Re: sshd signal 11 on -current

On Jan 19, 2024, at 16:39, bob prohaska <fbsd@www.zefox.net> wrote:

> On Fri, Jan 19, 2024 at 11:40:55AM -0800, Mark Millard wrote:
>> 
>> You may need to do local experiments where the
>> dsl_modem is not connected? Or may be move
>> switchA to the other side of the router temporarily?
>> 
>> Problem isolation may require things be off the
>> public WAN some of the time.
>> 
> 
> 
> In this case, ns2.zefox.net (along with ns1 and www)
> are at software EOL and need to be updated anyway. I
> might as well start with ns2  and see what happens. 
> If the problem persists, then there's something to
> investigate further. A setup similar to pelorus.zefox.org
> is likely to be used, which communicates via ssh
> with no trouble over the network as it is.
> 
>>> in this case 
>>> ns2.zefox.net and ns1.zefox.net seem to be affected. 
>>> No other hosts (so far) have reported that particular
>>> error.
>> 
>> If you are unable to isolate the smallest configurations
>> that fail and which Operating System combinations fail
>> I do not see how to get evidence for the range of
>> observed oddities.
>> 
> 
> At this point the need is for a nameserver that works and
> is maintainable. 12.4.4 works for now, but isn't maintainable.
> 
> 
>>> Subject to the proviso that www.zefox.net is headless
>>> and so must be accessed via ssh from some other host.
>> 
>> No tip session attached to www.zefox.net ? That
>> connection type is not via ssh to www.zefox.net .
>> 
> 
> The serial console server for ns2.zefox.net is
> nemesis.zefox.com. Ssh to nemesis followed by tip
> to ns2 works fine, no "Corrupted Mac..."errors even 
> when running grep -i ssh /var/log/messages.

Sounds like we have stopped for now.

>>> The choice of host seems to make no difference.
>> 
>> macOS vs. RasPiOS/Win10 ?
> 
> But, ns2 needs to communicate with modern hosts, not antiques. 
> All the servers in the trio (ns1, ns2, www) need to update
> past 12.4.4. Changing ns2 is inevitable. Going to aarch64
> seems prudent.
> 
> At the outset I thought I was seeing something much worse
> than stale software or maybe hardware. 

I do not expect armv7 to go away quickly. But it is some
time before 15.0-RELEASE . 14.* armv7 will likely
stick around while any of the 14.* are supported.

And, it turns out that today they got pkg base working
for armv7 as well: the base_latest packages are
present for main and stable/14 and base_weekly should
show up, as should releng/14.0 's base_relase_0 .

pkg base allows for tracking main or stable/14 or
releng/14.* without source code builds involved
if you do not want tailored system builds. That might
allow for smaller memory systems lasting longer.

But:

QUOTE
Packages will be provided from FreeBSD 14.0 onwards: 
    • as a technology preview of sorts, for experimentation and use by projects such as Doug Rabson's FreeBSD container images
    • not yet for things such as bsdinstall(8)
    • not yet intended for production environments. 
END QUOTE

See https://wiki.freebsd.org/PkgBase for some notes on
its use. (I've yet to try it. My armv7 and aarch64 main
[so: 15] builds are tailored in various ways.)

===
Mark Millard
marklmi at yahoo.com