On Tue, Feb 27, 2024 at 1:08 PM Andrew Cooper <[email protected]> wrote: > > On 27/02/2024 11:25 am, Roger Pau Monne wrote: > > diff --git a/xen/common/virtual_region.c b/xen/common/virtual_region.c > > index ddac5c9147e5..e3a4dc8540df 100644 > > --- a/xen/common/virtual_region.c > > +++ b/xen/common/virtual_region.c > > <snip> > > > > +#ifdef CONFIG_LIVEPATCH > > void unregister_virtual_region(struct virtual_region *r) > > { > > - /* Expected to be called from Live Patch - which has IRQs disabled. */ > > - ASSERT(!local_irq_is_enabled()); > > - > > remove_virtual_region(r); > > + > > + /* Assert that no CPU might be using the removed region. */ > > + rcu_barrier(); > > } > > > > -#if defined(CONFIG_LIVEPATCH) && defined(CONFIG_X86) > > +#ifdef CONFIG_X86 > > void relax_virtual_region_perms(void) > > { > > const struct virtual_region *region; > > @@ -117,6 +106,7 @@ void tighten_virtual_region_perms(void) > > rcu_read_unlock(&rcu_virtual_region_lock); > > } > > #endif > > +#endif > > Can we please have: > > #endif /* CONFIG_X86 */ > #endif /* CONFIG_LIVEPATCH */ > > before this gets any more complicated to follow. Can be fixed on commit. >
LGTM with this suggested change. Reviewed-by: Ross Lagerwall <[email protected]>
