[gPXE] Machines hangs when exiting from CLI

[Jonathan Andrews]

On Wed, 2010-03-31 at 16:00 -0400, Joshua Oreman wrote:
> On Wed, Mar 31, 2010 at 2:49 PM, Itay Gazit <itaygazit at gmail.com> wrote:
> > Hi Joshua, all,
> >
> > Continue to our discussion we had a while ago, I continued my testing on
> > entering CLI during POST.
> >
> > At this experiment I changed the main() to a blank function.
> >
> > I tested 19 different models machines. The results were that 5 of the 19
> > hanged when exiting from CLI!   Including machines from HP, IBM and
> > SuperMicro.
> >
> > My conclusion is that regardless the memory consumption, there is a major
> > problem in gPXE when exiting from gPXE during POST.
> >
> > As I mentioned before, having a reliable CLI during POST is important.
> >
> > I am willing to continue debug this issue, if you have any idea on how to
> > solve it then please let me know.
> 
> I appreciate the testing you've done.
> 
> This is a problem with the BIOS on these machines. On many BIOSes, it
> is simply not possible to enter protected mode during POST; attempting
> to do so will cause booting to ultimately fail. There are two
> solutions:
> - have gPXE provide only limited POST-time functionality with real-mode code;
> - don't try to use the CLI during POST
> 
> The first is extremely difficult. The second is extremely easy.
> 
> I don't understand - why is it so necessary that you enter the CLI
> during POST, as opposed to doing it when gPXE is invoked as a boot
> device?
Excuse my ignorance .... I don't see HOW you would enter a CLI before
POST. When I was programming against BIOS 20 years ago the "feature ROM"
hook was not activated until after POST. Before POST interrupts where
disabled - IRQ 1 (keyboard) and INT10(video)where simply not available,
INT10H video hook was only active as the last stage in POST.  In short
pre-POST is zero video, zero keybaord as far as I know ?

Cheers,
Jon