On Mon, 2018-07-30 at 18:26 +0200, Peter Zijlstra wrote: > > So for ARCH_NO_ACTIVE_MM we never touch ->active_mm and therefore > ->active_mm == ->mm.
Close, but not true for kernel threads, which have a NULL ->mm, but a non-null ->active_mm that gets passed to enter_lazy_tlb(). I stuck to the structure of your code, but ended up removing all the ifdefs because the final mmdrop requires that we actually keep track of the ->active_mm across potentially several kernel->kernel context switches. Ifdefs around the reference counting code are also not needed, because grab_lazy_mm and drop_lazy_mm already contain the equivalent of an ifdef themselves. By morning I should know what the test results look like. I expect they will be identical to my patches, since the refcounting is disabled completely anyway :) > > + /* > > + * kernel -> kernel lazy + transfer active > > + * user -> kernel lazy + mmgrab() active > > + * > > + * kernel -> user switch + mmdrop() active > > + * user -> user switch > > + */ > > + if (!next->mm) { // to > > kernel > > + enter_lazy_tlb(prev->active_mm, next); > > + > > #ifndef ARCH_NO_ACTIVE_MM > > + next->active_mm = prev->active_mm; > > + if (prev->mm) // from > > user > > + mmgrab(prev->active_mm); > > else > prev->active_mm = NULL; > > +#endif > > + } else { // to user > > + switch_mm_irqs_off(prev->active_mm, next->mm, > > next); > > + > > #ifndef ARCH_NO_ACTIVE_MM > > + if (!prev->mm) { // from > > kernel > > + /* will mmdrop() in finish_task_switch(). > > */ > > + rq->prev_mm = prev->active_mm; > > + prev->active_mm = NULL; > > + } > > +#endif > > -- All Rights Reversed.
signature.asc
Description: This is a digitally signed message part