On Wed, Sep 10, 2025 at 05:12:49AM +0100, Al Viro wrote: > On Wed, Sep 10, 2025 at 01:01:49PM +1000, NeilBrown wrote: > > On Tue, 09 Sep 2025, Al Viro wrote: > > > On Tue, Sep 09, 2025 at 02:43:21PM +1000, NeilBrown wrote: > > > > > > > d_instantiate(dentry, inode); > > > > - dget(dentry); > > > > -fail: > > > > - inode_unlock(d_inode(parent)); > > > > - return dentry; > > > > + return simple_end_creating(dentry); > > > > > > No. This is the wrong model - dget() belongs with d_instantiate() > > > here; your simple_end_creating() calling conventions are wrong. > > > > I can see that I shouldn't have removed the dget() there - thanks. > > It is not entirely clear why hypfs_create_file() returns with two > > references held to the dentry.... > > I see now one is added either to ->update_file or the list at > > hypfs_last_dentry, and the other is disposed of by kill_litter_super(). > > > > But apart from that one error is there something broader wrong with the > > patch? You say "the wrong model" but I don't see it. > > See below for hypfs: ... and see viro/vfs.git#work.persistency for the part of the queue that had order already settled down (I'm reshuffling the tail at the moment; hypfs commit is still in the leftovers pile - the whole thing used to have a really messy topology, with most of the prep work that used to be the cause of that topology already in mainline - e.g. rpc_pipefs series, securityfs one, etc.)
