On Thu, Jul 13, 2017 at 7:29 PM, Peter Zijlstra <peterz@xxxxxxxxxxxxx> wrote: > On Thu, Jul 13, 2017 at 07:09:53PM +0900, Byungchul Park wrote: >> On Thu, Jul 13, 2017 at 11:50:52AM +0200, Peter Zijlstra wrote: >> > wait_for_completion(&C); >> > atomic_inc_return(); >> > >> > mutex_lock(A1); >> > mutex_unlock(A1); >> > >> > >> > <IRQ> >> > spin_lock(B1); >> > spin_unlock(B1); >> > >> > ... >> > >> > spin_lock(B64); >> > spin_unlock(B64); >> > </IRQ> >> > >> > >> > mutex_lock(A2); >> > mutex_unlock(A2); >> > >> > complete(&C); >> > >> > >> > That gives: >> > >> > xhist[ 0] = A1 >> >> We have to rollback here later on irq_exit. >> >> The followings are ones for irq context. >> >> > xhist[ 1] = B1 >> > ... >> > xhist[63] = B63 >> > >> > then we wrap and have: >> > >> > xhist[0] = B64 >> > >> > then we rewind to 1 and invalidate to arrive at: >> > >> >> Now, whether xhist[0] has been overwritten or not is important. If yes, >> xhist[0] should be NULL, _not_ xhist[1], which is one for irq context so >> not interest at all. >> >> > xhist[ 0] = B64 >> > xhist[ 1] = NULL <-- idx >> >> Therefore, it should be, >> >> xhist[ 0] = NULL <- invalidate, cannot use it any more >> --- <- on returning back from irq context, start from here >> xhist[ 1] = B1 <-- obsolete history of irq > > Ah, so you rely on the same_context_xhlock() ? That doesn't work for > hist (formerly work). As I mentioned in cover-letter, I applied the rollback mechanism into work (of workqueue) as well. So it works even for hist. -- Thanks, Byungchul -- To unsubscribe, send a message with 'unsubscribe linux-mm' in the body to majordomo@xxxxxxxxx. For more info on Linux MM, see: http://www.linux-mm.org/ . Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
