ocfs2: Remove overzealous BUG_ON()
authorMark Fasheh <mark.fasheh@oracle.com>
Wed, 9 Aug 2006 20:23:08 +0000 (13:23 -0700)
committerMark Fasheh <mark.fasheh@oracle.com>
Wed, 20 Sep 2006 23:00:54 +0000 (16:00 -0700)
The truncate code was never supposed to BUG() on an allocator it doesn't
know about, but rather to ignore it. Right now, this does nothing, but when
we change our allocation paths to use all suballocator files, this will
allow current versions of the fs module to work fine.

Signed-off-by: Mark Fasheh <mark.fasheh@oracle.com>
fs/ocfs2/alloc.c

index edaab05..f43bc5f 100644 (file)
@@ -1717,17 +1717,29 @@ static int ocfs2_do_truncate(struct ocfs2_super *osb,
 
                        ocfs2_remove_from_cache(inode, eb_bh);
 
-                       BUG_ON(eb->h_suballoc_slot);
                        BUG_ON(el->l_recs[0].e_clusters);
                        BUG_ON(el->l_recs[0].e_cpos);
                        BUG_ON(el->l_recs[0].e_blkno);
-                       status = ocfs2_free_extent_block(handle,
-                                                        tc->tc_ext_alloc_inode,
-                                                        tc->tc_ext_alloc_bh,
-                                                        eb);
-                       if (status < 0) {
-                               mlog_errno(status);
-                               goto bail;
+                       if (eb->h_suballoc_slot == 0) {
+                               /*
+                                * This code only understands how to
+                                * lock the suballocator in slot 0,
+                                * which is fine because allocation is
+                                * only ever done out of that
+                                * suballocator too. A future version
+                                * might change that however, so avoid
+                                * a free if we don't know how to
+                                * handle it. This way an fs incompat
+                                * bit will not be necessary.
+                                */
+                               status = ocfs2_free_extent_block(handle,
+                                                                tc->tc_ext_alloc_inode,
+                                                                tc->tc_ext_alloc_bh,
+                                                                eb);
+                               if (status < 0) {
+                                       mlog_errno(status);
+                                       goto bail;
+                               }
                        }
                }
                brelse(eb_bh);