FYI: head -r357979 and powerd on Orange Pi+ 2ed (armv7,CortextA7): unable to kill at shutdown; no problem on old PowerMac G5

[Mark Millard]

[artifact.ci kernel bisect result included.]

On 2020-Feb-16, at 14:46, Mark Millard <marklmi at yahoo.com> wrote:

> [Also seen on prior head -r527529 context.]
> 
> I've got two contexts with powerd running
> (both have non-debug kernels and worlds):
> 
> A) A system for old PowerMac G5's. It does not and did not
>   have the problem.
> 
> B) A system for an Orange Pi+ 2ed (armv7, specifically
>   CortexA7) that does. (No other armv7 context around to
>   test.)
> 
> Recently shutdown has been doing the following if
> powerd has been running on the armv7 system:
> 
> . . .
> Stopping powerd.
> Waiting for PIDS: 815
> 90 second watchdog timeout expired. Shutdown terminated.
> Sun Feb 16 00:25:59 PST 2020
> Feb 16 00:25:59 OPiP2E init[1]: /etc/rc.shutdown terminated abnormally, going to single user mode
> Feb 16 00:25:59 OPiP2E syslogd: exiting on signal 15
> 2020-02-16T00:26:19.061233-08:00 OPiP2E init 1 - - some processes would not die; ps axl advised
> Waiting (max 60 seconds) for system process `vnlru' to stop... done
> Waiting (max 60 seconds) for system process `syncer' to stop... 
> Syncing disks, vnodes remaining... 1 0 0 0 done
> Waiting (max 60 seconds) for system thread `bufdaemon' to stop... done
> Waiting (max 60 seconds) for system thread `bufspacedaemon-0' to stop... done
> All buffers synced.
> Uptime: 6m17s
> 
> For reference, top previously listed:
> 
>  PID USERNAME    PRI NICE   SIZE    RES STATE    C   TIME     CPU COMMAND
> . . .
>  815 root        -100    0  4424K  2000K twsi     0   0:00   0.01% /usr/sbin/powerd
> . . .
> 
> I do have for my build (whitespace details might
> not be fully preserved):
> 
> # svnlite diff /usr/src/sys/gnu/dts/arm/sun8i-h3-orangepi-pc.dts 
> Index: /usr/src/sys/gnu/dts/arm/sun8i-h3-orangepi-pc.dts
> ===================================================================
> --- /usr/src/sys/gnu/dts/arm/sun8i-h3-orangepi-pc.dts	(revision 357979)
> +++ /usr/src/sys/gnu/dts/arm/sun8i-h3-orangepi-pc.dts	(working copy)
> @@ -102,7 +102,7 @@
> 		"Line Out", "LINEOUT",
> 		"MIC1", "Mic",
> 		"Mic",  "MBIAS";
> -	status = "okay";
> +	status = "disabled";
> };
> 
> 
> 
> but it has been a fairly long time since I've checked
> if "okay" prevents booting/operating.
> 
> 
> Note: Until there is an official fix checked in, the PowerMac
> G5 needs Ryan L.'s slb_zone_init patch from:
> 
> https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=244118
> 
> to boot anything at or after head -r357549. I also use other
> patches for PowerMacs, mostly personal, but they are unlikely
> to matter for why killing powerd works in this context.
> 

artifact.ci's head -r356602 does not have the problem.
artifact.ci's head -r356610 has the problem.

There are no armv7 builds in artifact.ci between those.

head -r356609 is a sys/dev/iicbux/twsi change and so
looks the most likely given the top result:

 PID USERNAME    PRI NICE   SIZE    RES STATE    C   TIME     CPU COMMAND
. . .
 815 root        -100    0  4424K  2000K twsi     0   0:00   0.01% /usr/sbin/powerd
. . .

with its -100 for PRI. (Previously 0 in the no-problem cases.)

I've no clue if it is powerd that is doing something
odd exposed by the -r356609 update vs. if it is the
update that is doing something odd and powerd just
happens to expose that it does.


===
Mark Millard
marklmi at yahoo.com
( dsl-only.net went
away in early 2018-Mar)