On Sat, Aug 15, 2020 at 07:18:39AM -0700, Paul E. McKenney wrote: > On Sat, Aug 15, 2020 at 10:42:50AM +0200, Peter Zijlstra wrote: > > On Sat, Aug 15, 2020 at 01:14:53AM +0200, Thomas Gleixner wrote: > > > > > #1 trivial fix is to force switching to an high prio thread or a soft > > > interrupt which does the allocation > > > > Yeah, push the alocation out to another context. I did consider it, but > > why bother? > > > > Also, raising a softirq can't be done from every context, that's a whole > > new problem. You can do irq_work I suppose, but not all architectures > > support the self-IPI yet. > > > > All in all, it's just more complexity than the fairly trivial > > __alloc_page_lockless(). > > > > Whichever way around, we can't rely on the allocation. > > One way to enforce that would be to put something like this at the > beginning of the __alloc_page_lockless() function: > > if (IS_ENABLED(CONFIG_PROVE_LOCKING) && (prandom_u32() & 0xffff)) > return NULL; Right, too early in the morning. :-/ This "slight" variation might include a bit of usefulness along with the convincing: if (IS_ENABLED(CONFIG_PROVE_LOCKING) && !(prandom_u32() & 0xff)) return NULL; Plus failing one out of 256 times is likely a better choice than once out of 65536 times, especially for the occasional caller of this function. Thanx, Paul > I am sure that there is a better choice than CONFIG_PROVE_LOCKING. > But whatever the choice, there is nothing quite like the occasional > allocation failure during testing to convince people that such failure > really can happen. > > Thanx, Paul
