diff options
Diffstat (limited to 'module/zfs')
| -rw-r--r-- | module/zfs/dmu.c | 52 | ||||
| -rw-r--r-- | module/zfs/dmu_recv.c | 229 | ||||
| -rw-r--r-- | module/zfs/zfs_ioctl.c | 36 | ||||
| -rw-r--r-- | module/zfs/zfs_onexit.c | 79 |
4 files changed, 19 insertions, 377 deletions
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c index 6eb935720..a21ac8d74 100644 --- a/module/zfs/dmu.c +++ b/module/zfs/dmu.c @@ -20,7 +20,7 @@ */ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2011, 2018 by Delphix. All rights reserved. + * Copyright (c) 2011, 2020 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013, Joyent, Inc. All rights reserved. * Copyright (c) 2016, Nexenta Systems, Inc. All rights reserved. @@ -1560,56 +1560,6 @@ dmu_return_arcbuf(arc_buf_t *buf) arc_buf_destroy(buf, FTAG); } -void -dmu_copy_from_buf(objset_t *os, uint64_t object, uint64_t offset, - dmu_buf_t *handle, dmu_tx_t *tx) -{ - dmu_buf_t *dst_handle; - dmu_buf_impl_t *dstdb; - dmu_buf_impl_t *srcdb = (dmu_buf_impl_t *)handle; - dmu_object_type_t type; - arc_buf_t *abuf; - uint64_t datalen; - boolean_t byteorder; - uint8_t salt[ZIO_DATA_SALT_LEN]; - uint8_t iv[ZIO_DATA_IV_LEN]; - uint8_t mac[ZIO_DATA_MAC_LEN]; - - ASSERT3P(srcdb->db_buf, !=, NULL); - - /* hold the db that we want to write to */ - VERIFY0(dmu_buf_hold(os, object, offset, FTAG, &dst_handle, - DMU_READ_NO_DECRYPT)); - dstdb = (dmu_buf_impl_t *)dst_handle; - datalen = arc_buf_size(srcdb->db_buf); - - DB_DNODE_ENTER(dstdb); - type = DB_DNODE(dstdb)->dn_type; - DB_DNODE_EXIT(dstdb); - - /* allocated an arc buffer that matches the type of srcdb->db_buf */ - if (arc_is_encrypted(srcdb->db_buf)) { - arc_get_raw_params(srcdb->db_buf, &byteorder, salt, iv, mac); - abuf = arc_loan_raw_buf(os->os_spa, dmu_objset_id(os), - byteorder, salt, iv, mac, type, - datalen, arc_buf_lsize(srcdb->db_buf), - arc_get_compression(srcdb->db_buf)); - } else { - /* we won't get a compressed db back from dmu_buf_hold() */ - ASSERT3U(arc_get_compression(srcdb->db_buf), - ==, ZIO_COMPRESS_OFF); - abuf = arc_loan_buf(os->os_spa, - DMU_OT_IS_METADATA(type), datalen); - } - - ASSERT3U(datalen, ==, arc_buf_size(abuf)); - - /* copy the data to the new buffer and assign it to the dstdb */ - bcopy(srcdb->db_buf->b_data, abuf->b_data, datalen); - dbuf_assign_arcbuf(dstdb, abuf, tx); - dmu_buf_rele(dst_handle, FTAG); -} - /* * When possible directly assign passed loaned arc buffer to a dbuf. * If this is not possible copy the contents of passed arc buf via diff --git a/module/zfs/dmu_recv.c b/module/zfs/dmu_recv.c index 9c82cecc6..817c85b93 100644 --- a/module/zfs/dmu_recv.c +++ b/module/zfs/dmu_recv.c @@ -21,7 +21,7 @@ /* * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved. * Copyright 2011 Nexenta Systems, Inc. All rights reserved. - * Copyright (c) 2011, 2018 by Delphix. All rights reserved. + * Copyright (c) 2011, 2020 by Delphix. All rights reserved. * Copyright (c) 2014, Joyent, Inc. All rights reserved. * Copyright 2014 HybridCluster. All rights reserved. * Copyright (c) 2018, loli10K <[email protected]>. All rights reserved. @@ -101,8 +101,6 @@ struct receive_writer_arg { boolean_t done; int err; - /* A map from guid to dataset to help handle dedup'd streams. */ - avl_tree_t *guid_to_ds_map; boolean_t resumable; boolean_t raw; /* DMU_BACKUP_FEATURE_RAW set */ boolean_t spill; /* DRR_FLAG_SPILL_BLOCK set */ @@ -123,13 +121,6 @@ struct receive_writer_arg { boolean_t or_byteorder; }; -typedef struct guid_map_entry { - uint64_t guid; - boolean_t raw; - objset_t *gme_os; - avl_node_t avlnode; -} guid_map_entry_t; - typedef struct dmu_recv_begin_arg { const char *drba_origin; dmu_recv_cookie_t *drba_cookie; @@ -1213,38 +1204,6 @@ dmu_recv_begin(char *tofs, char *tosnap, dmu_replay_record_t *drr_begin, } static int -guid_compare(const void *arg1, const void *arg2) -{ - const guid_map_entry_t *gmep1 = (const guid_map_entry_t *)arg1; - const guid_map_entry_t *gmep2 = (const guid_map_entry_t *)arg2; - - return (TREE_CMP(gmep1->guid, gmep2->guid)); -} - -static void -free_guid_map_onexit(void *arg) -{ - avl_tree_t *ca = arg; - void *cookie = NULL; - guid_map_entry_t *gmep; - - while ((gmep = avl_destroy_nodes(ca, &cookie)) != NULL) { - ds_hold_flags_t dsflags = DS_HOLD_FLAG_DECRYPT; - - if (gmep->raw) { - gmep->gme_os->os_raw_receive = B_FALSE; - dsflags &= ~DS_HOLD_FLAG_DECRYPT; - } - - dsl_dataset_disown(gmep->gme_os->os_dsl_dataset, - dsflags, gmep); - kmem_free(gmep, sizeof (guid_map_entry_t)); - } - avl_destroy(ca); - kmem_free(ca, sizeof (avl_tree_t)); -} - -static int receive_read(dmu_recv_cookie_t *drc, int len, void *buf) { int done = 0; @@ -1845,81 +1804,6 @@ receive_process_write_record(struct receive_writer_arg *rwa, return (EAGAIN); } -/* - * Handle a DRR_WRITE_BYREF record. This record is used in dedup'ed - * streams to refer to a copy of the data that is already on the - * system because it came in earlier in the stream. This function - * finds the earlier copy of the data, and uses that copy instead of - * data from the stream to fulfill this write. - */ -noinline static int -receive_write_byref(struct receive_writer_arg *rwa, - struct drr_write_byref *drrwbr) -{ - dmu_tx_t *tx; - int err; - guid_map_entry_t gmesrch; - guid_map_entry_t *gmep; - avl_index_t where; - objset_t *ref_os = NULL; - int flags = DMU_READ_PREFETCH; - dmu_buf_t *dbp; - - if (drrwbr->drr_offset + drrwbr->drr_length < drrwbr->drr_offset) - return (SET_ERROR(EINVAL)); - - /* - * If the GUID of the referenced dataset is different from the - * GUID of the target dataset, find the referenced dataset. - */ - if (drrwbr->drr_toguid != drrwbr->drr_refguid) { - gmesrch.guid = drrwbr->drr_refguid; - if ((gmep = avl_find(rwa->guid_to_ds_map, &gmesrch, - &where)) == NULL) { - return (SET_ERROR(EINVAL)); - } - ref_os = gmep->gme_os; - } else { - ref_os = rwa->os; - } - - if (drrwbr->drr_object > rwa->max_object) - rwa->max_object = drrwbr->drr_object; - - if (rwa->raw) - flags |= DMU_READ_NO_DECRYPT; - - /* may return either a regular db or an encrypted one */ - err = dmu_buf_hold(ref_os, drrwbr->drr_refobject, - drrwbr->drr_refoffset, FTAG, &dbp, flags); - if (err != 0) - return (err); - - tx = dmu_tx_create(rwa->os); - - dmu_tx_hold_write(tx, drrwbr->drr_object, - drrwbr->drr_offset, drrwbr->drr_length); - err = dmu_tx_assign(tx, TXG_WAIT); - if (err != 0) { - dmu_tx_abort(tx); - return (err); - } - - if (rwa->raw) { - dmu_copy_from_buf(rwa->os, drrwbr->drr_object, - drrwbr->drr_offset, dbp, tx); - } else { - dmu_write(rwa->os, drrwbr->drr_object, - drrwbr->drr_offset, drrwbr->drr_length, dbp->db_data, tx); - } - dmu_buf_rele(dbp, FTAG); - - /* See comment in restore_write. */ - save_resume_state(rwa, drrwbr->drr_object, drrwbr->drr_offset, tx); - dmu_tx_commit(tx); - return (0); -} - static int receive_write_embedded(struct receive_writer_arg *rwa, struct drr_write_embedded *drrwe, void *data) @@ -2607,13 +2491,6 @@ receive_process_record(struct receive_writer_arg *rwa, } break; } - case DRR_WRITE_BYREF: - { - struct drr_write_byref *drrwbr = - &rrd->header.drr_u.drr_write_byref; - err = receive_write_byref(rwa, drrwbr); - break; - } case DRR_WRITE_EMBEDDED: { struct drr_write_embedded *drrwe = @@ -2754,8 +2631,7 @@ resume_check(dmu_recv_cookie_t *drc, nvlist_t *begin_nvl) * NB: callers *must* call dmu_recv_end() if this succeeds. */ int -dmu_recv_stream(dmu_recv_cookie_t *drc, int cleanup_fd, - uint64_t *action_handlep, offset_t *voffp) +dmu_recv_stream(dmu_recv_cookie_t *drc, offset_t *voffp) { int err = 0; struct receive_writer_arg *rwa = kmem_zalloc(sizeof (*rwa), KM_SLEEP); @@ -2779,41 +2655,6 @@ dmu_recv_stream(dmu_recv_cookie_t *drc, int cleanup_fd, ASSERT0(drc->drc_os->os_encrypted && (drc->drc_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)); - /* if this stream is dedup'ed, set up the avl tree for guid mapping */ - if (drc->drc_featureflags & DMU_BACKUP_FEATURE_DEDUP) { - minor_t minor; - - if (cleanup_fd == -1) { - err = SET_ERROR(EBADF); - goto out; - } - err = zfs_onexit_fd_hold(cleanup_fd, &minor); - if (err != 0) { - cleanup_fd = -1; - goto out; - } - - if (*action_handlep == 0) { - rwa->guid_to_ds_map = - kmem_alloc(sizeof (avl_tree_t), KM_SLEEP); - avl_create(rwa->guid_to_ds_map, guid_compare, - sizeof (guid_map_entry_t), - offsetof(guid_map_entry_t, avlnode)); - err = zfs_onexit_add_cb(minor, - free_guid_map_onexit, rwa->guid_to_ds_map, - action_handlep); - if (err != 0) - goto out; - } else { - err = zfs_onexit_cb_data(minor, *action_handlep, - (void **)&rwa->guid_to_ds_map); - if (err != 0) - goto out; - } - - drc->drc_guid_to_ds_map = rwa->guid_to_ds_map; - } - /* handle DSL encryption key payload */ if (drc->drc_featureflags & DMU_BACKUP_FEATURE_RAW) { nvlist_t *keynvl = NULL; @@ -2980,9 +2821,6 @@ out: kmem_free(rwa, sizeof (*rwa)); nvlist_free(drc->drc_begin_nvl); - if ((drc->drc_featureflags & DMU_BACKUP_FEATURE_DEDUP) && - (cleanup_fd != -1)) - zfs_onexit_fd_rele(cleanup_fd); if (err != 0) { /* @@ -3083,6 +2921,7 @@ dmu_recv_end_sync(void *arg, dmu_tx_t *tx) dmu_recv_cookie_t *drc = arg; dsl_pool_t *dp = dmu_tx_pool(tx); boolean_t encrypted = drc->drc_ds->ds_dir->dd_crypto_obj != 0; + uint64_t newsnapobj; spa_history_log_internal_ds(drc->drc_ds, "finish receiving", tx, "snap=%s", drc->drc_tosnap); @@ -3148,7 +2987,7 @@ dmu_recv_end_sync(void *arg, dmu_tx_t *tx) dsl_dataset_phys(origin_head)->ds_flags &= ~DS_FLAG_INCONSISTENT; - drc->drc_newsnapobj = + newsnapobj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj; dsl_dataset_rele(origin_head, FTAG); @@ -3188,7 +3027,7 @@ dmu_recv_end_sync(void *arg, dmu_tx_t *tx) (void) zap_remove(dp->dp_meta_objset, ds->ds_object, DS_FIELD_RESUME_REDACT_BOOKMARK_SNAPS, tx); } - drc->drc_newsnapobj = + newsnapobj = dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj; } @@ -3203,9 +3042,9 @@ dmu_recv_end_sync(void *arg, dmu_tx_t *tx) * value. */ if (drc->drc_raw && drc->drc_ivset_guid != 0) { - dmu_object_zapify(dp->dp_meta_objset, drc->drc_newsnapobj, + dmu_object_zapify(dp->dp_meta_objset, newsnapobj, DMU_OT_DSL_DATASET, tx); - VERIFY0(zap_update(dp->dp_meta_objset, drc->drc_newsnapobj, + VERIFY0(zap_update(dp->dp_meta_objset, newsnapobj, DS_FIELD_IVSET_GUID, sizeof (uint64_t), 1, &drc->drc_ivset_guid, tx)); } @@ -3226,54 +3065,6 @@ dmu_recv_end_sync(void *arg, dmu_tx_t *tx) drc->drc_ds = NULL; } -static int -add_ds_to_guidmap(const char *name, avl_tree_t *guid_map, uint64_t snapobj, - boolean_t raw) -{ - dsl_pool_t *dp; - dsl_dataset_t *snapds; - guid_map_entry_t *gmep; - objset_t *os; - ds_hold_flags_t dsflags = (raw) ? 0 : DS_HOLD_FLAG_DECRYPT; - int err; - - ASSERT(guid_map != NULL); - - err = dsl_pool_hold(name, FTAG, &dp); - if (err != 0) - return (err); - gmep = kmem_alloc(sizeof (*gmep), KM_SLEEP); - err = dsl_dataset_own_obj(dp, snapobj, dsflags, gmep, &snapds); - - if (err == 0) { - err = dmu_objset_from_ds(snapds, &os); - if (err != 0) { - dsl_dataset_disown(snapds, dsflags, FTAG); - dsl_pool_rele(dp, FTAG); - kmem_free(gmep, sizeof (*gmep)); - return (err); - } - /* - * If this is a deduplicated raw send stream, we need - * to make sure that we can still read raw blocks from - * earlier datasets in the stream, so we set the - * os_raw_receive flag now. - */ - if (raw) - os->os_raw_receive = B_TRUE; - - gmep->raw = raw; - gmep->guid = dsl_dataset_phys(snapds)->ds_guid; - gmep->gme_os = os; - avl_add(guid_map, gmep); - } else { - kmem_free(gmep, sizeof (*gmep)); - } - - dsl_pool_rele(dp, FTAG); - return (err); -} - static int dmu_recv_end_modified_blocks = 3; static int @@ -3325,12 +3116,6 @@ dmu_recv_end(dmu_recv_cookie_t *drc, void *owner) drc->drc_tofs, drc->drc_tosnap); zvol_create_minor(snapname); kmem_strfree(snapname); - - if (drc->drc_guid_to_ds_map != NULL) { - (void) add_ds_to_guidmap(drc->drc_tofs, - drc->drc_guid_to_ds_map, - drc->drc_newsnapobj, drc->drc_raw); - } } return (error); } diff --git a/module/zfs/zfs_ioctl.c b/module/zfs/zfs_ioctl.c index 5c53eb637..2104bef71 100644 --- a/module/zfs/zfs_ioctl.c +++ b/module/zfs/zfs_ioctl.c @@ -27,7 +27,7 @@ * Copyright (c) 2014, 2016 Joyent, Inc. All rights reserved. * Copyright 2016 Nexenta Systems, Inc. All rights reserved. * Copyright (c) 2014, Joyent, Inc. All rights reserved. - * Copyright (c) 2011, 2018 by Delphix. All rights reserved. + * Copyright (c) 2011, 2020 by Delphix. All rights reserved. * Copyright (c) 2013 by Saso Kiselkov. All rights reserved. * Copyright (c) 2013 Steven Hartland. All rights reserved. * Copyright (c) 2014 Integros [integros.com] @@ -4765,9 +4765,9 @@ static boolean_t zfs_ioc_recv_inject_err; static int zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops, nvlist_t *localprops, nvlist_t *hidden_args, boolean_t force, - boolean_t resumable, int input_fd, dmu_replay_record_t *begin_record, - int cleanup_fd, uint64_t *read_bytes, uint64_t *errflags, - uint64_t *action_handle, nvlist_t **errors) + boolean_t resumable, int input_fd, + dmu_replay_record_t *begin_record, uint64_t *read_bytes, + uint64_t *errflags, nvlist_t **errors) { dmu_recv_cookie_t drc; int error = 0; @@ -4896,7 +4896,7 @@ zfs_ioc_recv_impl(char *tofs, char *tosnap, char *origin, nvlist_t *recvprops, nvlist_free(xprops); } - error = dmu_recv_stream(&drc, cleanup_fd, action_handle, &off); + error = dmu_recv_stream(&drc, &off); if (error == 0) { zfsvfs_t *zfsvfs = NULL; @@ -5088,13 +5088,10 @@ out: * zc_cookie file descriptor to recv from * zc_begin_record the BEGIN record of the stream (not byteswapped) * zc_guid force flag - * zc_cleanup_fd cleanup-on-exit file descriptor - * zc_action_handle handle for this guid/ds mapping (or zero on first call) * * outputs: * zc_cookie number of bytes read * zc_obj zprop_errflags_t - * zc_action_handle handle for this guid/ds mapping * zc_nvlist_dst{_size} error for each unapplied received property */ static int @@ -5137,8 +5134,7 @@ zfs_ioc_recv(zfs_cmd_t *zc) error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvdprops, localprops, NULL, zc->zc_guid, B_FALSE, zc->zc_cookie, &begin_record, - zc->zc_cleanup_fd, &zc->zc_cookie, &zc->zc_obj, - &zc->zc_action_handle, &errors); + &zc->zc_cookie, &zc->zc_obj, &errors); nvlist_free(recvdprops); nvlist_free(localprops); @@ -5171,15 +5167,14 @@ zfs_ioc_recv(zfs_cmd_t *zc) * "input_fd" -> file descriptor to read stream from (int32) * (optional) "force" -> force flag (value ignored) * (optional) "resumable" -> resumable flag (value ignored) - * (optional) "cleanup_fd" -> cleanup-on-exit file descriptor - * (optional) "action_handle" -> handle for this guid/ds mapping + * (optional) "cleanup_fd" -> unused + * (optional) "action_handle" -> unused * (optional) "hidden_args" -> { "wkeydata" -> value } * } * * outnvl: { * "read_bytes" -> number of bytes read * "error_flags" -> zprop_errflags_t - * "action_handle" -> handle for this guid/ds mapping * "errors" -> error for each unapplied received property (nvlist) * } */ @@ -5212,11 +5207,9 @@ zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) char tofs[ZFS_MAX_DATASET_NAME_LEN]; boolean_t force; boolean_t resumable; - uint64_t action_handle = 0; uint64_t read_bytes = 0; uint64_t errflags = 0; int input_fd = -1; - int cleanup_fd = -1; int error; snapname = fnvlist_lookup_string(innvl, "snapname"); @@ -5244,14 +5237,6 @@ zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) force = nvlist_exists(innvl, "force"); resumable = nvlist_exists(innvl, "resumable"); - error = nvlist_lookup_int32(innvl, "cleanup_fd", &cleanup_fd); - if (error && error != ENOENT) - return (error); - - error = nvlist_lookup_uint64(innvl, "action_handle", &action_handle); - if (error && error != ENOENT) - return (error); - /* we still use "props" here for backwards compatibility */ error = nvlist_lookup_nvlist(innvl, "props", &recvprops); if (error && error != ENOENT) @@ -5266,12 +5251,11 @@ zfs_ioc_recv_new(const char *fsname, nvlist_t *innvl, nvlist_t *outnvl) return (error); error = zfs_ioc_recv_impl(tofs, tosnap, origin, recvprops, localprops, - hidden_args, force, resumable, input_fd, begin_record, cleanup_fd, - &read_bytes, &errflags, &action_handle, &errors); + hidden_args, force, resumable, input_fd, begin_record, + &read_bytes, &errflags, &errors); fnvlist_add_uint64(outnvl, "read_bytes", read_bytes); fnvlist_add_uint64(outnvl, "error_flags", errflags); - fnvlist_add_uint64(outnvl, "action_handle", action_handle); fnvlist_add_nvlist(outnvl, "errors", errors); nvlist_free(errors); diff --git a/module/zfs/zfs_onexit.c b/module/zfs/zfs_onexit.c index bf86446d4..2a1332e71 100644 --- a/module/zfs/zfs_onexit.c +++ b/module/zfs/zfs_onexit.c @@ -20,7 +20,7 @@ */ /* * Copyright (c) 2010, Oracle and/or its affiliates. All rights reserved. - * Copyright (c) 2013 by Delphix. All rights reserved. + * Copyright (c) 2013, 2020 by Delphix. All rights reserved. */ #include <sys/types.h> @@ -171,80 +171,3 @@ zfs_onexit_add_cb(minor_t minor, void (*func)(void *), void *data, return (0); } - -static zfs_onexit_action_node_t * -zfs_onexit_find_cb(zfs_onexit_t *zo, uint64_t action_handle) -{ - zfs_onexit_action_node_t *match; - zfs_onexit_action_node_t *ap; - list_t *l; - - ASSERT(MUTEX_HELD(&zo->zo_lock)); - - match = (zfs_onexit_action_node_t *)(uintptr_t)action_handle; - l = &zo->zo_actions; - for (ap = list_head(l); ap != NULL; ap = list_next(l, ap)) { - if (match == ap) - break; - } - return (ap); -} - -/* - * Delete the callback, triggering it first if 'fire' is set. - */ -int -zfs_onexit_del_cb(minor_t minor, uint64_t action_handle, boolean_t fire) -{ - zfs_onexit_t *zo; - zfs_onexit_action_node_t *ap; - int error; - - error = zfs_onexit_minor_to_state(minor, &zo); - if (error) - return (error); - - mutex_enter(&zo->zo_lock); - ap = zfs_onexit_find_cb(zo, action_handle); - if (ap != NULL) { - list_remove(&zo->zo_actions, ap); - mutex_exit(&zo->zo_lock); - if (fire) - ap->za_func(ap->za_data); - kmem_free(ap, sizeof (zfs_onexit_action_node_t)); - } else { - mutex_exit(&zo->zo_lock); - error = SET_ERROR(ENOENT); - } - - return (error); -} - -/* - * Return the data associated with this callback. This allows consumers - * of the cleanup-on-exit interfaces to stash kernel data across system - * calls, knowing that it will be cleaned up if the calling process exits. - */ -int -zfs_onexit_cb_data(minor_t minor, uint64_t action_handle, void **data) -{ - zfs_onexit_t *zo; - zfs_onexit_action_node_t *ap; - int error; - - *data = NULL; - - error = zfs_onexit_minor_to_state(minor, &zo); - if (error) - return (error); - - mutex_enter(&zo->zo_lock); - ap = zfs_onexit_find_cb(zo, action_handle); - if (ap != NULL) - *data = ap->za_data; - else - error = SET_ERROR(ENOENT); - mutex_exit(&zo->zo_lock); - - return (error); -} |