sysinst, serial console on soekris

On Fri, Dec 29, 2006 at 06:54:17PM -0500, James K. Lowden
wrote:
> Mark E. Perkins wrote:
> > Is sysinst going to DTRT on it's own or ask
> > me if I'm using a serial console?  Or do I need to
drop into a shell and
> > run installboot after the installation is done
(and before I reboot)? 
> 
> In theory, yes, sysinst will DTRT.  

In practise as well....
You can set the baud rate (speed) to 0 with installboot, in
which case the
boot code will determine the current rate (by reading the
divisor registers)
Which is actually quite handy when pxebooting a kernel.
(use 'installboot -e' to modify the pxeboot iage)
 
> In practice, I twice told sysinst not to replace my
bootblocks, and it
> twice did (and said as much), forcing me to run
installboot afterwards. 

One day I'll get round to fixing that code...


	David

-- 
David Laight: davidl8s.co.uk

On 2007/01/01 12:28, David Laight wrote:
> On Fri, Dec 29, 2006 at 06:54:17PM -0500, James K.
Lowden wrote:
>> Mark E. Perkins wrote:
>>> Is sysinst going to DTRT on it's own or ask
>>> me if I'm using a serial console?  Or do I need
to drop into a shell and
>>> run installboot after the installation is done
(and before I reboot)? 
>> In theory, yes, sysinst will DTRT.  
> 
> In practise as well....
> You can set the baud rate (speed) to 0 with
installboot, in which case the
> boot code will determine the current rate (by reading
the divisor registers)
> Which is actually quite handy when pxebooting a kernel.
> (use 'installboot -e' to modify the pxeboot iage)

Here's how it went for me...  When sysinst got to the
bootblocks phase, it
appeared to offer to DTRT, suggesting serial console on com0
with baud rate 0
(this was a brand new drive, so there were no existing
bootblocks to fall back
on).  My notes say I accepted this, but when I was done with
the install, I
couldn't boot from wd0.  I tried booting to the install
kernel, dropping into
a shell, running installboot, but still could not boot from
the wd0.  I ended
up running through the entire installation again, this time
specifically
choosing com0 and 9600 baud (via the menu) at the bootblocks
phase.  After
that install, I could boot from wd0.

I'm not sure where I went wrong on the first install, but I
think that when I
ran installboot (from the shell I got via sysinst) I did not
specify the raw
device (which should have been obvious if I had done a more
careful RTFM for
installboot).

Mark

On Mon, Jan 01, 2007 at 04:38:26PM -0500, Mark E. Perkins
wrote:
> 
> I'm not sure where I went wrong on the first install,
but I think that when I
> ran installboot (from the shell I got via sysinst) I
did not specify the raw
> device (which should have been obvious if I had done a
more careful RTFM for
> installboot).

I've a fix for installboot for that in the pipeline (after I
found it
wasn't updating the boot area when regression testing it).

It appears that the 'you can't overwrite the disklabel' code
is rather
overzelous when you write to the sectors near the label
using the non-raw
device.  So although the write() suceeds, the data doesn't
get to the disk!

Calling fsync() and then doing a verify read shows that the
update has
been ignored.

My modifications to installboot should let the i386 code
work on non-512
byte media - where the label will be in the same physical
sector as the
boot code (so it has to enable writes to the label sector).

It is also worth noting that you can pxeboot a GENERIC
kernel and get it to
use the local disk as its root filesystem.
Just interrupt the downcount in pxeboot (you might need to
lengthen it, since
the delay function it uses isn't supported by the soekris
bios), then do:
> boot netbsd.gz -a
and type wd0a for the root filesystem (etc).

	David

-- 
David Laight: davidl8s.co.uk