Illumos #755: dmu_recv_stream builds incomplete guid_to_ds_map

An incomplete guid_to_ds_map would cause restore_write_byref() to fail
while receiving a de-duplicated backup stream.

Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Garrett D`Amore <garrett@nexenta.com>
Reviewed by: Gordon Ross <gwr@nexenta.com>
Approved by: Gordon Ross <gwr@nexenta.com>

References to Illumos issue and patch:
- https://www.illumos.org/issues/755
- https://github.com/illumos/illumos-gate/commit/ec5cf9d53a

Signed-off-by: Gunnar Beutner <gunnar@beutner.name>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #372

diff --git a/include/sys/dmu.h b/include/sys/dmu.h
index 20733f5023271b853519affb869b862228c7b0cf..da1aa30fc73ff7b9a98bed452e5a6620ff1ea7a2 100644 (file)
--- a/include/sys/dmu.h
+++ b/include/sys/dmu.h
@@ -721,6 +721,7 @@ typedef struct dmu_recv_cookie {
        char *drc_top_ds;
        boolean_t drc_newfs;
        boolean_t drc_force;
+       struct avl_tree *drc_guid_to_ds_map;
 } dmu_recv_cookie_t;
 
 int dmu_recv_begin(char *tofs, char *tosnap, char *topds, struct drr_begin *,

diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c
index ae0b36fc6380bd6f7756684218597013bb30cb4a..fad770e275c0a87e33b1a0af8bdb25be017a171f 100644 (file)
--- a/module/zfs/dmu_send.c
+++ b/module/zfs/dmu_send.c
@@ -21,6 +21,9 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  */
+/*
+ * Copyright 2011 Nexenta Systems, Inc. All rights reserved.
+ */
 
 #include <sys/dmu.h>
 #include <sys/dmu_impl.h>
@@ -833,61 +836,6 @@ guid_compare(const void *arg1, const void *arg2)
        return (0);
 }
 
-/*
- * This function is a callback used by dmu_objset_find() (which
- * enumerates the object sets) to build an avl tree that maps guids
- * to datasets.  The resulting table is used when processing DRR_WRITE_BYREF
- * send stream records.  These records, which are used in dedup'ed
- * streams, do not contain data themselves, but refer to a copy
- * of the data block that has already been written because it was
- * earlier in the stream.  That previous copy is identified by the
- * guid of the dataset with the referenced data.
- */
-int
-find_ds_by_guid(const char *name, void *arg)
-{
-       avl_tree_t *guid_map = arg;
-       dsl_dataset_t *ds, *snapds;
-       guid_map_entry_t *gmep;
-       dsl_pool_t *dp;
-       int err;
-       uint64_t lastobj, firstobj;
-
-       if (dsl_dataset_hold(name, FTAG, &ds) != 0)
-               return (0);
-
-       dp = ds->ds_dir->dd_pool;
-       rw_enter(&dp->dp_config_rwlock, RW_READER);
-       firstobj = ds->ds_dir->dd_phys->dd_origin_obj;
-       lastobj = ds->ds_phys->ds_prev_snap_obj;
-
-       while (lastobj != firstobj) {
-               err = dsl_dataset_hold_obj(dp, lastobj, guid_map, &snapds);
-               if (err) {
-                       /*
-                        * Skip this snapshot and move on. It's not
-                        * clear why this would ever happen, but the
-                        * remainder of the snapshot streadm can be
-                        * processed.
-                        */
-                       rw_exit(&dp->dp_config_rwlock);
-                       dsl_dataset_rele(ds, FTAG);
-                       return (0);
-               }
-
-               gmep = kmem_alloc(sizeof (guid_map_entry_t), KM_SLEEP);
-               gmep->guid = snapds->ds_phys->ds_guid;
-               gmep->gme_ds = snapds;
-               avl_add(guid_map, gmep);
-               lastobj = snapds->ds_phys->ds_prev_snap_obj;
-       }
-
-       rw_exit(&dp->dp_config_rwlock);
-       dsl_dataset_rele(ds, FTAG);
-
-       return (0);
-}
-
 static void
 free_guid_map_onexit(void *arg)
 {
@@ -1373,9 +1321,6 @@ dmu_recv_stream(dmu_recv_cookie_t *drc, vnode_t *vp, offset_t *voffp,
                        avl_create(ra.guid_to_ds_map, guid_compare,
                            sizeof (guid_map_entry_t),
                            offsetof(guid_map_entry_t, avlnode));
-                       (void) dmu_objset_find(drc->drc_top_ds, find_ds_by_guid,
-                           (void *)ra.guid_to_ds_map,
-                           DS_FIND_CHILDREN);
                        ra.err = zfs_onexit_add_cb(minor,
                            free_guid_map_onexit, ra.guid_to_ds_map,
                            action_handlep);
@@ -1387,6 +1332,8 @@ dmu_recv_stream(dmu_recv_cookie_t *drc, vnode_t *vp, offset_t *voffp,
                        if (ra.err)
                                goto out;
                }
+
+               drc->drc_guid_to_ds_map = ra.guid_to_ds_map;
        }
 
        /*
@@ -1524,6 +1471,30 @@ recv_end_sync(void *arg1, void *arg2, dmu_tx_t *tx)
        ds->ds_phys->ds_flags &= ~DS_FLAG_INCONSISTENT;
 }
 
+static int
+add_ds_to_guidmap(avl_tree_t *guid_map, dsl_dataset_t *ds)
+{
+       dsl_pool_t *dp = ds->ds_dir->dd_pool;
+       uint64_t snapobj = ds->ds_phys->ds_prev_snap_obj;
+       dsl_dataset_t *snapds;
+       guid_map_entry_t *gmep;
+       int err;
+
+       ASSERT(guid_map != NULL);
+
+       rw_enter(&dp->dp_config_rwlock, RW_READER);
+       err = dsl_dataset_hold_obj(dp, snapobj, guid_map, &snapds);
+       if (err == 0) {
+               gmep = kmem_alloc(sizeof (guid_map_entry_t), KM_SLEEP);
+               gmep->guid = snapds->ds_phys->ds_guid;
+               gmep->gme_ds = snapds;
+               avl_add(guid_map, gmep);
+       }
+
+       rw_exit(&dp->dp_config_rwlock);
+       return (err);
+}
+
 static int
 dmu_recv_existing_end(dmu_recv_cookie_t *drc)
 {
@@ -1564,6 +1535,8 @@ dmu_recv_existing_end(dmu_recv_cookie_t *drc)
 
 out:
        mutex_exit(&ds->ds_recvlock);
+       if (err == 0 && drc->drc_guid_to_ds_map != NULL)
+               (void) add_ds_to_guidmap(drc->drc_guid_to_ds_map, ds);
        dsl_dataset_disown(ds, dmu_recv_tag);
        (void) dsl_dataset_destroy(drc->drc_real_ds, dmu_recv_tag, B_FALSE);
        return (err);
@@ -1593,6 +1566,8 @@ dmu_recv_new_end(dmu_recv_cookie_t *drc)
                /* clean up the fs we just recv'd into */
                (void) dsl_dataset_destroy(ds, dmu_recv_tag, B_FALSE);
        } else {
+               if (drc->drc_guid_to_ds_map != NULL)
+                       (void) add_ds_to_guidmap(drc->drc_guid_to_ds_map, ds);
                /* release the hold from dmu_recv_begin */
                dsl_dataset_disown(ds, dmu_recv_tag);
        }