RE: [PATCH rfcv3 07/11] qemu: add hard reboot in QEMU driver

-----Original Message----- From: Daniel P. Berrangé <berrange(a)redhat.com&gt; Subject: Re: [PATCH rfcv3 07/11] qemu: add hard reboot in QEMU driver On Fri, Jan 12, 2024 at 09:20:19AM +0000, Duan, Zhenzhong wrote: > > > >-----Original Message----- > >From: Daniel P. Berrangé <berrange(a)redhat.com&gt; > >Subject: Re: [PATCH rfcv3 07/11] qemu: add hard reboot in QEMU driver > > > >On Mon, Nov 27, 2023 at 04:55:17PM +0800, Zhenzhong Duan wrote: > >> From: Chenyi Qiang <chenyi.qiang(a)intel.com&gt; > >> > >> Add the new flag > >VIR_DOMAIN_REBOOT_HARD/VIR_DOMAIN_SHUTDOWN_HARD to > >> carry out a hard reboot, which kills the QEMU process and creates a new > >> one with the same definition. > > > >AFAICT, the _HARD flags cause it to issue a QEMU monitor > >command todo an ACPI shutdown and then re-create QEMU. > > > >IOW, it seems like this is just the same as the existing > >_ACPI flags, and so I'm not sure why we need any new > >functionality at all. > > > >What is the existing ACPI shutdown & fake reboot not > >able to achieve ? > > ACPI shutdown & fake reboot send below QMP command in sequence: > > "system_powerdown", "system_reset", "cont". > > "system_reset" QMP command isn't supported by TDX for security reasons. > So we have to kill old QEMU and re-create new one. We are still doing the 'system_powerdown' step first though, to do a graceful shutdown, before re-creating QEMU, and the powerdown is what we refer to as "ACPI". So I think we don't need new public API constants. We should still use the _ACPI flag, and just "do the right thing" with choosing between system_reset vs re-creating QEMU depending on whether TDX/SEV are active.

That way the public API user experience is unchanged, only the internal impl varies.