On Fri, Oct 28, 2016 at 10:02:06AM -0500, Goldwyn Rodrigues wrote: > > > On 10/28/2016 09:42 AM, David Sterba wrote: > > On Thu, Oct 27, 2016 at 01:05:08PM -0500, Goldwyn Rodrigues wrote: > >> While deleting pending extents, we insert extents to the > >> extent_tree. However, because of corruption, this might already > >> be freed. We trickle the EEXISTS error and let the caller > >> decide what needs to be done with it. In this case, ignore, so > >> the repair may proceed. > >> > >> The primary motivation of this is to resolve this BUG_ON: > >> extent-tree.c:2731: alloc_reserved_tree_block: Assertion `ret` failed, > >> value 0 > >> > >> #2 assert_trace (val=0, line=2728, func=0x489510 <__func__.11558> > >> "alloc_reserved_tree_block", > >> filename=0x489065 "extent-tree.c", assertion=0x48cfb5 "ret") at > >> kerncompat.h:102 > >> #3 alloc_reserved_tree_block (trans=trans@entry=0x7e7240, > >> root=root@entry=0x6a88a0, root_objectid=2, > >> generation=175654, flags=0, key=key@entry=0x2bc5c70, level=0, > >> ins=ins@entry=0x7fffffffcf30) at extent-tree.c:2728 > >> #4 finish_current_insert (trans=trans@entry=0x7e7240, > >> extent_root=extent_root@entry=0x6a88a0) > >> at extent-tree.c:2108 > >> #5 __free_extent (trans=trans@entry=0x7e7240, root=root@entry=0x6a88a0, > >> bytenr=57127387136, > >> num_bytes=<optimized out>, parent=parent@entry=0, > >> root_objectid=<optimized out>, owner_objectid=0, > >> owner_offset=owner_offset@entry=0, refs_to_drop=refs_to_drop@entry=1) > >> at extent-tree.c:2410 > >> #6 del_pending_extents (trans=trans@entry=0x7e7240, extent_root=0x6a88a0) > >> at extent-tree.c:2448 > > > > Do you have a reproducer for that? > > I have a corrupt btrfs-image, not a test case though. Should I share > that? If yes, where should I upload it?
The image should not be large (compressed), if it's about to be stored in git. Minimizing can take time, I haven't analyzed the trace if there's an easier way to trigger it (eg corrupting a specific key). > >> Signed-off-by: Goldwyn Rodrigues <rgold...@suse.com> > > > > Applied, thanks. > > > > Ahh, just realized. Spell check s/EEXISTS/EEXIST/ Fixed. -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
