On 28.03.14 at 12:29,
wrote: On Fri, 28 Mar 2014, Jan Beulich wrote: On 28.03.14 at 11:54,
wrote: Le 28/03/2014 11:41, Alexander Graf a écrit : Am 21.03.2014 um 23:56 schrieb Guillaume Gardet
: Hi,
please find in attachment an update for the pending patch which disable Xen
on ARM since it is currently broken (compilation fails).
Is it broken upstream or is it broken due to our xen patches?
openSUSE Xen patches brake it, as mentionned here: http://lists.opensuse.org/opensuse-kernel/2014-03/msg00023.html
Now see - in that mail you point to an upstream patch as the presumed cause, plus XEN_HAVE_PV_GUEST_ENTRY is _intended_ to be undefined for ARM, which made me ignore it. Would it have been made clear that it's in our patches (with a more explicit description as to what the problem really is), I would have gone and tried to fix this there.
Looking at this in closer detail it's not even clear how to fix this though, since without a struct start_info instance there's nowhere to check for SIF_INITDOMAIN. Will need to consult xen-devel first...
ARM doesn't have a start_info. Linux manually populates a dummy start_info struct in arch/arm/xen/enlighten.c:xen_guest_init. SIF_INITDOMAIN and SIF_PRIVILEGED are set if xen_feature(XENFEAT_dom0).
But you realize that more than just that code (namely your fake start_info instance) wouldn't build anymore with the definition of struct start_info not visible. Anyway, until upstream is going to sync their headers and get this fixed, I'll have to re-expose the structure to ARM. I wish the hiding would have been done _after_ having switched upstream Linux away from using it... Jan -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org To contact the owner, e-mail: opensuse-kernel+owner@opensuse.org