Re: [RFC] situation with fput() locking (was Re: [PULL REQUEST] : ima-appraisal patches)

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

On Fri, Apr 20, 2012 at 03:54:38AM +0100, Al Viro wrote:

> and filp_close() would, if that turns out to be possible, call fput_nodefer()
> instead of fput().  If we *do* have places where we need deferral in
> filp_close() (and I'm fairly sure that any such place is a deadlock right
> now), well, we'll need a variant of filp_close() sans the call of fput...()
> and those places would call that, followed by full (deferring) fput().

BTW, some of those filp_close() are simply wrong - e.g. a big pile in
drivers/media/video/cx25821/*.c should be fput().

And yes, we have at least some under mutex - binder_lock held by
binder_ioctl(), which ends up doing binder_transaction() with its
failure cleanup hitting close_filp().  On an arbitrary struct file.
It *probably* doesn't deadlock on the spot, since binder_release()
itself contains a deferral mechanism (perhaps they'd spotted the
deadlock, perhaps there are other reasons to have it).