Re: [Qemu-devel] Re: [PATCH] Reboot CPU on triple fault

[Avi Kivity]

Jan Kiszka wrote:
> Kevin Wolf wrote:
>   
> > Jan Kiszka schrieb:
> >     
> > > As the same requirement came up here, I worked out the following patch.
> > > I feel a bit uneasy about it because
> > >
> > >  o I'm unsure if breaking out of the exception loop is OK this way.
> > >
> > >  o including necessary headers fails, mostly due to stdio redefinitions
> > >    in dyngen-exec.h.
> > >
> > > This patch behaves as it should, but only in one specific test case.
> > > Feedback is welcome.
> > >       
> > This is exactly the behaviour I'd like to have. I still needed a small
> > change to your patch, though: You shouldn't return 0 from
> > check_exception, after all it's not a divide error. I'm not sure if
> > EXCP_HLT is the right thing to return here, but the attached patch works
> > for me. Without this change it kept hanging in a loop forever writing
> > more dumps than I ever wanted. ;-)
> >     
>
> Your version still works for me (and makes more sense than return 0,
> indeed). So we just need someone to confirm that this is actually a
> clean way to leave the fault handling.
>
> And we need a consensus regarding fault reporting... What about
> confining the dump to the logfile, leaving the console alone?
>   

If logging is really necessary, have it disabled by default and add a 
switch to enable it.  This way ordinary (non-developers) don't need to 
worry about it, and developers can conveniently see it on stdout.

--
error compiling committee.c: too many arguments to function