On 7/23/2011 3:08 PM, Alexander Motin wrote:
> I've simulated one real world device failure condition, when SATA disk
> still reports its presence, but doesn't respond to any command. I've
> found that due to multiple command retries, each of which cause 30s
> timeout, bus reset and another retry/requeue, it may take ages to
> eventually drop the failed device. Odd thing that those retries continue
> even after XPT considered device lost and invalidated it.
> I've made a patch (http://people.freebsd.org/~mav/periph_noretry.patch)
> for cam_periph_error() to block any retries after periph was marked as
> invalid. With that patch all activity completes in 1-2 minutess, just
> after several timeouts, required to consider device loss.
> Can this way considered to be correct?
Yes, I like this.