On Thu, Nov 15, 2012 at 2:08 PM, Antonio Terceiro
<[email protected]> wrote:
> On Thu, Nov 15, 2012 at 04:20:11PM +1300, Michael Hudson-Doyle wrote:
>> Dave Pigott <[email protected]> writes:
>>
>> > On 12 Nov 2012, at 12:49, Alexander Sack <[email protected]> wrote:
>> >
>> >> On Mon, Nov 12, 2012 at 9:43 AM, Dave Pigott <[email protected]> wrote:
>> >>> Just the one since Saturday:
>> >>>
>> >>> ----------------
>> >>> snowball06
>> >>> ----------------
>> >>> http://validation.linaro.org/lava-server/scheduler/job/38435
>> >>>
>> >>> As noted by Andy, this is our most flakey snowball. I'm proposing 
>> >>> retiring it. It failed three times to reboot and get the network up. I 
>> >>> went on the board and had to hard reset to get the network to come up, 
>> >>> which suggests it gets into a weird state that needs a completely clean 
>> >>> boot.
>> >>
>> >> What is a clean boot? And why are we not doing a clean boot by default?
>> >
>> > Clean boot = hard reset
>> >
>> > Yes, I thought the same. I think we should always do a hard reset
>> > rather than a soft boot. It would mean the board state would be more
>> > likely to be consistent.
>>
>> The worry with always power cycling on master image targets is that
>> powering off sd cards is not particularly friendly, the firmware might
>> be shuffling blocks around as you do it and so even parts of the master
>> rootfs might be lost.  But maybe we should try it and see what happens.
>>
>> With the sd mux, we have to power cycle each time anyway.
>
> What about first sending a "halt" command to make sure everything is
> flushed properly, and only after that power cycling the board?

Yeah. That's what I would expect...



-- 
Alexander Sack
Technical Director, Linaro Platform Teams
http://www.linaro.org | Open source software for ARM SoCs
http://twitter.com/#!/linaroorg - http://www.linaro.org/linaro-blog

_______________________________________________
linaro-validation mailing list
[email protected]
http://lists.linaro.org/mailman/listinfo/linaro-validation

Reply via email to