Re: [Xen-devel] live migration fails (assert in shadow_hash_delete)

[Tim Deegan <Tim.Deegan@xxxxxxxxxx>]

At 10:27 +0000 on 26 Feb (1267180053), Jan Beulich wrote:
> >>> Tim Deegan <Tim.Deegan@xxxxxxxxxx> 26.02.10 10:52 >>>
> >The bug seems to have come in between 4.0.0 rc1 (20789) and 20822.
> >Bisecting between those is more fun because PV domain creation and migration
> >were broken in libxc then.  Reverting 20808 (the only cset there that
> >touches the shadow code) doesn't fix the problem.
> >
> >Selective backporting yesterday seemed to blame 20812 ("xend: NUMA: fix
> >division by zero on unpopulated nodes"), which seems unlikely.  I'll dig
> >further.
> 
> I'd think 20792 is a good candidate - a copy-and-paste mistake would
> cause the page subsequent to the one allocated to be overwritten.
> Will send a patch in a minute.

Thanks for that.

Keir, I'm still seeing (different) crashes on unstable tip even with
Jan's fix; the proximate cause is c/s 20954, which changes the paths
taken when log-dirty mode is turned off after the live migration.

Reverting c/s 20954 fixes migration for me and is probably the best
thing to get the 4.0 release schedule going again.   I'll try to find
the actual bug at some later date. 

Cheers,

Tim.

-- 
Tim Deegan <Tim.Deegan@xxxxxxxxxx>
Principal Software Engineer, XenServer Engineering
Citrix Systems UK Ltd.  (Company #02937203, SL9 0BG)

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