Ah, that explains why we seemed to be talking past each other. For
2.4.x, def_blk_fops.open() will bump the reference count, but in 2.5.x
it won't. So in 2.5.x, fs/devfs/base.c:devfs_open() shouldn't drop the
refcount after def_blk_fops.open() returns.
> > > You never answered my question, why you insist on managing the ops
> > > pointer. The far easier fix would be to simply remove this nonsense.
> > Because it's an optimsation, avoiding the need for looking up ops from
> > tables/lists. It's the sensible way of doing it. I've explained this
> > to others on the list, and in the FAQ. I'm not going to keep going
> > over it again and again.
> Optimization??? This would require any device had to be be opened
> _only_ through devfs, you are not seriously suggesting that???
Huh? Of course not! All I'm saying is that if you use devfs, the
optimisation will short-circuit the lookups.
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to email@example.com
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/