xen: how does XEN_PRIVILEGED_GUEST work?

xen: how does XEN_PRIVILEGED_GUEST work?

[Paul Bolle]

0) The Kconfig symbol XEN_PRIVILEGED_GUEST got added to the mainline
Linux repository in v2.6.37. It has never been used. It has always been
a nop.

1) Its Kconfig entry is preceded with this comment:
    # Dummy symbol since people have come to rely on the PRIVILEGED_GUEST
    # name in tools.

What does that mean?

2) On a related note: can anyone please explain commit
7c1bfd685bcdc822ab1d7411ea05c82bd2a7b260 ("xen/pci: Fix compiler error
when CONFIG_XEN_PRIVILEGED_GUEST is not set.") and commit
c00c8aa2d976e9ed1d12a57b42d6e9b27efb7abe ("xen/trace: Fix compile error
when CONFIG_XEN_PRIVILEGED_GUEST is not set"). Do they perhaps fix
problems in code not mainlined?


Paul Bolle

Re: xen: how does XEN_PRIVILEGED_GUEST work?

[Konrad Rzeszutek Wilk]

On Fri, Mar 22, 2013 at 11:37:35PM +0100, Paul Bolle wrote:
> 0) The Kconfig symbol XEN_PRIVILEGED_GUEST got added to the mainline
> Linux repository in v2.6.37. It has never been used. It has always been
> a nop.
> 
> 1) Its Kconfig entry is preceded with this comment:
>     # Dummy symbol since people have come to rely on the PRIVILEGED_GUEST
>     # name in tools.
> 
> What does that mean?

[root@phenom minutes]# cat /etc/grub.d/20_linux_xen |grep PRIV
        if (grep -qx "CONFIG_XEN_DOM0=y" "${config}" 2> /dev/null || grep -qx "CONFIG_XEN_PRIVILEGED_GUEST=y" "${config}" 2> /dev/null); then echo -n "$i " ; fi

> 
> 2) On a related note: can anyone please explain commit
> 7c1bfd685bcdc822ab1d7411ea05c82bd2a7b260 ("xen/pci: Fix compiler error
> when CONFIG_XEN_PRIVILEGED_GUEST is not set.") and commit

Hm, the change actually was for CONFIG_XEN_DOM0. Title is incorrect

> c00c8aa2d976e9ed1d12a57b42d6e9b27efb7abe ("xen/trace: Fix compile error
> when CONFIG_XEN_PRIVILEGED_GUEST is not set"). Do they perhaps fix
> problems in code not mainlined?

No. That one was triggered by some juicy random config that Randy's
randconfig found.

> 
> 
> Paul Bolle
> 

Re: xen: how does XEN_PRIVILEGED_GUEST work?

[Paul Bolle]

On Sat, 2013-03-23 at 09:22 -0400, Konrad Rzeszutek Wilk wrote:
> On Fri, Mar 22, 2013 at 11:37:35PM +0100, Paul Bolle wrote:
> > 1) Its Kconfig entry is preceded with this comment:
> >     # Dummy symbol since people have come to rely on the PRIVILEGED_GUEST
> >     # name in tools.
> > 
> > What does that mean?
> 
> [root@phenom minutes]# cat /etc/grub.d/20_linux_xen |grep PRIV
>         if (grep -qx "CONFIG_XEN_DOM0=y" "${config}" 2> /dev/null || grep -qx "CONFIG_XEN_PRIVILEGED_GUEST=y" "${config}" 2> /dev/null); then echo -n "$i " ; fi

Thanks.

Can userspace require the build system to keep using some Kconfig symbol
by doing tests like these? Anyhow, if that's the only tool then
XEN_PRIVILEGED_GUEST can be dropped. XEN_PRIVILIGED_GUEST should always
be equal to XEN_DOM0 so this if () test will evaluate identically with
the sub-test for CONFIG_XEN_PRIVILIGED_GUEST removed.


Paul Bolle

Re: xen: how does XEN_PRIVILEGED_GUEST work?

[Konrad Rzeszutek Wilk]

On Sat, Mar 23, 2013 at 02:59:35PM +0100, Paul Bolle wrote:
> On Sat, 2013-03-23 at 09:22 -0400, Konrad Rzeszutek Wilk wrote:
> > On Fri, Mar 22, 2013 at 11:37:35PM +0100, Paul Bolle wrote:
> > > 1) Its Kconfig entry is preceded with this comment:
> > >     # Dummy symbol since people have come to rely on the PRIVILEGED_GUEST
> > >     # name in tools.
> > > 
> > > What does that mean?
> > 
> > [root@phenom minutes]# cat /etc/grub.d/20_linux_xen |grep PRIV
> >         if (grep -qx "CONFIG_XEN_DOM0=y" "${config}" 2> /dev/null || grep -qx "CONFIG_XEN_PRIVILEGED_GUEST=y" "${config}" 2> /dev/null); then echo -n "$i " ; fi
> 
> Thanks.
> 
> Can userspace require the build system to keep using some Kconfig symbol
> by doing tests like these? Anyhow, if that's the only tool then

I don't know if there are rules written down for this sort of thing - but
I presume there are some utilities that check the /proc/config.gz to
see if certain options are enabled? This is not that much different from that.

> XEN_PRIVILEGED_GUEST can be dropped. XEN_PRIVILIGED_GUEST should always
> be equal to XEN_DOM0 so this if () test will evaluate identically with
> the sub-test for CONFIG_XEN_PRIVILIGED_GUEST removed.

That is the only tool that I am aware of.

But if we wanted to drop one of those CONFIG_ options I would be more happy
with dropping of the XEN_DOM0. As the initial domain is not that different
from any PV domain - except that it can do ACPI, VGA, and some EFI stuff.
In essence it is a priviliged guest type.

It might make more sense to have the XEN_PRIVILIGED_GUEST be the option
that that would define whether there should be a dependency on the ACPI, VGA, etc
stuff.

Unfortunatly the XEN_DOM0 seems to be so interleaved in the header/source code
that this would take some surgery to get right.

> 
> 
> Paul Bolle
>