Ceph » CephFS

** CID 1412578:  Null pointer dereferences  (FORWARD_NULL)
/home/brad/working/src/ceph/src/mds/Server.cc: 7032 in Server::_rename_prepare(boost::intrusive_ptr<MDRequestImpl> &, EMetaBlob *, ceph::buffer::list *, CDentry *, CDentry *, CDentry *)()

________________________________________________________________________________________________________
*** CID 1412578:  Null pointer dereferences  (FORWARD_NULL)
/home/brad/working/src/ceph/src/mds/Server.cc: 7032 in Server::_rename_prepare(boost::intrusive_ptr<MDRequestImpl> &, EMetaBlob *, ceph::buffer::list *, CDentry *, CDentry *, CDentry *)()
7026           force_journal_dest = true;
7027         } else
7028           force_journal_dest = _need_force_journal(srci, false);
7029       }
7030
7031       bool force_journal_stray = false;
>>>     CID 1412578:  Null pointer dereferences  (FORWARD_NULL)
>>>     Comparing "straydn" to null implies that "straydn" might be null.
7032       if (oldin && oldin->is_dir() && straydn && !straydn->is_auth())
7033         force_journal_stray = _need_force_journal(oldin, true);
7034
7035       if (linkmerge)
7036         dout(10) << " merging remote and primary links to the same inode" << dendl;
7037       if (silent)

From Brad Hubbard's static analysis roundup.

Looks like Coverity is complaining that we do a straydn NULL check here but later in the function do not check for NULL. I'm not sure if Coverity is smart enough to understand the assert(straydn) calls where we assume it cannot be NULL. There are three cases where we don't do the assert:

https://github.com/ceph/ceph/blob/9f259cb923e7474229bb2e5e54b59f3b7f068bc5/src/mds/Server.cc#L7164
https://github.com/ceph/ceph/blob/9f259cb923e7474229bb2e5e54b59f3b7f068bc5/src/mds/Server.cc#L7184-L7194
https://github.com/ceph/ceph/blob/9f259cb923e7474229bb2e5e54b59f3b7f068bc5/src/mds/Server.cc#L7278