Fwd: Re: Problems with SMP in amd64

[Giorgos Keramidas]

On 2006-02-08 10:39, John Baldwin <jhb at freebsd.org> wrote:
>On Wednesday 08 February 2006 02:54, Daniel Valencia wrote:
>> OK... it's Jan. 25th, at 8:14am.
>>
>> there are a bunch of commits at 8:14:45 that relate to the ata
>> drivers, and a couple at second 46 and 47... but that's it.
>
> Interesting.  Those changes are just before the /dev/pts
> changes that a lot of people are currently blaming for amd64
> instability.  It looks like 8:14am your time is 6:14pm PST?

I'm seeing random panics (unfortunately under X11, and I haven't
had the time to build a kernel with DDB_UNATTENDED yet) with a
kernel from 2006/01/26 01:30:00 UTC, but not as serious as the
ones after the pts changes.

What is funny is that reverting the pts changes to sys/kern only
from HEAD doesn't solve the syscons problems on FreeBSD/amd64
I've hit, so it is possible that the bug that has prevented HEAD
from being usable here for a few days exists in pre-pts code.

I'm now rolling back to 2006/01/25 00:00:00 UTC, which seems to
be the last date I could build a kernel & userland that was safe
to use on my laptop.

Then I'll start trying to narrow this down in the commits between:

    2006/01/25 00:00:00 UTC
    2006/01/26 01:30:00 UTC

But it may take a while, since I only have a single amd64 machine
here and it builds world+kernel in around 1.5 hours.

- Giorgos