Re: [Qemu-devel] [PATCH v2 1/2] vl: don't use RUN_STATE_PRECONFIG as initial state

[Eduardo Habkost]

On Tue, Jun 05, 2018 at 04:25:17PM +0200, Igor Mammedov wrote:
> On Tue, 5 Jun 2018 09:01:09 -0300
> Eduardo Habkost <address@hidden> wrote:
> 
> > On Tue, Jun 05, 2018 at 10:37:55AM +0200, Igor Mammedov wrote:
> > > On Mon, 4 Jun 2018 21:56:47 -0300
> > > Eduardo Habkost <address@hidden> wrote:  
> > > > On Mon, Jun 04, 2018 at 04:21:47PM +0200, Michal Privoznik wrote:
> > > > [...]  
> > > > > > @@ -3572,7 +3570,12 @@ int main(int argc, char **argv, char **envp)
> > > > > >                  }
> > > > > >                  break;
> > > > > >              case QEMU_OPTION_preconfig:
> > > > > > -                preconfig_exit_requested = false;
> > > > > > +                if (!runstate_check(RUN_STATE_NONE)) {
> > > > > > +                    error_report("'--preconfig' and '--incoming' 
> > > > > > options are "
> > > > > > +                                 "mutually exclusive");
> > > > > > +                    exit(EXIT_FAILURE);
> > > > > > +                }
> > > > > > +                runstate_set(RUN_STATE_PRECONFIG);    
> > > > > 
> > > > > Specifying --preconfig twice on the command line now fails with a very
> > > > > cryptic message (there's no --incoming).
> > > > >     
> > > > > >                  break;
> > > > > >              case QEMU_OPTION_enable_kvm:
> > > > > >                  olist = qemu_find_opts("machine");
> > > > > > @@ -3768,9 +3771,12 @@ int main(int argc, char **argv, char **envp)
> > > > > >                  }
> > > > > >                  break;
> > > > > >              case QEMU_OPTION_incoming:
> > > > > > -                if (!incoming) {
> > > > > > -                    runstate_set(RUN_STATE_INMIGRATE);
> > > > > > +                if (!runstate_check(RUN_STATE_NONE)) {
> > > > > > +                    error_report("'--preconfig' and '--incoming' 
> > > > > > options are "
> > > > > > +                                 "mutually exclusive");
> > > > > > +                    exit(EXIT_FAILURE);
> > > > > >                  }
> > > > > > +                runstate_set(RUN_STATE_INMIGRATE);    
> > > > > 
> > > > > Same here. Specifying --incoming twice fails with cryptic message. But
> > > > > one can argue that specifying --incoming twice is wrong anyway.
> > > > >     
> > > > 
> > > > Initially I was going to suggest simply not changing runstate
> > > > during option parsing to avoid this kind of problem, but maybe
> > > > this would be a nice way to implement the command-line parsing
> > > > rules:  
> > > Is there a big reason to try making early incoming transition hack nicer?
> > > I'd rather leave it as it is for now and fix it properly later
> > > (i.e. postponing the transition after machine_done point, which is on my
> > > todo list).  
> > 
> > Yeah, I'm not sure we want go towards encoding more knowledge in
> > the state machine, or keeping the system simple and encoding the
> > rules in more straightforward code+variables inside main().
> I've tried before to use code+variables for preconfig checks without
> introducing a new runstate and it turned out difficult to manage
> and the later changes often broke it. With new runstate end result
> was much cleaner.

Right.  I wasn't talking specifically about the new runstates
like PRECONFIG and NONE (which both make sense, IMO), but about
my suggestion below to also use runstate transition rules to
encode command-line validation and error messages.


> 
> > 
> > > > 
> > > >     case QEMU_OPTION_preconfig:
> > > >         /*
> > > >          * A INCOMING -> PRECONFIG transition would call:
> > > >          *   error_setg("--preconfig and --incoming options are 
> > > > mutually exclusive");
> > > >          */
> > > >         try_runstate_set(RUN_STATE_PRECONFIG, &error_fatal);
> > > >         break;
> > > >     case QEMU_OPTION_incoming:
> > > >         /*
> > > >          * A PRECONFIG -> INCOMING transition would also call:
> > > >          *   error_setg("--preconfig and --incoming options are 
> > > > mutually exclusive");
> > > >          *
> > > >          * Maybe a INCOMING -> INCOMING transition could
> > > >          * result in:
> > > >          *   error_setg("--incoming can't be specified twice");
> > > >          */
> > > >         try_runstate_set(RUN_STATE_INMIGRATE, &error_fatal);
> > > >         break;
> > > > 
> > > >   
> > > > > >                  incoming = optarg;
> > > > > >                  break;
> > > > > >              case QEMU_OPTION_only_migratable:    
> > > > > 
> > > > > Michal
> > > > >     
> > > >   
> > >   
> > 
> 

-- 
Eduardo