Re: [git pull] vfs and fs fixes

[Al Viro <viro@ZenIV.linux.org.uk>]

On Tue, Apr 17, 2012 at 03:08:26PM -0700, Linus Torvalds wrote:
> > Or I could increment that counter for all the conflicting operations and
> > rely on it instead of the i_mutex. ?I was trying to avoid adding
> > something like that (an inc, a dec, another error path) to every
> > operation. ?And hoping to avoid adding another field to struct inode.
> > Oh well.
> 
> We could just say that we can do a double inode lock, but then
> standardize on the order. And the only sane order is comparing inode
> pointers, not inode numbers like ext4 apparently does.
> 
> With a standard order, I don't think it would be at all wrong to just
> take the inode lock on rename.

In principle, yes, but have you tried to grep for i_mutex?  Note that
we have *another* place where multiple ->i_mutex might be held on
non-directories (and unless I'm missing something, ext4 move_extent.c
stuff doesn't play well with it): quota writes.  Which can, AFAICS,
happen while write(2) is holding ->i_mutex on a regular file.  So
it's not _that_ easy - we want something like "and quota file is goes
last", since there we don't get to change the locking order - the first
->i_mutex is taken too far outside.

I really don't like how messy i_mutex had become these days.  Right now
I'm staring at 700-odd lines all over the place where it's taken/released
and it's a wastebucket lock - used to protect random bits and scraps, with a
lot of filesystems, etc. using it for purposes of their own ;-/