On Tue, Sep 27, 2011 at 11:11:08AM -0400, Josef Bacik wrote: > Really this shouldn't happen at all, but at least this will keep the > thing from mounting and then we can debug the issue further without > bringing the users box to a halt. Is's not easy to say where the problem originates, we see just the unfixable result. IIRC there were delalloc and enospc involved, and some kind of race, but it's a guesswork after going through the surrounding code. Do you think it's possible to skip the inode rather than stopping the whole orphan cleanup? Orphan cleanup is called from several places and during whole lifecycle of a moutned system (open_ctree, recovery, snapshot, lookup_dentry). The broken inode will be there and prevent further orphan cleanup and reclaim of the used space (although this may be not that many bytes). Seems that btrfs_evict_inode fails to remove the item from the list (via btrfs_orphan_del). david -- To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
