[Bug 233863] r345425 on PowerMac G5 may require kern.smp.disabled=1 and must set usefdt=1 which causes net interface reorder

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=233863

--- Comment #6 from Mark Millard <[email protected]> ---
(In reply to Dennis Clarke from comment #5)

The following notes are currently based on variations on head -r335044 .
It has usefdt=1 enabled always (by changing the code) but does not have
smp disabled (even manually). The context does not involve reverting the
newer kernel maximum figure that I used to revert.

I have a hack that avoid the fans problem on the 2-socket/2cores-each G5
that I currently have access to. It also avoids problems with other
processes that also get stuck waiting for sleep. But it is a hack that is
not even approximately a general/proper solution to the original problem,
it just avoids the consequences. The hack involves an experimentally
found approximate bound that it tests against. That bound might vary
for related G5 contexts.

I've another hack that makes it rare that booting gets hung up: it
narrows a race condition's time frame. But I've have observed a few
examples of hang-up out of hundreds of boots. The hack basically
proves what is going on for the hang-ups based on the large change of
behavior for narrowing the race-condition's time frame.

There is a patch-review active for llvm's libunwind that makes it
handle r2 (the TOC register) correctly for powerpc64 both for
system-clang based buildworld's and devel/powerpc64-gcc based
buildworld's, both ELFv1 ABI and ELFv2 ABI. This makes thrown C++
exceptions work. (This does not cover WITH_LIB32= for exception
handling when WITH_LLVM_LIBUNWIND= is in use in the buildworld.)

-- 
You are receiving this mail because:
You are the assignee for the bug.
_______________________________________________
[email protected] mailing list
https://lists.freebsd.org/mailman/listinfo/freebsd-ppc
To unsubscribe, send any mail to "[email protected]"