Re: [PATCH v2 0/7] srcu: NMI-safe atomic SRCU readers, and their irqsave forms

[Paul E. McKenney]

On Wed, Nov 11, 2026 at 02:33:18PM -0500, Mathieu Desnoyers wrote:
> On 2026-11-11 14:02, David Woodhouse wrote:
> > The latter claim is only true in __srcu_read_lock_fast()/_fast_updown()
> > isn't it? Not in the atomic-SRCU. Unless we do something like this on
> > top?
> 
> It is true for the atomic readers as well.  Without
> CONFIG_NEED_SRCU_NMI_SAFE, include/linux/srcu.h defines
> __srcu_read_lock_nmisafe() as an inline that calls __srcu_read_lock(),
> and the same for unlock.  So x86 keeps this_cpu_inc() and smp_mb().
> Your diff selects the same functions by hand.
> 
> Only the architectures that select NEED_SRCU_NMI_SAFE get the atomic
> increments.

Even better!  ;-)

							Thanx, Paul

> > I set a test running (*without* the above).
> 
> Thanks!
> 
> Mathieu
> 
> -- 
> Mathieu Desnoyers
> EfficiOS Inc.
> https://www.efficios.com