tdb: use same expansion factor logic when expanding for new recovery area.

If we're expanding because the current recovery area is too small, we
expand only the amount we need.  This can quickly lead to exponential
growth when we have a slowly-expanding record (hence a
slowly-expanding transaction size).

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>

3 files changed, 34 insertions, 21 deletions

diff --git a/lib/tdb/common/io.c b/lib/tdb/common/io.c
index a2db3bf4bd..ac21e3f67a 100644
--- a/lib/tdb/common/io.c
+++ b/lib/tdb/common/io.c
@@ -313,41 +313,49 @@ static int tdb_expand_file(struct tdb_context *tdb, tdb_off_t size, tdb_off_t ad
 }
 
 
-/* expand the database at least size bytes by expanding the underlying
-   file and doing the mmap again if necessary */
-int tdb_expand(struct tdb_context *tdb, tdb_off_t size)
+/* You need 'size', this tells you how much you should expand by. */
+tdb_off_t tdb_expand_adjust(tdb_off_t map_size, tdb_off_t size, int page_size)
 {
-	struct tdb_record rec;
-	tdb_off_t offset, new_size, top_size, map_size;
-
-	if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
-		TDB_LOG((tdb, TDB_DEBUG_ERROR, "lock failed in tdb_expand\n"));
-		return -1;
-	}
-
-	/* must know about any previous expansions by another process */
-	tdb->methods->tdb_oob(tdb, tdb->map_size, 1, 1);
+	tdb_off_t new_size, top_size;
 
 	/* limit size in order to avoid using up huge amounts of memory for
 	 * in memory tdbs if an oddball huge record creeps in */
 	if (size > 100 * 1024) {
-		top_size = tdb->map_size + size * 2;
+		top_size = map_size + size * 2;
 	} else {
-		top_size = tdb->map_size + size * 100;
+		top_size = map_size + size * 100;
 	}
 
 	/* always make room for at least top_size more records, and at
 	   least 25% more space. if the DB is smaller than 100MiB,
 	   otherwise grow it by 10% only. */
-	if (tdb->map_size > 100 * 1024 * 1024) {
-		map_size = tdb->map_size * 1.10;
+	if (map_size > 100 * 1024 * 1024) {
+		new_size = map_size * 1.10;
 	} else {
-		map_size = tdb->map_size * 1.25;
+		new_size = map_size * 1.25;
 	}
 
 	/* Round the database up to a multiple of the page size */
-	new_size = MAX(top_size, map_size);
-	size = TDB_ALIGN(new_size, tdb->page_size) - tdb->map_size;
+	new_size = MAX(top_size, new_size);
+	return TDB_ALIGN(new_size, page_size) - map_size;
+}
+
+/* expand the database at least size bytes by expanding the underlying
+   file and doing the mmap again if necessary */
+int tdb_expand(struct tdb_context *tdb, tdb_off_t size)
+{
+	struct tdb_record rec;
+	tdb_off_t offset;
+
+	if (tdb_lock(tdb, -1, F_WRLCK) == -1) {
+		TDB_LOG((tdb, TDB_DEBUG_ERROR, "lock failed in tdb_expand\n"));
+		return -1;
+	}
+
+	/* must know about any previous expansions by another process */
+	tdb->methods->tdb_oob(tdb, tdb->map_size, 1, 1);
+
+	size = tdb_expand_adjust(tdb->map_size, size, tdb->page_size);
 
 	if (!(tdb->flags & TDB_INTERNAL))
 		tdb_munmap(tdb);
diff --git a/lib/tdb/common/tdb_private.h b/lib/tdb/common/tdb_private.h
index 3c6aabffc7..4fc7381aa0 100644
--- a/lib/tdb/common/tdb_private.h
+++ b/lib/tdb/common/tdb_private.h
@@ -271,6 +271,7 @@ tdb_off_t tdb_find_lock_hash(struct tdb_context *tdb, TDB_DATA key, uint32_t has
 			   struct tdb_record *rec);
 void tdb_io_init(struct tdb_context *tdb);
 int tdb_expand(struct tdb_context *tdb, tdb_off_t size);
+tdb_off_t tdb_expand_adjust(tdb_off_t map_size, tdb_off_t size, int page_size);
 int tdb_rec_free_read(struct tdb_context *tdb, tdb_off_t off,
 		      struct tdb_record *rec);
 bool tdb_write_all(int fd, const void *buf, size_t count);
diff --git a/lib/tdb/common/transaction.c b/lib/tdb/common/transaction.c
index 66ecbfdefb..f7d56a05a3 100644
--- a/lib/tdb/common/transaction.c
+++ b/lib/tdb/common/transaction.c
@@ -730,7 +730,11 @@ static int tdb_recovery_allocate(struct tdb_context *tdb,
 	*recovery_size = tdb_recovery_size(tdb);
 
 	/* round up to a multiple of page size */
-	*recovery_max_size = TDB_ALIGN(sizeof(rec) + *recovery_size, tdb->page_size) - sizeof(rec);
+	*recovery_max_size = tdb_expand_adjust(tdb->map_size,
+					       *recovery_size,
+					       tdb->page_size)
+		- sizeof(rec);
+
 	*recovery_offset = tdb->map_size;
 	recovery_head = *recovery_offset;