[syslinux] Distilling down a small set of useful system data

Kevin Landreth crackerjackmack at gmail.com
Wed Apr 27 15:51:16 PDT 2011


On Wed, Apr 27, 2011 at 5:00 PM, H. Peter Anvin <hpa at zytor.com> wrote:
>
> On 04/27/2011 02:48 PM, H. Peter Anvin wrote:
> > On 04/27/2011 02:41 PM, Erwan Velu wrote:
> >> If we speak about conditional booting, cpu is a very common requested
> >> feature. Ifcpu.c32 exists wheres a conditional ifdmi.c32 doesn't.
> >
> > Yes, that's true, but ifcpu64 was not sufficient.  Sending *ALL* the
> > data that ifcpu hinges in an unsolicited reply is a *lot*.
> >
> > Therein lies the dilemma...
>
> I do note that DMI-based configuration file selection *has* been a very
> commonly requested feature; not so much for conditional boot but for
> conditional menus.
>
> There might be a difference there, I don't really know.  I do recognize
> this as a shortcoming, I just don't have a good answer at this point.
> Ideas would be appreciated...
>

As a suggestion,  I believe the primary CPU information is returned
from eax in "00000001" [1]. So what if you just returned an comma
separated string of hex encoded registers eax,ebx,ecx,edx [2] ?  This
would always be within the 80 character limit as requested and usable
to some extent. While not consumable with the naked eye, it is
parseable and /fairly/ standardised.  One could always re-implement
the ifcpu.c32 parsing in whatever desired language on the HTTP server.

Just a suggestion, don't know if it helps any.

[1] Relevant data from "cpuid":
   eax in       eax          ebx          ecx         edx
00000001 00010676 01040800 000ce33d bfebfbff

[2] "0x00010676,0x01040800,0x000ce33d,0xbfebfbff"

---
Kevin Landreth




More information about the Syslinux mailing list