On Mon, Jan 11, 2021 at 01:04:24PM +0100, Peter Zijlstra wrote: > > +static DEFINE_PER_CPU(struct irq_work, late_wakeup_work) = > > + IRQ_WORK_INIT(late_wakeup_func); > > + > > /** > > * rcu_user_enter - inform RCU that we are resuming userspace. > > * > > @@ -692,9 +704,17 @@ noinstr void rcu_user_enter(void) > > struct rcu_data *rdp = this_cpu_ptr(&rcu_data); > > > > lockdep_assert_irqs_disabled(); > > - do_nocb_deferred_wakeup(rdp); > > + /* > > + * We may be past the last rescheduling opportunity in the entry code. > > + * Trigger a self IPI that will fire and reschedule once we resume to > > + * user/guest mode. > > + */ > > + if (do_nocb_deferred_wakeup(rdp) && need_resched()) > > + irq_work_queue(this_cpu_ptr(&late_wakeup_work)); > > + > > rcu_eqs_enter(true); > > } > > Do we have the guarantee that every architecture that supports NOHZ_FULL > has arch_irq_work_raise() on? Yes it's a requirement for NOHZ_FULL to work. But you make me realize this is tacit and isn't constrained anywhere in the code. I'm going to add HAVE_IRQ_WORK_RAISE and replace the weak definition with a config based. > > Also, can't you do the same thing you did earlier and do that wakeup > thing before we complete exit_to_user_mode_prepare() ? I do it for CONFIG_GENERIC_ENTRY but the other architectures have their own exit to user loop that I would need to audit and make sure that interrupts aren't ever re-enabled before resuming to user and there is no possible rescheduling point. I could manage to handle arm and arm64 but the others scare me: $ git grep HAVE_CONTEXT_TRACKING arch/csky/Kconfig: select HAVE_CONTEXT_TRACKING arch/mips/Kconfig: select HAVE_CONTEXT_TRACKING arch/powerpc/Kconfig: select HAVE_CONTEXT_TRACKING if PPC64 arch/riscv/Kconfig: select HAVE_CONTEXT_TRACKING arch/sparc/Kconfig: select HAVE_CONTEXT_TRACKING :-s
