Re: xsm: Consolidate xsm processing within domain control hypercall.

[Alex Williamson <alex.williamson@hp.com>]

On Tue, 2007-12-04 at 15:05 -0700, Alex Williamson wrote:
> On Tue, 2007-12-04 at 16:49 -0500, Mike D. Day wrote:
> > On 04/12/07 13:06 -0700, Alex Williamson wrote:
> > > 
> > >   Does this work right across a PV domain save/restore on x86?  On ia64
> > > I end up with "Domain-Unnamed" after I save a PV domain and another
> > > after I restore it, then do a shutdown.  Reverting this patch restores
> > > correct behavior.  Thanks,
> > 
> > I'll investigate this regression. Do you have any ideas as to why this
> > is happening?
> 
>    Nope, that's why I'm hoping x86 does something similar ;^)  Since I'm
> not using XSM_ENABLE, xsm_call should just be (0), which means the
> changed code should all be noops... but apparently something changes.
> BTW, xsm_call(domctl(domctl)) seems unnecessarily obfuscated.  Thanks,

Looks like the problem is that XEN_DOMCTL_destroydomain now always
returns -ESRCH.  The patch below seems to fix it although it may be just
as correct to hard code ret to 0.  I also took the liberty of returning
the error value xsm_domctl() returns, which seems to match previous
behavior.  Thanks,

	Alex

Signed-off-by: Alex Williamson <alex.williamson@hp.com>
---

diff -r 62451388f630 xen/common/domctl.c
--- a/xen/common/domctl.c	Tue Dec 04 11:52:10 2007 +0000
+++ b/xen/common/domctl.c	Tue Dec 04 17:11:07 2007 -0700
@@ -193,7 +193,8 @@ long do_domctl(XEN_GUEST_HANDLE(xen_domc
 
     spin_lock(&domctl_lock);
 
-    if ( xsm_domctl(op) )
+    ret = xsm_domctl(op);
+    if ( ret != 0 )
         goto domctl_out;
 
     switch ( op->cmd )
@@ -400,7 +401,7 @@ long do_domctl(XEN_GUEST_HANDLE(xen_domc
         ret = -ESRCH;
         if ( d != NULL )
         {
-            domain_kill(d);
+            ret = domain_kill(d);
             rcu_unlock_domain(d);
         }
     }