: case BIO_WRITE: case BIO_DELETE: zvol_strategy(bp); break; default: g_io_deliver(bp, EOPNOTSUPP); break; } } } extern boolean_t dataset_name_hidden(const char *name); static int zvol_create_snapshots(objset_t *os, const char *name) { uint64_t cookie, obj; char *sname; int error, len; cookie = obj = 0; sname = kmem_alloc(MAXPATHLEN, KM_SLEEP); #if 0 (void) dmu_objset_find(name, dmu_objset_prefetch, NULL, DS_FIND_SNAPSHOTS); #endif for (;;) { len = snprintf(sname, MAXPATHLEN, "%s@", name); if (len >= MAXPATHLEN) { dmu_objset_rele(os, FTAG); error = ENAMETOOLONG; break; } dsl_pool_config_enter(dmu_objset_pool(os), FTAG); error = dmu_snapshot_list_next(os, MAXPATHLEN - len, sname + len, &obj, &cookie, NULL); dsl_pool_config_exit(dmu_objset_pool(os), FTAG); if (error != 0) { if (error == ENOENT) error = 0; break; } error = zvol_create_minor(sname); if (error != 0 && error != EEXIST) { printf("ZFS WARNING: Unable to create ZVOL %s (error=%d).\n", sname, error); break; } } kmem_free(sname, MAXPATHLEN); return (error); } int zvol_create_minors(const char *name) { uint64_t cookie; objset_t *os; char *osname, *p; int error, len; if (dataset_name_hidden(name)) return (0); if ((error = dmu_objset_hold(name, FTAG, &os)) != 0) { printf("ZFS WARNING: Unable to put hold on %s (error=%d).\n", name, error); return (error); } if (dmu_objset_type(os) == DMU_OST_ZVOL) { dsl_dataset_long_hold(os->os_dsl_dataset, FTAG); dsl_pool_rele(dmu_objset_pool(os), FTAG); error = zvol_create_minor(name); if (error == 0 || error == EEXIST) { error = zvol_create_snapshots(os, name); } else { printf("ZFS WARNING: Unable to create ZVOL %s (error=%d).\n", name, error); } dsl_dataset_long_rele(os->os_dsl_dataset, FTAG); dsl_dataset_rele(os->os_dsl_dataset, FTAG); return (error); } if (dmu_objset_type(os) != DMU_OST_ZFS) { dmu_objset_rele(os, FTAG); return (0); } osname = kmem_alloc(MAXPATHLEN, KM_SLEEP); if (snprintf(osname, MAXPATHLEN, "%s/", name) >= MAXPATHLEN) { dmu_objset_rele(os, FTAG); kmem_free(osname, MAXPATHLEN); return (ENOENT); } p = osname + strlen(osname); len = MAXPATHLEN - (p - osname); #if 0 /* Prefetch the datasets. */ cookie = 0; while (dmu_dir_list_next(os, len, p, NULL, &cookie) == 0) { if (!dataset_name_hidden(osname)) (void) dmu_objset_prefetch(osname, NULL); } #endif cookie = 0; while (dmu_dir_list_next(os, MAXPATHLEN - (p - osname), p, NULL, &cookie) == 0) { dmu_objset_rele(os, FTAG); (void)zvol_create_minors(osname); if ((error = dmu_objset_hold(name, FTAG, &os)) != 0) { printf("ZFS WARNING: Unable to put hold on %s (error=%d).\n", name, error); return (error); } } dmu_objset_rele(os, FTAG); kmem_free(osname, MAXPATHLEN); return (0); } static void zvol_rename_minor(zvol_state_t *zv, const char *newname) { struct g_geom *gp; struct g_provider *pp; struct cdev *dev; ASSERT(MUTEX_HELD(&zfsdev_state_lock)); if (zv->zv_volmode == ZFS_VOLMODE_GEOM) { g_topology_lock(); pp = zv->zv_provider; ASSERT(pp != NULL); gp = pp->geom; ASSERT(gp != NULL); zv->zv_provider = NULL; g_wither_provider(pp, ENXIO); pp = g_new_providerf(gp, "%s/%s", ZVOL_DRIVER, newname); pp->flags |= G_PF_DIRECT_RECEIVE | G_PF_DIRECT_SEND; pp->sectorsize = DEV_BSIZE; pp->mediasize = zv->zv_volsize; pp->private = zv; zv->zv_provider = pp; g_error_provider(pp, 0); g_topology_unlock(); } else if (zv->zv_volmode == ZFS_VOLMODE_DEV) { struct make_dev_args args; if ((dev = zv->zv_dev) != NULL) { zv->zv_dev = NULL; destroy_dev(dev); if (zv->zv_total_opens > 0) { zv->zv_flags &= ~ZVOL_EXCL; zv->zv_total_opens = 0; zvol_last_close(zv); } } make_dev_args_init(&args); args.mda_flags = MAKEDEV_CHECKNAME | MAKEDEV_WAITOK; args.mda_devsw = &zvol_cdevsw; args.mda_cr = NULL; args.mda_uid = UID_ROOT; args.mda_gid = GID_OPERATOR; args.mda_mode = 0640; args.mda_si_drv2 = zv; if (make_dev_s(&args, &zv->zv_dev, "%s/%s", ZVOL_DRIVER, newname) == 0) zv->zv_dev->si_iosize_max = MAXPHYS; } strlcpy(zv->zv_name, newname, sizeof(zv->zv_name)); } void zvol_rename_minors(const char *oldname, const char *newname) { char name[MAXPATHLEN]; struct g_provider *pp; struct g_geom *gp; size_t oldnamelen, newnamelen; zvol_state_t *zv; char *namebuf; boolean_t locked = B_FALSE; oldnamelen = strlen(oldname); newnamelen = strlen(newname); DROP_GIANT(); /* See comment in zvol_open(). */ if (!MUTEX_HELD(&zfsdev_state_lock)) { mutex_enter(&zfsdev_state_lock); locked = B_TRUE; } LIST_FOREACH(zv, &all_zvols, zv_links) { if (strcmp(zv->zv_name, oldname) == 0) { zvol_rename_minor(zv, newname); } else if (strncmp(zv->zv_name, oldname, oldnamelen) == 0 && (zv->zv_name[oldnamelen] == '/' || zv->zv_name[oldnamelen] == '@')) { snprintf(name, sizeof(name), "%s%c%s", newname, zv->zv_name[oldnamelen], zv->zv_name + oldnamelen + 1); zvol_rename_minor(zv, name); } } if (locked) mutex_exit(&zfsdev_state_lock); PICKUP_GIANT(); } static int zvol_d_open(struct cdev *dev, int flags, int fmt, struct thread *td) { zvol_state_t *zv = dev->si_drv2; int err = 0; mutex_enter(&zfsdev_state_lock); if (zv->zv_total_opens == 0) err = zvol_first_open(zv); if (err) { mutex_exit(&zfsdev_state_lock); return (err); } if ((flags & FWRITE) && (zv->zv_flags & ZVOL_RDONLY)) { err = SET_ERROR(EROFS); goto out; } if (zv->zv_flags & ZVOL_EXCL) { err = SET_ERROR(EBUSY); goto out; } #ifdef FEXCL if (flags & FEXCL) { if (zv->zv_total_opens != 0) { err = SET_ERROR(EBUSY); goto out; } zv->zv_flags |= ZVOL_EXCL; } #endif zv->zv_total_opens++; + if (flags & (FSYNC | FDSYNC)) { + zv->zv_sync_cnt++; + if (zv->zv_sync_cnt == 1) + zil_async_to_sync(zv->zv_zilog, ZVOL_OBJ); + } mutex_exit(&zfsdev_state_lock); return (err); out: if (zv->zv_total_opens == 0) zvol_last_close(zv); mutex_exit(&zfsdev_state_lock); return (err); } static int zvol_d_close(struct cdev *dev, int flags, int fmt, struct thread *td) { zvol_state_t *zv = dev->si_drv2; mutex_enter(&zfsdev_state_lock); if (zv->zv_flags & ZVOL_EXCL) { ASSERT(zv->zv_total_opens == 1); zv->zv_flags &= ~ZVOL_EXCL; } /* * If the open count is zero, this is a spurious close. * That indicates a bug in the kernel / DDI framework. */ ASSERT(zv->zv_total_opens != 0); /* * You may get multiple opens, but only one close. */ zv->zv_total_opens--; + if (flags & (FSYNC | FDSYNC)) + zv->zv_sync_cnt--; if (zv->zv_total_opens == 0) zvol_last_close(zv); mutex_exit(&zfsdev_state_lock); return (0); } static int zvol_d_ioctl(struct cdev *dev, u_long cmd, caddr_t data, int fflag, struct thread *td) { zvol_state_t *zv; rl_t *rl; off_t offset, length; int i, error; boolean_t sync; zv = dev->si_drv2; error = 0; KASSERT(zv->zv_total_opens > 0, ("Device with zero access count in zvol_d_ioctl")); i = IOCPARM_LEN(cmd); switch (cmd) { case DIOCGSECTORSIZE: *(u_int *)data = DEV_BSIZE; break; case DIOCGMEDIASIZE: *(off_t *)data = zv->zv_volsize; break; case DIOCGFLUSH: zil_commit(zv->zv_zilog, ZVOL_OBJ); break; case DIOCGDELETE: if (!zvol_unmap_enabled) break; offset = ((off_t *)data)[0]; length = ((off_t *)data)[1]; if ((offset % DEV_BSIZE) != 0 || (length % DEV_BSIZE) != 0 || offset < 0 || offset >= zv->zv_volsize || length <= 0) { printf("%s: offset=%jd length=%jd\n", __func__, offset, length); error = EINVAL; break; } rl = zfs_range_lock(&zv->zv_znode, offset, length, RL_WRITER); dmu_tx_t *tx = dmu_tx_create(zv->zv_objset); error = dmu_tx_assign(tx, TXG_WAIT); if (error != 0) { sync = FALSE; dmu_tx_abort(tx); } else { sync = (zv->zv_objset->os_sync == ZFS_SYNC_ALWAYS); zvol_log_truncate(zv, tx, offset, length, sync); dmu_tx_commit(tx); error = dmu_free_long_range(zv->zv_objset, ZVOL_OBJ, offset, length); } zfs_range_unlock(rl); if (sync) zil_commit(zv->zv_zilog, ZVOL_OBJ); break; case DIOCGSTRIPESIZE: *(off_t *)data = zv->zv_volblocksize; break; case DIOCGSTRIPEOFFSET: *(off_t *)data = 0; break; case DIOCGATTR: { spa_t *spa = dmu_objset_spa(zv->zv_objset); struct diocgattr_arg *arg = (struct diocgattr_arg *)data; uint64_t refd, avail, usedobjs, availobjs; if (strcmp(arg->name, "GEOM::candelete") == 0) arg->value.i = 1; else if (strcmp(arg->name, "blocksavail") == 0) { dmu_objset_space(zv->zv_objset, &refd, &avail, &usedobjs, &availobjs); arg->value.off = avail / DEV_BSIZE; } else if (strcmp(arg->name, "blocksused") == 0) { dmu_objset_space(zv->zv_objset, &refd, &avail, &usedobjs, &availobjs); arg->value.off = refd / DEV_BSIZE; } else if (strcmp(arg->name, "poolblocksavail") == 0) { avail = metaslab_class_get_space(spa_normal_class(spa)); avail -= metaslab_class_get_alloc(spa_normal_class(spa)); arg->value.off = avail / DEV_BSIZE; } else if (strcmp(arg->name, "poolblocksused") == 0) { refd = metaslab_class_get_alloc(spa_normal_class(spa)); arg->value.off = refd / DEV_BSIZE; } else error = ENOIOCTL; break; } case FIOSEEKHOLE: case FIOSEEKDATA: { off_t *off = (off_t *)data; uint64_t noff; boolean_t hole; hole = (cmd == FIOSEEKHOLE); noff = *off; error = dmu_offset_next(zv->zv_objset, ZVOL_OBJ, hole, &noff); *off = noff; break; } default: error = ENOIOCTL; } return (error); } #endif /* illumos */