Re: v4.2-rc dcache regression, probably 75a6f82a0d10

[Hugh Dickins <hughd@google.com>]

On Fri, 31 Jul 2015, Hugh Dickins wrote:
> On Fri, 31 Jul 2015, Linus Torvalds wrote:
> > 
> > So leave it running a while longer, but maybe it's 4bf46a272647 like
> > Dominique suspects. Although I don't see how that could trigger
> > anything either..
> 
> I restarted with a slightly different version of the load this
> morning, which has sometimes shown the issue more easily - I thought
> it better to restart with a variant than persist with a run that
> might have settled into a protected pattern.  We'll see what that
> shows later on.

It showed nothing useful to this discussion: after an hour and a half
it had hung on some almost-certainly-unrelated issue that I've never
seen before - looked as if a jbd2 transaction never completed, some
tasks waiting for that, some waiting for f_pos_lock held by those
(which I hit when I tried to tail the output log).  Worry about
that another time, if it ever shows up again.

I think I'll try reinstating Al's commit, and hacking out that change
of David's in dentry_iput() that worried me (though the "especially
problematic" remark in his change description suggests that it is
an intentional and necessary change to suit unionmount).  See how
that goes.

Hugh