Failed attempt to boot a (non-debug) head -r339076 on an old PowerMac G5 "Quad Core" (built via devel/powerpc64-gcc): Waking up CPU 1

[Mark Millard]

On 2018-Oct-8, at 10:25 AM, Michael Tuexen <tuexen at fh-muenster.de> wrote:

>> On 8. Oct 2018, at 19:00, Mark Millard <marklmi at yahoo.com> wrote:
>> 
>> [kern.smp.disabled=1 is confirmed to enable booting
>> my head -r339076 build on teh "Quad Core", much like
>> the 11.2 reports for "2MHz dual-core".]
> Does 11.2 run with smp enabled?

[My wording was poor: "reports" in "11.2 reports with
smp enabled" was referring to reports by you and others
on the lists, not screen messages from attempting a
11.2 boot.]

No clue at this point for the G5 "Quad Core". I have no
11.x context established, and have not since head
went from 11 to 12, as I remember. Would trying to boot
an installer/live for 11.2 be sufficient to test
this?

A debug kernel build did not appear to report anything for
smp enabled for head. The boot -v did report "Waking up CPU
2" before it stopped echoing messages.

Is there a bugzilla for this old-multi-processor/multi-core
G5-PowerMac issue?

I moved the SSD to an old 2GHz PowerMac G5 with 2 single-core
processors (not: one dual-core processor). It also had the
problem. It did report the "Adding CPU 1" line as the
last message. (There should be another "SMP:" message after
that as I remember.)

Again kern.smp.disabled=1 allowed the boot. But top showed
both processors active. And the fan gradually wound up to a
sustained high speed even though the PowerMac kept operating.
The shutdown -p now had to wait for each of various processes
to timeout. It looks like kern.smp.disabled=1 is somewhat
messed up for some hardware.

(I did not look for such on the "Quad Core": its fan did not
prompt me to look at what was going on.)


I do not have access to other types of PowerMac G5
examples or any other type of powerpc64 computer. So
I can not directly check a single dual-core processor
test-case or any others.


> Best regards
> Michael
>> 
>> On 2018-Oct-8, at 3:09 AM, Michael Tuexen <tuexen at fh-muenster.de> wrote:
>> 
>>>> On 8. Oct 2018, at 09:48, Mark Millard via freebsd-ppc <freebsd-ppc at freebsd.org> wrote:
>>>> 
>>>> [I finally have access to old PowerMacs again. The below
>>>> is based on my environment for experimenting with building
>>>> via more modern C/C++ compilers with matching toolchains.
>>>> The build here was via a cross build from amd64 via
>>>> devel/powerpc64-gcc .]
>>>> 
>>>> When I attempt to boot the PowerMac G5 "Quad Core"
>>>> with my head -r339076 build it gets as far a reporting
>>>> (based on boot -v output):
>>>> 
>>>> . . .
>>>> ada0: (I'll not list the 5 ada0: lines)
>>>> GEOM: new disk cd0
>>>> GEOM: new disk ada0
>>>> Adding CPU 0, hwref=ff89d680, awake=1
>>>> Waking up CPU 1 (dev=ff89eb70)
>>>> 
>>>> That is the last of the output. After a while the fans
>>>> wind up. (The text is hand translated from a picture
>>>> of the screen.)
>>>> 
>>>> Without the "boot -v" the "ada0:" lines are the last to
>>>> show.
>>>> 
>>>> The build is a non-debug build (with symbols). I've
>>>> not tried a debug kernel build yet. It may be a while
>>>> before I will get to. Nor have I tried a gcc 4.2.1
>>>> based build yet.
>>>> 
>>>> At this point I do not have a clue what is going on.
>>>> This early in the boot I'm not sure how or if I'll
>>>> get a clue. (I've never had serial console access
>>>> on the old PowerMac's, which contributes to the messy
>>>> context.)
>>>> 
>>>> I do have in place in the variation used of src.conf:
>>>> LOADER_DEFAULT_INTERP=4th
>>>> 
>>>> The PowerMac G5 used does boot fine with the boot disk
>>>> I had experimented with back in May (or before).
>>> Does it boot when you disable SMP by adding
>>> kern.smp.disabled=1
>>> to /boot/loader.conf
>>> 
>>> This turned a non-booting/hanging up dual CPU G5 into a
>>> very stable (single core) machine...
>> 
>> Sure enough, kern.smp.disabled=1 lets my -r339076
>> devel/powerpc64-gcc based build boot on the G5
>> "Quad Core".
>> 
>> Unfortunately I jumped from -r333594 (May-13) to
>> -r339076 so I can not well narrow the range for
>> when things changed beyond that.
>> 
>> I see that the original report was for 11.2,
>> while I'm using head. Also that Justin reported
>> seeing the problem on a 2GHz dual-core too (on
>> 11.x I assume).
>> 
>> So we have . . .
>> 
>> 2GHz dual-core (11.2)
>> 2.5 GHz "Quad Core" (2 dual-core processors) (head)
>> 
>> as contexts so far, not that the internal issue
>> needs to be the same for the two, but it might be.
>> 
>> I wonder if 11 at, say, -r333583 (May-13) might
>> be before its problem. If so, it might be easier
>> for someone using 11.x variants to find when
>> it broke in 11 than to do so for head, presuming
>> the problem is common.
>> 
>> Is there a bugzilla for this?

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