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 328227) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil.h (revision 328228) @@ -1,447 +1,448 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2017 by Delphix. All rights reserved. * Copyright (c) 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; struct lwb; /* * 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 /* * 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) /* * 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_COMMIT 0 /* Commit marker (no on-disk state) */ #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_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, struct lwb *lwb, 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 void zil_commit_impl(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_lwb_add_block(struct lwb *lwb, const blkptr_t *bp); extern void zil_lwb_add_txg(struct lwb *lwb, uint64_t txg); 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/sys/zil_impl.h =================================================================== --- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil_impl.h (revision 328227) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zil_impl.h (revision 328228) @@ -1,217 +1,242 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2012, 2017 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ /* Portions Copyright 2010 Robert Milkowski */ #ifndef _SYS_ZIL_IMPL_H #define _SYS_ZIL_IMPL_H #include #include #ifdef __cplusplus extern "C" { #endif /* - * Possbile states for a given lwb structure. An lwb will start out in - * the "closed" state, and then transition to the "opened" state via a - * call to zil_lwb_write_open(). After the lwb is "open", it can - * transition into the "issued" state via zil_lwb_write_issue(). After - * the lwb's zio completes, and the vdev's are flushed, the lwb will - * transition into the "done" state via zil_lwb_write_done(), and the - * structure eventually freed. + * Possbile states for a given lwb structure. + * + * An lwb will start out in the "closed" state, and then transition to + * the "opened" state via a call to zil_lwb_write_open(). When + * transitioning from "closed" to "opened" the zilog's "zl_issuer_lock" + * must be held. + * + * After the lwb is "opened", it can transition into the "issued" state + * via zil_lwb_write_issue(). Again, the zilog's "zl_issuer_lock" must + * be held when making this transition. + * + * After the lwb's zio completes, and the vdev's are flushed, the lwb + * will transition into the "done" state via zil_lwb_write_done(). When + * transitioning from "issued" to "done", the zilog's "zl_lock" must be + * held, *not* the "zl_issuer_lock". + * + * The zilog's "zl_issuer_lock" can become heavily contended in certain + * workloads, so we specifically avoid acquiring that lock when + * transitioning an lwb from "issued" to "done". This allows us to avoid + * having to acquire the "zl_issuer_lock" for each lwb ZIO completion, + * which would have added more lock contention on an already heavily + * contended lock. + * + * Additionally, correctness when reading an lwb's state is often + * acheived by exploiting the fact that these state transitions occur in + * this specific order; i.e. "closed" to "opened" to "issued" to "done". + * + * Thus, if an lwb is in the "closed" or "opened" state, holding the + * "zl_issuer_lock" will prevent a concurrent thread from transitioning + * that lwb to the "issued" state. Likewise, if an lwb is already in the + * "issued" state, holding the "zl_lock" will prevent a concurrent + * thread from transitioning that lwb to the "done" state. */ typedef enum { LWB_STATE_CLOSED, LWB_STATE_OPENED, LWB_STATE_ISSUED, LWB_STATE_DONE, LWB_NUM_STATES } lwb_state_t; /* * Log write block (lwb) * * Prior to an lwb being issued to disk via zil_lwb_write_issue(), it * will be protected by the zilog's "zl_issuer_lock". Basically, prior * to it being issued, it will only be accessed by the thread that's * holding the "zl_issuer_lock". After the lwb is issued, the zilog's * "zl_lock" is used to protect the lwb against concurrent access. */ typedef struct lwb { zilog_t *lwb_zilog; /* back pointer to log struct */ blkptr_t lwb_blk; /* on disk address of this log blk */ boolean_t lwb_slog; /* lwb_blk is on SLOG device */ int lwb_nused; /* # used bytes in buffer */ int lwb_sz; /* size of block and buffer */ lwb_state_t lwb_state; /* the state of this lwb */ char *lwb_buf; /* log write buffer */ zio_t *lwb_write_zio; /* zio for the lwb buffer */ zio_t *lwb_root_zio; /* root zio for lwb write and flushes */ dmu_tx_t *lwb_tx; /* tx for log block allocation */ uint64_t lwb_max_txg; /* highest txg in this lwb */ list_node_t lwb_node; /* zilog->zl_lwb_list linkage */ list_t lwb_waiters; /* list of zil_commit_waiter's */ avl_tree_t lwb_vdev_tree; /* vdevs to flush after lwb write */ kmutex_t lwb_vdev_lock; /* protects lwb_vdev_tree */ hrtime_t lwb_issued_timestamp; /* when was the lwb issued? */ } lwb_t; /* * ZIL commit waiter. * * This structure is allocated each time zil_commit() is called, and is * used by zil_commit() to communicate with other parts of the ZIL, such * that zil_commit() can know when it safe for it return. For more * details, see the comment above zil_commit(). * * The "zcw_lock" field is used to protect the commit waiter against * concurrent access. This lock is often acquired while already holding * the zilog's "zl_issuer_lock" or "zl_lock"; see the functions * zil_process_commit_list() and zil_lwb_flush_vdevs_done() as examples * of this. Thus, one must be careful not to acquire the * "zl_issuer_lock" or "zl_lock" when already holding the "zcw_lock"; * e.g. see the zil_commit_waiter_timeout() function. */ typedef struct zil_commit_waiter { kcondvar_t zcw_cv; /* signalled when "done" */ kmutex_t zcw_lock; /* protects fields of this struct */ list_node_t zcw_node; /* linkage in lwb_t:lwb_waiter list */ lwb_t *zcw_lwb; /* back pointer to lwb when linked */ boolean_t zcw_done; /* B_TRUE when "done", else B_FALSE */ int zcw_zio_error; /* contains the zio io_error value */ } zil_commit_waiter_t; /* * Intent log transaction lists */ typedef struct itxs { list_t i_sync_list; /* list of synchronous itxs */ avl_tree_t i_async_tree; /* tree of foids for async itxs */ } itxs_t; typedef struct itxg { kmutex_t itxg_lock; /* lock for this structure */ uint64_t itxg_txg; /* txg for this chain */ itxs_t *itxg_itxs; /* sync and async itxs */ } itxg_t; /* for async nodes we build up an AVL tree of lists of async itxs per file */ typedef struct itx_async_node { uint64_t ia_foid; /* file object id */ list_t ia_list; /* list of async itxs for this foid */ avl_node_t ia_node; /* AVL tree linkage */ } itx_async_node_t; /* * Vdev flushing: during a zil_commit(), we build up an AVL tree of the vdevs * we've touched so we know which ones need a write cache flush at the end. */ typedef struct zil_vdev_node { uint64_t zv_vdev; /* vdev to be flushed */ avl_node_t zv_node; /* AVL tree linkage */ } zil_vdev_node_t; #define ZIL_PREV_BLKS 16 /* * Stable storage intent log management structure. One per dataset. */ struct zilog { kmutex_t zl_lock; /* protects most zilog_t fields */ struct dsl_pool *zl_dmu_pool; /* DSL pool */ spa_t *zl_spa; /* handle for read/write log */ const zil_header_t *zl_header; /* log header buffer */ objset_t *zl_os; /* object set we're logging */ zil_get_data_t *zl_get_data; /* callback to get object content */ lwb_t *zl_last_lwb_opened; /* most recent lwb opened */ hrtime_t zl_last_lwb_latency; /* zio latency of last lwb done */ uint64_t zl_lr_seq; /* on-disk log record sequence number */ uint64_t zl_commit_lr_seq; /* last committed on-disk lr seq */ uint64_t zl_destroy_txg; /* txg of last zil_destroy() */ uint64_t zl_replayed_seq[TXG_SIZE]; /* last replayed rec seq */ uint64_t zl_replaying_seq; /* current replay seq number */ uint32_t zl_suspend; /* log suspend count */ kcondvar_t zl_cv_suspend; /* log suspend completion */ uint8_t zl_suspending; /* log is currently suspending */ uint8_t zl_keep_first; /* keep first log block in destroy */ uint8_t zl_replay; /* replaying records while set */ uint8_t zl_stop_sync; /* for debugging */ kmutex_t zl_issuer_lock; /* single writer, per ZIL, at a time */ uint8_t zl_logbias; /* latency or throughput */ uint8_t zl_sync; /* synchronous or asynchronous */ int zl_parse_error; /* last zil_parse() error */ uint64_t zl_parse_blk_seq; /* highest blk seq on last parse */ uint64_t zl_parse_lr_seq; /* highest lr seq on last parse */ uint64_t zl_parse_blk_count; /* number of blocks parsed */ uint64_t zl_parse_lr_count; /* number of log records parsed */ itxg_t zl_itxg[TXG_SIZE]; /* intent log txg chains */ list_t zl_itx_commit_list; /* itx list to be committed */ uint64_t zl_cur_used; /* current commit log size used */ list_t zl_lwb_list; /* in-flight log write list */ avl_tree_t zl_bp_tree; /* track bps during log parse */ clock_t zl_replay_time; /* lbolt of when replay started */ uint64_t zl_replay_blks; /* number of log blocks replayed */ zil_header_t zl_old_header; /* debugging aid */ uint_t zl_prev_blks[ZIL_PREV_BLKS]; /* size - sector rounded */ uint_t zl_prev_rotor; /* rotor for zl_prev[] */ txg_node_t zl_dirty_link; /* protected by dp_dirty_zilogs list */ uint64_t zl_dirty_max_txg; /* highest txg used to dirty zilog */ }; typedef struct zil_bp_node { dva_t zn_dva; avl_node_t zn_node; } zil_bp_node_t; /* * Maximum amount of write data that can be put into single log block. */ #define ZIL_MAX_LOG_DATA (SPA_OLD_MAXBLOCKSIZE - sizeof (zil_chain_t) - \ sizeof (lr_write_t)) #define ZIL_MAX_COPIED_DATA \ ((SPA_OLD_MAXBLOCKSIZE - sizeof (zil_chain_t)) / 2 - sizeof (lr_write_t)) /* * Maximum amount of log space we agree to waste to reduce number of * WR_NEED_COPY chunks to reduce zl_get_data() overhead (~12%). */ #define ZIL_MAX_WASTE_SPACE (ZIL_MAX_LOG_DATA / 8) /* * Maximum amount of write data for WR_COPIED. Fall back to WR_NEED_COPY * as more space efficient if we can't fit at least two log records into * maximum sized log block. */ #define ZIL_MAX_COPIED_DATA ((SPA_OLD_MAXBLOCKSIZE - \ sizeof (zil_chain_t)) / 2 - sizeof (lr_write_t)) #ifdef __cplusplus } #endif #endif /* _SYS_ZIL_IMPL_H */ Index: head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio.h =================================================================== --- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio.h (revision 328227) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/sys/zio.h (revision 328228) @@ -1,653 +1,652 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2012, 2017 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright 2016 Toomas Soome */ #ifndef _ZIO_H #define _ZIO_H #include #include #include #include #include #include #include #include #ifdef __cplusplus extern "C" { #endif /* * Embedded checksum */ #define ZEC_MAGIC 0x210da7ab10c7a11ULL typedef struct zio_eck { uint64_t zec_magic; /* for validation, endianness */ zio_cksum_t zec_cksum; /* 256-bit checksum */ } zio_eck_t; /* * Gang block headers are self-checksumming and contain an array * of block pointers. */ #define SPA_GANGBLOCKSIZE SPA_MINBLOCKSIZE #define SPA_GBH_NBLKPTRS ((SPA_GANGBLOCKSIZE - \ sizeof (zio_eck_t)) / sizeof (blkptr_t)) #define SPA_GBH_FILLER ((SPA_GANGBLOCKSIZE - \ sizeof (zio_eck_t) - \ (SPA_GBH_NBLKPTRS * sizeof (blkptr_t))) /\ sizeof (uint64_t)) typedef struct zio_gbh { blkptr_t zg_blkptr[SPA_GBH_NBLKPTRS]; uint64_t zg_filler[SPA_GBH_FILLER]; zio_eck_t zg_tail; } zio_gbh_phys_t; enum zio_checksum { ZIO_CHECKSUM_INHERIT = 0, ZIO_CHECKSUM_ON, ZIO_CHECKSUM_OFF, ZIO_CHECKSUM_LABEL, ZIO_CHECKSUM_GANG_HEADER, ZIO_CHECKSUM_ZILOG, ZIO_CHECKSUM_FLETCHER_2, ZIO_CHECKSUM_FLETCHER_4, ZIO_CHECKSUM_SHA256, ZIO_CHECKSUM_ZILOG2, ZIO_CHECKSUM_NOPARITY, ZIO_CHECKSUM_SHA512, ZIO_CHECKSUM_SKEIN, #ifdef illumos ZIO_CHECKSUM_EDONR, #endif ZIO_CHECKSUM_FUNCTIONS }; /* * The number of "legacy" compression functions which can be set on individual * objects. */ #define ZIO_CHECKSUM_LEGACY_FUNCTIONS ZIO_CHECKSUM_ZILOG2 #define ZIO_CHECKSUM_ON_VALUE ZIO_CHECKSUM_FLETCHER_4 #define ZIO_CHECKSUM_DEFAULT ZIO_CHECKSUM_ON #define ZIO_CHECKSUM_MASK 0xffULL #define ZIO_CHECKSUM_VERIFY (1 << 8) #define ZIO_DEDUPCHECKSUM ZIO_CHECKSUM_SHA256 #define ZIO_DEDUPDITTO_MIN 100 /* * The number of "legacy" compression functions which can be set on individual * objects. */ #define ZIO_COMPRESS_LEGACY_FUNCTIONS ZIO_COMPRESS_LZ4 /* * The meaning of "compress = on" selected by the compression features enabled * on a given pool. */ #define ZIO_COMPRESS_LEGACY_ON_VALUE ZIO_COMPRESS_LZJB #define ZIO_COMPRESS_LZ4_ON_VALUE ZIO_COMPRESS_LZ4 #define ZIO_COMPRESS_DEFAULT ZIO_COMPRESS_OFF #define BOOTFS_COMPRESS_VALID(compress) \ ((compress) == ZIO_COMPRESS_LZJB || \ (compress) == ZIO_COMPRESS_LZ4 || \ (compress) == ZIO_COMPRESS_ON || \ (compress) == ZIO_COMPRESS_OFF) #define ZIO_FAILURE_MODE_WAIT 0 #define ZIO_FAILURE_MODE_CONTINUE 1 #define ZIO_FAILURE_MODE_PANIC 2 enum zio_flag { /* * Flags inherited by gang, ddt, and vdev children, * and that must be equal for two zios to aggregate */ ZIO_FLAG_DONT_AGGREGATE = 1 << 0, ZIO_FLAG_IO_REPAIR = 1 << 1, ZIO_FLAG_SELF_HEAL = 1 << 2, ZIO_FLAG_RESILVER = 1 << 3, ZIO_FLAG_SCRUB = 1 << 4, ZIO_FLAG_SCAN_THREAD = 1 << 5, ZIO_FLAG_PHYSICAL = 1 << 6, #define ZIO_FLAG_AGG_INHERIT (ZIO_FLAG_CANFAIL - 1) /* * Flags inherited by ddt, gang, and vdev children. */ ZIO_FLAG_CANFAIL = 1 << 7, /* must be first for INHERIT */ ZIO_FLAG_SPECULATIVE = 1 << 8, ZIO_FLAG_CONFIG_WRITER = 1 << 9, ZIO_FLAG_DONT_RETRY = 1 << 10, ZIO_FLAG_DONT_CACHE = 1 << 11, ZIO_FLAG_NODATA = 1 << 12, ZIO_FLAG_INDUCE_DAMAGE = 1 << 13, ZIO_FLAG_IO_ALLOCATING = 1 << 14, #define ZIO_FLAG_DDT_INHERIT (ZIO_FLAG_IO_RETRY - 1) #define ZIO_FLAG_GANG_INHERIT (ZIO_FLAG_IO_RETRY - 1) /* * Flags inherited by vdev children. */ ZIO_FLAG_IO_RETRY = 1 << 15, /* must be first for INHERIT */ ZIO_FLAG_PROBE = 1 << 16, ZIO_FLAG_TRYHARD = 1 << 17, ZIO_FLAG_OPTIONAL = 1 << 18, #define ZIO_FLAG_VDEV_INHERIT (ZIO_FLAG_DONT_QUEUE - 1) /* * Flags not inherited by any children. */ ZIO_FLAG_DONT_QUEUE = 1 << 19, /* must be first for INHERIT */ ZIO_FLAG_DONT_PROPAGATE = 1 << 20, ZIO_FLAG_IO_BYPASS = 1 << 21, ZIO_FLAG_IO_REWRITE = 1 << 22, ZIO_FLAG_RAW = 1 << 23, ZIO_FLAG_GANG_CHILD = 1 << 24, ZIO_FLAG_DDT_CHILD = 1 << 25, ZIO_FLAG_GODFATHER = 1 << 26, ZIO_FLAG_NOPWRITE = 1 << 27, ZIO_FLAG_REEXECUTED = 1 << 28, ZIO_FLAG_DELEGATED = 1 << 29, }; #define ZIO_FLAG_MUSTSUCCEED 0 #define ZIO_DDT_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_DDT_INHERIT) | \ ZIO_FLAG_DDT_CHILD | ZIO_FLAG_CANFAIL) #define ZIO_GANG_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_GANG_INHERIT) | \ ZIO_FLAG_GANG_CHILD | ZIO_FLAG_CANFAIL) #define ZIO_VDEV_CHILD_FLAGS(zio) \ (((zio)->io_flags & ZIO_FLAG_VDEV_INHERIT) | \ ZIO_FLAG_CANFAIL) enum zio_child { ZIO_CHILD_VDEV = 0, ZIO_CHILD_GANG, ZIO_CHILD_DDT, ZIO_CHILD_LOGICAL, ZIO_CHILD_TYPES }; enum zio_wait_type { ZIO_WAIT_READY = 0, ZIO_WAIT_DONE, ZIO_WAIT_TYPES }; /* * We'll take the number 122 and 123 to indicate checksum errors and * fragmentation. Those doesn't collide with any errno values as they * are greater than ELAST. */ #define ECKSUM 122 #define EFRAGS 123 typedef void zio_done_func_t(zio_t *zio); extern boolean_t zio_dva_throttle_enabled; extern const char *zio_type_name[ZIO_TYPES]; /* * A bookmark is a four-tuple that uniquely * identifies any block in the pool. By convention, the meta-objset (MOS) * is objset 0, and the meta-dnode is object 0. This covers all blocks * except root blocks and ZIL blocks, which are defined as follows: * * Root blocks (objset_phys_t) are object 0, level -1: . * ZIL blocks are bookmarked . * dmu_sync()ed ZIL data blocks are bookmarked . * dnode visit bookmarks are . * * Note: this structure is called a bookmark because its original purpose * was to remember where to resume a pool-wide traverse. * * Note: this structure is passed between userland and the kernel, and is * stored on disk (by virtue of being incorporated into other on-disk * structures, e.g. dsl_scan_phys_t). */ typedef struct zbookmark_phys { uint64_t zb_objset; uint64_t zb_object; int64_t zb_level; uint64_t zb_blkid; } zbookmark_phys_t; #define SET_BOOKMARK(zb, objset, object, level, blkid) \ { \ (zb)->zb_objset = objset; \ (zb)->zb_object = object; \ (zb)->zb_level = level; \ (zb)->zb_blkid = blkid; \ } #define ZB_DESTROYED_OBJSET (-1ULL) #define ZB_ROOT_OBJECT (0ULL) #define ZB_ROOT_LEVEL (-1LL) #define ZB_ROOT_BLKID (0ULL) #define ZB_ZIL_OBJECT (0ULL) #define ZB_ZIL_LEVEL (-2LL) #define ZB_DNODE_LEVEL (-3LL) #define ZB_DNODE_BLKID (0ULL) #define ZB_IS_ZERO(zb) \ ((zb)->zb_objset == 0 && (zb)->zb_object == 0 && \ (zb)->zb_level == 0 && (zb)->zb_blkid == 0) #define ZB_IS_ROOT(zb) \ ((zb)->zb_object == ZB_ROOT_OBJECT && \ (zb)->zb_level == ZB_ROOT_LEVEL && \ (zb)->zb_blkid == ZB_ROOT_BLKID) typedef struct zio_prop { enum zio_checksum zp_checksum; enum zio_compress zp_compress; dmu_object_type_t zp_type; uint8_t zp_level; uint8_t zp_copies; boolean_t zp_dedup; boolean_t zp_dedup_verify; boolean_t zp_nopwrite; } zio_prop_t; typedef struct zio_cksum_report zio_cksum_report_t; typedef void zio_cksum_finish_f(zio_cksum_report_t *rep, const void *good_data); typedef void zio_cksum_free_f(void *cbdata, size_t size); struct zio_bad_cksum; /* defined in zio_checksum.h */ struct dnode_phys; struct abd; struct zio_cksum_report { struct zio_cksum_report *zcr_next; nvlist_t *zcr_ereport; nvlist_t *zcr_detector; void *zcr_cbdata; size_t zcr_cbinfo; /* passed to zcr_free() */ uint64_t zcr_align; uint64_t zcr_length; zio_cksum_finish_f *zcr_finish; zio_cksum_free_f *zcr_free; /* internal use only */ struct zio_bad_cksum *zcr_ckinfo; /* information from failure */ }; typedef void zio_vsd_cksum_report_f(zio_t *zio, zio_cksum_report_t *zcr, void *arg); zio_vsd_cksum_report_f zio_vsd_default_cksum_report; typedef struct zio_vsd_ops { zio_done_func_t *vsd_free; zio_vsd_cksum_report_f *vsd_cksum_report; } zio_vsd_ops_t; typedef struct zio_gang_node { zio_gbh_phys_t *gn_gbh; struct zio_gang_node *gn_child[SPA_GBH_NBLKPTRS]; } zio_gang_node_t; typedef zio_t *zio_gang_issue_func_t(zio_t *zio, blkptr_t *bp, zio_gang_node_t *gn, struct abd *data, uint64_t offset); typedef void zio_transform_func_t(zio_t *zio, struct abd *data, uint64_t size); typedef struct zio_transform { struct abd *zt_orig_abd; uint64_t zt_orig_size; uint64_t zt_bufsize; zio_transform_func_t *zt_transform; struct zio_transform *zt_next; } zio_transform_t; typedef int zio_pipe_stage_t(zio_t *zio); /* * The io_reexecute flags are distinct from io_flags because the child must * be able to propagate them to the parent. The normal io_flags are local * to the zio, not protected by any lock, and not modifiable by children; * the reexecute flags are protected by io_lock, modifiable by children, * and always propagated -- even when ZIO_FLAG_DONT_PROPAGATE is set. */ #define ZIO_REEXECUTE_NOW 0x01 #define ZIO_REEXECUTE_SUSPEND 0x02 typedef struct zio_alloc_list { list_t zal_list; uint64_t zal_size; } zio_alloc_list_t; typedef struct zio_link { zio_t *zl_parent; zio_t *zl_child; list_node_t zl_parent_node; list_node_t zl_child_node; } zio_link_t; /* * Used for TRIM kstat. */ typedef struct zio_trim_stats { /* * Number of bytes successfully TRIMmed. */ kstat_named_t bytes; /* * Number of successful TRIM requests. */ kstat_named_t success; /* * Number of TRIM requests that failed because TRIM is not * supported. */ kstat_named_t unsupported; /* * Number of TRIM requests that failed for other reasons. */ kstat_named_t failed; } zio_trim_stats_t; extern zio_trim_stats_t zio_trim_stats; #define ZIO_TRIM_STAT_INCR(stat, val) \ atomic_add_64(&zio_trim_stats.stat.value.ui64, (val)); #define ZIO_TRIM_STAT_BUMP(stat) \ ZIO_TRIM_STAT_INCR(stat, 1); struct zio { /* Core information about this I/O */ zbookmark_phys_t io_bookmark; zio_prop_t io_prop; zio_type_t io_type; enum zio_child io_child_type; int io_cmd; zio_priority_t io_priority; uint8_t io_reexecute; uint8_t io_state[ZIO_WAIT_TYPES]; uint64_t io_txg; spa_t *io_spa; blkptr_t *io_bp; blkptr_t *io_bp_override; blkptr_t io_bp_copy; list_t io_parent_list; list_t io_child_list; zio_t *io_logical; zio_transform_t *io_transform_stack; /* Callback info */ zio_done_func_t *io_ready; zio_done_func_t *io_children_ready; zio_done_func_t *io_physdone; zio_done_func_t *io_done; void *io_private; int64_t io_prev_space_delta; /* DMU private */ blkptr_t io_bp_orig; /* Data represented by this I/O */ struct abd *io_abd; struct abd *io_orig_abd; uint64_t io_size; uint64_t io_orig_size; /* io_lsize != io_orig_size iff this is a raw write */ uint64_t io_lsize; /* Stuff for the vdev stack */ vdev_t *io_vd; void *io_vsd; const zio_vsd_ops_t *io_vsd_ops; uint64_t io_offset; hrtime_t io_timestamp; hrtime_t io_queued_timestamp; hrtime_t io_target_timestamp; avl_node_t io_queue_node; avl_node_t io_offset_node; avl_node_t io_alloc_node; zio_alloc_list_t io_alloc_list; #ifdef __FreeBSD__ struct bio *io_bio; #endif /* Internal pipeline state */ enum zio_flag io_flags; enum zio_stage io_stage; enum zio_stage io_pipeline; enum zio_flag io_orig_flags; enum zio_stage io_orig_stage; enum zio_stage io_orig_pipeline; enum zio_stage io_pipeline_trace; int io_error; int io_child_error[ZIO_CHILD_TYPES]; uint64_t io_children[ZIO_CHILD_TYPES][ZIO_WAIT_TYPES]; uint64_t io_child_count; uint64_t io_phys_children; uint64_t io_parent_count; uint64_t *io_stall; zio_t *io_gang_leader; zio_gang_node_t *io_gang_tree; void *io_executor; void *io_waiter; kmutex_t io_lock; kcondvar_t io_cv; /* FMA state */ zio_cksum_report_t *io_cksum_report; uint64_t io_ena; /* Taskq dispatching state */ taskq_ent_t io_tqent; avl_node_t io_trim_node; list_node_t io_trim_link; }; extern int zio_bookmark_compare(const void *, const void *); extern zio_t *zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done, void *priv, enum zio_flag flags); extern zio_t *zio_root(spa_t *spa, zio_done_func_t *done, void *priv, enum zio_flag flags); extern zio_t *zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, struct abd *data, uint64_t lsize, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb); extern zio_t *zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, struct abd *data, uint64_t size, uint64_t psize, const zio_prop_t *zp, zio_done_func_t *ready, zio_done_func_t *children_ready, zio_done_func_t *physdone, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb); extern zio_t *zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, struct abd *data, uint64_t size, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb); extern void zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite); extern void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp); extern zio_t *zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_done_func_t *done, void *priv, enum zio_flag flags); extern zio_t *zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd, uint64_t offset, uint64_t size, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags); extern zio_t *zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, struct abd *data, int checksum, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, boolean_t labels); extern zio_t *zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, struct abd *data, int checksum, zio_done_func_t *done, void *priv, zio_priority_t priority, enum zio_flag flags, boolean_t labels); extern zio_t *zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, uint64_t size, enum zio_flag flags); extern int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, blkptr_t *old_bp, uint64_t size, boolean_t *slog); extern void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp); extern void zio_flush(zio_t *zio, vdev_t *vd); extern zio_t *zio_trim(zio_t *zio, spa_t *spa, vdev_t *vd, uint64_t offset, uint64_t size); extern void zio_shrink(zio_t *zio, uint64_t size); extern int zio_wait(zio_t *zio); extern void zio_nowait(zio_t *zio); extern void zio_execute(zio_t *zio); extern void zio_interrupt(zio_t *zio); extern void zio_delay_init(zio_t *zio); extern void zio_delay_interrupt(zio_t *zio); extern zio_t *zio_walk_parents(zio_t *cio, zio_link_t **); extern zio_t *zio_walk_children(zio_t *pio, zio_link_t **); extern zio_t *zio_unique_parent(zio_t *cio); extern void zio_add_child(zio_t *pio, zio_t *cio); extern void *zio_buf_alloc(size_t size); extern void zio_buf_free(void *buf, size_t size); extern void *zio_data_buf_alloc(size_t size); extern void zio_data_buf_free(void *buf, size_t size); extern void zio_push_transform(zio_t *zio, struct abd *abd, uint64_t size, uint64_t bufsize, zio_transform_func_t *transform); extern void zio_pop_transforms(zio_t *zio); extern void zio_resubmit_stage_async(void *); extern zio_t *zio_vdev_child_io(zio_t *zio, blkptr_t *bp, vdev_t *vd, uint64_t offset, struct abd *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *priv); extern zio_t *zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, struct abd *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *priv); extern void zio_vdev_io_bypass(zio_t *zio); extern void zio_vdev_io_reissue(zio_t *zio); extern void zio_vdev_io_redone(zio_t *zio); extern void zio_checksum_verified(zio_t *zio); extern int zio_worst_error(int e1, int e2); extern enum zio_checksum zio_checksum_select(enum zio_checksum child, enum zio_checksum parent); extern enum zio_checksum zio_checksum_dedup_select(spa_t *spa, enum zio_checksum child, enum zio_checksum parent); extern enum zio_compress zio_compress_select(spa_t *spa, enum zio_compress child, enum zio_compress parent); -extern void zio_cancel(zio_t *zio); extern void zio_suspend(spa_t *spa, zio_t *zio); extern int zio_resume(spa_t *spa); extern void zio_resume_wait(spa_t *spa); /* * Initial setup and teardown. */ extern void zio_init(void); extern void zio_fini(void); /* * Fault injection */ struct zinject_record; extern uint32_t zio_injection_enabled; extern int zio_inject_fault(char *name, int flags, int *id, struct zinject_record *record); extern int zio_inject_list_next(int *id, char *name, size_t buflen, struct zinject_record *record); extern int zio_clear_fault(int id); extern void zio_handle_panic_injection(spa_t *spa, char *tag, uint64_t type); extern int zio_handle_fault_injection(zio_t *zio, int error); extern int zio_handle_device_injection(vdev_t *vd, zio_t *zio, int error); extern int zio_handle_label_injection(zio_t *zio, int error); extern void zio_handle_ignored_writes(zio_t *zio); extern hrtime_t zio_handle_io_delay(zio_t *zio); /* * Checksum ereport functions */ extern void zfs_ereport_start_checksum(spa_t *spa, vdev_t *vd, struct zio *zio, uint64_t offset, uint64_t length, void *arg, struct zio_bad_cksum *info); extern void zfs_ereport_finish_checksum(zio_cksum_report_t *report, const void *good_data, const void *bad_data, boolean_t drop_if_identical); extern void zfs_ereport_send_interim_checksum(zio_cksum_report_t *report); extern void zfs_ereport_free_checksum(zio_cksum_report_t *report); /* If we have the good data in hand, this function can be used */ extern void zfs_ereport_post_checksum(spa_t *spa, vdev_t *vd, struct zio *zio, uint64_t offset, uint64_t length, const void *good_data, const void *bad_data, struct zio_bad_cksum *info); /* Called from spa_sync(), but primarily an injection handler */ extern void spa_handle_ignored_writes(spa_t *spa); /* zbookmark_phys functions */ boolean_t zbookmark_subtree_completed(const struct dnode_phys *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block); int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2); #ifdef __cplusplus } #endif #endif /* _ZIO_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 328227) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zil.c (revision 328228) @@ -1,3213 +1,3258 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2017 by Delphix. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ /* Portions Copyright 2010 Robert Milkowski */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include /* * The ZFS Intent Log (ZIL) saves "transaction records" (itxs) of system * calls that change the file system. Each itx has enough information to * be able to replay them after a system crash, power loss, or * equivalent failure mode. These are stored in memory until either: * * 1. they are committed to the pool by the DMU transaction group * (txg), at which point they can be discarded; or * 2. they are committed to the on-disk ZIL for the dataset being * modified (e.g. due to an fsync, O_DSYNC, or other synchronous * requirement). * * In the event of a crash or power loss, the itxs contained by each * dataset's on-disk ZIL will be replayed when that dataset is first * instantianted (e.g. if the dataset is a normal fileystem, when it is * first mounted). * * As hinted at above, there is one ZIL per dataset (both the in-memory * representation, and the on-disk representation). The on-disk format * consists of 3 parts: * * - a single, per-dataset, ZIL header; which points to a chain of * - zero or more ZIL blocks; each of which contains * - zero or more ZIL records * * A ZIL record holds the information necessary to replay a single * system call transaction. A ZIL block can hold many ZIL records, and * the blocks are chained together, similarly to a singly linked list. * * Each ZIL block contains a block pointer (blkptr_t) to the next ZIL * block in the chain, and the ZIL header points to the first block in * the chain. * * Note, there is not a fixed place in the pool to hold these ZIL * blocks; they are dynamically allocated and freed as needed from the * blocks available on the pool, though they can be preferentially * allocated from a dedicated "log" vdev. */ /* * This controls the amount of time that a ZIL block (lwb) will remain * "open" when it isn't "full", and it has a thread waiting for it to be * committed to stable storage. Please refer to the zil_commit_waiter() * function (and the comments within it) for more details. */ int zfs_commit_timeout_pct = 5; /* * 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"); /* * Limit SLOG write size per commit executed with synchronous priority. * Any writes above that will be executed with lower (asynchronous) priority * to limit potential SLOG device abuse by single active ZIL writer. */ uint64_t zil_slog_bulk = 768 * 1024; SYSCTL_QUAD(_vfs_zfs, OID_AUTO, zil_slog_bulk, CTLFLAG_RWTUN, &zil_slog_bulk, 0, "Maximal SLOG commit size with sync priority"); static kmem_cache_t *zil_lwb_cache; static kmem_cache_t *zil_zcw_cache; #define LWB_EMPTY(lwb) ((BP_GET_LSIZE(&lwb->lwb_blk) - \ sizeof (zil_chain_t)) == (lwb->lwb_sz - lwb->lwb_nused)) 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 int zil_lwb_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); } static lwb_t * zil_alloc_lwb(zilog_t *zilog, blkptr_t *bp, boolean_t slog, 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_slog = slog; lwb->lwb_state = LWB_STATE_CLOSED; lwb->lwb_buf = zio_buf_alloc(BP_GET_LSIZE(bp)); lwb->lwb_max_txg = txg; lwb->lwb_write_zio = NULL; lwb->lwb_root_zio = NULL; lwb->lwb_tx = NULL; lwb->lwb_issued_timestamp = 0; 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); ASSERT(!MUTEX_HELD(&lwb->lwb_vdev_lock)); ASSERT(avl_is_empty(&lwb->lwb_vdev_tree)); - ASSERT(list_is_empty(&lwb->lwb_waiters)); + VERIFY(list_is_empty(&lwb->lwb_waiters)); return (lwb); } static void zil_free_lwb(zilog_t *zilog, lwb_t *lwb) { ASSERT(MUTEX_HELD(&zilog->zl_lock)); ASSERT(!MUTEX_HELD(&lwb->lwb_vdev_lock)); - ASSERT(list_is_empty(&lwb->lwb_waiters)); - - if (lwb->lwb_state == LWB_STATE_OPENED) { - avl_tree_t *t = &lwb->lwb_vdev_tree; - void *cookie = NULL; - zil_vdev_node_t *zv; - - while ((zv = avl_destroy_nodes(t, &cookie)) != NULL) - kmem_free(zv, sizeof (*zv)); - - ASSERT3P(lwb->lwb_root_zio, !=, NULL); - ASSERT3P(lwb->lwb_write_zio, !=, NULL); - - zio_cancel(lwb->lwb_root_zio); - zio_cancel(lwb->lwb_write_zio); - - lwb->lwb_root_zio = NULL; - lwb->lwb_write_zio = NULL; - } else { - ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED); - } - + VERIFY(list_is_empty(&lwb->lwb_waiters)); ASSERT(avl_is_empty(&lwb->lwb_vdev_tree)); ASSERT3P(lwb->lwb_write_zio, ==, NULL); ASSERT3P(lwb->lwb_root_zio, ==, NULL); + ASSERT3U(lwb->lwb_max_txg, <=, spa_syncing_txg(zilog->zl_spa)); + ASSERT(lwb->lwb_state == LWB_STATE_CLOSED || + lwb->lwb_state == LWB_STATE_DONE); /* * Clear the zilog's field to indicate this lwb is no longer * valid, and prevent use-after-free errors. */ if (zilog->zl_last_lwb_opened == lwb) zilog->zl_last_lwb_opened = NULL; kmem_cache_free(zil_lwb_cache, 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); ASSERT(spa_writeable(zilog->zl_spa)); 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); zilog->zl_dirty_max_txg = MAX(txg, zilog->zl_dirty_max_txg); } } /* * Determine if the zil is dirty in the specified txg. Callers wanting to * ensure that the dirty state does not change must hold the itxg_lock for * the specified txg. Holding the lock will ensure that the zil cannot be * dirtied (zil_itx_assign) or cleaned (zil_clean) while we check its current * state. */ boolean_t zilog_is_dirty_in_txg(zilog_t *zilog, uint64_t txg) { dsl_pool_t *dp = zilog->zl_dmu_pool; if (txg_list_member(&dp->dp_dirty_zilogs, zilog, txg & TXG_MASK)) return (B_TRUE); return (B_FALSE); } /* * Determine if the zil is dirty. The zil is considered dirty if it has * any pending itx records that have not been cleaned by zil_clean(). */ 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; boolean_t slog = FALSE; /* * 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); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); 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, &slog); if (error == 0) zil_init_log_chain(zilog, &blk); } /* * Allocate a log write block (lwb) for the first log block. */ if (error == 0) lwb = zil_alloc_lwb(zilog, &blk, slog, 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); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); 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(zilog->zl_spa, txg, &lwb->lwb_blk); zil_free_lwb(zilog, 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); } /* * When an itx is "skipped", this function is used to properly mark the * waiter as "done, and signal any thread(s) waiting on it. An itx can * be skipped (and not committed to an lwb) for a variety of reasons, * one of them being that the itx was committed via spa_sync(), prior to * it being committed to an lwb; this can happen if a thread calling * zil_commit() is racing with spa_sync(). */ static void zil_commit_waiter_skip(zil_commit_waiter_t *zcw) { mutex_enter(&zcw->zcw_lock); ASSERT3B(zcw->zcw_done, ==, B_FALSE); zcw->zcw_done = B_TRUE; cv_broadcast(&zcw->zcw_cv); mutex_exit(&zcw->zcw_lock); } /* * This function is used when the given waiter is to be linked into an * lwb's "lwb_waiter" list; i.e. when the itx is committed to the lwb. * At this point, the waiter will no longer be referenced by the itx, * and instead, will be referenced by the lwb. */ static void zil_commit_waiter_link_lwb(zil_commit_waiter_t *zcw, lwb_t *lwb) { + /* + * The lwb_waiters field of the lwb is protected by the zilog's + * zl_lock, thus it must be held when calling this function. + */ + ASSERT(MUTEX_HELD(&lwb->lwb_zilog->zl_lock)); + mutex_enter(&zcw->zcw_lock); ASSERT(!list_link_active(&zcw->zcw_node)); ASSERT3P(zcw->zcw_lwb, ==, NULL); ASSERT3P(lwb, !=, NULL); ASSERT(lwb->lwb_state == LWB_STATE_OPENED || lwb->lwb_state == LWB_STATE_ISSUED); list_insert_tail(&lwb->lwb_waiters, zcw); zcw->zcw_lwb = lwb; mutex_exit(&zcw->zcw_lock); } /* * This function is used when zio_alloc_zil() fails to allocate a ZIL * block, and the given waiter must be linked to the "nolwb waiters" * list inside of zil_process_commit_list(). */ static void zil_commit_waiter_link_nolwb(zil_commit_waiter_t *zcw, list_t *nolwb) { mutex_enter(&zcw->zcw_lock); ASSERT(!list_link_active(&zcw->zcw_node)); ASSERT3P(zcw->zcw_lwb, ==, NULL); list_insert_tail(nolwb, zcw); mutex_exit(&zcw->zcw_lock); } void zil_lwb_add_block(lwb_t *lwb, const blkptr_t *bp) { avl_tree_t *t = &lwb->lwb_vdev_tree; avl_index_t where; zil_vdev_node_t *zv, zvsearch; int ndvas = BP_GET_NDVAS(bp); int i; if (zfs_nocacheflush) return; mutex_enter(&lwb->lwb_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(&lwb->lwb_vdev_lock); } void zil_lwb_add_txg(lwb_t *lwb, uint64_t txg) { lwb->lwb_max_txg = MAX(lwb->lwb_max_txg, txg); } /* * This function is a called after all VDEVs associated with a given lwb * write have completed their DKIOCFLUSHWRITECACHE command; or as soon * as the lwb write completes, if "zfs_nocacheflush" is set. * * The intention is for this function to be called as soon as the * contents of an lwb are considered "stable" on disk, and will survive * any sudden loss of power. At this point, any threads waiting for the * lwb to reach this state are signalled, and the "waiter" structures * are marked "done". */ static void zil_lwb_flush_vdevs_done(zio_t *zio) { lwb_t *lwb = zio->io_private; zilog_t *zilog = lwb->lwb_zilog; dmu_tx_t *tx = lwb->lwb_tx; zil_commit_waiter_t *zcw; spa_config_exit(zilog->zl_spa, SCL_STATE, lwb); zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); mutex_enter(&zilog->zl_lock); /* * 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 zil_sync() to remove the lwb so * that it's not picked up as the next new one in * zil_process_commit_list(). zil_sync() will only remove the * lwb if lwb_buf is null. */ lwb->lwb_buf = NULL; lwb->lwb_tx = NULL; ASSERT3U(lwb->lwb_issued_timestamp, >, 0); zilog->zl_last_lwb_latency = gethrtime() - lwb->lwb_issued_timestamp; lwb->lwb_root_zio = NULL; lwb->lwb_state = LWB_STATE_DONE; if (zilog->zl_last_lwb_opened == lwb) { /* * 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. */ zilog->zl_commit_lr_seq = zilog->zl_lr_seq; } while ((zcw = list_head(&lwb->lwb_waiters)) != NULL) { mutex_enter(&zcw->zcw_lock); ASSERT(list_link_active(&zcw->zcw_node)); list_remove(&lwb->lwb_waiters, zcw); ASSERT3P(zcw->zcw_lwb, ==, lwb); zcw->zcw_lwb = NULL; zcw->zcw_zio_error = zio->io_error; ASSERT3B(zcw->zcw_done, ==, B_FALSE); zcw->zcw_done = B_TRUE; cv_broadcast(&zcw->zcw_cv); mutex_exit(&zcw->zcw_lock); } 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); } /* * This is called when an lwb write completes. This means, this specific * lwb was written to disk, and all dependent lwb have also been * written to disk. * * At this point, a DKIOCFLUSHWRITECACHE command hasn't been issued to * the VDEVs involved in writing out this specific lwb. The lwb will be * "done" once zil_lwb_flush_vdevs_done() is called, which occurs in the * zio completion callback for the lwb's root zio. */ static void zil_lwb_write_done(zio_t *zio) { lwb_t *lwb = zio->io_private; spa_t *spa = zio->io_spa; zilog_t *zilog = lwb->lwb_zilog; avl_tree_t *t = &lwb->lwb_vdev_tree; void *cookie = NULL; zil_vdev_node_t *zv; ASSERT3S(spa_config_held(spa, SCL_STATE, RW_READER), !=, 0); 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); abd_put(zio->io_abd); ASSERT3S(lwb->lwb_state, ==, LWB_STATE_ISSUED); mutex_enter(&zilog->zl_lock); lwb->lwb_write_zio = NULL; mutex_exit(&zilog->zl_lock); if (avl_numnodes(t) == 0) return; /* * If there was an IO error, we're not going to call zio_flush() * on these vdevs, so we simply empty the tree and free the * nodes. We avoid calling zio_flush() since there isn't any * good reason for doing so, after the lwb block failed to be * written out. */ if (zio->io_error != 0) { while ((zv = avl_destroy_nodes(t, &cookie)) != NULL) kmem_free(zv, sizeof (*zv)); return; } while ((zv = avl_destroy_nodes(t, &cookie)) != NULL) { vdev_t *vd = vdev_lookup_top(spa, zv->zv_vdev); if (vd != NULL) zio_flush(lwb->lwb_root_zio, vd); kmem_free(zv, sizeof (*zv)); } } /* * This function's purpose is to "open" an lwb such that it is ready to * accept new itxs being committed to it. To do this, the lwb's zio * structures are created, and linked to the lwb. This function is * idempotent; if the passed in lwb has already been opened, this * function is essentially a no-op. */ static void zil_lwb_write_open(zilog_t *zilog, lwb_t *lwb) { zbookmark_phys_t zb; zio_priority_t prio; ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); ASSERT3P(lwb, !=, NULL); EQUIV(lwb->lwb_root_zio == NULL, lwb->lwb_state == LWB_STATE_CLOSED); EQUIV(lwb->lwb_root_zio != NULL, lwb->lwb_state == LWB_STATE_OPENED); 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 (lwb->lwb_root_zio == NULL) { abd_t *lwb_abd = abd_get_from_buf(lwb->lwb_buf, BP_GET_LSIZE(&lwb->lwb_blk)); if (!lwb->lwb_slog || zilog->zl_cur_used <= zil_slog_bulk) prio = ZIO_PRIORITY_SYNC_WRITE; else prio = ZIO_PRIORITY_ASYNC_WRITE; lwb->lwb_root_zio = zio_root(zilog->zl_spa, zil_lwb_flush_vdevs_done, lwb, ZIO_FLAG_CANFAIL); ASSERT3P(lwb->lwb_root_zio, !=, NULL); lwb->lwb_write_zio = zio_rewrite(lwb->lwb_root_zio, zilog->zl_spa, 0, &lwb->lwb_blk, lwb_abd, BP_GET_LSIZE(&lwb->lwb_blk), zil_lwb_write_done, lwb, prio, ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE, &zb); ASSERT3P(lwb->lwb_write_zio, !=, NULL); lwb->lwb_state = LWB_STATE_OPENED; mutex_enter(&zilog->zl_lock); /* * The zilog's "zl_last_lwb_opened" field is used to * build the lwb/zio dependency chain, which is used to * preserve the ordering of lwb completions that is * required by the semantics of the ZIL. Each new lwb * zio becomes a parent of the "previous" lwb zio, such * that the new lwb's zio cannot complete until the * "previous" lwb's zio completes. * * This is required by the semantics of zil_commit(); * the commit waiters attached to the lwbs will be woken * in the lwb zio's completion callback, so this zio * dependency graph ensures the waiters are woken in the * correct order (the same order the lwbs were created). */ lwb_t *last_lwb_opened = zilog->zl_last_lwb_opened; if (last_lwb_opened != NULL && last_lwb_opened->lwb_state != LWB_STATE_DONE) { ASSERT(last_lwb_opened->lwb_state == LWB_STATE_OPENED || last_lwb_opened->lwb_state == LWB_STATE_ISSUED); ASSERT3P(last_lwb_opened->lwb_root_zio, !=, NULL); zio_add_child(lwb->lwb_root_zio, last_lwb_opened->lwb_root_zio); } zilog->zl_last_lwb_opened = lwb; mutex_exit(&zilog->zl_lock); } ASSERT3P(lwb->lwb_root_zio, !=, NULL); ASSERT3P(lwb->lwb_write_zio, !=, NULL); ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED); } /* * 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 }; /* * Start a log block write and advance to the next log block. * Calls are serialized. */ static lwb_t * zil_lwb_write_issue(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; boolean_t slog; ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); ASSERT3P(lwb->lwb_root_zio, !=, NULL); ASSERT3P(lwb->lwb_write_zio, !=, NULL); ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED); 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); /* * Since we are not going to create any new dirty data and we can even * help with clearing the existing dirty data, we should not be subject * to the dirty data based delays. * We (ab)use TXG_WAITED to bypass the delay mechanism. * One side effect from using TXG_WAITED is that dmu_tx_assign() can * fail if the pool is suspended. Those are dramatic circumstances, * so we return NULL to signal that the normal ZIL processing is not * possible and txg_wait_synced() should be used to ensure that the data * is on disk. */ error = dmu_tx_assign(tx, TXG_WAITED); if (error != 0) { ASSERT3S(error, ==, EIO); dmu_tx_abort(tx); return (NULL); } 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, &slog); 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 block (lwb). */ nlwb = zil_alloc_lwb(zilog, bp, slog, txg); } 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_write_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); spa_config_enter(zilog->zl_spa, SCL_STATE, lwb, RW_READER); zil_lwb_add_block(lwb, &lwb->lwb_blk); lwb->lwb_issued_timestamp = gethrtime(); lwb->lwb_state = LWB_STATE_ISSUED; zio_nowait(lwb->lwb_root_zio); zio_nowait(lwb->lwb_write_zio); /* * 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 *lrcb, *lrc; lr_write_t *lrwb, *lrw; char *lr_buf; uint64_t dlen, dnow, lwb_sp, reclen, txg; ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); ASSERT3P(lwb, !=, NULL); ASSERT3P(lwb->lwb_buf, !=, NULL); zil_lwb_write_open(zilog, lwb); lrc = &itx->itx_lr; lrw = (lr_write_t *)lrc; /* * A commit itx doesn't represent any on-disk state; instead * it's simply used as a place holder on the commit list, and * provides a mechanism for attaching a "commit waiter" onto the * correct lwb (such that the waiter can be signalled upon * completion of that lwb). Thus, we don't process this itx's * log record if it's a commit itx (these itx's don't have log * records), and instead link the itx's waiter onto the lwb's * list of waiters. * * For more details, see the comment above zil_commit(). */ if (lrc->lrc_txtype == TX_COMMIT) { + mutex_enter(&zilog->zl_lock); zil_commit_waiter_link_lwb(itx->itx_private, lwb); itx->itx_private = NULL; + mutex_exit(&zilog->zl_lock); return (lwb); } if (lrc->lrc_txtype == TX_WRITE && itx->itx_wr_state == WR_NEED_COPY) { dlen = P2ROUNDUP_TYPED( lrw->lr_length, sizeof (uint64_t), uint64_t); } else { dlen = 0; } reclen = lrc->lrc_reclen; zilog->zl_cur_used += (reclen + dlen); txg = lrc->lrc_txg; ASSERT3U(zilog->zl_cur_used, <, UINT64_MAX - (reclen + dlen)); cont: /* * If this record won't fit in the current log block, start a new one. * For WR_NEED_COPY optimize layout for minimal number of chunks. */ lwb_sp = lwb->lwb_sz - lwb->lwb_nused; if (reclen > lwb_sp || (reclen + dlen > lwb_sp && lwb_sp < ZIL_MAX_WASTE_SPACE && (dlen % ZIL_MAX_LOG_DATA == 0 || lwb_sp < reclen + dlen % ZIL_MAX_LOG_DATA))) { lwb = zil_lwb_write_issue(zilog, lwb); if (lwb == NULL) return (NULL); zil_lwb_write_open(zilog, lwb); ASSERT(LWB_EMPTY(lwb)); lwb_sp = lwb->lwb_sz - lwb->lwb_nused; ASSERT3U(reclen + MIN(dlen, sizeof (uint64_t)), <=, lwb_sp); } dnow = MIN(dlen, lwb_sp - reclen); lr_buf = lwb->lwb_buf + lwb->lwb_nused; bcopy(lrc, lr_buf, reclen); lrcb = (lr_t *)lr_buf; /* Like lrc, but inside lwb. */ lrwb = (lr_write_t *)lrcb; /* Like lrw, but inside lwb. */ /* * 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 (itx->itx_wr_state == WR_NEED_COPY) { dbuf = lr_buf + reclen; lrcb->lrc_reclen += dnow; if (lrwb->lr_length > dnow) lrwb->lr_length = dnow; lrw->lr_offset += dnow; lrw->lr_length -= dnow; } else { ASSERT(itx->itx_wr_state == WR_INDIRECT); dbuf = NULL; } /* * We pass in the "lwb_write_zio" rather than * "lwb_root_zio" so that the "lwb_write_zio" * becomes the parent of any zio's created by * the "zl_get_data" callback. The vdevs are * flushed after the "lwb_write_zio" completes, * so we want to make sure that completion * callback waits for these additional zio's, * such that the vdevs used by those zio's will * be included in the lwb's vdev tree, and those * vdevs will be properly flushed. If we passed * in "lwb_root_zio" here, then these additional * vdevs may not be flushed; e.g. if these zio's * completed after "lwb_write_zio" completed. */ error = zilog->zl_get_data(itx->itx_private, lrwb, dbuf, lwb, lwb->lwb_write_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. */ lrcb->lrc_seq = ++zilog->zl_lr_seq; lwb->lwb_nused += reclen + dnow; zil_lwb_add_txg(lwb, txg); ASSERT3U(lwb->lwb_nused, <=, lwb->lwb_sz); ASSERT0(P2PHASE(lwb->lwb_nused, sizeof (uint64_t))); dlen -= dnow; if (dlen > 0) { zilog->zl_cur_used += reclen; goto cont; } 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); 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_lr.lrc_seq = 0; /* defensive */ itx->itx_sync = B_TRUE; /* default is synchronous */ return (itx); } void zil_itx_destroy(itx_t *itx) { 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) { /* * In the general case, commit itxs will not be found * here, as they'll be committed to an lwb via * zil_lwb_commit(), and free'd in that function. Having * said that, it is still possible for commit itxs to be * found here, due to the following race: * * - a thread calls zil_commit() which assigns the * commit itx to a per-txg i_sync_list * - zil_itxg_clean() is called (e.g. via spa_sync()) * while the waiter is still on the i_sync_list * * There's nothing to prevent syncing the txg while the * waiter is on the i_sync_list. This normally doesn't * happen because spa_sync() is slower than zil_commit(), * but if zil_commit() calls txg_wait_synced() (e.g. * because zil_create() or zil_commit_writer_stall() is * called) we will hit this case. */ if (itx->itx_lr.lrc_txtype == TX_COMMIT) zil_commit_waiter_skip(itx->itx_private); 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); /* commit itxs should never be on the async lists. */ ASSERT3U(itx->itx_lr.lrc_txtype, !=, TX_COMMIT); 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); /* commit itxs should never be on the async lists. */ ASSERT3U(itx->itx_lr.lrc_txtype, !=, TX_COMMIT); 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. */ zfs_dbgmsg("zil_itx_assign: missed itx cleanup for " "txg %llu", itxg->itxg_txg); clean = itxg->itxg_itxs; } 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); } 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); /* * We don't want to dirty the ZIL using ZILTEST_TXG, because * zil_clean() will never be called using ZILTEST_TXG. Thus, we * need to be careful to always dirty the ZIL using the "real" * TXG (not itxg_txg) even when the SPA is frozen. */ zilog_dirty(zilog, dmu_tx_get_txg(tx)); 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; ASSERT3U(synced_txg, <, ZILTEST_TXG); 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); ASSERT3U(itxg->itxg_txg, !=, 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. */ ASSERT3P(zilog->zl_dmu_pool, !=, NULL); ASSERT3P(zilog->zl_dmu_pool->dp_zil_clean_taskq, !=, NULL); if (taskq_dispatch(zilog->zl_dmu_pool->dp_zil_clean_taskq, (void (*)(void *))zil_itxg_clean, clean_me, TQ_NOSLEEP) == 0) zil_itxg_clean(clean_me); } /* * This function will traverse the queue of itxs that need to be * committed, and move them onto the ZIL's 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; ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); if (spa_freeze_txg(zilog->zl_spa) != UINT64_MAX) /* ziltest support */ otxg = ZILTEST_TXG; else otxg = spa_last_synced_txg(zilog->zl_spa) + 1; /* * This is inherently racy, since there is nothing to prevent * the last synced txg from changing. That's okay since we'll * only commit things in the future. */ 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 we're adding itx records to the zl_itx_commit_list, * then the zil better be dirty in this "txg". We can assert * that here since we're holding the itxg_lock which will * prevent spa_sync from cleaning it. Once we add the itxs * to the zl_itx_commit_list we must commit it to disk even * if it's unnecessary (i.e. the txg was synced). */ ASSERT(zilog_is_dirty_in_txg(zilog, txg) || spa_freeze_txg(zilog->zl_spa) != UINT64_MAX); list_move_tail(commit_list, &itxg->itxg_itxs->i_sync_list); mutex_exit(&itxg->itxg_lock); } } /* * Move the async itxs for a specified object to commit into sync lists. */ 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; /* * This is inherently racy, since there is nothing to prevent * the last synced txg from changing. */ 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); } } /* * This function will prune commit itxs that are at the head of the * commit list (it won't prune past the first non-commit itx), and * either: a) attach them to the last lwb that's still pending * completion, or b) skip them altogether. * * This is used as a performance optimization to prevent commit itxs * from generating new lwbs when it's unnecessary to do so. */ static void zil_prune_commit_list(zilog_t *zilog) { itx_t *itx; ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); while (itx = list_head(&zilog->zl_itx_commit_list)) { lr_t *lrc = &itx->itx_lr; if (lrc->lrc_txtype != TX_COMMIT) break; mutex_enter(&zilog->zl_lock); lwb_t *last_lwb = zilog->zl_last_lwb_opened; if (last_lwb == NULL || last_lwb->lwb_state == LWB_STATE_DONE) { /* * All of the itxs this waiter was waiting on * must have already completed (or there were * never any itx's for it to wait on), so it's * safe to skip this waiter and mark it done. */ zil_commit_waiter_skip(itx->itx_private); } else { zil_commit_waiter_link_lwb(itx->itx_private, last_lwb); itx->itx_private = NULL; } mutex_exit(&zilog->zl_lock); list_remove(&zilog->zl_itx_commit_list, itx); zil_itx_destroy(itx); } IMPLY(itx != NULL, itx->itx_lr.lrc_txtype != TX_COMMIT); } static void zil_commit_writer_stall(zilog_t *zilog) { /* * When zio_alloc_zil() fails to allocate the next lwb block on * disk, we must call txg_wait_synced() to ensure all of the * lwbs in the zilog's zl_lwb_list are synced and then freed (in * zil_sync()), such that any subsequent ZIL writer (i.e. a call * to zil_process_commit_list()) will have to call zil_create(), * and start a new ZIL chain. * * Since zil_alloc_zil() failed, the lwb that was previously * issued does not have a pointer to the "next" lwb on disk. * Thus, if another ZIL writer thread was to allocate the "next" * on-disk lwb, that block could be leaked in the event of a * crash (because the previous lwb on-disk would not point to * it). * * We must hold the zilog's zl_issuer_lock while we do this, to * ensure no new threads enter zil_process_commit_list() until * all lwb's in the zl_lwb_list have been synced and freed * (which is achieved via the txg_wait_synced() call). */ ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); txg_wait_synced(zilog->zl_dmu_pool, 0); ASSERT3P(list_tail(&zilog->zl_lwb_list), ==, NULL); } /* * This function will traverse the commit list, creating new lwbs as * needed, and committing the itxs from the commit list to these newly * created lwbs. Additionally, as a new lwb is created, the previous * lwb will be issued to the zio layer to be written to disk. */ static void zil_process_commit_list(zilog_t *zilog) { spa_t *spa = zilog->zl_spa; list_t nolwb_waiters; lwb_t *lwb; itx_t *itx; ASSERT(MUTEX_HELD(&zilog->zl_issuer_lock)); /* * 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) return; list_create(&nolwb_waiters, sizeof (zil_commit_waiter_t), offsetof(zil_commit_waiter_t, zcw_node)); lwb = list_tail(&zilog->zl_lwb_list); if (lwb == NULL) { lwb = zil_create(zilog); } else { ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED); ASSERT3S(lwb->lwb_state, !=, LWB_STATE_DONE); } while (itx = list_head(&zilog->zl_itx_commit_list)) { lr_t *lrc = &itx->itx_lr; uint64_t txg = lrc->lrc_txg; ASSERT3U(txg, !=, 0); if (lrc->lrc_txtype == TX_COMMIT) { DTRACE_PROBE2(zil__process__commit__itx, zilog_t *, zilog, itx_t *, itx); } else { DTRACE_PROBE2(zil__process__normal__itx, zilog_t *, zilog, itx_t *, itx); } - /* - * This is inherently racy and may result in us writing - * out a log block for a txg that was just synced. This - * is ok since we'll end cleaning up that log block the - * next time we call zil_sync(). - */ boolean_t synced = txg <= spa_last_synced_txg(spa); boolean_t frozen = txg > spa_freeze_txg(spa); - if (!synced || frozen) { + /* + * If the txg of this itx has already been synced out, then + * we don't need to commit this itx to an lwb. This is + * because the data of this itx will have already been + * written to the main pool. This is inherently racy, and + * it's still ok to commit an itx whose txg has already + * been synced; this will result in a write that's + * unnecessary, but will do no harm. + * + * With that said, we always want to commit TX_COMMIT itxs + * to an lwb, regardless of whether or not that itx's txg + * has been synced out. We do this to ensure any OPENED lwb + * will always have at least one zil_commit_waiter_t linked + * to the lwb. + * + * As a counter-example, if we skipped TX_COMMIT itx's + * whose txg had already been synced, the following + * situation could occur if we happened to be racing with + * spa_sync: + * + * 1. we commit a non-TX_COMMIT itx to an lwb, where the + * itx's txg is 10 and the last synced txg is 9. + * 2. spa_sync finishes syncing out txg 10. + * 3. we move to the next itx in the list, it's a TX_COMMIT + * whose txg is 10, so we skip it rather than committing + * it to the lwb used in (1). + * + * If the itx that is skipped in (3) is the last TX_COMMIT + * itx in the commit list, than it's possible for the lwb + * used in (1) to remain in the OPENED state indefinitely. + * + * To prevent the above scenario from occuring, ensuring + * that once an lwb is OPENED it will transition to ISSUED + * and eventually DONE, we always commit TX_COMMIT itx's to + * an lwb here, even if that itx's txg has already been + * synced. + * + * Finally, if the pool is frozen, we _always_ commit the + * itx. The point of freezing the pool is to prevent data + * from being written to the main pool via spa_sync, and + * instead rely solely on the ZIL to persistently store the + * data; i.e. when the pool is frozen, the last synced txg + * value can't be trusted. + */ + if (frozen || !synced || lrc->lrc_txtype == TX_COMMIT) { if (lwb != NULL) { lwb = zil_lwb_commit(zilog, itx, lwb); } else if (lrc->lrc_txtype == TX_COMMIT) { ASSERT3P(lwb, ==, NULL); zil_commit_waiter_link_nolwb( itx->itx_private, &nolwb_waiters); } - } else if (lrc->lrc_txtype == TX_COMMIT) { - ASSERT3B(synced, ==, B_TRUE); - ASSERT3B(frozen, ==, B_FALSE); - - /* - * If this is a commit itx, then there will be a - * thread that is either: already waiting for - * it, or soon will be waiting. - * - * This itx has already been committed to disk - * via spa_sync() so we don't bother committing - * it to an lwb. As a result, we cannot use the - * lwb zio callback to signal the waiter and - * mark it as done, so we must do that here. - */ - zil_commit_waiter_skip(itx->itx_private); } list_remove(&zilog->zl_itx_commit_list, itx); zil_itx_destroy(itx); } if (lwb == NULL) { /* * This indicates zio_alloc_zil() failed to allocate the * "next" lwb on-disk. When this happens, we must stall * the ZIL write pipeline; see the comment within * zil_commit_writer_stall() for more details. */ zil_commit_writer_stall(zilog); /* * Additionally, we have to signal and mark the "nolwb" * waiters as "done" here, since without an lwb, we * can't do this via zil_lwb_flush_vdevs_done() like * normal. */ zil_commit_waiter_t *zcw; while (zcw = list_head(&nolwb_waiters)) { zil_commit_waiter_skip(zcw); list_remove(&nolwb_waiters, zcw); } } else { ASSERT(list_is_empty(&nolwb_waiters)); ASSERT3P(lwb, !=, NULL); ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED); ASSERT3S(lwb->lwb_state, !=, LWB_STATE_DONE); /* * At this point, the ZIL block pointed at by the "lwb" * variable is in one of the following states: "closed" * or "open". * * If its "closed", then no itxs have been committed to * it, so there's no point in issuing its zio (i.e. * it's "empty"). * * If its "open" state, then it contains one or more * itxs that eventually need to be committed to stable * storage. In this case we intentionally do not issue * the lwb's zio to disk yet, and instead rely on one of * the following two mechanisms for issuing the zio: * * 1. Ideally, there will be more ZIL activity occuring * on the system, such that this function will be * immediately called again (not necessarily by the same * thread) and this lwb's zio will be issued via * zil_lwb_commit(). This way, the lwb is guaranteed to * be "full" when it is issued to disk, and we'll make * use of the lwb's size the best we can. * * 2. If there isn't sufficient ZIL activity occuring on * the system, such that this lwb's zio isn't issued via * zil_lwb_commit(), zil_commit_waiter() will issue the * lwb's zio. If this occurs, the lwb is not guaranteed * to be "full" by the time its zio is issued, and means * the size of the lwb was "too large" given the amount * of ZIL activity occuring on the system at that time. * * We do this for a couple of reasons: * * 1. To try and reduce the number of IOPs needed to * write the same number of itxs. If an lwb has space * available in it's buffer for more itxs, and more itxs * will be committed relatively soon (relative to the * latency of performing a write), then it's beneficial * to wait for these "next" itxs. This way, more itxs * can be committed to stable storage with fewer writes. * * 2. To try and use the largest lwb block size that the * incoming rate of itxs can support. Again, this is to * try and pack as many itxs into as few lwbs as * possible, without significantly impacting the latency * of each individual itx. */ } } /* * This function is responsible for ensuring the passed in commit waiter * (and associated commit itx) is committed to an lwb. If the waiter is * not already committed to an lwb, all itxs in the zilog's queue of * itxs will be processed. The assumption is the passed in waiter's * commit itx will found in the queue just like the other non-commit * itxs, such that when the entire queue is processed, the waiter will * have been commited to an lwb. * * The lwb associated with the passed in waiter is not guaranteed to * have been issued by the time this function completes. If the lwb is * not issued, we rely on future calls to zil_commit_writer() to issue * the lwb, or the timeout mechanism found in zil_commit_waiter(). */ static void zil_commit_writer(zilog_t *zilog, zil_commit_waiter_t *zcw) { ASSERT(!MUTEX_HELD(&zilog->zl_lock)); ASSERT(spa_writeable(zilog->zl_spa)); - ASSERT0(zilog->zl_suspend); mutex_enter(&zilog->zl_issuer_lock); if (zcw->zcw_lwb != NULL || zcw->zcw_done) { /* * It's possible that, while we were waiting to acquire * the "zl_issuer_lock", another thread committed this * waiter to an lwb. If that occurs, we bail out early, * without processing any of the zilog's queue of itxs. * * On certain workloads and system configurations, the * "zl_issuer_lock" can become highly contended. In an * attempt to reduce this contention, we immediately drop * the lock if the waiter has already been processed. * * We've measured this optimization to reduce CPU spent * contending on this lock by up to 5%, using a system * with 32 CPUs, low latency storage (~50 usec writes), * and 1024 threads performing sync writes. */ goto out; } zil_get_commit_list(zilog); zil_prune_commit_list(zilog); zil_process_commit_list(zilog); out: mutex_exit(&zilog->zl_issuer_lock); } static void zil_commit_waiter_timeout(zilog_t *zilog, zil_commit_waiter_t *zcw) { ASSERT(!MUTEX_HELD(&zilog->zl_issuer_lock)); ASSERT(MUTEX_HELD(&zcw->zcw_lock)); ASSERT3B(zcw->zcw_done, ==, B_FALSE); lwb_t *lwb = zcw->zcw_lwb; ASSERT3P(lwb, !=, NULL); ASSERT3S(lwb->lwb_state, !=, LWB_STATE_CLOSED); /* * If the lwb has already been issued by another thread, we can * immediately return since there's no work to be done (the * point of this function is to issue the lwb). Additionally, we * do this prior to acquiring the zl_issuer_lock, to avoid * acquiring it when it's not necessary to do so. */ if (lwb->lwb_state == LWB_STATE_ISSUED || lwb->lwb_state == LWB_STATE_DONE) return; /* * In order to call zil_lwb_write_issue() we must hold the * zilog's "zl_issuer_lock". We can't simply acquire that lock, * since we're already holding the commit waiter's "zcw_lock", * and those two locks are aquired in the opposite order * elsewhere. */ mutex_exit(&zcw->zcw_lock); mutex_enter(&zilog->zl_issuer_lock); mutex_enter(&zcw->zcw_lock); /* * Since we just dropped and re-acquired the commit waiter's * lock, we have to re-check to see if the waiter was marked * "done" during that process. If the waiter was marked "done", * the "lwb" pointer is no longer valid (it can be free'd after * the waiter is marked "done"), so without this check we could * wind up with a use-after-free error below. */ if (zcw->zcw_done) goto out; ASSERT3P(lwb, ==, zcw->zcw_lwb); /* - * We've already checked this above, but since we hadn't - * acquired the zilog's zl_issuer_lock, we have to perform this - * check a second time while holding the lock. We can't call + * We've already checked this above, but since we hadn't acquired + * the zilog's zl_issuer_lock, we have to perform this check a + * second time while holding the lock. + * + * We don't need to hold the zl_lock since the lwb cannot transition + * from OPENED to ISSUED while we hold the zl_issuer_lock. The lwb + * _can_ transition from ISSUED to DONE, but it's OK to race with + * that transition since we treat the lwb the same, whether it's in + * the ISSUED or DONE states. + * + * The important thing, is we treat the lwb differently depending on + * if it's ISSUED or OPENED, and block any other threads that might + * attempt to issue this lwb. For that reason we hold the + * zl_issuer_lock when checking the lwb_state; we must not call * zil_lwb_write_issue() if the lwb had already been issued. + * + * See the comment above the lwb_state_t structure definition for + * more details on the lwb states, and locking requirements. */ if (lwb->lwb_state == LWB_STATE_ISSUED || lwb->lwb_state == LWB_STATE_DONE) goto out; ASSERT3S(lwb->lwb_state, ==, LWB_STATE_OPENED); /* * As described in the comments above zil_commit_waiter() and * zil_process_commit_list(), we need to issue this lwb's zio * since we've reached the commit waiter's timeout and it still * hasn't been issued. */ lwb_t *nlwb = zil_lwb_write_issue(zilog, lwb); ASSERT3S(lwb->lwb_state, !=, LWB_STATE_OPENED); /* * Since the lwb's zio hadn't been issued by the time this thread * reached its timeout, we reset the zilog's "zl_cur_used" field * to influence the zil block size selection algorithm. * * By having to issue the lwb's zio here, it means the size of the * lwb was too large, given the incoming throughput of itxs. By * setting "zl_cur_used" to zero, we communicate this fact to the * block size selection algorithm, so it can take this informaiton * into account, and potentially select a smaller size for the * next lwb block that is allocated. */ zilog->zl_cur_used = 0; if (nlwb == NULL) { /* * When zil_lwb_write_issue() returns NULL, this * indicates zio_alloc_zil() failed to allocate the * "next" lwb on-disk. When this occurs, the ZIL write * pipeline must be stalled; see the comment within the * zil_commit_writer_stall() function for more details. * * We must drop the commit waiter's lock prior to * calling zil_commit_writer_stall() or else we can wind * up with the following deadlock: * * - This thread is waiting for the txg to sync while * holding the waiter's lock; txg_wait_synced() is * used within txg_commit_writer_stall(). * * - The txg can't sync because it is waiting for this * lwb's zio callback to call dmu_tx_commit(). * * - The lwb's zio callback can't call dmu_tx_commit() * because it's blocked trying to acquire the waiter's * lock, which occurs prior to calling dmu_tx_commit() */ mutex_exit(&zcw->zcw_lock); zil_commit_writer_stall(zilog); mutex_enter(&zcw->zcw_lock); } out: mutex_exit(&zilog->zl_issuer_lock); ASSERT(MUTEX_HELD(&zcw->zcw_lock)); } /* * This function is responsible for performing the following two tasks: * * 1. its primary responsibility is to block until the given "commit * waiter" is considered "done". * * 2. its secondary responsibility is to issue the zio for the lwb that * the given "commit waiter" is waiting on, if this function has * waited "long enough" and the lwb is still in the "open" state. * * Given a sufficient amount of itxs being generated and written using * the ZIL, the lwb's zio will be issued via the zil_lwb_commit() * function. If this does not occur, this secondary responsibility will * ensure the lwb is issued even if there is not other synchronous * activity on the system. * * For more details, see zil_process_commit_list(); more specifically, * the comment at the bottom of that function. */ static void zil_commit_waiter(zilog_t *zilog, zil_commit_waiter_t *zcw) { ASSERT(!MUTEX_HELD(&zilog->zl_lock)); ASSERT(!MUTEX_HELD(&zilog->zl_issuer_lock)); ASSERT(spa_writeable(zilog->zl_spa)); - ASSERT0(zilog->zl_suspend); mutex_enter(&zcw->zcw_lock); /* * The timeout is scaled based on the lwb latency to avoid * significantly impacting the latency of each individual itx. * For more details, see the comment at the bottom of the * zil_process_commit_list() function. */ int pct = MAX(zfs_commit_timeout_pct, 1); #if defined(illumos) || !defined(_KERNEL) hrtime_t sleep = (zilog->zl_last_lwb_latency * pct) / 100; hrtime_t wakeup = gethrtime() + sleep; #else sbintime_t sleep = nstosbt((zilog->zl_last_lwb_latency * pct) / 100); sbintime_t wakeup = getsbinuptime() + sleep; #endif boolean_t timedout = B_FALSE; while (!zcw->zcw_done) { ASSERT(MUTEX_HELD(&zcw->zcw_lock)); lwb_t *lwb = zcw->zcw_lwb; /* * Usually, the waiter will have a non-NULL lwb field here, * but it's possible for it to be NULL as a result of * zil_commit() racing with spa_sync(). * * When zil_clean() is called, it's possible for the itxg * list (which may be cleaned via a taskq) to contain * commit itxs. When this occurs, the commit waiters linked * off of these commit itxs will not be committed to an * lwb. Additionally, these commit waiters will not be * marked done until zil_commit_waiter_skip() is called via * zil_itxg_clean(). * * Thus, it's possible for this commit waiter (i.e. the * "zcw" variable) to be found in this "in between" state; * where it's "zcw_lwb" field is NULL, and it hasn't yet * been skipped, so it's "zcw_done" field is still B_FALSE. */ IMPLY(lwb != NULL, lwb->lwb_state != LWB_STATE_CLOSED); if (lwb != NULL && lwb->lwb_state == LWB_STATE_OPENED) { ASSERT3B(timedout, ==, B_FALSE); /* * If the lwb hasn't been issued yet, then we * need to wait with a timeout, in case this * function needs to issue the lwb after the * timeout is reached; responsibility (2) from * the comment above this function. */ #if defined(illumos) || !defined(_KERNEL) clock_t timeleft = cv_timedwait_hires(&zcw->zcw_cv, &zcw->zcw_lock, wakeup, USEC2NSEC(1), CALLOUT_FLAG_ABSOLUTE); if (timeleft >= 0 || zcw->zcw_done) continue; #else int wait_err = cv_timedwait_sbt(&zcw->zcw_cv, &zcw->zcw_lock, wakeup, SBT_1NS, C_ABSOLUTE); if (wait_err != EWOULDBLOCK || zcw->zcw_done) continue; #endif timedout = B_TRUE; zil_commit_waiter_timeout(zilog, zcw); if (!zcw->zcw_done) { /* * If the commit waiter has already been * marked "done", it's possible for the * waiter's lwb structure to have already * been freed. Thus, we can only reliably * make these assertions if the waiter * isn't done. */ ASSERT3P(lwb, ==, zcw->zcw_lwb); ASSERT3S(lwb->lwb_state, !=, LWB_STATE_OPENED); } } else { /* * If the lwb isn't open, then it must have already * been issued. In that case, there's no need to * use a timeout when waiting for the lwb to * complete. * * Additionally, if the lwb is NULL, the waiter * will soon be signalled and marked done via * zil_clean() and zil_itxg_clean(), so no timeout * is required. */ IMPLY(lwb != NULL, lwb->lwb_state == LWB_STATE_ISSUED || lwb->lwb_state == LWB_STATE_DONE); cv_wait(&zcw->zcw_cv, &zcw->zcw_lock); } } mutex_exit(&zcw->zcw_lock); } static zil_commit_waiter_t * zil_alloc_commit_waiter() { zil_commit_waiter_t *zcw = kmem_cache_alloc(zil_zcw_cache, KM_SLEEP); cv_init(&zcw->zcw_cv, NULL, CV_DEFAULT, NULL); mutex_init(&zcw->zcw_lock, NULL, MUTEX_DEFAULT, NULL); list_link_init(&zcw->zcw_node); zcw->zcw_lwb = NULL; zcw->zcw_done = B_FALSE; zcw->zcw_zio_error = 0; return (zcw); } static void zil_free_commit_waiter(zil_commit_waiter_t *zcw) { ASSERT(!list_link_active(&zcw->zcw_node)); ASSERT3P(zcw->zcw_lwb, ==, NULL); ASSERT3B(zcw->zcw_done, ==, B_TRUE); mutex_destroy(&zcw->zcw_lock); cv_destroy(&zcw->zcw_cv); kmem_cache_free(zil_zcw_cache, zcw); } /* * This function is used to create a TX_COMMIT itx and assign it. This * way, it will be linked into the ZIL's list of synchronous itxs, and * then later committed to an lwb (or skipped) when * zil_process_commit_list() is called. */ static void zil_commit_itx_assign(zilog_t *zilog, zil_commit_waiter_t *zcw) { dmu_tx_t *tx = dmu_tx_create(zilog->zl_os); VERIFY0(dmu_tx_assign(tx, TXG_WAIT)); itx_t *itx = zil_itx_create(TX_COMMIT, sizeof (lr_t)); itx->itx_sync = B_TRUE; itx->itx_private = zcw; zil_itx_assign(zilog, itx, tx); dmu_tx_commit(tx); } /* * Commit ZFS Intent Log transactions (itxs) to stable storage. * * When writing ZIL transactions to the on-disk representation of the * ZIL, the itxs are committed to a Log Write Block (lwb). Multiple * itxs can be committed to a single lwb. Once a lwb is written and * committed to stable storage (i.e. the lwb is written, and vdevs have * been flushed), each itx that was committed to that lwb is also * considered to be committed to stable storage. * * When an itx is committed to an lwb, the log record (lr_t) contained * by the itx is copied into the lwb's zio buffer, and once this buffer * is written to disk, it becomes an on-disk ZIL block. * * As itxs are generated, they're inserted into the ZIL's queue of * uncommitted itxs. The semantics of zil_commit() are such that it will * block until all itxs that were in the queue when it was called, are * committed to stable storage. * * If "foid" is zero, this means all "synchronous" and "asynchronous" * itxs, for all objects in the dataset, will be committed to stable * storage prior to zil_commit() returning. If "foid" is non-zero, all * "synchronous" itxs for all objects, but only "asynchronous" itxs * that correspond to the foid passed in, will be committed to stable * storage prior to zil_commit() returning. * * Generally speaking, when zil_commit() is called, the consumer doesn't * actually care about _all_ of the uncommitted itxs. Instead, they're * simply trying to waiting for a specific itx to be committed to disk, * but the interface(s) for interacting with the ZIL don't allow such * fine-grained communication. A better interface would allow a consumer * to create and assign an itx, and then pass a reference to this itx to * zil_commit(); such that zil_commit() would return as soon as that * specific itx was committed to disk (instead of waiting for _all_ * itxs to be committed). * * When a thread calls zil_commit() a special "commit itx" will be * generated, along with a corresponding "waiter" for this commit itx. * zil_commit() will wait on this waiter's CV, such that when the waiter * is marked done, and signalled, zil_commit() will return. * * This commit itx is inserted into the queue of uncommitted itxs. This * provides an easy mechanism for determining which itxs were in the * queue prior to zil_commit() having been called, and which itxs were * added after zil_commit() was called. * * The commit it is special; it doesn't have any on-disk representation. * When a commit itx is "committed" to an lwb, the waiter associated * with it is linked onto the lwb's list of waiters. Then, when that lwb * completes, each waiter on the lwb's list is marked done and signalled * -- allowing the thread waiting on the waiter to return from zil_commit(). * * It's important to point out a few critical factors that allow us * to make use of the commit itxs, commit waiters, per-lwb lists of * commit waiters, and zio completion callbacks like we're doing: * * 1. The list of waiters for each lwb is traversed, and each commit * waiter is marked "done" and signalled, in the zio completion * callback of the lwb's zio[*]. * * * Actually, the waiters are signalled in the zio completion * callback of the root zio for the DKIOCFLUSHWRITECACHE commands * that are sent to the vdevs upon completion of the lwb zio. * * 2. When the itxs are inserted into the ZIL's queue of uncommitted * itxs, the order in which they are inserted is preserved[*]; as * itxs are added to the queue, they are added to the tail of * in-memory linked lists. * * When committing the itxs to lwbs (to be written to disk), they * are committed in the same order in which the itxs were added to * the uncommitted queue's linked list(s); i.e. the linked list of * itxs to commit is traversed from head to tail, and each itx is * committed to an lwb in that order. * * * To clarify: * * - the order of "sync" itxs is preserved w.r.t. other * "sync" itxs, regardless of the corresponding objects. * - the order of "async" itxs is preserved w.r.t. other * "async" itxs corresponding to the same object. * - the order of "async" itxs is *not* preserved w.r.t. other * "async" itxs corresponding to different objects. * - the order of "sync" itxs w.r.t. "async" itxs (or vice * versa) is *not* preserved, even for itxs that correspond * to the same object. * * For more details, see: zil_itx_assign(), zil_async_to_sync(), * zil_get_commit_list(), and zil_process_commit_list(). * * 3. The lwbs represent a linked list of blocks on disk. Thus, any * lwb cannot be considered committed to stable storage, until its * "previous" lwb is also committed to stable storage. This fact, * coupled with the fact described above, means that itxs are * committed in (roughly) the order in which they were generated. * This is essential because itxs are dependent on prior itxs. * Thus, we *must not* deem an itx as being committed to stable * storage, until *all* prior itxs have also been committed to * stable storage. * * To enforce this ordering of lwb zio's, while still leveraging as * much of the underlying storage performance as possible, we rely * on two fundamental concepts: * * 1. The creation and issuance of lwb zio's is protected by * the zilog's "zl_issuer_lock", which ensures only a single * thread is creating and/or issuing lwb's at a time * 2. The "previous" lwb is a child of the "current" lwb * (leveraging the zio parent-child depenency graph) * * By relying on this parent-child zio relationship, we can have * many lwb zio's concurrently issued to the underlying storage, * but the order in which they complete will be the same order in * which they were created. */ void zil_commit(zilog_t *zilog, uint64_t foid) { /* * We should never attempt to call zil_commit on a snapshot for * a couple of reasons: * * 1. A snapshot may never be modified, thus it cannot have any * in-flight itxs that would have modified the dataset. * * 2. By design, when zil_commit() is called, a commit itx will * be assigned to this zilog; as a result, the zilog will be * dirtied. We must not dirty the zilog of a snapshot; there's * checks in the code that enforce this invariant, and will * cause a panic if it's not upheld. */ ASSERT3B(dmu_objset_is_snapshot(zilog->zl_os), ==, B_FALSE); if (zilog->zl_sync == ZFS_SYNC_DISABLED) return; if (!spa_writeable(zilog->zl_spa)) { /* * If the SPA is not writable, there should never be any * pending itxs waiting to be committed to disk. If that * weren't true, we'd skip writing those itxs out, and * would break the sematics of zil_commit(); thus, we're * verifying that truth before we return to the caller. */ ASSERT(list_is_empty(&zilog->zl_lwb_list)); ASSERT3P(zilog->zl_last_lwb_opened, ==, NULL); for (int i = 0; i < TXG_SIZE; i++) ASSERT3P(zilog->zl_itxg[i].itxg_itxs, ==, NULL); return; } /* * If the ZIL is suspended, we don't want to dirty it by calling * zil_commit_itx_assign() below, nor can we write out * lwbs like would be done in zil_commit_write(). Thus, we * simply rely on txg_wait_synced() to maintain the necessary * semantics, and avoid calling those functions altogether. */ if (zilog->zl_suspend > 0) { txg_wait_synced(zilog->zl_dmu_pool, 0); return; } + zil_commit_impl(zilog, foid); +} + +void +zil_commit_impl(zilog_t *zilog, uint64_t foid) +{ /* * Move the "async" itxs for the specified foid to the "sync" * queues, such that they will be later committed (or skipped) * to an lwb when zil_process_commit_list() is called. * * Since these "async" itxs must be committed prior to this * call to zil_commit returning, we must perform this operation * before we call zil_commit_itx_assign(). */ zil_async_to_sync(zilog, foid); /* * We allocate a new "waiter" structure which will initially be * linked to the commit itx using the itx's "itx_private" field. * Since the commit itx doesn't represent any on-disk state, * when it's committed to an lwb, rather than copying the its * lr_t into the lwb's buffer, the commit itx's "waiter" will be * added to the lwb's list of waiters. Then, when the lwb is * committed to stable storage, each waiter in the lwb's list of * waiters will be marked "done", and signalled. * * We must create the waiter and assign the commit itx prior to * calling zil_commit_writer(), or else our specific commit itx * is not guaranteed to be committed to an lwb prior to calling * zil_commit_waiter(). */ zil_commit_waiter_t *zcw = zil_alloc_commit_waiter(); zil_commit_itx_assign(zilog, zcw); zil_commit_writer(zilog, zcw); zil_commit_waiter(zilog, zcw); if (zcw->zcw_zio_error != 0) { /* * If there was an error writing out the ZIL blocks that * this thread is waiting on, then we fallback to * relying on spa_sync() to write out the data this * thread is waiting on. Obviously this has performance * implications, but the expectation is for this to be * an exceptional case, and shouldn't occur often. */ DTRACE_PROBE2(zil__commit__io__error, zilog_t *, zilog, zil_commit_waiter_t *, zcw); txg_wait_synced(zilog->zl_dmu_pool, 0); } zil_free_commit_waiter(zcw); } /* * 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(spa, txg, &lwb->lwb_blk); zil_free_lwb(zilog, 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); } /* ARGSUSED */ static int zil_lwb_cons(void *vbuf, void *unused, int kmflag) { lwb_t *lwb = vbuf; list_create(&lwb->lwb_waiters, sizeof (zil_commit_waiter_t), offsetof(zil_commit_waiter_t, zcw_node)); avl_create(&lwb->lwb_vdev_tree, zil_lwb_vdev_compare, sizeof (zil_vdev_node_t), offsetof(zil_vdev_node_t, zv_node)); mutex_init(&lwb->lwb_vdev_lock, NULL, MUTEX_DEFAULT, NULL); return (0); } /* ARGSUSED */ static void zil_lwb_dest(void *vbuf, void *unused) { lwb_t *lwb = vbuf; mutex_destroy(&lwb->lwb_vdev_lock); avl_destroy(&lwb->lwb_vdev_tree); list_destroy(&lwb->lwb_waiters); } void zil_init(void) { zil_lwb_cache = kmem_cache_create("zil_lwb_cache", sizeof (lwb_t), 0, zil_lwb_cons, zil_lwb_dest, NULL, NULL, NULL, 0); zil_zcw_cache = kmem_cache_create("zil_zcw_cache", sizeof (zil_commit_waiter_t), 0, NULL, NULL, NULL, NULL, NULL, 0); } void zil_fini(void) { kmem_cache_destroy(zil_zcw_cache); 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_dirty_max_txg = 0; zilog->zl_last_lwb_opened = NULL; zilog->zl_last_lwb_latency = 0; mutex_init(&zilog->zl_lock, NULL, MUTEX_DEFAULT, NULL); mutex_init(&zilog->zl_issuer_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)); cv_init(&zilog->zl_cv_suspend, 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); 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_issuer_lock); mutex_destroy(&zilog->zl_lock); cv_destroy(&zilog->zl_cv_suspend); 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); ASSERT3P(zilog->zl_get_data, ==, NULL); ASSERT3P(zilog->zl_last_lwb_opened, ==, NULL); ASSERT(list_is_empty(&zilog->zl_lwb_list)); zilog->zl_get_data = get_data; return (zilog); } /* * Close an intent log. */ void zil_close(zilog_t *zilog) { lwb_t *lwb; uint64_t txg; if (!dmu_objset_is_snapshot(zilog->zl_os)) { zil_commit(zilog, 0); } else { ASSERT3P(list_tail(&zilog->zl_lwb_list), ==, NULL); ASSERT0(zilog->zl_dirty_max_txg); ASSERT3B(zilog_is_dirty(zilog), ==, B_FALSE); } mutex_enter(&zilog->zl_lock); lwb = list_tail(&zilog->zl_lwb_list); if (lwb == NULL) txg = zilog->zl_dirty_max_txg; else txg = MAX(zilog->zl_dirty_max_txg, lwb->lwb_max_txg); mutex_exit(&zilog->zl_lock); /* * We need to use txg_wait_synced() to wait long enough for the * ZIL to be clean, and to wait for all pending lwbs to be * written out. */ if (txg != 0) txg_wait_synced(zilog->zl_dmu_pool, txg); if (zilog_is_dirty(zilog)) zfs_dbgmsg("zil (%p) is dirty, txg %llu", zilog, txg); VERIFY(!zilog_is_dirty(zilog)); 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) { ASSERT3P(lwb, ==, list_tail(&zilog->zl_lwb_list)); ASSERT3S(lwb->lwb_state, !=, LWB_STATE_ISSUED); list_remove(&zilog->zl_lwb_list, lwb); zio_buf_free(lwb->lwb_buf, lwb->lwb_sz); zil_free_lwb(zilog, 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); + /* + * We need to use zil_commit_impl to ensure we wait for all + * LWB_STATE_OPENED and LWB_STATE_ISSUED lwb's to be committed + * to disk before proceeding. If we used zil_commit instead, it + * would just call txg_wait_synced(), because zl_suspend is set. + * txg_wait_synced() doesn't wait for these lwb's to be + * LWB_STATE_DONE before returning. + */ + zil_commit_impl(zilog, 0); + + /* + * Now that we've ensured all lwb's are LWB_STATE_DONE, we use + * txg_wait_synced() to ensure the data from the zilog has + * migrated to the main pool before calling zil_destroy(). + */ + txg_wait_synced(zilog->zl_dmu_pool, 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/zio.c =================================================================== --- head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio.c (revision 328227) +++ head/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zio.c (revision 328228) @@ -1,4156 +1,4142 @@ /* * CDDL HEADER START * * The contents of this file are subject to the terms of the * Common Development and Distribution License (the "License"). * You may not use this file except in compliance with the License. * * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE * or http://www.opensolaris.org/os/licensing. * See the License for the specific language governing permissions * and limitations under the License. * * When distributing Covered Code, include this CDDL HEADER in each * file and include the License file at usr/src/OPENSOLARIS.LICENSE. * If applicable, add the following below this CDDL HEADER, with the * fields enclosed by brackets "[]" replaced with your own identifying * information: Portions Copyright [yyyy] [name of copyright owner] * * CDDL HEADER END */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright (c) 2011, 2017 by Delphix. All rights reserved. * Copyright (c) 2011 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2014 Integros [integros.com] */ #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include #include SYSCTL_DECL(_vfs_zfs); SYSCTL_NODE(_vfs_zfs, OID_AUTO, zio, CTLFLAG_RW, 0, "ZFS ZIO"); #if defined(__amd64__) static int zio_use_uma = 1; #else static int zio_use_uma = 0; #endif SYSCTL_INT(_vfs_zfs_zio, OID_AUTO, use_uma, CTLFLAG_RDTUN, &zio_use_uma, 0, "Use uma(9) for ZIO allocations"); static int zio_exclude_metadata = 0; SYSCTL_INT(_vfs_zfs_zio, OID_AUTO, exclude_metadata, CTLFLAG_RDTUN, &zio_exclude_metadata, 0, "Exclude metadata buffers from dumps as well"); zio_trim_stats_t zio_trim_stats = { { "bytes", KSTAT_DATA_UINT64, "Number of bytes successfully TRIMmed" }, { "success", KSTAT_DATA_UINT64, "Number of successful TRIM requests" }, { "unsupported", KSTAT_DATA_UINT64, "Number of TRIM requests that failed because TRIM is not supported" }, { "failed", KSTAT_DATA_UINT64, "Number of TRIM requests that failed for reasons other than not supported" }, }; static kstat_t *zio_trim_ksp; /* * ========================================================================== * I/O type descriptions * ========================================================================== */ const char *zio_type_name[ZIO_TYPES] = { "zio_null", "zio_read", "zio_write", "zio_free", "zio_claim", "zio_ioctl" }; boolean_t zio_dva_throttle_enabled = B_TRUE; SYSCTL_INT(_vfs_zfs_zio, OID_AUTO, dva_throttle_enabled, CTLFLAG_RDTUN, &zio_dva_throttle_enabled, 0, ""); /* * ========================================================================== * I/O kmem caches * ========================================================================== */ kmem_cache_t *zio_cache; kmem_cache_t *zio_link_cache; kmem_cache_t *zio_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; kmem_cache_t *zio_data_buf_cache[SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT]; #ifdef _KERNEL extern vmem_t *zio_alloc_arena; #endif #define ZIO_PIPELINE_CONTINUE 0x100 #define ZIO_PIPELINE_STOP 0x101 #define BP_SPANB(indblkshift, level) \ (((uint64_t)1) << ((level) * ((indblkshift) - SPA_BLKPTRSHIFT))) #define COMPARE_META_LEVEL 0x80000000ul /* * The following actions directly effect the spa's sync-to-convergence logic. * The values below define the sync pass when we start performing the action. * Care should be taken when changing these values as they directly impact * spa_sync() performance. Tuning these values may introduce subtle performance * pathologies and should only be done in the context of performance analysis. * These tunables will eventually be removed and replaced with #defines once * enough analysis has been done to determine optimal values. * * The 'zfs_sync_pass_deferred_free' pass must be greater than 1 to ensure that * regular blocks are not deferred. */ int zfs_sync_pass_deferred_free = 2; /* defer frees starting in this pass */ SYSCTL_INT(_vfs_zfs, OID_AUTO, sync_pass_deferred_free, CTLFLAG_RDTUN, &zfs_sync_pass_deferred_free, 0, "defer frees starting in this pass"); int zfs_sync_pass_dont_compress = 5; /* don't compress starting in this pass */ SYSCTL_INT(_vfs_zfs, OID_AUTO, sync_pass_dont_compress, CTLFLAG_RDTUN, &zfs_sync_pass_dont_compress, 0, "don't compress starting in this pass"); int zfs_sync_pass_rewrite = 2; /* rewrite new bps starting in this pass */ SYSCTL_INT(_vfs_zfs, OID_AUTO, sync_pass_rewrite, CTLFLAG_RDTUN, &zfs_sync_pass_rewrite, 0, "rewrite new bps starting in this pass"); /* * An allocating zio is one that either currently has the DVA allocate * stage set or will have it later in its lifetime. */ #define IO_IS_ALLOCATING(zio) ((zio)->io_orig_pipeline & ZIO_STAGE_DVA_ALLOCATE) boolean_t zio_requeue_io_start_cut_in_line = B_TRUE; #ifdef illumos #ifdef ZFS_DEBUG int zio_buf_debug_limit = 16384; #else int zio_buf_debug_limit = 0; #endif #endif static void zio_taskq_dispatch(zio_t *, zio_taskq_type_t, boolean_t); void zio_init(void) { size_t c; zio_cache = kmem_cache_create("zio_cache", sizeof (zio_t), 0, NULL, NULL, NULL, NULL, NULL, 0); zio_link_cache = kmem_cache_create("zio_link_cache", sizeof (zio_link_t), 0, NULL, NULL, NULL, NULL, NULL, 0); if (!zio_use_uma) goto out; /* * For small buffers, we want a cache for each multiple of * SPA_MINBLOCKSIZE. For larger buffers, we want a cache * for each quarter-power of 2. */ for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) { size_t size = (c + 1) << SPA_MINBLOCKSHIFT; size_t p2 = size; size_t align = 0; int cflags = zio_exclude_metadata ? KMC_NODEBUG : 0; while (!ISP2(p2)) p2 &= p2 - 1; #ifdef illumos #ifndef _KERNEL /* * If we are using watchpoints, put each buffer on its own page, * to eliminate the performance overhead of trapping to the * kernel when modifying a non-watched buffer that shares the * page with a watched buffer. */ if (arc_watch && !IS_P2ALIGNED(size, PAGESIZE)) continue; #endif #endif /* illumos */ if (size <= 4 * SPA_MINBLOCKSIZE) { align = SPA_MINBLOCKSIZE; } else if (IS_P2ALIGNED(size, p2 >> 2)) { align = MIN(p2 >> 2, PAGESIZE); } if (align != 0) { char name[36]; (void) sprintf(name, "zio_buf_%lu", (ulong_t)size); zio_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, NULL, cflags); /* * Since zio_data bufs do not appear in crash dumps, we * pass KMC_NOTOUCH so that no allocator metadata is * stored with the buffers. */ (void) sprintf(name, "zio_data_buf_%lu", (ulong_t)size); zio_data_buf_cache[c] = kmem_cache_create(name, size, align, NULL, NULL, NULL, NULL, NULL, cflags | KMC_NOTOUCH | KMC_NODEBUG); } } while (--c != 0) { ASSERT(zio_buf_cache[c] != NULL); if (zio_buf_cache[c - 1] == NULL) zio_buf_cache[c - 1] = zio_buf_cache[c]; ASSERT(zio_data_buf_cache[c] != NULL); if (zio_data_buf_cache[c - 1] == NULL) zio_data_buf_cache[c - 1] = zio_data_buf_cache[c]; } out: zio_inject_init(); zio_trim_ksp = kstat_create("zfs", 0, "zio_trim", "misc", KSTAT_TYPE_NAMED, sizeof(zio_trim_stats) / sizeof(kstat_named_t), KSTAT_FLAG_VIRTUAL); if (zio_trim_ksp != NULL) { zio_trim_ksp->ks_data = &zio_trim_stats; kstat_install(zio_trim_ksp); } } void zio_fini(void) { size_t c; kmem_cache_t *last_cache = NULL; kmem_cache_t *last_data_cache = NULL; for (c = 0; c < SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT; c++) { if (zio_buf_cache[c] != last_cache) { last_cache = zio_buf_cache[c]; kmem_cache_destroy(zio_buf_cache[c]); } zio_buf_cache[c] = NULL; if (zio_data_buf_cache[c] != last_data_cache) { last_data_cache = zio_data_buf_cache[c]; kmem_cache_destroy(zio_data_buf_cache[c]); } zio_data_buf_cache[c] = NULL; } kmem_cache_destroy(zio_link_cache); kmem_cache_destroy(zio_cache); zio_inject_fini(); if (zio_trim_ksp != NULL) { kstat_delete(zio_trim_ksp); zio_trim_ksp = NULL; } } /* * ========================================================================== * Allocate and free I/O buffers * ========================================================================== */ /* * Use zio_buf_alloc to allocate ZFS metadata. This data will appear in a * crashdump if the kernel panics, so use it judiciously. Obviously, it's * useful to inspect ZFS metadata, but if possible, we should avoid keeping * excess / transient data in-core during a crashdump. */ void * zio_buf_alloc(size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; int flags = zio_exclude_metadata ? KM_NODEBUG : 0; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); if (zio_use_uma) return (kmem_cache_alloc(zio_buf_cache[c], KM_PUSHPAGE)); else return (kmem_alloc(size, KM_SLEEP|flags)); } /* * Use zio_data_buf_alloc to allocate data. The data will not appear in a * crashdump if the kernel panics. This exists so that we will limit the amount * of ZFS data that shows up in a kernel crashdump. (Thus reducing the amount * of kernel heap dumped to disk when the kernel panics) */ void * zio_data_buf_alloc(size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); if (zio_use_uma) return (kmem_cache_alloc(zio_data_buf_cache[c], KM_PUSHPAGE)); else return (kmem_alloc(size, KM_SLEEP | KM_NODEBUG)); } void zio_buf_free(void *buf, size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); if (zio_use_uma) kmem_cache_free(zio_buf_cache[c], buf); else kmem_free(buf, size); } void zio_data_buf_free(void *buf, size_t size) { size_t c = (size - 1) >> SPA_MINBLOCKSHIFT; VERIFY3U(c, <, SPA_MAXBLOCKSIZE >> SPA_MINBLOCKSHIFT); if (zio_use_uma) kmem_cache_free(zio_data_buf_cache[c], buf); else kmem_free(buf, size); } /* * ========================================================================== * Push and pop I/O transform buffers * ========================================================================== */ void zio_push_transform(zio_t *zio, abd_t *data, uint64_t size, uint64_t bufsize, zio_transform_func_t *transform) { zio_transform_t *zt = kmem_alloc(sizeof (zio_transform_t), KM_SLEEP); /* * Ensure that anyone expecting this zio to contain a linear ABD isn't * going to get a nasty surprise when they try to access the data. */ #ifdef illumos IMPLY(abd_is_linear(zio->io_abd), abd_is_linear(data)); #else IMPLY(zio->io_abd != NULL && abd_is_linear(zio->io_abd), abd_is_linear(data)); #endif zt->zt_orig_abd = zio->io_abd; zt->zt_orig_size = zio->io_size; zt->zt_bufsize = bufsize; zt->zt_transform = transform; zt->zt_next = zio->io_transform_stack; zio->io_transform_stack = zt; zio->io_abd = data; zio->io_size = size; } void zio_pop_transforms(zio_t *zio) { zio_transform_t *zt; while ((zt = zio->io_transform_stack) != NULL) { if (zt->zt_transform != NULL) zt->zt_transform(zio, zt->zt_orig_abd, zt->zt_orig_size); if (zt->zt_bufsize != 0) abd_free(zio->io_abd); zio->io_abd = zt->zt_orig_abd; zio->io_size = zt->zt_orig_size; zio->io_transform_stack = zt->zt_next; kmem_free(zt, sizeof (zio_transform_t)); } } /* * ========================================================================== * I/O transform callbacks for subblocks and decompression * ========================================================================== */ static void zio_subblock(zio_t *zio, abd_t *data, uint64_t size) { ASSERT(zio->io_size > size); if (zio->io_type == ZIO_TYPE_READ) abd_copy(data, zio->io_abd, size); } static void zio_decompress(zio_t *zio, abd_t *data, uint64_t size) { if (zio->io_error == 0) { void *tmp = abd_borrow_buf(data, size); int ret = zio_decompress_data(BP_GET_COMPRESS(zio->io_bp), zio->io_abd, tmp, zio->io_size, size); abd_return_buf_copy(data, tmp, size); if (ret != 0) zio->io_error = SET_ERROR(EIO); } } /* * ========================================================================== * I/O parent/child relationships and pipeline interlocks * ========================================================================== */ zio_t * zio_walk_parents(zio_t *cio, zio_link_t **zl) { list_t *pl = &cio->io_parent_list; *zl = (*zl == NULL) ? list_head(pl) : list_next(pl, *zl); if (*zl == NULL) return (NULL); ASSERT((*zl)->zl_child == cio); return ((*zl)->zl_parent); } zio_t * zio_walk_children(zio_t *pio, zio_link_t **zl) { list_t *cl = &pio->io_child_list; *zl = (*zl == NULL) ? list_head(cl) : list_next(cl, *zl); if (*zl == NULL) return (NULL); ASSERT((*zl)->zl_parent == pio); return ((*zl)->zl_child); } zio_t * zio_unique_parent(zio_t *cio) { zio_link_t *zl = NULL; zio_t *pio = zio_walk_parents(cio, &zl); VERIFY3P(zio_walk_parents(cio, &zl), ==, NULL); return (pio); } void zio_add_child(zio_t *pio, zio_t *cio) { zio_link_t *zl = kmem_cache_alloc(zio_link_cache, KM_SLEEP); /* * Logical I/Os can have logical, gang, or vdev children. * Gang I/Os can have gang or vdev children. * Vdev I/Os can only have vdev children. * The following ASSERT captures all of these constraints. */ ASSERT3S(cio->io_child_type, <=, pio->io_child_type); zl->zl_parent = pio; zl->zl_child = cio; mutex_enter(&cio->io_lock); mutex_enter(&pio->io_lock); ASSERT(pio->io_state[ZIO_WAIT_DONE] == 0); for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_children[cio->io_child_type][w] += !cio->io_state[w]; list_insert_head(&pio->io_child_list, zl); list_insert_head(&cio->io_parent_list, zl); pio->io_child_count++; cio->io_parent_count++; mutex_exit(&pio->io_lock); mutex_exit(&cio->io_lock); } static void zio_remove_child(zio_t *pio, zio_t *cio, zio_link_t *zl) { ASSERT(zl->zl_parent == pio); ASSERT(zl->zl_child == cio); mutex_enter(&cio->io_lock); mutex_enter(&pio->io_lock); list_remove(&pio->io_child_list, zl); list_remove(&cio->io_parent_list, zl); pio->io_child_count--; cio->io_parent_count--; mutex_exit(&pio->io_lock); mutex_exit(&cio->io_lock); kmem_cache_free(zio_link_cache, zl); } static boolean_t zio_wait_for_children(zio_t *zio, enum zio_child child, enum zio_wait_type wait) { uint64_t *countp = &zio->io_children[child][wait]; boolean_t waiting = B_FALSE; mutex_enter(&zio->io_lock); ASSERT(zio->io_stall == NULL); if (*countp != 0) { zio->io_stage >>= 1; ASSERT3U(zio->io_stage, !=, ZIO_STAGE_OPEN); zio->io_stall = countp; waiting = B_TRUE; } mutex_exit(&zio->io_lock); return (waiting); } static void zio_notify_parent(zio_t *pio, zio_t *zio, enum zio_wait_type wait) { uint64_t *countp = &pio->io_children[zio->io_child_type][wait]; int *errorp = &pio->io_child_error[zio->io_child_type]; mutex_enter(&pio->io_lock); if (zio->io_error && !(zio->io_flags & ZIO_FLAG_DONT_PROPAGATE)) *errorp = zio_worst_error(*errorp, zio->io_error); pio->io_reexecute |= zio->io_reexecute; ASSERT3U(*countp, >, 0); (*countp)--; if (*countp == 0 && pio->io_stall == countp) { zio_taskq_type_t type = pio->io_stage < ZIO_STAGE_VDEV_IO_START ? ZIO_TASKQ_ISSUE : ZIO_TASKQ_INTERRUPT; pio->io_stall = NULL; mutex_exit(&pio->io_lock); /* * Dispatch the parent zio in its own taskq so that * the child can continue to make progress. This also * prevents overflowing the stack when we have deeply nested * parent-child relationships. */ zio_taskq_dispatch(pio, type, B_FALSE); } else { mutex_exit(&pio->io_lock); } } static void zio_inherit_child_errors(zio_t *zio, enum zio_child c) { if (zio->io_child_error[c] != 0 && zio->io_error == 0) zio->io_error = zio->io_child_error[c]; } int zio_bookmark_compare(const void *x1, const void *x2) { const zio_t *z1 = x1; const zio_t *z2 = x2; if (z1->io_bookmark.zb_objset < z2->io_bookmark.zb_objset) return (-1); if (z1->io_bookmark.zb_objset > z2->io_bookmark.zb_objset) return (1); if (z1->io_bookmark.zb_object < z2->io_bookmark.zb_object) return (-1); if (z1->io_bookmark.zb_object > z2->io_bookmark.zb_object) return (1); if (z1->io_bookmark.zb_level < z2->io_bookmark.zb_level) return (-1); if (z1->io_bookmark.zb_level > z2->io_bookmark.zb_level) return (1); if (z1->io_bookmark.zb_blkid < z2->io_bookmark.zb_blkid) return (-1); if (z1->io_bookmark.zb_blkid > z2->io_bookmark.zb_blkid) return (1); if (z1 < z2) return (-1); if (z1 > z2) return (1); return (0); } /* * ========================================================================== * Create the various types of I/O (read, write, free, etc) * ========================================================================== */ static zio_t * zio_create(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, abd_t *data, uint64_t lsize, uint64_t psize, zio_done_func_t *done, void *private, zio_type_t type, zio_priority_t priority, enum zio_flag flags, vdev_t *vd, uint64_t offset, const zbookmark_phys_t *zb, enum zio_stage stage, enum zio_stage pipeline) { zio_t *zio; ASSERT3U(type == ZIO_TYPE_FREE || psize, <=, SPA_MAXBLOCKSIZE); ASSERT(P2PHASE(psize, SPA_MINBLOCKSIZE) == 0); ASSERT(P2PHASE(offset, SPA_MINBLOCKSIZE) == 0); ASSERT(!vd || spa_config_held(spa, SCL_STATE_ALL, RW_READER)); ASSERT(!bp || !(flags & ZIO_FLAG_CONFIG_WRITER)); ASSERT(vd || stage == ZIO_STAGE_OPEN); IMPLY(lsize != psize, (flags & ZIO_FLAG_RAW) != 0); zio = kmem_cache_alloc(zio_cache, KM_SLEEP); bzero(zio, sizeof (zio_t)); mutex_init(&zio->io_lock, NULL, MUTEX_DEFAULT, NULL); cv_init(&zio->io_cv, NULL, CV_DEFAULT, NULL); list_create(&zio->io_parent_list, sizeof (zio_link_t), offsetof(zio_link_t, zl_parent_node)); list_create(&zio->io_child_list, sizeof (zio_link_t), offsetof(zio_link_t, zl_child_node)); metaslab_trace_init(&zio->io_alloc_list); if (vd != NULL) zio->io_child_type = ZIO_CHILD_VDEV; else if (flags & ZIO_FLAG_GANG_CHILD) zio->io_child_type = ZIO_CHILD_GANG; else if (flags & ZIO_FLAG_DDT_CHILD) zio->io_child_type = ZIO_CHILD_DDT; else zio->io_child_type = ZIO_CHILD_LOGICAL; if (bp != NULL) { zio->io_bp = (blkptr_t *)bp; zio->io_bp_copy = *bp; zio->io_bp_orig = *bp; if (type != ZIO_TYPE_WRITE || zio->io_child_type == ZIO_CHILD_DDT) zio->io_bp = &zio->io_bp_copy; /* so caller can free */ if (zio->io_child_type == ZIO_CHILD_LOGICAL) zio->io_logical = zio; if (zio->io_child_type > ZIO_CHILD_GANG && BP_IS_GANG(bp)) pipeline |= ZIO_GANG_STAGES; } zio->io_spa = spa; zio->io_txg = txg; zio->io_done = done; zio->io_private = private; zio->io_type = type; zio->io_priority = priority; zio->io_vd = vd; zio->io_offset = offset; zio->io_orig_abd = zio->io_abd = data; zio->io_orig_size = zio->io_size = psize; zio->io_lsize = lsize; zio->io_orig_flags = zio->io_flags = flags; zio->io_orig_stage = zio->io_stage = stage; zio->io_orig_pipeline = zio->io_pipeline = pipeline; zio->io_pipeline_trace = ZIO_STAGE_OPEN; zio->io_state[ZIO_WAIT_READY] = (stage >= ZIO_STAGE_READY); zio->io_state[ZIO_WAIT_DONE] = (stage >= ZIO_STAGE_DONE); if (zb != NULL) zio->io_bookmark = *zb; if (pio != NULL) { if (zio->io_logical == NULL) zio->io_logical = pio->io_logical; if (zio->io_child_type == ZIO_CHILD_GANG) zio->io_gang_leader = pio->io_gang_leader; zio_add_child(pio, zio); } return (zio); } static void zio_destroy(zio_t *zio) { metaslab_trace_fini(&zio->io_alloc_list); list_destroy(&zio->io_parent_list); list_destroy(&zio->io_child_list); mutex_destroy(&zio->io_lock); cv_destroy(&zio->io_cv); kmem_cache_free(zio_cache, zio); } zio_t * zio_null(zio_t *pio, spa_t *spa, vdev_t *vd, zio_done_func_t *done, void *private, enum zio_flag flags) { zio_t *zio; zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private, ZIO_TYPE_NULL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL, ZIO_STAGE_OPEN, ZIO_INTERLOCK_PIPELINE); return (zio); } zio_t * zio_root(spa_t *spa, zio_done_func_t *done, void *private, enum zio_flag flags) { return (zio_null(NULL, spa, NULL, done, private, flags)); } void zfs_blkptr_verify(spa_t *spa, const blkptr_t *bp) { if (!DMU_OT_IS_VALID(BP_GET_TYPE(bp))) { zfs_panic_recover("blkptr at %p has invalid TYPE %llu", bp, (longlong_t)BP_GET_TYPE(bp)); } if (BP_GET_CHECKSUM(bp) >= ZIO_CHECKSUM_FUNCTIONS || BP_GET_CHECKSUM(bp) <= ZIO_CHECKSUM_ON) { zfs_panic_recover("blkptr at %p has invalid CHECKSUM %llu", bp, (longlong_t)BP_GET_CHECKSUM(bp)); } if (BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_FUNCTIONS || BP_GET_COMPRESS(bp) <= ZIO_COMPRESS_ON) { zfs_panic_recover("blkptr at %p has invalid COMPRESS %llu", bp, (longlong_t)BP_GET_COMPRESS(bp)); } if (BP_GET_LSIZE(bp) > SPA_MAXBLOCKSIZE) { zfs_panic_recover("blkptr at %p has invalid LSIZE %llu", bp, (longlong_t)BP_GET_LSIZE(bp)); } if (BP_GET_PSIZE(bp) > SPA_MAXBLOCKSIZE) { zfs_panic_recover("blkptr at %p has invalid PSIZE %llu", bp, (longlong_t)BP_GET_PSIZE(bp)); } if (BP_IS_EMBEDDED(bp)) { if (BPE_GET_ETYPE(bp) > NUM_BP_EMBEDDED_TYPES) { zfs_panic_recover("blkptr at %p has invalid ETYPE %llu", bp, (longlong_t)BPE_GET_ETYPE(bp)); } } /* * Pool-specific checks. * * Note: it would be nice to verify that the blk_birth and * BP_PHYSICAL_BIRTH() are not too large. However, spa_freeze() * allows the birth time of log blocks (and dmu_sync()-ed blocks * that are in the log) to be arbitrarily large. */ for (int i = 0; i < BP_GET_NDVAS(bp); i++) { uint64_t vdevid = DVA_GET_VDEV(&bp->blk_dva[i]); if (vdevid >= spa->spa_root_vdev->vdev_children) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } vdev_t *vd = spa->spa_root_vdev->vdev_child[vdevid]; if (vd == NULL) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } if (vd->vdev_ops == &vdev_hole_ops) { zfs_panic_recover("blkptr at %p DVA %u has hole " "VDEV %llu", bp, i, (longlong_t)vdevid); continue; } if (vd->vdev_ops == &vdev_missing_ops) { /* * "missing" vdevs are valid during import, but we * don't have their detailed info (e.g. asize), so * we can't perform any more checks on them. */ continue; } uint64_t offset = DVA_GET_OFFSET(&bp->blk_dva[i]); uint64_t asize = DVA_GET_ASIZE(&bp->blk_dva[i]); if (BP_IS_GANG(bp)) asize = vdev_psize_to_asize(vd, SPA_GANGBLOCKSIZE); if (offset + asize > vd->vdev_asize) { zfs_panic_recover("blkptr at %p DVA %u has invalid " "OFFSET %llu", bp, i, (longlong_t)offset); } } } zio_t * zio_read(zio_t *pio, spa_t *spa, const blkptr_t *bp, abd_t *data, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb) { zio_t *zio; zfs_blkptr_verify(spa, bp); zio = zio_create(pio, spa, BP_PHYSICAL_BIRTH(bp), bp, data, size, size, done, private, ZIO_TYPE_READ, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? ZIO_DDT_CHILD_READ_PIPELINE : ZIO_READ_PIPELINE); return (zio); } zio_t * zio_write(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, abd_t *data, uint64_t lsize, uint64_t psize, const zio_prop_t *zp, zio_done_func_t *ready, zio_done_func_t *children_ready, zio_done_func_t *physdone, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, const zbookmark_phys_t *zb) { zio_t *zio; ASSERT(zp->zp_checksum >= ZIO_CHECKSUM_OFF && zp->zp_checksum < ZIO_CHECKSUM_FUNCTIONS && zp->zp_compress >= ZIO_COMPRESS_OFF && zp->zp_compress < ZIO_COMPRESS_FUNCTIONS && DMU_OT_IS_VALID(zp->zp_type) && zp->zp_level < 32 && zp->zp_copies > 0 && zp->zp_copies <= spa_max_replication(spa)); zio = zio_create(pio, spa, txg, bp, data, lsize, psize, done, private, ZIO_TYPE_WRITE, priority, flags, NULL, 0, zb, ZIO_STAGE_OPEN, (flags & ZIO_FLAG_DDT_CHILD) ? ZIO_DDT_CHILD_WRITE_PIPELINE : ZIO_WRITE_PIPELINE); zio->io_ready = ready; zio->io_children_ready = children_ready; zio->io_physdone = physdone; zio->io_prop = *zp; /* * Data can be NULL if we are going to call zio_write_override() to * provide the already-allocated BP. But we may need the data to * verify a dedup hit (if requested). In this case, don't try to * dedup (just take the already-allocated BP verbatim). */ if (data == NULL && zio->io_prop.zp_dedup_verify) { zio->io_prop.zp_dedup = zio->io_prop.zp_dedup_verify = B_FALSE; } return (zio); } zio_t * zio_rewrite(zio_t *pio, spa_t *spa, uint64_t txg, blkptr_t *bp, abd_t *data, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, zbookmark_phys_t *zb) { zio_t *zio; zio = zio_create(pio, spa, txg, bp, data, size, size, done, private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_IO_REWRITE, NULL, 0, zb, ZIO_STAGE_OPEN, ZIO_REWRITE_PIPELINE); return (zio); } void zio_write_override(zio_t *zio, blkptr_t *bp, int copies, boolean_t nopwrite) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(zio->io_stage == ZIO_STAGE_OPEN); ASSERT(zio->io_txg == spa_syncing_txg(zio->io_spa)); /* * We must reset the io_prop to match the values that existed * when the bp was first written by dmu_sync() keeping in mind * that nopwrite and dedup are mutually exclusive. */ zio->io_prop.zp_dedup = nopwrite ? B_FALSE : zio->io_prop.zp_dedup; zio->io_prop.zp_nopwrite = nopwrite; zio->io_prop.zp_copies = copies; zio->io_bp_override = bp; } void zio_free(spa_t *spa, uint64_t txg, const blkptr_t *bp) { /* * The check for EMBEDDED is a performance optimization. We * process the free here (by ignoring it) rather than * putting it on the list and then processing it in zio_free_sync(). */ if (BP_IS_EMBEDDED(bp)) return; metaslab_check_free(spa, bp); /* * Frees that are for the currently-syncing txg, are not going to be * deferred, and which will not need to do a read (i.e. not GANG or * DEDUP), can be processed immediately. Otherwise, put them on the * in-memory list for later processing. */ if (zfs_trim_enabled || BP_IS_GANG(bp) || BP_GET_DEDUP(bp) || txg != spa->spa_syncing_txg || spa_sync_pass(spa) >= zfs_sync_pass_deferred_free) { bplist_append(&spa->spa_free_bplist[txg & TXG_MASK], bp); } else { VERIFY0(zio_wait(zio_free_sync(NULL, spa, txg, bp, BP_GET_PSIZE(bp), 0))); } } zio_t * zio_free_sync(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, uint64_t size, enum zio_flag flags) { zio_t *zio; enum zio_stage stage = ZIO_FREE_PIPELINE; ASSERT(!BP_IS_HOLE(bp)); ASSERT(spa_syncing_txg(spa) == txg); ASSERT(spa_sync_pass(spa) < zfs_sync_pass_deferred_free); if (BP_IS_EMBEDDED(bp)) return (zio_null(pio, spa, NULL, NULL, NULL, 0)); metaslab_check_free(spa, bp); arc_freed(spa, bp); if (zfs_trim_enabled) stage |= ZIO_STAGE_ISSUE_ASYNC | ZIO_STAGE_VDEV_IO_START | ZIO_STAGE_VDEV_IO_ASSESS; /* * GANG and DEDUP blocks can induce a read (for the gang block header, * or the DDT), so issue them asynchronously so that this thread is * not tied up. */ else if (BP_IS_GANG(bp) || BP_GET_DEDUP(bp)) stage |= ZIO_STAGE_ISSUE_ASYNC; flags |= ZIO_FLAG_DONT_QUEUE; zio = zio_create(pio, spa, txg, bp, NULL, size, size, NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, stage); return (zio); } zio_t * zio_claim(zio_t *pio, spa_t *spa, uint64_t txg, const blkptr_t *bp, zio_done_func_t *done, void *private, enum zio_flag flags) { zio_t *zio; dprintf_bp(bp, "claiming in txg %llu", txg); if (BP_IS_EMBEDDED(bp)) return (zio_null(pio, spa, NULL, NULL, NULL, 0)); /* * A claim is an allocation of a specific block. Claims are needed * to support immediate writes in the intent log. The issue is that * immediate writes contain committed data, but in a txg that was * *not* committed. Upon opening the pool after an unclean shutdown, * the intent log claims all blocks that contain immediate write data * so that the SPA knows they're in use. * * All claims *must* be resolved in the first txg -- before the SPA * starts allocating blocks -- so that nothing is allocated twice. * If txg == 0 we just verify that the block is claimable. */ ASSERT3U(spa->spa_uberblock.ub_rootbp.blk_birth, <, spa_first_txg(spa)); ASSERT(txg == spa_first_txg(spa) || txg == 0); ASSERT(!BP_GET_DEDUP(bp) || !spa_writeable(spa)); /* zdb(1M) */ zio = zio_create(pio, spa, txg, bp, NULL, BP_GET_PSIZE(bp), BP_GET_PSIZE(bp), done, private, ZIO_TYPE_CLAIM, ZIO_PRIORITY_NOW, flags, NULL, 0, NULL, ZIO_STAGE_OPEN, ZIO_CLAIM_PIPELINE); ASSERT0(zio->io_queued_timestamp); return (zio); } zio_t * zio_ioctl(zio_t *pio, spa_t *spa, vdev_t *vd, int cmd, uint64_t offset, uint64_t size, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags) { zio_t *zio; int c; if (vd->vdev_children == 0) { zio = zio_create(pio, spa, 0, NULL, NULL, 0, 0, done, private, ZIO_TYPE_IOCTL, ZIO_PRIORITY_NOW, flags, vd, 0, NULL, ZIO_STAGE_OPEN, ZIO_IOCTL_PIPELINE); zio->io_cmd = cmd; } else { zio = zio_null(pio, spa, NULL, NULL, NULL, flags); for (c = 0; c < vd->vdev_children; c++) zio_nowait(zio_ioctl(zio, spa, vd->vdev_child[c], cmd, offset, size, done, private, priority, flags)); } return (zio); } zio_t * zio_read_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, abd_t *data, int checksum, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, boolean_t labels) { zio_t *zio; ASSERT(vd->vdev_children == 0); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done, private, ZIO_TYPE_READ, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_READ_PHYS_PIPELINE); zio->io_prop.zp_checksum = checksum; return (zio); } zio_t * zio_write_phys(zio_t *pio, vdev_t *vd, uint64_t offset, uint64_t size, abd_t *data, int checksum, zio_done_func_t *done, void *private, zio_priority_t priority, enum zio_flag flags, boolean_t labels) { zio_t *zio; ASSERT(vd->vdev_children == 0); ASSERT(!labels || offset + size <= VDEV_LABEL_START_SIZE || offset >= vd->vdev_psize - VDEV_LABEL_END_SIZE); ASSERT3U(offset + size, <=, vd->vdev_psize); zio = zio_create(pio, vd->vdev_spa, 0, NULL, data, size, size, done, private, ZIO_TYPE_WRITE, priority, flags | ZIO_FLAG_PHYSICAL, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_WRITE_PHYS_PIPELINE); zio->io_prop.zp_checksum = checksum; if (zio_checksum_table[checksum].ci_flags & ZCHECKSUM_FLAG_EMBEDDED) { /* * zec checksums are necessarily destructive -- they modify * the end of the write buffer to hold the verifier/checksum. * Therefore, we must make a local copy in case the data is * being written to multiple places in parallel. */ abd_t *wbuf = abd_alloc_sametype(data, size); abd_copy(wbuf, data, size); zio_push_transform(zio, wbuf, size, size, NULL); } return (zio); } /* * Create a child I/O to do some work for us. */ zio_t * zio_vdev_child_io(zio_t *pio, blkptr_t *bp, vdev_t *vd, uint64_t offset, abd_t *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *private) { enum zio_stage pipeline = ZIO_VDEV_CHILD_PIPELINE; zio_t *zio; ASSERT(vd->vdev_parent == (pio->io_vd ? pio->io_vd : pio->io_spa->spa_root_vdev)); if (type == ZIO_TYPE_READ && bp != NULL) { /* * If we have the bp, then the child should perform the * checksum and the parent need not. This pushes error * detection as close to the leaves as possible and * eliminates redundant checksums in the interior nodes. */ pipeline |= ZIO_STAGE_CHECKSUM_VERIFY; pio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY; } /* Not all IO types require vdev io done stage e.g. free */ if (!(pio->io_pipeline & ZIO_STAGE_VDEV_IO_DONE)) pipeline &= ~ZIO_STAGE_VDEV_IO_DONE; if (vd->vdev_children == 0) offset += VDEV_LABEL_START_SIZE; flags |= ZIO_VDEV_CHILD_FLAGS(pio) | ZIO_FLAG_DONT_PROPAGATE; /* * If we've decided to do a repair, the write is not speculative -- * even if the original read was. */ if (flags & ZIO_FLAG_IO_REPAIR) flags &= ~ZIO_FLAG_SPECULATIVE; /* * If we're creating a child I/O that is not associated with a * top-level vdev, then the child zio is not an allocating I/O. * If this is a retried I/O then we ignore it since we will * have already processed the original allocating I/O. */ if (flags & ZIO_FLAG_IO_ALLOCATING && (vd != vd->vdev_top || (flags & ZIO_FLAG_IO_RETRY))) { metaslab_class_t *mc = spa_normal_class(pio->io_spa); ASSERT(mc->mc_alloc_throttle_enabled); ASSERT(type == ZIO_TYPE_WRITE); ASSERT(priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(flags & ZIO_FLAG_IO_REPAIR)); ASSERT(!(pio->io_flags & ZIO_FLAG_IO_REWRITE) || pio->io_child_type == ZIO_CHILD_GANG); flags &= ~ZIO_FLAG_IO_ALLOCATING; } zio = zio_create(pio, pio->io_spa, pio->io_txg, bp, data, size, size, done, private, type, priority, flags, vd, offset, &pio->io_bookmark, ZIO_STAGE_VDEV_IO_START >> 1, pipeline); ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV); zio->io_physdone = pio->io_physdone; if (vd->vdev_ops->vdev_op_leaf && zio->io_logical != NULL) zio->io_logical->io_phys_children++; return (zio); } zio_t * zio_vdev_delegated_io(vdev_t *vd, uint64_t offset, abd_t *data, uint64_t size, int type, zio_priority_t priority, enum zio_flag flags, zio_done_func_t *done, void *private) { zio_t *zio; ASSERT(vd->vdev_ops->vdev_op_leaf); zio = zio_create(NULL, vd->vdev_spa, 0, NULL, data, size, size, done, private, type, priority, flags | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_RETRY | ZIO_FLAG_DELEGATED, vd, offset, NULL, ZIO_STAGE_VDEV_IO_START >> 1, ZIO_VDEV_CHILD_PIPELINE); return (zio); } void zio_flush(zio_t *zio, vdev_t *vd) { zio_nowait(zio_ioctl(zio, zio->io_spa, vd, DKIOCFLUSHWRITECACHE, 0, 0, NULL, NULL, ZIO_PRIORITY_NOW, ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY)); } zio_t * zio_trim(zio_t *zio, spa_t *spa, vdev_t *vd, uint64_t offset, uint64_t size) { ASSERT(vd->vdev_ops->vdev_op_leaf); return (zio_create(zio, spa, 0, NULL, NULL, size, size, NULL, NULL, ZIO_TYPE_FREE, ZIO_PRIORITY_TRIM, ZIO_FLAG_DONT_AGGREGATE | ZIO_FLAG_CANFAIL | ZIO_FLAG_DONT_PROPAGATE | ZIO_FLAG_DONT_RETRY, vd, offset, NULL, ZIO_STAGE_OPEN, ZIO_FREE_PHYS_PIPELINE)); } void zio_shrink(zio_t *zio, uint64_t size) { ASSERT3P(zio->io_executor, ==, NULL); ASSERT3P(zio->io_orig_size, ==, zio->io_size); ASSERT3U(size, <=, zio->io_size); /* * We don't shrink for raidz because of problems with the * reconstruction when reading back less than the block size. * Note, BP_IS_RAIDZ() assumes no compression. */ ASSERT(BP_GET_COMPRESS(zio->io_bp) == ZIO_COMPRESS_OFF); if (!BP_IS_RAIDZ(zio->io_bp)) { /* we are not doing a raw write */ ASSERT3U(zio->io_size, ==, zio->io_lsize); zio->io_orig_size = zio->io_size = zio->io_lsize = size; } } /* * ========================================================================== * Prepare to read and write logical blocks * ========================================================================== */ static int zio_read_bp_init(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (BP_GET_COMPRESS(bp) != ZIO_COMPRESS_OFF && zio->io_child_type == ZIO_CHILD_LOGICAL && !(zio->io_flags & ZIO_FLAG_RAW)) { uint64_t psize = BP_IS_EMBEDDED(bp) ? BPE_GET_PSIZE(bp) : BP_GET_PSIZE(bp); zio_push_transform(zio, abd_alloc_sametype(zio->io_abd, psize), psize, psize, zio_decompress); } if (BP_IS_EMBEDDED(bp) && BPE_GET_ETYPE(bp) == BP_EMBEDDED_TYPE_DATA) { zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; int psize = BPE_GET_PSIZE(bp); void *data = abd_borrow_buf(zio->io_abd, psize); decode_embedded_bp_compressed(bp, data); abd_return_buf_copy(zio->io_abd, data, psize); } else { ASSERT(!BP_IS_EMBEDDED(bp)); } if (!DMU_OT_IS_METADATA(BP_GET_TYPE(bp)) && BP_GET_LEVEL(bp) == 0) zio->io_flags |= ZIO_FLAG_DONT_CACHE; if (BP_GET_TYPE(bp) == DMU_OT_DDT_ZAP) zio->io_flags |= ZIO_FLAG_DONT_CACHE; if (BP_GET_DEDUP(bp) && zio->io_child_type == ZIO_CHILD_LOGICAL) zio->io_pipeline = ZIO_DDT_READ_PIPELINE; return (ZIO_PIPELINE_CONTINUE); } static int zio_write_bp_init(zio_t *zio) { if (!IO_IS_ALLOCATING(zio)) return (ZIO_PIPELINE_CONTINUE); ASSERT(zio->io_child_type != ZIO_CHILD_DDT); if (zio->io_bp_override) { blkptr_t *bp = zio->io_bp; zio_prop_t *zp = &zio->io_prop; ASSERT(bp->blk_birth != zio->io_txg); ASSERT(BP_GET_DEDUP(zio->io_bp_override) == 0); *bp = *zio->io_bp_override; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (BP_IS_EMBEDDED(bp)) return (ZIO_PIPELINE_CONTINUE); /* * If we've been overridden and nopwrite is set then * set the flag accordingly to indicate that a nopwrite * has already occurred. */ if (!BP_IS_HOLE(bp) && zp->zp_nopwrite) { ASSERT(!zp->zp_dedup); ASSERT3U(BP_GET_CHECKSUM(bp), ==, zp->zp_checksum); zio->io_flags |= ZIO_FLAG_NOPWRITE; return (ZIO_PIPELINE_CONTINUE); } ASSERT(!zp->zp_nopwrite); if (BP_IS_HOLE(bp) || !zp->zp_dedup) return (ZIO_PIPELINE_CONTINUE); ASSERT((zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_DEDUP) || zp->zp_dedup_verify); if (BP_GET_CHECKSUM(bp) == zp->zp_checksum) { BP_SET_DEDUP(bp, 1); zio->io_pipeline |= ZIO_STAGE_DDT_WRITE; return (ZIO_PIPELINE_CONTINUE); } /* * We were unable to handle this as an override bp, treat * it as a regular write I/O. */ zio->io_bp_override = NULL; *bp = zio->io_bp_orig; zio->io_pipeline = zio->io_orig_pipeline; } return (ZIO_PIPELINE_CONTINUE); } static int zio_write_compress(zio_t *zio) { spa_t *spa = zio->io_spa; zio_prop_t *zp = &zio->io_prop; enum zio_compress compress = zp->zp_compress; blkptr_t *bp = zio->io_bp; uint64_t lsize = zio->io_lsize; uint64_t psize = zio->io_size; int pass = 1; EQUIV(lsize != psize, (zio->io_flags & ZIO_FLAG_RAW) != 0); /* * If our children haven't all reached the ready stage, * wait for them and then repeat this pipeline stage. */ if (zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_READY) || zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_READY)) return (ZIO_PIPELINE_STOP); if (!IO_IS_ALLOCATING(zio)) return (ZIO_PIPELINE_CONTINUE); if (zio->io_children_ready != NULL) { /* * Now that all our children are ready, run the callback * associated with this zio in case it wants to modify the * data to be written. */ ASSERT3U(zp->zp_level, >, 0); zio->io_children_ready(zio); } ASSERT(zio->io_child_type != ZIO_CHILD_DDT); ASSERT(zio->io_bp_override == NULL); if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg) { /* * We're rewriting an existing block, which means we're * working on behalf of spa_sync(). For spa_sync() to * converge, it must eventually be the case that we don't * have to allocate new blocks. But compression changes * the blocksize, which forces a reallocate, and makes * convergence take longer. Therefore, after the first * few passes, stop compressing to ensure convergence. */ pass = spa_sync_pass(spa); ASSERT(zio->io_txg == spa_syncing_txg(spa)); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!BP_GET_DEDUP(bp)); if (pass >= zfs_sync_pass_dont_compress) compress = ZIO_COMPRESS_OFF; /* Make sure someone doesn't change their mind on overwrites */ ASSERT(BP_IS_EMBEDDED(bp) || MIN(zp->zp_copies + BP_IS_GANG(bp), spa_max_replication(spa)) == BP_GET_NDVAS(bp)); } /* If it's a compressed write that is not raw, compress the buffer. */ if (compress != ZIO_COMPRESS_OFF && psize == lsize) { void *cbuf = zio_buf_alloc(lsize); psize = zio_compress_data(compress, zio->io_abd, cbuf, lsize); if (psize == 0 || psize == lsize) { compress = ZIO_COMPRESS_OFF; zio_buf_free(cbuf, lsize); } else if (!zp->zp_dedup && psize <= BPE_PAYLOAD_SIZE && zp->zp_level == 0 && !DMU_OT_HAS_FILL(zp->zp_type) && spa_feature_is_enabled(spa, SPA_FEATURE_EMBEDDED_DATA)) { encode_embedded_bp_compressed(bp, cbuf, compress, lsize, psize); BPE_SET_ETYPE(bp, BP_EMBEDDED_TYPE_DATA); BP_SET_TYPE(bp, zio->io_prop.zp_type); BP_SET_LEVEL(bp, zio->io_prop.zp_level); zio_buf_free(cbuf, lsize); bp->blk_birth = zio->io_txg; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; ASSERT(spa_feature_is_active(spa, SPA_FEATURE_EMBEDDED_DATA)); return (ZIO_PIPELINE_CONTINUE); } else { /* * Round up compressed size up to the ashift * of the smallest-ashift device, and zero the tail. * This ensures that the compressed size of the BP * (and thus compressratio property) are correct, * in that we charge for the padding used to fill out * the last sector. */ ASSERT3U(spa->spa_min_ashift, >=, SPA_MINBLOCKSHIFT); size_t rounded = (size_t)P2ROUNDUP(psize, 1ULL << spa->spa_min_ashift); if (rounded >= lsize) { compress = ZIO_COMPRESS_OFF; zio_buf_free(cbuf, lsize); psize = lsize; } else { abd_t *cdata = abd_get_from_buf(cbuf, lsize); abd_take_ownership_of_buf(cdata, B_TRUE); abd_zero_off(cdata, psize, rounded - psize); psize = rounded; zio_push_transform(zio, cdata, psize, lsize, NULL); } } /* * We were unable to handle this as an override bp, treat * it as a regular write I/O. */ zio->io_bp_override = NULL; *bp = zio->io_bp_orig; zio->io_pipeline = zio->io_orig_pipeline; } else { ASSERT3U(psize, !=, 0); } /* * The final pass of spa_sync() must be all rewrites, but the first * few passes offer a trade-off: allocating blocks defers convergence, * but newly allocated blocks are sequential, so they can be written * to disk faster. Therefore, we allow the first few passes of * spa_sync() to allocate new blocks, but force rewrites after that. * There should only be a handful of blocks after pass 1 in any case. */ if (!BP_IS_HOLE(bp) && bp->blk_birth == zio->io_txg && BP_GET_PSIZE(bp) == psize && pass >= zfs_sync_pass_rewrite) { ASSERT(psize != 0); enum zio_stage gang_stages = zio->io_pipeline & ZIO_GANG_STAGES; zio->io_pipeline = ZIO_REWRITE_PIPELINE | gang_stages; zio->io_flags |= ZIO_FLAG_IO_REWRITE; } else { BP_ZERO(bp); zio->io_pipeline = ZIO_WRITE_PIPELINE; } if (psize == 0) { if (zio->io_bp_orig.blk_birth != 0 && spa_feature_is_active(spa, SPA_FEATURE_HOLE_BIRTH)) { BP_SET_LSIZE(bp, lsize); BP_SET_TYPE(bp, zp->zp_type); BP_SET_LEVEL(bp, zp->zp_level); BP_SET_BIRTH(bp, zio->io_txg, 0); } zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; } else { ASSERT(zp->zp_checksum != ZIO_CHECKSUM_GANG_HEADER); BP_SET_LSIZE(bp, lsize); BP_SET_TYPE(bp, zp->zp_type); BP_SET_LEVEL(bp, zp->zp_level); BP_SET_PSIZE(bp, psize); BP_SET_COMPRESS(bp, compress); BP_SET_CHECKSUM(bp, zp->zp_checksum); BP_SET_DEDUP(bp, zp->zp_dedup); BP_SET_BYTEORDER(bp, ZFS_HOST_BYTEORDER); if (zp->zp_dedup) { ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); zio->io_pipeline = ZIO_DDT_WRITE_PIPELINE; } if (zp->zp_nopwrite) { ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); zio->io_pipeline |= ZIO_STAGE_NOP_WRITE; } } return (ZIO_PIPELINE_CONTINUE); } static int zio_free_bp_init(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio->io_child_type == ZIO_CHILD_LOGICAL) { if (BP_GET_DEDUP(bp)) zio->io_pipeline = ZIO_DDT_FREE_PIPELINE; } return (ZIO_PIPELINE_CONTINUE); } /* * ========================================================================== * Execute the I/O pipeline * ========================================================================== */ static void zio_taskq_dispatch(zio_t *zio, zio_taskq_type_t q, boolean_t cutinline) { spa_t *spa = zio->io_spa; zio_type_t t = zio->io_type; int flags = (cutinline ? TQ_FRONT : 0); ASSERT(q == ZIO_TASKQ_ISSUE || q == ZIO_TASKQ_INTERRUPT); /* * If we're a config writer or a probe, the normal issue and * interrupt threads may all be blocked waiting for the config lock. * In this case, select the otherwise-unused taskq for ZIO_TYPE_NULL. */ if (zio->io_flags & (ZIO_FLAG_CONFIG_WRITER | ZIO_FLAG_PROBE)) t = ZIO_TYPE_NULL; /* * A similar issue exists for the L2ARC write thread until L2ARC 2.0. */ if (t == ZIO_TYPE_WRITE && zio->io_vd && zio->io_vd->vdev_aux) t = ZIO_TYPE_NULL; /* * If this is a high priority I/O, then use the high priority taskq if * available. */ if (zio->io_priority == ZIO_PRIORITY_NOW && spa->spa_zio_taskq[t][q + 1].stqs_count != 0) q++; ASSERT3U(q, <, ZIO_TASKQ_TYPES); /* * NB: We are assuming that the zio can only be dispatched * to a single taskq at a time. It would be a grievous error * to dispatch the zio to another taskq at the same time. */ #if defined(illumos) || !defined(_KERNEL) ASSERT(zio->io_tqent.tqent_next == NULL); #else ASSERT(zio->io_tqent.tqent_task.ta_pending == 0); #endif spa_taskq_dispatch_ent(spa, t, q, (task_func_t *)zio_execute, zio, flags, &zio->io_tqent); } static boolean_t zio_taskq_member(zio_t *zio, zio_taskq_type_t q) { kthread_t *executor = zio->io_executor; spa_t *spa = zio->io_spa; for (zio_type_t t = 0; t < ZIO_TYPES; t++) { spa_taskqs_t *tqs = &spa->spa_zio_taskq[t][q]; uint_t i; for (i = 0; i < tqs->stqs_count; i++) { if (taskq_member(tqs->stqs_taskq[i], executor)) return (B_TRUE); } } return (B_FALSE); } static int zio_issue_async(zio_t *zio) { zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE); return (ZIO_PIPELINE_STOP); } void zio_interrupt(zio_t *zio) { zio_taskq_dispatch(zio, ZIO_TASKQ_INTERRUPT, B_FALSE); } void zio_delay_interrupt(zio_t *zio) { /* * The timeout_generic() function isn't defined in userspace, so * rather than trying to implement the function, the zio delay * functionality has been disabled for userspace builds. */ #ifdef _KERNEL /* * If io_target_timestamp is zero, then no delay has been registered * for this IO, thus jump to the end of this function and "skip" the * delay; issuing it directly to the zio layer. */ if (zio->io_target_timestamp != 0) { hrtime_t now = gethrtime(); if (now >= zio->io_target_timestamp) { /* * This IO has already taken longer than the target * delay to complete, so we don't want to delay it * any longer; we "miss" the delay and issue it * directly to the zio layer. This is likely due to * the target latency being set to a value less than * the underlying hardware can satisfy (e.g. delay * set to 1ms, but the disks take 10ms to complete an * IO request). */ DTRACE_PROBE2(zio__delay__miss, zio_t *, zio, hrtime_t, now); zio_interrupt(zio); } else { hrtime_t diff = zio->io_target_timestamp - now; DTRACE_PROBE3(zio__delay__hit, zio_t *, zio, hrtime_t, now, hrtime_t, diff); (void) timeout_generic(CALLOUT_NORMAL, (void (*)(void *))zio_interrupt, zio, diff, 1, 0); } return; } #endif DTRACE_PROBE1(zio__delay__skip, zio_t *, zio); zio_interrupt(zio); } /* * Execute the I/O pipeline until one of the following occurs: * * (1) the I/O completes * (2) the pipeline stalls waiting for dependent child I/Os * (3) the I/O issues, so we're waiting for an I/O completion interrupt * (4) the I/O is delegated by vdev-level caching or aggregation * (5) the I/O is deferred due to vdev-level queueing * (6) the I/O is handed off to another thread. * * In all cases, the pipeline stops whenever there's no CPU work; it never * burns a thread in cv_wait(). * * There's no locking on io_stage because there's no legitimate way * for multiple threads to be attempting to process the same I/O. */ static zio_pipe_stage_t *zio_pipeline[]; void zio_execute(zio_t *zio) { zio->io_executor = curthread; ASSERT3U(zio->io_queued_timestamp, >, 0); while (zio->io_stage < ZIO_STAGE_DONE) { enum zio_stage pipeline = zio->io_pipeline; enum zio_stage stage = zio->io_stage; int rv; ASSERT(!MUTEX_HELD(&zio->io_lock)); ASSERT(ISP2(stage)); ASSERT(zio->io_stall == NULL); do { stage <<= 1; } while ((stage & pipeline) == 0); ASSERT(stage <= ZIO_STAGE_DONE); /* * If we are in interrupt context and this pipeline stage * will grab a config lock that is held across I/O, * or may wait for an I/O that needs an interrupt thread * to complete, issue async to avoid deadlock. * * For VDEV_IO_START, we cut in line so that the io will * be sent to disk promptly. */ if ((stage & ZIO_BLOCKING_STAGES) && zio->io_vd == NULL && zio_taskq_member(zio, ZIO_TASKQ_INTERRUPT)) { boolean_t cut = (stage == ZIO_STAGE_VDEV_IO_START) ? zio_requeue_io_start_cut_in_line : B_FALSE; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, cut); return; } zio->io_stage = stage; zio->io_pipeline_trace |= zio->io_stage; rv = zio_pipeline[highbit64(stage) - 1](zio); if (rv == ZIO_PIPELINE_STOP) return; ASSERT(rv == ZIO_PIPELINE_CONTINUE); } } /* * ========================================================================== * Initiate I/O, either sync or async * ========================================================================== */ int zio_wait(zio_t *zio) { int error; ASSERT3P(zio->io_stage, ==, ZIO_STAGE_OPEN); ASSERT3P(zio->io_executor, ==, NULL); zio->io_waiter = curthread; ASSERT0(zio->io_queued_timestamp); zio->io_queued_timestamp = gethrtime(); zio_execute(zio); mutex_enter(&zio->io_lock); while (zio->io_executor != NULL) cv_wait(&zio->io_cv, &zio->io_lock); mutex_exit(&zio->io_lock); error = zio->io_error; zio_destroy(zio); return (error); } void zio_nowait(zio_t *zio) { ASSERT3P(zio->io_executor, ==, NULL); if (zio->io_child_type == ZIO_CHILD_LOGICAL && zio_unique_parent(zio) == NULL) { /* * This is a logical async I/O with no parent to wait for it. * We add it to the spa_async_root_zio "Godfather" I/O which * will ensure they complete prior to unloading the pool. */ spa_t *spa = zio->io_spa; zio_add_child(spa->spa_async_zio_root[CPU_SEQID], zio); } ASSERT0(zio->io_queued_timestamp); zio->io_queued_timestamp = gethrtime(); zio_execute(zio); } /* * ========================================================================== * Reexecute, cancel, or suspend/resume failed I/O * ========================================================================== */ static void zio_reexecute(zio_t *pio) { zio_t *cio, *cio_next; ASSERT(pio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(pio->io_orig_stage == ZIO_STAGE_OPEN); ASSERT(pio->io_gang_leader == NULL); ASSERT(pio->io_gang_tree == NULL); pio->io_flags = pio->io_orig_flags; pio->io_stage = pio->io_orig_stage; pio->io_pipeline = pio->io_orig_pipeline; pio->io_reexecute = 0; pio->io_flags |= ZIO_FLAG_REEXECUTED; pio->io_pipeline_trace = 0; pio->io_error = 0; for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_state[w] = 0; for (int c = 0; c < ZIO_CHILD_TYPES; c++) pio->io_child_error[c] = 0; if (IO_IS_ALLOCATING(pio)) BP_ZERO(pio->io_bp); /* * As we reexecute pio's children, new children could be created. * New children go to the head of pio's io_child_list, however, * so we will (correctly) not reexecute them. The key is that * the remainder of pio's io_child_list, from 'cio_next' onward, * cannot be affected by any side effects of reexecuting 'cio'. */ zio_link_t *zl = NULL; for (cio = zio_walk_children(pio, &zl); cio != NULL; cio = cio_next) { cio_next = zio_walk_children(pio, &zl); mutex_enter(&pio->io_lock); for (int w = 0; w < ZIO_WAIT_TYPES; w++) pio->io_children[cio->io_child_type][w]++; mutex_exit(&pio->io_lock); zio_reexecute(cio); } /* * Now that all children have been reexecuted, execute the parent. * We don't reexecute "The Godfather" I/O here as it's the * responsibility of the caller to wait on him. */ if (!(pio->io_flags & ZIO_FLAG_GODFATHER)) { pio->io_queued_timestamp = gethrtime(); zio_execute(pio); } } void -zio_cancel(zio_t *zio) -{ - /* - * Disallow cancellation of a zio that's already been issued. - */ - VERIFY3P(zio->io_executor, ==, NULL); - - zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; - zio->io_done = NULL; - - zio_nowait(zio); -} - -void zio_suspend(spa_t *spa, zio_t *zio) { if (spa_get_failmode(spa) == ZIO_FAILURE_MODE_PANIC) fm_panic("Pool '%s' has encountered an uncorrectable I/O " "failure and the failure mode property for this pool " "is set to panic.", spa_name(spa)); zfs_ereport_post(FM_EREPORT_ZFS_IO_FAILURE, spa, NULL, NULL, 0, 0); mutex_enter(&spa->spa_suspend_lock); if (spa->spa_suspend_zio_root == NULL) spa->spa_suspend_zio_root = zio_root(spa, NULL, NULL, ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE | ZIO_FLAG_GODFATHER); spa->spa_suspended = B_TRUE; if (zio != NULL) { ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); ASSERT(zio != spa->spa_suspend_zio_root); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ASSERT(zio_unique_parent(zio) == NULL); ASSERT(zio->io_stage == ZIO_STAGE_DONE); zio_add_child(spa->spa_suspend_zio_root, zio); } mutex_exit(&spa->spa_suspend_lock); } int zio_resume(spa_t *spa) { zio_t *pio; /* * Reexecute all previously suspended i/o. */ mutex_enter(&spa->spa_suspend_lock); spa->spa_suspended = B_FALSE; cv_broadcast(&spa->spa_suspend_cv); pio = spa->spa_suspend_zio_root; spa->spa_suspend_zio_root = NULL; mutex_exit(&spa->spa_suspend_lock); if (pio == NULL) return (0); zio_reexecute(pio); return (zio_wait(pio)); } void zio_resume_wait(spa_t *spa) { mutex_enter(&spa->spa_suspend_lock); while (spa_suspended(spa)) cv_wait(&spa->spa_suspend_cv, &spa->spa_suspend_lock); mutex_exit(&spa->spa_suspend_lock); } /* * ========================================================================== * Gang blocks. * * A gang block is a collection of small blocks that looks to the DMU * like one large block. When zio_dva_allocate() cannot find a block * of the requested size, due to either severe fragmentation or the pool * being nearly full, it calls zio_write_gang_block() to construct the * block from smaller fragments. * * A gang block consists of a gang header (zio_gbh_phys_t) and up to * three (SPA_GBH_NBLKPTRS) gang members. The gang header is just like * an indirect block: it's an array of block pointers. It consumes * only one sector and hence is allocatable regardless of fragmentation. * The gang header's bps point to its gang members, which hold the data. * * Gang blocks are self-checksumming, using the bp's * as the verifier to ensure uniqueness of the SHA256 checksum. * Critically, the gang block bp's blk_cksum is the checksum of the data, * not the gang header. This ensures that data block signatures (needed for * deduplication) are independent of how the block is physically stored. * * Gang blocks can be nested: a gang member may itself be a gang block. * Thus every gang block is a tree in which root and all interior nodes are * gang headers, and the leaves are normal blocks that contain user data. * The root of the gang tree is called the gang leader. * * To perform any operation (read, rewrite, free, claim) on a gang block, * zio_gang_assemble() first assembles the gang tree (minus data leaves) * in the io_gang_tree field of the original logical i/o by recursively * reading the gang leader and all gang headers below it. This yields * an in-core tree containing the contents of every gang header and the * bps for every constituent of the gang block. * * With the gang tree now assembled, zio_gang_issue() just walks the gang tree * and invokes a callback on each bp. To free a gang block, zio_gang_issue() * calls zio_free_gang() -- a trivial wrapper around zio_free() -- for each bp. * zio_claim_gang() provides a similarly trivial wrapper for zio_claim(). * zio_read_gang() is a wrapper around zio_read() that omits reading gang * headers, since we already have those in io_gang_tree. zio_rewrite_gang() * performs a zio_rewrite() of the data or, for gang headers, a zio_rewrite() * of the gang header plus zio_checksum_compute() of the data to update the * gang header's blk_cksum as described above. * * The two-phase assemble/issue model solves the problem of partial failure -- * what if you'd freed part of a gang block but then couldn't read the * gang header for another part? Assembling the entire gang tree first * ensures that all the necessary gang header I/O has succeeded before * starting the actual work of free, claim, or write. Once the gang tree * is assembled, free and claim are in-memory operations that cannot fail. * * In the event that a gang write fails, zio_dva_unallocate() walks the * gang tree to immediately free (i.e. insert back into the space map) * everything we've allocated. This ensures that we don't get ENOSPC * errors during repeated suspend/resume cycles due to a flaky device. * * Gang rewrites only happen during sync-to-convergence. If we can't assemble * the gang tree, we won't modify the block, so we can safely defer the free * (knowing that the block is still intact). If we *can* assemble the gang * tree, then even if some of the rewrites fail, zio_dva_unallocate() will free * each constituent bp and we can allocate a new block on the next sync pass. * * In all cases, the gang tree allows complete recovery from partial failure. * ========================================================================== */ static void zio_gang_issue_func_done(zio_t *zio) { abd_put(zio->io_abd); } static zio_t * zio_read_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { if (gn != NULL) return (pio); return (zio_read(pio, pio->io_spa, bp, abd_get_offset(data, offset), BP_GET_PSIZE(bp), zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark)); } static zio_t * zio_rewrite_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { zio_t *zio; if (gn != NULL) { abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE); zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); /* * As we rewrite each gang header, the pipeline will compute * a new gang block header checksum for it; but no one will * compute a new data checksum, so we do that here. The one * exception is the gang leader: the pipeline already computed * its data checksum because that stage precedes gang assembly. * (Presently, nothing actually uses interior data checksums; * this is just good hygiene.) */ if (gn != pio->io_gang_leader->io_gang_tree) { abd_t *buf = abd_get_offset(data, offset); zio_checksum_compute(zio, BP_GET_CHECKSUM(bp), buf, BP_GET_PSIZE(bp)); abd_put(buf); } /* * If we are here to damage data for testing purposes, * leave the GBH alone so that we can detect the damage. */ if (pio->io_gang_leader->io_flags & ZIO_FLAG_INDUCE_DAMAGE) zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; } else { zio = zio_rewrite(pio, pio->io_spa, pio->io_txg, bp, abd_get_offset(data, offset), BP_GET_PSIZE(bp), zio_gang_issue_func_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); } return (zio); } /* ARGSUSED */ static zio_t * zio_free_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { return (zio_free_sync(pio, pio->io_spa, pio->io_txg, bp, BP_IS_GANG(bp) ? SPA_GANGBLOCKSIZE : BP_GET_PSIZE(bp), ZIO_GANG_CHILD_FLAGS(pio))); } /* ARGSUSED */ static zio_t * zio_claim_gang(zio_t *pio, blkptr_t *bp, zio_gang_node_t *gn, abd_t *data, uint64_t offset) { return (zio_claim(pio, pio->io_spa, pio->io_txg, bp, NULL, NULL, ZIO_GANG_CHILD_FLAGS(pio))); } static zio_gang_issue_func_t *zio_gang_issue_func[ZIO_TYPES] = { NULL, zio_read_gang, zio_rewrite_gang, zio_free_gang, zio_claim_gang, NULL }; static void zio_gang_tree_assemble_done(zio_t *zio); static zio_gang_node_t * zio_gang_node_alloc(zio_gang_node_t **gnpp) { zio_gang_node_t *gn; ASSERT(*gnpp == NULL); gn = kmem_zalloc(sizeof (*gn), KM_SLEEP); gn->gn_gbh = zio_buf_alloc(SPA_GANGBLOCKSIZE); *gnpp = gn; return (gn); } static void zio_gang_node_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) ASSERT(gn->gn_child[g] == NULL); zio_buf_free(gn->gn_gbh, SPA_GANGBLOCKSIZE); kmem_free(gn, sizeof (*gn)); *gnpp = NULL; } static void zio_gang_tree_free(zio_gang_node_t **gnpp) { zio_gang_node_t *gn = *gnpp; if (gn == NULL) return; for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) zio_gang_tree_free(&gn->gn_child[g]); zio_gang_node_free(gnpp); } static void zio_gang_tree_assemble(zio_t *gio, blkptr_t *bp, zio_gang_node_t **gnpp) { zio_gang_node_t *gn = zio_gang_node_alloc(gnpp); abd_t *gbh_abd = abd_get_from_buf(gn->gn_gbh, SPA_GANGBLOCKSIZE); ASSERT(gio->io_gang_leader == gio); ASSERT(BP_IS_GANG(bp)); zio_nowait(zio_read(gio, gio->io_spa, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_gang_tree_assemble_done, gn, gio->io_priority, ZIO_GANG_CHILD_FLAGS(gio), &gio->io_bookmark)); } static void zio_gang_tree_assemble_done(zio_t *zio) { zio_t *gio = zio->io_gang_leader; zio_gang_node_t *gn = zio->io_private; blkptr_t *bp = zio->io_bp; ASSERT(gio == zio_unique_parent(zio)); ASSERT(zio->io_child_count == 0); if (zio->io_error) return; /* this ABD was created from a linear buf in zio_gang_tree_assemble */ if (BP_SHOULD_BYTESWAP(bp)) byteswap_uint64_array(abd_to_buf(zio->io_abd), zio->io_size); ASSERT3P(abd_to_buf(zio->io_abd), ==, gn->gn_gbh); ASSERT(zio->io_size == SPA_GANGBLOCKSIZE); ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); abd_put(zio->io_abd); for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; if (!BP_IS_GANG(gbp)) continue; zio_gang_tree_assemble(gio, gbp, &gn->gn_child[g]); } } static void zio_gang_tree_issue(zio_t *pio, zio_gang_node_t *gn, blkptr_t *bp, abd_t *data, uint64_t offset) { zio_t *gio = pio->io_gang_leader; zio_t *zio; ASSERT(BP_IS_GANG(bp) == !!gn); ASSERT(BP_GET_CHECKSUM(bp) == BP_GET_CHECKSUM(gio->io_bp)); ASSERT(BP_GET_LSIZE(bp) == BP_GET_PSIZE(bp) || gn == gio->io_gang_tree); /* * If you're a gang header, your data is in gn->gn_gbh. * If you're a gang member, your data is in 'data' and gn == NULL. */ zio = zio_gang_issue_func[gio->io_type](pio, bp, gn, data, offset); if (gn != NULL) { ASSERT(gn->gn_gbh->zg_tail.zec_magic == ZEC_MAGIC); for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { blkptr_t *gbp = &gn->gn_gbh->zg_blkptr[g]; if (BP_IS_HOLE(gbp)) continue; zio_gang_tree_issue(zio, gn->gn_child[g], gbp, data, offset); offset += BP_GET_PSIZE(gbp); } } if (gn == gio->io_gang_tree && gio->io_abd != NULL) ASSERT3U(gio->io_size, ==, offset); if (zio != pio) zio_nowait(zio); } static int zio_gang_assemble(zio_t *zio) { blkptr_t *bp = zio->io_bp; ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == NULL); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); zio->io_gang_leader = zio; zio_gang_tree_assemble(zio, bp, &zio->io_gang_tree); return (ZIO_PIPELINE_CONTINUE); } static int zio_gang_issue(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); ASSERT(BP_IS_GANG(bp) && zio->io_gang_leader == zio); ASSERT(zio->io_child_type > ZIO_CHILD_GANG); if (zio->io_child_error[ZIO_CHILD_GANG] == 0) zio_gang_tree_issue(zio, zio->io_gang_tree, bp, zio->io_abd, 0); else zio_gang_tree_free(&zio->io_gang_tree); zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; return (ZIO_PIPELINE_CONTINUE); } static void zio_write_gang_member_ready(zio_t *zio) { zio_t *pio = zio_unique_parent(zio); zio_t *gio = zio->io_gang_leader; dva_t *cdva = zio->io_bp->blk_dva; dva_t *pdva = pio->io_bp->blk_dva; uint64_t asize; if (BP_IS_HOLE(zio->io_bp)) return; ASSERT(BP_IS_HOLE(&zio->io_bp_orig)); ASSERT(zio->io_child_type == ZIO_CHILD_GANG); ASSERT3U(zio->io_prop.zp_copies, ==, gio->io_prop.zp_copies); ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(zio->io_bp)); ASSERT3U(pio->io_prop.zp_copies, <=, BP_GET_NDVAS(pio->io_bp)); ASSERT3U(BP_GET_NDVAS(zio->io_bp), <=, BP_GET_NDVAS(pio->io_bp)); mutex_enter(&pio->io_lock); for (int d = 0; d < BP_GET_NDVAS(zio->io_bp); d++) { ASSERT(DVA_GET_GANG(&pdva[d])); asize = DVA_GET_ASIZE(&pdva[d]); asize += DVA_GET_ASIZE(&cdva[d]); DVA_SET_ASIZE(&pdva[d], asize); } mutex_exit(&pio->io_lock); } static void zio_write_gang_done(zio_t *zio) { abd_put(zio->io_abd); } static int zio_write_gang_block(zio_t *pio) { spa_t *spa = pio->io_spa; metaslab_class_t *mc = spa_normal_class(spa); blkptr_t *bp = pio->io_bp; zio_t *gio = pio->io_gang_leader; zio_t *zio; zio_gang_node_t *gn, **gnpp; zio_gbh_phys_t *gbh; abd_t *gbh_abd; uint64_t txg = pio->io_txg; uint64_t resid = pio->io_size; uint64_t lsize; int copies = gio->io_prop.zp_copies; int gbh_copies = MIN(copies + 1, spa_max_replication(spa)); zio_prop_t zp; int error; int flags = METASLAB_HINTBP_FAVOR | METASLAB_GANG_HEADER; if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(pio->io_flags & ZIO_FLAG_NODATA)); flags |= METASLAB_ASYNC_ALLOC; VERIFY(refcount_held(&mc->mc_alloc_slots, pio)); /* * The logical zio has already placed a reservation for * 'copies' allocation slots but gang blocks may require * additional copies. These additional copies * (i.e. gbh_copies - copies) are guaranteed to succeed * since metaslab_class_throttle_reserve() always allows * additional reservations for gang blocks. */ VERIFY(metaslab_class_throttle_reserve(mc, gbh_copies - copies, pio, flags)); } error = metaslab_alloc(spa, mc, SPA_GANGBLOCKSIZE, bp, gbh_copies, txg, pio == gio ? NULL : gio->io_bp, flags, &pio->io_alloc_list, pio); if (error) { if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(pio->io_flags & ZIO_FLAG_NODATA)); /* * If we failed to allocate the gang block header then * we remove any additional allocation reservations that * we placed here. The original reservation will * be removed when the logical I/O goes to the ready * stage. */ metaslab_class_throttle_unreserve(mc, gbh_copies - copies, pio); } pio->io_error = error; return (ZIO_PIPELINE_CONTINUE); } if (pio == gio) { gnpp = &gio->io_gang_tree; } else { gnpp = pio->io_private; ASSERT(pio->io_ready == zio_write_gang_member_ready); } gn = zio_gang_node_alloc(gnpp); gbh = gn->gn_gbh; bzero(gbh, SPA_GANGBLOCKSIZE); gbh_abd = abd_get_from_buf(gbh, SPA_GANGBLOCKSIZE); /* * Create the gang header. */ zio = zio_rewrite(pio, spa, txg, bp, gbh_abd, SPA_GANGBLOCKSIZE, zio_write_gang_done, NULL, pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); /* * Create and nowait the gang children. */ for (int g = 0; resid != 0; resid -= lsize, g++) { lsize = P2ROUNDUP(resid / (SPA_GBH_NBLKPTRS - g), SPA_MINBLOCKSIZE); ASSERT(lsize >= SPA_MINBLOCKSIZE && lsize <= resid); zp.zp_checksum = gio->io_prop.zp_checksum; zp.zp_compress = ZIO_COMPRESS_OFF; zp.zp_type = DMU_OT_NONE; zp.zp_level = 0; zp.zp_copies = gio->io_prop.zp_copies; zp.zp_dedup = B_FALSE; zp.zp_dedup_verify = B_FALSE; zp.zp_nopwrite = B_FALSE; zio_t *cio = zio_write(zio, spa, txg, &gbh->zg_blkptr[g], abd_get_offset(pio->io_abd, pio->io_size - resid), lsize, lsize, &zp, zio_write_gang_member_ready, NULL, NULL, zio_write_gang_done, &gn->gn_child[g], pio->io_priority, ZIO_GANG_CHILD_FLAGS(pio), &pio->io_bookmark); if (pio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(pio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(!(pio->io_flags & ZIO_FLAG_NODATA)); /* * Gang children won't throttle but we should * account for their work, so reserve an allocation * slot for them here. */ VERIFY(metaslab_class_throttle_reserve(mc, zp.zp_copies, cio, flags)); } zio_nowait(cio); } /* * Set pio's pipeline to just wait for zio to finish. */ pio->io_pipeline = ZIO_INTERLOCK_PIPELINE; zio_nowait(zio); return (ZIO_PIPELINE_CONTINUE); } /* * The zio_nop_write stage in the pipeline determines if allocating a * new bp is necessary. The nopwrite feature can handle writes in * either syncing or open context (i.e. zil writes) and as a result is * mutually exclusive with dedup. * * By leveraging a cryptographically secure checksum, such as SHA256, we * can compare the checksums of the new data and the old to determine if * allocating a new block is required. Note that our requirements for * cryptographic strength are fairly weak: there can't be any accidental * hash collisions, but we don't need to be secure against intentional * (malicious) collisions. To trigger a nopwrite, you have to be able * to write the file to begin with, and triggering an incorrect (hash * collision) nopwrite is no worse than simply writing to the file. * That said, there are no known attacks against the checksum algorithms * used for nopwrite, assuming that the salt and the checksums * themselves remain secret. */ static int zio_nop_write(zio_t *zio) { blkptr_t *bp = zio->io_bp; blkptr_t *bp_orig = &zio->io_bp_orig; zio_prop_t *zp = &zio->io_prop; ASSERT(BP_GET_LEVEL(bp) == 0); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(zp->zp_nopwrite); ASSERT(!zp->zp_dedup); ASSERT(zio->io_bp_override == NULL); ASSERT(IO_IS_ALLOCATING(zio)); /* * Check to see if the original bp and the new bp have matching * characteristics (i.e. same checksum, compression algorithms, etc). * If they don't then just continue with the pipeline which will * allocate a new bp. */ if (BP_IS_HOLE(bp_orig) || !(zio_checksum_table[BP_GET_CHECKSUM(bp)].ci_flags & ZCHECKSUM_FLAG_NOPWRITE) || BP_GET_CHECKSUM(bp) != BP_GET_CHECKSUM(bp_orig) || BP_GET_COMPRESS(bp) != BP_GET_COMPRESS(bp_orig) || BP_GET_DEDUP(bp) != BP_GET_DEDUP(bp_orig) || zp->zp_copies != BP_GET_NDVAS(bp_orig)) return (ZIO_PIPELINE_CONTINUE); /* * If the checksums match then reset the pipeline so that we * avoid allocating a new bp and issuing any I/O. */ if (ZIO_CHECKSUM_EQUAL(bp->blk_cksum, bp_orig->blk_cksum)) { ASSERT(zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_NOPWRITE); ASSERT3U(BP_GET_PSIZE(bp), ==, BP_GET_PSIZE(bp_orig)); ASSERT3U(BP_GET_LSIZE(bp), ==, BP_GET_LSIZE(bp_orig)); ASSERT(zp->zp_compress != ZIO_COMPRESS_OFF); ASSERT(bcmp(&bp->blk_prop, &bp_orig->blk_prop, sizeof (uint64_t)) == 0); *bp = *bp_orig; zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; zio->io_flags |= ZIO_FLAG_NOPWRITE; } return (ZIO_PIPELINE_CONTINUE); } /* * ========================================================================== * Dedup * ========================================================================== */ static void zio_ddt_child_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp; zio_t *pio = zio_unique_parent(zio); mutex_enter(&pio->io_lock); ddp = ddt_phys_select(dde, bp); if (zio->io_error == 0) ddt_phys_clear(ddp); /* this ddp doesn't need repair */ if (zio->io_error == 0 && dde->dde_repair_abd == NULL) dde->dde_repair_abd = zio->io_abd; else abd_free(zio->io_abd); mutex_exit(&pio->io_lock); } static int zio_ddt_read_start(zio_t *zio) { blkptr_t *bp = zio->io_bp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_PSIZE(bp) == zio->io_size); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (zio->io_child_error[ZIO_CHILD_DDT]) { ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = ddt_repair_start(ddt, bp); ddt_phys_t *ddp = dde->dde_phys; ddt_phys_t *ddp_self = ddt_phys_select(dde, bp); blkptr_t blk; ASSERT(zio->io_vsd == NULL); zio->io_vsd = dde; if (ddp_self == NULL) return (ZIO_PIPELINE_CONTINUE); for (int p = 0; p < DDT_PHYS_TYPES; p++, ddp++) { if (ddp->ddp_phys_birth == 0 || ddp == ddp_self) continue; ddt_bp_create(ddt->ddt_checksum, &dde->dde_key, ddp, &blk); zio_nowait(zio_read(zio, zio->io_spa, &blk, abd_alloc_for_io(zio->io_size, B_TRUE), zio->io_size, zio_ddt_child_read_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio) | ZIO_FLAG_DONT_PROPAGATE, &zio->io_bookmark)); } return (ZIO_PIPELINE_CONTINUE); } zio_nowait(zio_read(zio, zio->io_spa, bp, zio->io_abd, zio->io_size, NULL, NULL, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark)); return (ZIO_PIPELINE_CONTINUE); } static int zio_ddt_read_done(zio_t *zio) { blkptr_t *bp = zio->io_bp; if (zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_PSIZE(bp) == zio->io_size); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (zio->io_child_error[ZIO_CHILD_DDT]) { ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = zio->io_vsd; if (ddt == NULL) { ASSERT(spa_load_state(zio->io_spa) != SPA_LOAD_NONE); return (ZIO_PIPELINE_CONTINUE); } if (dde == NULL) { zio->io_stage = ZIO_STAGE_DDT_READ_START >> 1; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_FALSE); return (ZIO_PIPELINE_STOP); } if (dde->dde_repair_abd != NULL) { abd_copy(zio->io_abd, dde->dde_repair_abd, zio->io_size); zio->io_child_error[ZIO_CHILD_DDT] = 0; } ddt_repair_done(ddt, dde); zio->io_vsd = NULL; } ASSERT(zio->io_vsd == NULL); return (ZIO_PIPELINE_CONTINUE); } static boolean_t zio_ddt_collision(zio_t *zio, ddt_t *ddt, ddt_entry_t *dde) { spa_t *spa = zio->io_spa; boolean_t do_raw = (zio->io_flags & ZIO_FLAG_RAW); /* We should never get a raw, override zio */ ASSERT(!(zio->io_bp_override && do_raw)); /* * Note: we compare the original data, not the transformed data, * because when zio->io_bp is an override bp, we will not have * pushed the I/O transforms. That's an important optimization * because otherwise we'd compress/encrypt all dmu_sync() data twice. */ for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { zio_t *lio = dde->dde_lead_zio[p]; if (lio != NULL) { return (lio->io_orig_size != zio->io_orig_size || abd_cmp(zio->io_orig_abd, lio->io_orig_abd, zio->io_orig_size) != 0); } } for (int p = DDT_PHYS_SINGLE; p <= DDT_PHYS_TRIPLE; p++) { ddt_phys_t *ddp = &dde->dde_phys[p]; if (ddp->ddp_phys_birth != 0) { arc_buf_t *abuf = NULL; arc_flags_t aflags = ARC_FLAG_WAIT; int zio_flags = ZIO_FLAG_CANFAIL | ZIO_FLAG_SPECULATIVE; blkptr_t blk = *zio->io_bp; int error; ddt_bp_fill(ddp, &blk, ddp->ddp_phys_birth); ddt_exit(ddt); /* * Intuitively, it would make more sense to compare * io_abd than io_orig_abd in the raw case since you * don't want to look at any transformations that have * happened to the data. However, for raw I/Os the * data will actually be the same in io_abd and * io_orig_abd, so all we have to do is issue this as * a raw ARC read. */ if (do_raw) { zio_flags |= ZIO_FLAG_RAW; ASSERT3U(zio->io_size, ==, zio->io_orig_size); ASSERT0(abd_cmp(zio->io_abd, zio->io_orig_abd, zio->io_size)); ASSERT3P(zio->io_transform_stack, ==, NULL); } error = arc_read(NULL, spa, &blk, arc_getbuf_func, &abuf, ZIO_PRIORITY_SYNC_READ, zio_flags, &aflags, &zio->io_bookmark); if (error == 0) { if (arc_buf_size(abuf) != zio->io_orig_size || abd_cmp_buf(zio->io_orig_abd, abuf->b_data, zio->io_orig_size) != 0) error = SET_ERROR(EEXIST); arc_buf_destroy(abuf, &abuf); } ddt_enter(ddt); return (error != 0); } } return (B_FALSE); } static void zio_ddt_child_write_ready(zio_t *zio) { int p = zio->io_prop.zp_copies; ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; zio_t *pio; if (zio->io_error) return; ddt_enter(ddt); ASSERT(dde->dde_lead_zio[p] == zio); ddt_phys_fill(ddp, zio->io_bp); zio_link_t *zl = NULL; while ((pio = zio_walk_parents(zio, &zl)) != NULL) ddt_bp_fill(ddp, pio->io_bp, zio->io_txg); ddt_exit(ddt); } static void zio_ddt_child_write_done(zio_t *zio) { int p = zio->io_prop.zp_copies; ddt_t *ddt = ddt_select(zio->io_spa, zio->io_bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; ddt_enter(ddt); ASSERT(ddp->ddp_refcnt == 0); ASSERT(dde->dde_lead_zio[p] == zio); dde->dde_lead_zio[p] = NULL; if (zio->io_error == 0) { zio_link_t *zl = NULL; while (zio_walk_parents(zio, &zl) != NULL) ddt_phys_addref(ddp); } else { ddt_phys_clear(ddp); } ddt_exit(ddt); } static void zio_ddt_ditto_write_done(zio_t *zio) { int p = DDT_PHYS_DITTO; zio_prop_t *zp = &zio->io_prop; blkptr_t *bp = zio->io_bp; ddt_t *ddt = ddt_select(zio->io_spa, bp); ddt_entry_t *dde = zio->io_private; ddt_phys_t *ddp = &dde->dde_phys[p]; ddt_key_t *ddk = &dde->dde_key; ddt_enter(ddt); ASSERT(ddp->ddp_refcnt == 0); ASSERT(dde->dde_lead_zio[p] == zio); dde->dde_lead_zio[p] = NULL; if (zio->io_error == 0) { ASSERT(ZIO_CHECKSUM_EQUAL(bp->blk_cksum, ddk->ddk_cksum)); ASSERT(zp->zp_copies < SPA_DVAS_PER_BP); ASSERT(zp->zp_copies == BP_GET_NDVAS(bp) - BP_IS_GANG(bp)); if (ddp->ddp_phys_birth != 0) ddt_phys_free(ddt, ddk, ddp, zio->io_txg); ddt_phys_fill(ddp, bp); } ddt_exit(ddt); } static int zio_ddt_write(zio_t *zio) { spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; uint64_t txg = zio->io_txg; zio_prop_t *zp = &zio->io_prop; int p = zp->zp_copies; int ditto_copies; zio_t *cio = NULL; zio_t *dio = NULL; ddt_t *ddt = ddt_select(spa, bp); ddt_entry_t *dde; ddt_phys_t *ddp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(BP_GET_CHECKSUM(bp) == zp->zp_checksum); ASSERT(BP_IS_HOLE(bp) || zio->io_bp_override); ASSERT(!(zio->io_bp_override && (zio->io_flags & ZIO_FLAG_RAW))); ddt_enter(ddt); dde = ddt_lookup(ddt, bp, B_TRUE); ddp = &dde->dde_phys[p]; if (zp->zp_dedup_verify && zio_ddt_collision(zio, ddt, dde)) { /* * If we're using a weak checksum, upgrade to a strong checksum * and try again. If we're already using a strong checksum, * we can't resolve it, so just convert to an ordinary write. * (And automatically e-mail a paper to Nature?) */ if (!(zio_checksum_table[zp->zp_checksum].ci_flags & ZCHECKSUM_FLAG_DEDUP)) { zp->zp_checksum = spa_dedup_checksum(spa); zio_pop_transforms(zio); zio->io_stage = ZIO_STAGE_OPEN; BP_ZERO(bp); } else { zp->zp_dedup = B_FALSE; BP_SET_DEDUP(bp, B_FALSE); } ASSERT(!BP_GET_DEDUP(bp)); zio->io_pipeline = ZIO_WRITE_PIPELINE; ddt_exit(ddt); return (ZIO_PIPELINE_CONTINUE); } ditto_copies = ddt_ditto_copies_needed(ddt, dde, ddp); ASSERT(ditto_copies < SPA_DVAS_PER_BP); if (ditto_copies > ddt_ditto_copies_present(dde) && dde->dde_lead_zio[DDT_PHYS_DITTO] == NULL) { zio_prop_t czp = *zp; czp.zp_copies = ditto_copies; /* * If we arrived here with an override bp, we won't have run * the transform stack, so we won't have the data we need to * generate a child i/o. So, toss the override bp and restart. * This is safe, because using the override bp is just an * optimization; and it's rare, so the cost doesn't matter. */ if (zio->io_bp_override) { zio_pop_transforms(zio); zio->io_stage = ZIO_STAGE_OPEN; zio->io_pipeline = ZIO_WRITE_PIPELINE; zio->io_bp_override = NULL; BP_ZERO(bp); ddt_exit(ddt); return (ZIO_PIPELINE_CONTINUE); } dio = zio_write(zio, spa, txg, bp, zio->io_orig_abd, zio->io_orig_size, zio->io_orig_size, &czp, NULL, NULL, NULL, zio_ddt_ditto_write_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark); zio_push_transform(dio, zio->io_abd, zio->io_size, 0, NULL); dde->dde_lead_zio[DDT_PHYS_DITTO] = dio; } if (ddp->ddp_phys_birth != 0 || dde->dde_lead_zio[p] != NULL) { if (ddp->ddp_phys_birth != 0) ddt_bp_fill(ddp, bp, txg); if (dde->dde_lead_zio[p] != NULL) zio_add_child(zio, dde->dde_lead_zio[p]); else ddt_phys_addref(ddp); } else if (zio->io_bp_override) { ASSERT(bp->blk_birth == txg); ASSERT(BP_EQUAL(bp, zio->io_bp_override)); ddt_phys_fill(ddp, bp); ddt_phys_addref(ddp); } else { cio = zio_write(zio, spa, txg, bp, zio->io_orig_abd, zio->io_orig_size, zio->io_orig_size, zp, zio_ddt_child_write_ready, NULL, NULL, zio_ddt_child_write_done, dde, zio->io_priority, ZIO_DDT_CHILD_FLAGS(zio), &zio->io_bookmark); zio_push_transform(cio, zio->io_abd, zio->io_size, 0, NULL); dde->dde_lead_zio[p] = cio; } ddt_exit(ddt); if (cio) zio_nowait(cio); if (dio) zio_nowait(dio); return (ZIO_PIPELINE_CONTINUE); } ddt_entry_t *freedde; /* for debugging */ static int zio_ddt_free(zio_t *zio) { spa_t *spa = zio->io_spa; blkptr_t *bp = zio->io_bp; ddt_t *ddt = ddt_select(spa, bp); ddt_entry_t *dde; ddt_phys_t *ddp; ASSERT(BP_GET_DEDUP(bp)); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); ddt_enter(ddt); freedde = dde = ddt_lookup(ddt, bp, B_TRUE); ddp = ddt_phys_select(dde, bp); ddt_phys_decref(ddp); ddt_exit(ddt); return (ZIO_PIPELINE_CONTINUE); } /* * ========================================================================== * Allocate and free blocks * ========================================================================== */ static zio_t * zio_io_to_allocate(spa_t *spa) { zio_t *zio; ASSERT(MUTEX_HELD(&spa->spa_alloc_lock)); zio = avl_first(&spa->spa_alloc_tree); if (zio == NULL) return (NULL); ASSERT(IO_IS_ALLOCATING(zio)); /* * Try to place a reservation for this zio. If we're unable to * reserve then we throttle. */ if (!metaslab_class_throttle_reserve(spa_normal_class(spa), zio->io_prop.zp_copies, zio, 0)) { return (NULL); } avl_remove(&spa->spa_alloc_tree, zio); ASSERT3U(zio->io_stage, <, ZIO_STAGE_DVA_ALLOCATE); return (zio); } static int zio_dva_throttle(zio_t *zio) { spa_t *spa = zio->io_spa; zio_t *nio; if (zio->io_priority == ZIO_PRIORITY_SYNC_WRITE || !spa_normal_class(zio->io_spa)->mc_alloc_throttle_enabled || zio->io_child_type == ZIO_CHILD_GANG || zio->io_flags & ZIO_FLAG_NODATA) { return (ZIO_PIPELINE_CONTINUE); } ASSERT(zio->io_child_type > ZIO_CHILD_GANG); ASSERT3U(zio->io_queued_timestamp, >, 0); ASSERT(zio->io_stage == ZIO_STAGE_DVA_THROTTLE); mutex_enter(&spa->spa_alloc_lock); ASSERT(zio->io_type == ZIO_TYPE_WRITE); avl_add(&spa->spa_alloc_tree, zio); nio = zio_io_to_allocate(zio->io_spa); mutex_exit(&spa->spa_alloc_lock); if (nio == zio) return (ZIO_PIPELINE_CONTINUE); if (nio != NULL) { ASSERT(nio->io_stage == ZIO_STAGE_DVA_THROTTLE); /* * We are passing control to a new zio so make sure that * it is processed by a different thread. We do this to * avoid stack overflows that can occur when parents are * throttled and children are making progress. We allow * it to go to the head of the taskq since it's already * been waiting. */ zio_taskq_dispatch(nio, ZIO_TASKQ_ISSUE, B_TRUE); } return (ZIO_PIPELINE_STOP); } void zio_allocate_dispatch(spa_t *spa) { zio_t *zio; mutex_enter(&spa->spa_alloc_lock); zio = zio_io_to_allocate(spa); mutex_exit(&spa->spa_alloc_lock); if (zio == NULL) return; ASSERT3U(zio->io_stage, ==, ZIO_STAGE_DVA_THROTTLE); ASSERT0(zio->io_error); zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, B_TRUE); } static int zio_dva_allocate(zio_t *zio) { spa_t *spa = zio->io_spa; metaslab_class_t *mc = spa_normal_class(spa); blkptr_t *bp = zio->io_bp; int error; int flags = 0; if (zio->io_gang_leader == NULL) { ASSERT(zio->io_child_type > ZIO_CHILD_GANG); zio->io_gang_leader = zio; } ASSERT(BP_IS_HOLE(bp)); ASSERT0(BP_GET_NDVAS(bp)); ASSERT3U(zio->io_prop.zp_copies, >, 0); ASSERT3U(zio->io_prop.zp_copies, <=, spa_max_replication(spa)); ASSERT3U(zio->io_size, ==, BP_GET_PSIZE(bp)); if (zio->io_flags & ZIO_FLAG_NODATA) { flags |= METASLAB_DONT_THROTTLE; } if (zio->io_flags & ZIO_FLAG_GANG_CHILD) { flags |= METASLAB_GANG_CHILD; } if (zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE) { flags |= METASLAB_ASYNC_ALLOC; } error = metaslab_alloc(spa, mc, zio->io_size, bp, zio->io_prop.zp_copies, zio->io_txg, NULL, flags, &zio->io_alloc_list, zio); if (error != 0) { spa_dbgmsg(spa, "%s: metaslab allocation failure: zio %p, " "size %llu, error %d", spa_name(spa), zio, zio->io_size, error); if (error == ENOSPC && zio->io_size > SPA_MINBLOCKSIZE) return (zio_write_gang_block(zio)); zio->io_error = error; } return (ZIO_PIPELINE_CONTINUE); } static int zio_dva_free(zio_t *zio) { metaslab_free(zio->io_spa, zio->io_bp, zio->io_txg, B_FALSE); return (ZIO_PIPELINE_CONTINUE); } static int zio_dva_claim(zio_t *zio) { int error; error = metaslab_claim(zio->io_spa, zio->io_bp, zio->io_txg); if (error) zio->io_error = error; return (ZIO_PIPELINE_CONTINUE); } /* * Undo an allocation. This is used by zio_done() when an I/O fails * and we want to give back the block we just allocated. * This handles both normal blocks and gang blocks. */ static void zio_dva_unallocate(zio_t *zio, zio_gang_node_t *gn, blkptr_t *bp) { ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp)); ASSERT(zio->io_bp_override == NULL); if (!BP_IS_HOLE(bp)) metaslab_free(zio->io_spa, bp, bp->blk_birth, B_TRUE); if (gn != NULL) { for (int g = 0; g < SPA_GBH_NBLKPTRS; g++) { zio_dva_unallocate(zio, gn->gn_child[g], &gn->gn_gbh->zg_blkptr[g]); } } } /* * Try to allocate an intent log block. Return 0 on success, errno on failure. */ int zio_alloc_zil(spa_t *spa, uint64_t txg, blkptr_t *new_bp, blkptr_t *old_bp, uint64_t size, boolean_t *slog) { int error = 1; zio_alloc_list_t io_alloc_list; ASSERT(txg > spa_syncing_txg(spa)); metaslab_trace_init(&io_alloc_list); error = metaslab_alloc(spa, spa_log_class(spa), size, new_bp, 1, txg, old_bp, METASLAB_HINTBP_AVOID, &io_alloc_list, NULL); if (error == 0) { *slog = TRUE; } else { error = metaslab_alloc(spa, spa_normal_class(spa), size, new_bp, 1, txg, old_bp, METASLAB_HINTBP_AVOID, &io_alloc_list, NULL); if (error == 0) *slog = FALSE; } metaslab_trace_fini(&io_alloc_list); if (error == 0) { BP_SET_LSIZE(new_bp, size); BP_SET_PSIZE(new_bp, size); BP_SET_COMPRESS(new_bp, ZIO_COMPRESS_OFF); BP_SET_CHECKSUM(new_bp, spa_version(spa) >= SPA_VERSION_SLIM_ZIL ? ZIO_CHECKSUM_ZILOG2 : ZIO_CHECKSUM_ZILOG); BP_SET_TYPE(new_bp, DMU_OT_INTENT_LOG); BP_SET_LEVEL(new_bp, 0); BP_SET_DEDUP(new_bp, 0); BP_SET_BYTEORDER(new_bp, ZFS_HOST_BYTEORDER); } else { zfs_dbgmsg("%s: zil block allocation failure: " "size %llu, error %d", spa_name(spa), size, error); } return (error); } /* * Free an intent log block. */ void zio_free_zil(spa_t *spa, uint64_t txg, blkptr_t *bp) { ASSERT(BP_GET_TYPE(bp) == DMU_OT_INTENT_LOG); ASSERT(!BP_IS_GANG(bp)); zio_free(spa, txg, bp); } /* * ========================================================================== * Read, write and delete to physical devices * ========================================================================== */ /* * Issue an I/O to the underlying vdev. Typically the issue pipeline * stops after this stage and will resume upon I/O completion. * However, there are instances where the vdev layer may need to * continue the pipeline when an I/O was not issued. Since the I/O * that was sent to the vdev layer might be different than the one * currently active in the pipeline (see vdev_queue_io()), we explicitly * force the underlying vdev layers to call either zio_execute() or * zio_interrupt() to ensure that the pipeline continues with the correct I/O. */ static int zio_vdev_io_start(zio_t *zio) { vdev_t *vd = zio->io_vd; uint64_t align; spa_t *spa = zio->io_spa; int ret; ASSERT(zio->io_error == 0); ASSERT(zio->io_child_error[ZIO_CHILD_VDEV] == 0); if (vd == NULL) { if (!(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) spa_config_enter(spa, SCL_ZIO, zio, RW_READER); /* * The mirror_ops handle multiple DVAs in a single BP. */ vdev_mirror_ops.vdev_op_io_start(zio); return (ZIO_PIPELINE_STOP); } if (vd->vdev_ops->vdev_op_leaf && zio->io_type == ZIO_TYPE_FREE && zio->io_priority == ZIO_PRIORITY_NOW) { trim_map_free(vd, zio->io_offset, zio->io_size, zio->io_txg); return (ZIO_PIPELINE_CONTINUE); } ASSERT3P(zio->io_logical, !=, zio); /* * We keep track of time-sensitive I/Os so that the scan thread * can quickly react to certain workloads. In particular, we care * about non-scrubbing, top-level reads and writes with the following * characteristics: * - synchronous writes of user data to non-slog devices * - any reads of user data * When these conditions are met, adjust the timestamp of spa_last_io * which allows the scan thread to adjust its workload accordingly. */ if (!(zio->io_flags & ZIO_FLAG_SCAN_THREAD) && zio->io_bp != NULL && vd == vd->vdev_top && !vd->vdev_islog && zio->io_bookmark.zb_objset != DMU_META_OBJSET && zio->io_txg != spa_syncing_txg(spa)) { uint64_t old = spa->spa_last_io; uint64_t new = ddi_get_lbolt64(); if (old != new) (void) atomic_cas_64(&spa->spa_last_io, old, new); } align = 1ULL << vd->vdev_top->vdev_ashift; if (!(zio->io_flags & ZIO_FLAG_PHYSICAL) && P2PHASE(zio->io_size, align) != 0) { /* Transform logical writes to be a full physical block size. */ uint64_t asize = P2ROUNDUP(zio->io_size, align); abd_t *abuf = NULL; if (zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE) abuf = abd_alloc_sametype(zio->io_abd, asize); ASSERT(vd == vd->vdev_top); if (zio->io_type == ZIO_TYPE_WRITE) { abd_copy(abuf, zio->io_abd, zio->io_size); abd_zero_off(abuf, zio->io_size, asize - zio->io_size); } zio_push_transform(zio, abuf, asize, abuf ? asize : 0, zio_subblock); } /* * If this is not a physical io, make sure that it is properly aligned * before proceeding. */ if (!(zio->io_flags & ZIO_FLAG_PHYSICAL)) { ASSERT0(P2PHASE(zio->io_offset, align)); ASSERT0(P2PHASE(zio->io_size, align)); } else { /* * For the physical io we allow alignment * to a logical block size. */ uint64_t log_align = 1ULL << vd->vdev_top->vdev_logical_ashift; ASSERT0(P2PHASE(zio->io_offset, log_align)); ASSERT0(P2PHASE(zio->io_size, log_align)); } VERIFY(zio->io_type == ZIO_TYPE_READ || spa_writeable(spa)); /* * If this is a repair I/O, and there's no self-healing involved -- * that is, we're just resilvering what we expect to resilver -- * then don't do the I/O unless zio's txg is actually in vd's DTL. * This prevents spurious resilvering with nested replication. * For example, given a mirror of mirrors, (A+B)+(C+D), if only * A is out of date, we'll read from C+D, then use the data to * resilver A+B -- but we don't actually want to resilver B, just A. * The top-level mirror has no way to know this, so instead we just * discard unnecessary repairs as we work our way down the vdev tree. * The same logic applies to any form of nested replication: * ditto + mirror, RAID-Z + replacing, etc. This covers them all. */ if ((zio->io_flags & ZIO_FLAG_IO_REPAIR) && !(zio->io_flags & ZIO_FLAG_SELF_HEAL) && zio->io_txg != 0 && /* not a delegated i/o */ !vdev_dtl_contains(vd, DTL_PARTIAL, zio->io_txg, 1)) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); zio_vdev_io_bypass(zio); return (ZIO_PIPELINE_CONTINUE); } if (vd->vdev_ops->vdev_op_leaf) { switch (zio->io_type) { case ZIO_TYPE_READ: if (vdev_cache_read(zio)) return (ZIO_PIPELINE_CONTINUE); /* FALLTHROUGH */ case ZIO_TYPE_WRITE: case ZIO_TYPE_FREE: if ((zio = vdev_queue_io(zio)) == NULL) return (ZIO_PIPELINE_STOP); if (!vdev_accessible(vd, zio)) { zio->io_error = SET_ERROR(ENXIO); zio_interrupt(zio); return (ZIO_PIPELINE_STOP); } break; } /* * Note that we ignore repair writes for TRIM because they can * conflict with normal writes. This isn't an issue because, by * definition, we only repair blocks that aren't freed. */ if (zio->io_type == ZIO_TYPE_WRITE && !(zio->io_flags & ZIO_FLAG_IO_REPAIR) && !trim_map_write_start(zio)) return (ZIO_PIPELINE_STOP); } vd->vdev_ops->vdev_op_io_start(zio); return (ZIO_PIPELINE_STOP); } static int zio_vdev_io_done(zio_t *zio) { vdev_t *vd = zio->io_vd; vdev_ops_t *ops = vd ? vd->vdev_ops : &vdev_mirror_ops; boolean_t unexpected_error = B_FALSE; if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); ASSERT(zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_FREE); if (vd != NULL && vd->vdev_ops->vdev_op_leaf && (zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_WRITE || zio->io_type == ZIO_TYPE_FREE)) { if (zio->io_type == ZIO_TYPE_WRITE && !(zio->io_flags & ZIO_FLAG_IO_REPAIR)) trim_map_write_done(zio); vdev_queue_io_done(zio); if (zio->io_type == ZIO_TYPE_WRITE) vdev_cache_write(zio); if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_device_injection(vd, zio, EIO); if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_label_injection(zio, EIO); if (zio->io_error) { if (zio->io_error == ENOTSUP && zio->io_type == ZIO_TYPE_FREE) { /* Not all devices support TRIM. */ } else if (!vdev_accessible(vd, zio)) { zio->io_error = SET_ERROR(ENXIO); } else { unexpected_error = B_TRUE; } } } ops->vdev_op_io_done(zio); if (unexpected_error) VERIFY(vdev_probe(vd, zio) == NULL); return (ZIO_PIPELINE_CONTINUE); } /* * For non-raidz ZIOs, we can just copy aside the bad data read from the * disk, and use that to finish the checksum ereport later. */ static void zio_vsd_default_cksum_finish(zio_cksum_report_t *zcr, const void *good_buf) { /* no processing needed */ zfs_ereport_finish_checksum(zcr, good_buf, zcr->zcr_cbdata, B_FALSE); } /*ARGSUSED*/ void zio_vsd_default_cksum_report(zio_t *zio, zio_cksum_report_t *zcr, void *ignored) { void *buf = zio_buf_alloc(zio->io_size); abd_copy_to_buf(buf, zio->io_abd, zio->io_size); zcr->zcr_cbinfo = zio->io_size; zcr->zcr_cbdata = buf; zcr->zcr_finish = zio_vsd_default_cksum_finish; zcr->zcr_free = zio_buf_free; } static int zio_vdev_io_assess(zio_t *zio) { vdev_t *vd = zio->io_vd; if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); if (vd == NULL && !(zio->io_flags & ZIO_FLAG_CONFIG_WRITER)) spa_config_exit(zio->io_spa, SCL_ZIO, zio); if (zio->io_vsd != NULL) { zio->io_vsd_ops->vsd_free(zio); zio->io_vsd = NULL; } if (zio_injection_enabled && zio->io_error == 0) zio->io_error = zio_handle_fault_injection(zio, EIO); if (zio->io_type == ZIO_TYPE_FREE && zio->io_priority != ZIO_PRIORITY_NOW) { switch (zio->io_error) { case 0: ZIO_TRIM_STAT_INCR(bytes, zio->io_size); ZIO_TRIM_STAT_BUMP(success); break; case EOPNOTSUPP: ZIO_TRIM_STAT_BUMP(unsupported); break; default: ZIO_TRIM_STAT_BUMP(failed); break; } } /* * If the I/O failed, determine whether we should attempt to retry it. * * On retry, we cut in line in the issue queue, since we don't want * compression/checksumming/etc. work to prevent our (cheap) IO reissue. */ if (zio->io_error && vd == NULL && !(zio->io_flags & (ZIO_FLAG_DONT_RETRY | ZIO_FLAG_IO_RETRY))) { ASSERT(!(zio->io_flags & ZIO_FLAG_DONT_QUEUE)); /* not a leaf */ ASSERT(!(zio->io_flags & ZIO_FLAG_IO_BYPASS)); /* not a leaf */ zio->io_error = 0; zio->io_flags |= ZIO_FLAG_IO_RETRY | ZIO_FLAG_DONT_CACHE | ZIO_FLAG_DONT_AGGREGATE; zio->io_stage = ZIO_STAGE_VDEV_IO_START >> 1; zio_taskq_dispatch(zio, ZIO_TASKQ_ISSUE, zio_requeue_io_start_cut_in_line); return (ZIO_PIPELINE_STOP); } /* * If we got an error on a leaf device, convert it to ENXIO * if the device is not accessible at all. */ if (zio->io_error && vd != NULL && vd->vdev_ops->vdev_op_leaf && !vdev_accessible(vd, zio)) zio->io_error = SET_ERROR(ENXIO); /* * If we can't write to an interior vdev (mirror or RAID-Z), * set vdev_cant_write so that we stop trying to allocate from it. */ if (zio->io_error == ENXIO && zio->io_type == ZIO_TYPE_WRITE && vd != NULL && !vd->vdev_ops->vdev_op_leaf) { vd->vdev_cant_write = B_TRUE; } /* * If a cache flush returns ENOTSUP or ENOTTY, we know that no future * attempts will ever succeed. In this case we set a persistent bit so * that we don't bother with it in the future. */ if ((zio->io_error == ENOTSUP || zio->io_error == ENOTTY) && zio->io_type == ZIO_TYPE_IOCTL && zio->io_cmd == DKIOCFLUSHWRITECACHE && vd != NULL) vd->vdev_nowritecache = B_TRUE; if (zio->io_error) zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (vd != NULL && vd->vdev_ops->vdev_op_leaf && zio->io_physdone != NULL) { ASSERT(!(zio->io_flags & ZIO_FLAG_DELEGATED)); ASSERT(zio->io_child_type == ZIO_CHILD_VDEV); zio->io_physdone(zio->io_logical); } return (ZIO_PIPELINE_CONTINUE); } void zio_vdev_io_reissue(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); ASSERT(zio->io_error == 0); zio->io_stage >>= 1; } void zio_vdev_io_redone(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_DONE); zio->io_stage >>= 1; } void zio_vdev_io_bypass(zio_t *zio) { ASSERT(zio->io_stage == ZIO_STAGE_VDEV_IO_START); ASSERT(zio->io_error == 0); zio->io_flags |= ZIO_FLAG_IO_BYPASS; zio->io_stage = ZIO_STAGE_VDEV_IO_ASSESS >> 1; } /* * ========================================================================== * Generate and verify checksums * ========================================================================== */ static int zio_checksum_generate(zio_t *zio) { blkptr_t *bp = zio->io_bp; enum zio_checksum checksum; if (bp == NULL) { /* * This is zio_write_phys(). * We're either generating a label checksum, or none at all. */ checksum = zio->io_prop.zp_checksum; if (checksum == ZIO_CHECKSUM_OFF) return (ZIO_PIPELINE_CONTINUE); ASSERT(checksum == ZIO_CHECKSUM_LABEL); } else { if (BP_IS_GANG(bp) && zio->io_child_type == ZIO_CHILD_GANG) { ASSERT(!IO_IS_ALLOCATING(zio)); checksum = ZIO_CHECKSUM_GANG_HEADER; } else { checksum = BP_GET_CHECKSUM(bp); } } zio_checksum_compute(zio, checksum, zio->io_abd, zio->io_size); return (ZIO_PIPELINE_CONTINUE); } static int zio_checksum_verify(zio_t *zio) { zio_bad_cksum_t info; blkptr_t *bp = zio->io_bp; int error; ASSERT(zio->io_vd != NULL); if (bp == NULL) { /* * This is zio_read_phys(). * We're either verifying a label checksum, or nothing at all. */ if (zio->io_prop.zp_checksum == ZIO_CHECKSUM_OFF) return (ZIO_PIPELINE_CONTINUE); ASSERT(zio->io_prop.zp_checksum == ZIO_CHECKSUM_LABEL); } if ((error = zio_checksum_error(zio, &info)) != 0) { zio->io_error = error; if (error == ECKSUM && !(zio->io_flags & ZIO_FLAG_SPECULATIVE)) { zfs_ereport_start_checksum(zio->io_spa, zio->io_vd, zio, zio->io_offset, zio->io_size, NULL, &info); } } return (ZIO_PIPELINE_CONTINUE); } /* * Called by RAID-Z to ensure we don't compute the checksum twice. */ void zio_checksum_verified(zio_t *zio) { zio->io_pipeline &= ~ZIO_STAGE_CHECKSUM_VERIFY; } /* * ========================================================================== * Error rank. Error are ranked in the order 0, ENXIO, ECKSUM, EIO, other. * An error of 0 indicates success. ENXIO indicates whole-device failure, * which may be transient (e.g. unplugged) or permament. ECKSUM and EIO * indicate errors that are specific to one I/O, and most likely permanent. * Any other error is presumed to be worse because we weren't expecting it. * ========================================================================== */ int zio_worst_error(int e1, int e2) { static int zio_error_rank[] = { 0, ENXIO, ECKSUM, EIO }; int r1, r2; for (r1 = 0; r1 < sizeof (zio_error_rank) / sizeof (int); r1++) if (e1 == zio_error_rank[r1]) break; for (r2 = 0; r2 < sizeof (zio_error_rank) / sizeof (int); r2++) if (e2 == zio_error_rank[r2]) break; return (r1 > r2 ? e1 : e2); } /* * ========================================================================== * I/O completion * ========================================================================== */ static int zio_ready(zio_t *zio) { blkptr_t *bp = zio->io_bp; zio_t *pio, *pio_next; zio_link_t *zl = NULL; if (zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_READY) || zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_READY)) return (ZIO_PIPELINE_STOP); if (zio->io_ready) { ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(bp->blk_birth == zio->io_txg || BP_IS_HOLE(bp) || (zio->io_flags & ZIO_FLAG_NOPWRITE)); ASSERT(zio->io_children[ZIO_CHILD_GANG][ZIO_WAIT_READY] == 0); zio->io_ready(zio); } if (bp != NULL && bp != &zio->io_bp_copy) zio->io_bp_copy = *bp; if (zio->io_error != 0) { zio->io_pipeline = ZIO_INTERLOCK_PIPELINE; if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(IO_IS_ALLOCATING(zio)); ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); /* * We were unable to allocate anything, unreserve and * issue the next I/O to allocate. */ metaslab_class_throttle_unreserve( spa_normal_class(zio->io_spa), zio->io_prop.zp_copies, zio); zio_allocate_dispatch(zio->io_spa); } } mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_READY] = 1; pio = zio_walk_parents(zio, &zl); mutex_exit(&zio->io_lock); /* * As we notify zio's parents, new parents could be added. * New parents go to the head of zio's io_parent_list, however, * so we will (correctly) not notify them. The remainder of zio's * io_parent_list, from 'pio_next' onward, cannot change because * all parents must wait for us to be done before they can be done. */ for (; pio != NULL; pio = pio_next) { pio_next = zio_walk_parents(zio, &zl); zio_notify_parent(pio, zio, ZIO_WAIT_READY); } if (zio->io_flags & ZIO_FLAG_NODATA) { if (BP_IS_GANG(bp)) { zio->io_flags &= ~ZIO_FLAG_NODATA; } else { ASSERT((uintptr_t)zio->io_abd < SPA_MAXBLOCKSIZE); zio->io_pipeline &= ~ZIO_VDEV_IO_STAGES; } } if (zio_injection_enabled && zio->io_spa->spa_syncing_txg == zio->io_txg) zio_handle_ignored_writes(zio); return (ZIO_PIPELINE_CONTINUE); } /* * Update the allocation throttle accounting. */ static void zio_dva_throttle_done(zio_t *zio) { zio_t *lio = zio->io_logical; zio_t *pio = zio_unique_parent(zio); vdev_t *vd = zio->io_vd; int flags = METASLAB_ASYNC_ALLOC; ASSERT3P(zio->io_bp, !=, NULL); ASSERT3U(zio->io_type, ==, ZIO_TYPE_WRITE); ASSERT3U(zio->io_priority, ==, ZIO_PRIORITY_ASYNC_WRITE); ASSERT3U(zio->io_child_type, ==, ZIO_CHILD_VDEV); ASSERT(vd != NULL); ASSERT3P(vd, ==, vd->vdev_top); ASSERT(!(zio->io_flags & (ZIO_FLAG_IO_REPAIR | ZIO_FLAG_IO_RETRY))); ASSERT(zio->io_flags & ZIO_FLAG_IO_ALLOCATING); ASSERT(!(lio->io_flags & ZIO_FLAG_IO_REWRITE)); ASSERT(!(lio->io_orig_flags & ZIO_FLAG_NODATA)); /* * Parents of gang children can have two flavors -- ones that * allocated the gang header (will have ZIO_FLAG_IO_REWRITE set) * and ones that allocated the constituent blocks. The allocation * throttle needs to know the allocating parent zio so we must find * it here. */ if (pio->io_child_type == ZIO_CHILD_GANG) { /* * If our parent is a rewrite gang child then our grandparent * would have been the one that performed the allocation. */ if (pio->io_flags & ZIO_FLAG_IO_REWRITE) pio = zio_unique_parent(pio); flags |= METASLAB_GANG_CHILD; } ASSERT(IO_IS_ALLOCATING(pio)); ASSERT3P(zio, !=, zio->io_logical); ASSERT(zio->io_logical != NULL); ASSERT(!(zio->io_flags & ZIO_FLAG_IO_REPAIR)); ASSERT0(zio->io_flags & ZIO_FLAG_NOPWRITE); mutex_enter(&pio->io_lock); metaslab_group_alloc_decrement(zio->io_spa, vd->vdev_id, pio, flags); mutex_exit(&pio->io_lock); metaslab_class_throttle_unreserve(spa_normal_class(zio->io_spa), 1, pio); /* * Call into the pipeline to see if there is more work that * needs to be done. If there is work to be done it will be * dispatched to another taskq thread. */ zio_allocate_dispatch(zio->io_spa); } static int zio_done(zio_t *zio) { spa_t *spa = zio->io_spa; zio_t *lio = zio->io_logical; blkptr_t *bp = zio->io_bp; vdev_t *vd = zio->io_vd; uint64_t psize = zio->io_size; zio_t *pio, *pio_next; metaslab_class_t *mc = spa_normal_class(spa); zio_link_t *zl = NULL; /* * If our children haven't all completed, * wait for them and then repeat this pipeline stage. */ if (zio_wait_for_children(zio, ZIO_CHILD_VDEV, ZIO_WAIT_DONE) || zio_wait_for_children(zio, ZIO_CHILD_GANG, ZIO_WAIT_DONE) || zio_wait_for_children(zio, ZIO_CHILD_DDT, ZIO_WAIT_DONE) || zio_wait_for_children(zio, ZIO_CHILD_LOGICAL, ZIO_WAIT_DONE)) return (ZIO_PIPELINE_STOP); /* * If the allocation throttle is enabled, then update the accounting. * We only track child I/Os that are part of an allocating async * write. We must do this since the allocation is performed * by the logical I/O but the actual write is done by child I/Os. */ if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING && zio->io_child_type == ZIO_CHILD_VDEV) { ASSERT(mc->mc_alloc_throttle_enabled); zio_dva_throttle_done(zio); } /* * If the allocation throttle is enabled, verify that * we have decremented the refcounts for every I/O that was throttled. */ if (zio->io_flags & ZIO_FLAG_IO_ALLOCATING) { ASSERT(zio->io_type == ZIO_TYPE_WRITE); ASSERT(zio->io_priority == ZIO_PRIORITY_ASYNC_WRITE); ASSERT(bp != NULL); metaslab_group_alloc_verify(spa, zio->io_bp, zio); VERIFY(refcount_not_held(&mc->mc_alloc_slots, zio)); } for (int c = 0; c < ZIO_CHILD_TYPES; c++) for (int w = 0; w < ZIO_WAIT_TYPES; w++) ASSERT(zio->io_children[c][w] == 0); if (bp != NULL && !BP_IS_EMBEDDED(bp)) { ASSERT(bp->blk_pad[0] == 0); ASSERT(bp->blk_pad[1] == 0); ASSERT(bcmp(bp, &zio->io_bp_copy, sizeof (blkptr_t)) == 0 || (bp == zio_unique_parent(zio)->io_bp)); if (zio->io_type == ZIO_TYPE_WRITE && !BP_IS_HOLE(bp) && zio->io_bp_override == NULL && !(zio->io_flags & ZIO_FLAG_IO_REPAIR)) { ASSERT(!BP_SHOULD_BYTESWAP(bp)); ASSERT3U(zio->io_prop.zp_copies, <=, BP_GET_NDVAS(bp)); ASSERT(BP_COUNT_GANG(bp) == 0 || (BP_COUNT_GANG(bp) == BP_GET_NDVAS(bp))); } if (zio->io_flags & ZIO_FLAG_NOPWRITE) VERIFY(BP_EQUAL(bp, &zio->io_bp_orig)); } /* * If there were child vdev/gang/ddt errors, they apply to us now. */ zio_inherit_child_errors(zio, ZIO_CHILD_VDEV); zio_inherit_child_errors(zio, ZIO_CHILD_GANG); zio_inherit_child_errors(zio, ZIO_CHILD_DDT); /* * If the I/O on the transformed data was successful, generate any * checksum reports now while we still have the transformed data. */ if (zio->io_error == 0) { while (zio->io_cksum_report != NULL) { zio_cksum_report_t *zcr = zio->io_cksum_report; uint64_t align = zcr->zcr_align; uint64_t asize = P2ROUNDUP(psize, align); char *abuf = NULL; abd_t *adata = zio->io_abd; if (asize != psize) { adata = abd_alloc_linear(asize, B_TRUE); abd_copy(adata, zio->io_abd, psize); abd_zero_off(adata, psize, asize - psize); } if (adata != NULL) abuf = abd_borrow_buf_copy(adata, asize); zio->io_cksum_report = zcr->zcr_next; zcr->zcr_next = NULL; zcr->zcr_finish(zcr, abuf); zfs_ereport_free_checksum(zcr); if (adata != NULL) abd_return_buf(adata, abuf, asize); if (asize != psize) abd_free(adata); } } zio_pop_transforms(zio); /* note: may set zio->io_error */ vdev_stat_update(zio, psize); if (zio->io_error) { /* * If this I/O is attached to a particular vdev, * generate an error message describing the I/O failure * at the block level. We ignore these errors if the * device is currently unavailable. */ if (zio->io_error != ECKSUM && vd != NULL && !vdev_is_dead(vd)) zfs_ereport_post(FM_EREPORT_ZFS_IO, spa, vd, zio, 0, 0); if ((zio->io_error == EIO || !(zio->io_flags & (ZIO_FLAG_SPECULATIVE | ZIO_FLAG_DONT_PROPAGATE))) && zio == lio) { /* * For logical I/O requests, tell the SPA to log the * error and generate a logical data ereport. */ spa_log_error(spa, zio); zfs_ereport_post(FM_EREPORT_ZFS_DATA, spa, NULL, zio, 0, 0); } } if (zio->io_error && zio == lio) { /* * Determine whether zio should be reexecuted. This will * propagate all the way to the root via zio_notify_parent(). */ ASSERT(vd == NULL && bp != NULL); ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); if (IO_IS_ALLOCATING(zio) && !(zio->io_flags & ZIO_FLAG_CANFAIL)) { if (zio->io_error != ENOSPC) zio->io_reexecute |= ZIO_REEXECUTE_NOW; else zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; } if ((zio->io_type == ZIO_TYPE_READ || zio->io_type == ZIO_TYPE_FREE) && !(zio->io_flags & ZIO_FLAG_SCAN_THREAD) && zio->io_error == ENXIO && spa_load_state(spa) == SPA_LOAD_NONE && spa_get_failmode(spa) != ZIO_FAILURE_MODE_CONTINUE) zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; if (!(zio->io_flags & ZIO_FLAG_CANFAIL) && !zio->io_reexecute) zio->io_reexecute |= ZIO_REEXECUTE_SUSPEND; /* * Here is a possibly good place to attempt to do * either combinatorial reconstruction or error correction * based on checksums. It also might be a good place * to send out preliminary ereports before we suspend * processing. */ } /* * If there were logical child errors, they apply to us now. * We defer this until now to avoid conflating logical child * errors with errors that happened to the zio itself when * updating vdev stats and reporting FMA events above. */ zio_inherit_child_errors(zio, ZIO_CHILD_LOGICAL); if ((zio->io_error || zio->io_reexecute) && IO_IS_ALLOCATING(zio) && zio->io_gang_leader == zio && !(zio->io_flags & (ZIO_FLAG_IO_REWRITE | ZIO_FLAG_NOPWRITE))) zio_dva_unallocate(zio, zio->io_gang_tree, bp); zio_gang_tree_free(&zio->io_gang_tree); /* * Godfather I/Os should never suspend. */ if ((zio->io_flags & ZIO_FLAG_GODFATHER) && (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) zio->io_reexecute = 0; if (zio->io_reexecute) { /* * This is a logical I/O that wants to reexecute. * * Reexecute is top-down. When an i/o fails, if it's not * the root, it simply notifies its parent and sticks around. * The parent, seeing that it still has children in zio_done(), * does the same. This percolates all the way up to the root. * The root i/o will reexecute or suspend the entire tree. * * This approach ensures that zio_reexecute() honors * all the original i/o dependency relationships, e.g. * parents not executing until children are ready. */ ASSERT(zio->io_child_type == ZIO_CHILD_LOGICAL); zio->io_gang_leader = NULL; mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_DONE] = 1; mutex_exit(&zio->io_lock); /* * "The Godfather" I/O monitors its children but is * not a true parent to them. It will track them through * the pipeline but severs its ties whenever they get into * trouble (e.g. suspended). This allows "The Godfather" * I/O to return status without blocking. */ zl = NULL; for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) { zio_link_t *remove_zl = zl; pio_next = zio_walk_parents(zio, &zl); if ((pio->io_flags & ZIO_FLAG_GODFATHER) && (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND)) { zio_remove_child(pio, zio, remove_zl); zio_notify_parent(pio, zio, ZIO_WAIT_DONE); } } if ((pio = zio_unique_parent(zio)) != NULL) { /* * We're not a root i/o, so there's nothing to do * but notify our parent. Don't propagate errors * upward since we haven't permanently failed yet. */ ASSERT(!(zio->io_flags & ZIO_FLAG_GODFATHER)); zio->io_flags |= ZIO_FLAG_DONT_PROPAGATE; zio_notify_parent(pio, zio, ZIO_WAIT_DONE); } else if (zio->io_reexecute & ZIO_REEXECUTE_SUSPEND) { /* * We'd fail again if we reexecuted now, so suspend * until conditions improve (e.g. device comes online). */ zio_suspend(spa, zio); } else { /* * Reexecution is potentially a huge amount of work. * Hand it off to the otherwise-unused claim taskq. */ #if defined(illumos) || !defined(_KERNEL) ASSERT(zio->io_tqent.tqent_next == NULL); #else ASSERT(zio->io_tqent.tqent_task.ta_pending == 0); #endif spa_taskq_dispatch_ent(spa, ZIO_TYPE_CLAIM, ZIO_TASKQ_ISSUE, (task_func_t *)zio_reexecute, zio, 0, &zio->io_tqent); } return (ZIO_PIPELINE_STOP); } ASSERT(zio->io_child_count == 0); ASSERT(zio->io_reexecute == 0); ASSERT(zio->io_error == 0 || (zio->io_flags & ZIO_FLAG_CANFAIL)); /* * Report any checksum errors, since the I/O is complete. */ while (zio->io_cksum_report != NULL) { zio_cksum_report_t *zcr = zio->io_cksum_report; zio->io_cksum_report = zcr->zcr_next; zcr->zcr_next = NULL; zcr->zcr_finish(zcr, NULL); zfs_ereport_free_checksum(zcr); } /* * It is the responsibility of the done callback to ensure that this * particular zio is no longer discoverable for adoption, and as * such, cannot acquire any new parents. */ if (zio->io_done) zio->io_done(zio); mutex_enter(&zio->io_lock); zio->io_state[ZIO_WAIT_DONE] = 1; mutex_exit(&zio->io_lock); zl = NULL; for (pio = zio_walk_parents(zio, &zl); pio != NULL; pio = pio_next) { zio_link_t *remove_zl = zl; pio_next = zio_walk_parents(zio, &zl); zio_remove_child(pio, zio, remove_zl); zio_notify_parent(pio, zio, ZIO_WAIT_DONE); } if (zio->io_waiter != NULL) { mutex_enter(&zio->io_lock); zio->io_executor = NULL; cv_broadcast(&zio->io_cv); mutex_exit(&zio->io_lock); } else { zio_destroy(zio); } return (ZIO_PIPELINE_STOP); } /* * ========================================================================== * I/O pipeline definition * ========================================================================== */ static zio_pipe_stage_t *zio_pipeline[] = { NULL, zio_read_bp_init, zio_write_bp_init, zio_free_bp_init, zio_issue_async, zio_write_compress, zio_checksum_generate, zio_nop_write, zio_ddt_read_start, zio_ddt_read_done, zio_ddt_write, zio_ddt_free, zio_gang_assemble, zio_gang_issue, zio_dva_throttle, zio_dva_allocate, zio_dva_free, zio_dva_claim, zio_ready, zio_vdev_io_start, zio_vdev_io_done, zio_vdev_io_assess, zio_checksum_verify, zio_done }; /* * Compare two zbookmark_phys_t's to see which we would reach first in a * pre-order traversal of the object tree. * * This is simple in every case aside from the meta-dnode object. For all other * objects, we traverse them in order (object 1 before object 2, and so on). * However, all of these objects are traversed while traversing object 0, since * the data it points to is the list of objects. Thus, we need to convert to a * canonical representation so we can compare meta-dnode bookmarks to * non-meta-dnode bookmarks. * * We do this by calculating "equivalents" for each field of the zbookmark. * zbookmarks outside of the meta-dnode use their own object and level, and * calculate the level 0 equivalent (the first L0 blkid that is contained in the * blocks this bookmark refers to) by multiplying their blkid by their span * (the number of L0 blocks contained within one block at their level). * zbookmarks inside the meta-dnode calculate their object equivalent * (which is L0equiv * dnodes per data block), use 0 for their L0equiv, and use * level + 1<<31 (any value larger than a level could ever be) for their level. * This causes them to always compare before a bookmark in their object * equivalent, compare appropriately to bookmarks in other objects, and to * compare appropriately to other bookmarks in the meta-dnode. */ int zbookmark_compare(uint16_t dbss1, uint8_t ibs1, uint16_t dbss2, uint8_t ibs2, const zbookmark_phys_t *zb1, const zbookmark_phys_t *zb2) { /* * These variables represent the "equivalent" values for the zbookmark, * after converting zbookmarks inside the meta dnode to their * normal-object equivalents. */ uint64_t zb1obj, zb2obj; uint64_t zb1L0, zb2L0; uint64_t zb1level, zb2level; if (zb1->zb_object == zb2->zb_object && zb1->zb_level == zb2->zb_level && zb1->zb_blkid == zb2->zb_blkid) return (0); /* * BP_SPANB calculates the span in blocks. */ zb1L0 = (zb1->zb_blkid) * BP_SPANB(ibs1, zb1->zb_level); zb2L0 = (zb2->zb_blkid) * BP_SPANB(ibs2, zb2->zb_level); if (zb1->zb_object == DMU_META_DNODE_OBJECT) { zb1obj = zb1L0 * (dbss1 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT)); zb1L0 = 0; zb1level = zb1->zb_level + COMPARE_META_LEVEL; } else { zb1obj = zb1->zb_object; zb1level = zb1->zb_level; } if (zb2->zb_object == DMU_META_DNODE_OBJECT) { zb2obj = zb2L0 * (dbss2 << (SPA_MINBLOCKSHIFT - DNODE_SHIFT)); zb2L0 = 0; zb2level = zb2->zb_level + COMPARE_META_LEVEL; } else { zb2obj = zb2->zb_object; zb2level = zb2->zb_level; } /* Now that we have a canonical representation, do the comparison. */ if (zb1obj != zb2obj) return (zb1obj < zb2obj ? -1 : 1); else if (zb1L0 != zb2L0) return (zb1L0 < zb2L0 ? -1 : 1); else if (zb1level != zb2level) return (zb1level > zb2level ? -1 : 1); /* * This can (theoretically) happen if the bookmarks have the same object * and level, but different blkids, if the block sizes are not the same. * There is presently no way to change the indirect block sizes */ return (0); } /* * This function checks the following: given that last_block is the place that * our traversal stopped last time, does that guarantee that we've visited * every node under subtree_root? Therefore, we can't just use the raw output * of zbookmark_compare. We have to pass in a modified version of * subtree_root; by incrementing the block id, and then checking whether * last_block is before or equal to that, we can tell whether or not having * visited last_block implies that all of subtree_root's children have been * visited. */ boolean_t zbookmark_subtree_completed(const dnode_phys_t *dnp, const zbookmark_phys_t *subtree_root, const zbookmark_phys_t *last_block) { zbookmark_phys_t mod_zb = *subtree_root; mod_zb.zb_blkid++; ASSERT(last_block->zb_level == 0); /* The objset_phys_t isn't before anything. */ if (dnp == NULL) return (B_FALSE); /* * We pass in 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT) for the * data block size in sectors, because that variable is only used if * the bookmark refers to a block in the meta-dnode. Since we don't * know without examining it what object it refers to, and there's no * harm in passing in this value in other cases, we always pass it in. * * We pass in 0 for the indirect block size shift because zb2 must be * level 0. The indirect block size is only used to calculate the span * of the bookmark, but since the bookmark must be level 0, the span is * always 1, so the math works out. * * If you make changes to how the zbookmark_compare code works, be sure * to make sure that this code still works afterwards. */ return (zbookmark_compare(dnp->dn_datablkszsec, dnp->dn_indblkshift, 1ULL << (DNODE_BLOCK_SHIFT - SPA_MINBLOCKSHIFT), 0, &mod_zb, last_block) <= 0); } Index: head/sys/cddl/contrib/opensolaris =================================================================== --- head/sys/cddl/contrib/opensolaris (revision 328227) +++ head/sys/cddl/contrib/opensolaris (revision 328228) Property changes on: head/sys/cddl/contrib/opensolaris ___________________________________________________________________ Modified: svn:mergeinfo ## -0,0 +0,1 ## Merged /vendor-sys/illumos/dist:r328227