Re: [RFC PATCH 0/2] Fix up the NFS mmap code

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

On Fri, Jan 08, 2010 at 06:22:01PM -0800, Linus Torvalds wrote:
> 
> 
> On Sat, 9 Jan 2010, Al Viro wrote:
> > 
> > Well, sure - it steps on i_mutex-before-mmmap_sem first from ls somewhere and
> > records the ordering for posterity.  Then NFS steps into mmap() (on a
> > different inode) and gets conflicting ordering.
> 
> Look closer: the inodes for directories and for non-directories have 
> i_mutex in different lockdep classes.
> 
> So that "on a different inode" thing should have made it a non-issue, 
> since there is no actual chain back. There is "mmap_sem -> 
> i_mutex_regular_file" (for mmap) and there is "i_mutex_directory -> 
> mmap_sem" (for filldir), but that isn't an ABBA.
> 
> The problem _seems_ to be (if I read Andi's chain correctly) that a 
> directory hasn't gone through the i_mutex_dir_key change, so filldir ends 
> up being counted against the default i_mutex_key.

Interesting...  There's nfs_update_inode(), but it ought to scream bloody
murder on the type change (and its return value is ignore by nfs_fhget(),
BTW).