On 19.10.2023 18:21, David Woodhouse wrote: > --- a/xen/drivers/char/ns16550.c > +++ b/xen/drivers/char/ns16550.c > @@ -586,6 +586,8 @@ static void __init cf_check ns16550_endboot(struct > serial_port *port) > > if ( uart->remapped_io_base ) > return; > + if (!hardware_domain) > + return; > rv = ioports_deny_access(hardware_domain, uart->io_base, uart->io_base + > 7); > if ( rv != 0 ) > BUG();
Considering this is the only use of hardware_domain in the driver, fixing it like this (with style adjusted) is certainly an option. I'd like to ask though whether it makes sense for execution to make it here in shim mode, where there's no serial port anyway (unless we start supporting pass-through [of possibly even non-PCI devices] for the shim). I wonder whether we wouldn't want to bypass ns16550_init() - likely as well as the EHCI and XHCI ones - instead, until (if ever) such passing through was actually deemed necessary to support. Jan
