[xfs-masters] RFC: Fix f_flags races without the BKL

[Al Viro]

On Mon, Dec 29, 2008 at 04:13:52AM -0700, Jonathan Corbet wrote:
> Accesses to the f_flags field have always involved a read-modify-write
> operation, and have always been racy in the absence of the BKL.  The recent
> BKL-removal work made this problem worse, but it has been there for a very
> long time.  The race is quite small, and, arguably, has never affected
> anybody, but it's still worth fixing.
> 
> After pondering for a while, I couldn't come up with anything better than a
> global file->f_flags mutex.  There's no point in bloating struct file with
> a mutex just for this purpose; it's hard to imagine that there will be any
> real contention for this lock.

Bloating with mutex is over the top, indeed, but why can't we simply keep
a pointer to fasync_struct in there?  Do we ever have a struct file with
several fasync_struct?  They'd have to be on different queues and I don't
see any cases where that would happen...