Add ddt_object_count() error handling

The interface for the ddt_zap_count() function assumes it can
never fail.  However, internally ddt_zap_count() is implemented
with zap_count() which can potentially fail.  Now because there
was no way to return the error to the caller a VERIFY was used
to ensure this case never happens.

Unfortunately, it has been observed that pools can be damaged in
such a way that zap_count() fails.  The result is that the pool can
not be imported without hitting the VERIFY and crashing the system.

This patch reworks ddt_object_count() so the error can be safely
caught and returned to the caller.  This allows a pool which has
be damaged in this way to be safely rewound for import.

Signed-off-by: Brian Behlendorf <[email protected]>
Closes #910

2 files changed, 22 insertions, 15 deletions

diff --git a/module/zfs/ddt.c b/module/zfs/ddt.c
index 895422b50..6221157f8 100644
--- a/module/zfs/ddt.c
+++ b/module/zfs/ddt.c
@@ -82,13 +82,14 @@ ddt_object_destroy(ddt_t *ddt, enum ddt_type type, enum ddt_class class,
 	spa_t *spa = ddt->ddt_spa;
 	objset_t *os = ddt->ddt_os;
 	uint64_t *objectp = &ddt->ddt_object[type][class];
+	uint64_t count;
 	char name[DDT_NAMELEN];
 
 	ddt_object_name(ddt, type, class, name);
 
 	ASSERT(*objectp != 0);
-	ASSERT(ddt_object_count(ddt, type, class) == 0);
 	ASSERT(ddt_histogram_empty(&ddt->ddt_histogram[type][class]));
+	VERIFY(ddt_object_count(ddt, type, class, &count) == 0 && count == 0);
 	VERIFY(zap_remove(os, DMU_POOL_DIRECTORY_OBJECT, name, tx) == 0);
 	VERIFY(zap_remove(os, spa->spa_ddt_stat_object, name, tx) == 0);
 	VERIFY(ddt_ops[type]->ddt_op_destroy(os, *objectp, tx) == 0);
@@ -102,6 +103,7 @@ ddt_object_load(ddt_t *ddt, enum ddt_type type, enum ddt_class class)
 {
 	ddt_object_t *ddo = &ddt->ddt_object_stats[type][class];
 	dmu_object_info_t doi;
+	uint64_t count;
 	char name[DDT_NAMELEN];
 	int error;
 
@@ -124,7 +126,11 @@ ddt_object_load(ddt_t *ddt, enum ddt_type type, enum ddt_class class)
 	if (error)
 		return (error);
 
-	ddo->ddo_count = ddt_object_count(ddt, type, class);
+	error = ddt_object_count(ddt, type, class, &count);
+	if (error)
+		return (error);
+
+	ddo->ddo_count = count;
 	ddo->ddo_dspace = doi.doi_physical_blocks_512 << 9;
 	ddo->ddo_mspace = doi.doi_fill_count * doi.doi_data_block_size;
 
@@ -138,6 +144,7 @@ ddt_object_sync(ddt_t *ddt, enum ddt_type type, enum ddt_class class,
 {
 	ddt_object_t *ddo = &ddt->ddt_object_stats[type][class];
 	dmu_object_info_t doi;
+	uint64_t count;
 	char name[DDT_NAMELEN];
 
 	ddt_object_name(ddt, type, class, name);
@@ -150,8 +157,9 @@ ddt_object_sync(ddt_t *ddt, enum ddt_type type, enum ddt_class class,
 	 * Cache DDT statistics; this is the only time they'll change.
 	 */
 	VERIFY(ddt_object_info(ddt, type, class, &doi) == 0);
+	VERIFY(ddt_object_count(ddt, type, class, &count) == 0);
 
-	ddo->ddo_count = ddt_object_count(ddt, type, class);
+	ddo->ddo_count = count;
 	ddo->ddo_dspace = doi.doi_physical_blocks_512 << 9;
 	ddo->ddo_mspace = doi.doi_fill_count * doi.doi_data_block_size;
 }
@@ -208,13 +216,14 @@ ddt_object_walk(ddt_t *ddt, enum ddt_type type, enum ddt_class class,
 	    ddt->ddt_object[type][class], dde, walk));
 }
 
-uint64_t
-ddt_object_count(ddt_t *ddt, enum ddt_type type, enum ddt_class class)
+int
+ddt_object_count(ddt_t *ddt, enum ddt_type type, enum ddt_class class,
+    uint64_t *count)
 {
 	ASSERT(ddt_object_exists(ddt, type, class));
 
 	return (ddt_ops[type]->ddt_op_count(ddt->ddt_os,
-	    ddt->ddt_object[type][class]));
+	    ddt->ddt_object[type][class], count));
 }
 
 int
@@ -1124,11 +1133,13 @@ ddt_sync_table(ddt_t *ddt, dmu_tx_t *tx, uint64_t txg)
 	}
 
 	for (type = 0; type < DDT_TYPES; type++) {
-		uint64_t count = 0;
+		uint64_t add, count = 0;
 		for (class = 0; class < DDT_CLASSES; class++) {
 			if (ddt_object_exists(ddt, type, class)) {
 				ddt_object_sync(ddt, type, class, tx);
-				count += ddt_object_count(ddt, type, class);
+				VERIFY(ddt_object_count(ddt, type, class,
+				    &add) == 0);
+				count += add;
 			}
 		}
 		for (class = 0; class < DDT_CLASSES; class++) {
diff --git a/module/zfs/ddt_zap.c b/module/zfs/ddt_zap.c
index 1733448a8..65b14ab63 100644
--- a/module/zfs/ddt_zap.c
+++ b/module/zfs/ddt_zap.c
@@ -138,14 +138,10 @@ ddt_zap_walk(objset_t *os, uint64_t object, ddt_entry_t *dde, uint64_t *walk)
 	return (error);
 }
 
-static uint64_t
-ddt_zap_count(objset_t *os, uint64_t object)
+static int
+ddt_zap_count(objset_t *os, uint64_t object, uint64_t *count)
 {
-	uint64_t count = 0;
-
-	VERIFY(zap_count(os, object, &count) == 0);
-
-	return (count);
+	return zap_count(os, object, count);
 }
 
 const ddt_ops_t ddt_zap_ops = {