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

[Daniel P . Berrangé]

On Thu, Aug 04, 2022 at 03:20:59PM +0200, Jason A. Donenfeld wrote:
> On Thu, Aug 04, 2022 at 03:13:20PM +0200, Paolo Bonzini wrote:
> > Using a property makes it possible to use the normal compat property
> > mechanism instead of ad hoc code; it avoids parameter proliferation
> > in x86_load_linux; and allows shipping the code even if it is
> > disabled by default.
> 
> Strong NACK from me here.
> 
> If this kind of thing is off by default, it's as good as useless. Indeed
> it shouldn't even be a knob at all. Don't do this.

You're misunderstanding the patch. This remains on by default for
 the 7.1 machine type.

The patch is merely exposing a knob so that users can override the
built-in default if they need to. Imagine if we had shipped this
existing code before today's bugs were discovered.  The knob
proposed her would allow users to turn off the broken pieces.
This is a good thing.

> Rather, let's fix the bug. The code as-is -- going back to the 2016 DTB
> addition -- is problematic and needs to be fixed. So let's fix that.
> Trying to cover up the problem with a default-off knob just ensures this
> stuff will never be made to work right.

It isn't covering up the problem, just providing a workaround
option, should another bug be discovered after release. We
still need to fix current discussed problems of course.


With regards,
Daniel
-- 
|: https://berrange.com      -o-    https://www.flickr.com/photos/dberrange :|
|: https://libvirt.org         -o-            https://fstop138.berrange.com :|
|: https://entangle-photo.org    -o-    https://www.instagram.com/dberrange :|