* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2011 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
- * Copyright (c) 2012, Joyent, Inc. All rights reserved.
- * Copyright (c) 2013 by Delphix. All rights reserved.
+ * Copyright (c) 2014, Joyent, Inc. All rights reserved.
+ * Copyright (c) 2011, 2014 by Delphix. All rights reserved.
*/
#include <sys/dmu.h>
#include <sys/zfs_onexit.h>
#include <sys/dmu_send.h>
#include <sys/dsl_destroy.h>
+#include <sys/blkptr.h>
+#include <sys/dsl_bookmark.h>
+#include <sys/zfeature.h>
/* Set this tunable to TRUE to replace corrupt data with 0x2f5baddb10c */
int zfs_send_corrupt_data = B_FALSE;
{
struct drr_free *drrf = &(dsp->dsa_drr->drr_u.drr_free);
+ /*
+ * When we receive a free record, dbuf_free_range() assumes
+ * that the receiving system doesn't have any dbufs in the range
+ * being freed. This is always true because there is a one-record
+ * constraint: we only send one WRITE record for any given
+ * object+offset. We know that the one-record constraint is
+ * true because we always send data in increasing order by
+ * object,offset.
+ *
+ * If the increasing-order constraint ever changes, we should find
+ * another way to assert that the one-record constraint is still
+ * satisfied.
+ */
+ ASSERT(object > dsp->dsa_last_data_object ||
+ (object == dsp->dsa_last_data_object &&
+ offset > dsp->dsa_last_data_offset));
+
+ /*
+ * If we are doing a non-incremental send, then there can't
+ * be any data in the dataset we're receiving into. Therefore
+ * a free record would simply be a no-op. Save space by not
+ * sending it to begin with.
+ */
+ if (!dsp->dsa_incremental)
+ return (0);
+
if (length != -1ULL && offset + length < offset)
length = -1ULL;
}
static int
-dump_data(dmu_sendarg_t *dsp, dmu_object_type_t type,
+dump_write(dmu_sendarg_t *dsp, dmu_object_type_t type,
uint64_t object, uint64_t offset, int blksz, const blkptr_t *bp, void *data)
{
struct drr_write *drrw = &(dsp->dsa_drr->drr_u.drr_write);
+ /*
+ * We send data in increasing object, offset order.
+ * See comment in dump_free() for details.
+ */
+ ASSERT(object > dsp->dsa_last_data_object ||
+ (object == dsp->dsa_last_data_object &&
+ offset > dsp->dsa_last_data_offset));
+ dsp->dsa_last_data_object = object;
+ dsp->dsa_last_data_offset = offset + blksz - 1;
/*
* If there is any kind of pending aggregation (currently either
drrw->drr_offset = offset;
drrw->drr_length = blksz;
drrw->drr_toguid = dsp->dsa_toguid;
- drrw->drr_checksumtype = BP_GET_CHECKSUM(bp);
- if (zio_checksum_table[drrw->drr_checksumtype].ci_dedup)
- drrw->drr_checksumflags |= DRR_CHECKSUM_DEDUP;
- DDK_SET_LSIZE(&drrw->drr_key, BP_GET_LSIZE(bp));
- DDK_SET_PSIZE(&drrw->drr_key, BP_GET_PSIZE(bp));
- DDK_SET_COMPRESS(&drrw->drr_key, BP_GET_COMPRESS(bp));
- drrw->drr_key.ddk_cksum = bp->blk_cksum;
+ if (BP_IS_EMBEDDED(bp)) {
+ /*
+ * There's no pre-computed checksum of embedded BP's, so
+ * (like fletcher4-checkummed blocks) userland will have
+ * to compute a dedup-capable checksum itself.
+ */
+ drrw->drr_checksumtype = ZIO_CHECKSUM_OFF;
+ } else {
+ drrw->drr_checksumtype = BP_GET_CHECKSUM(bp);
+ if (zio_checksum_table[drrw->drr_checksumtype].ci_dedup)
+ drrw->drr_checksumflags |= DRR_CHECKSUM_DEDUP;
+ DDK_SET_LSIZE(&drrw->drr_key, BP_GET_LSIZE(bp));
+ DDK_SET_PSIZE(&drrw->drr_key, BP_GET_PSIZE(bp));
+ DDK_SET_COMPRESS(&drrw->drr_key, BP_GET_COMPRESS(bp));
+ drrw->drr_key.ddk_cksum = bp->blk_cksum;
+ }
if (dump_bytes(dsp, dsp->dsa_drr, sizeof (dmu_replay_record_t)) != 0)
return (SET_ERROR(EINTR));
return (0);
}
+static int
+dump_write_embedded(dmu_sendarg_t *dsp, uint64_t object, uint64_t offset,
+ int blksz, const blkptr_t *bp)
+{
+ char buf[BPE_PAYLOAD_SIZE];
+ struct drr_write_embedded *drrw =
+ &(dsp->dsa_drr->drr_u.drr_write_embedded);
+
+ if (dsp->dsa_pending_op != PENDING_NONE) {
+ if (dump_bytes(dsp, dsp->dsa_drr,
+ sizeof (dmu_replay_record_t)) != 0)
+ return (EINTR);
+ dsp->dsa_pending_op = PENDING_NONE;
+ }
+
+ ASSERT(BP_IS_EMBEDDED(bp));
+
+ bzero(dsp->dsa_drr, sizeof (dmu_replay_record_t));
+ dsp->dsa_drr->drr_type = DRR_WRITE_EMBEDDED;
+ drrw->drr_object = object;
+ drrw->drr_offset = offset;
+ drrw->drr_length = blksz;
+ drrw->drr_toguid = dsp->dsa_toguid;
+ drrw->drr_compression = BP_GET_COMPRESS(bp);
+ drrw->drr_etype = BPE_GET_ETYPE(bp);
+ drrw->drr_lsize = BPE_GET_LSIZE(bp);
+ drrw->drr_psize = BPE_GET_PSIZE(bp);
+
+ decode_embedded_bp_compressed(bp, buf);
+
+ if (dump_bytes(dsp, dsp->dsa_drr, sizeof (dmu_replay_record_t)) != 0)
+ return (EINTR);
+ if (dump_bytes(dsp, buf, P2ROUNDUP(drrw->drr_psize, 8)) != 0)
+ return (EINTR);
+ return (0);
+}
+
static int
dump_spill(dmu_sendarg_t *dsp, uint64_t object, int blksz, void *data)
{
{
struct drr_freeobjects *drrfo = &(dsp->dsa_drr->drr_u.drr_freeobjects);
+ /* See comment in dump_free(). */
+ if (!dsp->dsa_incremental)
+ return (0);
+
/*
* If there is a pending op, but it's not PENDING_FREEOBJECTS,
* push it out, since free block aggregation can only be done for
if (dump_bytes(dsp, DN_BONUS(dnp), P2ROUNDUP(dnp->dn_bonuslen, 8)) != 0)
return (SET_ERROR(EINTR));
- /* free anything past the end of the file */
+ /* Free anything past the end of the file. */
if (dump_free(dsp, object, (dnp->dn_maxblkid + 1) *
- (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), -1ULL))
+ (dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT), -1ULL) != 0)
return (SET_ERROR(EINTR));
if (dsp->dsa_err != 0)
return (SET_ERROR(EINTR));
return (0);
}
+static boolean_t
+backup_do_embed(dmu_sendarg_t *dsp, const blkptr_t *bp)
+{
+ if (!BP_IS_EMBEDDED(bp))
+ return (B_FALSE);
+
+ /*
+ * Compression function must be legacy, or explicitly enabled.
+ */
+ if ((BP_GET_COMPRESS(bp) >= ZIO_COMPRESS_LEGACY_FUNCTIONS &&
+ !(dsp->dsa_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA_LZ4)))
+ return (B_FALSE);
+
+ /*
+ * Embed type must be explicitly enabled.
+ */
+ switch (BPE_GET_ETYPE(bp)) {
+ case BP_EMBEDDED_TYPE_DATA:
+ if (dsp->dsa_featureflags & DMU_BACKUP_FEATURE_EMBED_DATA)
+ return (B_TRUE);
+ break;
+ default:
+ return (B_FALSE);
+ }
+ return (B_FALSE);
+}
+
#define BP_SPAN(dnp, level) \
(((uint64_t)dnp->dn_datablkszsec) << (SPA_MINBLOCKSHIFT + \
(level) * (dnp->dn_indblkshift - SPA_BLKPTRSHIFT)))
/* ARGSUSED */
static int
backup_cb(spa_t *spa, zilog_t *zilog, const blkptr_t *bp,
- const zbookmark_t *zb, const dnode_phys_t *dnp, void *arg)
+ const zbookmark_phys_t *zb, const dnode_phys_t *dnp, void *arg)
{
dmu_sendarg_t *dsp = arg;
dmu_object_type_t type = bp ? BP_GET_TYPE(bp) : DMU_OT_NONE;
if (zb->zb_object != DMU_META_DNODE_OBJECT &&
DMU_OBJECT_IS_SPECIAL(zb->zb_object)) {
return (0);
- } else if (bp == NULL && zb->zb_object == DMU_META_DNODE_OBJECT) {
+ } else if (zb->zb_level == ZB_ZIL_LEVEL) {
+ /*
+ * If we are sending a non-snapshot (which is allowed on
+ * read-only pools), it may have a ZIL, which must be ignored.
+ */
+ return (0);
+ } else if (BP_IS_HOLE(bp) &&
+ zb->zb_object == DMU_META_DNODE_OBJECT) {
uint64_t span = BP_SPAN(dnp, zb->zb_level);
uint64_t dnobj = (zb->zb_blkid * span) >> DNODE_SHIFT;
err = dump_freeobjects(dsp, dnobj, span >> DNODE_SHIFT);
- } else if (bp == NULL) {
+ } else if (BP_IS_HOLE(bp)) {
uint64_t span = BP_SPAN(dnp, zb->zb_level);
err = dump_free(dsp, zb->zb_object, zb->zb_blkid * span, span);
} else if (zb->zb_level > 0 || type == DMU_OT_OBJSET) {
err = dump_spill(dsp, zb->zb_object, blksz, abuf->b_data);
(void) arc_buf_remove_ref(abuf, &abuf);
+ } else if (backup_do_embed(dsp, bp)) {
+ /* it's an embedded level-0 block of a regular object */
+ int blksz = dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT;
+ err = dump_write_embedded(dsp, zb->zb_object,
+ zb->zb_blkid * blksz, blksz, bp);
} else { /* it's a level-0 block of a regular object */
uint32_t aflags = ARC_WAIT;
arc_buf_t *abuf;
int blksz = BP_GET_LSIZE(bp);
+ ASSERT3U(blksz, ==, dnp->dn_datablkszsec << SPA_MINBLOCKSHIFT);
+ ASSERT0(zb->zb_level);
if (arc_read(NULL, spa, bp, arc_getbuf_func, &abuf,
ZIO_PRIORITY_ASYNC_READ, ZIO_FLAG_CANFAIL,
&aflags, zb) != 0) {
}
}
- err = dump_data(dsp, type, zb->zb_object, zb->zb_blkid * blksz,
+ err = dump_write(dsp, type, zb->zb_object, zb->zb_blkid * blksz,
blksz, bp, abuf->b_data);
(void) arc_buf_remove_ref(abuf, &abuf);
}
}
/*
- * Releases dp, ds, and fromds, using the specified tag.
+ * Releases dp using the specified tag.
*/
static int
dmu_send_impl(void *tag, dsl_pool_t *dp, dsl_dataset_t *ds,
- dsl_dataset_t *fromds, int outfd, vnode_t *vp, offset_t *off)
+ zfs_bookmark_phys_t *fromzb, boolean_t is_clone, boolean_t embedok,
+ int outfd, vnode_t *vp, offset_t *off)
{
objset_t *os;
dmu_replay_record_t *drr;
dmu_sendarg_t *dsp;
int err;
uint64_t fromtxg = 0;
-
- if (fromds != NULL && !dsl_dataset_is_before(ds, fromds)) {
- dsl_dataset_rele(fromds, tag);
- dsl_dataset_rele(ds, tag);
- dsl_pool_rele(dp, tag);
- return (SET_ERROR(EXDEV));
- }
+ uint64_t featureflags = 0;
err = dmu_objset_from_ds(ds, &os);
if (err != 0) {
- if (fromds != NULL)
- dsl_dataset_rele(fromds, tag);
- dsl_dataset_rele(ds, tag);
dsl_pool_rele(dp, tag);
return (err);
}
uint64_t version;
if (zfs_get_zplprop(os, ZFS_PROP_VERSION, &version) != 0) {
kmem_free(drr, sizeof (dmu_replay_record_t));
- if (fromds != NULL)
- dsl_dataset_rele(fromds, tag);
- dsl_dataset_rele(ds, tag);
dsl_pool_rele(dp, tag);
return (SET_ERROR(EINVAL));
}
if (version >= ZPL_VERSION_SA) {
- DMU_SET_FEATUREFLAGS(
- drr->drr_u.drr_begin.drr_versioninfo,
- DMU_BACKUP_FEATURE_SA_SPILL);
+ featureflags |= DMU_BACKUP_FEATURE_SA_SPILL;
}
}
#endif
+ if (embedok &&
+ spa_feature_is_active(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA)) {
+ featureflags |= DMU_BACKUP_FEATURE_EMBED_DATA;
+ if (spa_feature_is_active(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS))
+ featureflags |= DMU_BACKUP_FEATURE_EMBED_DATA_LZ4;
+ } else {
+ embedok = B_FALSE;
+ }
+
+ DMU_SET_FEATUREFLAGS(drr->drr_u.drr_begin.drr_versioninfo,
+ featureflags);
+
drr->drr_u.drr_begin.drr_creation_time =
- ds->ds_phys->ds_creation_time;
+ dsl_dataset_phys(ds)->ds_creation_time;
drr->drr_u.drr_begin.drr_type = dmu_objset_type(os);
- if (fromds != NULL && ds->ds_dir != fromds->ds_dir)
+ if (is_clone)
drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_CLONE;
- drr->drr_u.drr_begin.drr_toguid = ds->ds_phys->ds_guid;
- if (ds->ds_phys->ds_flags & DS_FLAG_CI_DATASET)
+ drr->drr_u.drr_begin.drr_toguid = dsl_dataset_phys(ds)->ds_guid;
+ if (dsl_dataset_phys(ds)->ds_flags & DS_FLAG_CI_DATASET)
drr->drr_u.drr_begin.drr_flags |= DRR_FLAG_CI_DATA;
- if (fromds != NULL)
- drr->drr_u.drr_begin.drr_fromguid = fromds->ds_phys->ds_guid;
+ if (fromzb != NULL) {
+ drr->drr_u.drr_begin.drr_fromguid = fromzb->zbm_guid;
+ fromtxg = fromzb->zbm_creation_txg;
+ }
dsl_dataset_name(ds, drr->drr_u.drr_begin.drr_toname);
-
- if (fromds != NULL) {
- fromtxg = fromds->ds_phys->ds_creation_txg;
- dsl_dataset_rele(fromds, tag);
- fromds = NULL;
+ if (!dsl_dataset_is_snapshot(ds)) {
+ (void) strlcat(drr->drr_u.drr_begin.drr_toname, "@--head--",
+ sizeof (drr->drr_u.drr_begin.drr_toname));
}
dsp = kmem_zalloc(sizeof (dmu_sendarg_t), KM_SLEEP);
dsp->dsa_proc = curproc;
dsp->dsa_os = os;
dsp->dsa_off = off;
- dsp->dsa_toguid = ds->ds_phys->ds_guid;
+ dsp->dsa_toguid = dsl_dataset_phys(ds)->ds_guid;
ZIO_SET_CHECKSUM(&dsp->dsa_zc, 0, 0, 0, 0);
dsp->dsa_pending_op = PENDING_NONE;
+ dsp->dsa_incremental = (fromzb != NULL);
+ dsp->dsa_featureflags = featureflags;
mutex_enter(&ds->ds_sendstream_lock);
list_insert_head(&ds->ds_sendstreams, dsp);
kmem_free(dsp, sizeof (dmu_sendarg_t));
dsl_dataset_long_rele(ds, FTAG);
- dsl_dataset_rele(ds, tag);
return (err);
}
int
dmu_send_obj(const char *pool, uint64_t tosnap, uint64_t fromsnap,
- int outfd, vnode_t *vp, offset_t *off)
+ boolean_t embedok, int outfd, vnode_t *vp, offset_t *off)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
}
if (fromsnap != 0) {
+ zfs_bookmark_phys_t zb;
+ boolean_t is_clone;
+
err = dsl_dataset_hold_obj(dp, fromsnap, FTAG, &fromds);
if (err != 0) {
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (err);
}
+ if (!dsl_dataset_is_before(ds, fromds, 0))
+ err = SET_ERROR(EXDEV);
+ zb.zbm_creation_time =
+ dsl_dataset_phys(fromds)->ds_creation_time;
+ zb.zbm_creation_txg = dsl_dataset_phys(fromds)->ds_creation_txg;
+ zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid;
+ is_clone = (fromds->ds_dir != ds->ds_dir);
+ dsl_dataset_rele(fromds, FTAG);
+ err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone, embedok,
+ outfd, vp, off);
+ } else {
+ err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE, embedok,
+ outfd, vp, off);
}
-
- return (dmu_send_impl(FTAG, dp, ds, fromds, outfd, vp, off));
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
}
int
-dmu_send(const char *tosnap, const char *fromsnap,
+dmu_send(const char *tosnap, const char *fromsnap, boolean_t embedok,
int outfd, vnode_t *vp, offset_t *off)
{
dsl_pool_t *dp;
dsl_dataset_t *ds;
- dsl_dataset_t *fromds = NULL;
int err;
+ boolean_t owned = B_FALSE;
- if (strchr(tosnap, '@') == NULL)
- return (SET_ERROR(EINVAL));
- if (fromsnap != NULL && strchr(fromsnap, '@') == NULL)
+ if (fromsnap != NULL && strpbrk(fromsnap, "@#") == NULL)
return (SET_ERROR(EINVAL));
err = dsl_pool_hold(tosnap, FTAG, &dp);
if (err != 0)
return (err);
- err = dsl_dataset_hold(dp, tosnap, FTAG, &ds);
+ if (strchr(tosnap, '@') == NULL && spa_writeable(dp->dp_spa)) {
+ /*
+ * We are sending a filesystem or volume. Ensure
+ * that it doesn't change by owning the dataset.
+ */
+ err = dsl_dataset_own(dp, tosnap, FTAG, &ds);
+ owned = B_TRUE;
+ } else {
+ err = dsl_dataset_hold(dp, tosnap, FTAG, &ds);
+ }
if (err != 0) {
dsl_pool_rele(dp, FTAG);
return (err);
}
if (fromsnap != NULL) {
- err = dsl_dataset_hold(dp, fromsnap, FTAG, &fromds);
+ zfs_bookmark_phys_t zb;
+ boolean_t is_clone = B_FALSE;
+ int fsnamelen = strchr(tosnap, '@') - tosnap;
+
+ /*
+ * If the fromsnap is in a different filesystem, then
+ * mark the send stream as a clone.
+ */
+ if (strncmp(tosnap, fromsnap, fsnamelen) != 0 ||
+ (fromsnap[fsnamelen] != '@' &&
+ fromsnap[fsnamelen] != '#')) {
+ is_clone = B_TRUE;
+ }
+
+ if (strchr(fromsnap, '@')) {
+ dsl_dataset_t *fromds;
+ err = dsl_dataset_hold(dp, fromsnap, FTAG, &fromds);
+ if (err == 0) {
+ if (!dsl_dataset_is_before(ds, fromds, 0))
+ err = SET_ERROR(EXDEV);
+ zb.zbm_creation_time =
+ dsl_dataset_phys(fromds)->ds_creation_time;
+ zb.zbm_creation_txg =
+ dsl_dataset_phys(fromds)->ds_creation_txg;
+ zb.zbm_guid = dsl_dataset_phys(fromds)->ds_guid;
+ is_clone = (ds->ds_dir != fromds->ds_dir);
+ dsl_dataset_rele(fromds, FTAG);
+ }
+ } else {
+ err = dsl_bookmark_lookup(dp, fromsnap, ds, &zb);
+ }
if (err != 0) {
dsl_dataset_rele(ds, FTAG);
dsl_pool_rele(dp, FTAG);
return (err);
}
+ err = dmu_send_impl(FTAG, dp, ds, &zb, is_clone, embedok,
+ outfd, vp, off);
+ } else {
+ err = dmu_send_impl(FTAG, dp, ds, NULL, B_FALSE, embedok,
+ outfd, vp, off);
}
- return (dmu_send_impl(FTAG, dp, ds, fromds, outfd, vp, off));
+ if (owned)
+ dsl_dataset_disown(ds, FTAG);
+ else
+ dsl_dataset_rele(ds, FTAG);
+ return (err);
}
int
* fromsnap must be an earlier snapshot from the same fs as tosnap,
* or the origin's fs.
*/
- if (fromds != NULL && !dsl_dataset_is_before(ds, fromds))
+ if (fromds != NULL && !dsl_dataset_is_before(ds, fromds, 0))
return (SET_ERROR(EXDEV));
/* Get uncompressed size estimate of changed data. */
if (fromds == NULL) {
- size = ds->ds_phys->ds_uncompressed_bytes;
+ size = dsl_dataset_phys(ds)->ds_uncompressed_bytes;
} else {
uint64_t used, comp;
err = dsl_dataset_space_written(fromds, ds,
/* temporary clone name must not exist */
error = zap_lookup(dp->dp_meta_objset,
- ds->ds_dir->dd_phys->dd_child_dir_zapobj, recv_clone_name,
+ dsl_dir_phys(ds->ds_dir)->dd_child_dir_zapobj, recv_clone_name,
8, 1, &val);
if (error != ENOENT)
return (error == 0 ? EBUSY : error);
/* new snapshot name must not exist */
error = zap_lookup(dp->dp_meta_objset,
- ds->ds_phys->ds_snapnames_zapobj, drba->drba_cookie->drc_tosnap,
- 8, 1, &val);
+ dsl_dataset_phys(ds)->ds_snapnames_zapobj,
+ drba->drba_cookie->drc_tosnap, 8, 1, &val);
if (error != ENOENT)
return (error == 0 ? EEXIST : error);
+ /*
+ * Check snapshot limit before receiving. We'll recheck again at the
+ * end, but might as well abort before receiving if we're already over
+ * the limit.
+ *
+ * Note that we do not check the file system limit with
+ * dsl_dir_fscount_check because the temporary %clones don't count
+ * against that limit.
+ */
+ error = dsl_fs_ss_limit_check(ds->ds_dir, 1, ZFS_PROP_SNAPSHOT_LIMIT,
+ NULL, drba->drba_cred);
+ if (error != 0)
+ return (error);
+
if (fromguid != 0) {
dsl_dataset_t *snap;
- uint64_t obj = ds->ds_phys->ds_prev_snap_obj;
+ uint64_t obj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
/* Find snapshot in this dir that matches fromguid. */
while (obj != 0) {
dsl_dataset_rele(snap, FTAG);
return (SET_ERROR(ENODEV));
}
- if (snap->ds_phys->ds_guid == fromguid)
+ if (dsl_dataset_phys(snap)->ds_guid == fromguid)
break;
- obj = snap->ds_phys->ds_prev_snap_obj;
+ obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_dataset_rele(snap, FTAG);
}
if (obj == 0)
dsl_dataset_rele(snap, FTAG);
} else {
/* if full, most recent snapshot must be $ORIGIN */
- if (ds->ds_phys->ds_prev_snap_txg >= TXG_INITIAL)
+ if (dsl_dataset_phys(ds)->ds_prev_snap_txg >= TXG_INITIAL)
return (SET_ERROR(ENODEV));
- drba->drba_snapobj = ds->ds_phys->ds_prev_snap_obj;
+ drba->drba_snapobj = dsl_dataset_phys(ds)->ds_prev_snap_obj;
}
return (0);
uint64_t fromguid = drrb->drr_fromguid;
int flags = drrb->drr_flags;
int error;
+ uint64_t featureflags = DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo);
dsl_dataset_t *ds;
const char *tofs = drba->drba_cookie->drc_tofs;
return (SET_ERROR(EINVAL));
/* Verify pool version supports SA if SA_SPILL feature set */
- if ((DMU_GET_FEATUREFLAGS(drrb->drr_versioninfo) &
- DMU_BACKUP_FEATURE_SA_SPILL) &&
- spa_version(dp->dp_spa) < SPA_VERSION_SA) {
+ if ((featureflags & DMU_BACKUP_FEATURE_SA_SPILL) &&
+ spa_version(dp->dp_spa) < SPA_VERSION_SA)
+ return (SET_ERROR(ENOTSUP));
+
+ /*
+ * The receiving code doesn't know how to translate a WRITE_EMBEDDED
+ * record to a plan WRITE record, so the pool must have the
+ * EMBEDDED_DATA feature enabled if the stream has WRITE_EMBEDDED
+ * records. Same with WRITE_EMBEDDED records that use LZ4 compression.
+ */
+ if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA) &&
+ !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_EMBEDDED_DATA))
+ return (SET_ERROR(ENOTSUP));
+ if ((featureflags & DMU_BACKUP_FEATURE_EMBED_DATA_LZ4) &&
+ !spa_feature_is_enabled(dp->dp_spa, SPA_FEATURE_LZ4_COMPRESS))
return (SET_ERROR(ENOTSUP));
- }
error = dsl_dataset_hold(dp, tofs, FTAG, &ds);
if (error == 0) {
if (error != 0)
return (error);
+ /*
+ * Check filesystem and snapshot limits before receiving. We'll
+ * recheck snapshot limits again at the end (we create the
+ * filesystems and increment those counts during begin_sync).
+ */
+ error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
+ ZFS_PROP_FILESYSTEM_LIMIT, NULL, drba->drba_cred);
+ if (error != 0) {
+ dsl_dataset_rele(ds, FTAG);
+ return (error);
+ }
+
+ error = dsl_fs_ss_limit_check(ds->ds_dir, 1,
+ ZFS_PROP_SNAPSHOT_LIMIT, NULL, drba->drba_cred);
+ if (error != 0) {
+ dsl_dataset_rele(ds, FTAG);
+ return (error);
+ }
+
if (drba->drba_origin != NULL) {
dsl_dataset_t *origin;
error = dsl_dataset_hold(dp, drba->drba_origin,
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(EINVAL));
}
- if (origin->ds_phys->ds_guid != fromguid) {
+ if (dsl_dataset_phys(origin)->ds_guid != fromguid) {
dsl_dataset_rele(origin, FTAG);
dsl_dataset_rele(ds, FTAG);
return (SET_ERROR(ENODEV));
VERIFY0(dsl_dataset_own_obj(dp, dsobj, dmu_recv_tag, &newds));
dmu_buf_will_dirty(newds->ds_dbuf, tx);
- newds->ds_phys->ds_flags |= DS_FLAG_INCONSISTENT;
+ dsl_dataset_phys(newds)->ds_flags |= DS_FLAG_INCONSISTENT;
/*
* If we actually created a non-clone, we need to create the
drc->drc_tosnap = tosnap;
drc->drc_tofs = tofs;
drc->drc_force = force;
+ drc->drc_cred = CRED();
if (drrb->drr_magic == BSWAP_64(DMU_BACKUP_MAGIC))
drc->drc_byteswap = B_TRUE;
}
static void *
-restore_read(struct restorearg *ra, int len)
+restore_read(struct restorearg *ra, int len, char *buf)
{
- void *rv;
int done = 0;
+ if (buf == NULL)
+ buf = ra->buf;
+
/* some things will require 8-byte alignment, so everything must */
ASSERT0(len % 8);
ssize_t resid;
ra->err = vn_rdwr(UIO_READ, ra->vp,
- (caddr_t)ra->buf + done, len - done,
+ buf + done, len - done,
ra->voff, UIO_SYSSPACE, FAPPEND,
RLIM64_INFINITY, CRED(), &resid);
}
ASSERT3U(done, ==, len);
- rv = ra->buf;
if (ra->byteswap)
- fletcher_4_incremental_byteswap(rv, len, &ra->cksum);
+ fletcher_4_incremental_byteswap(buf, len, &ra->cksum);
else
- fletcher_4_incremental_native(rv, len, &ra->cksum);
- return (rv);
+ fletcher_4_incremental_native(buf, len, &ra->cksum);
+ return (buf);
}
noinline static void
break;
case DRR_OBJECT:
DO64(drr_object.drr_object);
- /* DO64(drr_object.drr_allocation_txg); */
DO32(drr_object.drr_type);
DO32(drr_object.drr_bonustype);
DO32(drr_object.drr_blksz);
DO64(drr_write_byref.drr_key.ddk_cksum.zc_word[3]);
DO64(drr_write_byref.drr_key.ddk_prop);
break;
+ case DRR_WRITE_EMBEDDED:
+ DO64(drr_write_embedded.drr_object);
+ DO64(drr_write_embedded.drr_offset);
+ DO64(drr_write_embedded.drr_length);
+ DO64(drr_write_embedded.drr_toguid);
+ DO32(drr_write_embedded.drr_lsize);
+ DO32(drr_write_embedded.drr_psize);
+ break;
case DRR_FREE:
DO64(drr_free.drr_object);
DO64(drr_free.drr_offset);
#undef DO32
}
+static inline uint8_t
+deduce_nblkptr(dmu_object_type_t bonus_type, uint64_t bonus_size)
+{
+ if (bonus_type == DMU_OT_SA) {
+ return (1);
+ } else {
+ return (1 +
+ ((DN_MAX_BONUSLEN - bonus_size) >> SPA_BLKPTRSHIFT));
+ }
+}
+
noinline static int
restore_object(struct restorearg *ra, objset_t *os, struct drr_object *drro)
{
- int err;
+ dmu_object_info_t doi;
dmu_tx_t *tx;
void *data = NULL;
+ uint64_t object;
+ int err;
if (drro->drr_type == DMU_OT_NONE ||
!DMU_OT_IS_VALID(drro->drr_type) ||
return (SET_ERROR(EINVAL));
}
- err = dmu_object_info(os, drro->drr_object, NULL);
+ err = dmu_object_info(os, drro->drr_object, &doi);
if (err != 0 && err != ENOENT)
return (SET_ERROR(EINVAL));
+ object = err == 0 ? drro->drr_object : DMU_NEW_OBJECT;
if (drro->drr_bonuslen) {
- data = restore_read(ra, P2ROUNDUP(drro->drr_bonuslen, 8));
+ data = restore_read(ra, P2ROUNDUP(drro->drr_bonuslen, 8), NULL);
if (ra->err != 0)
return (ra->err);
}
- if (err == ENOENT) {
- /* currently free, want to be allocated */
- tx = dmu_tx_create(os);
- dmu_tx_hold_bonus(tx, DMU_NEW_OBJECT);
- err = dmu_tx_assign(tx, TXG_WAIT);
- if (err != 0) {
- dmu_tx_abort(tx);
- return (err);
+ /*
+ * If we are losing blkptrs or changing the block size this must
+ * be a new file instance. We must clear out the previous file
+ * contents before we can change this type of metadata in the dnode.
+ */
+ if (err == 0) {
+ int nblkptr;
+
+ nblkptr = deduce_nblkptr(drro->drr_bonustype,
+ drro->drr_bonuslen);
+
+ if (drro->drr_blksz != doi.doi_data_block_size ||
+ nblkptr < doi.doi_nblkptr) {
+ err = dmu_free_long_range(os, drro->drr_object,
+ 0, DMU_OBJECT_END);
+ if (err != 0)
+ return (SET_ERROR(EINVAL));
}
+ }
+
+ tx = dmu_tx_create(os);
+ dmu_tx_hold_bonus(tx, object);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err != 0) {
+ dmu_tx_abort(tx);
+ return (err);
+ }
+
+ if (object == DMU_NEW_OBJECT) {
+ /* currently free, want to be allocated */
err = dmu_object_claim(os, drro->drr_object,
drro->drr_type, drro->drr_blksz,
drro->drr_bonustype, drro->drr_bonuslen, tx);
- dmu_tx_commit(tx);
- } else {
- /* currently allocated, want to be allocated */
+ } else if (drro->drr_type != doi.doi_type ||
+ drro->drr_blksz != doi.doi_data_block_size ||
+ drro->drr_bonustype != doi.doi_bonus_type ||
+ drro->drr_bonuslen != doi.doi_bonus_size) {
+ /* currently allocated, but with different properties */
err = dmu_object_reclaim(os, drro->drr_object,
drro->drr_type, drro->drr_blksz,
- drro->drr_bonustype, drro->drr_bonuslen);
+ drro->drr_bonustype, drro->drr_bonuslen, tx);
}
if (err != 0) {
+ dmu_tx_commit(tx);
return (SET_ERROR(EINVAL));
}
- tx = dmu_tx_create(os);
- dmu_tx_hold_bonus(tx, drro->drr_object);
- err = dmu_tx_assign(tx, TXG_WAIT);
- if (err != 0) {
- dmu_tx_abort(tx);
- return (err);
- }
-
dmu_object_set_checksum(os, drro->drr_object, drro->drr_checksumtype,
tx);
dmu_object_set_compress(os, drro->drr_object, drro->drr_compress, tx);
if (dmu_object_info(os, obj, NULL) != 0)
continue;
- err = dmu_free_object(os, obj);
+ err = dmu_free_long_object(os, obj);
if (err != 0)
return (err);
}
struct drr_write *drrw)
{
dmu_tx_t *tx;
+ dmu_buf_t *bonus;
+ arc_buf_t *abuf;
void *data;
int err;
!DMU_OT_IS_VALID(drrw->drr_type))
return (SET_ERROR(EINVAL));
- data = restore_read(ra, drrw->drr_length);
- if (data == NULL)
- return (ra->err);
-
if (dmu_object_info(os, drrw->drr_object, NULL) != 0)
return (SET_ERROR(EINVAL));
+ if (dmu_bonus_hold(os, drrw->drr_object, FTAG, &bonus) != 0)
+ return (SET_ERROR(EINVAL));
+
+ abuf = dmu_request_arcbuf(bonus, drrw->drr_length);
+
+ data = restore_read(ra, drrw->drr_length, abuf->b_data);
+ if (data == NULL) {
+ dmu_return_arcbuf(abuf);
+ dmu_buf_rele(bonus, FTAG);
+ return (ra->err);
+ }
+
tx = dmu_tx_create(os);
dmu_tx_hold_write(tx, drrw->drr_object,
drrw->drr_offset, drrw->drr_length);
err = dmu_tx_assign(tx, TXG_WAIT);
if (err != 0) {
+ dmu_return_arcbuf(abuf);
+ dmu_buf_rele(bonus, FTAG);
dmu_tx_abort(tx);
return (err);
}
DMU_OT_BYTESWAP(drrw->drr_type);
dmu_ot_byteswap[byteswap].ob_func(data, drrw->drr_length);
}
- dmu_write(os, drrw->drr_object,
- drrw->drr_offset, drrw->drr_length, data, tx);
+ dmu_assign_arcbuf(bonus, drrw->drr_offset, abuf, tx);
dmu_tx_commit(tx);
+ dmu_buf_rele(bonus, FTAG);
return (0);
}
int err;
guid_map_entry_t gmesrch;
guid_map_entry_t *gmep;
- avl_index_t where;
+ avl_index_t where;
objset_t *ref_os = NULL;
dmu_buf_t *dbp;
err = dmu_buf_hold(ref_os, drrwbr->drr_refobject,
drrwbr->drr_refoffset, FTAG, &dbp, DMU_READ_PREFETCH);
- if (err)
+ if (err != 0)
return (err);
tx = dmu_tx_create(os);
return (0);
}
+static int
+restore_write_embedded(struct restorearg *ra, objset_t *os,
+ struct drr_write_embedded *drrwnp)
+{
+ dmu_tx_t *tx;
+ int err;
+ void *data;
+
+ if (drrwnp->drr_offset + drrwnp->drr_length < drrwnp->drr_offset)
+ return (EINVAL);
+
+ if (drrwnp->drr_psize > BPE_PAYLOAD_SIZE)
+ return (EINVAL);
+
+ if (drrwnp->drr_etype >= NUM_BP_EMBEDDED_TYPES)
+ return (EINVAL);
+ if (drrwnp->drr_compression >= ZIO_COMPRESS_FUNCTIONS)
+ return (EINVAL);
+
+ data = restore_read(ra, P2ROUNDUP(drrwnp->drr_psize, 8), NULL);
+ if (data == NULL)
+ return (ra->err);
+
+ tx = dmu_tx_create(os);
+
+ dmu_tx_hold_write(tx, drrwnp->drr_object,
+ drrwnp->drr_offset, drrwnp->drr_length);
+ err = dmu_tx_assign(tx, TXG_WAIT);
+ if (err != 0) {
+ dmu_tx_abort(tx);
+ return (err);
+ }
+
+ dmu_write_embedded(os, drrwnp->drr_object,
+ drrwnp->drr_offset, data, drrwnp->drr_etype,
+ drrwnp->drr_compression, drrwnp->drr_lsize, drrwnp->drr_psize,
+ ra->byteswap ^ ZFS_HOST_BYTEORDER, tx);
+
+ dmu_tx_commit(tx);
+ return (0);
+}
+
static int
restore_spill(struct restorearg *ra, objset_t *os, struct drr_spill *drrs)
{
drrs->drr_length > SPA_MAXBLOCKSIZE)
return (SET_ERROR(EINVAL));
- data = restore_read(ra, drrs->drr_length);
+ data = restore_read(ra, drrs->drr_length, NULL);
if (data == NULL)
return (ra->err);
*/
VERIFY0(dmu_objset_from_ds(drc->drc_ds, &os));
- ASSERT(drc->drc_ds->ds_phys->ds_flags & DS_FLAG_INCONSISTENT);
+ ASSERT(dsl_dataset_phys(drc->drc_ds)->ds_flags & DS_FLAG_INCONSISTENT);
featureflags = DMU_GET_FEATUREFLAGS(drc->drc_drrb->drr_versioninfo);
*/
pcksum = ra.cksum;
while (ra.err == 0 &&
- NULL != (drr = restore_read(&ra, sizeof (*drr)))) {
+ NULL != (drr = restore_read(&ra, sizeof (*drr), NULL))) {
if (issig(JUSTLOOKING) && issig(FORREAL)) {
ra.err = SET_ERROR(EINTR);
goto out;
ra.err = restore_write_byref(&ra, os, &drrwbr);
break;
}
+ case DRR_WRITE_EMBEDDED:
+ {
+ struct drr_write_embedded drrwe =
+ drr->drr_u.drr_write_embedded;
+ ra.err = restore_write_embedded(&ra, os, &drrwe);
+ break;
+ }
case DRR_FREE:
{
struct drr_free drrf = drr->drr_u.drr_free;
* the snap before drc_ds, because drc_ds can not
* have any snaps of its own).
*/
- uint64_t obj = origin_head->ds_phys->ds_prev_snap_obj;
- while (obj != drc->drc_ds->ds_phys->ds_prev_snap_obj) {
+ uint64_t obj;
+
+ obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
+ while (obj !=
+ dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
dsl_dataset_t *snap;
error = dsl_dataset_hold_obj(dp, obj, FTAG,
&snap);
error = dsl_destroy_snapshot_check_impl(
snap, B_FALSE);
}
- obj = snap->ds_phys->ds_prev_snap_obj;
+ obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_dataset_rele(snap, FTAG);
if (error != 0)
return (error);
}
}
error = dsl_dataset_clone_swap_check_impl(drc->drc_ds,
- origin_head, drc->drc_force);
+ origin_head, drc->drc_force, drc->drc_owner, tx);
if (error != 0) {
dsl_dataset_rele(origin_head, FTAG);
return (error);
}
error = dsl_dataset_snapshot_check_impl(origin_head,
- drc->drc_tosnap, tx, B_TRUE);
+ drc->drc_tosnap, tx, B_TRUE, 1, drc->drc_cred);
dsl_dataset_rele(origin_head, FTAG);
if (error != 0)
return (error);
error = dsl_destroy_head_check_impl(drc->drc_ds, 1);
} else {
error = dsl_dataset_snapshot_check_impl(drc->drc_ds,
- drc->drc_tosnap, tx, B_TRUE);
+ drc->drc_tosnap, tx, B_TRUE, 1, drc->drc_cred);
}
return (error);
}
* Destroy any snapshots of drc_tofs (origin_head)
* after the origin (the snap before drc_ds).
*/
- uint64_t obj = origin_head->ds_phys->ds_prev_snap_obj;
- while (obj != drc->drc_ds->ds_phys->ds_prev_snap_obj) {
+ uint64_t obj;
+
+ obj = dsl_dataset_phys(origin_head)->ds_prev_snap_obj;
+ while (obj !=
+ dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj) {
dsl_dataset_t *snap;
VERIFY0(dsl_dataset_hold_obj(dp, obj, FTAG,
&snap));
ASSERT3P(snap->ds_dir, ==, origin_head->ds_dir);
- obj = snap->ds_phys->ds_prev_snap_obj;
+ obj = dsl_dataset_phys(snap)->ds_prev_snap_obj;
dsl_destroy_snapshot_sync_impl(snap,
B_FALSE, tx);
dsl_dataset_rele(snap, FTAG);
/* set snapshot's creation time and guid */
dmu_buf_will_dirty(origin_head->ds_prev->ds_dbuf, tx);
- origin_head->ds_prev->ds_phys->ds_creation_time =
+ dsl_dataset_phys(origin_head->ds_prev)->ds_creation_time =
drc->drc_drrb->drr_creation_time;
- origin_head->ds_prev->ds_phys->ds_guid =
+ dsl_dataset_phys(origin_head->ds_prev)->ds_guid =
drc->drc_drrb->drr_toguid;
- origin_head->ds_prev->ds_phys->ds_flags &=
+ dsl_dataset_phys(origin_head->ds_prev)->ds_flags &=
~DS_FLAG_INCONSISTENT;
dmu_buf_will_dirty(origin_head->ds_dbuf, tx);
- origin_head->ds_phys->ds_flags &= ~DS_FLAG_INCONSISTENT;
+ dsl_dataset_phys(origin_head)->ds_flags &=
+ ~DS_FLAG_INCONSISTENT;
dsl_dataset_rele(origin_head, FTAG);
dsl_destroy_head_sync_impl(drc->drc_ds, tx);
+
+ if (drc->drc_owner != NULL)
+ VERIFY3P(origin_head->ds_owner, ==, drc->drc_owner);
} else {
dsl_dataset_t *ds = drc->drc_ds;
/* set snapshot's creation time and guid */
dmu_buf_will_dirty(ds->ds_prev->ds_dbuf, tx);
- ds->ds_prev->ds_phys->ds_creation_time =
+ dsl_dataset_phys(ds->ds_prev)->ds_creation_time =
drc->drc_drrb->drr_creation_time;
- ds->ds_prev->ds_phys->ds_guid = drc->drc_drrb->drr_toguid;
- ds->ds_prev->ds_phys->ds_flags &= ~DS_FLAG_INCONSISTENT;
+ dsl_dataset_phys(ds->ds_prev)->ds_guid =
+ drc->drc_drrb->drr_toguid;
+ dsl_dataset_phys(ds->ds_prev)->ds_flags &=
+ ~DS_FLAG_INCONSISTENT;
dmu_buf_will_dirty(ds->ds_dbuf, tx);
- ds->ds_phys->ds_flags &= ~DS_FLAG_INCONSISTENT;
+ dsl_dataset_phys(ds)->ds_flags &= ~DS_FLAG_INCONSISTENT;
}
- drc->drc_newsnapobj = drc->drc_ds->ds_phys->ds_prev_snap_obj;
+ drc->drc_newsnapobj = dsl_dataset_phys(drc->drc_ds)->ds_prev_snap_obj;
/*
* Release the hold from dmu_recv_begin. This must be done before
* we return to open context, so that when we free the dataset's dnode,
gmep = kmem_alloc(sizeof (*gmep), KM_SLEEP);
err = dsl_dataset_hold_obj(dp, snapobj, gmep, &snapds);
if (err == 0) {
- gmep->guid = snapds->ds_phys->ds_guid;
+ gmep->guid = dsl_dataset_phys(snapds)->ds_guid;
gmep->gme_ds = snapds;
avl_add(guid_map, gmep);
dsl_dataset_long_hold(snapds, gmep);
}
int
-dmu_recv_end(dmu_recv_cookie_t *drc)
+dmu_recv_end(dmu_recv_cookie_t *drc, void *owner)
{
+ drc->drc_owner = owner;
+
if (drc->drc_newfs)
return (dmu_recv_new_end(drc));
else
return (dmu_recv_existing_end(drc));
}
+
+/*
+ * Return TRUE if this objset is currently being received into.
+ */
+boolean_t
+dmu_objset_is_receiving(objset_t *os)
+{
+ return (os->os_dsl_dataset != NULL &&
+ os->os_dsl_dataset->ds_owner == dmu_recv_tag);
+}
+
+#if defined(_KERNEL)
+module_param(zfs_send_corrupt_data, int, 0644);
+MODULE_PARM_DESC(zfs_send_corrupt_data, "Allow sending corrupt data");
+#endif
projects / mirror_zfs.git / blobdiff