Re: [PATCH v2] pc: add property for Linux setup_data random number seed

[Michael S. Tsirkin]

On Fri, Aug 05, 2022 at 10:16:37AM +0200, Paolo Bonzini wrote:
> On 8/5/22 09:01, Michael S. Tsirkin wrote:
> > > -    if (!legacy_no_rng_seed) {
> > > +    if (x86ms->linuxboot_randomness != ON_OFF_AUTO_OFF &&
> > > +        (protocol >= 0x209 || x86ms->linuxboot_randomness == 
> > > ON_OFF_AUTO_ON)) {
> > Hmm so if user requested "on" but protocol is too old then we just
> > ignore it silently? I'd rather we failed initialization.
> > So:
> > 
> > if (x86ms->linuxboot_randomness == ON_OFF_AUTO_ON &&
> >      protocol < 0x209) {
> >     fprintf(stderr, "qemu: Linux kernel too old to load a dtb\n");
> >     exit(1);
> > }
> 
> It doesn't ignore the "on" setting; it passes the seed anyway even if the
> protocol is too old.  Basically, a kernel that is too old to support setup
> data is treated the same as a kernel that supports setup data but doesn't
> know about the seed datum.

Oh it's an || not an &&. You are right. I needed more coffee.

> It seemed the more sensible implementation because anyway you cannot know if
> the kernel will use the datum.
> 
> Paolo

OK then.

-- 
MST