Ingo Molnar wrote:
* Jeremy Fitzhardinge <jeremy@xxxxxxxx> wrote:
Maybe should this force-paravirt mode be a kernel debugging option?
No, just a plain CONFIG_PARAVIRT to turn it on and off independently
of everything else.
it should be a debugging option in the way that it's dependent on
KERNEL_DEBUG and KERNEL_EXPERIMENTAL.
EXPERIMENTAL certainly, but not DEBUG. Sure, CONFIG_PARAVIRT is
somewhat useless if you don't also configure Xen/lguest/etc, but running
a paravirt system on bare hardware will probably be the most common way
a paravirt kernel gets run (a distro will always enable it and compile
in whatever hypervisor support they want, but most people will end up
just running it on bare hardware). Therefore, we should try to be as
sure as possible that PARAVIRT works well (ie, indistinguishable from
non-PARAVIRT) on bare hardware.
i dont think you shold worry about this - i think 64-bit guest support
for specific hypervisors could be added even after the merge window, if
it's reasonably isolated (which i'd expect it to be). It clearly cannot
break anything else so it's the same category as new driver or new
hardware support - more relaxed integration rules. Then we can remove
the KERNEL_DEBUG and KERNEL_EXPERIMENTAL dependency as well.
Good, I'm glad you feel that way.
J
_______________________________________________
Virtualization mailing list
Virtualization@xxxxxxxxxxxxxxxxxxxxxxxxxx
https://lists.linux-foundation.org/mailman/listinfo/virtualization
