lkml.org 
[lkml]   [1996]   [May]   [24]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: The end of booting as we know it
[Axel wrote...]
[snip]

> 3. From Here to There
> =====================
>
> We need a standard format for files that contain a complete memory
> image. Every participating OS should have two new shutdown options:

I doubt that every OS manufacturer holds your view point. That can't
even get together and standardize on EBCDIC or ASCII and now you want
them to standardize boot sequences??

> either save the current state to a specified file and do an
> reboot/halt, or save the current state to a specified file and slurp
> in another state from another specified file. In the case of Linux,
> this would mean that the shutdown command sends a special signal to
> the kernel, which would cause the kernel to flush the buffers and then
> save the memory state to disk. The boot loaders should also know about
> this format; it should be possible to use lilo or loadlin to slurp in
> a memory image. The memory image filenames for the various OS's would
> be given to lilo in the lilo.conf file, and lilo decides at boot time
> whether to do an ordinary boring boot or to slurp in the memory
> image. After having been slurped in, the image file must immediately
> be marked as "outdated" to prevent it from being used ever again. Same
> goes for loadlin.

the boot loader does use BIOS (which you earlier did away with) more
over your system sounds like it has a lot of over head in keeping track
of the available images and such. boot executables are severely
limited in size and as it is now LILO is packed to the gills as I
guess is the bootsect.c portion of the kernels.

additionally the necessary steps for booting up different workstations
(DEC alpha for instance) are quite different from Intel PCs and would
therefore probably cuase problems in standardizing.

>
>
> 4. Problems
> ===========
>
> Some hardware needs to be initialized. One could either try to be
> ambitious and save state information or bring the problematic devices
> in a well-defined state (well-defined across OS's, that is) before
> writing out the memory file, or do it the simple minded way like the
> laptop BIOSes.
>
>
> Axel
>

Note: I'm certainly no boot guru... I can't even get an embedded
controller to boot linux (I think since it doesn't have an A20 handler
or something like that) so take everything I said with a grain of salt
and enlighten me if I said something wrong.

I'm also personally not phased or troubled by the 30 seconds of time
out of my life that is takes to boot my machine. (which is
considerably faster than Windoze95 takes)

- Jeff




\
 
 \ /
  Last update: 2005-03-22 13:37    [W:0.058 / U:3.088 seconds]
©2003-2020 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site