Xen devel - Jan 2011 - what could cause crash on restore from save

GPLPV is crashing very shortly after restoring from a save. The bug
check code is 0x7F (0xd, 0, 0, 0) which is described as "some other
exception". This doesn''t seem to happen if I omit the actual
suspend
hypercall, so I''m wondering if I''m not accounting for
something
somewhere.

Any suggestions?

Thanks

James

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Does your VM have multiple vCPUs? Did you corral them into some known code to
make sure everything was in a known state? How ''shortly'' after
return from suspend are things dying?

  Paul
> -----Original Message-----
> From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel-
> bounces@lists.xensource.com] On Behalf Of James Harper
> Sent: 23 January 2011 21:47
> To: xen-devel@lists.xensource.com
> Subject: [Xen-devel] what could cause crash on restore from save
> 
> GPLPV is crashing very shortly after restoring from a save. The bug
> check code is 0x7F (0xd, 0, 0, 0) which is described as "some other
> exception". This doesn''t seem to happen if I omit the actual
suspend
> hypercall, so I''m wondering if I''m not accounting for
something
> somewhere.
> 
> Any suggestions?
> 
> Thanks
> 
> James
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

> 
> Does your VM have multiple vCPUs?
Just the one in my minimal test case.
> Did you corral them into some known code to
> make sure everything was in a known state?
For MP, yes. CPU0 waits until all others are spinning at HIGH_LEVEL
before making the suspend hypercall. This all used to work just fine but
it''s been a while since I tested it.
> How ''shortly'' after return from suspend are things dying?
Pretty much as soon as usermode code gets a chance to execute. Sometimes
vbd is part way through initialisation but most of the time its well
before that.

I''ve minimised things into leaving all my code in a stalled state (eg
set things up but turn off the xen interrupt so no requests are ever
answered) and it still hangs so I''m pretty sure it''s not
something I''m
doing on resume, so either it''s something I''m doing on
suspend, or it''s
something xen is doing to my memory.

I''ve also tried doing everything _except_ making the hypercall and that
all works fine (eg xenbus gets shut down and then resumed etc).

The windows debugger doesn''t give me anything particularly useful, and
at least one thread ends up executing where no known code is, so I''m
wondering if I''m doing something that is preventing memory from being
restored correctly. I''m trying to build some tests for that now.

James
>   Paul
> 
> > -----Original Message-----
> > From: xen-devel-bounces@lists.xensource.com [mailto:xen-devel-
> > bounces@lists.xensource.com] On Behalf Of James Harper
> > Sent: 23 January 2011 21:47
> > To: xen-devel@lists.xensource.com
> > Subject: [Xen-devel] what could cause crash on restore from save
> >
> > GPLPV is crashing very shortly after restoring from a save. The bug
> > check code is 0x7F (0xd, 0, 0, 0) which is described as "some
other
> > exception". This doesn''t seem to happen if I omit the
actual suspend
> > hypercall, so I''m wondering if I''m not accounting
for something
> > somewhere.
> >
> > Any suggestions?
> >
> > Thanks
> >
> > James
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xensource.com
> > http://lists.xensource.com/xen-devel
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

> 
> I''ve also tried doing everything _except_ making the hypercall and
that
> all works fine (eg xenbus gets shut down and then resumed etc).
> 
Additionally, ''xm save -c'' also seems to work fine.

James


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel