Re: [PATCH v3 06/17] s390x: protvirt: Add migration blocker

[Cornelia Huck]

On Thu, 20 Feb 2020 12:24:23 +0100
Janosch Frank <address@hidden> wrote:

> On 2/20/20 11:48 AM, Cornelia Huck wrote:
> > On Fri, 14 Feb 2020 10:16:25 -0500
> > Janosch Frank <address@hidden> wrote:
> >   
> >> Migration is not yet supported.
> >>
> >> Signed-off-by: Janosch Frank <address@hidden>
> >> ---
> >>  hw/s390x/s390-virtio-ccw.c | 16 ++++++++++++++++
> >>  1 file changed, 16 insertions(+)
> >>
> >> diff --git a/hw/s390x/s390-virtio-ccw.c b/hw/s390x/s390-virtio-ccw.c
> >> index 5fa4372083..d64724af91 100644
> >> --- a/hw/s390x/s390-virtio-ccw.c
> >> +++ b/hw/s390x/s390-virtio-ccw.c
> >> @@ -42,6 +42,9 @@
> >>  #include "hw/s390x/tod.h"
> >>  #include "sysemu/sysemu.h"
> >>  #include "hw/s390x/pv.h"
> >> +#include "migration/blocker.h"
> >> +
> >> +static Error *pv_mig_blocker;
> >>  
> >>  S390CPU *s390_cpu_addr2state(uint16_t cpu_addr)
> >>  {
> >> @@ -373,6 +376,7 @@ static void s390_machine_reset(MachineState *machine)
> >>      CPUState *cs, *t;
> >>      S390CPU *cpu;
> >>      S390CcwMachineState *ms = S390_CCW_MACHINE(machine);
> >> +    static Error *local_err;
> >>  
> >>      /* get the reset parameters, reset them once done */
> >>      s390_ipl_get_reset_request(&cs, &reset_type);
> >> @@ -422,6 +426,17 @@ static void s390_machine_reset(MachineState *machine)
> >>          }
> >>          run_on_cpu(cs, s390_do_cpu_reset, RUN_ON_CPU_NULL);
> >>  
> >> +        if (!pv_mig_blocker) {
> >> +            error_setg(&pv_mig_blocker,
> >> +                       "protected VMs are currently not migrateable.");
> >> +        }
> >> +        migrate_add_blocker(pv_mig_blocker, &local_err);  
> > 
> > If I'm not lost in the context, that's during PV_RESET. I'm a bit
> > confused why you'd add the blocker here?  
> 
> Where would you want me to add it?
> It's here where we switch into secure mode and I need to block before
> switching and unblock if it fails.
> 
> When having the blocker in diag.c, I'd have a hard time unblocking on a
> PV switch fail.
> 
> >   
> >> +        if (local_err) {
> >> +            error_report_err(local_err);
> >> +            error_free(pv_mig_blocker);
> >> +            exit(1);  
> > 
> > Why the exit()? Can't you fail the call?  
> 
> Well, if that fails and we go protected, I wouldn't be protected agains
> migrations, right?

No, I meant not go protected, if that's possible.

> 
> >   
> >> +        }
> >> +
> >>          if (s390_machine_pv_secure(ms)) {

Ok, I think what confuses me is this call: it reads as if you actually
tear down things if the machine is secure. Call it
s390_machine_pv_make_secure() to make sure it is actively doing
something and not checking a previously set value?

> >>              CPU_FOREACH(t) {
> >>                  s390_pv_vcpu_destroy(t);
> >> @@ -430,6 +445,7 @@ static void s390_machine_reset(MachineState *machine)
> >>              ms->pv = false;
> >>  
> >>              s390_machine_inject_pv_error(cs);
> >> +            migrate_del_blocker(pv_mig_blocker);
> >>              s390_cpu_set_state(S390_CPU_STATE_OPERATING, cpu);
> >>              return;
> >>          }  
> > 
> >   
> 
>

pgpmby5MV6kaK.pgp
Description: OpenPGP digital signature