after running the installer on a PowerMac G5 no root filesystem found

[Andreas Tobler]

On 15.11.18 21:47, Mark Millard wrote:
> On 2018-Nov-15, at 12:13, Andreas Tobler <andreast-list at fgznet.ch> wrote:
> 
>> On 15.11.18 21:05, Michael Tuexen wrote:
>>>> . . .
>>> OK, I did some testing. Reverting r334498 and not disabling smp results
>>> in a usable, stable system. However, there are two things I observed that
>>> are not working as good as with disabling smp:
>>> 1. After a while the fans get louder. Running sysctl -a dev.fcu.0
>>>     gets them down again. After a while they will start to get
>>>     louder again.
>>>     When SMP is disabled, this is not an issue.
>>
>> pmac_therm gets stuck, we already had that, r329080 tried to fix this.
>> It worked a while but now things are back to worse.
> 
> In my context somethings things go for weeks without
> this happening. I also have more trouble with such from
> debug kernel builds than from non-debug ones. I normally
> run non-debug kernels unless I'm trying to track down
> some issue.
> 
> For non-debug kernels rebooting (shutdown -r now) does
> seem to initially clear the condition.
> 
> This has been a fairly long term issue.
> 
>>> 2. When shutting down the system,
>>>     Waiting (max 60 seconds) for system thread `bufdaemon' to stop...
>>>     Waiting (max 60 seconds) for system thread `bufspacedaemon-1' to stop...
>>>     Waiting (max 60 seconds) for system thread `bufspacedaemon-0' to stop...
>>>     times out.
>>>     This is also not an issue when SMP is disabled.
>>> So there are some issue with SMP even when reverting r334498.
>>
>> I second the two points above, iow I see the same on different SMP G5 running 64-bit. For me both are related.
> 
> I've seen this and, as I remember, it has been under debug kernels
> instead of non-debug ones. Since I normally run non-debug kernels,
> I normally do not see this.
> 
> The powerpc's involved are still based on -r339076. The prior
> context  with much use was long before that, predating -r334498
> as well. I do not remember seeing this before the jump to
> -r339076 .
> 
> So I expect this is somewhat new.
> 
> 
> I can not comment much on having smp disabled. The only time
> that I've had such was the jump to -r339076 until I identified
> reverting -r334498 as how to get back to a nearly
> historically-normal context.

I usually run GENERIC64-NODEBUG and it is happening. I'm on r340261+, 
maybe I should allocate a weekend for bisecting once more :(