W dniu 08.01.2019 o?12:51, Stefan Bethke pisze:> Am 08.01.2019 um 10:34 schrieb Marek Zarychta <zarychtam at
plan-b.pwste.edu.pl>:
>> W dniu 03.01.2019 o 14:13, Stefan Bethke pisze:
>>>> I have under supervision a few old servers running 11.2-STABLE.
The
>>>> hardware is almost for retirement, but still in working
condition. It's
>>>> all old Nocona NetBurst microarchitecture. I have recently
tried do
>>>> upgrade OS two of them to 12.0-STABLE, but failed. When I use
old
>>>> bootloader the boot freezes on blue highlighted
"Booting" stage, when I
>>>> tried to use 12 loader, it freezes earlier, on loading kernel
modules.
>>>> The kernel was compiled from fresh sources for CPUTYPE?=nocona.
>>>> 11.2-STABLE is fine with this optimization and the same kernel
boots
>>>> fine on newer hardware.
>>>>
>>>> It is fair, that 11 EOL is expected September 30, 2021 and
these servers
>>>> will likely be retired before this date, but some questions
arise:
>>>>
>>>> Is such old hardware still supported? Is it possible (how to)
debug the
>>>> booting process?
>>> The first step is to try with known-good bits: can you boot these
machines off the 12.0 ISO or memstick images? Can you load your kernel and
modules with the loader from the ISO/memstick? Does GENERIC built without any
flags work?
>>>
>>> If any of these don?t work, try to be as specific as possible when
reporting problems. For example, the exact make of mainboard (kenv output) and
the BIOS version, and any relevant BIOS settings are likely important for
problems regarding the loader. If the kernel and modules load, you can try a
verbose boot to see better how far the kernel gets.
>>>
>>> I?d be really surprised if the CPUs themselves would cause trouble.
>>
>> The first step is done. The affected hardware doesn't boot from
official
>> 12.0-RELEASE CD either. Loader also freezes at the stage of loading
>> kernel modules. These servers are old Maxdata Platinum 500 and 3200.
>> Some time ago I have submitted dmesgs to NYC BUG dmesg
repository[1][2].
>>
>> Both configurations are fine with 11-STABLE, so I am not going to
>> upgrade them and I am replying only FYI.
>>
>>
>> [1] https://dmesgd.nycbug.org/index.cgi?do=view&id=3790
<https://dmesgd.nycbug.org/index.cgi?do=view&id=3790>
>> [2] https://dmesgd.nycbug.org/index.cgi?do=view&id=4111
<https://dmesgd.nycbug.org/index.cgi?do=view&id=4111>
> I think it would be great to get some input from someone familiar with the
new loader. I?ve cc?ed Warner, Kyle and Toomas, as they were listed in the
quarterly status report.
>
After recent MFCs LUA loader became available in 11-STABLE branch. I
have tested it on this old hardware using freshly built r344482. I must
admit the LUA loader on this hardware loads and boots 11-STABLE kernel
flawlessly in root on ZFS environment. The only caveat was that
kernel.old haven't been listed under "5" key, but it wasn't
tested
intensely so the error could be temporary or nonexistent.
Nothing changed with regard to booting FreeBSD 12 on this hardware. I
have tried to boot the recent 12-STABLE kernel. The kernel and all
modules were loaded (in the same manner as previously loader 4th did),
but instead of booting machine froze at this stage, so the breakage is
elsewhere.
--
Marek Zarychta
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: OpenPGP digital signature
URL:
<http://lists.freebsd.org/pipermail/freebsd-stable/attachments/20190225/605762fd/attachment.sig>