Xen devel - Jul 2008 - Re: [Regression] Changeset #17532 breaks xenner

On Tue, 2008-07-15 at 18:28 +0200, Alexander Graf wrote:
> Hi,
> 
> while attempting to get xenner running, I realized that changeset #17532 
> (xemnstored: Fix xenstored abort when connection dropped.) actually 
> triggers the problem it is supposed to fix.
Well, it did fix the problem. I''m not saying it didn''t
introduce or
expose another one though ;-)
> To easily reproduce the issue, run xenstored with that patch applied and 
> start xenner. As soon as xenner quits (I usually killed it because my 
> guest hung), xenstored abort()ed on me.
I''m afraid I don''t have a xenner setup. Please could you
collect a
backtrace of the crash. Also perhaps it would be worth instrumenting up
the ref counting functions to log incr/decr to spot where the count goes
wrong.
> 
> The problem does not exist with xenstored from 3.2.1, did trigger with 
> current unstable though.
> 
> 
> Thank you for looking into this,
> 
> Alex

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

Ian Campbell writes ("[Xen-devel] Re: [Regression] Changeset #17532 breaks
xenner"):
> On Tue, 2008-07-15 at 18:28 +0200, Alexander Graf wrote:
> > To easily reproduce the issue, run xenstored with that patch applied
and
> > start xenner. As soon as xenner quits (I usually killed it because my 
> > guest hung), xenstored abort()ed on me.
> 
> I''m afraid I don''t have a xenner setup. Please could you
collect a
> backtrace of the crash. Also perhaps it would be worth instrumenting up
> the ref counting functions to log incr/decr to spot where the count goes
> wrong.
Does xenner connect to xenstored via an AF_UNIX socket or via a shared
memory ring ?  If the former I would suggest that attaching to
xenstored with strace, and watching it die, would be one good thing to
do.

Ian.

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

Ian Jackson wrote:
> Ian Campbell writes ("[Xen-devel] Re: [Regression] Changeset #17532
breaks xenner"):
>> On Tue, 2008-07-15 at 18:28 +0200, Alexander Graf wrote:
>>> To easily reproduce the issue, run xenstored with that patch
applied and
>>> start xenner. As soon as xenner quits (I usually killed it because
my
>>> guest hung), xenstored abort()ed on me.
>> I''m afraid I don''t have a xenner setup. Please could
you collect a
>> backtrace of the crash. Also perhaps it would be worth instrumenting up
>> the ref counting functions to log incr/decr to spot where the count
goes
>> wrong.
> 
> Does xenner connect to xenstored via an AF_UNIX socket or via a shared
> memory ring ?
Both ;)

Guests talks to xenstored via shared page, and xenner itself via socket
(like xend, for device setup etc).  backend daemons talk via socket too.

cheers,
  Gerd

-- 
http://kraxel.fedorapeople.org/xenner/

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

Gerd Hoffmann writes ("Re: [Xen-devel] Re: [Regression] Changeset #17532
breaks xenner"):
> Guests talks to xenstored via shared page, and xenner itself via socket
> (like xend, for device setup etc).  backend daemons talk via socket too.
Right.  If xenner is talking via a socket then you should be able to
capture the exact sequence of events as seen by xenstored, when you
kill xenner, with strace.

If you could post the relevant parts of the result to the list we
could take a look at the entrails.

Ian.

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