Possible nohz-full/RCU issue in arm64 KVM

[Paul E. McKenney]

On Fri, Dec 17, 2021 at 05:20:21PM +0100, Nicolas Saenz Julienne wrote:
> Hi Paul,
> 
> On Fri, 2021-12-17 at 08:07 -0800, Paul E. McKenney wrote:
> > On Fri, Dec 17, 2021 at 04:54:22PM +0100, Paolo Bonzini wrote:
> > > On 12/17/21 15:38, Mark Rutland wrote:
> > > > For example kvm_guest_enter_irqoff() calls guest_enter_irq_off() which calls
> > > > vtime_account_guest_enter(), but kvm_guest_exit_irqoff() doesn't call
> > > > guest_exit_irq_off() and the call to vtime_account_guest_exit() is open-coded
> > > > elsewhere. Also, guest_enter_irq_off() conditionally calls
> > > > rcu_virt_note_context_switch(), but I can't immediately spot anything on the
> > > > exit side that corresponded with that, which looks suspicious.
> > > 
> > > rcu_note_context_switch() is a point-in-time notification; it's not strictly
> > > necessary, but it may improve performance a bit by avoiding unnecessary IPIs
> > > from the RCU subsystem.
> > > 
> > > There's no benefit from doing it when you're back from the guest, because at
> > > that point the CPU is just running normal kernel code.
> > 
> > Do scheduling-clock interrupts from guest mode have the "user" parameter
> > set?  If so, that would keep RCU happy.
> 
> Are you referring to the user_mode() check in irqentry_enter()? If so I don't
> think it'll help, arm64 doesn't use that function. It directly calls
> enter_from_{user,kernel}_mode() through its custom entry/exit routines.

I am talking about rcu_sched_clock_irq(), which is called from
update_process_times(), which is called from various places depending
on .config.  These call sites pass in either user_mode(regs) or
user_mode(get_irq_regs()).

Huh.  Maybe I should be looking into using user_mode(get_irq_regs())
in other places within RCU.  Except that I bet the possibility of RCU
being invoked from NMI handlers makes this a bit tricky.

							Thanx, Paul