diff options
| author | Ned Bass <[email protected]> | 2017-09-05 16:09:15 -0700 |
|---|---|---|
| committer | Brian Behlendorf <[email protected]> | 2017-09-05 16:09:15 -0700 |
| commit | 65dcb0f67a4d72ee4e1e534703db5caacf1ec85f (patch) | |
| tree | eff9601d9162ede41c171824f2a264fe8e281246 /module/zfs/dmu_send.c | |
| parent | c8811dec7044a126650c7e2d9f3404680ae115b5 (diff) | |
Handle new dnode size in incremental backup stream
When receiving an incremental backup stream, call
dmu_object_reclaim_dnsize() if an object's dnode size differs between
the incremental source and target. Otherwise it may appear that a
dnode which has shrunk is still occupying slots which are in fact
free. This will cause a failure to receive new objects that should
occupy the now-free slots.
Add a test case to verify that an incremental stream containing
objects with changed dnode sizes can be received without error. This
test case fails without this change.
Reviewed-by: Brian Behlendorf <[email protected]>
Signed-off-by: Ned Bass <[email protected]>
Closes #6366
Closes #6576
Diffstat (limited to 'module/zfs/dmu_send.c')
| -rw-r--r-- | module/zfs/dmu_send.c | 8 |
1 files changed, 5 insertions, 3 deletions
diff --git a/module/zfs/dmu_send.c b/module/zfs/dmu_send.c index 359ef99d3..882926b75 100644 --- a/module/zfs/dmu_send.c +++ b/module/zfs/dmu_send.c @@ -2482,11 +2482,13 @@ receive_object(struct receive_writer_arg *rwa, struct drr_object *drro, } 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) { + drro->drr_bonuslen != doi.doi_bonus_size || + drro->drr_dn_slots != (doi.doi_dnodesize >> DNODE_SHIFT)) { /* currently allocated, but with different properties */ - err = dmu_object_reclaim(rwa->os, drro->drr_object, + err = dmu_object_reclaim_dnsize(rwa->os, drro->drr_object, drro->drr_type, drro->drr_blksz, - drro->drr_bonustype, drro->drr_bonuslen, tx); + drro->drr_bonustype, drro->drr_bonuslen, + drro->drr_dn_slots << DNODE_SHIFT, tx); } if (err != 0) { dmu_tx_commit(tx); |