Zero pad bytes when allocating a ZIL record

When allocating a record, we round up the allocation size to a multiple
of 8.  In this case, any padding bytes should be zeroed, otherwise the
contents of uninitialized memory are written to the ZIL.

This was found using KMSAN.

Reviewed-by: Brian Behlendorf <[email protected]>
Reviewed-by: Alexander Motin <[email protected]>
Signed-off-by: Mark Johnston <[email protected]>
Closes #12383

Author: markjdb

module/zfs/zil.c

@@ -1783,18 +1783,19 @@ zil_lwb_commit(zilog_t *zilog, itx_t *itx, lwb_t *lwb)
 }
 
 itx_t *
-zil_itx_create(uint64_t txtype, size_t lrsize)
+zil_itx_create(uint64_t txtype, size_t olrsize)
 {
-	size_t itxsize;
+	size_t itxsize, lrsize;
 	itx_t *itx;
 
-	lrsize = P2ROUNDUP_TYPED(lrsize, sizeof (uint64_t), size_t);
+	lrsize = P2ROUNDUP_TYPED(olrsize, sizeof (uint64_t), size_t);
 	itxsize = offsetof(itx_t, itx_lr) + lrsize;
 
 	itx = zio_data_buf_alloc(itxsize);
 	itx->itx_lr.lrc_txtype = txtype;
 	itx->itx_lr.lrc_reclen = lrsize;
 	itx->itx_lr.lrc_seq = 0;	/* defensive */
+	bzero((char *)&itx->itx_lr + olrsize, lrsize - olrsize);
 	itx->itx_sync = B_TRUE;		/* default is synchronous */
 	itx->itx_callback = NULL;
 	itx->itx_callback_data = NULL;