On Tue, Aug 22, 2017 at 06:22:36PM +0900, Byungchul Park wrote:
> On Tue, Aug 22, 2017 at 11:06:03AM +0200, Peter Zijlstra wrote:
> > So I did the below little hack, which basically wipes the entire lock
> > history when we start a work and thereby disregards/looses the
> > dependency on the work 'lock'.
> >
> > It makes my test box able to boot and build a kernel on XFS, so while I
> > see what you're saying (I think), it doesn't appear to instantly show.
> >
> > Should I run xfstests or something to further verify things are OK? Does
> > that need a scratch partition (I keep forgetting how to run that stuff
> > :/).
> >
> > ---
> > diff --git a/kernel/locking/lockdep.c b/kernel/locking/lockdep.c
> > index 66011c9f5df3..de91cdce9460 100644
> > --- a/kernel/locking/lockdep.c
> > +++ b/kernel/locking/lockdep.c
> > @@ -4756,10 +4756,14 @@ void crossrelease_hist_start(enum xhlock_context_t
> > c)
> > {
> > struct task_struct *cur = current;
> >
> > - if (cur->xhlocks) {
> > - cur->xhlock_idx_hist[c] = cur->xhlock_idx;
> > - cur->hist_id_save[c] = cur->hist_id;
> > - }
> > + if (!cur->xhlocks)
> > + return;
> > +
> > + if (c == XHLOCK_PROC)
> > + invalidate_xhlock(&xhlock(cur->xhlock_idx));
>
> We have to detect dependecies if it exists, even in the following case:
>
> oooooooiiiiiiiiiiiiiiiiiii.........
> |<- range for commit ->|
>
> where
> o: acquisition outside of each work,
> i: acquisition inside of each work,
>
> With yours, we can never detect dependecies wrt 'o'.
There really shouldn't be any o's when you call
crossrelease_hist_start(XHLOCK_PROC), it should denote the bottom of a
context, see:
https://lkml.kernel.org/r/20170301104328.gd6...@twins.programming.kicks-ass.net
And in that respect you placed the calls wrongly in process_one_work(),
except it now works out nicely to disregard these fake locks.
