[ipxe-devel] ipxelinux.0

[Martin Zwerschke]

On 03.03.2011 00:02, H. Peter Anvin wrote:
> On 02/26/2011 05:34 AM, Michael Brown wrote:
>> On Saturday 26 Feb 2011 10:08:13 Martin Zwerschke wrote:
>>> I used gpxelinux.0 before, to be able to net boot a mixed environment of
>>> PXE and gPXE -clients from a http-sever.
>>>
>>> As far as I understood, breaking an infinite boot-chain could succeed
>>> this way even without a proxy-user-class, cause gpxelinux does not boot
>>> anything else then the kernel or c32 given in it's
>>> "pxelinux.cfg/default" file.
>>>
>>> Why don't you create an  "ipxelinux.0" ?  To me the concept of
>>> "gpxelinux.0" seemed  a nice try.
>> Good idea.  I've created a syslinux branch at
>>
>>    http://git.ipxe.org/people/mcb30/syslinux.git/shortlog/refs/heads/ipxelinux
>>
>> This will build ipxelinux.0.  I haven't tested it yet.  I suspect it may not
>> work straight away, since I think there's an outstanding bug regarding the use
>> of use-cached.
>>
>> Feel free to give it a try and let us know if it works!
>>
> For what it's worth, I intend to merge iPXE into Syslinux as soon as
> there is a numbered version.  Otherwise bug reports become almost
> impossible to deal with.
>
> 	-hpa
>

I gave feed back to Michael Brown already, unfortunately it did not CC 
to the list.

Reasons for ipxelinux.0  =>
0) booting with PXE-only-netcards (e.g. motherboard-integraded) can 
chainload ipxelinux.0 and instantly fetch kernels initrds from HTTP-Server
without having to fork for user-class
1)  on Windows-desktop-versions there  seems to be no good (free) 
DHCP-Server, that offers DHCP-user classes
2) DHC-user-classes are a bit complicated to configure
3) you can boot ipxelinux.0 "over" existing old versions of gPXE or iPXE 
in ROMs you don't want to re-program,
so you can use actual iPXE-version

Here my answer to Michael Brown again:

Am 26.02.2011 14:34, schrieb Michael Brown:
> On Saturday 26 Feb 2011 10:08:13 Martin Zwerschke wrote:
>> I used gpxelinux.0 before, to be able to net boot a mixed environment of
>> PXE and gPXE -clients from a http-sever.
>>
>> As far as I understood, breaking an infinite boot-chain could succeed
>> this way even without a proxy-user-class, cause gpxelinux does not boot
>> anything else then the kernel or c32 given in it's
>> "pxelinux.cfg/default" file.
>>
>> Why don't you create an  "ipxelinux.0" ?  To me the concept of
>> "gpxelinux.0" seemed  a nice try.
> Good idea.  I've created a syslinux branch at
>
> http://git.ipxe.org/people/mcb30/syslinux.git/shortlog/refs/heads/ipxelinux 
>
>
> This will build ipxelinux.0.  I haven't tested it yet.  I suspect it 
> may not
> work straight away, since I think there's an outstanding bug regarding 
> the use
> of use-cached.
>
> Feel free to give it a try and let us know if it works!
>
> Michael
>
Wow, you are REALLY fast!

I gave it a try (as I'm no linux-man it was a bit of a skill for me).
2 build-errors occured (something 'bout "undionly.kkpxe").
Tried it with "make -i" so it went through in spite of this and created 
a "ipxelinux.0" after all.

With this one i managed to boot Thinstation 2.2.2gRC in a virtualbox 
(this has got a pure PXE emulation).
It
-  pulled "ipxelinux.0" from my tftpd (Jounin)     and -as configured in 
my "pxelinux.cfg"
-  chainloaded vmlinuz and initrd from my IIS-webserver
- successfully started Thinstation.

To my impression, http-transfer goes slower than it should - and if i 
try to chainload vmlinuz/initrd from the TFTP-Server, it gets stuck 
loading vmlinuz.

I believe you are right with this pending "use-cached"-bug.

But all in all it gets really promising!

btw.:  Thinstation creates a "vesamenu.c32" module, that can be run with 
"pxelinux.0" but not with "gpxelinux.0" or as well with "ipxelinux.0".
Do you know the reason for that behaviour?

Thanks a lot and so long,

Martin


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ipxe.org/pipermail/ipxe-devel/attachments/20110303/153a45dc/attachment.htm>