MFC r277419:

Allow skipping dmu_buf_will_dirty() call in dsl_dir_transfer_space().

dsl_dir_transfer_space() is mostly called after dsl_dir_diduse_space(),
which already calls dmu_buf_will_dirty() for the same dbuf and tx, so
its duplicate call in those cases will change nothing, only spend time.

Skipping this call by four times reduces time spent in dbuf_write_done()
and descendants, updating dataset statistics with several congested lock
acquisitions.  When rewriting 8K zvol blocks at 1GB/s rate, this reduces
CPU time spent inside dbuf_write_done(), according to profiling, from 45%
of 683K samples to 18% of 422K.

1 files changed, 4 insertions, 3 deletions

diff --git a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_dir.c b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_dir.c
index 5a891d1b..1862f8c 100644
--- a/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_dir.c
+++ b/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/dsl_dir.c
@@ -1389,7 +1389,7 @@ dsl_dir_diduse_space(dsl_dir_t *dd, dd_used_t type,
 		    accounted_delta, compressed, uncompressed, tx);
 		dsl_dir_transfer_space(dd->dd_parent,
 		    used - accounted_delta,
-		    DD_USED_CHILD_RSRV, DD_USED_CHILD, tx);
+		    DD_USED_CHILD_RSRV, DD_USED_CHILD, NULL);
 	}
 }
 
@@ -1397,7 +1397,7 @@ void
 dsl_dir_transfer_space(dsl_dir_t *dd, int64_t delta,
     dd_used_t oldtype, dd_used_t newtype, dmu_tx_t *tx)
 {
-	ASSERT(dmu_tx_is_syncing(tx));
+	ASSERT(tx == NULL || dmu_tx_is_syncing(tx));
 	ASSERT(oldtype < DD_USED_NUM);
 	ASSERT(newtype < DD_USED_NUM);
 
@@ -1405,7 +1405,8 @@ dsl_dir_transfer_space(dsl_dir_t *dd, int64_t delta,
 	    !(dsl_dir_phys(dd)->dd_flags & DD_FLAG_USED_BREAKDOWN))
 		return;
 
-	dmu_buf_will_dirty(dd->dd_dbuf, tx);
+	if (tx != NULL)
+		dmu_buf_will_dirty(dd->dd_dbuf, tx);
 	mutex_enter(&dd->dd_lock);
 	ASSERT(delta > 0 ?
 	    dsl_dir_phys(dd)->dd_used_breakdown[oldtype] >= delta :