On 30/05/17 10:56, Auger Eric wrote: > Hi, > > On 03/05/2017 12:46, Marc Zyngier wrote: >> Now that we're able to safely handle common sysreg access, let's >> give the user the opportunity to enable it by passing a specific >> command-line option (vgic_v3.common_trap). >> >> Signed-off-by: Marc Zyngier <marc.zyngier@xxxxxxx> > > What is the rationale behind trapping those common sysreg accesses. This > is not enabled by default for some erratas. Is it to complete the > trapping modality. May be worth to explain. That's strictly for completeness. Let's assume some hypothetical HW has a bug on the EOI path, where the physical interrupt wouldn't get properly deactivated on the physical distributor. In order to deal with that when EOImode==1 in the guest, you'd need to trap DIR as well. Same thing if PMR was broken as well. Also, we could implement some form of tracing of all the GICv3 accessors if we wanted such a functionality (trying to find something more gratifying than basic errata workaround here... ;-). > > Besides > > Reviewed-by: Eric Auger <eric.auger@xxxxxxxxxx> Thanks, M. -- Jazz is not dead. It just smells funny...
