Fabrics Bug: keep alive timer re-initialized when active

[Sagi Grimberg]

Hi Berck (or Nash :))

> During fabric cable pulls, sometimes the keep alive timer gets
> re-initialized while it's already active.  It appears to be a race as it
> doesn't happen every time, but can be reproduced fairly consistently
> with a dozen or cable pulls.  Without object debugging enabled, the
> resulting crash is a pretty opaque dereference of a poisoned object.
> See attached for the ODEBUG output.

Thanks for reporting.

> We're seeing this on 4.12, but I couldn't find any reference to it being
> fixed later.

Some code in the reconnect area was indeed changed in the area of auto
reconnects, can you try the latest nvme-4.15 branch
(http://git.infradead.org/nvme.git)?

If the issue still persist there, the only way this can happen is if
reconnect and reset are concurrently running together.