Pi3 shutdown -p [Was: Re: Shutdown -r under -current hangs on RPi3]

From: bob prohaska <fbsd_at_www.zefox.net>
Date: Mon, 02 Oct 2023 16:17:34 UTC
[subject changed to reflect topic drift]
On Sun, Oct 01, 2023 at 09:12:17PM -0700, Mark Millard wrote:
> 
> My understanding is that the RPi3B's SOC can be powered off without
> cutting the external power, leaving much of the board also powered
> off despite external power. (Simplified wording.) In other words,
> "shutdown -p  now" has a chance of being made to largely work, if I
> understand right.
> 

Ahh, that's borne out with one of my Pi3s. It appears to be a
one-way ticket, however: Ethernet status LEDs off, power LEDs
on, no console response. Moreover, after power-cycling, boot 
stoped at the loader with an OK prompt. The next boot -r came 
up with a live console but still stopped at the loader prompt, 
proceeding only after two "boot" commands. The first was 
unrecognized, the second ran.  It seems as if something is 
contaminating the console input, causing a pause at the loader 
prompt or corrupting the initial input.

It's perhaps unfortunate, but the first Pi3 described gets
its terminal service from the second, both of which have
ftdi usb-serial adapters. Both exhibit stray characters when
establishing a serial console connection to another host.

> If the USB disk is bus powered, that stuck-on may indicate that the
> SOC/board is still powered. (Not just in the external supply
> sense, but operationally using that external power still.)
> 

In both cases the disks are on a powered hub. The hub and Pi
are on the same cord and so power-cycle together, but the
disk isn't reliant the Pi3's power. 

For some months now the first particular Pi3 seems to display console
output at what looks like a 9600 baud rate, though the connection
via cu or tip uses 115200 baud. I've ignored it thus far, but
maybe there's more going on than meets the eye. It's the same Pi3
that displayed the duplicate MAC address problem.

Thanks for writing!

bob  prohaska