Re: [PATCH v4 1/2] tpm: convert tpmdev options processing to new visitor format

[James Bottomley]

On Fri, 2022-12-30 at 12:01 -0500, Stefan Berger wrote:
> On 12/30/22 10:24, James Bottomley wrote:
[...]
> > @@ -2906,9 +2893,7 @@ void qemu_init(int argc, char **argv)
> >                   break;
> >   #ifdef CONFIG_TPM
> >               case QEMU_OPTION_tpmdev:
> > -                if (tpm_config_parse(qemu_find_opts("tpmdev"),
> > optarg) < 0) {
> > -                    exit(1);
> > -                }
> > +                tpm_config_parse(optarg);
> 
> The patches don't apply to upstream's master.

I think it depends how you apply them.  If you use git, they do except
a minor merge conflict in tpm_passthrough.c

More seriously there's now a compile failure in tpm_mssim.c because of
the lost has_X for X pointer options, but it's also easily fixable.

> This used to exit() on failure but doesn't do this anymore, though it
> probably should.

Actually it still does.  I converted it to the standard &error_fatal
way of doing this, which will cause an exit(1) if we get an error.  The
error_fatal construct seems to have been done precisely to cure this
type of return value threading.

James