Re: [Etherboot-users] Re: http://rom-o-matic.net updated with Etherboot-5.4.0

On Apr 27, 2005, at 11:48 PM, isac wrote:
> No, we shouldn't need mknbi for etherbooting memtest86(v3.2). The 
> binary file named 'memtest' we get (along with memtest.bin) when 
> building memtest86 (from the source tarball) is already a valid nbi 
> for both eb5.2.4 and eb5.4.0 (the two eb versions I've tried). If the 
> nfl menu is not that important for Sam who seemed not that comfortable 
> with pxelinux, he could simply do this:
>
>     etherboot (5.2.x or 5.4.0) -> memtest

Does this actually work for you?

     Etherboot-5.4.0 -> memtest

fails for me (booting from a floppy).  It attempts to load the ELF 
file, but memtest claims it gets an unexpected interrupt and aborts.

Perhaps the ELF loading code is not working properly in 5.4.0.

> I believe it is an nfl-related issue more than anything else. In fact, 
> it (mk{nbi,elf}-nfl) never worked (for me at least) since eb5.3.x.

Yes, this is something that we need to discuss to figure out whether 
NBI is actually supportable now that Ken has lost interest in Etherboot 
hacking, since he was pretty much the only maintainer of mknbi.

The PXELINUX model has merit because once you wrap a 
kernel/file/code-segment with mknbi-*, later if you find a bug in 
mknbi, all the files wrapped with that version of mknbi may be broken, 
so you may have to re-wrap them.

PXELINUX loads native kernel files and disk images, so if there is a 
bug in PXELINUX, you only have to get a newer PXELINUX.0 file.  Your 
images are not affected.

It may be time for NBI format to be retired, and for us to require 
people to use PXELINUX and MEMDISK to load images.

If anyone has good reasons for keeping NBI support, and is willing to 
support mknbi-* and NBI loading in Etherboot, going forward, I'd really 
like to hear from you.  Right now, as things stand, it seems 
unsupportable from a practical standpoint, and PXELINUX seems able to 
do what most people need done, and is well-supported.

> That's  why I'm still stuck with 5.2.4 today (which is more than 
> enough for me :). Unfortunately, I don't know enough about bootloaders 
> to be able to fix it. The only thing I noticed about memtest86 is that 
> its 'load' and 'start' address look very different from that of a 
> regular linux kernel. Here's what I got by doing a `disnbi memtest`:
> Type: ELF
> Start address:          00010000
> Flags:
> Segment number 0
> Load address:           00010000
> Image length:           91796
> Memory length:          91796
>
> Could nfl be doing something bad there?

It's possible.  I think Etherboot should be able to load an ELF file at 
64K.

Debugging ELF support is something we'll be working on in 5.5.  Feel 
free to join in.  Maybe we can figure this out together.

Thanks for you help,

Marty

-- 
     Try: http://rom-o-matic.net/ to make Etherboot images instantly.

    Name: Marty Connor
US Mail: Entity Cyber, Inc.; P.O. Box 391827;
          Cambridge, MA 02139; USA
   Voice: (617) 491-6935; Fax: (617) 491-7046
   Email: [email protected]
     Web: http://www.etherboot.org/



-------------------------------------------------------
SF.Net email is sponsored by: Tell us your software development plans!
Take this survey and enter to win a one-year sub to SourceForge.net
Plus IDC's 2005 look-ahead and a copy of this survey
Click here to start!  http://www.idcswdc.com/cgi-bin/survey?id=105hix