[Bug 233579] ppc64 r341455 will panic on boot with usefdt=1

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233579

--- Comment #13 from Mark Millard <marklmi26-fbsd at yahoo.com> ---
(In reply to Mark Millard from comment #12)

(PowerMac G5 "4 core" (system total) context, with 16 GiByte of
RAM.)

I tested head -r341836 using usefdt=1 mode based on a build the modern
/usr/src/sys/powerpc/include/vmparam.h (but still smp enabled) and the
results of repeated boot attempts were mixed: many boot failures but
some boot successes. In some cases it took dozens of attempts to get
a success. I also had some example of a bunch of successful boots in
a row.

The failures were: stopping during the "Waking up CPU" sequence where,
other than for CPU 0, "Adding CPU" messages did not show up and and
at least one  "Waking up CPU" did not show up before the boot hung up.
But each hang up at least one did show before the hang-up.

Note: No panics occurred, just hang ups.


Side note on the shutdown isde of things:

There is also a shutdown issue of some the buf*maeomon* shutdowns timing
out instead of reporting "done". Which timeout varies. But this specific
point is also true of using the older VM_MAX_KERNEL_ADDRESS value: just
the usefdt=1 use seems sufficient for this to be an issue. The combination
of the old VM_MAX_KERNEL_ADDRESS and not using usefdt mode dos not show
this behavior in my testing.

-- 
You are receiving this mail because:
You are the assignee for the bug.