| Re: Creating bootable CD for Solaris 8 x86
Bump... no further interest?
Sunny wrote:
>
>
> Sunny wrote:
>
>>
>>
>> Beezer wrote:
>>
>>> Sunny wrote...
>>>
>>>
>>>>> Thanks Sunny, I already saw this doc, but unfortunately it's for Sparc
>>>>> platforms, the CD structure for Intel is different I believe...
>>>>
>>>>
>>>>
>>>> Doh! Of course, it has to be El Torito... think before posting, Sunny!
>>>>
>>>> You'll need a copy of the El Torito specification (easily googled),
>>>> and a hex editor so you can analyse sector 17 and the boot catalog
>>>> to see how the Sun CD works. The techniques outlined in the
>>>> Blueprint should still be applicable for customising the boot slice
>>>> behavior, but assembling the components into an image will need to
>>>> follow El Torito specs.
>>>>
>>>> This is an interesting project... wish I had time for it :-(
>>>
>>>
>>>
>>>
>>> Wish I had time as well ;-)
>>>
>>> What puzzles me is why my CD doesn't work, when all I've done is dd
>>> bits off the SunCD and reassemble them. I think I must be missing
>>> something...
>>
>>
>>
>> You are, or rather your CD is - it's the first 512 bytes of sector 0
>>
>> I took a look at sector 17 today. The boot catalog specifies no
>> emulation, just load 4 sectors starting at the beginning of sector 0 -
>> but when you dd s1, (or s2 for that matter) it skips the first 512
>> bytes of sector 0. On the other hand, dd of s0 gets everything.
>>
>> If I grab the first 512 bytes of sector 0 using CDRWIN on Windows, and
>> prepend it to copies of s1 and s0 obtained with dd, the resulting CD
>> is a bootable copy of the original.
>>
>> I don't understand why dd skips the first block when it wasn't told to
>> (thereby throwing the whole CD off by 512 bytes and rendering it
>> nonbootable), and I haven't yet found a way to extract those bytes on
>> Solaris. Pointing dd at the raw device made no difference...
>>
>> Anyone?
>
>
> I successfully built and installed from a modified Software 1 of 2 CD
> with embedded flash archive today :-) ... but I lost my main Sol9 x86
> installation (and most of my notes) in the process - see my post
> "suninstall gets root disk wrong" for the unfortunate details :-(
>
> The procedure is pretty much as described in the Sun Blueprint for
> building a bootable installation from CD on Sparc, except for the need
> to tell mkisofs to do the El Torito stuff and fix the mess it makes when
> you do.
>
> I used cpio to take s1 off the original CD, deleted the Product
> directory to make room, then customised .install_config for jumpstart
> installation of the flash archive (which I also stored in .install_config)
>
> mkisofs arguments to rebuild s1 were:
>
> mkisofs -no-emul-boot -boot-load-seg 0 -boot-load-size 4 -G bootseg -f
> -b .bootimage -c .catalog -d -L -l -N -p Sunny -P Sunny -o
> sol9x86_803.s1 s1
>
> Where bootseg is a file containing sectors 0 thru 16 from the original
> CD. I still haven't figured out how to extract these on Solaris, so I
> grabbed them on Windows.
>
> mkisofs insists on a boot image (-b .bootimage), despite having been
> told we want no emulation and given the load segment and sector count.
> It then proceeds to point the boot catalog at the boot image instead of
> the load segment, which has to be fixed with a hex editor later. I
> suspect Sun had the same problem when they built the original CD, as
> .bootimage exists at the root of the original, and is a 1.44MB (floppy
> sized) file containing nothing but nulls.
>
> With any luck I'll be able to repeat today's results, despite the loss
> of my notes, once I rebuild my installation - in which case I'll post a
> complete recipe.
>
> Still hoping someone can suggest how to read raw sectors off the
> original CD from Solaris... I'd prefer the final recipe not require the
> use of Windows tools :-) Life would also be easier if mkisofs could
> somehow be convinced to point the boot catalog at the specified load
> segment. Any ideas?
>
| 
