On Sunday, September 29, 2002, at 08:42 PM, Kevin Keith Woo wrote:

> On Sun, 29 Sep 2002, Bill Studenmund wrote:
>>>
>>> If I remove the ":0" from the boot-device environment variable, I get
>>> "loader: unrecognised client program format" errors, which indicate
>>> that OF 2.4 can't find OFWBOOT.XCF. Yet if I add ":0" or ":1" (I 
>>> tried
>>> ":1" recently), OFWBOOT.CXF appears to load, because I now see:
>>
>> Very weird.
>>
>
> Out of curiosity, has anyone suffered/experienced the above phenomenon 
> before? It sounds like no-one else has seen this behaviour before.
>
>>
>> I think the problems you're encountering now have little to do with 
>> OF 2.4
>> specifically (the :0 vs :1 bit does). One big problem is that we don't
>> really support opening a file from an hfs partition. If you renamed
>> netbsd_g.gz (why _g AND .gz?) to just netbsd, I think it'd work 
>> better.
>>
>
> The files are on an ISO9660 partition. I didn't make an hfs partition 
> on the C i'm currently trying to boot. The file name was originally 
> netbsd_generic.gz, so when I made a pure ISO9660 (I omitted 
> 31-character extensions. Should I allow 31-character filenames?) the 
> filename of the kernel was truncated to netbsd_g.gz. I'll rename the 
> kernel file to netbsd on a new ISO, and try then new ISO.
>

one thing to try/check is the following line that works on my G4 dual 
to boot from a hard drive I stuffed
into the zip bay. I don't really recommend this for the unix partition 
though, as it runs using pio vice dma.

boot zip:,\ofwboot.xcf zip:11,/netbsd

note the following:
1) ofwboot.xcf is on the hfs partition that will boot macos 9.2 also.
2) partition 11 is MY a partition for netbsd, with type "A/UX Root", 
yours may be different.
3) note especially that the part of the boot line for netbsd uses a / 
vice a \... the \ will not work
4) some or all of the above may not apply to your version of OF, but it 
took me quite a while
	and a random glimpse at someone elses email re the / to get the 
booting working properly
	after reading the install docs repeatedly. I always managed to boot 
the original kernel from
	the hfs partition using the \ character and no partition number 
though. The hassle was getting
	my new netbsd file over to the hfs+ partition to try it out.
5) other things besides pdisk that might be useful is the cdrecord 
stuff packaged up so that one can
	rebuild the system and then make a new install cd, or perhaps working 
firewire hard drives under
	netbsd.
6) one other glitch I ran into, is with the new -current rescue 
directory, where I tried to make a "live" cd
	that I could use for repair work should I need it. Unfortunately, the 
incantation I used with mkisofs
	made a separate file for each of the hard links in that directory, 
eating up half of the cd I my memory
	serves correctly. Anyone know the proper direction to wave the magic 
wand to accomplish this?



>> The problem is if we are passed ide1/disk@0:1,\NETBSD_G.GZ as a name, 
>> we
>> first try to see if OF can open the whole thing (I'm not sure why it
>> can't).
>
> I don't know why either. Perhaps we'll leave this issue for now, and 
> perhaps we'll get some ideas later.
>
>> Then we try to have OF open "ide1/disk@0", and look for a NetBSD
>> disklabel,
>
> I didn't realise OFWBOOT.XCF looks for a NetBSD disklabel, so I didn't 
> create any NetBSD disklabels on the ISO. Might this cause problems?
>
>> look in partition 1 therein, and look for NETBSD_G.GZ in there.
>> We really want to also try to have OF open "ide1/disk0:1" and look for
>> NETBSD_G.GZ in there.
>>
>
> On an ISO image, what is the difference between partition 0 (aka 
> "ide1/dis@0:0" (correct me if i'm wrong)) and partition 1 (aka 
> "ide1/disk@0:1" (again, correct me if i'm wrong))? I see identical 
> directory listings when I invoke "dir ide1/disk@0:0",
>
> Regards,
>
>     Kevin
>
> kkwoo@sdf1.org
> SDF Public Access UNIX System - http://sdf.lonestar.org
>