Re: Blind-access patch for Linux 1.3.x
Albert Cahalan (albert@ccs.neu.edu)
Thu, 7 Mar 1996 15:28:33 -0500 (EST)
>
> I'm the most uncomfortable with the changes to the bootsector. We only
> have so many configuration bytes available, and using a full byte for
> the blind option seems to not be the best use of that resource.
>
> Question: is it really critical that all of the hardware probe messages
> always be accessible via the blind-mode console? How about this as a
> proposal:
>
> All kernels will have support for turning on/off "blind mode" via the
> escape sequences. Init will recognize the kernel command line argument
> "blind=1", and if it sees it, it will send the escape sequence to all of
> the console. A blind person who wishes to see the boot messages gets
> Linux running using the above procedure, and then recompiles a kernel
> whose default mode is "blind mode on". That will allow the blind person
> to see the hardware probing boot messages.
This is not good enough because it is not possible to
know what went wrong during an install.
The problem is that this stuff must go in the boot sector.
Why just one sector? There could be:
0: magic number, floppy code
1: config options
2: other junk that needs the BIOS
3: more
4: more
5: ...
Actually, I do not think the kernel should even start in 16-bit mode.
The 386+ boot stuff for Linux/BSD/Amoeba/Mach posted here was great.
For now though, it would be good to just move stuff into a second sector.