[oi-dev] OI 2017-05 snapshot: AMD Athlon 64 (K8) compatibility issues

[Jean-Pierre André]

Hi,

Thank you for your reply. Below I am adding a few comments
about it.

However, the issue is not what I thought, so I am to open
a new thread about it.

Toomas Soome wrote:
>
>> On 4. juuni 2017, at 20:13, Jean-Pierre André
>> <jean-pierre.andre at wanadoo.fr <mailto:jean-pierre.andre at wanadoo.fr>>
>> wrote:
>>
>> Jean-Pierre André wrote:
>>> ken mays via oi-dev wrote:
>>>> On that note, be aware of prior notice on AMD64 Athlon 64-based
>>>> computers and the use of software from:
>>>
>>> I have such a computer, which I intended to upgrade to 2017.04
>>>
>>> Is booting on a live-usb a valid test for presence/absence of
>>> further problems ahead ? (IOW is the problem in the new loader or
>>> in the kernel ?)
>>
>> As nobody told whether the issue lies within the new loader or
>> in the kernel, and bug #8169 pointed at an issue in the loader,
>> I tried updating the current installation with the grub loader
>> kept in place (BE_HAS_GRUB=true in /etc/default/be).
>>
>> After the update, I can only boot into the new BE in single user
>> mode. Two kinds of errors are displayed :
>>
>> 1) svc:/system/metainit: default:
>>        method "lib/svc/method/svc-metainit"
>>        failed with exit status 127
>> (shown three times then transitioned to maintenance).
>
>
> thats issue about SMF, just run svccfg delete
> svc:/system/metainit:default - it does not affect booting.

Done. On reboot the warning disappears.

>>
>> 2) APIC Error interrupt on CPU 1. Status 0 = 0, Status 1 = 80
>> This second error is traditional on this computer and is
>> normally avoided by the boot option "acpi-user-options=0x2"
>> which was in place.
>
> you can always test by: ok boot -B acpi-user-options=0x2 , or check the
> loader options menu;
>
> also prtconf -v and tr ‘\0’ ‘\n’ < /system/boot/environment  from OS.
>
> This issue is worth to be filed in illumos.org/issues
> <http://illumos.org/issues> if not done already.

The prtconf output shows some possibly related differences :

-        name='acpi-status' type=int items=1
-            value=00000013
+        name='acpi-enum' type=string items=1
+            value='off'
          name='acpi-user-options' type=int items=1
-            value=00000008
+            value=00000002

It at least shows the acpi-user-options is taken into account
(the values 2 and 8 were known to work and the difference
above only means I tried both values).

I thought the APIC error meant the CPU1 did not receive any
interrupts, but apparently both CPUs are running.

I could not find any file in /system/boot and I could not
guess which file you really meant. I issued the command "set"
to get the current environment, but I did not see anything
noteworthy.

>>
>> Are these the errors which I was warned against ?
>>
>> Have they something to do with bug #8169 ?
>>
>> Do they mean I am kicked out, or is there a way to update
>> to the latest state before the bug was created ?
>
> 8169 is fixed in later builds, it was unfortunate the ISO’s got
> published just a bit before the issue and the fix. However, the 8169 is
> only affecting systems with ACPI v1 and only if you boot the latest
> iso/usb, pkg update will give you the fix. The 8169 does make kernel
> startup to fail as the boot loader was passing wrong size for ACPI table
> - if you got the single user, you are not affected.
>
> However, if you did end up in single user mode, it means that some
> essential service did fail for some reason, does svcs -vx reveal anything?
>
> rgds,
> toomas
>
>