Messages in this thread | | | Date | Mon, 1 Mar 2010 19:51:53 +0100 | From | Oleg Nesterov <> | Subject | Re: [PATCH 20/43] workqueue: reimplement work flushing using linked works |
| |
On 03/02, Tejun Heo wrote: > > On 03/01/2010 11:53 PM, Oleg Nesterov wrote: > > > > and in this case we put this barrier at the head of ->scheduled list. > > > > This means, this barrier will run after that work W, not before it? > > Yes, the barrier will run after the target work as it should.
You are right. it will run after target work == current_work and before the next pending work W.
Because,
> > Hmm. And what if there are no pending works but ->current_work == target ? > > Again, we add the barrier to ->scheduled, but in this case worker_thread() > > can't even notice ->scheduled is not empty because it only checks ->worklist? > > A worker always checks ->scheduled after a work is finished.
Yes! I missed this, thanks.
> > insert_wq_barrier() also does: > > > > unsigned long *bits = work_data_bits(target); > > ... > > *bits |= WORK_STRUCT_LINKED; > > > > perhaps this needs atomic_long_set(), although I am not sure this really > > matters. > > Yeah, well, work->data access is pretty messed up. At this point, > there's no reason for atomic_long_t to begin with.
grep, grep, grep... arch/sparc/lib/atomic32.c uses spinlocks for atomic_set() and ___set_bit(). Probably that is why atomic_long_set() is really needed to avoid the race with test_and_set_bit(PENDING).
Oleg.
| |