sparc64 9.1-beta1 test results

[Kurt Lidl]

On Thu, Jul 19, 2012 at 03:42:00PM +0200, Marius Strobl wrote:
> On Wed, Jul 18, 2012 at 11:40:49PM -0400, Kurt Lidl wrote:
> > I downloaded the sparc64 .iso image from the recently built 9.1-beta1
> > test distribution.
> > 
> > I burned it to a CD-ROM, and then booted it on my test Netra-T1 105.
> > (440Mhz CPU, 1GB memory, 2 x 18GB disks)
> > 
> > I booted via 'boot cdrom' at the PROM monitor.
> > 
> > I accepted all the default options - UFS install, entire disk, install
> > all sets, turn on all services offered by default, told it to setup
> > a dump device...
> > 
> > A couple of notes:
> > 
> > At the end of the install cycle, the installer asks if you want to
> > reboot the machine.  I selected yes, and then the machine rebooted.
> > But it attempted to reboot the cd-rom media again.  Obviously, it
> > was using the same "boot cdrom" be which I selected the original boot
> > media.
> 
> That's typical PROM behavior.
> 
> > 
> > I sent a break to the serial console, then did a "reset", so the machine
> > would attempt to boot the first device in the "boot-device" list of
> > devices.
> > 
> > On the first boot of the freshly installed system, I noted a few errors:
> > 
> > Boot device: rootdisk  File and args:
> > 
> > >> FreeBSD/sparc64 boot block
> >     Boot path:
> >    Boot loader: /pci at 1f,0/pci at 1,1/scsi at 2/disk at 0,0:a
> > Consoles: Open Firmware console
> > 
> > FreeBSD/sparc64 bootstrap loader, Revision 1.0
> > (root at heller.cse.buffalo.edu, Thu Jul 12 14:04:35 UTC 2012)
> > bootpath="/pci at 1f,0/pci at 1,1/scsi at 2/disk at 0,0:a"
> > Loading /boot/defaults/loader.conf
> > [...]
> > Trying to mount root from ufs:/dev/da0a [rw]...
> > Setting hostuuid: 34d846dd-d126-11e1-95bc-080020da684a.
> > Setting hostid: 0x52a3fdee.
> > No suitable dump device was found.
> > Entropy harvesting: interrupts ethernet point_to_point kickstart.
> > Starting file system checks:
> > 
> > It looks like the setting of 'dumpdev="AUTO"' in the /etc/rc.conf
> > doesn't work properly on the sparc64 machines.
> 
> 
> Do you have swap configured? I don't see why 'dumpdev="AUTO"' shouldn't
> work otherwise or why this should be a sparc64 specific problem.

Yes, the default "AUTO" configuration of the disks from the installer
ended up giving me about 855MB of swap space.  (Not actually enough
to crash dump onto, but I just accepted the default.)

On my i386/amd64 machines, I've always set the dumpdev explicitly to
a gpt style partiton (e.g. 'dumpdev="/dev/mirror/swap"').

> > 
> > [...]
> > Creating and/or trimming log files.
> > Starting syslogd.
> > realpath: /dev/dumpdev: No such file or directory
> > /etc/rc: WARNING: Dump device does not exist.  Savecore not run.
> > ELF ldconfig path: /lib /usr/lib /usr/lib/compat
> > Clearing /tmp (X related).
> > Updating motd:.
> > Starting ntpd.
> > Starting powerd.
> > powerd: lookup freq: No such file or directory
> > /etc/rc: WARNING: failed to start powerd
> > [...]
> > 
> > More fallout of the dumpdev not working, I think.
> > And then a failure to start the powerd daemon.
> 
> Only a subset of the CPUs found in supported sun4u machines have
> frequency stepping capabilities, also only in incompatible ways
> and nobody has implemented support for that so far. However,
> AFAIK powered isn't enabled by default by bsdinstall(8) so it
> looks like you enabled it explicitly.

Yes, that's what I meant when I said:
	"turn on all services offered by default"
"powered" was an option presented to me, so I turned it on.
When it didn't work correctly, I thought I should document the
failure.

> > [...]
> > Wed Jul 18 18:17:28 EDT 2012
> > 
> > FreeBSD/sparc64 (sol.pix.net) (ttyu0)
> > 
> > login: Jul 18 18:17:28 sol getty[1533]: open /dev/ttyv2: No such file or directory
> > Jul 18 18:17:28 sol getty[1536]: open /dev/ttyv5: No such file or directory
> > Jul 18 18:17:28 sol getty[1534]: open /dev/ttyv3: No such file or directory
> > Jul 18 18:17:28 sol getty[1538]: open /dev/ttyv7: No such file or directory
> > Jul 18 18:17:28 sol getty[1532]: open /dev/ttyv1: No such file or directory
> > Jul 18 18:17:28 sol getty[1531]: open /dev/ttyv0: No such file or directory
> > Jul 18 18:17:28 sol getty[1537]: open /dev/ttyv6: No such file or directory
> > Jul 18 18:17:28 sol getty[1535]: open /dev/ttyv4: No such file or directory
> > Jul 18 18:17:28 sol getty[1541]: open /dev/ttyu2: No such file or directory
> > [...]
> > 
> > The installed /etc/ttys file specifies a bunch of virtual terminals that
> > do not exist on the sparc64 machines.  It would be cool if someone
> > could get this fixed before the 9.1-RC1 release was cut.
> 
> The sun4u machines models are very diverse; some have a video console
> and some a serial console on one of ttyu[0-2]. The default /etc/ttys
> for sparc64 tries to enable all these known console devices so it
> works regardless of the actual model. If the warnings bug you, you
> need disabled the respective devices in /etc/ttys. There's no
> appropriate infrastructure in FreeBSD to only configure the device(s)
> actually needed on a given machine. Also, that would be a bit of
> over-engineered for avoiding a one-time edit of a config file ...

Oh, I'm perfectly capable of editting the configuration files to
make the various daemons not whine, I'm just trying to provide some
feedback on what the 9.1-BETA1 release looks like on the sparc64
machines. Both for documentation for others who might try to install
it, as well as point out some potential (but minor) issues.

I reported the DMA breakage of the CD-ROM before 9.0 (which is what
prevents anyone from installing the 9.0 image from the release CD-ROM)
weeks before 9.0 image finalized (I think to current-users) and got
no feedback, acknowledgement that the problem existed, etc.  So,
I'm using the port-specific mailing list this time.

-Kurt