> On Fri, 2012-10-26 at 23:59 +0000, Zou, Yi wrote:
>
> <-- snip -->
>
>
> Mmmmm, not sure on this one. There have definitely been changes in the
> TCM active I/O shutdown codepath to support tcm_qla2xxx active I/O
> shutdown starting in v3.5 code, so if pre v3.5 code is working as
> expected it might very well be it.
>
> I'm happy to have a look at this some point in the next week to try and
> reproduce in vn2vn mode.
>
> > Anyway, I would like to mimic your setup to see if I can reproduce it.
> >
>
> Sure, the latest target-pending/master HEAD should easily reproduce with
> slub_debug=FPUZ.
>
> Thanks Yi!
>
> --nab
>
Nick,
With help from Mark, the original issue I was seeing seems fixed by the following:
c0158ca workqueue: cancel_delayed_work() should return %false if work item is idle
which was causing refcnt on exchange being messed up. However there may be other
issues but let me know if you are still hitting this on rc3.
Thanks,
yi
��.n��������+%������w��{.n�����{������ܨ}���Ơz�j:+v�����w����ޙ��&�)ߡ�a����z�ޗ���ݢj��w�f
